Bidirectional LSTM (Long Short Term Memory) network microgrid scheduling method under constraint condition

The invention discloses a bidirectional LSTM (Long Short Term Memory) network micro-grid scheduling method under constraint conditions, which can establish a mapping relation between a micro-grid operation scene and a scheduling decision result based on a micro-grid optimization scheduling model of a bidirectional LSTM network, and train a bidirectional LSTM model by adding constraint condition data to a reference prediction scheme. Therefore, a more stable optimal scheduling strategy is generated. By comparing the performance of different models on the mean square error and the threshold over-limit average value and the performance of the models on different strategies, the mean square error and the threshold over-limit average value of the prediction data are remarkably reduced, the risk that the scheduling scheme exceeds the bearing capacity of the micro-grid under the limited condition is reduced, and the scheduling performance of the micro-grid is improved. The problem that when a deep learning model is