A Physics Informed Machine Learning Method for Power System Model Parameter Optimization

TL;DR

This paper introduces a physics-informed machine learning approach utilizing automatic differentiation and gradient descent to identify and optimize uncertain parameters in power system models, demonstrated on a single machine system.

Contribution

It presents a novel gradient-based optimization method leveraging automatic differentiation for power system parameter identification and optimization, applicable to various system configurations.

Findings

Effective parameter identification demonstrated on a single machine system.

Method shows potential for broad applicability in power system modeling.

Gradient descent with automatic differentiation enables efficient optimization.

Abstract

This paper proposes a gradient descent based optimization method that relies on automatic differentiation for the computation of gradients. The method uses tools and techniques originally developed in the field of artificial neural networks and applies them to power system simulations. It can be used as a one-shot physics informed machine learning approach for the identification of uncertain power system simulation parameters. Additionally, it can optimize parameters with respect to a desired system behavior. The paper focuses on presenting the theoretical background and showing exemplary use-cases for both parameter identification and optimization using a single machine infinite busbar system. The results imply a generic applicability for a wide range of problems.