A Reinforcement Learning-Based Method to Coordinated Multi-energy Optimization

Abstract

Aiming at the current problems of emission reduction and reliability to be optimized in multi-energy co-optimization related research results, a multi-energy co-optimization method in integrated energy system based on reinforcement learning is proposed. With the objective function of lowest total system cost, highest reliability and highest emission reduction rate, and the constraints of reliability, heat balance, equipment operation, energy storage and demand response, a multi-energy collaborative optimization model is constructed. The objective model is solved by a reinforcement learning algorithm, which uses the fast optimization performance of reinforcement learning to gradually approach the theoretical optimal solution, dynamically maintains the optimal solution size according to the adaptive grid density method, and optimizes the diversity of the optimal solution set by adaptive chaos optimization, and finally selects the best update particle for the state space by the optimal solution selection scheme. The algorithm stops when the conditions of optimal solution or maximum number of iterations are met, and the optimal solution is output to obtain a multi-energy collaborative optimization scheme that meets the target model. The experiments show that this method can effectively improve the system reliability and has strong robustness in emission reduction and environmental protection.