Techniques of applying wavelet transform into combined model for short-term load forecasting

Techniques of applying wavelet transform into combined model for short-term load forecasting are presented in this paper. The analysis shows that the load can be described by the corresponding components in the time frequency domain. It allows the decomposition of a load signal into different levels of resolution scales with the wavelet transform. Incorporating the masking parameter in the corresponding scale, useless information could be effectively removed, and a denoising algorithm can also be developed. It is found that even the model works well for certain load components, it will be not suitable for other components because it cannot consider every factor. So different combined forecast methods are chosen in each scale. The data preprocessing, different combined forecast methods adopted in different scales are introduced. Finally, the forecasting results can be obtained by the reconstruction of the forecast results in different scales. Case studies demonstrate that the proposed method can offer high forecasting precision. (c) 2005 Elsevier B.V. All rights reserved.