MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641124760 A) filed by C Veeksha; Aditi Raj; Dr. Roopsahree H R; Dr. Asha M; Dr. Raghavendra Y M; Dr. Manjula G; Dr. Bharathi R; Dr. Rajath A N; Dr. Rajashekar M B; Sahana D M; Reshma R; and Likitha B on October 19, 2026, for Multi-Agent Reinforcement Learning For Energy Optimization In Smart Grids.
Inventors include C Veeksha; Aditi Raj; Dr. Roopsahree H R; Dr. Asha M; Dr. Raghavendra Y M; Dr. Manjula G; Dr. Bharathi R; Dr. Rajath A N; Dr. Rajashekar M B; Sahana D M; Reshma R; and Likitha B.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: A computer-implemented smart-grid energy optimization system coordinates solar generation, wind generation, battery storage, utility-grid exchange and flexible electricity demand through five cooperative reinforcement-learning agents. Fifteen-minute operational data is synchronized and standardized to create agent observations that include renewable resource availability, demand, market price, battery charge state, and time characteristics. Every agent chooses an action tailored to a specific resource from a distinct set of actions. Dynamic feasibility masks restrict actions that are not in line with battery operational limits or the availability of exportable energy. A centralized energy-flow system transforms collaborative activities into renewable use, battery charging and discharging, grid imports and exports, curtailment, and unmet demand. A communal economic benefit takes into account operating expenses, export income, and delivery fines. Distinct actor policies are developed through multi-agent proximal policy optimization featuring a centralized critic and then applied for decentralized inference. The same logic of feasibility and energy flow is upheld throughout training and application, resulting in interpretable simulated energy dispatch choices. FIG.1
Disclaimer: Curated by HT Syndication.
