Method for LSTM-Based Cascade Hydropower Plant Scheduling

With the increase of the number of cascade hydropower plants and the increase of the discrete number of state variables in the time period to obtain more high-precision solutions, the calculation time increases exponentially, which makes it difficult to optimize through physical models. Therefore, a data-driven cascade hydropower plant scheduling method with self-learning abilities is proposed. Firstly, perform cluster pre-processing of the historical dispatching data of hydropower plants in the scheduling scope with K-means algorithm; secondly, establish a deep learning model of cascade hydropower plant schedules based on long short-term memory (LSTM), and build the mapping model among system load, water regimen and hydropower plant schedule through historical data training, to make a decision; finally, continuously correct the model by accumulating the historical data, so that it has the ability of self-evolution and self-learning. We make case analysis based on the actual power grid data, and the calculation results show the effectiveness of the proposed method.