Rethink AI-based Power Grid Control: Diving Into Algorithm Design

TL;DR

This paper critically analyzes DRL-based voltage control in power grids, introduces an imitation learning approach that improves generalization and reduces training time, and demonstrates its superior performance over traditional RL methods.

Contribution

It proposes a novel imitation learning method for power grid control that bypasses reinforcement learning, enhancing robustness and efficiency.

Findings

The imitation learning approach outperforms previous RL agents.

The proposed method has strong generalization capabilities.

Training time is significantly reduced.

Abstract

Recently, deep reinforcement learning (DRL)-based approach has shown promisein solving complex decision and control problems in power engineering domain.In this paper, we present an in-depth analysis of DRL-based voltage control fromaspects of algorithm selection, state space representation, and reward engineering.To resolve observed issues, we propose a novel imitation learning-based approachto directly map power grid operating points to effective actions without any interimreinforcement learning process. The performance results demonstrate that theproposed approach has strong generalization ability with much less training time.The agent trained by imitation learning is effective and robust to solve voltagecontrol problem and outperforms the former RL agents.