The Power System Load Forecasting Based On Wavelet Neural Network

In power system the purpose on increasing the precision of load forecasting makes it possible in improving security operation condition and saving economic consumption. Nowadays most of mid-term load forecasting methods are derived from traditional linear statistics theory. Due to the complicacy and uncertainty of load forecasting make it difficult to obtain high accuracy forecasting result.Wavelet neural network based on wavelet theory is a novel feed-forward neural network and have many better character. On review the research of national and international power system load forecasting status this paper explore power system load forecasting correlated wavelet theory which is developed fast in the last few years. Basing on the character analyses of electrical load frequency spectrum combined with wavelet transform and Fourier transform we determine wavelet function and wavelet time series analytical level.A kind of"Decomposition-Reconstruction-Prediction-Synthesis"wavelet neural network method is proposed in power system load forecasting. Through the wavelet transform the monthly power load series are decomposed and reconstructed on the original sequence scale. We obtain different bandwidth sub-sequences which have the same length with original sequence. And then these sub-sequences are predicted by BP neural networks. The finally predict result can be obtained from the synthesis forecasting value of sub-sequences. These methods integrate the merits of time-frequency localization features in wavelet analysis and non-linearity approach in neural networks and have better approach and forecast ability.The predicting experiment shows that the proposed method possesses high forecasting accuracy and better adaptability than traditional forecasting methods when small quantity monthly load series dada are used. We improve the lower accuracy status in the condition of shortage training set sample and give out a new idea in small quantity dada power load forecasting.