Short-term Load Forecasting Of Power Users Based On RBF Neural Network Trained By Particle Filter Algorithm And System Implementation

Short-term power load forecasting is the basic basis for power supply companies to carry out power dispatching and formulate power trading plans.With the development of the power spot market,the importance of power load forecasting for the development of power supply enterprises will become more prominent.At present,power supply enterprises generally use the classical load forecasting methods such as regression analysis to complete the power load forecasting.The low prediction accuracy can only be performed by using historical data for offline fitting,and the prediction real-time performance is not high.The new method of researching load forecasting realizes the online prediction of user’s power load,So this paper mainly includes the following parts:1.An online recursive prediction method for short-term load forecasting is proposed.A new method based on particle filter and its improved algorithm is adopted.The calculation speed meets the real-time requirements.2.The multi-dimensional input state model of electric load is proposed.By analyzing the factors affecting electric load at different time points,the state model based on RBF neural network is built,and the influence of external factors such as weather is superimposed,which improves the accuracy of load forecasting.3.Develop short-term load forecasting system based on MATLAB: The App designer component is based on MATLAB.Finally,it is packaged and deployed as a common.exe file for Windows system.This thesis studies the online recursive prediction algorithm and realizes the high accuracy and low delay of the user’s power load with the help of MATLAB platform.The online forecasting function greatly improves the user’s power load forecasting business capability of the unit.