Electric Power Dispatch Instruction Generation Method Based on Sequence-to-Sequence Model

In view of the fact that the existing power dispatch instruction generation methods are difficult to meet the accuracy requirements in the actual application of distribution networks, this paper proposes a method for generating electric power dispatching instructions based on a sequence-to-sequence model (Seq2Seq). This method uses a long short-term memory network (LSTM) to build an Encoder-Decoder structure, and adds an attention mechanism and a beam search algorithm in the decoding process. The comparative experiment shows that the Encoder-Decoder structure model proposed in this paper improves the F1 score by 7.14% and 10.37% respectively compared with the similar models built by double-layer and single-layer recurrent neural networks (RNN). The ablation experiment further verifies the effectiveness of adding the attention mechanism and the beam search algorithm in the decoding process for generating electric power dispatching instructions.