An Ensemble Approach for Multi-Step Ahead Energy Forecasting of Household Communities

This paper addresses the estimation of household communities' overall energy usage and solar production, considering different prediction horizons. Forecasting electricity demand generation communities can help enrich information available to grid operators better plan their short-term supply. Moreover, households will increasingly need know more about patterns make wiser decisions on appliance energy-trading programs. The main issues address here are volatility load consumption induced by behaviour variability in output influenced cells specifications, several meteorological variables, contextual factors such as time calendar information. To these issues, we propose a predicting approach that first considers highly influential and, second, benefits from an ensemble learning method where one Gradient Boosted Regression Tree algorithm is combined with Sequence-to-Sequence LSTM networks. We conducted experiments public dataset provided Ausgrid Australian distributor collected over three years. proposed model's performance was compared those contributing learners conventional ensembles. obtained results have demonstrated potential predictor improve multi-step forecasting providing stable forecasts accurate estimations under day types conditions.