MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641027318 A) filed by Keelagaram Gunaprasad; Dr. S Parameswari; Dr. Manam Ravindra; B. Karthikprabu; Stephy Angelin J; S Manikandan; Mahendiran C R; and Dr. R. Sampathkumar, Puttur, Andra Pradesh, on March 9, for 'deep spectral learning fourier transforms for electric vehicle battery state estimation and degradation prediction.'
Inventor(s) include Keelagaram Gunaprasad; Dr. S Parameswari; Dr. Manam Ravindra; B. Karthikprabu; Stephy Angelin J; S Manikandan; Mahendiran C R; and Dr. R. Sampathkumar.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "Accurate estimation of battery states and early prediction of degradation are critical challenges for the reliable operation, safety, and longevity of electric vehicles (EVs). Conventional data-driven battery management approaches primarily rely on time-domain features extracted from voltage, current, and temperature signals, which often fail to capture hidden frequency-domain characteristics associated with electrochemical aging and dynamic operating conditions. This work presents a Deep Spectral Learning framework integrating Fourier Transform-based feature extraction with deep neural architectures for robust battery state estimation and degradation prediction. The proposed method transforms raw battery signals into the spectral domain using Fast Fourier Transform (FFT), enabling the model to identify frequency-specific signatures linked to state of charge (SoC), state of health (SoH), and aging progression. A deep learning architecture comprising spectral convolution layers and temporal attention mechanisms learns both stationary and non-stationary battery behaviors under diverse drive cycles. By fusing time-frequency representations, the model achieves enhanced generalization across varying temperatures, load profiles, and battery chemistries. Experimental validation demonstrates that the proposed spectral-deep framework significantly improves estimation accuracy and early degradation detection compared to conventional LSTM and purely time-domain approaches. The method offers a scalable, data- efficient, and real-time-compatible solution for next-generation battery management systems (BMS). This research establishes deep spectral learning as a promising paradigm for intelligent EV energy storage diagnostics and predictive maintenance."
Disclaimer: Curated by HT Syndication.
