Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1279

Mixed Algorithm of SINDy and HAVOK for

Measure-Based Analysis of Power System with
Inverter-based Resource
Reza Saeed Kandezy and John Ning Jiang
School of Electrical and Computer Engineering
the University of Oklahoma Norman, OK, USA

Abstract:- Artificial intelligence and machine learning is module-level nonlinearity thus the power system remains
enhancing electric grids by offering data analysis tools first order, i.e., linearizable [5]. The current state-of-the -art
that can be used to operate the power grid more reliably. ML algorithm showed remarkable performance regarding
However, the complex nonlinear dynamics, particularly model identifications and system analysis of linear and
when coupled with multi-scale interactions among nonlinear first-order system. Increase in the penetration of
Inverter-based renewable energy Resources, calls for renewable energy, specifically IBRs, in power system
effective algorithms for power system application. caused a shift in the nature of the system by introducing
coupled multi-scale nonlinear dynamic. This changed
This paper presents affective novel algorithm to introduced and interactive complex nonlinearity on system-
detect various nonlinear dynamics, which is built upon: level to the grid which elevate the power system to the
the Sparse Identification of Nonlinear Dynamics method second-order system. The generic ML algorithms struggle
for nonlinear dynamics detection; and Hankel to capture systems' multi-scale temporal and spatial
Alternative View of Koopman method for multi-scale complexities, further limiting their accuracy and
decomposition. We show that, by an appropriate effectiveness in modeling and analysis of power system
integration of the strengths of the two, the mixed derived by IBRs, where fast and slow dynamics capture
algorithm not only can detect the nonlinearity, but also localized phenomena and overall spatial patterns [5] [6].
it distinguishes the nonlinearity caused by coupled
Inverter-based resources from the more familiar ones Since the introduction of SINDy algorithm in 2016 [7]
caused synchronous generators. This shows that the it has seen widespread use in model identification across
proposal algorithm can be a promising application of various disciplines, showing remarkable performance by
artificial intelligence and machine learning for data explicitly identifying underlying governing equations
measure-based analysis to support operation of power through sparse regression techniques, which leads to
system with integrated renewables. interpretable models while effectively addressing model
complexity [7], [8]. The SINDy algorithm has been used in
Keywords:- HAVOK, Inverter-Based Resources, Machine limited number of the field of power system analysis [9-12].
Learning, Measure-Based Method, Model Identification, Notably, these studies predominantly analyzed the general
Multi-Scale Dynamics, Non-Linear Dynamics, Power power system and relied on first-order system models, as
System, Sindy. exemplified by [9] in 2020 paper on power system
applications.
I. INTRODUCTION
SINDy is a measure-based method specifically
Although machine learning (ML) algorithm and designed to discover governing equations or mathematical
artificial intelligent (AI) enhanced the analysis of the system models from observed data. Several studies in different
by addressing the limitations of traditional models to disciplines demonstrated the performance and practicality
capture the complex nonlinear dynamics of power systems, of SINDy in model identifications [13]. By leveraging
the integration of inverter-based energy resources (IBRs) compressed sensing and sparse modeling principles, SINDy
with coupled muti-scale dynamics necessitate an alternative offers robustness and the potential for generalization,
approach utilizing more advanced algorithms to improve allowing for identifying key dynamical features with
modeling accuracy and system optimization [1-4]. In the relatively few measurements. However, it's important to
conventional view, the IBRs technology will introduce note that SINDy can be sensitive to noise. At its core, the
nonlinearity into the power system, however, the generic SINDy assumes the system as linearizable system,
nonlinearity introduced by the IBRs classified into the i.e., first-order nonlinear system. Therefore, generic SINDy

IJISRT24MAR1279 www.ijisrt.com 1677

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1279

faces challenge in coupled multi-scale systems where framework of generic SINDy method and supporting
multiple variables interact strongly [7], [8]. These systems HAVOK decomposition in power system. Section III
often involve intricate relationships, and accurately demonstrates the proposed methods' performance by
capturing these dynamics requires disentanglement of the evaluating and examining the obtained outcomes. The final
multi-scale dynamics [14]. This intricate and multifaceted segment comprises the concluding remarks, emphasizing
nonlinear effects of integrating IBRs and renewables, the results' significance and potential research directions for
manifesting as second-order nonlinearity, was not future studies.
considered in [9-12].
II. SINDY ALGORITHM AND SUPPORTING
HAVOK is a decomposition technique approximating HAVOK DECOMPOSITION
chaotic dynamics as an intermittently forced linear system,
combining principles from delay embedding and Koopman A. Developed Method based on Generic SINDy Algorithm:
theory [14]. It leverages real-time data covariance analysis The dynamics of power system can be described by the
to uncover the underlying structures within high- following general form [15]:
dimensional data. The method employs time-delay
embeddings to capture latent variables and intrinsic 𝑑𝑣(𝑡)
= 𝑓(𝑣(𝑡)), (1)
measurement coordinates, rooted in the theory of Koopman 𝑑𝑡

operators [14]. By describing the evolution of observables

Where 𝑣(𝑡) ∈ 𝑅𝑛 represents the system's voltage at
as an infinite-dimensional linear process, Koopman
operators offer an alternative perspective for analyzing time 𝑡 and 𝑓(𝑣(𝑡)) encompasses the dynamic constraints
dynamical systems without explicitly solving their governing the system's equations, including parameters,
underlying differential equations. The HAVOK method time dependence, and external forcing. To determine the
extends this framework by utilizing time-delay embeddings function 𝑓 from available data, a time history of the system's
and intrinsic measurement coordinates, enabling the voltage, denoted as 𝑣(𝑡), is collected. The derivative of
disentanglement of intricate couplings in multiscale 𝑣(𝑡), denoted as 𝑣̇ (𝑡), is directly or numerically
systems. approximated. The data is sampled at various time
instances, {𝑡1 , 𝑡2 , . . . , 𝑡𝑚 } and organized into 𝑉 and 𝑉̇
The combination of SINDy with HAVOK addresses matrices:
the challenges posed by complex coupled multi-scale
𝑇
interactions and dynamics within system such as power | | |
grids integrated with intermittent IBRs. This integrated 𝑣(𝑡
𝑉 = [ 1 ),𝑣(𝑡 )
2 ,..., 𝑣(𝑡𝑚 )] , (2)
approach leverages SINDy to identify the underlying | | |
governing equations and utilizes HAVOK to exploit the
Hankel matrix structure of the data, enabling the extraction 𝑇
| | |
of informative patterns and features that characterize the
power grid's behavior. 𝑉̇ = [𝑣̇ (𝑡1 ), 𝑣̇ (𝑡2 ), . . . , 𝑣̇ (𝑡𝑚 )] . (3)
| | |
The contributions of our work are summarized in two
aspects, The next step in the approach involves defining a
library of candidate nonlinear functions, denoted as 𝛩(𝑉),
 We present a novel mixed algorithm that can be used for where 𝑉 is the data matrix that contains observed data points
data-based analysis. This algorithm combines the of the system voltage. These functions can include
advantages of two powerful methods recently developed polynomials, trigonometric, exponentials, logarithmic
in the field of data science for the detection and analysis functions, and other suitable nonlinear expressions.
of complex multi-scale nonlinear dynamics in power
systems with IBRs. 𝛩(𝑉) = [1, 𝑉, 𝑉 2 , 𝑉 3 , . . . , 𝑠𝑖𝑛(𝑉), 𝑐𝑜𝑠(𝑉), . . . ]. (4)
 Through a demonstrative study showing the
effectiveness of identifying various nonlinear dynamics Assuming that only a few of these nonlinearities are
with different characteristics, we demonstrate, in a active in each row of 𝑓, a sparse regression problem is
broader sense, the necessity of integrating different formulated to determine the sparse vectors of coefficients,
AI/ML data analytics to the development of more
effective measurement-based tools for power system | | |
analysis. 𝛯 = [𝜉1 , 𝜉2 , . . . , 𝜉𝑛 ], (5)
| | |
The subsequent sections of this manuscript are
organized as follows: Section II describe the developed which indicate the active nonlinearities.
Mathematically, this can be expressed as:

IJISRT24MAR1279 www.ijisrt.com 1678

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1279

𝑉̇ = 𝛩(𝑉)𝛯. (6) Where 𝑢(𝑡) signifies a vector comprising the first 𝑟 −

1 eigen-time-delay coordinates, 𝐴 denotes a matrix
Given the voltage matrix 𝑉 and the library of candidate capturing the linear dynamics, and 𝐵 represents a matrix
functions 𝛩(𝑉), The method formulates the sparse characterizing the coupling between the eigen-time-delay
regression problem as follows: coordinates and the forcing term 𝑢𝑟 (𝑡). In the case of
chaotic systems, the linear model based on the first 𝑟 − 1
𝑚𝑖𝑛𝑖𝑚𝑖𝑧𝑒 ||𝑉̇ − 𝛩(𝑉)𝛯||2 + 𝜆 ||𝛯||₁ (7) terms provide a commendable approximation, while the
forcing term encapsulates the nonlinear and chaotic
where 𝛯 is the sparse coefficients representing the behavior of coupled multi-scale dynamic.
importance or relevance of each term in the library, || ⋅ ||₂
denotes the L2 norm, || ⋅ ||₁ represents the L1 norm, and 𝜆 It has to be noted that in scenarios where the intervals
is a regularization parameter that controls the trade-off between multi-scale dynamics' time constants are
between data fidelity and sparsity. The first term ensures significantly disparate, the necessity for sampling becomes
that the model predictions, obtained by multiplying 𝛩(𝑉) considerably demanding [16]. By employing burst
with 𝛯, are close to the observed data, while the second term sampling, the data requirements for SINDy remain
encourages a sparse solution by promoting a minimal relatively consistent, even as the temporal scales diverge
number of nonzero coefficients [7]. [17]. This approach reduces the sampling rate, which proves
advantageous when limitations exist on the number of
B. Adaptation of HAVOK decomposition on power system: samples that can be acquired due to bandwidth restrictions
Consider a record of the voltage on a single point 𝑣(𝑡), [16].
where we generate a delay embedding vector denoted as
𝑉(𝑡). This vector is comprised of delayed measurements of III. DEMONSTRATION STUDY
𝑣(𝑡), at different time points and can be formally expressed
as: In this paper, conducted on an IEEE 15-bus power grid
(depicted in Figure 1), the developed method, based on
𝑉(𝑡) = [𝑣(𝑡), 𝑣(𝑡 − 𝜏), 𝑣(𝑡 − 2𝜏), . . . , 𝑣(𝑡 − (𝑚 − 1)𝜏)]𝑇 , (8) generic SINDy algorithm and HAVOK decomposition, was
deployed to scrutinize voltage waveforms and uncover
Where 𝜏 signifies the time delay between system dynamics. The algorithm's performance was
measurements and m corresponds to the embedding assessed across diverse and intricate operational scenarios.
dimension. The deliberate selection of this particular power system
configuration serves as a foundation for showcasing
We conduct the Singular Value Decomposition (SVD) of SINDy's performance under varying conditions and
the matrix 𝑉, which is constructed from the time series of a highlights the practicality of HAVOK in tackling the
single measurement voltage 𝑣(𝑡) to obtain intrinsic challenges presented by complex, multi-scale dynamics.
measurement coordinates. The SVD of V yields the The evaluation of the method in the large-scale systems and
following decomposition: further experimental analysis are reserved for future
investigations.
𝑉 = 𝑌𝛴𝑈 ∗ , (9)
A. Test Setup:
𝑌 and 𝑈 represent orthogonal matrices, the ∙∗ represent The system architecture consists of 15 buses,
conjugate transpose, and 𝛴 is a diagonal matrix comprising representing distinct nodes within the power system
singular values. 𝑌 columns correspond to the eigen-time- network, and these buses are interconnected through
delay coordinates, which capture the essential structure of branches that symbolize the power transmission lines. Each
the system's dynamics. The resultant linear model within the bus has a unique set of parameters and attributes and is
HAVOK method can be succinctly expressed as: connected to neighboring buss via branches characterized
by specific impedance, which govern the dynamics of
𝑑
𝑢(𝑡) = 𝐴𝑢(𝑡) + 𝐵𝑢𝑟 (𝑡), (10) power flow among the interconnected buses. Table 1
𝑑𝑡 provides a comprehensive overview of the network
configuration and its branch parameters.

IJISRT24MAR1279 www.ijisrt.com 1679

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1279

Fig 1: Single Line Diagram of Implemented IEEE 15 Bus Network

Table 1: Implemented IEEE 15 Bus Network Configuration

Line index From bus To bus r+xi (Ω) Node index Pload+Qload (kW+kVAR)
1 1 2 1.53+1.778i 2 100+60
2 2 3 1.037+1.071i 3 90+40
3 3 4 1.224+1.428i 4 120+80
4 4 5 1.262+1.499i 5 60+30
5 5 9 1.176+1.335i 6 60+20
6 6 10 1.1+0.6188i 7 200+100
7 7 6 1.174+0.2351i 8 200+100
8 8 7 1.174+0.74i 9 60+20
9 9 8 1.174+ 0.74i 10 60+20
10 10 11 1.15+ 0.065i 11 45+30
11 11 12 1.274+1.522i 12 60+35
12 12 13 1.274+1.522i 13 60+35
13 13 14 1.075+1.522i 14 120+80
14 14 15 1.075+ 1.522i 15 60+10

This investigation explores a comprehensive set of an extra load is connected to Bus 14, disconnected at 𝑡 =
power system conditions, encompassing both abrupt 8𝑠.
changes (faults) and gradual changes (load variations), in
the context of conventional synchronous generators (SG) The simulation duration of 10 s and total sample count
and IBRs. The study encompasses three distinct scenarios, of 200,000 (20,000 sample per second) are determined to
representing both single- and multi-dynamic systems. The capture temporal dynamics faithfully. Furthermore, a
first scenario examines a system solely supplied by a polynomial library is constructed with a precise third order
synchronous generator at Bus 1 and Bus 3. The second polynomial and regularization parameter (0.8).
scenario incorporates the integration of an IBR at Bus 3,
sharing the load demand equally with the synchronous In multiscale dynamic conditions, the SINDy
generator at Bus 1. In the third scenario, the network is enhanced by HAVOK decomposition were applied. The
subjected to a 100% penetration of IBRs, where the demand essential parameters such as the number of samples per burst
is supplied by two IBRs located at Bus 1 and Bus 3. Each (1000 samples per burst), the optimal hard threshold value
scenario spans 10 seconds, with the synchronous generators 4
for singular values ( ) [18], and the sampling time interval
and IBRs initiated at 𝑡 = 0𝑠. At 𝑡 = 3.3𝑠, a three-phase-to- √3
(50 microseconds) play crucial roles in this process.
ground fault occurs at Bus 10, cleared after four cycles of
the fundamental frequency (60 Hz). Furthermore, at 𝑡 = 7𝑠,

IJISRT24MAR1279 www.ijisrt.com 1680

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1279

B. Model Identification using Generic SINDY Continuing in this scenario, involving sudden and
gradual changes to system parameters in transient states, the
 General Power System (SG driven): generic algorithm captures the system nonlinear dynamics
The results in the first scenario demonstrate SINDy's and tracks changes, while experiencing a minor fluctuation
accuracy in estimating the system model. For the steady- during these transitions, resulting in slight increases in
state dynamics of the power system supplied by SG, the approximation errors. However, the results shows that the
sparse model adeptly replicates the dynamics observed in generic algorithm swiftly recovers its accuracy once
measurements. As depicted in Figure 2.a, the reconstructed changes are detected, as evident in faults and load
waveform resulting from identified model closely aligns variations, shown in Figure 2.b and Figure 2.c, respectively.
with the actual data collected from the power network, all
presented per unit. Notably, the algorithm not only correctly  Power System Integrated with IBRs:
identifies the terms governing the dynamics but also When IBRs were introduced into the power grid, the
accurately determines the associated coefficients, with performance of generic algorithm in model identification
deviations well within a remarkable 0.03%. deteriorated since it could not distinguish the multi-scale
nonlinear dynamics. The result obtain in this scenario
indicates that the algorithm’s errors experience a notable
increase.

Fig 2: The Actual Measurement and Generic Algorithm Approximation during a) Steady State, b) Fault and c) Load Change in
Power System Supplied by SG

Looking into the detail, by comparing the voltage of the approximated voltage from, shown with dash lines in
variations approximated through the generic algorithm and Figure 3, indicates the inclusion of harmonic distortions in
the actual measurement it is observed that the generic the final result, comparing to the voltage waveform
algorithm failed to distinguish the nonlinearity caused by approximated in previous scenario, i.e., first-order system.
fast dynamics, in coupled multi-scaled dynamics. As Furthermore, the phase, frequency and amplitudes of the
presented in Figure 3, the generic algorithm did not approximated voltages shows deviation from cycle to cycle.
recognize the fast dynamics in the voltage. Detailed analysis

IJISRT24MAR1279 www.ijisrt.com 1681

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1279

Fig 3: The Actual Dynamic and Generic Algorithm Approximation during Steady State in Power System Supplied by SG and
IBRs.

The investigation continued by increase of the IBR of IBRs in power system led to a five-fold increase in errors
penetration to 100 % and the result, obtained from generic compared to scenarios with conventional SG, while
algorithm was subjected to the same analysis. After achieving a 100% penetration of IBRs resulted in
analyzing the voltage waveform approximated by generic approximately seven times higher errors, as illustrated in
algorithm in second and third scenario the error from actual Figure 4.
data was calculated. The findings indicate 50 % penetration

Fig 4: The Errors between Actual Dynamic and Generic Algorithm Approximation during Steady State and Transient in Power
System

The introduction of harmonic distortion and deviation C. Model Identification using Enhanced Mixed Algorithm
in voltage, frequency and phase by the generic algorithm in (Generic SINDy and HAVOK Decomposition):
power its approximation is due to inability of recognizing The second and third scenarios were repeated using the
the fast dynamics. The generic algorithm could not identify enhanced algorithm, using combination of generic SINDy
the second-order nonlinearity in the system, hence, translate and HAVOK decomposition. The voltage was
it into inaccurate level of first-order nonlinearity showing approximated using the mixed algorithm and has been
that the presence of IBRs in the power grid introduces compared to the actual data in both scenarios. Figure 5
complexities or dynamics not adequately captured by present the approximated voltage and the actual
generic algorithm. measurement for the voltage in the third scenario (100%
penetration of IBRs) for steady and transient states. The
errors calculated for the second and third scenarios were
1.14 and 1.79 times of the base error (1.14𝜖 and 1.79𝜖
where 𝜖 is the error captured in the first scenario).

IJISRT24MAR1279 www.ijisrt.com 1682

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1279

Fig 5: The Actual Measurement and Mixed Algorithm Approximation during a) Steady State and b) Transient
State in Power System Supplied by IBR

The results demonstrate a notable enhancement in the Although the effectiveness of the algorithm is
approximation of the power system with second-order demonstrated with case studies using a small-scale IEEE
nonlinearity. The mixed algorithm was able to recognize the model, the benefits of integrating various data analytic tools
fast dynamic within the coupled multi-dynamic systems. are clearly shown from a broader perspective.
The enhanced algorithm was able to capture and follow the
dynamic in steady state, and transient state, shown in Figure Currently, we are collaborating with power and utility
5.a and Figure 5.b respectively. Note that the absence of companies to develop larger-scale models and obtain real
voltage fluctuations in the fault incident are due to the data from measurements to apply the mixed algorithm in
limitation of the IBRs in providing fault current, leading to building measurement-based tools for detecting nonlinear
fast and hardly recognizable impulses. This indicates that dynamics, such as sub-synchronous oscillations and
the combined method can characterize the complex resonance, as well as complex transient energy waves
interactions and dynamics present in the system. caused by inverter interactions.

IV. CONCLUSION ACKNOWLEDGMENT

This paper presents the concept and generic We express our profound appreciation to all those who
framework of a mixed algorithm that could be used for contributed directly or indirectly to the successful
measurement-based power system analysis. To address the completion of this research paper with their valuable time,
challenging impact of coupled multi-scale dynamics of insights, recommendations, and support.
IBRs on the complex nonlinear dynamics experienced in the
power grid, we integrate the HAVOK decomposition REFERENCES
method with the powerful SINDy analytics recently
developed in data science, along with a set of illustrative [1]. Deng, Zhuofu, Binbin Wang, Yanlu Xu, Tengteng
case studies in power system analysis. Xu, Chenxu Liu, and Zhiliang Zhu. "Multi-scale
convolutional neural network with time-cognition
The case studies not only demonstrate the for multi-step short-term load forecasting." IEEE
effectiveness of detecting various nonlinear dynamics in Access 7 (2019): 88058-88071.
power systems but also clearly demonstrate the promising [2]. Hong, Seng-Phil. "Different Numerical Techniques,
capability of the mixed algorithm to separate the nonlinear Modeling and Simulation in Solving Complex
dynamics induced by coupled IBRs from those caused by Problems." Journal of Machine and
synchronous machines. Computing (2023): 058-068.
[3]. Viberg, Mats. "Subspace-based methods for the
identification of linear time-invariant
systems." Automatica 31, no. 12 (1995): 1835-1851.

IJISRT24MAR1279 www.ijisrt.com 1683

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1279

[4]. Caponetto, Riccardo. Fractional order systems: [16]. Champion, Kathleen P., Steven L. Brunton, and J.
modeling and control applications. Vol. 72. World Nathan Kutz. "Discovery of nonlinear multiscale
Scientific, 2010. systems: Sampling strategies and
[5]. Steele, Brenden, Reza SaeedKandezy, Peter Huang, embeddings." SIAM Journal on Applied Dynamical
and John N. Jiang. "Study on Impact of Modulational Systems 18, no. 1 (2019): 312-333.
Instability of Inverters on Electrical Energy [17]. Kaheman, Kadierdan, J. Nathan Kutz, and Steven L.
Systems." In 2022 IEEE Kansas Power and Energy Brunton. "SINDy-PI: a robust algorithm for parallel
Conference (KPEC), pp. 1-6. IEEE, 2022. implicit sparse identification of nonlinear
[6]. Geiser, Juergen. Multicomponent and Multiscale dynamics." Proceedings of the Royal Society A 476,
Systems. Springer International Publishing, 2016. no. 2242 (2020): 20200279.
[7]. Brunton, Steven L., Joshua L. Proctor, and J. Nathan [18]. Josse, Julie, and Sylvain Sardy. "Adaptive shrinkage
Kutz. "Discovering governing equations from data of singular values." Statistics and Computing 26
by sparse identification of nonlinear dynamical (2016): 715-724.
systems." Proceedings of the national academy of
sciences 113, no. 15 (2016): 3932-3937.
[8]. Fasel, Urban, Eurika Kaiser, J. Nathan Kutz, Bingni
W. Brunton, and Steven L. Brunton. "SINDy with
control: A tutorial." In 2021 60th IEEE Conference
on Decision and Control (CDC), pp. 16-21. IEEE,
2021.
[9]. Stanković, Alex M., Aleksandar A. Sarić, Andrija T.
Sarić, and Mark K. Transtrum. "Data-driven
symbolic regression for identification of nonlinear
dynamics in power systems." In 2020 IEEE Power &
Energy Society General Meeting (PESGM), pp. 1-5.
IEEE, 2020.
[10]. Lakshminarayana, Subhash, Saurav Sthapit, and
Carsten Maple. "Application of physics-informed
machine learning techniques for power grid
parameter estimation." Sustainability 14, no. 4
(2022): 2051.
[11]. Hamid, Asif, Danish Rafiq, Shahkar Ahmad Nahvi,
and Mohammad Abid Bazaz. "Power Grid parameter
estimation using Sparse Identification of Nonlinear
Dynamics." In 2022 International Conference on
Intelligent Controller and Computing for Smart
Power (ICICCSP),. IEEE, 2022.
[12]. Lakshminarayana, Subhash, Saurav Sthapit, and
Carsten Maple. "A Comparison of Data-Driven
Techniques for Power Grid Parameter
Estimation." arXiv preprint
arXiv:2107.03762 (2021).
[13]. Brunton, Steven L., and J. Nathan Kutz. "Methods
for data-driven multiscale model discovery for
materials." Journal of Physics: Materials 2, no. 4
(2019): 044002.
[14]. Fasel, Urban, Eurika Kaiser, J. Nathan Kutz, Bingni
W. Brunton, and Steven L. Brunton. "SINDy with
control: A tutorial." In 2021 60th IEEE Conference
on Decision and Control (CDC), pp. 16-21. IEEE,
2021.
[15]. Machowski, Jan, Zbigniew Lubosny, Janusz W.
Bialek, and James R. Bumby. Power system
dynamics: stability and control. John Wiley & Sons,
2020.

IJISRT24MAR1279 www.ijisrt.com 1684