Household Load Forecasting Based on a Pre-Processing Non-Intrusive Load Monitoring Techniques

Abstract

The new vision for moving the power system to a smart grid enables a variety of smart applications at different power system infrastructure's levels. Households represent a massive section of the grid infrastructure which could not be considered as a smart grid without smart households integrated into it. Short Term Load Forecasting (STLF) is the essential tool needed in the management and control techniques required for households to be smart. STLF at this level of the grid is very challenging due to the high percentage of uncertainty in the load demand, influenced by customer behavior, which is too stochastic to predict. In this paper, a new approach for STLF of household load demand is employed based on artificial neural network (ANN) and a pre-processing stage of a Non-Intrusive Load Monitoring (NILM) techniques. The NILM techniques extract the individual load pattern from the available historical aggregated load demand. These new features increase the training data window for the ANN forecaster and achieve a significant enhancement for its prediction performance. By comparing the new approach with the state of the art techniques in household load forecasting, the proposed method outperforms feed-forward artificial neural network (FFANN) regarding RMSE. Two techniques of NILM were used to emphasize the correlation between the NILM disaggregation accuracy performance and the load forecasting enhancement performance.