Learning Optimization for Dispatch of Interregional Power Grid under Uncertain Environment

This study investigates the dispatch problem in interregional power grids with uncertain renewable energy sources and loads. A knowledge-transfer-based hierarchical learning optimization method is designed to search for the optimal dispatch strategy. First, the model of various sources, loads, and tie-lines are established according to each components characteristics in the system. The dispatch process of an interregional power grid is correspondingly divided into tie-line dispatch in the upper layer and both generator units and flexible load dispatch in the lower layer. The dispatch optimization problems of both layers are formulated with a hierarchical Markov decision process model. Finally, a learning optimization method is adopted to obtain the dispatch strategy, in which a pre-learning technique is incorporated. A case study is performed to validate the effectiveness of the proposed method, and the simulation results show that the pre-learning technique improves the learning speed and reduces the cost of relearning.