The rising demand for electricity necessitates the development of advanced predictive models to enable effective load management and promote sustainable energy usage. In smart home applications, real-time monitoring and optimization of appliance-level energy consumption are crucial for enhancing energy efficiency. The proposed study analyzes residential electricity usage patterns and forecasts appliance-specific energy consumption using artificial intelligence (AI) techniques. A comprehensive evaluation involving eight machine learning algorithms was conducted to predict energy consumption across various time intervals such as hourly, daily, weekly, monthly, and quarterly. The objective is to identify the most suitable algorithm for each time frame to enhance energy optimization and scheduling. The analysis was implemented using Python on the Google Colab platform, and model performance was assessed using standard evaluation metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), R-squared, and accuracy. The results revealed that the performance of forecasting models varies depending on the time interval. For hourly forecasts, a time series model achieved an accuracy of 99.96%. Daily predictions using the GBM + Optuna model reached 95.63% accuracy. Weekly forecasts with the XGBoost model achieved 98.34% accuracy, while monthly and quarterly predictions using the NGBoost model reached 99.76% and 99.99% accuracy, respectively. These findings provide valuable insights for energy planners and homeowners to optimize electricity consumption. In conclusion, this research demonstrates the effectiveness of AI-based time series forecasting in managing residential energy usage and promoting efficient energy consumption practices.
Devanathan et al. (Fri,) studied this question.