Improving Microgrid Energy Demand Forecasting Using Convolutional Neural Networks

Microgrids are localized energy systems that operate independently or in conjunction with the main power grid, integrating various energy sources for a reliable and resilient power supply. Microgrid energy demand refers to the electrical power required to meet consumer needs, influenced by factors such as time, weather, and habits of consumers. Accurate forecasting of energy demand is crucial for efficient microgrid operation, resource management, and optimization. Machine learning offers advantages over traditional methods in forecasting, such as the ability to handle complex relationships, adaptability to changes, and scalability. This paper proposes a Convolutional Neural Network (CNN) based approach to forecast microgrid energy demand.
The study considers a dataset of multi-year power generation, consumption, and storage data in a microgrid and compares different approaches considering the metrics: Mean Squared Error, Mean Absolute Error, and R Squared Value. The results demonstrate that the CNN model performs well compared to other models. This study seeks to advance the application of machine learning in microgrid management.