wang-fujin/PINN4SOH

What it solves

This project provides a method for the stable modeling and prognosis of lithium-ion battery degradation, specifically focusing on State of Health (SOH) estimation. It aims to overcome the volatility and instability often found in standard regression models for battery health prediction.

How it works

The project implements a Physics-Informed Neural Network (PINN), which integrates physical laws and battery degradation physics into the neural network architecture. This approach allows the model to maintain stability and accuracy when predicting battery degradation compared to traditional Multi-Layer Perceptrons (MLP) or Convolutional Neural Networks (CNN).

Who it’s for

Researchers and engineers working on battery management systems (BMS), energy storage, and the prognosis of lithium-ion battery health.

Highlights

• Uses a Physics-Informed Neural Network (PINN) to improve stability in SOH modeling.
• Includes a demo using the XJTU battery dataset.
• Provides comparison scripts to evaluate the model against standard MLP and CNN architectures.
• Supports multiple battery datasets including XJTU, TJU, HUST, and MIT.