Algorithm Design for Improved LSTM Neural Network Microgrid Load Prediction Based on Expert Rules
A modified long short-term memory(LSTM)neural network load prediction algorithm based on expert rules was designed to address the problems of low efficiency and accuracy in load forecasting in microgrids caused by high system complexity,obvious coupling phenomena,and disturbance effects in traditional BP neural network algorithms.The use of LSTM neural networks to replace the original BP neural network load prediction algorithm,as well as the use of historical time series data to predict the load in microgrids,improved the accuracy of prediction;the expert rules were introduced for real-time tuning of membership parameters in LSTM neural networks;by utilizing expert rules to organize and summarize prior knowledge,and matching corresponding specialization mechanisms to optimize parameters,the accuracy and efficiency of load forecasting were improved.Building a digital simulation model for microgrid load prediction in MATLAB simulation platform,and compared with traditional BP neural network algorithm and model prediction algorithm,the improved LSTM neural network can improve the accuracy and efficiency of load prediction to over 95％,improve the shortcomings of BP neural network in load prediction,and enhance the accuracy and efficiency of load prediction.
万方数据
维普
