Study On Short-term Power Load Forecasting Method Based On Wavelet Neural Network

Electricity is a kind of indispensable energy in our daily life, at the same time lots of electricalequipments have been widely used in production and life. And how to using electrical energy effectivelybecomes a huge problem for large voltage users. Power load forecasting is not only applied in high side ofpower grid, but also for low voltage users, it provides effective basis for power planning, monitoring andeffective using.The target of this thesis is to making short-term power load forecasting for a24-hour continuousoperation lab of a large monitor enterprise in Tianjin. The load of this lab makes a periodic transformation,and the power consumption is different for different subjects. Load forecasting objects are so volatile andthe factors which influence the accuracy are various. This paper proposes a new kind of algorithm model toachieve the load forecasting. It’s based on the neural networks combined with wavelet transform. In orderto verify the validity of the wavelet neural network model, we use the traditional BP neural network (BP),loose model wavelet neural network (WNN1) and fusion model wavelet neural network (WNN2) to realizeload forecasting. We introduce three kinds of algorithm models in detail. After normalization processing ofthe same input data, we make the short-term load forecasting, and compare the error and the forecastingaccuracy of three kinds of models. We discover that only WNN2network model can meet the requirementsof the short-term prediction accuracy. WNN2has better learning ability and faster convergence speed thanthe other methods.In this paper, the prediction software is designed in a mixed programing method by using VC6.0andMatlab. Algorithm programming in Matlab has features of simple, flexible and better extensibility; MatlabCOM generator components is used to transform the algorithm program in Matlab to the function whichcan be called for VC. At the same time, we combined VC with SQL Server to achieve the storage of thedata and information in database.By using the short-term forecasting algorithm and the forecast software designed for low voltagepower users, we can predict the daily load accurately, and provide a better basis for effective utilization ofenergy.