摘要
新能源出力和用户负荷的随机性使得综合能源系统的优化调度问题充满挑战,本工作提出一种考虑冷热电互补及储能系统的多园区综合能源协调优化调度方法。首先,引入虚拟电厂技术,将综合能源系统内部资源聚合,考虑园区内的能量储存装置以及园区间的能量传输装置,构建虚拟电厂框架下的多园区综合能源系统结构;其次,建立热电联供单元、储能系统、冷热管道等内部资源模型,以最大化系统收益为目标,建立多虚拟电厂综合能源系统协调调度模型;最后,构建深度强化学习框架,再将Q值迁移方法引入到深度强化学习算法中,基于改进的深度确定性策略梯度算法进行连续状态和动作空间下的优化调度决策。通过算例分析,验证了所提方法的有效性,结果表明基于改进算法的多园区综合能源系统协调优化调度能够有效实现资源的合理分配以及各个园区间的供能互补,降低了系统运行成本。
The randomness of new energy output and user load makes the optimal dispatching of the integrated energy system full of challenges,and a multi-park integrated energy coordinated optimal dispatching method considering the complementary cooling,heating and power and energy storage systems is proposed.First,the virtual power plant technology was introduced to aggregate the internal resources of the integrated energy system,and the energy storage devices in the park and the energy transmission devices between the parks were considered,and a multi-park integrated energy system structure under the framework of the virtual power plant was constructed;secondly,the thermal power system was established.Internal resource models such as co-supply units,energy storage systems,and cooling and heating pipelines,with the goal of maximizing system revenue,established a coordinated scheduling model for a multi-virtual power plant integrated energy system;finally,a deep reinforcement learning framework was built,and the Q value was migrated The method is introduced into the deep reinforcement learning algorithm.Based on the improved deep deterministic policy gradient algorithm,the optimal scheduling decision is made in the continuous state and action space.The effectiveness of the proposed method is verified through the analysis of calculation examples,and the results show that the coordinated and optimized dispatch of the multi-park integrated energy system based on the improved algorithm can effectively realize the reasonable allocation of resources and the complementary energy supply between the parks,and reduce the operating cost of the system.
作者
李昊
刘畅
苗博
张静
LI Hao;LIU Chang;MIAO Bo;ZHANG Jing(China Electric Power Research Institute,Beijing 100192,China)
出处
《储能科学与技术》
CAS
CSCD
北大核心
2022年第5期1482-1491,共10页
Energy Storage Science and Technology
基金
国家电网有限公司总部科技项目资助:客户侧园区多元能效感知与数据驱动优化调控方法研究(YD71-20-016)。
关键词
多园区综合能源系统
储能系统
Q值迁移
深度强化学习
优化调度
multi-park integrated energy system
energy storage system
Q value migration
deep reinforcement learning
optimal scheduling