Research And Implementation Of LSTM-based Power Analysis And Load Forecasting System

In recent years,under the background of current electricity marketization,the generation mode of production on demand is gradually replacing the production mode of generation on schedule,and power load forecasting has received more and more attention as an important reference indicator for power dispatch and generation planning.The traditional load forecasting methods,which are good for smooth data forecasting,all need to be improved in terms of forecasting accuracy when facing the increasingly complex factors affecting load changes.In this article,we propose a Bi LSTM load forecasting model(WOA-Attention-Bi LSTM)with whale algorithm to optimize the model parameters and integrate the attention mechanism,design experiments to verify that the model has higher forecasting accuracy,and implement the forecasting algorithm optional power load forecasting system.The main research results of this article are as follows:(1)Power load characteristics analysis and data pre-processing are carried out.Based on the theoretical knowledge of electric load,the law of load change and the factors affecting load change are analyzed,and the Pearson correlation coefficient method is used to compare the correlation,eliminate the factors with very weak correlation,and pre-process the collected data.(2)An LSTM-based electric load forecasting model was designed.Firstly,experiments were designed to determine the key parameter of time step in power load forecasting,and then several sets of experiments were designed to improve the model step by step,and the Attention-Bil STM model with the introduction of the attention mechanism,and the WOA-Attention-Bi LSTM load forecasting model with optimized parameters by the whale algorithm were constructed,and the experimental results were compared,and the model was compared with the The average relative error is reduced by 1.2%,RMSE is reduced by 71.6979 MW,and MAE is reduced by 68.039 MW,compared with the prediction results of the model without parameter optimization,the error is smaller,the fit are higher,and the prediction effect is the best,which verifies the superiority of the WOA-Attention-Bi LSTM algorithm model,and also verifies that the introduction of attention mechanism and whale algorithm for model parameter optimization.(3)Designing an LSTM-based electric load forecasting system.Using Vue framework front-end system interface design and Mysql back-end data management,we realize the functions of data storage and management,calling Python model files for forecasting,and completing load forecasting results display,etc.All models used in this article can also be selected and used in the system.The WOA-Attention-Bi LSTM model proposed in this article is experimentally verified to have higher prediction accuracy in load forecasting,and the designed and implemented load forecasting system makes the electric load forecasting more convenient and accurate.