网格化知识迁移学习算法及其在碳能复合流优化中的应用

建立了计及碳责任分摊的碳能复合流优化模型,并提出了一种网格化知识迁移学习算法,以便实现电网的低碳、经济、安全最优运行。算法采用二值编码的方式实现连续-离散空间的转换,以解决连续状态-动作空间的学习和维数灾难问题;从优化任务的状态信息和最优Q值之间的关系从发,构建了知识迁移的基本框架;为了避免在弱联系环境下,整体性提取状态特征信息给学习网络带来干扰,影响迁移学习的准确性,提出了一种网格化信息提取方式,分散式地对各局部特征进行提取和迁移。最后,通过IEEE 118节点系统的碳能复合流优化仿真验证了算法的有效性。

Grid Knowledge Transfer Learning Algorithm and Its Application in Carbon-Energy Combined-Flow Optimization

This paper establishes a carbon-energy combined-flow optimization model with carbon responsibility sharing, and proposes a grid knowledge transfer learning algorithm to realize the low-carbon, economical and safe optimal operation of power grid.The algorithm uses the binary coding method to realize the continuous-discrete space conversion, in order to solve the continuous state-action space learning and dimension disaster problem.This paper constructs the basic framework of knowledge migration from the relationship between the state information of the optimization task and the optimal Q value.In order to avoid the interference of the state feature information in the weak connection environment to the learning network, which affects the accuracy of the migration learning, this paper proposes a kind of grid information extraction method for decentralized extraction and migration of each local feature.Finally, the effectiveness of this algorithm is verified by the carbon-energy combined-flow optimization model of IEEE 118-bus system.