A review of reinforcement learning based approaches for industrial demand response

Abstract

Industrial demand response plays a key role in mitigating the operational challenges of smart grid brought by massive proliferation of distributed energy resources. However, industrial plants have complex and intertwined processes, which provides barriers for their participation in industrial demand response programs. This is in part due to the complexity and uncertainties of approximating systems models. More recently, reinforcement learning has emerged as a data-driven control technique for sequential decision-making under uncertainty. This emergence is strongly coupled with the abundance of data offered by advanced information technologies. The potential of applying reinforcement learning in industrial demand response is identified in this work by comparing pivotal aspects of reinforcement learning with the requirements of industrial demand response schemes.