Short and Very Short Term Firm-level Load Forecasting for Warehouses: A Comparison of Machine Learning and Deep Learning Models

• DOI: 10.20944/preprints202201.0107.v1
• Publication Date: 2022-01-11T09:49:25Z
• Authors: Andrea Maria N. C...; Pedro Rafael X. d...; Patricia Takako Endo; Pierangelo Rosati; Theo Lynn
• Keywords: Gradient boosting

Commercial buildings are a significant consumer of energy worldwide. Logistics facilities, and specifically warehouses, common building type yet under-researched in the demand-side forecasting literature. Warehouses have an idiosyncratic profile when compared to other commercial industrial with reliance on small number systems. As such, warehouse owners operators increasingly entering performance contracts service companies (ESCOs) minimise environmental impact, reduce costs, improve competitiveness. ESCOs require accurate forecasts their consumption so that precautionary mitigation measures can be taken. This paper explores three machine learning models (Support Vector Regression (SVR), Random Forest, Extreme Gradient Boosting (XGBoost)), deep (Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU)), classical time series model, Autoregressive Integrated Moving Average (ARIMA) for predicting daily consumption. The dataset comprises 8,040 records generated over 11-month period from January November 2020 non-refrigerated logistics facility located Ireland. grid search method was used identify best configurations each model. proposed XGBoost outperform both very short load (VSTLF) term (STLF); ARIMA model performed worst.