Division of Economics and Business Working Paper Series

A Primer on Optimal Power Flow: Theory, Formulation, and Practical Examples
Stephen Frank Steffen Rebennack

Working Paper 2012-14 http://econbus.mines.edu/working-papers/wp201214.pdf Colorado School of Mines Division of Economics and Business 1500 Illinois Street Golden, CO 80401

October 2012

c 2012 by the listed authors. All rights reserved.

Colorado School of Mines Division of Economics and Business Working Paper No. 2012-14 October 2012 Title: A Primer on Optimal Power Flow: Theory, Formulation, and Practical Examples∗

Author(s): Stephen Frank Department of Electrical Engineering & Computer Science Colorado School of Mines Golden, CO 80401-1887 stephen.frank@ieee.org Steffen Rebennack Division of Economics and Business Colorado School of Mines Golden, CO 80401-1887 srebenna@mines.edu

ABSTRACT The set of optimization problems in electric power systems engineering known collectively as Optimal Power Flow (OPF) is one of the most practically important and well-researched subfields of constrained nonlinear optimization. OPF has enjoyed a rich history of research, innovation, and publication since its debut five decades ago. Nevertheless, entry into OPF research is a daunting task for the uninitiated—both due to the sheer volume of literature and because OPF’s familiarity within the electric power systems community has led authors to assume a great deal of prior knowledge that readers unfamiliar with electric power systems may not possess. This primer provides a practical introduction to OPF from an Operations Research perspective; it describes a complete and concise basis of knowledge for beginning OPF research. The primer is tailored for the Operations Researcher who has experience with optimization but little knowledge of Electrical Engineering. Topics include power systems modeling, the power flow equations, typical OPF formulations, and data exchange for OPF. AMS subject classifications: 90-01, 90C26, 90C30, 90C90 Keywords: power flow, optimal power flow, electric power systems analysis, electrical engineering, nonlinear programming, optimization, operations research.

This work is supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-1057607.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49


1. Introduction. The set of optimization problems in electric power systems engineering known collectively as Optimal Power Flow (OPF) is one of the most practically important and well-researched subfields of constrained nonlinear optimization. In 1962, Carpentier [8] introduced OPF as an extension to the problem of optimal economic dispatch (ED) of generation in electric power systems. Carpentier’s key contribution was the inclusion of the electric power flow equations in the ED formulation. Today, the defining feature of OPF remains the presence of the power flow equations in the set of equality constraints. In general, OPF includes any optimization problem which seeks to optimize the operation of an electric power system (specifically, the generation and transmission of electricity) subject to the physical constraints imposed by electrical laws and engineering limits on the decision variables. This general framework encompasses dozens of optimization problems for power systems planning and operation [11, 36, 37]. As illustrated in Figure 1.1, OPF may be applied to decision making at nearly any planning horizon in power systems operation and control—from long-term transmission network capacity planning to minute-by-minute adjustment of real and reactive power dispatch [14, 32, 37]. To date, thousands of articles and hundreds of textbook entries have been written about OPF. In its maturation over the past five decades, OPF has served as a practical proving ground for many popular nonlinear optimization algorithms, including gradient methods [4, 10, 22], Newton-type methods [29], sequential linear programming [3, 27], sequential quadratic programming [7], and both linear and nonlinear interior point methods [15, 30, 31]. These OPF algorithms, among others, are reviewed in several surveys [17–19, 36], including one that we recently published [11, 12]. Although OPF spans both Operations Research and Electrical Engineering, the accessibility of the OPF literature is skewed heavily toward the Electrical Engineering community. Both conventional power flow (PF) and OPF have become so familiar within the electric power systems community that the recent literature assumes a great deal of prior knowledge on the part of the reader. For example, while conducting our recent survey we found that few papers even include a full OPF formulation, much less explain the particulars of the objective function or constraints. Even introductory textbooks [26, 32, 37] require a strong background in power systems analysis, specifically regarding the form, construction, and solution of the electrical power flow equations. Although many Electrical Engineers have this prior knowledge, an Operations Researcher likely will not. We believe this accessibility gap has been detrimental to the involvement of the Operations Research community in OPF research; our impression is that most OPF papers continue to be published in engineering journals by Electrical Engineers. What is missing from the literature—and what we provide in this primer—is a practical introduction to OPF from an Operations Research perspective. The goal of this primer is to describe the tool set required to formulate, solve, and analyze a practical OPF problem. Other introductory texts for OPF focus heavily on optimization theory and tailored solution algorithms. In contrast, this primer places an emphasis on the theory and mechanics of the OPF formulation—the least documented aspect of OPF—using practical, illustrative examples. Because we have written this primer for the Operations Researcher, we assume that the reader has significant experience with nonlinear optimization but little or no background in Electrical Engineering. Specifically, this paper requires a solid understanding of



Planning Horizon Years Months Weeks Days Hours Minutes Real-Time

asin Incre Decr

del g Mo

n lutio Reso

easin g

Unce rtain


Long-term Generation Scheduling Capacity Expansion Planning Real Power Dispatch Reactive Power Dispatch Reactive Power Planning

Unit Commitment SecurityConstrained Unit Commitment

Economic Dispatch Classic Optimal Power Flow

Automatic Generation Control

SecurityConstrained Unit Commitment

Optimal Reactive Power Flow

Voltage Control

Fig. 1.1. Optimization in power system operation via incremental planning. Long-term planning procedures make high level decisions based on coarse system models. Short-term procedures fine tune earlier decisions, using detailed models but a more limited decision space. Bold text indicates planning procedures which incorporate variants of optimal power flow.

50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66

• linear algebra [16], • complex number theory [13, 16], • analysis of differential equations in the frequency domain [16], and • linear and nonlinear optimization theory and application [20, 24]. Readers who also have a working knowledge of electrical circuits [21] and electric power systems analysis [14] will find the development of the power flow equations familiar. Other readers may wish to expand their understanding by consulting a good power systems text such as [14] or [32]. However, prior familiarity with power flow is not strictly required in order to follow the development presented in this primer. The primer begins with a guide to OPF notation in §2, including notational differences between electric power systems engineers and Operations Researchers. Sections 3 and 4 introduce the modeling of electric power systems and the power flow equations; these fundamental topics are omitted in most other introductory OPF texts. Building upon the previous sections, §5 discusses OPF formulations; this section includes full formulations for several of the decision processes shown in Figure 1.1. §6 provides a descriptive guide to two common file formats for exchanging PF and OPF data. Finally, §7 concludes the primer.

we have added clarifying remarks. the branch between buses i and k is denoted by (i. In the optimization community.1. for example 120 V indicates 120 Volts. Notation.3. In the examples.4. bold roman font (A) indicates a set.2. Dimensions. but P 2 indicates (net) power squared.2. x) vector function of equality constraints h(u. and Sets. Throughout this primer. In Electrical Engineering. rather than i or ˆ ı.1. designates the imaginary unit. Remark 2. j . x) vector function of inequality constraints The symbols e and j represent mathematical constants: e Euler’s number (the base of the natural √ logarithm).4 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 STEPHEN FRANK AND STEFFEN REBENNACK 2. the superscript L differentiates load power P L from net power P . For this reason. italic roman font (A) indicates a variable or parameter. The following dimensions and indices are used in the OPF formulations within this primer: N total number of system buses (nodes) L total number of system branches (arcs) M number of system PQ buses i. General.71828 j the imaginary unit or 90◦ operator. and a tilde over a symbol (a) indicates a phasor quantity (complex number). k ) or ik . while numeric superscripts indicate mathematical operations. electrical units are specified where applicable using regular roman font. Remark 2. e ≈ 2. Where there are significant differences. this primer uses notation that follows the conventions of electric power systems engineering rather than the Operations Research community. Indices. In order to remain consistent with the existing body of OPF literature. In this primer. We use the following general notation for optimization formulations: u vector of control variables (independent decision variables) x vector or state variables (dependent decision variables) ξ vector of uncertain parameters f (u. x) objective function (scalar) g (u. This differs from the common use in Operations Research of e as the unit vector and j as an index. k indices corresponding to system buses and branches c contingency case index t time period index Remark 2. For example. System branches are indexed as arcs between buses. we avoid the use of j as an index throughout this primer. For example. c typically refers to a vector of objective function coefficients. j = −1 Remark 2. Letter case does not differentiate variables from parameters. Symbolic superscripts are used as qualifiers to differentiate similar variables. however. Subscripts indicate indices. a given quantity may be a variable in some cases and a parameter in others. We use L to indicate the number of system branches because B is reserved for the bus susceptance matrix. we use upper case C for . 2. The following electrical units are used in this primer: V Volt (unit of electrical voltage) A Ampere (unit of electrical current) W Watt (unit of real electrical power) VA Volt-Ampere (unit of apparent electrical power) VAR Volt-Ampere Reactive (unit of reactive electrical power) 2. The unit for a quantity follows the numeric quantity and is separated by a space.

In power systems analysis. we use H and K to represent sets of controllable phase-shifting and tap-changing transformers rather than S (often used to designate sources or scenarios) and T (often used to designate time periods).A PRIMER ON OPTIMAL POWER FLOW 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 5 objective function coefficients and reserve lower case c for the contingency case index of security-constrained economic dispatch as described in §5. as two real-valued variables in rectangular form a + jb. we adopt the following sets in this primer: N set of system buses (nodes) L set of system branches (arcs) M set of load (PQ) buses G set of controllable generation buses H set of branches with controllable phase-shifting transformers K set of branches with controllable tap-changing transformers Q set of planned locations (buses) for new reactive power sources C set of power system contingencies for contingency analysis T set of time-periods for multi-period OPF Remark 2.5.2. all of these notations are found in the OPF literature. The letters H and K otherwise have no special association with phase-shifting and tap-changing transformers. or as two real-valued variables in polar form c∠γ .3. electrical quantities are represented in the frequency domain as phasor quantities (complex numbers).) For convenience. There is no standard set notation within the OPF literature. we document the usual symbols and relationships used for the electrical quantities.1. Complex numbers may be represented as a single complex variable.) Here. Admittance. For clarity. (Many authors do not use sets in their formulations. Electrical Quantities. however. (Complex number notation is explained in more detail in §3.1. Zik Rik Xik yik gik bik Sh yik Sh gik bSh ik S yi S gi bS i 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 Yik Yik θik Gik 2. 2.2. some notational exceptions exist in the literature. complex impedance of branch ik resistance of branch ik (real component of Zik ) reactance of branch ik (imaginary component of Zik ) Zik = Rik + jXik complex series admittance of branch ik series conductance of branch ik (real component of yik ) series susceptance of branch ik (imaginary component of yik ) yik = 1/Zik = gik + jbik complex shunt admittance of branch ik Sh shunt conductance of branch ik (real component of yik ) Sh shunt susceptance of branch ik (imaginary component of yik ) Sh Sh Sh yik = gik + jbik complex shunt admittance at bus i S shunt conductance at bus i (real component of yi ) S shunt susceptance at bus i (imaginary component of yi ) S S S yi = gi + jbi complex ik th element of the bus admittance matrix magnitude of ik th element of the bus admittance matrix angle of ik th element of the bus admittance matrix conductance of ik th element of the bus admittance matrix (real component of .3.

we summarize several fundamental concepts required for power systems analysis and the development of the power flow equations as presented in §4. many of which may be unfamiliar to the Operations Researcher. Other. complex (phasor) current injected at bus i magnitude of current injected at bus i complex (phasor) current in branch ik . and B : the former represents the values corresponding to individual system branch elements. Scholarly literature discussing OPF assumes a working knowledge of power systems models and electrical concepts. but we include it here to disambiguate the complex quantities from their associated (real-valued) magnitudes. directed from bus i to bus k magnitude of current in branch ik 172 173 174 175 176 177 178 179 180 181 182 183 Ii Ii Iik Iik PiL QL i L Si 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 2. I . Note the distinction between lowercase y . and Y .3. and b and uppercase Y . G. g . . Current. The phasor indicator is omitted for complex power S . phase shift of phase-shifting transformer in branch ik tap ratio of tap-changing transformer in branch ik 201 202 203 204 205 206 207 208 3. load (demand) real power at bus i load (demand) reactive power at bus i load (demand) complex power at bus i L Si = PiL + jQL i G Pi generator (supply) real power at bus i QG generator (supply) reactive power at bus i i G Si generator (supply) complex power at bus i G Si = PiG + jQG i Pi net real power injection at bus i (Pi = PiG − PiL ) L Qi net reactive power injection at bus i (Qi = QG i − Qi ) Si net complex power injection at bus i G L Si = Si − Si = Pi + jQi Remark 2. while the latter refers to the admittance matrix which models the interaction of all system branches.3. the indicator is also often omitted for V . y .7. Fundamental Concepts.3. Bik Vi Vi δi Ei Fi 2. as S is always understood to be a complex quantity.2. ϕik Tik 2. Power.3. complex (phasor) voltage at bus i voltage magnitude at bus i voltage angle at bus i real component of complex voltage at bus i imaginary component of complex voltage at bus i Vi = Vi ∠δi = Ei + jFi 2.5. In the literature.6.3.6 164 165 166 167 168 169 170 171 STEPHEN FRANK AND STEFFEN REBENNACK Yik ) susceptance of ik th element of the bus admittance matrix (imaginary component of Yik ) Yik = Yik ∠θik = Gik + jBik Remark 2.4. Voltage. In this section.

3). ch n a Br Branch (2. (2. (3. the electric power system is analyzed in the frequency domain under the assumption of sinusoidal steady-state operation. Electric power systems are modeled as a network of electrical nodes (buses) interconnected via admittances (branches) that represent transmission lines.1. and all system impedances are fixed. A phasor transforms a sinusoidal time-domain signal into a complex exponential in the frequency domain. This algebraic representation is much easier to solve. Phasor Quantities. Steady-state sinusoidal voltages and currents can be fully characterized by their magnitude and phase shift using phasors.2. The network in Figure 3.1 has N = 5 buses and L = 6 branches. 4). with corresponding sets N = {1. while branches are referenced as arcs between nodes (i. k ) ∈ L. Under these conditions. System Representation. and similar power systems equipment. 209 210 211 212 213 3. all voltage and current waveforms are sinusoids with fixed magnitude. Buses are referenced by node with index i ∈ N. 3) Branch (3. (1. Bus and branch indices in an example 5-bus electrical network.1.4) Bus 4 Br an ch an r B ch (1. 214 215 216 217 218 219 220 221 For power flow analysis.1. 2). 3. and phase shift. 4. the differential equations governing power system operation reduce to a set of complex algebraic equations involving the phasor representation of the system electrical quantities. 5). frequency. 4). k ∈ N.4 . At sinusoidal steady-state.5) Bus 3 Bus 5 Fig. Example 3. 5} and L = {(1. 3. where i.A PRIMER ON OPTIMAL POWER FLOW 7 Bus 2 2) (1. (4. The phasor cejγ may be written as c∠γ in polar coordinates or as a + jb in rectangular Branch (4. using the relationship c sin (2πf t + γ ) Time Domain ←→ cejγ Frequency Domain The frequency f of the signal is fixed and therefore omitted from the phasor notation. 3.5) Bus 1 (3 ) . 2. (3. 5)} . cables. transformers.

Complex power is a phasor quantity consisting of real power and reactive power. For a sinusoid.2. ≈ 103. while in rectangular coordinates it is V = 120 cos −30° + j 120 sin −30° V. in Electrical Engineering. instantaneous electrical power fluctuates as the voltage and current magnitudes change over time. The time domain voltage waveform √ 120 2 sin(377t − 30°) V represents the standard outlet voltage in the United States with the angle referenced. Real power represents real work.9 − j 60. Ch.3. power systems engineers use the concept of complex power to characterize these time domain power fluctuations. . 3. Remark 3. a2 + b2 . In polar coordinates. 2 224 Example 3. The product of voltage and current is electrical power. a c= 222 223 Freitag and Busam [13] provide an overview of complex number theory. however. This is done so that frequency domain power calculations yield the correct values without √ the need for an additional scaling factor. according to Euler’s formula. b = c sin γ. In AC power systems. Complex Power. for instance. In Electrical Engineering.3. b γ = arctan . where.0 V. Thus. a = c cos γ. while O’Malley [21. voltage and current phasors are expressed as root-mean-square (RMS) quantities rather than peak quantities. 11] discusses the use of phasor quantities in Electrical Engineering. this phasor is V = 120∠−30° V. that is. a time-domain voltage waveform VPk sin (2πf t + δ ) has the frequency domain phasor VPk jδ √ e . This voltage has the frequency domain phasor V = 120e−j 30° V. a net transfer of energy from source to load.8 STEPHEN FRANK AND STEFFEN REBENNACK coordinates. 225 226 227 228 229 230 231 232 Note that the RMS voltage magnitude of 120 V is the familiar quantity. to the high side of a utility distribution transformer. For frequency domain analysis. the RMS magnitude is 1/ 2 times the peak magnitude.

while reactive power transfer occurs when voltage and current are 90° out of phase (that is. The magnitude of complex power. . By convention. 233 234 235 236 237 238 239 240 Reactive power. on the other hand. An arbitrary AC current i(t) can be represented by the sum of direct current id (t) (in phase with the voltage) and quadrature current iq (t) (orthogonal to the voltage). represents circulating energy—an cyclic exchange of energy that averages zero net energy transfer over time. complex power S can be computed from1 S = V I ∗ = P + jQ 241 242 243 and consists of orthogonal components P (real power) and Q (reactive power). current i(t) lags voltage v (t) by 30°. reactive power is considered positive when current lags voltage. Complex and 1 Here and elsewhere in this primer. the symbol ∗ denotes complex conjugation rather than an optimal value. Therefore. orthogonal). In the figure. as illustrated in Figure 3.2.2. Real power transfer occurs when voltage and current are in phase. is called the apparent power and is often used to specify power systems equipment and transmission line ratings. 3.A PRIMER ON OPTIMAL POWER FLOW 9 v (t ) Total Current t Quadrature i (t ) = i d(t ) + i q(t ) i d(t ) i q(t ) Direct Instantaneous Power Real t Power v (t )i (t ) v (t )i d(t ) v (t )i q(t ) Reactive Power Fig. This use is typical in Electrical Engineering and consistent with most OPF literature. Direct current produces real power and quadrature current reactive power. |S |. Conceptual illustration of real and reactive power using time domain waveforms.

The perunit value of an SI quantity x on a given base xBase is x . and the reactive power is 120 VAR. The instantaneous power is p(t) = v (t)i(t) = 480 sin(377t) sin(377t − 30°) W. while per-unit quantities are dimensionless. base quantities are given in the SI system (Volts.). The complex power draw of the appliance is S = (120 V∠0°) (2 A∠−30°) . and reactive power has units of Volt-Amperes Reactive (VAR). respectively. Electric power systems quantities are usually expressed as a ratio of the actual quantity to a reference. Indeed. Watts. if any. power systems texts frequently mix per-unit and SI (metric) units. Per-unit quantities are unitless and are labeled using a designation. 3. for instance. = 240 VA ∠30°. As a result.”. real power has units of Watts (W). Remark 3. the average power is pAvg = 0 1 60 240 cos(30°) − 240 cos (2 (377t) − 30°) dt W. Amperes. of “p. this practice is called the per-unit system. Both real and reactive power affect power systems operation and are therefore modeled in PF and OPF.5. both because it does not perform real work and because it is the imaginary part of S . quantity. In power systems analysis.4. etc. 15] and other introductory circuits texts provide a more complete overview of complex power. Reactive power is sometimes called imaginary power. 248 249 The apparent power draw of the appliance is 240 VA. Ch. Or.4. Over time. reporting voltage in per-unit and power in MW. but this assumption is rarely stated explicitly. by using trigonometric identities.8 W.10 244 245 246 247 STEPHEN FRANK AND STEFFEN REBENNACK apparent power have units of Volt-Amperes (VA).8 W + j 120 VAR. Ohms. We can verify that the real power is correct by examining the average power in the time domain. 250 251 252 253 254 255 256 257 258 259 260 261 which is identical to the real power P computed in the frequency domain. = 240 cos(30°) ≈ 207.8 W. ≈ 207. Example 3. Nearly all OPF literature assumes a working knowledge of per-unit on the part of the reader. √ i(t) = 2 2 sin(377t − 30°) A. or base. the real power is 207. xpu = xBase . The Per-Unit System. p(t) = 240 cos(30°) − 240 cos (2 (377t) − 30°) W. A small electrical appliance draws 2 A∠−30° from a 120 V∠0° source. per-unit can be a significant source of confusion when working with practical OPF formulations.u. O’Malley [21. The voltage and current waveforms are √ v (t) = 120 2 sin(377t) V. ∗ = (120 V · 2 A) ∠ (0° + 30°) .

) Once two system bases are specified.u.u. IBase = √ 3 VBase VBase ZBase = √ = 3 IBase √ 3 IBase = YBase = VBase 2 VBase . the voltage ratio of most system transformers becomes 1:1 in per-unit. If a nominally 12.85 kV. simplifying the development of the system admittance matrix. and the remaining three-phase system bases are calculated according to SBase . the voltage and power bases are specified. Per-unit quantities have consistent magnitudes on the order of 1. 4. power systems engineers typically set SBase to one of 10. However.u. 2 In per-unit. 2.0. most notably 1. (i) per-unit and SI quantities cannot be mixed in calculations. 3.95 · 12.2 All calculations that can be performed in the SI system can also be performed in per-unit.1. In power flow analysis. The use of per-unit eliminates the need to apply voltage and current scaling factors at the majority of system transformers. The use of per-unit automatically adjusts for the phase shift of three-phase transformers (Wye-Delta or Delta-Wye). 100.u.6. Example 3. For example. VBase = Line-to-line Voltage. For mathematical convenience. . which improves the numerical stability of power flow calculations. and apparent power (VA) share a common base with units of VA. (For example. then its voltage level in SI units is VSI = 0. When VBase is selected in this way. The use of per-unit eliminates the need to distinguish between single-phase and three-phase electrical quantities. see §4.95–1.5 MW.47 kV is a common distribution voltage level in the United States. regardless of the SI voltage.. on a 1000 MVA base is equal to 150 MW. on a 10 MVA base is equal to 1. real power (W). a power of 0.15 p. and 5.. the others are fixed exactly. or 1000 MVA and select VBase as the nominal line-to-line voltage at each bus. reactive power (VAR). per-unit voltage should always lie within the approximate range 0. The per-unit system is easier to interpret at a glance. but 0. 2 VBase ZBase 281 282 283 284 285 286 SBase is constant throughout the power system.47 kV.47 kV feeder is operated at 0. the per-unit system has several advantages over the SI system of measurement. and (ii) all per-unit calculations must be performed on a consistent set of bases. SBase SBase 1 = .15 p. = 11. SBase = Three-phase Power.A PRIMER ON OPTIMAL POWER FLOW 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 11 Correct interpretation of the SI value of a per-unit quantity requires knowledge of the base quantity. but VBase is distinct for each system bus. 12.05 p.95 p. With a proper selection of system bases.

.12 STEPHEN FRANK AND STEFFEN REBENNACK Similarly. .1) YN 1 YN N 294 295 296 297 298 299 is the N × N complex bus admittance matrix. IN is an N -dimensional vector of phasor currents injected at each system bus. . . Therefore.2) . . Y = . indicate the system power base.059 p. The power values must be converted to per-unit prior to evaluating the power flow equations. the elements of Y were considered constant.. 12. and specify all other quantities directly in per-unit.47 kV feeder operating at 13. and Y11  . relationships for single-phase bases. I represents the current supplied by generators and demanded by loads. .u. .1) into the complex power flow equation S =V ◦ YV ∗ . In power systems analysis. . it is more convenient to work with power flows than currents because (i) injected powers are independent of system voltage angle while injected currents are not. .. we develop the power flow equations and present the mechanics of their construction. . Sarma. a common convention is to specify source and load power in SI units. In practical power systems.. V = V1 . . 4. including instructions for base conversions. The voltages V are state variables which fully characterize the system operation for a given matrix Y . The Power Flow Equations. .  (4. and transformers. . but in newer OPF formulations Y may contain both constants and control (decision) variables. while Y models transmission lines. In this section.  . 287 288 289 290 291 292 293 In OPF.47 kV = 1. and Overbye [14] provide additional discussion of the per-unit system.2 kV . .  Y1N  . VN is an N -dimension vector of phasor voltages at each system bus. a nominally 12. Traditionally. Glover. and practical examples.2 kV (another common distribution voltage level) has a per-unit voltage of Vpu = 13.. cables. The steady-state operation of an alternating current (AC) electrical network is governed by the matrix equation I =YV where I = I1 .. see §6. power systems engineers transform (4. and (ii) working directly with power allows straightforward computation of required electrical energy. but usually no other conversions are necessary. (4.

At each bus i. Y is sparse.Total Ii = Vi yi + k:(i. the total injected power is the difference between the generation and the load.3) S. we first document the types of branch elements used in power flow analysis and then develop a general expression for the entries of Y that satisfies (4.1.i)∈L Vi − Vk yki (4. and Kirchoff’s voltage law (KVL) to a steady-state AC electrical network. Yik = 0.1 shows an example network consisting of two buses i and k and a single branch (i. From Ohm’s law and KVL. the current injected from outside sources (such as connected generators or loads) required to satisfy KCL. Branch (i. O’Malley [21] provides a concise summary of these electrical laws. In matrix form. k ) connects bus i to bus k . k ) has series admittance . k ) between them. Figure 4.4) Typically. L Qi = QG i − Qi .1. The Admittance Matrix. only a single branch (i. OPF data generally does not give Y directly. that is. in which case the offdiagonal elements become Yik = Yki = −yik . Kirchoff’s current law (KCL). S. G L Si = Si − Si . Typically. Ii is the net current flowing out of the bus through all connected branches. load real and reactive power are fixed while generation real and reactive power are control variables with minimum and maximum limits. Example two-bus network illustrating the definitions of bus voltage and injected current. Thus. where S = P + jQ is a vector of complex power injections at each bus and ◦ denotes element-wise vector multiplication.3) when the elements of Y are defined as Yii = Yik = − 305 306 307 308 309 310 Admittances directly connected to bus i Admittances directly connected between bus i and bus k (4. Example 4. At each bus. k ) and yi is the total shunt admittance from bus i to neutral. If there is no connection between buses i and k . In this section. (4.1) is equivalent to (4. and therefore we summarize the mechanics of its construction here. 4. 300 301 Pi = PiG − PiL .A PRIMER ON OPTIMAL POWER FLOW 13 Bus i ei V e I i yiS e yik e ykS e e I k Bus k ek V Fig.k)∈L 302 303 304 Vi − Vk yik + k:(k. 4. having dimension N × N but only N + 2L nonzero entries.1. The bus admittance matrix Y forms the core of the power flow equations.4). The elements of Y are derived from the application of Ohm’s law.Total where yik is the admittance of branch (i.

The Π branch series admittance yik for inclusion in Y is yik = gik bik 312 313 314 315 Xik Rik 1 = 2 2 − j R2 + X 2 . and phase-shifting transformers. yik .5) Branch shunt susceptance bSh ik is related to but distinct from the net shunt susceptance at buses i and k . Off-Nominal Transformers. branch shunt susceptance is usually omitted.3. 4.3) for each bus yields the matrix equation Ii Ik 311 = S yik + yi −yik −yik S yik + yk Vi Vk .2.2. in power flow analysis. tap-changing transformers. Π branch model for cables and transmission lines. the branch shunt susceptance bSh ik is divided into two S at each end of the line. Any transformer that does not have exactly a 1:1 voltage ratio in per-unit is an off-nominal transformer. the voltage ratio across the transformer exactly equals change in system voltage base across the transformer (a 1:1 voltage ratio in per-unit. Proper modeling of off-nominal transformers is a key 316 317 318 319 320 321 322 323 324 325 326 327 328 329 . Cables and Transmission Lines. Rik + jXik Rik + Xik ik ik Rik = 2 2 . as equal parts and added to bus shunt susceptances bS and b i k illustrated in Figure 4. Rik + Xik (4. Power cables and transmission lines are modeled as Π branch circuits (Figure 4. For this network. Specifically.2. 4. which is sometimes given as “line charging” reactive power. transformer branches are almost always modeled with zero shunt susceptance and often with zero series resistance as well. Transformers. 4.2). Off-nominal transformers require modified entries in Y to account for the additional voltage magnitude or phase angle change relative to the nominal case. This category includes fixed-tap transformers with off-nominal turns ratios. 4. For short lines. and each bus also has a shunt admittance.1. Rik + Xik Rik = 2 2 . the branch model for a transformer with nominal turns ratios is identical to the Π branch circuit for a transmission line (Figure 4. writing (4. Most power systems transformers have nominal turns ratios. Because the per-unit system automatically accounts for the turns ratio.1.14 STEPHEN FRANK AND STEFFEN REBENNACK Rik Bus i bik ∕ 2 Sh Xik Bus k Sh bik ∕ 2 Fig. that is. However. The line characteristics are specified as a series impedance Rik + jXik and a branch shunt admittance jbSh ik . with no phase shift).2).1.1.

4.7) . In OPF. bus i is the tap bus and bus k is the impedance bus or Z bus. Off-nominal transformer branch model. 1 = yik Vk − Vi . k ) in Y . The transformer turns ratio in per-unit is a:1.6) Similarly. where a is a complex exponential consisting of magnitude T and phase shift ϕ. Iki = yik Vk − Vi . this topic is neglected in most introductory OPF texts. Selecting T = 1 and ϕ = 0 yields the nominal turns ratio. a 1 1 = ∗ yik Vi − Vk . the expression for current Iik is developed as follows: Iik = 1 I . k ) are required such that Iik Iki = Yii Yki Yik Ykk Vi Vk . a 1 = − yik Vi + yik Vk .3. 330 331 skill in developing practical OPF algorithms. while controllable ϕ models a phase shifting transformer. In order to include the effects of off-nominal transformer (i. a (4. In the model.A PRIMER ON OPTIMAL POWER FLOW 15 Bus i ei V e I ik eik I' ei V' a:1 yik e e I ki Bus k ek V Fig. a∗ ik 1 = ∗ yik Vi − Vk .3 displays the standard model for off-nominal transformers found in practical power flow and OPF software. 332 333 334 335 such that Vi = aVi and Iik = Iik /a∗ . partial admittance matrix entries for branch (i. aa a (4. either T or ϕ (or both) may be a control variable: controllable T models an on-load tap changer. Unfortunately. a a 1 1 = ∗ yik Vi − ∗ yik Vk . Using the turns ratio definitions given above and Ohm’s law. Figure 4. a = T ejϕ .

4. then the following correction procedure is required for each off-nominal branch (i. a 346 347 348 349 350 351 352 When the procedure is complete.8) The differing expressions for Iik and Iki in (4. (4. For notational convenience. A transformer with off-nominal magnitude only (real valued a) leaves Y a symmetric matrix.1.2. Construction Equations for Admittance Matrix. The partial diagonal term corresponding to branch (i. aa∗ 343 344 345 where yik is the uncorrected partial diagonal admittance term.i)∈L 1 Sh yki + yki .16 STEPHEN FRANK AND STEFFEN REBENNACK In matrix form. as given by the replacement procedure Yik ←− − 1 yik .3. but a phase-shifting transformer (complex a) does not. If Y has previously been constructed according to (4. When constructing the full admittance matrix Y . rather than performed as a replacement procedure. k ) in Ykk remains unchanged. 3.4. the corrections are made in the initial construction of Y .4) and the correction procedure given in §4. this correction procedure is written with respect to an already constructed admittance matrix Y . In practice. as given by the replacement procedure Old Yii ←− Yii − yik + 1 yik . In general.9) (4. a∗ 4. expressions (4. as given by the replacement procedure 1 Yki ←− − yik .k)∈L 2 1 Sh yik + yik 2 + k:(k. a shunt admittance yik . the relationships of (4.7) become   1 1 y − y ik Iik Vi  ∗ ik a∗  =  aa1  Iki Vk − yik yik a 336 337 338 339 340 341 342 .1.1. The off-diagonal admittance matrix entry Yki is divided by a.4) with off-nominal voltage ratios neglected.k)∈L 1 yik − a∗ ik k:(k.6) and (4. Using (4. Remark 4. aik i=k . 2.4. the entries of Y become S Yii = yi + 1 |aik | k:(i. The off-diagonal admittance matrix entry Yik is divided by a∗ . the effects of the off-nominal turns ratios are included directly in Y .8) must be preserved. k ) in Yii is divided by 2 aa∗ = |a| .i)∈L 1 yki . as given in §4. and a complex turns ratio (nominal or off-nominal) jϕik aik = Tik e .10) Yik = − k:(i. reversing the two leads to considerable error. any of the branch elements described above can be represented by a series admittance Sh yik . 2 (4. The partial diagonal admittance term corresponding to branch (i. k ): 1.8) indicate the importance of the difference between the tap bus i and the Z bus k .

145 0. S Gii = gi + k:(i. 4) is a phase-shifting transformer and branch (4.320 0.k)∈L 1 (gik cos ϕik − bik sin ϕik ) Tik − k:(k. . The remaining branches have series admittances y24 ≈ 5.1 provides a set of branch data for the 5-bus example system of Figure 3.000 0.i)∈L 1 (−gki sin ϕki + bki cos ϕki ) Tki i = k.11) Gik = − k:(i.5). 0. we first compute the series admittance yik of each branch using (4. the series admittance of branch (1.189.032 0.023 0.9)–(4.040 0.000 − j 3.i)∈L 1 (gki cos ϕki + bki sin ϕki ) .k)∈L 1 2 Tik 1 Sh gik + gik 2 + k:(k.1452 y12 ≈ 0. In addition to the branch data. 5) has an off-nominal voltage ratio. All quantities except phase angles are given in perunit. Tki 1 2 Tik 1 bik + bSh 2 ik + k:(k.000 0.30 pu and bus 3 has a shunt conductance of 0.1. y35 ≈ 0.i)∈L i=k 1 bki + bSh .145 −j ≈ 1. Series Resistance Rik 0.000 − j 3.000 0.i)∈L 1 Sh gki + gki .1452 0.3.k)∈L Bik = − k:(i.0232 + 0. 3) is y13 = 0.824.000 0.125.010 0.000 0.12) (4.14) Example 4.333. Table 4.0232 + 0.000 Voltage Ratio Tik · · · · 0.300 0.10) can also be separated until real and imaginary parts using the definition Y = G + jB and the identity aik = Tik (cos ϕik + j sin ϕik ).1 Branch impedance data for Example 4.067 − j 6. y34 ≈ 0.000 Series Reactance Xik 0.294 − j 3. (4.260 0.023 0. For example. bus 2 has a shunt susceptance of j 0.k)∈L 1 (gik sin ϕik + bik cos ϕik ) Tik − 353 354 355 356 k:(k.020 0. To compute the admittance matrix for this system. Dots indicate nominal voltage ratios and phase angles.727.0° · · From Bus i 1 1 2 3 3 4 To Bus k 2 3 4 4 5 5 where a = 1 for any branch with a nominal turns ratio. Equations (4. 2 (4.A PRIMER ON OPTIMAL POWER FLOW 17 Table 4. Note that branch (3.006 0.05 pu.500 Shunt Susceptance bSh ik 0.13) Bii = bS i + k:(i.3.660 − j 30.98 · Phase Angle ϕik · · · −3. 2 ki (4.

AC Power Flow. (0.824 ≈ −0.18 and STEPHEN FRANK AND STEFFEN REBENNACK y45 ≈ 0. and (3.0 if rounding errors are neglected.04 + − + 0.0° ≈ 0.0° = 0. 5).294 − j 3.0e−j 3. δ ) = Vi k=1 N Vk Gik cos (δi − δk ) + Bik sin (δi − δk ) Vk Gik sin (δi − δk ) − Bik cos (δi − δk ) (4. Vi = Vi ∠δi .19 0 −1. equation (4.9)–(4. 3).83 5.95 − j 36.2) can be decomposed into a set of equations for the real and reactive power injections by evaluating the real and imaginary parts of S . which can be written in several equivalent forms depending on whether the voltages and admittance matrix elements are expressed in polar or rectangular coordinates. (Note that a34 a∗ 34 = 1. Yik = Gik + jBik : N Pi (V.33 5.19 −0.999 + j 0. For a given Y . 4. y34 and y35 have off-nominal turns ratios a34 = 1.9802 An example off-diagonal element is Y34 .00 358 359 360 361 362 363 364 365 Verification of the remaining matrix elements is left as an exercise for the reader.052) (0. 0. δ ) = Vi k=1 (4. diagonal element Y33 consists of summing the admittances of branches (1. Selection of polar coordinates for voltage.) Therefore.07 + j 6.66 − j 33.49 + j 3.052 and a35 = 0.824 j 3.41 − j 13.000.999 − j 0.19 . 3).294 − j 3.07 − j 10. For example. 0.05. In the literature.07 + j 6.83.052)  0  0  0.41 − j 13.125 0.00 0.15) Qi (V. (3.067 − j 6.19 0 −5.10) is Y34 ≈ −  0.16) . the most common forms of the AC power flow equations are (in order) 1.00 − j 5. 2 (0. This process yields a pair of equations called the AC power flow equations.80 0.73  0 0 0.00 + j 2. 357 Verification of these values is left as an exercise for the reader.00 + j 3.999 − j 0. the full expression for Y33 is Y33 ≈ (1.09 + j 3.000 − j 2.00 + j 3.73 0 1.66 + j 30.01 0. plus the contributions of the shunt conductance at bus 3 and the shunt susceptance of branch (1.00 + j 2.98e−j 0.33  Y ≈  −1. and rectangular coordinates for admittance.980. which from (4.66 + j 30.78.999 + j 0.10). we construct Y using (4.78 −0. respectively.13 The full admittance matrix is 1.2.22 0 −5.00 + j 3.052) 0.04  0. 4).727) + j ≈ 1. Next.00 + j 3.09 + j 3.

there is little difference between the selection of polar or rectangular coordinates for the admittance matrix. However. δ ) = Vi k=1 (4. In voltage rectangular coordinates. Yik = Gik + jBik : N Pi (E.1.19) Qi (E. if the voltage magnitude is fixed (for instance at a PV bus.17) Qi (V. 37] provide exact derivations of these three forms of the AC power flow equations. From an OPF perspective. These approximations are also useful in the development of the DC power flow equations.A PRIMER ON OPTIMAL POWER FLOW 19 2.18) 3.4). Vi ≥ Vimin . and rectangular coordinates for admittance. however. Vi ≤ Vimax . 32. Selection of rectangular coordinates for voltage. Rectangular coordinates also facilitate the inclusion of transformer voltage ratios and phase angles as decision variables. .3 Each form of the equations involves real-valued quantities only.20) Power systems texts [14. In rectangular coordinates. Yik = Yik ∠θik : N Pi (V. F ) = k=1 N Gik (Ei Ek + Fi Fk ) + Bik (Fi Ek − Ei Fk ) Gik (Fi Ek − Ei Fk ) − Bik (Ei Ek + Fi Fk ) (4. see §4. see §4. Ei i i Similarly. all forms are equivalent and give the exact solution to the power flow under the assumptions outlined in §3. The more important distinction is the choice of polar or rectangular coordinates for voltage. Ei i i 3 A fourth form—selection of rectangular coordinates for voltage and polar coordinates for admittance—is theoretically possible but has no advantages for practical use. Rectangular coordinates are more common in practice because they facilitate the use of certain approximations in fast-decoupled solution methods for conventional PF [37]. Vi = Ei + jFi .3. Selection of polar coordinates for voltage. Vi = Vi ∠δi . and polar coordinates for admittance. neither of these advantages strongly affects the AC power flow equations as used in most OPF formulations. voltage magnitude limits require the functional inequality constraints 2 + F 2 ≥ V min . However. δ ) = Vi k=1 N Vk Yik cos (δi − δk − θik ) Vk Yik sin (δi − δk − θik ) (4. on the other hand. then in polar coordinates Vi can be replaced with a constant value. F ) = k=1 366 367 368 369 370 371 372 373 374 375 376 377 (4. Ei i i 2 + F 2 ≤ V max . The advantage of voltage polar coordinates is that constraints on the voltage magnitude can be enforced directly. a fixed voltage magnitude requires the equality constraint 2 + F2 = V .

4. δ ) = V1 ≈ k=1 1.73V1 V3 sin (δ1 − δ3 ).20 STEPHEN FRANK AND STEFFEN REBENNACK Table 4.07V1 V3 cos (δ1 − δ3 ) + 3.07V12 Vk G1k cos (δ1 − δk ) + B1k sin (δ1 − δk ) .20) are quadratic. use of voltage polar coordinates leads to a reduction of variables while use of voltage rectangular coordinates leads to an increase in (non-linear.15). The elimination of trigonometric functions speeds evaluation of the equations.33V1 V2 sin (δ1 − δ2 ) + 6. Polar Coords.33V1 V2 sin (δ1 − δ2 ) + 6. we can write the real and reactive power flow equations for any bus in the 5-bus example system.2 summarizes the differences between the two voltage coordinate choices.19)–(4. these computational advantages outweigh the disadvantages associated with enforcing voltage magnitude constraints. . for a fixed voltage magnitude. The resulting power flow equations (4. which presents several advantages [30]: 1. Using the admittance matrix developed in Example 4. polar coordinates are preferred both for conventional PF and most OPF formulations. Bold entries indicate the superior characteristic. cos (δ1 − δ1 ) − 1. Example 4. however. For this reason. From (4.3. 3. There is.07V12 − 1. + 3. non-convex) equality constraints. Table 4. The 2nd order Taylor series expansion of a quadratic function is exact. Variable limit Variable elimination by substitution N +M −1 Trigonometric Trigonometric Trigonometric Rectangular Coords.04V12 sin (δ1 − δ1 ) ≈ 1. one compelling reason to use voltage rectangular coordinates: expressing voltage in rectangular coordinates eliminates trigonometric functions from the power flow equations. the real power injection at bus 1 is 382 383 384 385 386 387 388 389 390 391 392 393 394 5 P1 (V. This simplifies the application of Newton’s method to the KKT conditions of the OPF formulation. 2. Nonlinear functional inequality constraint Nonlinear functional equality constraint 2N − 2 Quadratic Linear Constant Voltage magnitude limit Fixed voltage magnitude # of variables in conventional PF Nature of PF equations 1st derivative of PF equations 2nd derivative of PF equations 378 379 380 381 Thus. In some cases.07V1 V3 cos (δ1 − δ3 ) − 10. this yields an efficiency advantage in higher-order interior-point algorithms for OPF.73V1 V3 sin (δ1 − δ3 ).2 Comparison of the selection of voltage polar coordinates versus voltage rectangular coordinates for the power flow equations. The Hessian matrix for a quadratic function is constant and need be evaluated only once.

All system branch resistances are approximately zero. The AC power flow equations are nonlinear. Applying these assumptions to (4. (4. while the DC power flow approximation is Pik ≈ −bik sin (δi − δk ). DC power flow models real power transfer quite accurately.15).4 The conventional development of the DC power flow equations requires several assumptions regarding the power system [26.23) 418 419 420 421 The bik term dominates the exact expression because Vi2 ≈ Vi Vk cos (δi − δk ) and therefore the first two terms in (4. However.3. for OPF it implies both a nonlinear formulation and non-convexity in the feasible region. For conventional PF. the reactive power injection at bus 1 is 5 Q1 (V. We observe that (4. 2.04V12 cos (δ1 − δ1 ) ≈ −1.33V1 V2 cos (δ1 − δ2 ) − 6. The system bus voltages are approximately equal to 1. from (4.16).07V1 V3 sin (δ1 − δ3 ) + 10.0. (4. DC Power Flow. (4. Reactive power flow is neglected.15) produces the DC power flow equation N Pi (δ ) ≈ 412 413 414 415 416 k=1 Bik (δi − δk ) (4.07V12 Vk G1k sin (δ1 − δk ) − B1k cos (δ1 − δk ) .04V12 − 3. As a result. Evaluation of the remaining buses is left as an excercise for the reader.73V1 V3 cos (δ1 − δ3 ). The differences between adjacent bus voltage angles are small. This assumption requires that there is sufficient reactive power generation in the system to maintain a level voltage profile.23) overestimates the magnitude of the branch power transfer (4. sin (δ1 − δ1 ) − 1.22) 417 cf. the DC power flow equations still model an AC power system. 4. that is.A PRIMER ON OPTIMAL POWER FLOW 21 Similarly. the assumptions required for DC power flow can lead to significant errors for stressed systems. . the transmission system is assumed to be lossless. 4.21) Under normal system operating conditions. 37]: 1.22) largely cancel. such that sin(δi − δk ) ≈ δi − δk and cos(δi − δk ) ≈ 1. this nonlinearity requires the use of an iterative numerical method. It has been successfully used in many OPF applications that require rapid and robust solutions. 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 − 3.07V1 V3 sin (δ1 − δ3 ) + 10.73V1 V3 cos (δ1 − δ3 ). However. all θik = ±90° and all Gik = 0. In order to simplify the system representation. δ ) = V1 ≈ k=1 1.33V1 V2 cos (δ1 − δ2 ) − 6.22) if 4 The DC power flow is so named because the equations resemble the power flow in a direct current (DC) network. power systems engineers have developed a linear approximation to the power flow equations. The exact equation for branch power transfer is Pik = gik Vi2 − gik Vi Vk cos (δi − δk ) − bik Vi Vk sin (δi − δk ). This approximation is called DC power flow. 3.

0.590 p.90 ∠ −20°.05 · 0. Remark 4. In particular.684 p. but have no realistic physical interpretation.5.0 p.u.95 · 0. ≈ 0.95 ∠ 0° and Vk = 0.90 cos (0° + 20°) − 2. (These numbers do not represent normal operation. Thus. we refer the interested reader to [25. especially if the network is large [28]. Even under normal operation. Each system bus has four variables (real power injection Pi . are allowable in emergency conditions.0 sin (0° + 20°). For further discussion regarding the advantages and disadvantages of DC power flow. a basic understanding of these methods is helpful when reviewing OPF literature.16) or (4. The solution method for voltage rectangular coordinates is similar. = 0. Here. Solution Methods for Conventional PF.05 · 0.4. Operating voltages as low as 0.u. and voltage angle δi ) and is governed by two equations: either (4. as magnitudes are by definition nonnegative. a unique solution to the conventional PF requires fixing the values of two out of four variables at each bus. Therefore. we discuss the solution of the AC power flow equations with voltage polar coordinates. in practical DC power flow models an estimate of the losses must be reintroduced using approximate methods. Zhu [37] provides a good summary.) The exact power transfer for this line is Pik = 0.u. Consider a transmission line from bus i to bus k with admittance 0.15)–(4.95 · 0. the DC power flow equations should not be used for OPF under stressed system conditions unless they have been carefully evaluated for accuracy in the system under test. Even though the power flow equations are nonlinear. or (ii) The angle difference between the buses is too large. Let Vi = 0. the approximation of a lossless transmission network can also lead to significant errors in generator scheduling.952 − 0. but are plausible for a stressed power system. (An example would be any solution which returns a negative voltage magnitude. The approximate power transfer is Pik ≈ −2. reactive power injection Qi . Therefore.9 p. heavily loaded transmission lines.) . and marginal fuel cost estimates.6. Example 4. 28]. branch power flow estimates.18). Therefore. Many practical OPF algorithms incorporate aspects of conventional PF solution methods.0 · 0.90 sin (0° + 20°).05 − j 2. large differences in voltage in different areas of the system can lead to significant error [28].17)–(4. Other solutions may exist mathematically. voltage magnitude Vi . Throughout the rest of this primer.. Depressed voltages and larger than normal angle differences are common in stressed power systems. Power transfer errors for certain critically loaded branches can be much higher than the average branch error. and angle differences of up to ±30° can occur on long. most of this error is attributable to the voltage difference.22 422 423 424 425 426 427 428 429 430 431 432 433 434 435 STEPHEN FRANK AND STEFFEN REBENNACK (i) The bus voltages at either end of the branch are depressed relative to the assumed value of 1. 4. there exists only one physically meaningful solution for most power systems models given an equal number of equations and unknowns. we use the AC power flow equations. 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 The error in the approximate power transfer is 16%. Observation (ii) follows from the relationship |sin (δi − δk )| ≤ |δi − δk |. including loss approximation methods.u.

If all voltage magnitudes and angles in the system are known.3 Power system bus types and characteristics for conventional power flow. Newton’s method is commonly used to solve this system. the voltage magnitude and angle are fixed and the power injections are free.15) at each PV and PQ bus and the reactive power injection equation (4. ∆V (4. There are M PQ buses in the system. Bus Type # of buses in system Known quantities Unknown quantities # of equations in conventional PF Slack 1 δ.16) at each PQ bus.26). Q δ. This equation set consists of the real power injection equation (4. the power injections are fixed while the voltage magnitude and angle are free. Newton’s method is locally quadratically convergent. 37]. then the power injections are fully determined.25) (4. V P.24) for the ∆δ and ∆V required to correct the mismatch in the power flow equations computed from (4. or PV bus. V δ.26) ∆Qi = 481 482 483 484 485 486 − Qi (V.25)–(4. (This corresponds to allowing a local source of reactive power to regulate the voltage to a desired setpoint. given a sufficiently good starting point.24) where J is the Jacobian matrix of the system. Table 4.) There are N − M − 1 PV buses in the system.3 summarizes the three bus types. .A PRIMER ON OPTIMAL POWER FLOW Table 4. QG i − QL i (4. 32. Voltage-Controlled Bus At a voltage controlled bus. V 2 PV N −M −1 P. This is done by solving N + M − 1 simultaneous nonlinear equations with known right hand side values. the method reliably finds the correct solution to the PF equations. Assigning buses in this way establishes an equal number of equations and unknowns. Newton’s method consists of iteratively solving (4. There is only one slack bus in a power system. ∆Q ∂δ ∂V ∆δ ≈J . Solving the power flow therefore requires determining N − 1 voltage angles (corresponding to the PQ and PV buses) and M voltage magnitudes (corresponding to the PQ buses only). At each iteration. The 1st order Taylor series approximation about the current estimate of V and δ yields   ∂P ∂P  ∂δ ∂V  ∆δ ∆P  ≈  ∂Q ∂Q  ∆V . Load Bus At a load bus. Q 0 PQ M P. Therefore. δ ) . δ ) . Newton’s method for conventional PF is described more fully in power systems texts [14. all system buses are assigned to one of three bus types: Slack Bus At the slack bus. the mismatches in the power flow equations are ∆Pi = PiG − PiL − Pi (V. the real power injection and voltage magnitude are fixed while the reactive power injection and the voltage angle are free. Q 1 23 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 In conventional PF. or PQ bus.

the convergence of the power flow equations depends strongly on the selection of a starting point.24 487 488 489 490 491 492 493 494 495 496 497 498 STEPHEN FRANK AND STEFFEN REBENNACK In OPF. or may not converge at all. Decoupled Power Flow versus Decoupled OPF. It provides a direct calculation of portions of the Hessian matrix of the Lagrangian function in OPF (see [29]). this is called a “cold start” or a “flat start”. Even when using the DC power flow equations (which require no trigonometric function evaluations). and 4. 2. and all voltage angles to zero. The alternative is a “hot start”. In DC power flow.5. in which the voltages and angles are initialized to the solution of a pre-solved power flow. When this method is used. 4. For computation. It provides sensitivities in the power flow injections with respect to the state variables. In both conventional power flow and OPF. We discuss a few others here. for two reasons: 1. System Initialization.3.21). the slack bus angle is fixed. In our experience. Power flow software typically handles these conversions transparently. Nearly all optimization software and descriptive languages—including AMPL and GAMS—implement trigonometric functions in radians. there is no distinction between PV and PQ buses because all voltage magnitudes are considered to be 1. As with the AC power flow.5. radians must be used. the Jacobian matrix J plays several important roles: 1. 499 (This is simply a solved matrix representation of (4. the OPF algorithm fixes u and derives x by solving a conventional PF. Given a starting point far from the correct solution. real power 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 . In the absence of a starting point. 4. Power systems engineers usually report angles in degrees.u.0 p. If degrees are used. It provides the linearization of the power flow equations required for successive linear programming (SLP) OPF algorithms. real power injections are strongly coupled to voltage angles and reactive power injections are strongly coupled to voltage magnitudes. 4.2. including in data files for OPF (see §6). not degrees.1. there are several practical and computational aspects of OPF stemming from the power flow equations that can cause confusion. At each search step. 10]. Thus. which is discussed in §3. 3. It is therefore often used to improve computational efficiency in computing the KKT conditions for the Lagrangian function. Practical Considerations. the powers computed from DC power flow will have a scaling error of 180/π .) 4.4. these angles must be converted to radians. Hot starts are often used in online OPF to minimize computation time and ensure that the search begins from the current system operating condition. Degrees versus Radians. Because the DC power flow equations are linear. accepting input and giving output in degrees. the decision variables are often partitioned into a set of control (independent) variables u and a set of state (dependent) variables x [7. Conversely.5.0. 2. it can be difficult to remember that generalpurpose optimization software requires an explicit conversion. standard practice is to initialize all voltage magnitudes to 1. In practical power systems. the power flow equations may converge to a meaningless solution. One of these is the use of the per-unit system. they may be solved directly for the voltage angles using δ = B −1 P.5.

Most of these variants. Decoupled OPF also takes advantage of the strong P -δ and Q-V relationships by formulating a real subproblem and a reactive subproblem. because of the approximated Jacobian matrix. any power systems optimization problem which includes the power flow equations in the set of equality constraints is an OPF problem. it is not always clear whether decoupled OPF is in use or whether a decoupled power flow procedure is used within the solution algorithm for a coupled OPF. Unlike decoupled power flow. the term OPF now encompasses an extremely wide variety of formulations. This feature has led to the development of decoupled solution methods for the power flow equations [14. The optima of the subproblems are assumed to be independent. this is a significant computational advantage. However. The power system is modeled as a set of buses N connected by a set of branches L.25)–(4. Optimal Power Flow. The operating cost of each generator is a (typically quadratic) function of its real output power: Ci PiG . Decoupled power flow therefore is locally convergent to the exact solution to the power flow. more iterations are required for convergence [14]. ∂δ ∂Q ∆Q ≈ ∆V.) Here.1. focusing only on novel enhancements or algorithmic development. Because of the implications for the OPF solution quality. In the OPF literature. Although decoupled power flow uses an approximate update method. (This is so common that most OPF papers omit the core of the formulation entirely. however. The objective is to minimize the total cost of generation. Decoupled OPF is therefore distinctly different from decoupled power flow in that the decoupled OPF solution is inexact. we first present the classical formulation and then briefly discuss several common extensions. these assumptions should therefore be evaluated for accuracy if a decoupled OPF approach is considered. ∂V ∆P ≈ 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 This allows the use of separate Newton updates for δ and V with correspondingly smaller matrices. The most basic decoupling method is to use a set of approximate Taylor series expansions of the form ∂P ∆δ. Broadly speaking. 37]. 5. Zhu [37] discusses several decoupled power flow variants in detail. build upon the classic formulation of Carpentier [8] and Dommel and Tinney [10].29]. Controllable generators are located at a subset G of the system buses. The classical OPF formulation of Dommel and Tinney is an extension of economic dispatch (ED): its objective is to minimize the total cost of electricity generation while maintaining the electric power system within safe operating limits.A PRIMER ON OPTIMAL POWER FLOW 25 injections are weakly coupled to voltage magnitudes and reactive power injections are weakly coupled to voltage angles. however. and the reactive subproblem solves for the optimal values of Q and V while holding P and δ constant [9. The error is a function of the accuracy of the decoupling assumptions. Remark 4. 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 . 5.7. the careful reader should try to discern which is the case. Classical Formulation. Thus. many with tailored solution methods [11]. decoupled OPF solves the subproblems sequentially rather than simultaneously: the real subproblem solves for the optimal values of P and δ while holding Q and V constant. it still uses exact real and reactive power mismatches ∆P and ∆Q from (4.26) and updates both V and δ at each iteration.

The control variable vector u becomes G u = PiG :i∈G . If the system contains controllable phase-shifting or tap-changing transformers. where H and K are the sets of branches with controllable-phase shifting transformers and tap-changing transformers. δ ) = − QL i G. Qi (V.3).15)–(4. ϕ.max ≤ QG Qi i ≤ Qi Vimin ≤ Vi ≤ Vimax min max δi ≤ δi ≤ δi ∀ i ∈ N.min G. Since ϕ and T alter the elements of admittance matrix Y . more recent OPF formulations also consider branch current limits.18). ∀ i ∈ G.min G. In (5. (5.max G Pi ≤ P i ≤ Pi G. .0∠0.6) (5. T ) and Qi (V.2) (5.2)–(5.9) Although not considered in the earliest papers. . Tik:ik∈K . By Ohm’s law.16) or (4. the branch current limits require functional inequality constraints. Qi:i∈G and the vector of state variables (dependent decision variables) is x = (δ2 . The formulation is also augmented with bound constraints on the phase angles max ϕmin ik ≤ ϕik ≤ ϕik ∀ ik ∈ H (5. then the corresponding phase angles and tap ratios are introduced into the set of control variables. the current magnitude in branch ik is Iik = Vi − Vk yik .4) (5. T ). The vector of control variables (independent decision variables) is G u = PiG :i∈G . δN . VN ) .2) and (5. .17)–(4.26 STEPHEN FRANK AND STEFFEN REBENNACK The classical form of the formulation is min i∈ G Ci PiG . 563 564 The voltage magnitude and angle at the system slack bus (by convention. ∀ i ∈ G. ∀ i ∈ N. . Pi (V. δ ) and Qi (V. δ. . δ ) represent the power flow equations in polar form—either (4. Qi:i∈G .1) (5. respectively. δ ) = PiG − PiL QG i ∀ i ∈ N. .t. δ. ∀ i ∈ N. respectively.10) . Unlike the previous bounds. Pi (V. ϕik:ik∈H .8) and the tap ratios min max Tik ≤ Tik ≤ Tik ∀ ik ∈ K.3) (5. (5.3) become functions of ϕ and T : Pi (V. usually to V1 = 1. (5. bus 1) are fixed. ϕ. the left hand sides of (5.5) (5. V2 . .7) s. .

Example 5. Generator Real Power PiG.11) must be corrected for the off-nominal turns ratio. except that we assign ϕ34 and T35 to be decision variables representing a phase shifting transformer and an on-load tap changer.max QG i ∞ 0.95 Max. Dots indicates zero values.11).11) Rather than bounding the square of the current as is given in (5. the classical OPF formulation is difficult to solve. many formulations bound the total real and reactive power flow entering the line. Bus Voltage Vimin 1. Remark 5.min QG i −∞ −0.2)– (5. ϕ34 and T35 have limits −30. or branch current limits.20 Max. variable tap ratios.2.400 0. We now develop the classical OPF formulation for the 5-bus example system first presented in Figure 3. Vi is replaced by Vi = Vi /Tik and δi is replaced by δi = δi − ϕik . Generator Real Power PiG.05 Max. Thus.05 1. (5. not a maximum power. The power flow constraints (5.129 Min. OPF problems have historically been solved using tailored algorithms rather than general purpose solvers.95 0. which is technically a maximum current.2.2 Generator data for Example 5. ⇔ ⇔ 565 566 567 568 569 570 571 572 573 574 575 576 (Vi cos δi − Vk cos δk ) + (Vi sin δi − Vk sin δk ) ≤ (Vi cos δi − Vk cos δk ) + (Vi sin δi − Vk sin δk ) ≤ 2 2 2 2 max Iik . For this example.00 1.1 Bus data for Example 5.00 0.0° . yik max (Iik ) 2 yik 2 ∀ ik ∈ L.min −∞ 0.05 1. However.40 0. we can constrain the branch current using max Vi − Vk yik ≤ Iik . Bus Voltage Vimax 1.05 27 Bus i 1 2 3 4 5 Table 5. Generator Reactive Power .2. All quantities are given in per-unit.max ∞ 0.1. For branches with off-nominal turns ratios.239 Load Reactive Power QL i · · · 0.95 0.11) gives a more exact representation of the true constraint. Min. Load Real Power PiL · · · 0.10 0. Generator Reactive Power .95 0.40 Min.1. Even without variable phase angles.A PRIMER ON OPTIMAL POWER FLOW Table 5.20 −0. For this reason.1. and the presence of trigonometric functions complicates the construction of approximations.900 0. (5.0° ≤ ϕ34 ≤ 30. the tap bus voltage Vi and angle δi in (5.3) are both nonlinear and non-convex. In this case. respectively.05 1.30 0. the branch impedance data are as given in Table 4.20 Bus i 1 3 4 where yik is the magnitude of the branch admittance. All quantities are given in per-unit.

and generation) given in Tables 5. Given this data.2.13. 3. 4).29 cos ϕ34 − 3. it is first necessary to re-write Y from Example 4. (1.82 sin ϕ34 + j (3. T35 Y43 = −0.82 sin ϕ34 + j (3. H = {(3.9)–(4.20P3 + 0. . 3). and K = {(3. T35 Y44 .30P4 + 0.0.35P1 . using (4. are G G C1 P1 = 0. 4} . G = {1. To develop the full formulation. 5). (3. G G G C3 P3 = 0.29 sin ϕ34 ) .50 577 2 .10) and simplifying.05. (4.29 sin ϕ34 ) .1 and 5. and the remaining matrix entries are unchanged. (3. ∗ ∗ (Note that a34 a∗ 34 = 1. 5} .82 cos ϕ34 + 0. load. 5)} . L = {(1. 5)} . (2. Let a34 = cos ϕ34 + j sin ϕ34 and a35 = T35 . and 1/a34 = cos ϕ34 + j sin ϕ34 = a34 . Y33 = 1.40 P3 G G C4 P4 = 0. .82 cos ϕ34 − 0. Y55 . 1/a34 = cos ϕ34 − j sin ϕ34 = a34 . G 2 P4 where the PiG are expressed in per-unit. 2).) Then. 4. the sets defining the formulation are: N = {1. 1 Y35 = j · 3.13. The system power base is 100 MW.53 − j 1 2 · 3.13. 3. in thousands of dollars.3 to explicitly include ϕ34 and T35 . The three generator cost functions. T35 and Y53 = j 578 Y34 = −0. 2.29 cos ϕ34 + 3.41 − j 10. 4). 1 · 3.28 and STEPHEN FRANK AND STEFFEN REBENNACK 0.95 ≤ T35 ≤ 1. Consider the bus data (voltage limits. 4)} .

73V3 sin (−δ3 ) .900 = 5.66V2 V4 cos (δ2 − δ4 ) G P3 = 1.30P4 + 0.41V32 − 1.33V2 cos (δ2 ) − 30.19V4 V2 sin (δ4 − δ2 ) + (−0.33V2 sin (δ2 ) + 30.13V52 − V5 V3 cos (δ5 − δ3 ) − 2.66V2 V4 sin (δ2 − δ4 ) + 33.66V22 − 5.07V3 cos (δ3 ) = 1.29 sin ϕ34 ) V4 V3 sin (δ4 − δ3 ) + 2.t. −0. T35 QG 3 = −1.82 sin ϕ34 ) V3 V4 cos (δ3 − δ4 ) + (3. 3.22V22 − 3.239 = T35 QG 1 = 10.129 = 5.00V4 V5 cos (δ4 − δ5 ) .A PRIMER ON OPTIMAL POWER FLOW 29 Bus 1 is the system slack bus.29 cos ϕ34 − 3.35P1 + 0.29 cos ϕ34 + 3.50 P4 2 .940 = −5.7).40 P3 G P1 2 G G + 0.82 sin ϕ34 ) V4 V3 sin (δ4 − δ3 ) − (3.20P3 + 0.82 cos ϕ34 − 0.) Following (5.07 − 1.82 sin ϕ34 ) V3 V4 sin (δ3 − δ4 ) + 10.00V5 V4 sin (δ5 − δ4 ) .8).00V4 V5 sin (δ4 − δ5 ) .82 sin ϕ34 ) V4 V3 cos (δ4 − δ3 ) + (−0.19V2 V4 sin (δ2 − δ4 ) . and (5.66V4 V2 cos (δ4 − δ2 ) + 30.29 cos ϕ34 − 3. we round all numerical values to two decimal places.73V3 cos (δ3 ) 3.13 −0.07V3 sin (δ3 ) + (−0.0∠0. (This rounding does not affect model feasibility because sufficient degrees of freedom exist in the state variables.19V2 V4 cos (δ2 − δ4 ) . T35 + 36.33V2 cos (−δ2 ) − 6. the full formulation is min s.73V3 cos (−δ3 ) .13 2 T35 V32 − 6.13 V3 V5 cos (δ3 − δ5 ) .29 sin ϕ34 ) V4 V3 cos (δ4 − δ3 ) − 2.07V3 cos (−δ3 ) + 3.07V3 sin (−δ3 ) − 3.33V2 sin (−δ2 ) + 6.95V42 − 5.29 cos ϕ34 + 3. (5. To construct the formulation.1)–(5.66V4 V2 sin (δ4 − δ2 ) + (−0.9).0°.82 cos ϕ34 + 0.73V3 sin (δ3 ) + (3.00V5 V4 cos (δ5 − δ4 ) . + 3. + 6.29 sin ϕ34 ) V3 V4 cos (δ3 − δ4 ) − QG 4 − 0.82 cos ϕ34 − 0. 3. 0 = −5.13 V3 V5 sin (δ3 − δ5 ) .13 V5 V3 sin (δ5 − δ3 ) + 2.04 − 1. G G G 0. and therefore V1 is fixed to 1.19V4 V2 cos (δ4 − δ2 ) .01V42 − 30. 3.29 sin ϕ34 ) V3 V4 sin (δ3 − δ4 ) 3. 0 = 5.53 + − (3.82 cos ϕ34 + 0. + T35 G P4 − 0.

64°.20. ≈ 0. we implemented three versions of the classical formulation (5.67°.50°.053. i ∈ {2. ≈ 0.4041438. δ2 . ≈ −0. δ3 . 5} .15)–(4. ≈ 0. −180.0°. and (iii) rectangular voltage coordinates with rectangular admittance coordinates (4. V3 ≈ 0. V5 ≈ 0. with objective function value 0. T35 and the vector of state variables is x = (δ2 . Q4 . 3. −0.058.38°.387.59°. The model is publicly available in the GAMS model library [1].200. G 0.20).20 ≤ QG 3 ≤ 0. ≈ 0. If the controllable phase-shifting and tapchanging transformers are instead fixed to ϕ34 = −3. ≈ 0. Voltage angles δ1 .4016596. ≈ 0. (ii) polar voltage coordinates with polar admittance coordinates (4. δ5 ≈ −8. the optimal solution becomes V2 ≈ 0. The optimal solution for this formulation is V2 ≈ 0. For this formulation. a cost increase of approximately 0. and δ4 are restricted to one full sweep of the unit circle. δ5 . G P1 QG 1 582 583 584 585 586 587 588 589 δ2 ≈ −7. V2 . δ4 . ≈ 0.20 ≤ QG 4 ≤ 0. V5 ) . G P3 QG 3 δ3 ≈ −1. T35 ≈ 0. Q3 . V5 ≈ 0. i ∈ {2. −30.2. ϕ34 ≈ 12.40. QG 1 . G P4 QG 4 δ4 ≈ −13. P3 . V4 ≈ 0.249. V3 .86°. V4 .959. 0. ≈ 0. ϕ34 .968. The model is not yet available for download in the GAMS model library as stated here.9) in the GAMS modeling language.22°.98. To obtain the optimal solution for Example 5. G P3 QG 3 δ3 ≈ −4. 3.200.05. with objective function value 0.05.970.10 ≤ P3 ≤ 0.983.13°.192. δ5 ≈ −9. 4. ≈ 0. the model yielded 5 Note to the reviewer: the GAMS model is attached at the end of the article.6%. .30 STEPHEN FRANK AND STEFFEN REBENNACK G 0.981.0°. −0.957.05 ≤ P4 ≤ 0.18). G P4 QG 4 δ4 ≈ −8.0° and T35 = 0.127. The three versions each use a different form of the power flow equations: (i) polar voltage coordinates with rectangular admittance coordinates (4. V3 ≈ 0. ≈ −0. and branch current limits are neglected. G Slack bus generator powers P1 and QG 1 are unrestricted.0° ≤ δi ≤ 180.95. δ3 .0° ≤ ϕ34 ≤ 30. 5} .16).19)–(4.964.195. the vector of control variables is G G G G G u = P1 . G P1 QG 1 δ2 ≈ −12.5 For the example. ≈ 0. but will be finalized and made available when this paper is published.95 ≤ Vi ≤ 1.95 ≤ T35 ≤ 1.946.072. P4 .40.30.20°.17)–(4.1)–(5.950.947. V4 ≈ 0. 4. 579 580 581 0.

. f . x0 ). SCED has interesting connections to other areas of optimization. system voltages are allowed to dip further during an emergency than under normal operating conditions. xc ) ≤ 0 (5. Security-constrained economic dispatch (SCED). 5.1. Besides the classical ED formulation. NC } is the set of contingencies to consider. where NC is the number of contingencies.t. For example. SCED can be expressed in a general way as min s. they introduce NC additional sets of state variables x and accompanying sets of power flow constraints. xc ) = 0 hc (u. for each contingency c ∈ C. Typically. SCED seeks an optimal solution that remains feasible under any of a pre-specified set of likely contingency events. gc (u.12) ∀ c ∈ C. the slack bus real and reactive power are treated as state variables because they must be allowed to change for each contingency in order for the system to remain feasible. security-constrained unit commitment (SCUC). (ii) The contingency state variables xc must remain within limits. the post-contigency power flow must remain feasible for the original decision variables u: (i) The power flow equations must have a solution. and (iii) Any inequality constraints.2. The motivation for SCED is theoretically similar to that of Robust Optimization (RO) [6]. Security-Constrained Economic Dispatch. the optimal solution to SCED will be no better than the optimal solution without considering contingencies. is an OPF formulation which includes power system contingency constraints [4]. 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 where C = {1.2. operators immediately begin re-configuring the system to return all branches and buses to normal operating limits. such as branch flow limits. 5. and reactive power planning (RPP). . x0 ) ≤ 0. SCED is a restriction of the classic OPF formulation: for the same objective function.A PRIMER ON OPTIMAL POWER FLOW 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 31 identical optimal solutions using three local nonlinear solvers. increasing the stress on the remaining network. The justification for the restriction is that SCED mitigates the risk of a system failure (blackout) should one of the contingencies occur. and h are defined as the objective function. the limits on the contingency-dependent state variables xc and other functional inequality constraints are relaxed for the contingency cases compared to the base case. Apart from the contingency index. and verified as globally optimal using the global solver LINDOGlobal. g0 (u. A contingency is defined as an event which removes one or more generators or transmission lines from the power system. Special Applications. respectively.1. .6 However.3. Each contingency has a distinct admittance matrix Yc with less connectivity than the original system. SNOPT. Remark 5. SCED formulations typically have the same objective function and decision variables u as the classic formulation. . sometimes referred to as security-contrained optimal power flow (SCOPF). must be satisfied. and CONOPT. h0 (u. several other OPF variants are common in both industry and research. although 6 In SCED. MINOS. g . . optimal reactive power flow (ORPF). These include securityconstrained economic dispatch (SCED). ∀ c ∈ C. x0 ) = 0. f (u. The relaxation of system limits is justified because operation under a contingency is temporary: when a contingency occurs. and inequality constraints in the classical OPF formulation of §5. equality contstraints. In other words.

2. ∀ t ∈ T. δt ) = QG it min max δi ≤ δit ≤ δi max ϕmin ik ≤ ϕikt ≤ ϕik min max Tik ≤ Tikt ≤ Tik max Iikt (Vt . unit commitment (UC) refers to the scheduling of generating units such that total operating cost is minimized. Constraint (5. research on SCUC has accelerated only with the advent of faster computing capabilities. and other scheduling constraints. PiDown i∈ G 649 650 651 652 653 wit PiG.16)–(5. In SCUC. δt ) ≤ Iik G wit PiG. ramp rate limits.18) (5. ∀ t ∈ T. we introduce a time index t ∈ T and a set of binary control variables wit to the OPF formulation.32 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 STEPHEN FRANK AND STEFFEN REBENNACK RO typically addresses continuous uncertain parameters rather than discrete scenarios. Each wit indicates whether or not generator i is committed for time period t. ∀ ik ∈ H. G L Pit (Vt . In electric power systems operation. ∀ t ∈ T.17) (5. SCED lends itself well to parallelization and decomposition algorithms [23]. mixed-integer nonlinear programming (MINLP) problem.22) is a compact expression of (5.11) with an added time index.20) (5.13) ∀ i ∈ N. ∀ i ∈ N.min ≤ Pit ≤ wit PiG. ∀ i ∈ N. (5.max wit QG ≤ QG it ≤ wit Qi i Vimin ≤ Vit ≤ Vimax − QL it ∀ i ∈ N. UC differs from ED in that it operates across multiple time periods and schedules the on-off status of each generator in addition to its power output. Because of its complexity. ∀ t ∈ T. Current limit constraint (5. ∀ t ∈ T.2. ∀ i ∈ G. The modified formulation becomes min t∈ T i ∈ G G SU wit Ci Pit + Ci wit (1 − wi. ∀ ik ∈ K. because the constraints are separable for a fixed u. ∀ t ∈ T. ∀ t ∈ T. ∀ i ∈ G.15) (5. limits on generator cycling. ∀ t ∈ T.t −1 i∈ G − G Pit The objective function (5. In other words. δt ) = Pit − Pit Qit (Vt . Additionally.17) are modified such that uncommitted units must have zero real and reactive power generation. The generation limits (5. UC is a large-scale. In SCUC. Many UC formulations relax certain aspects of the problem in order to obtain a mixed-integer linear program (MILP) instead—for instance by using linearized cost functions. UC must address generator startup and shutdown time and costs.max ≤ G Pit − G Pi.24) s. ∀ t ∈ T.t−1 . multi-period.16) (5. (5.21) (5.min G. 5.max . ∀ t ∈ T.t−1 ) SD + Ci (1 − wit ) wi.14) (5. a power flow is applied at each time period to ensure that the scheduled generation satisfies not only the scheduling constraints but also system voltage and branch flow limits. SCUC ensures that the UC algorithm produces a generation schedule that can be physically realized in the power system. ∀ i ∈ G.22) (5.t. the formulation becomes security-constrained unit commitment (SCUC).19) (5.23) specifies positive and negative .23) (5. If the power flow equations are added to the UC problem. reserve margin requirements. ∀ t ∈ T.13) includes terms for unit startup costs C SU and shutdown costs C SD in addition to the generation costs in each time period. Security-Constrained Unit Commitment. ≤ PiUp ≥ PReserve ∀ ik ∈ L.

Zhu [37.13)–(5. respectively. Optimal Reactive Power Flow. or other form of ED. seeks to optimize the system reactive power generation in order to minimize the total system losses. 10] discusses several approximate ORPF formulations and their solution methods. Because of the scale and presence of binary decision variables. ∀ i ∈ N.24) is one of many possible formulations. One motivation for using ORPF is the reduction of the variable space compared to fully coupled OPF [9]. sometimes this constraint is written such that PReserve is a fraction of the total load in each time period.28) (5. The vector of control variables is u = P1 . V ) is identical to the classical formulation. 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 while the vector of state variables x = (δ. .26) (5. ∀ ik ∈ L.2. ∀ ik ∈ H. ∀ i ∈ G. ∀ i ∈ N. Zhu [37. Some formulations include more precise ramp limits and startup and shutdown characteristics. δ ) = QG i − Qi . UC. Reactive power planning (RPP) extends the ORPF problem to the optimal allocation of new reactive power sources—such as capacitor banks—within a power system in order to minimize either system losses or total costs. 7] and Bai and Wei [5] provide more discussion of SCUC. ∀ i ∈ N. RPP modifies ORPF to include a set of possible new reactive power sources. including detailed formulations.t.min G. δ ) = PiG − PiL L Qi (V. QG i:i∈G . Tik:ik∈K . another is the ability to reschedule reactive power to optimally respond to changes in the system load without changing the system real power setpoints. P1 . ϕik:ik∈H .32) (5.3. δ ) ≤ Iik ∀ i ∈ N. SCUC is one of the most difficult power systems optimization problems.max QG ≤ QG i ≤ Qi i Vimin ≤ Vi ≤ Vimax min max δi ≤ δi ≤ δi max ϕmin ik ≤ ϕik ≤ ϕik min max Tik ≤ Tik ≤ Tik max Iik (V. In ORPF. others include constraints governing generator minimum uptime and downtime. the system real power generation is determined a priori. P1 . Many interior point algorithms for OPF have focused specifically on ORPF [11].A PRIMER ON OPTIMAL POWER FLOW 654 655 656 657 658 659 660 661 662 663 33 generator ramp limits P Up and P Down .30) (5.27) (5. these are physical limitations of the generators.25) (5. Reactive Power Planning.24) requires a spinning reserve margin of at least PReserve . Constraint (5. The combinatorial nature of installing new reactive power sources has inspired many papers which apply heuristic methods to RPP [12]. In ORPF. the presence or absence of each new source is modeled with a binary variable. all real power load and generation is fixed except for the real power at the slack bus. 5. (5.2. Minimizing P1 is therefore equivalent to minimizing total system loss. ch. 5. The SCUC formulation (5. from the outcome of (for example) a DC OPF algorithm. A basic ORPF formulation is min s.33) ∀ ik ∈ K. Optimal reactive power flow (ORPF).29) (5. also known as reactive power dispatch or VAR control.4. ch.31) (5. Pi (V.

the goal is to assist the reader in interpreting and applying available published data. ∀ i ∈ Q. ∀ i ∈ N.42) (5. New.40) (5. Each IEEE CDF data card consists of plain text with fields delimited by character column. (5.39) (5. ∀ i ∈ N. The modified vector of control variables is New u = P1 . Tik:ik∈K .44) Some variants of RPP also include real power dispatch in the decision variables or include multiple load scenarios.41) (5. .max subject to limits Qi and Qi . The format includes sections.36) (5. New Qi is the amount of reactive power provided by each new reactive power source. branch data. or “cards”.35) (5. (5. ∀ i ∈ N.min . Zhang et al. and interchange data.max New ≤ Qi wi Qi ≤ wi QNew i Vimin ≤ Vi ≤ Vimax min max δi ≤ δi ≤ δi max ϕmin ik ≤ ϕik ≤ ϕik min max Tik ≤ Tik ≤ Tik max Iik (V. It has since been used to archive and exchange power systems test case data for the purpose of testing conventional PF and OPF algorithms.min G. 6. C 1 ( P1 ) + i∈ Q Install wi Ci . ∀ ik ∈ K. 6. 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 (5. δ ) = + QNew − QL i i G. bus. Only the title. ϕik:ik∈H . RPP optimizes with respect to uncertain future conditions— typically reactive power requirements for worst-case scenarios.34) ∀ i ∈ N. 35] review both formulations and solution techniques for RPP.7 for title data. the CDF data was exchanged among utilities by mail on paper card media. This uncertainty. This section summarizes the structure of these formats and their relationship to the classical OPF formulation. and branch data are relevant for classical OPF as described in this primer. The IEEE Common Data Format. Qi:i∈Q .t. The full specification for the IEEE CDF can be found in [33] and an abbreviated description is available at [2]. make RPP a very challenging optimization problem [34].34 STEPHEN FRANK AND STEFFEN REBENNACK A basic RPP formulation which minimizes total costs is min s. QG i:i∈G . bus data. together with the problem complexity.37) (5. and wi is a binary variable governing the decision to install each new reactive power source. By necessity.max Qi ≤ QG ≤ Q i i New. ∀ ik ∈ H. Data Exchange.38) (5.min New. Install where Ci represents the capital cost of each new reactive power source i ∈ Q. The IEEE Common Data Format (CDF) was first developed in order to standardize the exchange of PF case data among utilities [33]. δ ) ≤ Iik ∀ i ∈ G. A number of publicly available test cases for OPF are distributed in one or both of these two formats [2. The title data card is a single line which includes summary information for 7 Originally. ∀ ik ∈ L. [34. Two common formats for the exchange of power flow and OPF case data are the IEEE Common Data Format [33] and the MATPOWER Case Format [38]. δ ) = PiG − PiL QG i Qi (V.1. 38].43) Pi (V. wi:i∈Q . loss zone data.

Nominalvalued and unused fields in the data have zero entries.min QG i Vimin S gi bS i 15 99–106 16 107–114 17 115–122 18 124–127 a Optional field b 0=PQ.1 Field specification for IEEE Commmon Data Format bus data. MW MVAR MW MVAR kV p.u.) 2. 1=PQ (within voltage limits). MVAR p. the bus . p. p. Reactive Power or Min. or variable used in the classical OPF formulation given in §5.u. Some IEEE CDF fields specify final variable values (for instance.) Field 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Columns 1–4 6–17 19–20 21–23 26 28–33 34–40 41–49 50–58 59–67 68–75 77–83 85–90 91–98 Field Name Bus Number Bus Name Bus Area Loss Zone Numbera Bus Type Voltage Magnitude Voltage Angle Load Real Power Load Reactive Power Gen.A PRIMER ON OPTIMAL POWER FLOW 35 Table 6. Reactive Power or Max.0 in the branch voltage ratio field should be interpreted as a nominal tap ratio (T = 1. MVAR p.0). 2=PV (within VAR limits). However.u. Specialb Vi δi PiL QL i PiG QG i Vi (Specialc ) . The fields include a mixture of SI and per-unit quantities. respectively. The fields bus type (bus field 5) and branch type (branch field 6) specify system control methods. Real Power Gen.u. The field structure therefore has several limitations: 1. voltages Vi ) for conventional power flow. including the power base SBase in MVA.1. a value of 0. deg. Voltage Magnituded Bus Shunt Conductance Bus Shunt Susceptance Remote Bus Number Data Type Integer Text Integer Integer Integer Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Integer Units Quantity in OPF i (bus index) p. parameter. (Some fields are used indirectly via inclusion in Y . Each line within the card gives the data for a single bus or branch. these fields should be understood as a feasible or near feasible starting point rather than an optimal solution. for example. and ending with the flag characters -999. Tables 6. beginning with the characters BUS DATA FOLLOWS and BRANCH DATA FOLLOWS. The sixth column maps the field to an index.u. For OPF. (Due to rounding. The IEEE CDF format is adapted for the compact exchange of system control data rather than OPF data. Conversion of all quantities to per-unit is required prior to use in an OPF formulation. the reported solution may not be strictly feasible. voltage limits if bus type is 1 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 the case. Voltage Magnituded Min. This quirk of the specification requires some caution in processing the data. Reactive Power Base Voltagea Desired Voltage Max.u. and are therefore of limited use in OPF.2 list the IEEE CDF field specifications for bus and branch data.1 and 6. respectively. The bus and branch data cards follow. 3=Swing c Indicates target voltage magnitude for voltage-controlled (PV) buses d Gives reactive power limits if bus type is 2.max QG i Vimax .

u. MVA MVA MVA Specialb Rik Xik bSh ik max c Iik p. deg. 4=Fixed T and controllable ϕ c Conversion to per-unit current (using rated branch voltage) is required d Gives voltage tap limits or step if branch type is 2 or 3. p. Voltage Tap or Max. MW Transfere Data Type Integer Integer Integer Integer Integer Integer Numeric Numeric Numeric Numeric Numeric Numeric Integer Integer Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Numeric Units Quantity in OPF i (from bus index) k (to bus index) p. Voltage or Max. Phase Angled Max. p.u. 2=Controllable T and fixed ϕ (voltage control). deg. p. MVar Transfer or Min. MVar Transfer or Max. 3=Controllable T and fixed ϕ (MVAR control). (Some fields are used indirectly via inclusion in Y . MVar Tik ϕik min Tik ϕmin ik max Tik ϕmax ik 18 98–104 19 105–111 20 113–119 21 120–126 Numeric MW field b 0=Transmission line. p. Voltage Tap or Min. p. MVAR limits if branch type is 3.u. 1=Fixed T and ϕ. deg. MW Transfere Max.u. MVar MW p. The sixth column maps the field to an index.36 STEPHEN FRANK AND STEFFEN REBENNACK Table 6. MW limits if branch type is 4 a Optional . or variable used in the classical OPF formulation given in §5.) Field 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Columns 1–4 6–9 11–12 13–15 17 19 20–29 30–39 41–49 51–55 57–61 63–67 69–72 74 77–82 84–90 91–97 Field Name Tap Bus Number Z Bus Number Line Areaa Loss Zone Numbera Circuit Number Branch Type Branch Resistance Branch Reactance Branch Shunt Susceptance Line Rating 1a Line Rating 2a Line Rating 3a Control Bus Number Side Voltage Ratio Phase Angle Min. phase angle limits or step if branch type is 4 e Gives voltage limits if branch type is 2.u. Phase Angled Tap Step Size or Phase Angle Step Sized Min.u.1. Voltage or Min.u.2 Field specification for IEEE Commmon Data Format branch data.u. parameter.u. deg. p.

then MATPOWER case data contains nearly all the information necessary to formulate the classical OPF problem as described in §5. bus fields 14 and 15 give voltage limits Vimax and Vimin . the IEEE CDF lacks voltage limits for PV buses and reactive power generation limits at PQ buses. For example. practical aspects of Optimal Power Flow formulations. Tables 6. MATPOWER [39] is an open-source software package for MATLAB8 including functions for both conventional PF and OPF. The remaining fields are matrices.0. Given these limitations. .1. . For IEEE CDF fields which depend on the bus and branch types. in ascending order.4. If gencost is included. Like the IEEE CDF. give the coordinate pairs (P0 . The user must supply limits for these controls if they exist in the formulation. bus. The user must supply (or assume) values for the incomplete data. Conclusion. branch. the MATPOWER case structure uses a mixture of SI and per-unit quantities and specifies nominal-valued branch tap ratios as 0 instead of 1. and 6. For the reader interested in learning more. Column 1 specifies the type of cost model: 1 for piecewise linear or 2 for polynomial. . give the cost coefficients Cn−1 . The format is described in detail in the MATPOWER manual [38]. . respectively. For example. C0 in descending order. beginning with column 5. . Inc. . publicly archived IEEE CDF case data is most useful for obtaining the network structure and associated bus and branch admittance data. For PV buses. 6. The IEEE CDF lacks other data required for OPF. Columns 2 and 3 give the generator startup and shutdown costs. C ) that generate the piecewise linear cost function. the field structure prevents these data from being available. these same fields . The MATPOWER case format is a set of standard matrix structures used to store power systems case data and closely resembles the IEEE CDF. the data are sufficient for conventional PF but incomplete for OPF. baseMVA is a scalar giving the system power base SBase in MVA. However. 6. respectively. but the column structure provides a flexible description of the generator cost function. . The gencost matrix has the same number of rows as the gen matrix. The corresponding polynomial cost model is Cn−1 P n−1 + . MATPOWER case data consists of a MATLAB structure with fields baseMVA. column 4 specifies the number of coordinate pairs n of the form (P. + C1 P + C0 .2. The next n columns. In this primer. The next 2n columns. . gen.A PRIMER ON OPTIMAL POWER FLOW 721 722 723 724 725 726 727 728 729 730 731 732 733 37 and branch types govern the interpretation of certain other fields in the IEEE CDF.max . column 4 specifies the number n of polynomial cost coefficients. . The units of C are $/hr and the units of P are MW.5 describe the bus. C0 ). . 7. we have addressed the basic. therefore.min instead give reactive power generation limits QG and QG . and gen matrices. branch. including generator real power limits and cost data. as described in the table footnotes. MATPOWER makes no provision for including transformer tap ratios or phase shifting transformer angles in the set of decision variables. limits on these variables are not present in the data structure. 4. . for PQ buses. (Pn−1 . • For a polynomial cost model. and gencost. The interpretation of column numbers 4 and greater depends on the type of cost model: • For a piecewise linear cost model. i i 3. particu8 MATLAB 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 is a popular technical computing environment produced by The MathWorks.3. beginning with column 5. MATPOWER Case Format. Cn−1 ). The units are such that the cost evaluates to dollars $/hr for power given in MW.

or variable used in the classical OPF formulation given in §5. Rik 4 Branch Reactance Numeric p.) Column Field Description Data Type Units 1 Bus Number Integer 2 Bus Type Integer 3 Load Real Power Numeric MW 4 Load Reactive Power Numeric MVAR 5 Bus Area Integer 6 Bus Shunt Conductance Numeric MWb 7 Bus Shunt Susceptance Numeric MVARb 8 Voltage Magnitude Numeric p. a 1=PV.37].u. 2. parameter. 13 Min. 27. Experiment with the GAMS OPF formulations provided to accompany Ex- . The fifth column maps the field to an index. bSh ik max c 6 Line Rating (Long-term) Numeric MVA Iik 7 Line Rating (Short-term) Numeric MVA 8 Line Rating (Emergency) Numeric MVA 9 Voltage Ratio Numeric p. 29]. detailed formulations and also provide lists of relevant references. for instance [4.u. Review the survey papers on OPF from the past several decades. 10.38 STEPHEN FRANK AND STEFFEN REBENNACK Table 6.17–19]. Voltage Magnitude Numeric p. Voltage Magnitude Numeric p.3 Field specification for bus data matrix in MATPOWER case data (input fields only). for instance [11.u. Read the classical papers on OPF. parameter. These papers provide a detailed discussion of the foundations of OPF and provide context for more recent work. or variable used in the classical OPF formulation given in §5. we recommend any of the following: 1. The fifth column maps the field to an index. Xik 5 Branch Shunt Susceptance Numeric p. Review textbooks which describe the OPF problem [32.1.0 p. 10 Base Voltage Numeric kV 11 Loss Zone Integer 12 Max.) Column Field Description Data Type Units Quantity in OPF 1 Tap Bus Number Integer i (from bus index) 2 Z Bus Number Integer k (to bus index) 3 Branch Resistance Numeric p. (Some fields are used indirectly via inclusion in Y .u.1. Tik 10 Phase Angle Numeric deg.u.u. Reading the older surveys prior to the more recent ones provides insight into how OPF has developed over time. 3=Swing.4 Field specification for branch data matrix in MATPOWER case data (input fields 1–11 only). 4=Isolated b Specified as a MW or MVAR demand for V = 1.12. ϕik 11 Branch Status Binary a Conversion to per-unit current (using rated branch voltage) is required 768 769 770 771 772 773 774 775 776 777 778 larly regarding optimization algorithms than have been used for OPF. (Some fields are used indirectly via inclusion in Y .u.u. These textbooks provide clear. Quantity in OPF i (bus index) Speciala PiL QL i S gi bS i Vi δi Vimax Vimin Table 6. 3. 2=PQ. 4. 9 Voltage Angle Numeric deg.

Burchett. We hope that this primer encourages a similar level of engagement within the Operations Research community. 7 Gen. and C. Contribution to the economic dispatch problem. M. 745–751. Available: http://www. or variable used in the classical OPF formulation given in §5. Internet site. [5] X. 3 (2009).washington. 182–197. 1–7. Stott. Real Power Numeric MW PiG. pp. 697–711.u. C. 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 .W. D. parameter. IEEE Transactions on Power Systems. Available: http://gams.F. Stott. [3] O. [8] J. pp.B. 464–501. and G. Optimal load flow with steady-state security. [7] R. [6] D.edu/ research/pstca/. Reactive Power Numeric MVAR Qi 6 Voltage Setpoint Numeric p. Bulletin de la Soci´ et´ e Fran¸ caise des Electriciens. Korres. Tinney. Delkis. 53 (2011). pp. 431–447. particularly in the development of new.ee. Bright. Real Power Numeric MW PiG. install and experiment with the OPF capabilities available in MATPOWER [39]. 101 (1982). In recent years. efficient OPF algorithms. Reactive Power Numeric MVAR QG i . 8 (1962).com/modlib/modlib. Semi-definite programming-based method for security-constrained unit commitment with operational and optimal power flow constraints. Acknowledgment. IEEE Transactions on Power Apparatus and Systems. [10] H. Wirgau.2.H. which are available in the GAMS model library [1].A. pp. pp. 3722–3732.C. Happ. Further developments in LP-based optimal power flow. SIAM Review. 5 (1990).A PRIMER ON OPTIMAL POWER FLOW 39 Table 6. IEEE Transactions on Power Apparatus and Systems. Theory and applications of robust optimization. 1866–1876. We thank Kathryn Schumacher of the University of Michigan for her valuable comments and suggestions during the drafting of this primer. Carpentier. The material presented in this primer should provide a sufficient foundation for understanding the content of the references cited in this list. Column Field Description Data Type Units Quantity in OPF 1 Bus Number Integer i (generator index) 2 Gen. IEEE Transactions on Power Systems.htm.min 5 Min. Bai and H. pp. Alsac and B. MVA Basea Numeric MVA 8 Generator Statusb Binary 9 Max. Reactive Power Numeric MVAR QG i G. Real Power Numeric MW PiG 3 Gen. Caramanis. and K.min a Defaults to system power base S Base b 0 indicates generator out of service (remove from OPF formulation) 779 780 781 782 783 784 785 786 787 ample 5. Alternatively. Optimal power flow solutions. IET Generation. Transmission & Distribution. Large scale optimal power flow. Decoupled Optimal Load Flow Using Linear or Quadratic Programming. Either software will provide insight into the practical challenges of OPF. [9] G. H. 87 (1968). PAS-93 (1974). OPF has become one of the most widely researched topics in electric power systems engineering. pp. Contaxis. PWRS-I (1986).5 Field specification for generator data matrix in MATPOWER case data (input fields 1–10 only). J. REFERENCES [1] GAMS Model Library. Brown.max 10 Min. Wei. IEEE Transactions on Power Apparatus and Systems. pp. Bertsimas. Dommel and W.C. and B. [2] Power Systems Test Case Archive. Praise. Internet site.max 4 Max.1. Alsac. [4] O. The fifth column maps the field to an index.

6 (1972). Stott and E. and O. . [27] B. Hobson. 2006. 2nd ed. A. June 2-16 2005. 1996.J. MATPOWER 4. tba.E. tba (2012). McGrawHill. Optimal power flow by Newton approach. Upper Saddle River. IEEE Transactions on Power Systems. M. Stamford. [14] J. Inc. pp. 2008. in IEEE Power Engineering Society General Meeting. Hughes. Power System Security Control Calculation Using Linear Programming Parts I and II. Fundamentals of Electric Power Systems.A.W. 2011. Momoh. Rebennack. NY. [36] X. Power Sys[38] R. Thomas.1 User’s Manual. Optimization Principles: Practical Applications to the Operation and Markets of the [26] Electric Power Industry.J. Prentice Hall. and L. [28] B. Hoboken.I. 2 ed. J. Wright. Granville. 2011. Jardim. [23] W. Steponavice. C.P. 14 (1999).D. Stott. Numerical Optimization.S. Optimization in Operations Research..-P. Operation. Tu. Murillo-Sa tems Engineering Research Center (PSERC). A survey of the optimal power flow literature. Zhang.M. MATPOWER: Steady-State [39] R. Common Format For Exchange of Solved Load Flow Data. D. 2 ed. ch. 2864– 2880. 9 (1994). Issues in the path toward an RTO and standard markets. and F. [24] R. and Control. Proceedings of the IEEE. pp.H. 64–70. 2. A Review of Selected Optimal Power Flow Literature to 1993 Part I: NonLinear and Quadratic Programming Approaches. Sun. IEEE Transactions on Power Apparatus and Systems. IEEE Transactions on Power Systems. and L. An Interior-Point Method For Nonlinear Optimal Power Flow Using Voltage Rectangular Coordinates. 18 (2003). pp. 2422–2428. Optimization Of Power System Operation. [17] M. 2 ed. Quintana. . 105–111. PAS-97 (1978). pp. 105–111. IEEE Transactions on Power Systems. Nocedal and S. [31] L. A. Piscataway. 2 ed. Optimal Power-Flow Solutions for Power System Planning. and A. IEEE Transactions on Power Systems. Wiley-IEEE. [13] Eberhard Freitag and Rolf Busam. Zhang. Optimal Power Flow: A Bibliographic Survey I. Bree. Schaum’s Outline Series. in IEEE Power Engineering Society General Meeting. Energy Systems.. Schaum’s Outline of Basic Circuit Analysis. NY. pp. [15] S. Glover. New York.J. A tutorial description of an interior point method and its applications to security-constrained economic dispatch. 2005.. O’Malley.F. Frank. El-Hawary.. and R. 1998. Survey of Reactive Power Planning Methods. p. [33] Working Group on a Common Format for Exchange of Solved Load Flow Data. IEEE Transactions on Power Apparatus and Systems. pp. M. CT. Rardin. 2003. Sarma. NY. Prentice Hall. 2177–2186. A Review of Selected Optimal Power Flow Literature to 1993 Part II: Newton. Energy Systems. Vargas. pp. Inc. 1713–1731. Tolbert.E. 103 (1984).. Peschon. New York. and R. Wiley-IEEE Press. Optimal reactive dispatch through interior point method. Optimal Power Flow: A Bibliographic Survey II. [20] J. Wood and B. Power System Analysis and Design. 136–146. ´ nchez.D.40 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 STEPHEN FRANK AND STEFFEN REBENNACK [11] S. Zimmerman and C.D. Springer. IEEE Transactions on Power Systems. Wollenberg. Vannelli. M. 1. Review of Reactive Power Planning: Objectives.M. [21] J. pp.H.L. and S. pp. pp. Rau. p. Quintana. 1997.F. Brewer. 4 ed. PAS-92 (1973). and R. Zhang and L. 2010. [16] M. 1211–1218. Zimmerman.L.E. New York. pp.E. vol. Complex Analysis. A. John Wiley & Sons. NJ. 1430–1440. pp. [37] J. [18] J. 435–443. 1315–1323. Zhu. [22] J. NJ. Flueck. pp. Adapa. 2009.. Non-Deterministic and Hybrid Meth[12] ods. Tinney. Advanced Engineering Mathematics. F. [34] W. pp. John Wiley & Sons. NJ. A new parallel algorithm for security constrained optimal power flow with a nonlinear interior point method. Adapa. DC power flow revisited. 1290–1300. I. Momoh. B. [35] W. Cengage Learning.S. 13 (1998). V. and T. Formulations and Deterministic Methods. Ashley. 1916–1925. tba. [30] G. 2009.D. and Algorithms. 1–52. Hajdu. Power Generation. Overbye. 8 (1993). Murillo-Sa . tba (2012). [29] D. Tolbert. [19] J. and W.S. Springer. [32] A. Linear Programming and Interior Point Methods. Li. pp. Upper Saddle River. 762–770. B. Alsac.J. Constraints. IEEE Transactions on Power Systems. Qiu. 1 ed. Torres and V.. ´ nchez. IEEE Transactions on Power Apparatus and Systems. Huneault and F. IEEE Transactions on Power Systems. Greenberg.. 22 (2007). 24 (2009). pp. Galiana. IEEE Transactions on Power Systems.. 14 (1999). IEEE Transactions on Power Systems. [25] N. El-Hawary. NJ. 6 (1991).

26 (2011).A PRIMER ON OPTIMAL POWER FLOW 873 874 41 Operations. Planning. . 12–19. pp. IEEE Transactions on Power Systems. and Analysis Tools for Power Systems Research and Education.

Sign up to vote on this title
UsefulNot useful