Energeneous – A smart energy management system

Abstract

The thesis introduces Energeneous, a Smart Energy Management System (EMS) designed to optimise solar energy usage in Finnish households by integrating machine learning-based forecasting and automated decision-making. The system utilizes Random Forest and XGBoost algorithms to predict hourly electricity prices and solar energy production, drawing on real-time data from ENTSO-E Transparency Platform and OpenWeatherMap API. By combining these fore-casts with user-defined inputs such as battery capacity and consumption patterns, the EMS commends optimal actions weather to buy, sell, or store solar energy to maximise cost efficiency and energy independence. The model training process employs TimeSeriesSplit cross-validation to maintain chronological integrity, and performance is assessed by using MAE, RMSE, and R-Square metrics. Results showing solid alignment between the Predicted and actual values that confirms the model’s effectiveness in capturing complex temporal and weather-driven patterns. The system also includes a user-friendly interface and cloud-based deployment, sup-porting real-time operation and potential integration with smart home infrastructure. While challenges such as limited access to real-time household data remain, the EMS demonstrate practical viability in the Nordic context, offering an intelligent, data-driven approach to man-aging distributed solar resources and navigating Finland’s dynamic market.