You are on page 1of 8

Bulk Power System Dynamics and Control IV Restructuring, August 24-28, Santorini, Greece.

Optimization of Ancillary Services for System Security

Madeleine Gibescu Chen-Ching Liu
Department of Electrical Engineering, Box 352500 University of Washington, Seattle, WA 98195-2500, U.S.A. Abstract: The optimization method proposed in this paper is intended to aid the Independent System Operator (ISO) working under open access in calculating the necessary ancillary services for system security and reliability. A nonlinear optimization problem is formulated that enables the ISO to make least-cost decisions for purchasing ancillary services. The adequacy of ancillary services is validated by timedomain, midterm stability simulations. At the same time, the ISO is responsible for adjusting the schedules resulting from the market clearing process to eliminate congestion. The DC formulation for congestion management is contrasted with an AC formulation. This new formulation enables minimumcost reactive power market clearing, taking into consideration not only line flows, but also voltage constraints. The sequential approach congestion management followed by ancillary service selection is compared with a simultaneous approach. It is shown that, due to the coupling among control variables through unit capacity constraints, optimality can be achieved only by the ISO simultaneously handling congestion and ancillary services. Keywords: electricity market, power system security, transmission congestion, ancillary services, interior point optimization. 1. Introduction In a competitive market for electric energy, system reliability remains a critical issue. Regardless of the industry structure, a grid operating authority, subsequently referred to as the Independent System Operator (ISO), is expected to perform operational planning studies to determine the ancillary services that are needed to maintain the security level for the anticipated energy transactions, network configuration, and disturbance scenarios. The focus of this paper is the design of a decision support system to help the ISO manage the grid to achieve the following goals: least interference with the unconstrained energy market equilibrium, ensuring a reasonable level of system steady-state and dynamic security, selecting sufficient ancillary services such that the required level of system security is met at a minimum additional cost to system users. The open access environment discussed is similar to the one proposed for the California deregulation [1], which features decoupling of the market and securityrelated functions and also allows multiple separate forward markets for energy. The ISO receives as input the energy schedules resulting from an unconstrained market clearing mechanism. This work proposes that the process of clearing various ancillary service markets be done simultaneously with meeting the transmission system security constraints and the market separation constraints. A summary of the market rules in various countries follows. In California [1], congestion management is performed by the ISO, starting from the preferred schedules as submitted by all market players. The tool used is a linear program that minimizes the sum of schedule adjustments weighted by the adjustment bids, enforcing (at present) only inter-zonal flow constraints. Regarding ancillary services, day-ahead and hour-ahead markets currently exist for regulation, spinning, non-spinning and replacement reserves. The same resource capacity can be bid simultaneously in all of these markets, since the ISO evaluates the bids in each market sequentially, in the order that they were enumerated above. Allocation is based on minimizing cost for each type of ancillary service, subject to system-wide requirements and ramp-rate constraints. In UK [2-3], the National Grid Company uses supply bids and the demand forecast to establish a least-cost half-hourly schedule for generation and a system marginal price, taking into account only intact network constraints for major interfaces. The market clearing mechanism is based on a modified heuristic unit commitment program, where fuel costs have been replaced by bid prices. The output of this computation is passed on to a DC power flow based thermal security assessment module and also to an AC power flow based voltage security assessment module. Resulting constraints are then fed back to the original scheduling program, while requirements for regulation and reserve, along with their respective capacity bids are now incorporated. Reactive reserves are allocated separately, given the active power schedule, such that target voltage levels are met. Finally, the Pool Purchase Price is set, which includes an "uplift" due to the out-of-merit redispatch to satisfy transmission and generation capacity constraints and also charges for reserve, frequency control and reactive power as ancillary services.

In Norway [4], the unconstrained equilibrium is established at the market cross of supply and demand bid curves. In the case of congestion, the market is split into two areas separated by the congested interface and generation on the sending end is ramped down, while generation on the receiving end is ramped up, such that the interface flow is set at the constrained level. The market income generated by the price difference between the two areas times the interface flow is passed over to the ISO. The ISO in turn purchases the necessary generation adjustments from the regulating market, based on a merit order list. The regulating market [5] operates in a 15-minute time frame and is the mechanism used to maintain generation-load balance. After energy schedules have been set, Statnett (the national bulk transmission company) receives bid blocks for up and down regulation with associated prices and establishes the merit order list. Price is determined at the end of each hour, as equal to the most expensive block ramped up for deficit of power or cheapest block ramped down in the case of excess of power. Blocks out of merit order may be called upon in the face of transmission network limitations. Fixed compensations are awarded for frequency response and also for reactive power provision by generators operating outside a prescribed power factor range. Fast power reserves are paid a percentage of the hourly spot price for capacity reservation and the regulating market marginal price for energy output in the event of reserve activation. The New Zealand electricity market uses a DC power flow model with losses approximated as piece-wise linear functions of branch flows - to simultaneously clear the energy and reserve day-ahead markets, subject to line flow, generation capacity, and inter-temporal up and down ramp rate constraints [6]. Regulation and Frequency Response, Spinning and Supplemental Reserve, and Reactive Supply and Voltage Control are all community services required for system security [7], hence it is anticipated that the ISO will mediate in the markets for these services. Although not in operation

at this time, proposals for the creation of reactive power markets have been made, e.g. in [8]. The interdependence between the security margin, the energy schedules and the ancillary services, as well as its multi-objective and combinatorial nature, make the problem of choosing the optimal energy and ancillary service (AS) configuration difficult. The environment in which the ISO operates, together with the necessary tools for managing system security is depicted in fig. 1. A Power Exchange (PX) is a market in which suppliers and consumers place their bids for buying and selling electric energy; the PX aggregates these bids into unique supply and demand price-quantity curves and determines the equilibrium point at the market cross. Scheduling Coordinators (SCs) are also allowed to operate their own markets independent of the PX, however, they must submit balanced schedule information to the ISO. Both SCs and PX participants may submit adjustment bids to the ISO, for the purpose of congestion management, and also bids for providing ancillary services. Optimal Power Flow (OPF) formulations tailored to the needs of vertically integrated utilities were focused on minimizing objectives such as: fuel cost, losses, deviations in control device setpoints, violations in voltage levels and/or line flows, load shedding, or maximizing loadability or reactive reserve margins. An extensive review of OPF formulations and solution techniques is given in [9]. In addition to satisfying the transmission constraints for the planned network configuration, constraints for the n-1 post-contingency system configuration may be added, thus increasing the complexity of the problem, known in the literature as the security-constrained OPF [10]. Less conservative strategies take into account the possibility of operator intervention to reschedule controls immediately after an outage has occurred, while placing time-related limits on the movement of control variables [11]. New optimization formulations are needed in the new environment, as described in section 2.

Power Exchange

Preferred Schedules ISO

Congestion Management + AS Auction

AS Bids Adjustment Bids Scheduling Coordinator

Bilateral Trades

Dynamic Performance Validation

Final Schedules AS Allocation

Fig. 1 Generic Open Access Environment

2. Incorporating Ancillary Services into SecurityConstrained Optimization A new optimal power flow objective is formulated, such that deviations from unconstrained market equilibrium are minimized and also minimum cost of reactive power, regulation and spinning reserve services are incorporated. Market separation in the case of multiple forward markets is modeled as an additional constraint. The definitions and models below are based on [1], [7]. Regulation and Frequency Response requires generators on AGC to balance control area supply and demand, and maintain frequency and tie line flows at scheduled values. This ancillary service is procured competitively, by the ISO, for the shared benefit of all system users. To this effect, the portion of the objective function that deals with minimum cost regulation service procurement can be written as the sum of three terms - one for capacity reservation, one for the actual energy output from the regulating unit, and the last one representing a premium proportional to the speed of response of the chosen unit:

Pi res (MW) is the amount of reserve cleared; cires ($/MW)

is the bid price of capacity reservation for reserve;


($/MWh) is the bid price for the energy output from reserve; res (h) is the time requirement for the activation of reserve; res (0,1] is the probability of spinning reserve activation. The constraints associated with the regulation and reserve services incorporate maximum amount bids, system-wide reliability requirements, and ramp rate constraints (3-5). In (3), the simplifying assumption that up and down regulation blocks are of equal amount and have equal bid prices is made.

0 Pi reg Pi reg ,max 0 Pi res Pi res ,max




Reg _ req

C reg = (cireg Pi reg + eireg Pi reg reg + ui aireg Ri reg )




Pi res Res _ req

Pi reg Ri reg Pi res Ri res


Pi reg (MW) is the amount of regulation cleared at each bus,

each scheduling coordinator;

cireg ($/MW) is the bid price eireg

of capacity reservation for regulation purposes;

($/MWh) is the bid price of the energy output from regulation sources; aireg ($/MWh/min) is the bid price for the regulation adder, Ri is the generation ramp rate (MW/min); of response of regulation; ui is a binary decision variable, equal to 1 if provider i has been selected.

reg (h) is the time requirement for the speed

Spinning Reserve involves generation synchronized to the system (usually responsive to AGC requests), fully available in 10 minutes and sustainable for at least 20 minutes. The portion of the objective function responsible for ensuring minimum cost procurement of reserve is the sum of two terms one for capacity reservation and one for the energy output from the reserve unit, the latter weighted by the probability of supply deficit due to unplanned generator outages:

Reactive Supply and Voltage Control involves the provision of reactive power from generation sources, including the ability to continually adjust voltages in response to system changes. This is a community service, since voltagereactive power control is planned on an area-wide basis, in conjunction with transmission system based reactive resources (capacitors, inductors, SVCs). The latter are included as control variables but are not part of the objective function formulation. Since the allocation of reactive power resources determines node voltage levels, which in turn impact steady-state security, this problem is naturally coupled with the congestion management process via the nodal power balance and line flow and voltage limits. The cost of security can thus be modeled as:

Cs =

iG , jD , kQ

Q G ciG PiG + c D PjD + c k Qk j


where: ciG ($/MW) and

c D ($/MW) are the implicit bid prices for j

Cres = (cires Pi res + res eires Pi res res )



transmission by generators and dispatchable loads, Q respectively; ck ($/MVAr) are the reactive power bid
G D prices; Pi , Pj (MW) are the adjustments in supply and

demand side injections due to transmission constraints; and G Qk (MVAr) are the reactive power bid amounts. The constraints associated with the security subproblem are the reactive power bid constraints and the adjustment range constraints for both supply and demand-side,

Qimin QiG Qimax Pi



Pi Pi


Power Factor Correction means locally supplying reactive power to correct the power factor of a customer load to meet pre-specified standards. There will not be an ISOmediated power factor correction market, however the ISO must be informed of the magnitude and location of power factor correction devices, to be modeled as additional shunt reactive injections. In this work, shunt reactors and capacitors are merely incorporated in the bus admittance matrix. 2.1 Choice of Optimization Tool This work uses an off-the-shelf infeasible primal-dual interior point method applied to a sequence of quadratic approximations to the original problem [12]. A recent paper [13] obtains satisfactory results using the interior point method for economic dispatch, combining network constraints for each time period with inter-temporal ramping constraints. A reactive power dispatch for the Spanish HV network [14] was also formulated as a nonlinear programming problem, with transmission loss minimization as the objective, subject to node voltage inequality constraints, and implemented successfully using a primal-dual interior point algorithm. It has been shown [12-14] that interior point methods scale well with size, since the speed of convergence is roughly invariable and the only computation time increase with system size is due to the increased complexity of matrix operations. However, sparsity techniques for factorization, such as optimal ordering, are efficient in dealing with large Jacobian and Hessian matrices. The method used by LOQO does not require convexity, however the objective function and constraints must be smooth, i.e. twice continuously differentiable at every point visited by the algorithm. Hence, in the proposed formulation, the regulation adder term is not taken into account in (11), due to the discrete nature of the decision variables ui . 2.2 Discussion of Alternative Formulations 2.2.1 Simultaneous versus Sequential Optimization. The sequential formulation can be summarized as: T min f1 ( x ) = c x x, subject to : (13) g(x) = 0 x min x x max where the optimal solution to (13), x*, is fed as a parameter into (14): T min f 2 ( y ) = c y y, subject to :
y x

and also, nodal power balance and voltage and line flow constraints:

Pi (V , ) = Pi G Pi D Qi (V , ) = QiG QiD V




max Pline (V , ) Pline


for intact and "n-1" network configurations. Given a set of energy contracts and a set of spinning reserve, frequency regulation and reactive power ancillary service bids, the goal is to devise a minimum cost solution to all network users. Aggregating all the above-mentioned objectives into a single objective function yields:


V , iG , jD , kQ

Q G ciG PiG + c D P jD + c k Qk + j

(cires Pi res + res eires Pires res ) + (cireg Pireg + eireg Pireg reg + ui a ireg Ri reg )]


To the above-enumerated constraints, the following coupling constraints on generation capacity must be added, assuming equal ramp-up and ramp down regulation capability:

Pi + Pi


+ 0.5Pi




Pi min Pi G 0.5 Pi reg


This optimization problem belongs to the class of linear objective, non-linearly constrained, mixed discrete and continuous variable problems. Loss Compensation is the replacement of active power losses associated with a particular energy transaction on the transmission system. Since this is not a security-related service, the ISO will not operate a loss compensation market; however, the ISO must be informed of the means for providing this service, as it augments nodal power injections in power flow simulations. Treatment of losses in the proposed formulation will be discussed in section 2.2.2.

h1(x,y) = ax * + by 0 h2 ( y ) 0 y min y y max


In contrast, the simultaneous formulation attempts to solve: min f1 ( x ) + f 2 ( y ), subject to :

x, y

g(x) = 0 h1(x,y) = ax + by 0 h2 ( y ) 0 x min x x max


y min y y max Optimality can be achieved if and only if problem (15) is solved, as opposed to (13), followed by (14). 2.2.2 DC versus AC Power Flow Model. For the sake of comparison with the established California market approach, the DC model follows [15], with the market separation constraints reproduced below:

Fig. 2 PTI 20 bus Test System


G i ,k


Both SCs submit adjustment bids for congestion management at all six generator buses and also bids for all types of ancillary services. In the simultaneous approach, problem (11) is solved, subject to (3-5), (7-10), (12) and (17). In the sequential approach, problem (6) is solved first, subject to (7-10) and (17), then the resulting values for active and reactive injections are set as inputs to the minimization of (1)+(2), subject to (3-5) and (12). Table 1 shows the values of the objective function at the optimal solution point for various problem formulations. In the DC formulation (columns 1 and 3), the two terms represent congestion costs and reserve plus regulation costs, respectively. In the AC formulation (columns 2 and 4), the three terms represent congestion costs, reactive power costs and reserve plus regulation costs, respectively.
TABLE 1 OPTIMAL OBJECTIVE VALUE (CONGESTION + ANCILLARY SERVICE COSTS) ($*100) System Method Sequential DC 3 bus 21.40+ 20.43 21.45+ 19.01 AC 3 bus 19.88+0.89+ 18.39 19.93+0.89+ 17.49 DC 20 bus 307.35+ 74.33 308.15+ 68.70 AC 20 bus 390.41+55.49+ 74.15 391.89+55.29+ 68.38

D i ,k

, k = 1,2 SC 1


where SC is the total number of scheduling coordinators. In a lossy system, equality (16) cannot be enforced; however, to accommodate for losses while still maintaining market separation, the following set of inequality constraints is used instead:


G i ,k


D i ,k

, k = 1,2 SC


In addition, allocation of reactive support in order to maintain a desired range for the voltage profile can be dealt with only in an AC formulation. 3. Simulation Results The test systems used were the 3-bus system introduced in [15] and the 20-bus PTI/PSSe test system, shown in fig. 2. For the sake of brevity, detailed results are presented for the 20-bus system only. Assume two scheduling coordinators, SC1 and SC2, and three major congestion zones. The limiting constraint is the transfer capability from area 3 into area 1. Both SC1 and SC2 have generation available at buses 101, 102, 201, 206, 301 and 308 and serve loads at 6 load buses, out of which only two are shown in fig. 2. In this example, no congestion management or ancillary service bids are submitted by the demand side. All node bus voltage constraints are enforced, in addition to the seven tie Area 3 Area 2 line flow constraints, intact network configuration only. G201 G301 G308 G101


The difference between the AC and DC formulations is insignificant in the case of the 3-bus system (compare columns 1 and 2 in table 1), since this system can be closely approximated by a lossless network. The DC approximation becomes inaccurate for the larger, heavily loaded 20-bus system (column 3 versus 4 in table 1), which has nonnegligible branch resistances and significant reactive flows. Zonal prices for real and reactive energy, together with marginal values for reserve and regulation, resulting from the values of the Lagrange multipliers associated with constraints (9) and (4) at the optimal solution point, are presented in table 2. Zonal prices for each SC are defined as the weighted average of nodal prices (i for real power and i for reactive power) in a zone, with weights given by the ratio of the SC load at each node (real or reactive) to total real load in a zone: Pi ,D k Zonal _ P k = i , k SC iZone Zone _ Load (18) QiD ,k k Zonal _ Q = i , k SC iZone Zone _ Load

Area 1 G102





The real power flow pattern for both SC1 and SC2 is from area 3 into area 1 and from area 1 into area 2 (as indicated in fig. 2). It can be observed from table 2 that zonal prices increase in the direction of the power transfer (resulting in congestion charges paid by the scheduling coordinators to the ISO), with the largest price gradient between areas 3 and 1, across the congested interface. In contrast, the reactive power flows from both areas 2 and 3 into area 1.

Zonal Zone 1 Energy Zone 2 Prices Zone 3 Zonal Zone 1 Reactive Zone 2 Prices Zone 3 Marginal Value of Reserve Marginal Value of Regulation

Sequential DC 20 bus AC 20 bus SC1 SC2 SC1 SC2 6.81 9.26 15.71 16.79 7.34 9.51 15.72 17.00 3.09 5.61 10.75 12.20 24.09 21.37 N/A N/A 13.01 25.56 19.41 17.52 2.94 7.06 4.15 7.06

Simultaneous DC 20 bus AC 20 bus SC1 SC2 SC1 SC2 6.81 9.26 15.54 16.62 7.34 9.51 15.56 16.83 3.09 5.61 10.73 12.18 23.76 21.08 N/A N/A 12.83 25.21 19.16 17.29 2.94 6.7
3529.32 (MW) 420.0 (MW)

4.2 5.58
800.0 (MW)

Tables 3 and 4 present the amount of energy, reserve and regulation cleared at each bus, for each SC, for the simultaneous, respectively sequential formulations. Although the value of the congestion-related objective is roughly the same for both simultaneous and sequential formulations (last column in table 1), differences in the values of control variables at the optimal solution point, as seen in tables 3 and 4, lead to a more expensive regulation service (last row in table 2) in the sequential case. This is because the sequential formulation lacks foresight of the existence of unit capacity constraints when it performs congestion management.
TABLE 3 ENERGY, RESERVE AND REGULATION AMOUNTS (MW) CLEARED SIMULTANEOUS METHOD Gen. Energy Regulation Reserve Bus No. SC1 SC2 SC1 SC2 SC1 SC2 101 975.0 516.43 10.0 50.0 20.0 100.0 102 60.0 120.0 240.0 201 39.73 206 850.0 243.57 100.0 100.0 301 553.54 0.27 308 331.06 140.0 300.0 Total 3529.60 (MW) 420.0 (MW) 800.0 (MW) TABLE 4 ENERGY, RESERVE AND REGULATION AMOUNTS (MW) CLEARED SEQUENTIAL METHOD Gen. Energy Regulation Reserve Bus No. SC1 SC2 SC1 SC2 SC1 SC2 101 1000.0 542.94 50.0 100.0 102 240.0 201 60.0 206 850.0 251.15 100.0 45.0 100.0 301 552.33 75.0 308 332.90 150.0 300.0

4. Impact of Ancillary Services on Dynamic Security The modeling techniques for energy contracts and ancillary services for dynamic security assessment using timedomain simulations have been developed in our previous work [16]. The resulting optimal energy and ancillary service configuration can be tested for dynamic performance using the contract interpreter described in [16] and a time-domain simulator such as ETMSP [17]. The data required for each energy and ancillary service contract, in order to select appropriate device models, is described in table 5. The impact of the type and location of automatic voltage controllers on the systems vulnerability to voltage collapse is illustrated below. The 20-bus PTI system is augmented with a 10-bus, 3-machine system, known as the voltage collapse test system [18], where the two load buses are identified with buses 153 and 154 in the original PTI test system (fig. 2).
TABLE 5 - CONTRACT AND ANCILLARY SERVICES DATA SPECIFICATION Energy Contract e1. Receipt Bus e2. Delivery Bus e3. MW Amount e4. Power Factor at Delivery e5. Loss Flag e6. Percent Loss Ancillary Service Contract a1. Physical Provider Bus a2. Unit ID a3. Voltage-Reactive Power Control Flag a4. Voltage Setpoint a5. MVAr Limits a6. Frequency Regulation Flag

Compensation a7. Regulated Area Bus 152

The dynamic models used for exciters, over-excitation limiters (OXLs), and turbine-governors (TGs) in the dynamic simulation are summarized in table 6. An UnderLoad Tap Changer (ULTC) is modeled between buses 152 and 153, regulating the voltage at bus 154.
TABLE 6 ETMSP GENERATOR MODELS USED Generator Exciter OXL Bus No. Model Model 3 101 Generator Commutator 3 102 Generator Commutator 3 201 Alternator Fed 206 Self-Excited DC 301 Alternator Fed 308 Alternator Fed TG Model steam steam hydro steam Bus 151

Fig. 4 - Load Bus Voltages - Unstable Case

The voltage collapse phenomenon has been linked to the inability of generators to provide reactive support following system disturbances that occur under heavy power transfer conditions. This study shows the effect on bus voltage profiles of maintaining automatic voltage regulation at key generators in the system. Figures 3 and 4 show the timedomain simulations for generator reactive power and representative load bus voltages for the case when one generator in area 1 (at bus 102) is on manual control. The disturbance is a three-phase fault at a bus electrically close to generator 156, cleared by the opening of a major doublecircuit line linking the generator with the remote load center represented by L153 and L154. This case results in a voltage collapse, due to limited reactive support available from the generator that is switched to manual control. For the same energy contract configuration, figures 5 and 6 show the same quantities for the case where all three generators in area 1 are on AVR. The second case leads to a new stable, albeit low, steady-state voltage profile. The OXL at generator 201 operates in both cases to reduce the field current.

Generator 201

Generator 101

Generator 102

Fig. 5 - Generator Reactive Power - Stable Case (all AVR)

Bus 152

Bus 151

Generator 201

Fig. 6 - Load Bus Voltages - Stable Case (all AVR) Generator 101 Generator 102

5. Conclusions A new OPF formulation was proposed in this paper, reengineered to satisfy the requirements of a market environment. It was shown that simultaneously performing congestion management and ancillary service allocation

Fig. 3 - Generator Reactive Power - Unstable Case

leads to a least-cost solution, thus benefiting all market participants. However, dynamic security aspects cannot be incorporated into a steady-state optimization tool, hence time-domain simulations are needed to further validate the OPF results. 6. Acknowledgments The authors would like to thank Prof. Robert J. Vanderbei from Statistics and Operation Research, Princeton University, for providing LOQO, the code for primal-dual interior point optimization, and also for his assistance in navigating through its multitude of capabilities. This research was partially supported by the National Science Foundation, grant ECS-9612636. 7. References [1] WEPEX, California Independent System Operator and Power Exchange, FERC Filing, Mar. 1997. [2] T.D. Russell, A.L. Smart, "The System Operator in an Open Electricity Market: Meeting the Challenge", Paper 120-03, Impact of Open Trading on Power Systems, Proceedings of the CIGRE Symposium, Tours, France, 8-10 Jun. 1997. [3] T.D. Russell, "Evolution of Power System Operating Practice in England and Wales Post Privatization", 29th Universities Power Engineering Conference, Sept. 16., 1994, pp. 98-101. [4] I. Glende, E. Westre, Transmission Pricing in Norway, Proceedings of CIGRE, Tokyo, Japan, May 1995. [5] G. Gjerde, S.A. Fismen, T. Sletten, "The System Operator Responsibility in a Deregulated Power Market", Paper No. 110-06, Impact of Open Trading on Power Systems, Proceedings of the CIGRE Symposium, Tours, France, 8-10 Jun. 1997. [6] T. Alvey, D. Goodwin, X. Ma, D. Streiffert, D. Sun, "A Security-Constrained Bid Clearing System for the New Zealand Wholesale Electricity Market", IEEE Transactions on Power Systems, Vol. 13, No. 2, May 1998, pp. 340-346. [7] NERC, Defining Interconnected Operations Services under Open Access, IOS Working Group Final Report, Mar. 1997.

[8] J. L. Sancha, J.L. Fernandez, A. Cortes, "A Proposal for Reactive Power Ancillary Service Market", Paper 500-03, Impact of Open Trading on Power Systems, Proceedings of the CIGRE Symposium, Tours, France, 8-10 Jun. 1997. [9] M. Huneault, F.D. Galiana, "A Survey of the Optimal Power Flow Literature", IEEE Transactions on Power Systems, Vol. 6, No. 2, May 1991, pp. 762-770. [10] B. Stott, O. Alsac, A.J. Monticelli, Security Analysis and Optimization, Proceedings of the IEEE, Vol. 75, No. 12, pp. 1623-1644, 1987. [11] A. Monticelli, M.V.F. Pereira, S. Granville, SecurityConstrained Optimal Power Flow with PostContingency Corrective Rescheduling, IEEE Transactions on Power Systems, Vol. 2, No. 1, Feb. 1987, pp. 175-182. [12] R.J. Vanderbei, D.F. Shanno, An Interior Point Algorithm for Non-convex Nonlinear Programming, Statistics and Operations Research Report No. SOR-97-21, Princeton University, 1997. [13] G. Irisarri, L.M. Kimball, K.A. Clements, A. Bagchi, "Economic Dispatch with Network and Ramping Constraints", IEEE Transactions on Power Systems, Vol. 13, No. 1, Feb. 1998, pp. 236-242. [14] J.L. Martinez-Ramos, A. Gomez-Exposito, V.H. Quintana, "Reactive Power Optimization by Interior Point Methods: Implementation Issues", Proceedings of the Twelfth Power Systems Computation Conference, Dresden, Germany, 20-23 Aug. 1996, Vol. 2, pp. 844850. [15] P.R Gribik, G.A. Angelidis, R.R. Kovacs, Transmission Access and Pricing with Multiple Separate Forward Markets, PE-221-PWRS-0-111997, IEEE PES Winter Meeting, 1998. [16] M. Gibescu, C.C. Liu, Modeling of Energy Transactions and Ancillary Services for Dynamic Security Assessment, in preparation. [17] B. Gao, G.K. Morison, P. Kundur, Towards the Development of a Systematic Approach for Voltage Stability Assessment of Large-Scale Power Systems, 95SM 526-4 PWRS, IEEE PES Summer Meeting, 1995. [18] P. Kundur, Power System Stability and Control, pp. 581-627, McGraw-Hill, 1994.