Neural networks short-term forecasting of electricity price based on wavelet decomposition and homo-layer series combination

Short-term electricity price forecasting based on wavelet decomposition is mainly to forecast each subsequence after power price sample is decomposed and reconstruct each forecast result to get the final forecasting price.On this basis,tariff and load samples are decomposed by multi-resolution wavelet to 2 layers,and then remove the noise signals.Then homo-layer series of power price and load are combined as the input of neural networks.According to time frequency character,neural network model is designed and built to forecast.At last forecast results of each sub series are reconstructed to get forecast price.In the numerical example analysis,the data in PJM market from March 2007 to February 2008 is adopted.Error duration curves are drawn to verify the effectiveness and feasibility of the proposed forecasting model in contrasting with alternatives.