Research On The Power Load Forecasting Method Based On CNN-BILSTM

Electric load forecasting is one of the key technologies in the smart grid which mainly explores the changing laws and influencing factors of electric load and predicts the load value at a certain time in the future.Accurate power load forecasting provides a scientific basis for power planning,power dispatch,load control and economic operation,thus it has great significance to the stable and reliable operation of smart grids.The key issue the power load forecasting facing at present is how to accurately establish the model and express the characteristics of the power load influencing factors in order to improve the forecasting accuracy.This thesis proposes a power load forecasting model of CNN-BiLSTM combined neural network comprehensively utilizing the feature expression ability of convolutional neural network(CNN)and the capability to process time sequence relationship of bidirectional long shot-term memory(BiLSTM)though taking into account the periodicity and randomness of power load data.In addition,an online update method is proposed to solve the online update problem of the firmware aiming at the problem including complex,time-consuming,labor-intensive,and low-reliability of the on-site update process of the underlying firmware application software of the online power load forecasting platform.The main research is summarized as follows:(1)This thesis systematically studies the influence of random factors and periodic laws on the power load and the CNN-BiLSTM combined neural network model is designed,analyzed and optimized depending on the public power load data in a certain area as a sample The results show that the root mean square error of the prediction results of the CNN-BiLSTM combined neural network is reduced by 42.83%,39.71%and 35.85%compared with the three baseline models of LightGBM,LSTM and CNN-LSTM to verify the effectiveness of the CNN-BiLSTM model proposed in this paper.(2)The design of the boot program for the online upgrade of the underlying firmware of the power load forecasting platform has been carried out in order to ensure the security of grid data,the real-time maintenance of the grid and the further improvement of control capacity for the smart grid.Taking the dsPIC30F6014 single-chip microcomputer as the platform,the BootLoader program design framework and its communication protocol are introduced detailedly in this thesis,and the design of the corresponding upper computer software is also carried out.The results prove that the developed firmware boot program has complete functions,convenient process and reliable operation.