International Research Journal of Education and Technology

Energy intake prediction is crucial for energy supply companies, allowing thermo adjust pricing and anticipate high-demand periods based on consumer behaviour. Forecasting predicts future values of time series data. Load forecasting estimates energy consumption using customer behaviour data, with short-term forecasts helping to prevent overloads and improve network reliability. The system consists of two models: a network for load forecasting, designed to capture temporal patterns in energy consumption, and an XGBoost model for price forecasting, which is particularly effective with tabular data. A meta-learner combines the outputs of these two models to improve accuracy. Data is pre-processed and organized by key features, such as seasonality. Deployments can take place in real-time using Docker and SQLite3 as lightweight databases. Additionally, a REST API allows utilities to integrate with utility software, while a Grafana dashboard visualizes both historical and real-time forecasts. This system enables energy providers to optimize resource allocation, anticipate peak demand, and manage pricing strategies effectively. Ultimately, it enhances grid management and supports the development of a more resilient energy infrastructure.