Intelligent management and resilience assessment of integrated energy system in industrial park based on digital twin and deep reinforcement learning

Abstract

The integrated energy system (IES) of industrial park is the key carrier to achieve the “double carbon” goal, but its characteristics such as strong coupling of multi-energy flow, high proportion of new energy access and multiple uncertainties pose severe challenges to the intelligent operation and resilience of the system. In order to realize the economical, reliable and low-carbon operation of the system in complex environment, this paper proposes an intelligent management and resilience evaluation framework that integrates digital twin (DT) and deep reinforcement learning, and designs a deep reinforcement learning agent based on flexible actor-critic algorithm. Moreover, its reward function integrates operating cost, new energy consumption and resilience indicators, so as to realize the collaborative optimal dispatching of multi-energy flow of cooling, heat, electricity and gas. The simulation results on the public dataset show that compared with the traditional model predictive control method, the proposed method reduces the average daily operating cost of the system by 22.1%, increases the new energy consumption rate to 96.8%, and increases the system toughness index by 35.7% under the simulated load change and equipment failure scenarios. In general, this study not only provides an explainable intelligent solution for real-time optimization decision-making of IES, but also the “perception-decision-evaluation” integrated framework constructed in this paper provides theoretical tools and practical references for the planning and operation of high-toughness parks.