Integrating machine learning models into a real-time energy forecasting system

Abstract

Energy markets experience viability, fluctuations, and complexity because of their resources, which are renewable energy sources. This enables the challenges that stakeholders face to make informed and effective decisions and planning for the market that can affect consumer behavior and other industries as well. A critical problem in this case is having accurate and real-time forecasting to address the challenges. To solve this problem, this thesis develops a real-time energy forecasting system that contains advanced machine learning and deep learning models, Bidirectional Long Short-Term Memory (BiLSTM), and Autoencoder Convolutional LSTM (AE-CLSTM). The proposed solution considers scalability, maintainability, adaptability, and availability for handling large, volatile datasets and dynamic market conditions. Results illustrate the system’s effectiveness by showing the improvements in prediction and comparing them to the traditional models. These findings highlight the potential of the proposed system to enhance decision-making and planning in energy markets, ultimately contributing to their stability and sustainability.