MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641111007 A) filed by Pragati Engineering College on September 16, 2026, for A Hybrid Machine Learning And Temporal Deep Learning Framework For Short-Term Energy Demand Forecasting.
Inventors include Dr. Kola Satyanarayana; Mr. M Sunil Raj; Mr. A. Phani Bhaskar; and Mr. D Prakasa Rao.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention provides a hybrid forecasting framework for short-term prediction of electrical energy demand that combines classical machine learning with temporal deep learning to address the shortcomings of relying solely on one type of model. The framework consists of a data acquisition and preprocessing module for generating engineered tabular features (lagged demand, rolling statistics, cyclical time encodings, calendar and weather variables) and normalized sequential load data from historical demand data. Tabular features are passed into a machine learning prediction branch, which consists of an ensemble of gradient-boosted decision trees that produces a first demand prediction as well as a feature importance profile. At the same time, a temporal deep learning prediction branch, which uses a recurrent neural network or temporal convolutional network, is used to process the sequential load data, and a second prediction of the load demand is obtained, which can better reflect the short-term and long-term load demand. The adaptive fusion module uses weights that are calculated dynamically from the respective rolling forecasting error of each branch to provide a combination of both forecasts, thus giving more weight to the better-performing branch. All modules are retrained periodically with more recent data to remain flexible with changing consumption patterns. The invention is applicable for deployment in a utility, microgrid and/or energy management and provides improved accuracy, anomaly robustness and interpretability.
Disclaimer: Curated by HT Syndication.
