Short-term Load Forecasting Based On Neural Network

The improvement of power load forecasting precision is the key to improve the economy and safety of system operation.For the scientific and effective development of power system generation and power supply plan,we must give full play to short-term load forecasting in power grid operation,unit start and stop and power dispatch,power exchange and other important aspects.Traditional power load forecasting algorithm is not ideal,especially in the non-linear relationship between the input and output,so when using traditional power load forecasting algorithm,combined with other algorithms,including modern intelligent algorithm,can improve the prediction effect more effectively.In the aspect of improving the accuracy of power load forecasting,this paper based on the threshold of high-frequency component signal,divide and set the intermediate frequency signal deeply,carry out the unscented Kalman filter and the extended Kalman filter respectively on the low frequency,intermediate frequency and high frequency signal,thus improving the input data quality of neural network.In the meantime,for the low-frequency signal using incremental way to reduce input noise,and then design the frequency range prediction fusion algorithm,as the output of neural network training and prediction indicators.Finally,the traditional forecasting method of gray value and improved neural network are compared with TUKF-WNN forecasting results,which proves that the TUKF-WNN prediction method has higher prediction accuracy and better adaptability.A short-term load-forecasting model considering real-time electricity price isproposed.In the context of gradual opening up of the electricity market,real-time electricity price will have a certain impact on the load curve,the traditional method of forecasting the majority did not take into account the price.In this paper,based on the neural network,and different experimental cases were compared,effectively breaking the neural network of the inherent limitations and improve the reliability of load forecasting.Considering the multiple influencing factors of short-term load forecasting including electricity price,a multi-variable time series load-forecasting model is proposed.Phase space reconstruction of multivariate,the establishment of a comprehensive global forecasting model,and proved by experiments to further improving the accuracy of prediction...