Input data importance analysis for various machine learning techniques in task of short-term electricity demand forecasting in industrial plant

The sensible goal of reducing the industrial carbon footprint necessitates comprehensive solutions. These solutions should consider aspects such as balancing energy demand with renewable generation, improving energy efficiency, and shaping demand. Research on these solutions is one of the goals of the DIEGO project (details in acknowledgments), with selected excerpts presented in this paper. The publication analyses the importance of input variables for electricity demand forecasts, considering lagged and aggregated variables. Additionally, it presents forecast results using various predictive models and concludes by describing differences between variables that could impact the construction of predictive models. The presented work is part of greater undertaking, the results of which will be published subsequently.