A framework application for improving energy demand forecasting using digital twining

Abstract

Energy management became an essential mindset toward sustainability. Analyzing energy demand rates facilitates the adoption of efficient energy management solutions. Moreover, forecasting future energy demand is needed to support decision makers in long-term strategic planning. Hence, energy consumption prediction using Artificial Intelligence (AI) becomes an important research field recently. However, forecasting models need to be trained on rich data to provide precise predictions. It sometimes can be difficult as irregularities could have very low frequency in the data. Therefore, this research paper proposes an innovative data construction framework to enhance the performance of AI-based energy demand forecasting solutions using Digital Twin Technology (DT). Nowadays, DT becomes an emerging technology in the industrial and energy sectors. In this research study, DT technology is utilized in constructing an extensive dataset that covers various scenarios to train AI models. This approach reduces the overestimation and underestimation of the prediction model and ensures the model generalization. In addition, the research study considers the impact of different external factors such as climatic factors, socioeconomic factors, etc. As a part of the framework, the performance of the trained models will be evaluated and validated using cross-validation and other evaluation metrics.