Idea
A Gaussian Process-based detection method for real-time lithium plating monitoring in lithium-ion batteries benefiting battery manufacturers and EV operators.
Research Paper
Core Innovation
This paper introduces a Gaussian Process model to directly infer the derivative of charge-voltage curves with calibrated uncertainty, overcoming noise and bias issues in traditional methods. It enables robust, noise-aware lithium plating detection without ad hoc smoothing and supports scalable online implementation for embedded systems.
Market Size (TAM)
$20–50B TAM for lithium-ion battery diagnostics and management; $2–10B SAM from electric vehicle and consumer electronics battery manufacturers. Driven by increasing EV adoption and demand for battery safety and longevity.
Potential Customers & Pain Points
- Battery Manufacturers Needing Early Degradation Detection
- Electric Vehicle Operators Seeking Safety and Longevity
- Battery Management System Developers Requiring Accurate Real-Time Diagnostics
Business Model
Licensing the detection algorithm as an API or embedded software module to battery manufacturers and BMS developers; offering consulting and integration services.
Competitive Landscape
- Twaice
- Cadenza Innovation
- Battery Informatics
Implementation Challenges
- Integration with existing battery management systems
- Validation across diverse battery chemistries and form factors
- Real-time computational constraints in embedded environments
Validation Strategy
- Conduct extended field tests on commercial EV battery packs under varied conditions
- Collaborate with battery manufacturers for pilot integration in BMS hardware
- Benchmark detection accuracy against established electrochemical and safety tests
Research Paper Overview
Machine Learning Detection of Lithium Plating in Lithium-ion Cells: A Gaussian Process Approach
Summary
This paper proposes a Gaussian Process framework to detect lithium plating in lithium-ion cells by modeling the charge-voltage relationship as a stochastic process. It infers the derivative dQ/dV analytically with uncertainty quantification, avoiding noise amplification and bias from traditional finite differencing methods. Validated experimentally across various C-rates and temperatures, the method reliably identifies plating peaks and correlates with capacity fade, enabling real-time detection suitable for embedded battery management systems.