Intelligent Forecasting of Power Consumption of Smart Grids with Machine Learning and Full-Stack Implementation on Machine Learning
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Understanding the amount of electricity that people will consume becomes extremely important in operating a smart grid, scheduling the utilization of renewable energy, as well as ensuring that the entire process remains stable. In this project, we employed machine-learning based on the forecasting of electricity demand based on a 5 year history of hourly smart-grid data. We have compared a variety of ensemble models, such as Random Forest, Extra Trees, Gradient Boosting, LightGBM, and XGBoost, by using two error indicators, including Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The boosting models were most accurate; XGBoost presented the lowest MAE of 2236.70, and Gradient Boosting presented the lowest RMSE of 3192.55. Attempts to optimize XGBoost using Optuna were not that effective, so the default settings were good. A full-stack system consisting of Fast API (backend) and Streamlit (front end) was used to test the winning model. The system is correct that is, MAE is not much more than 2.7 percent of the average demand, and it demonstrates that machine-learning may be used to forecast smart-grid use in real-time
Fields of Study & Subjects
Energy Load and Power Forecasting
Smart Grid Energy Management
Smart Grid and Power Systems
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 30 March 2026
- Language
- EN