Wavelets based analysis of non-uniformly sampled data for power load forecasting

In this paper, the compactly supported orthonormal symmetrical wavelets is used to estimate non-uniformly sampled and non-gaussian noise corrupted load consumption signals for the purpose of load forecasting in a typical electrical utility network. Power load forecasting is an important function of utility management and present methods invariably rely heavily on past historical load curves which are collected from the grid via various monitors placed at several nodes. Wavelet technology is proposed in this paper to recover irregularly sampled data, for denoising, compression and subsequent extraction of evolutionary trends in the signal in various time windows. Simple algorithms are outlined for each stage in a modular load signal analysis scheme.