MUMBAI, India, Feb. 6 -- Intellectual Property India has published a patent application (202541092605 A) filed by N. Dhivya Devi; and Dr. Elankurisil. S. A, Chengalpattu, Tamil Nadu, on Sept. 26, 2025, for 'intelligent adaptive control system for coupled inductor dc-dc converters in electric hybrid vehicles using deep reinforcement learning and predictive maintenance.'
Inventor(s) include N. Dhivya Devi; and Dr. Elankurisil. S. A.
The application for the patent was published on Feb. 6, under issue no. 06/2026.
According to the abstract released by the Intellectual Property India: "The present invention discloses an intelligent adaptive control system for coupled inductor DC-DC converters specifically designed for electric hybrid vehicle applications, integrating advanced artificial intelligence methodologies to achieve superior power conversion performance and predictive maintenance capabilities. The system employs a deep reinforcement learning framework using Proximal Policy Optimization algorithm that processes an 18-dimensional state space including electrical parameters, thermal conditions, component health indices, and predicted load demands to generate optimal control actions for duty cycle and switching frequency modulation. The core innovation combines real-time adaptive control with comprehensive health monitoring through multi-modal sensor fusion. An LSTM-based load forecasting module predicts power demand patterns 0.3ms ahead with 8% mean absolute percentage error, enabling proactive control adjustments. A CNN-based thermal analysis module processes 224 224 thermal images to detect hotspots and thermal anomalies with 94% precision, while a Variational Autoencoder analyzes electrical telemetry for fault detection with 3% false positive rates. The physics-informed reward function simultaneously optimizes converter efficiency, voltage ripple minimization, thermal management, and component longevity through weighted multi-objective optimization. Barrier function-based constraint handling ensures safe operation within physical limits during both training and deployment phases. Experimental validation using a 5kW prototype demonstrates exceptional performance improvements: 12-18% efficiency enhancement achieving 96.8% peak efficiency, 60% faster dynamic response with 2ms settling time, 22-28 C component temperature reduction, and 45% component lifetime extension. The integrated health monitoring system provides 150-200 hours early fault warning with 87% RUL prediction accuracy, resulting in 68% unplanned downtime reduction and 40% maintenance cost savings. The system is implemented on a hybrid DSP-FPGA platform combining Texas Instruments TMS320F28379D for real-time control and Xilinx Zynq UltraScale+ for AI acceleration, making it suitable for automotive-grade applications requiring high reliability and performance. KEYWORDS: Deep Reinforcement Learning, Coupled Inductor DC-DC Converter, Electric Hybrid Vehicles, Proximal Policy Optimization, LSTM Load Forecasting, CNN Thermal Analysis, Variational Autoencoder, Predictive Maintenance, Health Index, Remaining Useful Life, Multi-modal Sensor Fusion, Physics-informed Control, Barrier Functions, Real-time Optimization, Power Electronics, Automotive Applications, Artificial Intelligence, Machine Learning, Thermal Management, Fault Detection."
Disclaimer: Curated by HT Syndication.
