Advanced Statistical Learning on Short Term Load Process Forecasting

TL;DR

This paper compares advanced machine learning models, specifically LSTM and GRU, for short-term electricity load forecasting, demonstrating their superior accuracy over other models in a chemical production setting.

Contribution

It introduces the application of LSTM and GRU models to short-term load forecasting and shows their improved performance over traditional models.

Findings

LSTM and GRU outperform other models in forecasting accuracy.

Forecasting up to 2 days ahead at 15-minute intervals is feasible.

Models are validated using the Diebold-Mariano test.

Abstract

Short Term Load Forecast (STLF) is necessary for effective scheduling, operation optimization trading, and decision-making for electricity consumers. Modern and efficient machine learning methods are recalled nowadays to manage complicated structural big datasets, which are characterized by having a nonlinear temporal dependence structure. We propose different statistical nonlinear models to manage these challenges of hard type datasets and forecast 15-min frequency electricity load up to 2-days ahead. We show that the Long-short Term Memory (LSTM) and the Gated Recurrent Unit (GRU) models applied to the production line of a chemical production facility outperform several other predictive models in terms of out-of-sample forecasting accuracy by the Diebold-Mariano (DM) test with several metrics. The predictive information is fundamental for the risk and production management of electricity consumers.