Short-Term Load Forecasting Based On Improved BP Neural Network

This thesis carries on to the meaning and actual states of Load forecasting to say all first, analyzing the characteristic of load and Load forecasting, tallying up composition and sorts and load periodic change regulations of load, analyzing every kind of factor of load of impact. Analyzed history load of XuZhou, consider every kind of impact, apply improved BP neural network, create short-term load forecasting models. At input load, the type of date and every kind of weather factors that considered the close by date in the variable, enter to unify the transaction to the input quantity, to the temperature, rain to shine on etc. the factor to put forward the special disposal with optics. The training of the network applies the Levenberg- Marquardt calculate way, putting up the velocity and forecasting accuracies of the neural network on the very big degree.On the foundation of the over - face research, applied MATLAB 6.5 software, use the neural network tool box plait distance, carry out short-term load forecasting. Pass applied good foreground for imitating true forecasting, prove the neural network forecasting models that this thesis create, can well consider every kind of factor of impact, predict accuracy very good, and can further apply in actual Load forecasting.