Power System Quasi-Steady State Estimation: An Echo State Network Approach

TL;DR

Problem: track quasi-steady-state points in power systems undergoing slow injection changes. Approach: a reservoir-computing QSSE using Echo State Networks with online ridge regression and LMS training to learn nonlinear power-flow mappings from measurements. Findings: ESN achieves comparable or better estimation accuracy than WLS, EKF, and PF on IEEE 14-bus and 300-bus systems, with substantially lower computation times. Significance: provides a fast, model-free QSSE method suitable for real-time, large-scale grid monitoring.

Abstract

The operating point of a power system may change due to slow enough variations of the power injections. Rotating machines in the bulk system can absorb smooth changes in the dynamic states of the system. In this context, we present a novel reservoir computing (RC) method for estimating power system quasi-steady states. By exploiting the behavior of an RC-based recurrent neural network, the proposed method can capture the inherent nonlinearities in the power flow equations. Our approach is compared with traditional methods, including least squares, Kalman filtering, and particle filtering. We demonstrate the estimation performance for all the methods under normal operation and sudden load change. Extensive experiments tested on the standard IEEE 14-bus and 300-bus cases corroborate the merit of the proposed approach.

Figures (1)

• Figure 1: The architecture of an echo state network.