Probabilistic Forecasting of Household Electrical Load Using Artificial Neural Networks

The emergence of Demand Response tariffs creates incentives for residential consumers to optimise their electricity consumption. This optimisation requires forecasts of electrical load on a single-household level. However, these forecasts are subject to high errors when using state-of-the-art point-forecasting methods. Therefore, this paper presents probabilistic forecasts using density-estimating Artificial Neural Networks (ANN). Two different probabilistic ANNs models, namely, Mixture Density Networks (MDN) and Softmax Regression Networks (SRN) are implemented and compared on three different datasets over a broad range of hyper-parameter configurations: temporal dataset granularity, input configurations and ANN architecture. The evaluation shows that both ANN models generate reliable forecasts of the probability density over the future consumption, which significantly outperform an unconditional benchmarking model. Furthermore, the experiments demonstrate that a decreased dataset granularity and lagged input improve the forecasts, while using additional calendar inputs and increasing the length of lagged inputs had little effect.