Energy load forecasting: one-step ahead hybrid model utilizing ensembling

Abstract In the light of adverse effects climate change, data analysis and Machine Learning (ML) techniques can provide accurate forecasts, which enable efficient scheduling operation energy usage. Especially in built environment, Energy Load Forecasting (ELF) enables Distribution System Operators or Aggregators to accurately predict demand generation trade-offs. This paper focuses on developing comparing predictive algorithms based historical from a near Zero Building. involves load, as well temperature data, are used develop evaluate various base ML methodologies, including Artificial Neural Networks Decision-trees, their combination. Each algorithm is fine-tuned tested, accounting for unique characteristics, such presence photovoltaics, order produce robust approach One-Step-Ahead ELF. To this end, novel hybrid model utilizing ensemble methods was developed. It combines multiple outputs utilized train meta-model voting regressor. acts normalizer any new input. An experimental comparison against unseen other approaches, showed promising forecasting results (mean absolute percentage error = 5.39%), particularly compared algorithms.