SlideShare a Scribd company logo
Ph.D. Pre-Registration Seminar
Presented by
Abhinav Kumar
Roll: 185EE08
Under the Supervision of
Dr. S.K.Mallik
National Institute of Technology Patna
Ph.D. Full-time
Department of Electrical Engineering
November 16, 2023
Presentation Outline 2
Coursework Status
Introduction
Literature Survey
Research Gap
Research Objectives
Works Undertaken
References
Coursework Status 3
Coursework Status
▶ Courses completed at NIT Patna:
ˆ Advanced Electrical Engineering Lab II
ˆ Research Methodology
ˆ SCADA and Energy Management
ˆ Advanced Electrical Engg Lab-I
ˆ Seminar and Technical Report Writing
▶ CGPA: 7.81 for 16 credit course
Introduction 4
Load Modelling
▶ A simplified mathematical model and an equivalent circuit to represent the
entire load connected at any load bus terminal [1].
▶ Why?
ˆ Load modeling is needed for continuous monitoring of load variation with respect
to generation and transmission.
Introduction.... 5
▶ Types of load models [2].
ˆ Static load models
* Static models express the active and reactive power at any instant of time as func-
tions of bus voltage magnitudes and frequency
* ZIP load model, exponential load model, frequency-dependent load model are types
of static load models.
ˆ Dynamic load models
* Dynamic load models express the active and reactive powers as a function of voltage
and time.
* Induction motor, exponential recovery load model, non-linear tansig function are
types of dynamic load models.
ˆ Composite load models
* Composite load model is the combination of static and dynamic load models
Introduction.... 6
▶ Methodologies
ˆ Measurement-based approach
* Method which takes the measurement of a load terminal as an input and apply the
parameter tuning algorithm for parameter estimation.
ˆ Component-based approach
* Method which classify the load as its load class then estimate the parameters of
load models
ˆ ANN-based model
* The load model which is designed by updating the weights of ANN model by
tracking true load characteristics.
Introduction.... 7
▶ Parameter tuning algorithms
ˆ The algorithm to estimate the load model parameter by minimizing the difference
between estimated and true load characteristics.
ˆ In literature, there are number of parameter tuning algorithms are investigated,
namely, Genetic Algorithm, Improved Particle swarm optimization, Support vec-
tor machine, Variable Projection method etc.
Literature Survey
Literature Survey 9
S.No. Title Remark Research Gap
1.
J. Gil-Aguirre, S. Perez-Londono, and J. Mora-Florez,“A measurement-
based load modelling methodology for electric vehicle fast-charging sta-
tions,” Electric Power Systems Research, vol. 176, p. 105934, 2019 [3]
.
Classify the load
model which
is suitable for
EVFCS
Dynamic and
composite load
models are not
considered
2.
M. E.-N. Jahromi and M. T. Ameli, “Measurement-based modelling of com-
posite load using genetic algorithm,” Electric Power Systems Research, vol.
158, pp. 82–91, 2018 [4]
.
Parameter Identifi-
cation of composite
load based on GA
Estimation
will be more
profound with
latest tuning
techniques.
3.
V. Vignesh, S. Chakrabarti, and S. C. Srivastava, “Power system load mod-
elling under large and small disturbances using phasor measurement units
data,” IET Generation, Transmission
Distribution,vol. 9, no. 12, pp. 1316–1323, 2015 [5]
.
Parameter Identifi-
cation of compos-
ite load under large
and small distur-
bances
Estimation
will be more
accurate with
deep learning
techniques.
4.
S. M. H. Rizvi, S. K. Sadanandan, and A. K. Srivastava, “Real-time zip
load parameter tracking using sensitivity-based adaptive window and vari-
able elimination with realistic synchrophasor data,” IEEE Transactions on
Industry Applications, vol. 57, no. 6, pp. 6525–6536, 2021 [6]
.
Parameter Identifi-
cation of ZIP load
in Real-time
Estimation and
noise cancel-
lation will be
done with latest
techniques.
Literature Survey.... 10
S.No. Title Remark Research Gap
5.
J. Xie, Z. Ma, K. Dehghanpour, Z. Wang, Y. Wang, R. Diao, and D. Shi,
“Imitation and transfer q-learning-based parameter identification for com-
posite load modeling,” IEEE Transactions on Smart Grid, vol. 12, no. 2,
pp. 1674–1684,2020 [7]
.
Parameter Iden-
tification of com-
posite load in
simulation
Identification
should be done
with deep learn-
ing methods.
6.
H. Renmu, M. Jin, and D. J. Hill, “Composite load modeling via measure-
ment approach,” IEEE Transactions on power systems, vol. 21, no. 2, pp.
663–672, 2006 [8].
Parameter Iden-
tification of com-
posite load in
simulation
noise cancel-
lation should
be applied for
accuracy.
7.
L. Chavarro-Barrera, S. Perez-Londo ´ no, and J. Mora-Fl ˜ orez, “An adap-
tive approach for dynamic load modeling in microgrids,” IEEE Transactions
on Smart Grid, vol. 12, pp. 2834–2843, 2021 [9]
.
Parameter Iden-
tification of Dy-
namic load in
microgrid
static and com-
posite load
model should
also be consid-
ered
9.
G. A. Barzegkar-Ntovom, T. A. Papadopoulos, and E. O. Kontis, “Robust
framework for online parameter estimation of dynamic equivalent models
using measurements,” IEEE Transactions on Power Systems, vol. 36, pp.
2380–2389, 2021 [10]
.
Parameter Iden-
tification of dy-
namic load using
measurements
Static and dy-
namic equivalent
should be taken
Literature Survey.... 11
S.No. Title Remark Research Gap
8.
F. Conte, F. D’Agostino, S. Massucco, F. Silvestro, C. Bossi, and M. Cabi-
ati, “Experimental validation of a dynamic equivalent model for micro-
grids,” 2020 IEEE International Conference on Environment and Electrical
Engineering and 2020 IEEE Industrial and Commercial Power Systems Eu-
rope (EEEIC / I
CPS Europe), pp. 1–6, 2020 [11]
.
Parameter Iden-
tification of dy-
namic load in
microgrid exper-
imentally
static and composite load
model should also be con-
sidered in simulation.
10.
M. M. Conner, M. R. Ebinger, and F. F. Knowlton, “Evaluating coyote
management strategies using a spatially explicit, individual-based, socially
structured population model,” Ecological Modelling, vol. 219, pp. 234–247,
2008 [12].
coyote optimiza-
tion methodolgy.
coyote strategies should
be applied to estimate
dynamic load model.
11.
A. Pai, Energy function analysis for power system stability. Springer Science
& Business Media, 1989 [13]
.
Complete clarifi-
cation of NE 39
bus system.
should be used to analyze
static,dynamic and com-
posite load models.
12.
M. Srivastava, C. L. Anderson, and J. H. Freed, “A new wavelet denoising
method for selecting decomposition levels and noise thresholds,” IEEE ac-
cess, vol. 4, pp. 3862–3877, 2016. [14]
.
Wavelet de-
noising method
for noise can-
cellation of
measurements
wavelet denoising
method should be
applied to denoise the
load measurements
Literature Survey.... 12
S.No. Title Remark Research Gap
13.
Hiyama T, Tokieda M, Hubbi W and Andou H 1997 Artificial neural network
based dynamic load modeling IEEE Trans. Power Syst. 12 1576–83 [15]
.
ANN based dy-
namic load mod-
eling
more advanced
deep learning
method should
be used for load
modeling.
14.
R. C. Staudemeyer and E. R. Morris, “Understanding lstm–a tutorial
into long short-term memory recurrent neural networks,” arXiv preprint
arXiv:1909.09586, 2019 [16]
.
Understanding
the concept of
long-short term
memory deep
learning method
LSTM is not
used regression
method for ex-
ponential load
model.
15.
F. A. Gers, J. Schmidhuber, and F. Cummins, “Learning to forget: Con-
tinual prediction with lstm,” Neural computation, vol. 12, no. 10, pp.
2451–2471, 2000 [17]
.
Application of
LSTM method
LSTM should be
used as param-
eter tuning al-
gorithm of load
models.
16.
K. Hatipoglu, I. Fidan, and G. Radman, “Investigating effect of voltage
changes on static zip load model in a microgrid environment,” in 2012 North
American power symposium (NAPS). IEEE, 2012, pp. 1–5 [18]
.
Stability of mi-
crogrid in which
zip load is used
as load
Dynamic load is
not considered.
Literature Survey.... 13
S.No. Title Remark Research Gap
17.
C. Zheng et al., ”A Novel Equivalent Model of Active Distribution Net-
works Based on LSTM,” in IEEE Transactions on Neural Networks and
Learning Systems, vol. 30, no. 9, pp. 2611-2624, Sept. 2019, doi:
10.1109/TNNLS.2018.2885219 [19]
.
Designing Ac-
tive Distribution
network using
LSTM
LSTM is not
used as regres-
sion for load
modelling
18.
C. Wang, Z. Wang, J. Wang, and D. Zhao, “Svm-based parameter identifi-
cation for composite zip and electronic load modeling,” IEEE Transactions
on Power Systems, vol. 34, no. 1, pp. 182–193, 2018 [20]
.
parameter es-
timation using
Support vector
machine
SVM technique
should be used
for static load
modelling
19.
X. Wang, Y. Wang, D. Shi, J. Wang, and Z. Wang, “Two-stage wecc com-
posite load modeling: A double deep q-learning networks approach,” IEEE
Transactions on Smart Grid, vol. 11, no. 5, pp. 4331–4344, 2020 [21]
.
parameter es-
timation of
WECC load
using double
deep Q learning
system is very
complex and
time taking
20.
H. Yang, J. Jiang, G. Chen, and J. Zhao, “Dynamic load identification
based on deep convolution neural network,”Mechanical Systems and Signal
Processing, vol. 185, p. 109757, 2023 [22]
.
dynamic load
modeling using
deep convolu-
tional network
static load
model should
also be consid-
ered.
Research Gap 14
▶ A number of load model is available in literature which needs to be investigated.
▶ For accurate estimation of load model, there should be parameter tuning. For
this purpose, a number of parameter tuning algorithm needs to be investigated.
Research Objectives 15
▶ Investigation of ZIP load using opposition based differential evolution optimiza-
tion.
▶ Investigation of exponential load model using Long-short term memory method.
▶ Investigation of dynamic load modeling using dimensional-growth coyote opti-
mization.
Measurement-based ZIP load modelling using
Opposition based differential evolution optimization
Measurement-based ZIP load modelling using opposition based differential evolution optimization... 17
▶ ZIP load model is static
load model which shows volt-
age power relationship at the
load terminal as shown in (1)
▶ It constitutes of constant
impedance load, constant
current load and constant
power load.
▶ Vm and V0 are measured and
initial voltage at load bus
terminal.
Pzip = Pz

Vm
V0
2
+ Pi

Vm
V0

+ Pp
Qzip = Qz

Vm
V0
2
+ Qi

Vm
V0

+ Qq









(1)
Measurement-based ZIP load modelling using opposition based differential evolution optimization... 18
Parameters Definition
Pz, Pi, Pp
Proportional coefficients of the constant
impedance, constant current, and con-
stant power in static active load.
Qz, Qi, Qq
Proportional coefficients of the constant
impedance, constant current,and con-
stant power in static reactive load.
V True Load
Characteristics
Load Model
Parameter Tuning Algorithm
+
-
Y=[P Q]
Yest=[Pest Qest]
e
Identification
+
+
Noise (Pn, Qn)
Fig 1. Block diagram of Measurement-based
approach
Measurement-based ZIP load modelling using opposition based differential evolution optimization... 19
Parameter Tuning Algorithm
▶ Opposition based differential evolution optimization (ODEO) is used as the
parameter tuning algorithm.
▶ The steps covered in this technique is mentioned below
ˆ Generate a random population as trial vectors.
ˆ Perform mutation to generate donor vectors.
ˆ Perform crossover to generate target vectors.
ˆ set the jumping coefficients to jump the solution when it attains local minima.
Measurement-based ZIP load modelling using opposition based differential evolution optimization... 20
Estimated value of ZIP load parameters at
load bus 30
Parameters True ODEO GWO PSO
Pp 0.4 0.42380 0.45284 0.45506
Pz 0.3 0.28620 0.25478 0.25965
Pi 0.3 0.28980 0.29308 0.28593
Qq 0.4 0.43860 0.50595 0.40291
Qz 0.3 0.24720 0.24272 0.32433
Qi 0.3 0.31920 0.25243 0.27262
Performance of tuning algorithm for case1
Methods No. of iterations Time/iteration(sec.) Accuracy
ODEO 859 3.8 4.84 × 10−5
PSO 1620 3.77 4.9719 × 10−5
GWO 1630 4.07 4.9 × 10−5
Measurement-based ZIP load modelling using opposition based differential evolution optimization... 21
Fig 2. Active power of ZIP load Fig 3. Reactive power of ZIP load
Measurement-based ZIP load modelling using opposition based differential evolution optimization... 22
▶ In this work, a measurement-based approach is employed for estimating the pa-
rameters of the ZIP load using a ODEO, PSO, GWO parameter tuning method.
▶ These algorithms are tested and validated on NE 39-bus system.
▶ The results confirm the superior performance of proposed ODEO in terms of
parameter estimation, convergence speed, and accuracy compared to PSO and
GWO.
Exponential load modelling using long-short term
memory (LSTM) back-propagation deep learning
method
Exponential load modelling using LSTM 24
▶ Exponential load model is
voltage-power relationship at
load bus as shown in (2)
▶ The nature of load varies
with variation in α and β.
▶ For composite load system,
the value of α and β usually
ranges between 0.5 and 1.8 
1.5 and 6 respectively.
Pexp =

Vm
V0
α
Qexp =

Vm
V0
β









(2)
▶ α and β are the parameters of the load model.
▶ Vm is the measured voltage and V0 is the ini-
tial voltage of load terminal
Exponential load modelling using LSTM.. 25
▶ Measurements are taken at load bus
30 by applying step disturbance at
generator 4 [23].
▶ A gaussian noise of standard devia-
tion 0.001 and mean 0 is added to the
true measurements.
▶ Wavelet based denoising filter is ap-
plied to denoise the noisy measure-
ment as shown in block diagram rep-
resented in Fig 4.
▶ The amplitude above threshold value
are considered as noise and hence it
is elliminated.
Fig 4. Block diagram of
Wavelet based denoising
Exponential load modelling using LSTM.. 26
Fig 5. Voltage of load
terminal after and before
denoising
Fig 6. Active power of load
terminal after and before
denoising
Fig 7. Reactive Power of
load terminal after and
before denoising
Exponential load modelling using LSTM.. 27
▶ The filtered measurements are com-
pared with the estimated output of
parameter tuning algorithm as shown
in Fig 8.
▶ Parameter tuning algorithm under-
goes iteratively to achieve the min-
imal error between filterd and esti-
mated output.
▶ LSTM is used as parameter tuning al-
gorithm Fig 8. Block diagram of measurement-based approach
Exponential load modelling using LSTM.. 28
▶ Cell State (Ct)
▶ Forget Gate (ft)
▶ Input gate layer (it)
▶ Output gate layer (ot)
Fig 9. LSTM Layer architecture
Exponential load modelling using LSTM.. 29
Fig 10. LSTM Layer architecture
Estimated value of Exponential load
parameters at load bus 30
Parameters True LSTM
α 1.2 1.2011247
β 3.5 3.507059
Exponential load modelling using LSTM.. 30
Fig 11. Active power Tracking using LSTM Fig 12. Reactive Power tracking using LSTM
Exponential load modelling using LSTM.. 31
▶ In this work a measurement based approach is employed to estimate the pa-
rameters of exponential load using LSTM as parameter tuning algorithm.
▶ LSTM as a parameter tuning works superbly in parameter etimation and active
 reactive power tracking.
▶ In future research, it is advisable to investigate more number of load models
available in literature.
Proposed Research Topic 32
▶ Investigation of parameter tuning methods for estimating various load models
in power system.
References I 33
[1] A. Arif, Z. Wang, J. Wang, B. Mather, H. Bashualdo, and D. Zhao, “Load modeling—a review,” IEEE Transactions on
Smart Grid, vol. 9, no. 6, pp. 5986–5999, 2017.
[2] P. S. Kundur and O. P. Malik, Power system stability and control. McGraw-Hill Education, 2022.
[3] J. Gil-Aguirre, S. Perez-Londoño, and J. Mora-Flórez, “A measurement-based load modelling methodology for electric
vehicle fast-charging stations,” Electric Power Systems Research, vol. 176, p. 105934, 2019.
[4] M. E.-N. Jahromi and M. T. Ameli, “Measurement-based modelling of composite load using genetic algorithm,” Electric
Power Systems Research, vol. 158, pp. 82–91, 2018. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S0378779617305023
[5] V. Vignesh, S. Chakrabarti, and S. C. Srivastava, “Power system load modelling under large and small disturbances
using phasor measurement units data,” IET Generation, Transmission  Distribution, vol. 9, no. 12, pp. 1316–1323, 2015.
[6] S. M. H. Rizvi, S. K. Sadanandan, and A. K. Srivastava, “Real-time zip load parameter tracking using sensitivity-based
adaptive window and variable elimination with realistic synchrophasor data,” IEEE Transactions on Industry
Applications, vol. 57, no. 6, pp. 6525–6536, 2021.
[7] J. Xie, Z. Ma, K. Dehghanpour, Z. Wang, Y. Wang, R. Diao, and D. Shi, “Imitation and transfer q-learning-based
parameter identification for composite load modeling,” IEEE Transactions on Smart Grid, vol. 12, no. 2, pp. 1674–1684,
2020.
[8] H. Renmu, M. Jin, and D. J. Hill, “Composite load modeling via measurement approach,” IEEE Transactions on power
systems, vol. 21, no. 2, pp. 663–672, 2006.
[9] L. Chavarro-Barrera, S. Pérez-Londoño, and J. Mora-Flórez, “An adaptive approach for dynamic load modeling in
microgrids,” IEEE Transactions on Smart Grid, vol. 12, pp. 2834–2843, 2021.
References II 34
[10] G. A. Barzegkar-Ntovom, T. A. Papadopoulos, and E. O. Kontis, “Robust framework for online parameter estimation of
dynamic equivalent models using measurements,” IEEE Transactions on Power Systems, vol. 36, pp. 2380–2389, 2021.
[11] F. Conte, F. D’Agostino, S. Massucco, F. Silvestro, C. Bossi, and M. Cabiati, “Experimental validation of a dynamic
equivalent model for microgrids,” 2020 IEEE International Conference on Environment and Electrical Engineering and
2020 IEEE Industrial and Commercial Power Systems Europe (EEEIC / ICPS Europe), pp. 1–6, 2020.
[12] M. M. Conner, M. R. Ebinger, and F. F. Knowlton, “Evaluating coyote management strategies using a spatially
explicit, individual-based, socially structured population model,” Ecological Modelling, vol. 219, pp. 234–247, 2008.
[13] A. Pai, Energy function analysis for power system stability. Springer Science  Business Media, 1989.
[14] M. Srivastava, C. L. Anderson, and J. H. Freed, “A new wavelet denoising method for selecting decomposition levels
and noise thresholds,” IEEE access, vol. 4, pp. 3862–3877, 2016.
[15] T. Hiyama, M. Tokieda, W. Hubbi, and H. Andou, “Artificial neural network based dynamic load modeling,” IEEE
Transactions on Power Systems, vol. 12, no. 4, pp. 1576–1583, Nov. 1997.
[16] R. C. Staudemeyer and E. R. Morris, “Understanding lstm–a tutorial into long short-term memory recurrent neural
networks,” arXiv preprint arXiv:1909.09586, 2019.
[17] F. A. Gers, J. Schmidhuber, and F. Cummins, “Learning to forget: Continual prediction with lstm,” Neural computation,
vol. 12, no. 10, pp. 2451–2471, 2000.
[18] K. Hatipoglu, I. Fidan, and G. Radman, “Investigating effect of voltage changes on static zip load model in a microgrid
environment,” in 2012 North American power symposium (NAPS). IEEE, 2012, pp. 1–5.
[19] C. Zheng, S. Wang, Y. Liu, C. Liu, W. Xie, C. Fang, and S. Liu, “A novel equivalent model of active distribution
networks based on lstm,” IEEE transactions on neural networks and learning systems, vol. 30, no. 9, pp. 2611–2624, 2019.
References III 35
[20] C. Wang, Z. Wang, J. Wang, and D. Zhao, “Svm-based parameter identification for composite zip and electronic load
modeling,” IEEE Transactions on Power Systems, vol. 34, no. 1, pp. 182–193, 2018.
[21] X. Wang, Y. Wang, D. Shi, J. Wang, and Z. Wang, “Two-stage wecc composite load modeling: A double deep q-learning
networks approach,” IEEE Transactions on Smart Grid, vol. 11, no. 5, pp. 4331–4344, 2020.
[22] H. Yang, J. Jiang, G. Chen, and J. Zhao, “Dynamic load identification based on deep convolution neural network,”
Mechanical Systems and Signal Processing, vol. 185, p. 109757, 2023.
[23] P. W. Sauer, M. A. Pai, and J. H. Chow, Power system dynamics and stability: with synchrophasor measurement and
power system toolbox. John Wiley  Sons, 2017.
Publications 36
▶ Abhinav Kumar, Sanjeev Kumar Mallik, “Measurement-based ZIP load mod-
elling using Opposition based differential evolution optimization,” Engineering
Research Express, Volume 5, Article Number 3, DOI 10.1088/2631-8695/ace81c
(published).
▶ Abhinav Kumar, Sanjeev Kumar Mallik, “Design and Development of Dynamic
Load Modeling For Microgrid Using DG-COA approach,” Journal of Circuits,
Systems, and Computers (Under review).
▶ Abhinav Kumar, Sanjeev Kumar Mallik, “Exponential load modelling using
long-short term memory back-propagation deep learning method ,” (Prepared).
“If you have knowledge, let others
light their candles in it.”
Margaret Fuller
Thank You!

More Related Content

PDF
Load modelling in distributed generation planning
PDF
Ak methodology nnloadmodeling
PDF
An intelligent hybrid short-term load forecasting model for smartpower gridsM...
PDF
Using Z-Transform for Electric Loads Modeling in the Presence of Harmonics
PDF
Short term load forecasting using hybrid neuro wavelet model
PDF
Daily Peak Load Forecast Using Artificial Neural Network
DOCX
mehtodalgy.docx
PDF
The effect of load modelling on phase balancing in distribution networks usin...
Load modelling in distributed generation planning
Ak methodology nnloadmodeling
An intelligent hybrid short-term load forecasting model for smartpower gridsM...
Using Z-Transform for Electric Loads Modeling in the Presence of Harmonics
Short term load forecasting using hybrid neuro wavelet model
Daily Peak Load Forecast Using Artificial Neural Network
mehtodalgy.docx
The effect of load modelling on phase balancing in distribution networks usin...

Similar to Investigation of parameter tuning methods for estimating various load models in power systems (20)

PDF
Nov 2018 Table of contents; current issue -International Journal on Soft Comp...
PDF
A040101001006
PDF
Short term residential load forecasting using long short-term memory recurre...
PPT
LoadModeling.ppt
PPT
LoadModeling (1).ppt
PDF
A Critical Review on Employed Techniques for Short Term Load Forecasting
PDF
Electricity load forecasting by artificial neural network model
PDF
Neural wavelet based hybrid model for short-term load forecasting
PDF
Performance assessment of an optimization strategy proposed for power systems
PDF
Iv3515241527
PDF
Intelligent methods in load forecasting
PDF
Short Term Load Forecasting: One Week (With & Without Weekend) Using Artifici...
PDF
H011137281
PDF
International Journal of Computational Engineering Research(IJCER)
PPTX
Five most used load forecasting models
PDF
Short Term Electrical Load Forecasting by Artificial Neural Network
PDF
A multi-layer-artificial-neural-network-architecture-design-for-load-forecast...
PDF
Advance Data Mining - Analysis and forecasting of power factor for optimum el...
Nov 2018 Table of contents; current issue -International Journal on Soft Comp...
A040101001006
Short term residential load forecasting using long short-term memory recurre...
LoadModeling.ppt
LoadModeling (1).ppt
A Critical Review on Employed Techniques for Short Term Load Forecasting
Electricity load forecasting by artificial neural network model
Neural wavelet based hybrid model for short-term load forecasting
Performance assessment of an optimization strategy proposed for power systems
Iv3515241527
Intelligent methods in load forecasting
Short Term Load Forecasting: One Week (With & Without Weekend) Using Artifici...
H011137281
International Journal of Computational Engineering Research(IJCER)
Five most used load forecasting models
Short Term Electrical Load Forecasting by Artificial Neural Network
A multi-layer-artificial-neural-network-architecture-design-for-load-forecast...
Advance Data Mining - Analysis and forecasting of power factor for optimum el...
Ad

Recently uploaded (20)

PDF
BIO-INSPIRED HORMONAL MODULATION AND ADAPTIVE ORCHESTRATION IN S-AI-GPT
PDF
A SYSTEMATIC REVIEW OF APPLICATIONS IN FRAUD DETECTION
PDF
UNIT no 1 INTRODUCTION TO DBMS NOTES.pdf
PPTX
Current and future trends in Computer Vision.pptx
PPTX
Artificial Intelligence
PDF
Human-AI Collaboration: Balancing Agentic AI and Autonomy in Hybrid Systems
PDF
Enhancing Cyber Defense Against Zero-Day Attacks using Ensemble Neural Networks
PDF
Level 2 – IBM Data and AI Fundamentals (1)_v1.1.PDF
PPT
Introduction, IoT Design Methodology, Case Study on IoT System for Weather Mo...
PDF
Automation-in-Manufacturing-Chapter-Introduction.pdf
PDF
null (2) bgfbg bfgb bfgb fbfg bfbgf b.pdf
PDF
COURSE DESCRIPTOR OF SURVEYING R24 SYLLABUS
PDF
86236642-Electric-Loco-Shed.pdf jfkduklg
PPTX
communication and presentation skills 01
PPTX
UNIT - 3 Total quality Management .pptx
PDF
R24 SURVEYING LAB MANUAL for civil enggi
PDF
Exploratory_Data_Analysis_Fundamentals.pdf
PDF
SMART SIGNAL TIMING FOR URBAN INTERSECTIONS USING REAL-TIME VEHICLE DETECTI...
PDF
737-MAX_SRG.pdf student reference guides
PDF
Visual Aids for Exploratory Data Analysis.pdf
BIO-INSPIRED HORMONAL MODULATION AND ADAPTIVE ORCHESTRATION IN S-AI-GPT
A SYSTEMATIC REVIEW OF APPLICATIONS IN FRAUD DETECTION
UNIT no 1 INTRODUCTION TO DBMS NOTES.pdf
Current and future trends in Computer Vision.pptx
Artificial Intelligence
Human-AI Collaboration: Balancing Agentic AI and Autonomy in Hybrid Systems
Enhancing Cyber Defense Against Zero-Day Attacks using Ensemble Neural Networks
Level 2 – IBM Data and AI Fundamentals (1)_v1.1.PDF
Introduction, IoT Design Methodology, Case Study on IoT System for Weather Mo...
Automation-in-Manufacturing-Chapter-Introduction.pdf
null (2) bgfbg bfgb bfgb fbfg bfbgf b.pdf
COURSE DESCRIPTOR OF SURVEYING R24 SYLLABUS
86236642-Electric-Loco-Shed.pdf jfkduklg
communication and presentation skills 01
UNIT - 3 Total quality Management .pptx
R24 SURVEYING LAB MANUAL for civil enggi
Exploratory_Data_Analysis_Fundamentals.pdf
SMART SIGNAL TIMING FOR URBAN INTERSECTIONS USING REAL-TIME VEHICLE DETECTI...
737-MAX_SRG.pdf student reference guides
Visual Aids for Exploratory Data Analysis.pdf
Ad

Investigation of parameter tuning methods for estimating various load models in power systems

  • 1. Ph.D. Pre-Registration Seminar Presented by Abhinav Kumar Roll: 185EE08 Under the Supervision of Dr. S.K.Mallik National Institute of Technology Patna Ph.D. Full-time Department of Electrical Engineering November 16, 2023
  • 2. Presentation Outline 2 Coursework Status Introduction Literature Survey Research Gap Research Objectives Works Undertaken References
  • 3. Coursework Status 3 Coursework Status ▶ Courses completed at NIT Patna: ˆ Advanced Electrical Engineering Lab II ˆ Research Methodology ˆ SCADA and Energy Management ˆ Advanced Electrical Engg Lab-I ˆ Seminar and Technical Report Writing ▶ CGPA: 7.81 for 16 credit course
  • 4. Introduction 4 Load Modelling ▶ A simplified mathematical model and an equivalent circuit to represent the entire load connected at any load bus terminal [1]. ▶ Why? ˆ Load modeling is needed for continuous monitoring of load variation with respect to generation and transmission.
  • 5. Introduction.... 5 ▶ Types of load models [2]. ˆ Static load models * Static models express the active and reactive power at any instant of time as func- tions of bus voltage magnitudes and frequency * ZIP load model, exponential load model, frequency-dependent load model are types of static load models. ˆ Dynamic load models * Dynamic load models express the active and reactive powers as a function of voltage and time. * Induction motor, exponential recovery load model, non-linear tansig function are types of dynamic load models. ˆ Composite load models * Composite load model is the combination of static and dynamic load models
  • 6. Introduction.... 6 ▶ Methodologies ˆ Measurement-based approach * Method which takes the measurement of a load terminal as an input and apply the parameter tuning algorithm for parameter estimation. ˆ Component-based approach * Method which classify the load as its load class then estimate the parameters of load models ˆ ANN-based model * The load model which is designed by updating the weights of ANN model by tracking true load characteristics.
  • 7. Introduction.... 7 ▶ Parameter tuning algorithms ˆ The algorithm to estimate the load model parameter by minimizing the difference between estimated and true load characteristics. ˆ In literature, there are number of parameter tuning algorithms are investigated, namely, Genetic Algorithm, Improved Particle swarm optimization, Support vec- tor machine, Variable Projection method etc.
  • 9. Literature Survey 9 S.No. Title Remark Research Gap 1. J. Gil-Aguirre, S. Perez-Londono, and J. Mora-Florez,“A measurement- based load modelling methodology for electric vehicle fast-charging sta- tions,” Electric Power Systems Research, vol. 176, p. 105934, 2019 [3] . Classify the load model which is suitable for EVFCS Dynamic and composite load models are not considered 2. M. E.-N. Jahromi and M. T. Ameli, “Measurement-based modelling of com- posite load using genetic algorithm,” Electric Power Systems Research, vol. 158, pp. 82–91, 2018 [4] . Parameter Identifi- cation of composite load based on GA Estimation will be more profound with latest tuning techniques. 3. V. Vignesh, S. Chakrabarti, and S. C. Srivastava, “Power system load mod- elling under large and small disturbances using phasor measurement units data,” IET Generation, Transmission Distribution,vol. 9, no. 12, pp. 1316–1323, 2015 [5] . Parameter Identifi- cation of compos- ite load under large and small distur- bances Estimation will be more accurate with deep learning techniques. 4. S. M. H. Rizvi, S. K. Sadanandan, and A. K. Srivastava, “Real-time zip load parameter tracking using sensitivity-based adaptive window and vari- able elimination with realistic synchrophasor data,” IEEE Transactions on Industry Applications, vol. 57, no. 6, pp. 6525–6536, 2021 [6] . Parameter Identifi- cation of ZIP load in Real-time Estimation and noise cancel- lation will be done with latest techniques.
  • 10. Literature Survey.... 10 S.No. Title Remark Research Gap 5. J. Xie, Z. Ma, K. Dehghanpour, Z. Wang, Y. Wang, R. Diao, and D. Shi, “Imitation and transfer q-learning-based parameter identification for com- posite load modeling,” IEEE Transactions on Smart Grid, vol. 12, no. 2, pp. 1674–1684,2020 [7] . Parameter Iden- tification of com- posite load in simulation Identification should be done with deep learn- ing methods. 6. H. Renmu, M. Jin, and D. J. Hill, “Composite load modeling via measure- ment approach,” IEEE Transactions on power systems, vol. 21, no. 2, pp. 663–672, 2006 [8]. Parameter Iden- tification of com- posite load in simulation noise cancel- lation should be applied for accuracy. 7. L. Chavarro-Barrera, S. Perez-Londo ´ no, and J. Mora-Fl ˜ orez, “An adap- tive approach for dynamic load modeling in microgrids,” IEEE Transactions on Smart Grid, vol. 12, pp. 2834–2843, 2021 [9] . Parameter Iden- tification of Dy- namic load in microgrid static and com- posite load model should also be consid- ered 9. G. A. Barzegkar-Ntovom, T. A. Papadopoulos, and E. O. Kontis, “Robust framework for online parameter estimation of dynamic equivalent models using measurements,” IEEE Transactions on Power Systems, vol. 36, pp. 2380–2389, 2021 [10] . Parameter Iden- tification of dy- namic load using measurements Static and dy- namic equivalent should be taken
  • 11. Literature Survey.... 11 S.No. Title Remark Research Gap 8. F. Conte, F. D’Agostino, S. Massucco, F. Silvestro, C. Bossi, and M. Cabi- ati, “Experimental validation of a dynamic equivalent model for micro- grids,” 2020 IEEE International Conference on Environment and Electrical Engineering and 2020 IEEE Industrial and Commercial Power Systems Eu- rope (EEEIC / I CPS Europe), pp. 1–6, 2020 [11] . Parameter Iden- tification of dy- namic load in microgrid exper- imentally static and composite load model should also be con- sidered in simulation. 10. M. M. Conner, M. R. Ebinger, and F. F. Knowlton, “Evaluating coyote management strategies using a spatially explicit, individual-based, socially structured population model,” Ecological Modelling, vol. 219, pp. 234–247, 2008 [12]. coyote optimiza- tion methodolgy. coyote strategies should be applied to estimate dynamic load model. 11. A. Pai, Energy function analysis for power system stability. Springer Science & Business Media, 1989 [13] . Complete clarifi- cation of NE 39 bus system. should be used to analyze static,dynamic and com- posite load models. 12. M. Srivastava, C. L. Anderson, and J. H. Freed, “A new wavelet denoising method for selecting decomposition levels and noise thresholds,” IEEE ac- cess, vol. 4, pp. 3862–3877, 2016. [14] . Wavelet de- noising method for noise can- cellation of measurements wavelet denoising method should be applied to denoise the load measurements
  • 12. Literature Survey.... 12 S.No. Title Remark Research Gap 13. Hiyama T, Tokieda M, Hubbi W and Andou H 1997 Artificial neural network based dynamic load modeling IEEE Trans. Power Syst. 12 1576–83 [15] . ANN based dy- namic load mod- eling more advanced deep learning method should be used for load modeling. 14. R. C. Staudemeyer and E. R. Morris, “Understanding lstm–a tutorial into long short-term memory recurrent neural networks,” arXiv preprint arXiv:1909.09586, 2019 [16] . Understanding the concept of long-short term memory deep learning method LSTM is not used regression method for ex- ponential load model. 15. F. A. Gers, J. Schmidhuber, and F. Cummins, “Learning to forget: Con- tinual prediction with lstm,” Neural computation, vol. 12, no. 10, pp. 2451–2471, 2000 [17] . Application of LSTM method LSTM should be used as param- eter tuning al- gorithm of load models. 16. K. Hatipoglu, I. Fidan, and G. Radman, “Investigating effect of voltage changes on static zip load model in a microgrid environment,” in 2012 North American power symposium (NAPS). IEEE, 2012, pp. 1–5 [18] . Stability of mi- crogrid in which zip load is used as load Dynamic load is not considered.
  • 13. Literature Survey.... 13 S.No. Title Remark Research Gap 17. C. Zheng et al., ”A Novel Equivalent Model of Active Distribution Net- works Based on LSTM,” in IEEE Transactions on Neural Networks and Learning Systems, vol. 30, no. 9, pp. 2611-2624, Sept. 2019, doi: 10.1109/TNNLS.2018.2885219 [19] . Designing Ac- tive Distribution network using LSTM LSTM is not used as regres- sion for load modelling 18. C. Wang, Z. Wang, J. Wang, and D. Zhao, “Svm-based parameter identifi- cation for composite zip and electronic load modeling,” IEEE Transactions on Power Systems, vol. 34, no. 1, pp. 182–193, 2018 [20] . parameter es- timation using Support vector machine SVM technique should be used for static load modelling 19. X. Wang, Y. Wang, D. Shi, J. Wang, and Z. Wang, “Two-stage wecc com- posite load modeling: A double deep q-learning networks approach,” IEEE Transactions on Smart Grid, vol. 11, no. 5, pp. 4331–4344, 2020 [21] . parameter es- timation of WECC load using double deep Q learning system is very complex and time taking 20. H. Yang, J. Jiang, G. Chen, and J. Zhao, “Dynamic load identification based on deep convolution neural network,”Mechanical Systems and Signal Processing, vol. 185, p. 109757, 2023 [22] . dynamic load modeling using deep convolu- tional network static load model should also be consid- ered.
  • 14. Research Gap 14 ▶ A number of load model is available in literature which needs to be investigated. ▶ For accurate estimation of load model, there should be parameter tuning. For this purpose, a number of parameter tuning algorithm needs to be investigated.
  • 15. Research Objectives 15 ▶ Investigation of ZIP load using opposition based differential evolution optimiza- tion. ▶ Investigation of exponential load model using Long-short term memory method. ▶ Investigation of dynamic load modeling using dimensional-growth coyote opti- mization.
  • 16. Measurement-based ZIP load modelling using Opposition based differential evolution optimization
  • 17. Measurement-based ZIP load modelling using opposition based differential evolution optimization... 17 ▶ ZIP load model is static load model which shows volt- age power relationship at the load terminal as shown in (1) ▶ It constitutes of constant impedance load, constant current load and constant power load. ▶ Vm and V0 are measured and initial voltage at load bus terminal. Pzip = Pz Vm V0 2 + Pi Vm V0 + Pp Qzip = Qz Vm V0 2 + Qi Vm V0 + Qq          (1)
  • 18. Measurement-based ZIP load modelling using opposition based differential evolution optimization... 18 Parameters Definition Pz, Pi, Pp Proportional coefficients of the constant impedance, constant current, and con- stant power in static active load. Qz, Qi, Qq Proportional coefficients of the constant impedance, constant current,and con- stant power in static reactive load. V True Load Characteristics Load Model Parameter Tuning Algorithm + - Y=[P Q] Yest=[Pest Qest] e Identification + + Noise (Pn, Qn) Fig 1. Block diagram of Measurement-based approach
  • 19. Measurement-based ZIP load modelling using opposition based differential evolution optimization... 19 Parameter Tuning Algorithm ▶ Opposition based differential evolution optimization (ODEO) is used as the parameter tuning algorithm. ▶ The steps covered in this technique is mentioned below ˆ Generate a random population as trial vectors. ˆ Perform mutation to generate donor vectors. ˆ Perform crossover to generate target vectors. ˆ set the jumping coefficients to jump the solution when it attains local minima.
  • 20. Measurement-based ZIP load modelling using opposition based differential evolution optimization... 20 Estimated value of ZIP load parameters at load bus 30 Parameters True ODEO GWO PSO Pp 0.4 0.42380 0.45284 0.45506 Pz 0.3 0.28620 0.25478 0.25965 Pi 0.3 0.28980 0.29308 0.28593 Qq 0.4 0.43860 0.50595 0.40291 Qz 0.3 0.24720 0.24272 0.32433 Qi 0.3 0.31920 0.25243 0.27262 Performance of tuning algorithm for case1 Methods No. of iterations Time/iteration(sec.) Accuracy ODEO 859 3.8 4.84 × 10−5 PSO 1620 3.77 4.9719 × 10−5 GWO 1630 4.07 4.9 × 10−5
  • 21. Measurement-based ZIP load modelling using opposition based differential evolution optimization... 21 Fig 2. Active power of ZIP load Fig 3. Reactive power of ZIP load
  • 22. Measurement-based ZIP load modelling using opposition based differential evolution optimization... 22 ▶ In this work, a measurement-based approach is employed for estimating the pa- rameters of the ZIP load using a ODEO, PSO, GWO parameter tuning method. ▶ These algorithms are tested and validated on NE 39-bus system. ▶ The results confirm the superior performance of proposed ODEO in terms of parameter estimation, convergence speed, and accuracy compared to PSO and GWO.
  • 23. Exponential load modelling using long-short term memory (LSTM) back-propagation deep learning method
  • 24. Exponential load modelling using LSTM 24 ▶ Exponential load model is voltage-power relationship at load bus as shown in (2) ▶ The nature of load varies with variation in α and β. ▶ For composite load system, the value of α and β usually ranges between 0.5 and 1.8 1.5 and 6 respectively. Pexp = Vm V0 α Qexp = Vm V0 β          (2) ▶ α and β are the parameters of the load model. ▶ Vm is the measured voltage and V0 is the ini- tial voltage of load terminal
  • 25. Exponential load modelling using LSTM.. 25 ▶ Measurements are taken at load bus 30 by applying step disturbance at generator 4 [23]. ▶ A gaussian noise of standard devia- tion 0.001 and mean 0 is added to the true measurements. ▶ Wavelet based denoising filter is ap- plied to denoise the noisy measure- ment as shown in block diagram rep- resented in Fig 4. ▶ The amplitude above threshold value are considered as noise and hence it is elliminated. Fig 4. Block diagram of Wavelet based denoising
  • 26. Exponential load modelling using LSTM.. 26 Fig 5. Voltage of load terminal after and before denoising Fig 6. Active power of load terminal after and before denoising Fig 7. Reactive Power of load terminal after and before denoising
  • 27. Exponential load modelling using LSTM.. 27 ▶ The filtered measurements are com- pared with the estimated output of parameter tuning algorithm as shown in Fig 8. ▶ Parameter tuning algorithm under- goes iteratively to achieve the min- imal error between filterd and esti- mated output. ▶ LSTM is used as parameter tuning al- gorithm Fig 8. Block diagram of measurement-based approach
  • 28. Exponential load modelling using LSTM.. 28 ▶ Cell State (Ct) ▶ Forget Gate (ft) ▶ Input gate layer (it) ▶ Output gate layer (ot) Fig 9. LSTM Layer architecture
  • 29. Exponential load modelling using LSTM.. 29 Fig 10. LSTM Layer architecture Estimated value of Exponential load parameters at load bus 30 Parameters True LSTM α 1.2 1.2011247 β 3.5 3.507059
  • 30. Exponential load modelling using LSTM.. 30 Fig 11. Active power Tracking using LSTM Fig 12. Reactive Power tracking using LSTM
  • 31. Exponential load modelling using LSTM.. 31 ▶ In this work a measurement based approach is employed to estimate the pa- rameters of exponential load using LSTM as parameter tuning algorithm. ▶ LSTM as a parameter tuning works superbly in parameter etimation and active reactive power tracking. ▶ In future research, it is advisable to investigate more number of load models available in literature.
  • 32. Proposed Research Topic 32 ▶ Investigation of parameter tuning methods for estimating various load models in power system.
  • 33. References I 33 [1] A. Arif, Z. Wang, J. Wang, B. Mather, H. Bashualdo, and D. Zhao, “Load modeling—a review,” IEEE Transactions on Smart Grid, vol. 9, no. 6, pp. 5986–5999, 2017. [2] P. S. Kundur and O. P. Malik, Power system stability and control. McGraw-Hill Education, 2022. [3] J. Gil-Aguirre, S. Perez-Londoño, and J. Mora-Flórez, “A measurement-based load modelling methodology for electric vehicle fast-charging stations,” Electric Power Systems Research, vol. 176, p. 105934, 2019. [4] M. E.-N. Jahromi and M. T. Ameli, “Measurement-based modelling of composite load using genetic algorithm,” Electric Power Systems Research, vol. 158, pp. 82–91, 2018. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0378779617305023 [5] V. Vignesh, S. Chakrabarti, and S. C. Srivastava, “Power system load modelling under large and small disturbances using phasor measurement units data,” IET Generation, Transmission Distribution, vol. 9, no. 12, pp. 1316–1323, 2015. [6] S. M. H. Rizvi, S. K. Sadanandan, and A. K. Srivastava, “Real-time zip load parameter tracking using sensitivity-based adaptive window and variable elimination with realistic synchrophasor data,” IEEE Transactions on Industry Applications, vol. 57, no. 6, pp. 6525–6536, 2021. [7] J. Xie, Z. Ma, K. Dehghanpour, Z. Wang, Y. Wang, R. Diao, and D. Shi, “Imitation and transfer q-learning-based parameter identification for composite load modeling,” IEEE Transactions on Smart Grid, vol. 12, no. 2, pp. 1674–1684, 2020. [8] H. Renmu, M. Jin, and D. J. Hill, “Composite load modeling via measurement approach,” IEEE Transactions on power systems, vol. 21, no. 2, pp. 663–672, 2006. [9] L. Chavarro-Barrera, S. Pérez-Londoño, and J. Mora-Flórez, “An adaptive approach for dynamic load modeling in microgrids,” IEEE Transactions on Smart Grid, vol. 12, pp. 2834–2843, 2021.
  • 34. References II 34 [10] G. A. Barzegkar-Ntovom, T. A. Papadopoulos, and E. O. Kontis, “Robust framework for online parameter estimation of dynamic equivalent models using measurements,” IEEE Transactions on Power Systems, vol. 36, pp. 2380–2389, 2021. [11] F. Conte, F. D’Agostino, S. Massucco, F. Silvestro, C. Bossi, and M. Cabiati, “Experimental validation of a dynamic equivalent model for microgrids,” 2020 IEEE International Conference on Environment and Electrical Engineering and 2020 IEEE Industrial and Commercial Power Systems Europe (EEEIC / ICPS Europe), pp. 1–6, 2020. [12] M. M. Conner, M. R. Ebinger, and F. F. Knowlton, “Evaluating coyote management strategies using a spatially explicit, individual-based, socially structured population model,” Ecological Modelling, vol. 219, pp. 234–247, 2008. [13] A. Pai, Energy function analysis for power system stability. Springer Science Business Media, 1989. [14] M. Srivastava, C. L. Anderson, and J. H. Freed, “A new wavelet denoising method for selecting decomposition levels and noise thresholds,” IEEE access, vol. 4, pp. 3862–3877, 2016. [15] T. Hiyama, M. Tokieda, W. Hubbi, and H. Andou, “Artificial neural network based dynamic load modeling,” IEEE Transactions on Power Systems, vol. 12, no. 4, pp. 1576–1583, Nov. 1997. [16] R. C. Staudemeyer and E. R. Morris, “Understanding lstm–a tutorial into long short-term memory recurrent neural networks,” arXiv preprint arXiv:1909.09586, 2019. [17] F. A. Gers, J. Schmidhuber, and F. Cummins, “Learning to forget: Continual prediction with lstm,” Neural computation, vol. 12, no. 10, pp. 2451–2471, 2000. [18] K. Hatipoglu, I. Fidan, and G. Radman, “Investigating effect of voltage changes on static zip load model in a microgrid environment,” in 2012 North American power symposium (NAPS). IEEE, 2012, pp. 1–5. [19] C. Zheng, S. Wang, Y. Liu, C. Liu, W. Xie, C. Fang, and S. Liu, “A novel equivalent model of active distribution networks based on lstm,” IEEE transactions on neural networks and learning systems, vol. 30, no. 9, pp. 2611–2624, 2019.
  • 35. References III 35 [20] C. Wang, Z. Wang, J. Wang, and D. Zhao, “Svm-based parameter identification for composite zip and electronic load modeling,” IEEE Transactions on Power Systems, vol. 34, no. 1, pp. 182–193, 2018. [21] X. Wang, Y. Wang, D. Shi, J. Wang, and Z. Wang, “Two-stage wecc composite load modeling: A double deep q-learning networks approach,” IEEE Transactions on Smart Grid, vol. 11, no. 5, pp. 4331–4344, 2020. [22] H. Yang, J. Jiang, G. Chen, and J. Zhao, “Dynamic load identification based on deep convolution neural network,” Mechanical Systems and Signal Processing, vol. 185, p. 109757, 2023. [23] P. W. Sauer, M. A. Pai, and J. H. Chow, Power system dynamics and stability: with synchrophasor measurement and power system toolbox. John Wiley Sons, 2017.
  • 36. Publications 36 ▶ Abhinav Kumar, Sanjeev Kumar Mallik, “Measurement-based ZIP load mod- elling using Opposition based differential evolution optimization,” Engineering Research Express, Volume 5, Article Number 3, DOI 10.1088/2631-8695/ace81c (published). ▶ Abhinav Kumar, Sanjeev Kumar Mallik, “Design and Development of Dynamic Load Modeling For Microgrid Using DG-COA approach,” Journal of Circuits, Systems, and Computers (Under review). ▶ Abhinav Kumar, Sanjeev Kumar Mallik, “Exponential load modelling using long-short term memory back-propagation deep learning method ,” (Prepared).
  • 37. “If you have knowledge, let others light their candles in it.” Margaret Fuller Thank You!