基于深度强化学习的风电场储能系统预测决策一体化调度

风电场储能系统的优化控制可以提高风电场作为发电商在电力市场中的竞争力。文中提出基于深度强化学习的储能系统预测决策一体化调度方法,令高维度的风电场原始测量数据直接驱动储能系统。与预测、决策相分离的传统调度模式相比,预测决策一体化调度模式将风电功率预测与储能系统动作决策相融合,避免了预测阶段中有效决策信息的损失,使风电的随机规律无须通过数学假设等方式被人为刻画,避免了建模误差。其次,引入深度强化学习Rainbow算法优化风电场测量数据与储能系统动作指令之间端到端的控制策略,该策略具备动态统筹多时段系统收益的能力。最后,基于风电场历史数据的算例分析,验证了所提一体化调度模式的优越性和深度强化学习应对不确定性问题的有效性。

Prediction and Decision Integrated Scheduling of Energy Storage System in Wind Farm Based on Deep Reinforcement Learning

The optimal control of the energy storage system (ESS) in the wind farm can improve the competitiveness of the wind farm, which is a power producer in the electricity market. In this paper, a prediction and decision integrated scheduling method of ESS of the wind farm is proposed based on deep reinforcement learning, in order to make the original high-dimensional measurement data of the wind farm directly drive the ESS. Compared with the traditional scheduling model that separates prediction and decision, the prediction and decision integrated scheduling model combines the wind power prediction and ESS operation decision together, to avoid the loss of effective decision-making information in the prediction period. Stochastic laws of wind power do not need to be artificially portrayed through mathematical assumptions and modeling errors are avoided. Then, Rainbow algorithm, a deep reinforcement learning algorithm, is introduced to optimize the end-to-end control strategy between the wind farm measurement data and the ESS operation instruction, which can dynamically coordinate multi-period system profits. Finally, by analyzing case studies based on the wind farm historical data, the superiority of the proposed integrated scheduling mode and the effectiveness of deep reinforcement learning in dealing with the uncertainty problems are verified.