Transformer-Based Model for Electrical Load Forecasting

Amongst energy-related CO2 emissions, electricity is the largest single contributor, and with proliferation of electric vehicles other developments, energy use expected to increase. Load forecasting essential for combating these issues as it balances demand production contributes management. Current state-of-the-art solutions such recurrent neural networks (RNNs) sequence-to-sequence algorithms (Seq2Seq) are highly accurate, but most studies examine them on a data stream. On hand, in natural language processing (NLP), transformer architecture has become dominant technique, outperforming RNN Seq2Seq while also allowing parallelization. Consequently, this paper proposes transformer-based load by modifying NLP workflow, adding N-space transformation, designing novel technique handling contextual features. Moreover, contrast studies, we evaluate proposed solution different streams under various horizons input window lengths order ensure result reproducibility. Results show that approach successfully handles time series outperforms models.