Smart Energy Monitor: A Java-Based Application with Artificial Intelligence for Household Electricity Consumption Prediction

Mounia Stitou, Mohamed Ben Jouad

EnerWise is a household energy management prototype that combines appliance monitoring, cost estimation, recommendations, and one-hour-ahead energy consumption forecasting. A customized six-month synthetic dataset was developed for six household appliances. Following an initial Random Forest experiment (MAE = 0.1927 kWh, MSE = 0.1262, and R² = 0.3825), the temporal consistency of the dataset and the input features were improved. The resulting model achieved an R² of 0.5397. Hyperparameter tuning using RandomizedSearchCV and TimeSeriesSplit further improved the performance, achieving an MAE of 0.1502kWh, an MSE of 0.0797, and an R² of 0.5828. RNN, LSTM, and SSM models were also evaluated. The selected model was integrated into a Vue.js–Spring Boot–H2–FastAPI architecture.