MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202621079620 A) filed by Symbiosis International Deemed University on June 27, 2026, for Machine Learning Based Energy Demand Forecasting System For Smart Grid.
Inventors include Aditya Sharma; Akshita Mishra; Aaron Frank Sanjay; Aditya Shelar; and Dr. Archana Chaudhari.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: ABSTRACT MACHINE LEARNING BASED ENERGY DEMAND FORECASTING SYSTEM FOR SMART GRID The present invention discloses a machine learning based energy demand forecasting system (100) for smart grids comprising a smart meter data acquisition module (110) for collecting three-phase measurements, a data pre-processing module (120) for outlier capping and gap filling, a feature engineering module (130) for deriving thirty temporal features including lag features at eight time offsets and rolling statistics over four window sizes with cyclical calendar encodings, a multi-model training module (140) for parallel training of SARIMA (142), LSTM (144) and XGBoost (146) models on chronological data, a forecasting and evaluation module (150) for generating hourly predictions and computing per-month metrics including RMSE, MAE, R-squared and sMAPE, and a visualization module (160) for comparative graphical output. The system provides a unified comparative pipeline with explainable AI through SHAP analysis enabling transparent demand forecasting for residential smart grids. [
Disclaimer: Curated by HT Syndication.
