Short-term load forecasting method based on wavelet and reconstructed phase space

This paper proposed a multi-scale prediction model for short-term electric load, which makes merit of wavelet decomposition and different local approximation algorithm in reconstructed phase space. The load series are non-stationary, observed from a chaotic dynamical system. With the view of wavelet multi-resolution analysis, the original load series are decomposed with A Trous algorithm into a relatively stationary approximation and a set of detail coefficients, which are taken as independent dynamical subsystems. The subsystems are coped with different local approximation algorithms, producing coefficients at prediction point. The coefficients are reconstructed with wavelet method to obtain prediction value. Case study demonstrates that the proposed model has a good performance for less-step predictions, compared with the result from direct chaotic prediction method to the original series.