Enhancing Load Forecasting using Deep Learning Techniques: A Comparative Study of Time Series Models.

Abstract

Accurate load prediction is of utmost importance in order to efficiently allocate resources and plan systems in various domains. This research presents a comparative analysis of statistical methods and deep learning approaches for tasks related to load forecasting. The study focuses on assessing the performance of both traditional statistical methods and deep learning techniques in scenarios involving both univariate and multivariate forecasting. This research aims to explore and compare the potential of transformer-based models, LSTM, AR/VAR in load forecasting. By integrating these approaches and conducting real-world validation, the research aims to provide valuable insights and recommendations for optimizing load forecasting strategies and contributing to the advancement of energy management practices. The time series models have been assessed using evaluation metrics.