基于改进TD3的综合能源优化调度研究

Research on Integrated Energy Optimal Scheduling Based on Improved TD3

针对综合能源系统的经济优化调度问题,提出一种基于优先经验回放机制与绝对均值法的双延迟深度确定性策略梯度算法(TD3),优先经验回放机制通过区分样本价值,优化采样过程,绝对均值法计算TD误差,确保样本价值的可靠性。以系统总运行成本为指标,构建系统调度模型,并设计环境状态、调度动作和奖励函数。采用某高校微电网算例仿真,结果表明所提算法较TD3算法、深度确定性策略梯度算法(DDPG)和CPLEX求解器能更有效的协调设备出力,提升系统的经济性。

Aiming at the economic optimal scheduling problem of the integrated energy system,a twin delayed deep deterministic strategy gradient algorithm(TD3)based on priority experience playback mechanism and absolute mean method is proposed.The priority experience playback mechanism optimizes the sampling process by distinguis⁃hing the sample value,and the absolute mean method calculates the TD error to ensure the reliability of the sample value.Taking the total operation cost of the system as the index,the system scheduling model is constructed,and the environment state,scheduling action and reward function are designed.The simulation results of a microgrid in a uni⁃versity show that the proposed algorithm can coordinate the output of equipment more effectively and improve the e⁃conomy of the system than the TD3 algorithm,deep deterministic policy gradient(DDPG)algorithm and CPLEX sol⁃ver.