Distributed Collaborative Optimal Dispatch of Multi-park Integrated Energy System Based on Bilayer Imitation Learning

Aiming at the multiple uncertainty factors such as source-load and electricity price as well as privacy protection problems in the collaborative dispatch of multi-park integrated energy system, a bilayer distributed collaborative optimization model is proposed. The upper layer of the model adopts the communication neural network (CommNet) to decide the energy storage actions of each park based on real-time information, and conducts supervised training through imitation learning, enabling the agent to obtain the integrated function of prediction and decision-making. In the lower layer, each park adopts the alternating direction method of multipliers (ADMM) to carry out distributed optimization, and obtains the actions of other equipment and the power interaction between the parks, forming a complete multi-park distributed collaborative optimal operation scheme for the current period. This paper also proposes an inter-park transaction mechanism considering the real-time electricity price of the upper power grid and the supply and demand relationship of the multi-park integrated energy system to protect the interests of each park. The numerical example proves that the proposed method does not rely on the accurate prediction of the uncertainty factors such as source-load and electricity price, and can achieve performance close to the theoretically optimal strategy on the premise of protecting the data privacy in each park. © 2022 Automation of Electric Power Systems Press. All rights reserved.