NeuroCellX: An AI-Driven System for Health Monitoring and Lifespan Prediction of Electric Vehicle Batteries

This article introduces the open-source project NeuroCellX, an electric vehicle Battery Management System (BMS) that integrates machine learning and visualization technologies. It addresses the problem that traditional BMS struggles to accurately predict the State of Health (SOH), State of Charge (SOC), and Remaining Useful Life (RUL) of batteries. Through an interactive dashboard, it provides real-time monitoring and prediction to support user decision-making and industry applications.

With the explosive growth of the global electric vehicle market, the Battery Management System (BMS) has become a core technology that determines user experience and vehicle residual value. Battery aging is affected by multiple factors such as temperature and charge-discharge cycles. Traditional BMS relies on simple voltage-current models, making it difficult to accurately predict SOH and RUL, leading to increased user range anxiety, difficulty in used car valuation, ambiguous battery replacement timing, and potential safety hazards.

NeuroCellX is an AI-driven battery health monitoring web application open-sourced by developer SathsaraniBS. Its core functions include:

1. SOC Prediction: Compensates for temperature and aging effects to provide more accurate power readings;
2. SOH Assessment: Analyzes features such as charge-discharge curves and internal resistance changes to determine battery health levels;
3. RUL Prediction: Estimates remaining charge-discharge cycles or driving mileage based on historical data and degradation trends. The system adopts a front-end and back-end separation architecture, with the back-end using Python (FastAPI/Uvicorn) and the front-end based on the npm ecosystem.

The back-end uses FastAPI + Uvicorn (asynchronous processing, automatic API documentation generation) and isolates dependencies through a virtual environment; the front-end is based on the npm ecosystem, supporting interactive visualization and responsive design. The machine learning models are presumed to use time-series models (LSTM/GRU), ensemble learning (Random Forest/XGBoost), or Physics-Informed Neural Networks (PINN) to adapt to the characteristics of battery data.

The system has a built-in RBAC permission system, supporting four roles: administrator, engineer, analyst, and ordinary user, with differentiated permissions for each role. The interactive visualization dashboard provides real-time monitoring (alerts for key parameters), historical trends (SOH/SOC change curves), prediction displays (RUL confidence intervals), and comparative analysis (comparison of multiple batteries/vehicles) to support decision-making.

Application scenarios include:

• Passenger vehicles: Integrated into in-vehicle systems or apps to provide range estimation, charging recommendations, and used car health reports;
• Commercial fleets: Optimize scheduling, preventive maintenance, and operational cost prediction;
• Battery echelon utilization: Determine the feasibility of secondary battery use and plan full-life cycle management.

Current limitations: Dependence on high-quality training data, insufficient data for edge scenarios (extreme temperatures, fast charging abuse), and real-time performance to be optimized. Improvement directions include federated learning (joint training under privacy protection), digital twins (fine simulation of battery status), and edge AI deployment (lightweight model offline inference).

NeuroCellX transforms complex electrochemical problems into data science problems, providing a reference for electric vehicle battery management. As EV penetration increases, such systems will evolve from 'nice-to-have' to a necessity. Its open-source release provides a valuable implementation case for technical exploration in the field, worthy of learning and reference by relevant engineers.