You are on page 1of 120

Sevilla, Junio 2014

Analysis of Distribution
Systems with Distributed
Resources

Juan A. Martinez Velasco


Universitat Politècnica de Catalunya
Barcelona
Bibliography
J.A. Martinez and J. Mahseredjian, “Load flow calculations in
distribution systems with distributed resources. A review,” IEEE PES
General Meeting, Detroit, July 2011.
J.A. Martinez and J. Martin-Arnedo, “Distribution load flow calculations
using time driven and probabilistic approaches,” IEEE PES General
Meeting, Detroit, July 2011.
R.C. Dugan, Reference Guide. The Open Distribution System Simulator
(OpenDSS), EPRI, June 2013.
Several Power Point presentations by R. Dugan
J. Mahseredjian, S. Dennetière, L. Dubé, B. Khodabakhchian, and L.
Gérin-Lajoie, “On a new approach for the simulation of transients in
power systems,” Electric Power Systems Research, vol. 77, no. 11, pp.
1514-1520, September 2007.
L. Gérin-Lajoie and J. Mahseredjian, “Simulation of an extra large
network in EMTP: From electromagnetic to electromechanical
transients,” Int. Conf. Power Syst. Transients (IPST), Kyoto, Japan, Jun.
2-6, 2009.
Bibliography
C. Dufour, J. Mahseredjian, and J. Bélanger, “A combined state-space
nodal method for the simulation of power system transients,” IEEE
Trans. on Power Delivery, vol. 26, no. 2, pp. 928-935, April 2011.
I. Kocar, J. Mahseredjian, U. Karaagac, G. Soykan, and O. Saad,
“Multiphase load flow solution of large scale distribution systems using
the MANA approach,” IEEE Trans. on Power Delivery, vol. 29, no. 2, pp.
908-915, April 2014.
V. Spitsa, R. Salcedo, X. Ran, J. Martinez, R. Uosef, F. de León, D.
Czarkowski, and Z. Zabar, “three–phase time–domain simulation of very
large distribution networks,” IEEE Trans. on Power Delivery, vol. 27, no.
2, pp. 677-687, April 2012.
F. de León, D. Czarkowski, V. Spitsa, J. A. Martinez, T. Noda, R. Iravani,
X. Wang, A. Davoudi, G. W. Chang, A. Mehrizi-Sani, and I. Kocar,
“Development of data translators for interfacing power-flow programs
with EMTP-type programs: Challenges and lessons learned,” IEEE
Trans. on Power Delivery, vol. 28, no. 2, pp. 1192-1201, April 2013.
Introduction
 Distributed Resources (DR)
 Benefits and challenges
 Models for Load Flow calculations
 Load Flow calculations in distribution systems with DRs
 needed to assess voltage profile, power flows and losses, and
to analyze voltage regulation issues associated with DR
interconnection
 complex analysis with intermittent resources and energy
storage
 models and solution techniques depend on the simulation tool
 Tools for load flow calculations must
 combine new analysis capabilities with models for representing
the various DR technologies
 include capabilities for time driven calculations (i.e., performed
during a time period)
 support three-phase models
The Smart Grid

 Smart Grid means basically


 Communications and control
• Not represented in Distribution System Analysis
(DSA)
 Distributed Resources
• Generation, Storage, Demand Response
 Monitoring
 Protection
 Energy Efficiency
Smart Grid Features
 Distributed Energy Resources
 Generation
 Renewable Generation
 Energy Storage
 Demand Response
 Communications and Control
 AMI deployed throughout the system
 High-speed communications to Metering and Controls
 Effective Distribution State Estimation
 Improved Reliability
 Automated Fault Location
 Automated Restoration
 Planning (switch locations)
 Improve asset utilization
 Improved Energy Efficiency
 End-use Efficiency
 Delivery Efficiency
 At the Planning Stage
 Operationally - Active voltage regulation
Analysis Tools
 Impact of Smart Grid (SG) on DSA?
 What DSA framework is needed to support all
features of the SG?
 Will there be a need for DSA if everything is
monitored thoroughly?
 What could we do if we know more about the
system?
 How will merging of planning, monitoring and
DSE change DSA tools?
 What Kind of Analysis Tools are Needed?
 Expected: Convergence of distribution
monitoring and Distribution State Estimation
(DSE) into DMS
State of the Art in
Distribution System Analysis
 Full 3-phase analysis
 Some can do more than 3 phases (OpenDSS)
 Primarily peak demand capacity problem
 Static power flow
 A few perform sequential time simulations
 Tools designed for uniprocessors
 Mostly satisfactory for now
 Several tools can perform meshed network analysis
 Weakly-meshed or highly-meshed
 Many tools exploit radial nature of typical MV feeders for
computational efficiencies
 Increased demand for full network capabilities
 Harmonic analysis is an optional feature
 Frequency-domain tools are dominant in distribution
planning
State of the Art in
Distribution System Analysis

 Dynamics analysis is not common in


distribution planning
 Planning and operations tools are generally
separate modules
 LV system is often ignored
 This is changing
 Modeling of end-use loads is generally with
time-invariant ZIP models
Needs Envisioned by EPRI
 Sequential time simulation
 Meshed network solution capability
 Better modeling of Smart Grid controllers
 Advanced load and generation modeling
 High phase order modeling ( >3 phases)
 Stray voltage (NEV), crowded ROWs, etc.
 Integrated harmonics
 NEV requires 1st and 3rd
 User-defined (scriptable) behavior
 Dynamics for DG evaluations
EPRI’s Vision
 Distribution planning and distribution
management systems (DMS) with
access to real time loading and control
data will converge into a unified set of
analysis tools.
 Real-time analysis and planning
analysis will merge into common tools.
 Distribution system analysis tools will
continue to play an important role,
although they might appear in a much
different form than today.
DSA Key Challenges
 Very Large System Models
 Meshed (Looped) Network Systems
 Expanded Model
(including Transmission-Subtransmission)
 Detailed DG Modeling
 DG Protection
 Generator and Inverter Models
 Merging Planning and Real-Time Analysis
 Communication System Simulation
 Large Volume of AMI Data
 AMI-based Decision Making
 Time Series Simulations
 Regulatory Time Pressures
 Distribution State Estimation 12
Key Modelling Capabilities
 Distributed generation modeling
 Time series simulations
 Efficiency studies
 Meshed networks
 Large systems
 Parallel computing
 Distribution state estimation
 Protective relay simulation
 AMI Load data
 Modeling controllers
 Modeling comm.
 Work flow integration
Time Sequential Simulation
 Electric vehicle charging (minutes, hours)
 Solar and wind generation (seconds)
 Dispatchable generation (minutes to hours)
 Storage simulations (minutes to hours)
 Energy efficiency (hours)
 Distribution state estimation (seconds, minutes)
 End use load models (minutes to hours)
 End use thermal models (minutes to hours)
Analysis with Distributed Generation
 Voltage rise and regulation
 Voltage fluctuations
 Protective relaying and control functions
 Impact on short-circuit analysis
 Impact on fault location and clearing practices
 Interconnection transformer
 Transformer configuration
 Harmonics
 Response to system imbalances
 e.g. open-conductor faults due to failing splices
Modelling for Unbalances and
More Than 3 Phases
 Symmetrical component model and an unbalanced
phase-domain model can yield quite different
results.
 A symmetrical component model uses only positive-
and zero-sequence impedances – assumes balance
 Asymmetries yield impedances that are not
balanced between phases.
 Many distribution system analysis tools can perform
full 3-phase analysis;
 A few programs can go beyond 3-phases.
– Many circuits include multiple feeders sharing right-of-
ways
Example

54.4 A
FEEDER A Shield FEEDER B
Feeder A Feeder B

543.7 A 380.0 A A B C A B C
A A

525.2 A 364.6 A
B B

509.6 A 348.6 A
C C

515 A 519 A 518 A 357 A 359 A 359 A


%I2/I1= 3.85% %I2/I1= 4.09%

IAVG=517 A IAVG=358 A
Unbalanced model

Symmetrical Component Model


I2 = Negative Sequence
I1 = Positive Sequence
Distribution State Estimation (DSE)
 Key feature for Smart Grid
 Robust estimation more feasible with AMI,
sensors
 Transmission state estimation mature
 Barriers to DSE
 Low X/R
 Phase unbalances
 Magnitude (V, I) measurements in line sensors
 Communications latency and bandwidth
 Non-coincident samples
 Insufficient samples to make feeder observable
Large Systems

 A key capability
 5000 – 10000 bus systems are routine today
 Smart Grid requires solution of multiple
feeders simultaneously
 Goal:
 100,000 to 800,000 nodes
 Parallel computing could enable this
 Requires new algorithms
Contents
 Models for Load Flow calculations
 Power delivery components
 Power conversion components
 Load Flow Techniques
 Deterministic Load Flow methods
 Probabilistic Load Flow methods
 OpenDSS and Case Studies
 EMTP load flow and Case Studies
Studies and Models
 Distribution systems are three-phase and radial
 Single-phase models must be also considered
 Distribution systems can run at either MV or LV
 Generation units can be stochastic (intermittent and
non-dispatchable) or deterministic (non-intermittent
and dispatchable)
 Generating machine models must be included in
unbalanced three-phase load flow or when load flow
calculations are used as initialization in time-domain
tools (i.e., EMT-type tools)
 If the machine is connected via a static converter, the
converter and its control units can be also included in
the model
Studies and Models
 Conventional synchronous machine-based generating
unit is represented as a PV node
 if voltage control is not allowed, then the node becomes PQ
 internal impedance matrix must be included in unbalanced
three-phase calculations
 constraints about power ramps may be considered
 Load representation
 from the simplest voltage-independent PQ node to the most
sophisticated stochastic time-varying dynamic voltage-
dependent load curve
 Without energy storage devices, history terms are not
usually required, unless constraints about power
ramps are considered
 The states of energy storage devices at one time step
affect the states at the next step, and depend on the
states at the previous step
Models for LF Calculations
 Models are by default adequate for unbalanced load
flow calculations (i.e., three-phase models)
 Important aspect: the study zone
 In general, a single feeder plus a constant voltage source
connected at its root node
 When load flow calculations are used for initialization in
further studies (e.g., electromagnetic transients), the
substation or the distribution transformer plus the network
equivalent of the supply side may be required
 Components may be classified into two groups:
 Power delivery components: they transport energy
They can be represented by either their admittance or their
impedance matrices, depending on the solution technique
used for load flow calculations
 Power conversion components: they convert electrical
energy to other form of energy, or vice-versa
They are represented as a single multiphase terminal block
Models for LF Calculations
Lines and Cables
 Representation: Multiphase pi model with parameters
calculated at power frequency and considering the
effect of earth return currents
 This model may be divided into the series impedance matrix
and the parallel line charging capacitances
 Two-terminal model of a LV four-wire overhead line section
will have four nodes plus ground
 An underground line section consisting of three concentric
cables will have six nodes plus ground; in some cases Kron
reduction is applied and a three-phase model is derived
 Models for one-phase lines/cables can be required
 Approximate models of overhead distribution lines
neglect the shunt capacitances, or are derived from
the positive and zero sequence impedances
Models for LF Calculations
Transformers
 Three-phase transformer models, considering any
winding connection (even unusual connections)
 Approaches for modeling windings compute either the
impedance matrix or the admittance matrix
 The core may be ignored or treated as a load that can
be connected at either side of the transformer
Primary
Secondary

Winding
model The core can be modeled as
the usual linear parallel R-X
model or by empirically-deter-
Core mined nonlinear functions
model
Models for LF Calculations
Voltage Regulators
 A voltage regulator consists of an autotransformer
and a tap-changing mechanism whose position is
determined by a control circuit
 Model implementation is complex; it has to include the
control parameters of the compensator circuit (voltage
level, bandwith, R-X settings, time delay)
 Most load flow solutions use a simplified model in
which the voltage regulator is represented by an
impedance in series with an ideal transformer having
taps in the secondary
 The tap setting is determined depends on the solution
method. (e.g., backward/forward sweep method,
Newton-like solution methods)
Models for LF Calculations
Capacitors
 Usually installed to control voltage or reduce losses
 A three-phase capacitor bank may be delta- or wye-
connected (grounded or ungrounded)
 In load flow equations, a capacitor bank can be
represented as a PV node or as a constant impedance/
admittance
Models for LF Calculations
Generators
 A synchronous generator has been traditionally
modeled as a fixed voltage source (when connected to
the slack node), as a PV source (whose terminal
voltage is known), or as a PQ node (whose terminal
voltage is unknown)
 All these models neglect the internal impedances
 The internal impedances of a rotating machine unit
can have some effect on system unbalance due to the
rotor turn and the coupling between machine winding
impedances
 Internal impedances must be included when
calculating unbalanced three-phase power flows
 A synchronous generator can be represented by its
Norton or its Thevenin equivalent (see next slide)
Models for LF Calculations
Generators
 Models developed for wind power generators
 represent the machine as a balanced PQ node
 or calculate P and Q from the wind speed and the power
input curve taking into account the type of wind turbine (Q
can be set to zero)
 In case of unbalance, machine equations can be
included in the load flow equations
 Single-phase DG models: constant power factor
model, constant voltage model, or variable reactive
power model Positive sequence
voltage source

A synchronous generator can


Internal
be represented by its Norton
admittance/
or its Thevenin equivalent impedance
matrix
Models for LF Calculations
Energy Resources
 Two groups: random and non-random
 Wind
 Basically characterized by the wind speed
 Several approaches for wind speed forecast; in
general, Weibull or Rayleigh pdfs are used
 Factors that affect the wind resource: air density,
turbulence, auto-correlation factor, diurnal pattern
strength, hour of peak wind speed
 Several approaches to obtain wind power from
wind forecast
 Power generations from wind turbines can be
strongly correlated among adjacent wind farms
due to the similar wind speed at the area
Models for LF Calculations
Energy Resources
 Two groups: random and non-random
 Sun
 Characterized by insolation and site parameters:
geographical (latitude and longitude, ground
reflectance), environmental (ambient temperature)
 Insolation may be expressed as average total solar
radiation on the horizontal surface or average
clearness index (ratio of total horizontal radiation
at ground surface to extraterrestrial radiation)
 The insolation on the panel surface is composed of
direct radiation and diffuse radiation, each of
which depends on the clearness index
Models for LF Calculations
Loads
 Actual load: (i) it may be unbalanced; (ii) it is voltage-
dependent; (iii) it is random and time-variable
 Aggregate load models are usually balanced, but
single-phase tapped lines add unbalance
 Statistical characterization of loads, including phase
unbalance, derived from measurements, (assumed
Gaussian or beta distribution)
 Deterministic detailed residential models representing
individual appliances and human behavior patterns
have been developed
 The electric vehicle (EV) is another important load
 EV model has to include a battery model and a statistical
distribution of the number of plug-in vehicles in one node
Models for LF Calculations
Energy Storage Devices
 Not easy task due to the various storage technologies
 Energy storage can be useful to balance the system
 Rather simple models have been used to date
 Characteristics of an energy storage unit: energy
capacity, charging and discharging power capacity,
charging and discharging efficiency, minimum and
maximum state of charge, and control strategy
 An energy storage device must be linked to a control
algorithm that will decide the power flow and its
direction
 Simplest control strategy (with intermittent sources):
the load power is compared to the source power, and
the difference is fed or extracted from the storage unit
Load Flow Techniques
 An efficient and robust solution method must be able
to solve the load flow problem for distribution systems
with several thousands of nodes and several voltage
levels, whose topology can be radial or meshed, and
to which unbalanced loads and distributed energy
resources of random nature may be connected
 Deterministic solution techniques
Backward/forward sweep methods
Newton-Raphson type methods
Gauss-Seidel or fixed-point type methods
Load flow solver for EMT initialization
 Probabilistic methods
Numerical solution methods
Analytical solution methods
Probabilistic load flow for distribution systems
Deterministic Methods
Backward/Forward Sweep Methods
 The most popular for load flow calculations in
radial distribution systems
 First version (1967) considered only PQ nodes
and a single-phase model
 Several modifications have been introduced to
solve
 systems with a weakly meshed topology
 systems with voltage dependent loads
 three-phase systems
 three-phase four-wire systems, including neutral
grounding
 radial and meshed systems (in which only a
backward sweep is required)
 systems with distributed generation
Deterministic Methods
Newton-Raphson Type Methods
 Conventional Newton-Raphson methods may
fail to converge when solving the load flow
problem for ill conditioned systems
 Several approaches have been developed to
solve this problem
 One method was presented as a subroutine for
optimal capacitor sizing
 Up to three different algorithms derived from this
one were later proposed
 The current injection method, whose formulation
is based on the nodal current injections written in
rectangular coordinates and considering voltage-
dependent loads
 It was further modified to cope with unbalanced
three-phase distribution systems
Deterministic Methods
Gauss-Seidel Methods
 One of the first methods used a Gauss implicit Zbus
approach for solving three-phase unbalanced systems
 An optimal ordering scheme and triangular factoriza-
tion of the admittance matrix was later suggested
 A phase-decoupled solution method, also based on
the Gauss implicit Zbus approach was later presented
 This decoupling technique was further applied to
asymmetrical line and transformer models to reduce
the computation time
 Methods based on a Gauss-like approach and using a
reference loop frame have been recently presented for
the solution of three-phase unbalanced distribution
systems
Deterministic Methods
Load Flow Solver for EMTP Initialization
 Load-flow solutions used for initialization in EMTP-like
tools must be circuit-based and capable of handling
multiphase networks with arbitrary topologies and
model setups
 The objective is to use the same network as the one
used in the time-domain solution
 EMTP-like tools are based on nodal analysis
 An initial load-flow method for EMTP-like tools using
nodal analysis was presented in 1993
 Newer method uses modified-augmented-nodal formu-
lation
 Completely general and can handle any network topology
 Equations for model connections are derived automatically
without case-by-case treatment
Probabilistic Methods
 Uncertainties must be taken into consideration
 Load uncertainty is not usually very high and it can be
generally modeled by means of a normal distribution
 Proliferation of renewable resources poses new challenges;
(e.g., wind variability is very high and cannot be represented
by a normal distribution)
 Solutions: probabilistic approach, application of fuzzy
logic, interval analysis
 Probabilistic load flow
 First proposed in 1974
 Further developed and applied into power system analysis
and design
 Requires inputs specified by means of probability density
functions to obtain system states and power flows in terms of
a density function, and to reflect system uncertainties in the
outcome
Probabilistic Methods
Numerical Solution Methods
 Load flow based on the Monte Carlo method
 Main features of a Monte Carlo simulation: random
number generation and random sampling
 This technique consists of solving a deterministic
power flow for a large number of times with inputs of
different combinations, using the nonlinear form of the
load flow equations
 The use of the exact load flow equations is the reason
why results obtained from this approach are usually
taken as a reference for results obtained from other
probabilistic solutions to check their accuracy
 Main disadvantage: The large amount of computation
time it may take due to the large number of load flow
solutions that may be required
Probabilistic Methods
Analytical Solution Methods
 Main principle: Do arithmetic with density functions of
random input variables so that density functions of
random state variables and line flows can be obtained
 Difficulties: (a) load flow equations are nonlinear, (b)
input variables may be correlated
 Simplifications
 load flow equations are linearized
 power variables totally independent or linearly-correlated
 normal distribution and discrete distribution assumed for
both load and generation
 network configuration and parameter are constant
 Different methods proposed to mitigate the error
caused by the linearization of the load flow equations
 Two typical solutions: multi-linearization, quadratic
probabilistic load flow
Probabilistic Methods
Analytical Solution Methods
 Point estimation methods are approximate methods
 Efficiency can be improved by combining cumulants
and Gram-Charlier expansion theory
 This theory can express the density function of a
standardized variable of any type by a sum of corresponding
weighted standard normal distributions at different orders of
derivatives)
 The weights can be calculated through different orders of
cumulants
 Cumulative density functions can be obtained directly
without integrating probability density functions
 It has the disadvantage of the necessary linearization but it
may be generalized for dependent random variables
 For non-normal distributions, this approach may show
convergence problems; the Cornish-Fisher expansion is an
alternative without more computational burden
Probabilistic Methods
Methods for Distribution Networks
 DR connection challenges:
 the continuous and usually large variation of the
power generation from DG may result in a large
error if linearized load flow equations around the
mean values are used
 wind forecast cannot be based on a normal
distribution
 there can be power generation interdependence
between different DG units (e.g., correlated wind
power generation from adjacent wind farms), and
interdependence between generation and load
(e.g., PV generation and air conditioning)
 the analysis must be carried out over an arbitrary
time period when storage devices and some loads
(e.g., the electric vehicle) are present
Probabilistic Methods
Methods for Distribution Networks
 Modeling wind power generation interdependence is of
great importance due to the fact that generations from
wind turbines are strongly correlated among adjacent
wind farms
 Theoretical expressions need to be further developed
in order to account for the interdependence among
stochastic generations and loads
 Several works on probabilistic load flow studies in
radial distribution systems with stochastic wind
generation have been presented
 Several three-phase and single-phase load flow
versions together with the Monte Carlo approach have
been developed
 Not many works have been dedicated to study
distribution systems with energy storage units
OpenDSS Capabilities
 OpenDSS is a simulation tool for electric utility
distribution systems, implemented as both a stand-alone
executable program and a COM DLL that can be driven
from some software platforms
 The executable version adds a basic user interface onto
the solution engine to assist users in developing scripts
and viewing solutions
 OpenDSS supports frequency-domain analyses and
some new types of analyses required to meet future
needs
 Built-in solution capabilities include: snapshot and time
driven power flow, harmonics, fault current study,
dynamics, parametric and probabilisitic studies
 Other solution techniques and capabilities (e.g.,
graphics) can be added through the COM interface
OpenDSS Capabilities
 OpenDSS use a general multi-phase AC circuit model,
and includes models for distributed energy resources, or
unusual transformer configurations
 OpenDSS components are classified into two groups
Power delivery components (e.g., lines, cables, transformers and
voltage regulators), represented as a two multiphase terminal block
Power conversion components (e.g., generation units, energy
storage devices and loads), they convert electrical energy to other
form of energy, or viceversa.
 Aspects to be considered:
(i) The simulator itself, with its capabilities, its command language
syntax, and options
(ii) Input data, required to specify load and generation curves, over a
given time period, or random load and generation values
(iii) Output results that can be generated as text files, CSV files or
plots, which can display system variables or the system
performance (e.g., by showing overloaded components)
OpenDSS Capabilities
 Support tools can be used to generate load and
intermittent (wind, solar) generation curves, mainly when
the period to be analyze is too long (e.g., one year)
 Some can be also required when a probabilistic
approach is to be used for generating random values of
loads and generations according to a specific probability
density function (e.g., Weibull)
 Load and generation curves for time driven calculations
have been generated using capabilities of HOMER
(Hybrid Optimization Model for Electric Renewables)
 EXCEL has been used for generating random values
according to a given probability density function
 OpenDSS can generate a great variety of text and CSV
files, which can be easily manipulated to display
simulation results (e.g., voltages, currents, powers,
overloads, over- and undervoltages, or energy losses)
OpenDSS Capabilities

OpenDSS structure and tasks

Data Input Simulation Output Results


Distribution Equipment Databases DSS Simulation Results
(Lines, Cables, Transformers, Voltage (Snapshot Mode Simulation)
Regulators, Capacitors, Generators)
Distribution
DSS Graphical Outputs
System
Generation of Load Power Demand Simulator DSS
and Generation Power Generation Graphical
Output
Shapes (Wind, Solar) Outputs
Files

Supporting Tools for


Supporting Tools for Loadshape Generation
Visualization
(HOMER, Excel, Matlab, Others)
(TOP, Excel, Matlab, Others)
Test System
100

230 kV 80

Load (kW)
60

40

4.8 kV 799 20
724
0
722 0 6 12 18 24

712 707 Hour


701
742 713 704
7
720
6
705 702
Daily Radiation

714
(kWh/m²/d)

5
4
3
706
2
729 744 727 100
703 718
1
0
725

Power (kW)
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
80
728
Photovoltaic 730 60
100
generation
80
40
Power (kW)

60 732 708 709 1 2 3 4 5 6 7


731 Day
40

20 736
733
0
1 2 3 4 5 6 7
Day
710 734

711
735
737 738 741
740
12 70
60
Wind Speed (m/s)

9
50
Wind

Power (kW)
40
6 generation
30

3 20
10
0 0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 1 2 3 4 5 6 7
Day
Case Studies

1. Test system without generation and with a


time-varying load. The time period is one
year.
2. Test system with renewable generation.
3. Test system with only solar generation.
4. Test system with renewable generation and
energy storage.
5. Test system with renewable generation.
probabilistic load flow calculation based on
the Monte Carlo method.
Case Study 1
Maximum Node Loads
Node Max. Power (kVA) Power Factor Node Max. Power (kVA) Power Factor
799 ----- ----- 730 85 0.92
701 230 0.90 731 72 0.90
712 105 0.89 732 46 0.93
713 92 0.90 733 92 0.95
714 28 0.91 734 42 0.90
718 88 0.95 735 120 0.92
720 65 0.93 736 52 0.90
722 205 0.90 737 64 0.92
724 46 0.91 738 195 0.95
725 60 0.90 740 85 0.90
727 86 0.92 741 52 0.94
728 192 0.95 742 15 0.91
729 65 0.90 744 38 0.95
Case Study 1
 Main goal: To compare results when using
different approaches to represent loads
 voltage independent and
 constant impedances (i.e., quadratic P and Q)
 Simulation results
 differences are not very important for the kVAs
measured at Node 728
 differences may be up to 10% for the power flow
measured at the feeder head
 energy supplied during the entire year is respec-
tively 8300 and 7723 MWh with the constant PQ
model and the quadratic PQ; i.e., the difference is
a 7%
 minimum voltages with the two models are res-
pectively 0.915 and 0.924 pu
Case Study 1
Daily Profile
1.0

0.8
0.6

0.4
0.2
0.0
0 6 12 18 24
Hour

Load profiles

Monthly load
Case Study 1
60
Constant P-Q
50
Power (kVA)

40

30 Load at Node 728


20 (kVA) – Phase A
10 Quadratic P-Q

0
500 520 540 560 580 600
Hour

700
Constant P-Q
600

500
Power (kVA)

400
Power flow from the
300 substation terminal
200
Quadratic P-Q
(kVA) - Node 799 –
100 Phase A
0
1080 1085 1090 1095 1100
Hour
Case Study 2
 Two renewable generators are connected to the
system
 A 400 kW wind power generator is attached to Node
740
 A 200 kW photovoltaic generator is attached to Node
728
 In both cases, the power factor of the injected power
is unity
 HOMER capabilities were initially used to obtain the
wind speed and the solar radiation
 These profiles were later used to derive the shapes
of the power injected by both generators
 The two load models considered in the previous
study were applied again
Case Study 2
12
Wind Speed (m/s)

6
Wind resource –
3
Wind speed (m/s)
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

7
Daily Radiation (kWh/m²/d)

Solar resource –
5

3
Daily radiation
2 (kWh/m2/d)
1

0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Case Study 2
150
Wind
120
Photovoltaic
Power (kW)

90
Distributed
60
generation (kVA)
30
– Phase A
0
1800 1820 1840 1860 1880 1900
Hour

600
Constant P-Q
500
Power flow from
Power (kVA)

400

300
the substation
200
terminal (kVA) -
100
Quadratic P-Q Node 799 –
0 Phase A
1080 1085 1090 1095 1100
Hour
Case Study 2
 Differences are again about 10%
 The values of the energy supplied from
the substation are now 7336 and 6792
MWh (i.e., embedded generators take
10% of the required energy)
 Differences between models are again
of about 7%
 The connection of generators raises the
minimum voltages to 0.922 and 0.930 pu
Case Study 3
 Battery model implemented in HOMER
 Considered factors: nominal capacity, state of
charge of batteries, rated power of the rectifier,
power available from the grid

108 120 132 144 156


Hour

Profile of the electric vehicle load


(Loadshape generated with HOMER)
Case Study 3
 Battery charge is made during some hours
that remain fixed along the entire year
 In this example the charge is not completed
at the time it is stopped
 The assumption that all vehicles will charge
following the same mode and at the same
time is not realistic
 This feature is used to analyze the impact of
a combination of solar resources and large
battery charging loads
Case Study 3
 Influence of embedded generators
 EV Loadshape, see Figure
 Load peaks , see initial Table
 Two different time tables are considered
(a) nightly charging, from 10 pm until 6 am
(b) daily charging, from 8 am until 4 pm
 Embedded generators connected to nodes
728 and 740
 Only solar resource (see previous figure)
 Rated power of the photovoltaic generators
will vary depending on the new load peak
Electric Vehicle Load Substation Peak
DG Peak Hour
Load shape Load (kVA)

12:00 am
No 1813
February
Nightly
6:00 am
Yes 1697
Low (Nodes March
718, 728, 740) 12:00 am
No 2131
February
Daily
11:00 am
Yes 1867
August
11:00 pm
No 1673
June
Nightly
11:00 pm
Medium (Nodes Yes 1673
June
712, 718, 728,
730, 733, 740) 12:00 am
No 2178
February
Daily
12:00 am
Yes 1944
January
Case Study 3
 Table shows the peak value of the active power
supplied from the substation terminals
with low electric vehicle load, the peak power of the
generators is respectively 400 kW (Node 740) and
200 kW (Node 728)
with medium electric vehicle load, the peak power of
the generators is respectively 500 kW (Node 740) and
400 kW (Node 728)
 Conclusion: night loading is the best choice to
release substation burden
 Table II includes information about the time at which
the substation load is higher for each case; this can
be useful for analyzing the impact of an increasing
photovoltaic generation
Case Study 3
 Future scenarios could be different
a very high electric vehicle load during night hours could make
the system unfeasible since many of its sections would not
withstand the load, unless nightly embedded generation was
available (e.g., wind generation) or a high percentage of this
load was shifted to daily hours
a significant penetration of photovoltaic generation could also
motivate this load shift
 If energy storage is installed, then part of this solar generation
could be used during night hours, see next study
 Vehicle-to-grid (V2G): Some electric vehicles could partially
discharge stored energy to the grid and act as a distributed
source of storage
 The electric vehicle load could be mostly connected to the low
voltage grid, and its impact to higher voltage levels could be
lower than assumed here
 Exploring the feasibility of all these scenarios will require
simulation capabilities like those applied in this study
 More complicated scenarios can arise if price signals are
included in the study
Case Study 4
 Peak rated power of PV generator increased to 300
kW
 Energy storage device is connected to its terminal
 Storage specification:
rated energy storage capacity = 2000 kWh
power output rating = 300 kW
minimum storage capacity hold in reserve is 20%
 Rated power of the interconnection transformer
remains the same (i.e., 200 kVA)
 The storage device will be controlled to keep the
regulating transformer loading below 100 kVA
As the photovoltaic power decreases below 100 kW,
storage device starts discharging until the energy
level reaches minimum capacity or cycle starts again
 Ideal storage device (i.e., round trip efficiency is
100%)
Case Study 4
100

Generator
80
Per phase power
Power (kVA)

60
Transformer generation and power
40 flow through the
20 interconnection
0
transformer
0 20 40 60 80 100
Hour

100 1500
Photovoltaic
80 Generation Stored 1200
Energy
Energy (kVWh)
Power (kVA)

60 900
Per phase power
40 600 generation and stored
20 300 energy
0 0
0 20 40 60 80 100
Hour
Case Study 4

Simulation results
 The power flowing through the transformer does not
increase above 100 kW (figures show per phase
results, power factor of generation is unity)
 The charge of the energy storage device begins as
soon as the power generated by the photovoltaic
plant increases above 100 kW, and starts discharging
as soon as it decreases below 100 kW
 When the generation does not reach the minimum
value during daytime, the stored energy remains at
its minimum level
Case Study 4
 Energy storage can be installed to solve key issues in power
systems (e.g., load leveling and shifting, VAR support and
voltage regulation, relief of overloaded lines, release of system
capacity)
 The storage device has been used to level the power generation
curve and prevent transformer overloading
 The intermittent power injected by the solar array can become
non-intermittent, and a flat power flow curve across the
transformer can be obtained, although a large size storage
device will be usually required
 Energy storage can reduce the need for conventional
generation reserve when renewables are unavailable
 Other scenarios can be considered (e.g., storing the excess
available energy that has not been consumed can contribute to
an increase of the overall efficiency of the system)
Case Study 5
 Load and generation values are generally estimated
or calculated with some uncertainties
Load uncertainty is not usually very high and it can
be generally modeled as a normal distribution
Wind variability is very high, and cannot be
represented by a normal distribution
Wind power generation interdependence is important
due to the fact that generations from wind turbines
are strongly correlated among adjacent wind farms
 Probabilistic calculation may be considered
 Probabilistic load flow can be solved by using either
a numerical or an analytical method
 The numerical option uses a Monte Carlo method, a
capability implemented in OpenDSS and selected in
this work
Case Study 5
 Table shows the probability density functions
assumed for loads and generations, as well as the
parameters selected for each load and each
generator
 A Gaussian distribution with a mean and a standard
deviation of 50% and 5% of the peak load has been
assumed for all node loads and for the PV generation
 A Weibull distribution is generally assumed for the
wind speed
 This distribution is further used to derive the power
generation from the power curve of the turbine,
taking into account air density ratio
 The mean and the standard deviation of the wind
power generation are also 50 and 5% of the rated
peak power
Case Study 5

Probability Density Functions

Probability Density
Node Parameters
Function
Mean = 50% of the peak load
Loads Gaussian/Normal Standard deviation = 5% of
the peak load (see Table I)
Mean = 200 kW
Wind
Weibull Standard deviation = 5% of
generation
the mean power generation
Mean = 100 kW
Photovoltaic
Gaussian/Normal Standard deviation = 5% of
generation
the mean power generation
Case Study 5
40

30
Frequency

20
Per phase load
variation at Node 728
10
1000 samples
0
20 25 30 35 40 45
Power (kVA)

40

30
Frequency

20
Per phase wind
power variation
10
1000 samples
0
40 45 50 55 60 65 70 75 80 85
Power (kW)
Case Study 5
 Simulation results derived from 1000 and 5000
samples with and without distributed generation
 As expected, the results after increasing the number
of samples are closer to the theoretical distribution
 However, the mean values of the power from the
substation are not significantly improved
 For instance, the mean power supplied from the
substation per phase without distributed generation
is respectively 406 and 407 kVA after 1000 samples
and 5000 samples, being the standard deviations
10.6 and 9.8 kVA, respectively
 When generators are connected, the mean value of
the substation power decreases to 314 and 315 kVA
after simulating respectively 1000 and 5000 samples,
being standard deviations 12.2 and 10.6 kVA
Case Study 5
Per phase power flow at substation terminal
1000 samples
50

40
Frequency

30
Without
20

10
distributed
0
generation
370 380 390 400 410 420 430 440 450
Power (kVA)

50

40
Frequency

30
With
20
distributed
10

0
generation
270 280 290 300 310 320 330 340 350
Power (kVA)
EMTP Equations for Steady-sate
Solution
 Nodal Analysis
 It is based on Kirchhoff’s current law; the admittance matrix is
used for computing the sum of currents entering each node
Yn vn in
vn is the vector of node voltages and the members of in hold the
sum of currents entering each node.
 The network has a ground node at zero voltage.
 The network may contain voltage sources (that is, known node
voltages), so the equation must be partitioned
T
Yn vn in Ys vs vn vn vs
where Yn is the coefficient matrix of unknown node voltages vn ,
in holds the sum of currents entering nodes with unknown
voltage, Ys Yn and relates to known voltages v s .
EMTP Equations for Steady-sate
Solution
 Nodal Analysis
 Nodal equations have several limitations.
 They are restricted to modeling devices with the admittance
matrix representation of and ideal voltage sources connected to
ground.
 It is not possible to directly model branch relations such as ideal
transformer units or ideal sources without a ground node.
 Ideal transformer units are used as primitive devices for building
three-phase transformer models.
 It is not possible to model ideal switches in using a fixed rank
matrix.
 Devices with voltage and current relations cannot be directly
represented .
 These limitations can be eliminated using modified-augmented-
nodal analysis (MANA).
EMTP Equations for Steady-sate
Solution
 Augmented Nodal Analysis (MANA)
 Nodal equations are augmented to include generic device
equations
AN xN = bN Yn AN
xn contains both unknown voltage and current quantities
bn contains known current and voltage quantities
 The matrix AN is not necessarily symmetric.
 These equations can be also written explicitly as
Yn Ac vn in T T
1 xN = vn ix bN in vx
Ar Ad ix vx
 Matrices Ar, Ac, and Ad (augmented portion, row, column and
diagonal coefficients) are used to enter model equations which
are not or cannot be included in Yn.
 ix is the vector of unknown currents in device models, vx is the
vector of known voltages.
EMTP Equations for Steady-sate
Solution
 Augmented Nodal Analysis (MANA)
 The MANA formulation is generic and can easily accommodate
the juxtaposition of arbitrary component models in arbitrary
network topologies with any number of wires and nodes.
 It is not limited to the usage of the unknown variables presented
above and can be augmented to use different types of unknown
and known variables.
 Steady-State Solution
 The steady-state version is based on complex numbers, and the
equations are simply rewritten using capital letters to represent
complex numbers (phasors).
 If a network contains sources at different frequencies, the
equations can be solved at each frequency and assuming that
the network is linear (linear models are used for all devices).
 The solutions of xN can be combined to derive the harmonic
steady-state solution in the form of a Fourier series.
EMTP Equations for Steady-sate
Solution
 State-Space Analysis
 Some EMTP-like tools are based on the following equations
x Ax Bu
y Cx Du
 x is the vector of state variables, u is the vector of inputs and
y is the vector of outputs.
 Matrices A, B, C and D are the state matrices.
 These matrices can be calculated for given ideal switch
positions and piecewise linear device segments.
 Each topological change requires updating the state
matrices.
 More generic version of the state-space equations

y C x D u D1 u
EMTP Equations for Steady-sate
Solution
 State-Space Analysis
 To obtain steady-state conditions the equations can
be transformed via the Laplace operator
1
X sI A BU
s j
1
Y C sI A B D s D1 U
 Tilde-upper-case vectors are used to denote phasors
and I is the identity matrix.
EMTP Equations for Steady-sate
Solution
 Hybrid Analysis
 A generic method for formulating network equations.
 It can be used to derive other formulations and to relate and
combine different formulations methods for eliminating
numerical limitations.
 Equations characterizing an N–port
Iˆ H H Vˆ H H ˆ
V
v vv vi v va vb a
ˆ
V i
Hiv Hii Iˆ i Hia Hib Iˆ b

V̂ and Î are respectively the vector for extracted port voltages


and the vector for extracted port currents.
 The subscripts a, b, v and i are used to denote independent
voltage source voltage ports, independent current source ports,
voltage ports and current ports, respectively.
EMTP Equations for Steady-sate
Solution
 Hybrid Analysis
 The extracted ports can be of any type, linear or
nonlinear.
 The matrix H is used to relate port variables through
network connectivity.
 The ports are generic components, either capacitors,
inductors of nonlinear devices.
 The term nonlinear includes linear functions or
particular cases, such as short-circuits and open-
circuits. It can also contain entire circuits.
EMTP Equations for Steady-sate
Solution
 Hybrid Analysis
 If all extracted ports are nonlinear, then it is possible
to relate them through a nonlinear function
ˆ Iˆ , V
Φ ˆ 0
T T
with I = Iˆ v Iˆ i ˆ
and V = V ˆ
V
v i

 If all nonlinear branches are modeled as current


ports, then

ˆ
V Hii Iˆ i Hia Hib a


i
b
EMTP Equations for Steady-sate
Solution
 Hybrid Analysis
 This equation is identical to the compensation
method with
T
Zth = Hii ˆ ˆ
Vth H ia H ib Va I b
 The compensation method can only use current ports
for representing nonlinear devices, may not be
conformal to network’s topological proper-tree and
may become ill-conditioned.
 Hybrid analysis is more generic that the
compensation method and can avoid ill-conditioning
problems by remaining conformal to the topological
proper-tree through the usage of a combination of
nonlinear voltage and current ports.
EMTP Equations for Steady-sate
Solution
 Hybrid Analysis
 Automatic formulation from MANA: Hybrid analysis equations
can be automatically formulated from MANA equations.
 If is augmented to extract and include nonlinear port currents I
and V voltages, then is modified as follows
Yn Ac An
xN bN
Ar Ad 0 =
I V
ATn 0 0
An is the adjacency matrix of ports.
 Gaussian elimination is applied at this stage in to extract the
network relations between I , V and independent variables.
 The extracted equations are written symbolically as
PI MV SbN
EMTP Equations for Steady-sate
Solution
 Hybrid Analysis
 Generation of state-space equations: If nonlinear and linear
components are separated in the hybrid formulation, then
I vx H vx vx H vxi x H v x vn H v x in Vvx H v xa H v xb
Vi x H ix vx H ixix H i x vn H i x in I ix H i xa H i xb Va
I vn H vn v x H vn i x H vn vn H vn i n Vvn H vn a H vn b Ib
Vin H in v x H in i x H i n vn H in in I in H in a H inb
The hatted notation is dropped to simplify the presentation.
 This formulation allows to separate nonlinear components from
linear components.
 n stands for generic nonlinear or nodal in the particular case, x
stands for state variables, v is for voltage ports, i is for current
ports, a indicates independent voltage sources and b indicates
independent current sources. generic nonlinear.
EMTP Equations for Steady-sate
Solution
 State-Space Nodal (SSN) method
 This method combines state-space groups with nodal
(MANA) equations.
 State-space equations are redefined as follows
uit
xt ˆ x Bˆ u
A Bˆ ki Bˆ kn
t
t k t k t
unt t

yit t
Ck i Dk ii Dk in uit t
xt t
y nt t
Ck n Dk ni Dk nn unt t

 The subscripts i and n refer respectively to internal


sources (injections) and external nodal injections.
EMTP Equations for Steady-sate
Solution
 State-Space Nodal (SSN) method
 Upon combination of equations

y nt ˆ x
Ckn A ˆ u
B ˆ u
B Dkni uit ˆ
Ckn B Dknn unt
t k t k t ki it t t kn t

 This equation can be written as


y nt t y k hist Wkn unt t
where the subscript hist denotes known variables.
 When yn represents current injections (entering a group) and un
is for node voltages, then ykhist represents history current
sources and Wkn is an admittance matrix. This is called a V-type
SSN group and it is a Norton equivalent.
 When yn represents voltages and un holds currents entering a
group, then ykhist represents history voltage sources and Wkn is
an impedance matrix. This is referred to as an I-type SSN group
and it is a Thevenin equivalent.
EMTP Equations for Steady-sate
Solution
 State-Space Nodal (SSN) method
 In general, it is possible to have both types of groups (V-type
and I-type) by rewriting equation as follows

i Vn i khist v nV
Wkn
v In v khist i nI
t t t t

where the superscripts V and I denote respectively V-type and I-


type relations.
 This equation is referred to as a mixed-type group equation. It is
equivalent to equation by noticing that V-type relations are
voltage ports, I-type relations are current ports and Wkn Qn .
 This fact confirms the generality of hybrid analysis and the
theoretical foundations of the SSN method.
EMTP Equations for Steady-sate
Solution
 State-Space Nodal (SSN) method
 Steady-State Solution: It is found for initializing the time-domain
solution. In addition to the complex version of ,
 The state-space groups are solved using the complex versions
1
X H Bki Ui Bkn Un

Yi Cki H Bki Dkii Ui Cki H Bkn Dkin Un

Yn Ckn H Bki Dkni Ui Ckn H Bkn Dknn Un


where tilde-upper-case vectors are used to denote phasors, I is
the identity matrix and
1
H sI Ak
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 EMTP Load-flow module is used to compute the operating
conditions of the power system.
 It must employ a multiphase solution since the objective is to use the same
network topology and data and initialize the time-domain network.
 The network is not necessarily balanced, the applied models are usually
more complex and there could be a variation of models that must be
correctly initialized for smooth transition into time-domain.
 The sources (synchronous machines or other types of
generation) are replaced by PQ, PV or slack bus constraints.
 The loads are replaced by PQ constraints.
 All network components must provide a load-flow solution
model. In some cases it is the same as the one used in the
steady-state representation.
 The system of equations is used for representing the passive
network equations and must be augmented with load-flow
constraint equations.
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 Modified-augmented load-flow equations
A LF
N AI Δx
= fLF
LLA Ld ΔxLF

 These equations are real, since real and imaginary parts must be
represented separately for load-flow constraints.
 ALF
N is constructed from the original complex version of AN by
separating real and imaginary parts of each element
 AI is a connectivity matrix for accounting for load-flow devices
 LLA and Ld provide load-flow device constraint equations.
 The vector function fLF must be minimized (zero) for finding the
load-flow solution
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 The solution of
fLF (x) = 0
can be found using the Newton method

J ( j ) Δx( j ) = fLF (x( j ) )


where (j) is the iteration count, J is the Jacobian
matrix and x is used for updating the vector x
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 Expanded equation
( j) ( j)
Yn Ac A IL A IG 0 fn
Ar Ad 0 0 0 Vn fx
JL 0 J LI 0 0 Ix fL
YG 0 0 BGI BGE IL = fGI 3
J GPQ 0 0 J GPQI 0 IG fGPQ
J GPV 0 0 0 0 EG fGPV
J GSL 0 0 0 0 fGSL
 The left hand side matrix is the Jacobian matrix of the
solved nonlinear function fLF
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 The sub-matrices AIL and AIG are adjacency matrices
for interfacing the load (IL) and generator (IG) currents
with the corresponding passive network nodes.
 The following sub-matrices are used to include load-
flow constraints for any number of phases:
 JL and JLI for PQ-loads,
 YG, BGI and BGE for generator current equations,
 JGPQ and JGPQI for PQ-type generators,
 JGPV for PV-type generators,
 JGSL for slack-type generators (buses).
 The unknown supplementary vectors are defined as
follows: IL is for PQ-load currents, IG is for generator
currents and represents generator internal voltages.
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 The linear network coefficient submatrices are taken directly
from the real version (1) and remain unchanged.
 The right-hand side of (1) is eliminated due to the differentiation process for
the derivation of the Jacobian matrix.
 There is no need to derive separate equations for linear network devices.
 For the ideal transformer case the constraint equation is taken directly from

Vk 2 Vm2 gVk1 gVm1 fx

Vk 2 Vm 2 g Vk1 g Vm1 f x( j ) 2
 This means that the existing submatrices Ar and Ac are not affected and
already constitute the Jacobian terms allowing to account for all single-
phase and three-phase transformer configurations.
 When the tap positions are not defined, it is also possible to represent the
transformation ratio g in (2) as a variable and introduce additional Jacobian
terms in (3).
 All submatrices used for the construction of are directly reused in (1).
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 Initialization from Load-Flow Equations
 It is performed by reverting (3) to its linearized version (1).
 All loads are replaced by equivalent impedances calculated
at nominal voltages and all generators are replaced by ideal
sources behind their impedances.
 This approach yields sufficiently good initial conditions for
the fast convergence of (3).
 Initialization from Steady-State Solution
 Upon convergence of (3), all steady-state phasors become
available.
 The synchronous machine phasors are used to calculate
internal state variables.
 The asynchronous machine requires the calculation of slip
for a given mechanical power or torque.
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 Initialization from Steady-State Solution
 The steady-state module starts with the load-flow solution
and replaces all devices by lumped equivalents to proceed
with a phasor solution.
 This is achieved with the complex version of (1).
 The resulting solution is the same as with the load-flow
module, except that now the device models have access to
internal phasors for proceeding with initialization.
 Steady-state solution phasors are used for initializing all
state-variables at the time-point .
 The solution at t = 0 is only from the steady-state and all
history terms for all devices are initialized for the first
solution time-point.
Multiphase Load-Flow Solution and
Initialization in EMTP-like Tools
 Initialization from Steady-State Solution
 If the network contains harmonic sources or nonlinearities, it
is necessary to perform a harmonic load-flow.
 Finding the harmonic steady-state solution can have a
significant impact on computing time under some particular
conditions.
 When the network is linear, then the initialization with
harmonics through the linear steady-state constitutes a
simple superposition of all harmonic solutions.
 In some special conditions (e.g., different rotor frequencies),
initialization is possible by solving equations independently.
 Complex and yet unsolved problems
 Initialization with power electronic devices
 Initialization of control systems
Case Study
 The following case study was carried out with the
EMTP-RV
 The presentation summarizes
 The main features of the Test System
 The Modelling Approach followed in the implementation of
the test system in EMTP-RV
 The development of Power-Flow to EMTP-RV Translator
 Simulation Results and a Comparison between results
derived from a power flow simulator and the EMTP-RV
Sources: V. Spitsa, R. Salcedo, X. Ran, J. Martinez, R. Uosef, F. de León,
D. Czarkowski, and Z. Zabar, “three–phase time–domain simulation of
very large distribution networks,” IEEE Trans. on Power Delivery, vol. 27,
no. 2, pp. 677-687, April 2012.
F. de León, D. Czarkowski, V. Spitsa, J. A. Martinez, T. Noda, R. Iravani,
X. Wang, A. Davoudi, G. W. Chang, A. Mehrizi-Sani, and I. Kocar,
“Development of data translators for interfacing power-flow programs
with EMTP-type programs: Challenges and lessons learned,” IEEE Trans.
on Power Delivery, vol. 28, no. 2, pp. 1192-1201, April 2013.
Test System
Test System
 Electrical power is supplied to the area substation that has a
number of parallel transformers equipped with tap changers.
 A backup transformer (TR13) exists which operates when one of
the main transformers is out of service.
 The tap changers have scheduling which depends on active
power output of the transformers to the loads.
 The transformers are gathered into groups and can be
synchronized by closing substation breakers at the
synchronization (SYN) buses.
 There are dozens of primary feeders connecting the area
substation to a secondary grid through network transformers.
 The feeders have breakers equipped with overcurrent relays.
 The network transformers operate with fixed taps.
Test System
 Network protectors (NPs) are installed on the secondary side of
each one of the network transformers.
 NPs prevent backfeeding from the secondary grid into the
primary network.
 The secondary grid represents interconnections of underground
cables and overhead lines supplying electrical loads.
 Some distribution networks can have even more complex
architectures since they include radial feeders, spot networks,
and secondary grids at different voltage levels linked together by
additional network transformers.
 Manual modeling of such large and complicated networks using
the EMTP graphical user interface (GUI) is impractical.
 An automatic modeling approach was developed at the
Polytechnic Institute of New York University, Brooklyn, NY, to
create models of distribution networks having very large size.
Test System
Actual distribution network that supplies power in Manhattan, NYC.
The network has more than 2503 branches.
Total number of three-phase nodes = 2333.
Transmission-side voltage at area substation = 69 kV.
The primary operates at 13.8 kV whereas the secondary grid
consume power at voltage levels of 120/208 V.
Test System
Operational conditions are summarized in the table II.
Active and reactive power demands given in the table describe an
aggregate load of the distribution network.
Total active and reactive powers stand for power consumption from
the transmission system.
Modelling Approach
 EMTP-like tools use a graphical user interface (GUI) to translate
the graphical model into a description language called netlist.
 The netlist provides the connectivity of the design.
 The EMTP-RV GUI was used only to derive detailed prototype
models for each group of network elements (i.e., one model for
all network transformers, one model for all breakers, etc.).
 The created prototype models were converted into the subcircuit
netlist files and the netlist generation process was automated.
 The input text data files of the load-flow program, which contain
specifications of network elements and connectivity information,
were used to calculate parameters, update subcircuit prototype
models, and reproduce network architecture in MATLAB.
 A special script PF-EMTP Translator was developed.
 The translator output is a complete network netlist which can be
automatically loaded into the EMTP to perform the simulation.
 The correctness was verified by comparing results derived from
the EMTP-RV to those calculated using the load-flow program.
Modelling Approach
 The following prototype models were derived:
 Area substation transformer with tap changers
 Network transformers
 Unit substation transformers
 Intermittent energy resources
 Circuit breakers
 Network protectors
 Overcurrent protection
 Overvoltage protection
 Undervoltage protection
 Directional power protection
 Directional overcurrent protection.
 Some built-in EMTP models were adopted:
 RLC branches
 PI-sections
 Grounding zigzag transformers
 Electrical loads
 Ideal switches
 Synchronous machines
 Induction machines.
A flowchart of the data
translation from a power
-flow database to EMTP
Block diagram of the
PF-EMTP translator
Simulation and Validation
 Time-domain models must be validated because even small
modelling errors may result in significant changes in steady-
state and dynamic behaviour of large complex systems.
 All developed custom models were tested individually using
time-domain simulations.
 A tool was written in MATLAB to compare connectivity of the
system, branch impedances, currents, and node voltages
calculated using the EMTP time-domain simulations and PVL
load-flow program.
 Comparison results for peak loading are shown in the figures.
 A very good match exists between the results since most of the
relative differences in the rms values of the primary feeder
currents shown in the first figure are within 1%. The maximum
relative difference is 3.5%. It corresponds to the absolute current
difference of only 0.07 A.
 The maximum absolute difference of 0.92 A introduces a 0.6% of
the relative error in the rms value of the primary feeder current.
Simulation and Validation
 Differences in the secondary grid currents are given in the
second figure; most differences are within a few percent.
 The largest relative difference of 18.1% corresponds to 0.64 A of
the absolute current difference.
 The maximum absolute difference is 18 A, but it makes up
slightly less than 1% of the relative difference.
 The small differences are attributed to two reasons:
 numeric inaccuracy of the specific database impedances used for the netlist
generation
 the display of few significant digits in the load-flow results of PVL.
 The maximum relative difference of voltages in all nodes is
0.029%. Similar comparison results were obtained for the light
loading case proving the validity of the derived dynamic model.
Simulation Results
Comparison of primary currents calculated using a commercial PF
program and the EMTP time-domain simulator
Simulation Results
Comparison of secondary currents calculated using a commercial
PF program and the EMTP time-domain simulator
Simulation and Validation
 The validation of the EMTP simulation results against the power-
flow program output has been carried out for the cases of three-
phase short circuits in the primary and secondary subnetworks.
 Three fault locations were chosen in the primary subnetwork:
 head of the feeder (area substation bus);
 middle of the feeder branch 2;
 end of the feeder branch 3 (primaries of most distant network transformer).
 Three-phase short circuits were simulated at service boxes and
transformer vaults in the secondary subnetwork.
 A comparison of the fault currents is given in the table.
 The maximum relative difference between node voltages was
below 0.05% for all of the cases of the three-phase short circuits
in primary and secondary subnetworks.
Simulation Results
Comparison of the Tree-Phase Short-Circuit Currents

Faults in the Primary Subnetwork


Power Flow EMTP Current, Relative
FAULT LOCATION
Current, [kA] [kA] Difference, [%]
Beginning of the feeder 30.5 30.7 0.65
Middle of the feeder
19.3 19.7 2.07
branch 2
End of the feeder
19.8 20.3 2.52
branch 3
Faults in the Secondary Subnetwork
Power Flow EMTP Current, Relative
FAULT LOCATION
Current, [kA] [kA] Difference, [%]
Service Box 14 14.5 3.57
Vault 81.4 84.6 3.93
Simulation and Validation

 The load-flow solution of the EMTP-RV was not used because


this tool does not have steady-state models (for use in the load
flow) of network protectors.
 The time-domain steady-state results of the EMTP are compared
at every step of the automatic network reconfiguration.
 The power-flow program is capable of updating the system
model iteratively based on the status of every network protector
in the network.
 From the obtained currents and voltages, it was possible to
confirm that hundreds of these devices operate correctly in time
domain and in accordance with PVL.
Conclusions
 Load flow methods for distribution systems with DR
penetration
 Important aspects: models and solution techniques
 Three-phase modeling is generally required (e.g., due
to unbalance)
 A reliable load flow method should be capable of
solving large three-phase unbalance distribution
systems with any topology
 Advanced load flow packages have to
 consider the random nature of some inputs (load,
generation)
 perform the calculations over a period of time
 Not much experience is presently available on the
solution of systems in which all the aspects and
models analyzed are adequately represented
Conclusions
 The application of OpenDSS to Load Flow
calculations in distribution systems has been
presented
 OpenDSS offers a flexible and powerful plat-
form for load flow analysis with capabilities
for special studies
 Capabilities implemented in other tools (e.g.,
HOMER) can be used for preparing the input
file required in time driven calculations.
 Some basic tools were developed by the
authors to help users with the preparation of
the input file
Conclusions
 The application of an EMTP-like tool for load
flow calculations in distribution systems has
been presented
 This application is based on the development
of a PowerFlow-EMTP translator
 EMTP models derived from power flow data-
bases were accurate enough
 Recommendations
 Data Inventory
 Multi-File Input – Single-File Output
 Data and Model Verification
 Automated Data Validation
 Automatic Post-Processing of Results
Conclusions
 Test studies have shown the importance that new
modeling and simulation capabilities
(i) load must be adequately represented (i.e., by considering a
correct voltage dependence), since significant differences may
result from different load models
(ii) energy storage can be used, among other applications, to
level intermittent power generation and component overloading
(iii) exploring the impact of future power demand (e.g., the
electric vehicle) requires capabilities for synthesizing realistic/
accurate loadshapes
(iv) a time driven calculation is a common feature to all these
studies
 An adequate representation of loads and generations
in probabilistic studies is an open research field
 Many other aspects could be considered (e.g., certain
loads can show similar behavior and there can be
correlation between active and reactive powers; there
can also be correlation between load and generation,
or between some generators)

You might also like