A Framework for Analytical Power Flow Solution Using Gaussian Process Learning

Pareek, Parikshit; Nguyen, Hung D.

doi:10.1109/tste.2021.3116544

Cited by 11 publications

(9 citation statements)

References 60 publications

(138 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, evaluating the rate of state change is difficult in power grids, in particular for P2P dominant grids. The first cause of this difficulty is the non-linear, implicit nature of the power flow equations i.e., the absence of a direct state-injection relationship [4]. Second is that in P2P energy trading setting, the operator (e.g., DSO) does not have complete injection information.…”

Section: Critical Prosumer Identificationmentioning

confidence: 99%

“…To solve the problem of local sensitivity, we employ the recently developed closed-form power flow (CFPF) approximation [4] to obtain state-sensitivity function (SSF)-valid for an injection subspace or hypercube. The SSF is an approximate closed-form function of state-sensitivity with injection as input.…”

Section: A State-sensitivity Functionmentioning

confidence: 99%

“…The SSF is an approximate closed-form function of state-sensitivity with injection as input. If we consider voltage magnitude function at jth node V j (s) as a state function (using [4]), then the SSF is obtained by partial differentiation as ∂V j (s)/∂s. In the process of CFPF approximation, prior knowledge about power flow and application-accuracy requirements are used to select the kernel function k(•, •) [4].…”

Section: A State-sensitivity Functionmentioning

confidence: 99%

See 2 more Smart Citations

Locating Critical Prosumers in P2P Dominant Grids Using State-Sensitivity Function

Pareek

Sampath

Nguyen

et al. 2023

IEEE Trans. Smart Grid

Self Cite

View full text Add to dashboard Cite

This letter presents the idea of ranking critical prosumers in using a novel state-sensitivity function (SSF). The SSF is designed to work in a given nodal power injection hypercube, and thus the proposed ranking method captures the effect of injection subspace. An element-wise maximum Jacobian is obtained, that quantifies the worst-case sensitivity of each injection on each state. Results from various 33-Bus system configurations highlight the proposed method's capability in ranking state-injection risk pairs, providing node voltage risk levels and quantifying combined risk levels in peer-to-peer dominant grids.

show abstract

Section: Critical Prosumer Identificationmentioning

confidence: 99%

Section: A State-sensitivity Functionmentioning

confidence: 99%

Section: A State-sensitivity Functionmentioning

confidence: 99%

See 1 more Smart Citation

Locating Critical Prosumers in P2P Dominant Grids Using State-Sensitivity Function

Pareek

Sampath

Nguyen

et al. 2023

IEEE Trans. Smart Grid

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, the linear models cannot extract non-linear features of the PF function and suffer from accuracy limitations. In addition, [13] and [14] use Gaussian process regression, while it can only obtain the PDF of one target quantity for one-time training. This property prevents its application in middle and large-scale systems if the PDFs of all buses' voltage phasors are required.…”

Section: Introductionmentioning

confidence: 99%

Physics-guided Residual Learning for Probabilistic Power Flow Analysis

Chen¹

2023

Preprint

View full text Add to dashboard Cite

Probabilistic power flow (PPF) analysis is critical to power system operation and planning. It focuses on obtaining probability descriptions of system states with stochastic power injections (e.g., renewable generations and load demands). Monte Carlo Simulations are widely used in PPF analysis. Given random samples of power injections, they repeatedly run power flow (PF) solvers to obtain their corresponding voltage phasors. However, the cumulative computational time is heavy because many simulations are necessary for obtaining accurate probability descriptions. Therefore, reducing the computational time of individual PF analysis can relieve the total computational burden of PPF analysis. Inspired by residual neural networks (ResNets) structure, we propose a novel neural network framework designed for PF analysis. We add an extra fully connected linear layer between the inputs and outputs to the multilayer perceptions (MLPs) structure. In addition, based on our proposed framework, we design three methods to initialize the weights of the shortcut connection layer according to the physical characteristics of AC-PF equations. Numerical tests show that our proposed methods achieve higher accuracy in estimating voltage phasors and branch flows than existing methods. In addition, three meticulously designed initialization schemes help converge faster than random initialization in the training process.

show abstract

“…However, these linear models suffer from accuracy limitations because they cannot extract non-linear features of the PF functions. In addition, [16] and [17] use Gaussian process regression, which can only obtain the PDF of a single target quantity in one shot. This property prevents its application in medium-and large-scale power systems if the PDFs of all buses' voltage phasors are required.…”

Section: Introductionmentioning

confidence: 99%

Physics-Guided Residual Learning for Probabilistic Power Flow Analysis

Chen,

Zhang

2023

IEEE Access

View full text Add to dashboard Cite

Probabilistic power flow (PPF) analysis is critical to power system operation and planning. PPF aims at obtaining probabilistic descriptions of the state of the system with stochastic power injections (e.g., renewable power generation and load demands). Given power injection samples, numerical methods repeatedly run classic power flow (PF) solvers to find the voltage phasors. However, the computational burden is heavy due to many PF simulations. Recently, many data-driven based PF solvers have been proposed due to the availability of sufficient measurements. This paper proposes a novel neural network (NN) framework which can accurately approximate the non-linear AC-PF equations. The trained NN works as a rapid PF solver, significantly reducing the heavy computational burden in classic PPF analysis. Inspired by residual learning, we develop a fully connected linear layer between the input and output in the multilayer perceptron (MLP). To improve the NN training convergence, we propose three schemes to initialize the NN weights of the shortcut connection layer based on the physical characteristics of AC-PF equations. Specifically, two model-based methods require the knowledge of system topology and line parameters, while the purely data-driven method can work without power grid parameters. Numerical tests on five benchmark systems show that our proposed approaches achieve higher accuracy in estimating voltage phasors than existing methods. In addition, three meticulously designed initialization schemes help the NN training process converge faster, which is appealing under limited training time.INDEX TERMS Data-driven, neural network, probabilistic power flow, physics-guided initialization, residual learning.

show abstract

A Framework for Analytical Power Flow Solution Using Gaussian Process Learning

Cited by 11 publications

References 60 publications

Locating Critical Prosumers in P2P Dominant Grids Using State-Sensitivity Function

Locating Critical Prosumers in P2P Dominant Grids Using State-Sensitivity Function

Physics-guided Residual Learning for Probabilistic Power Flow Analysis

Physics-Guided Residual Learning for Probabilistic Power Flow Analysis

Contact Info

Product

Resources

About