🔋 Battery Temperature Forecasting Web App

A machine learning–driven web application for forecasting lithium-ion battery temperature behavior across charge/discharge cycles using historical degradation data.

This implementation aligns with a related research publication:
📄 Research Paper: https://doi.org/10.1109/i-PACT65952.2025.11308020

Technical Overview

This project implements an end-to-end pipeline that combines data preprocessing, model training, API-based inference, and client-side visualization.

• Dataset: NASA Ames Prognostics Center B0005 Li-ion battery degradation dataset
• Problem Formulation: Univariate time-series forecasting of battery temperature
• Model: Linear Regression trained on sequential discharge cycle data
• Inference: Forecasts temperature values for the next 50 cycles given an initial input
• Deployment: Model served via a Flask REST API and consumed by a lightweight web interface

System Architecture

1. Battery discharge data is preprocessed and normalized using MinMax scaling
2. A regression model is trained to learn temporal temperature trends
3. The trained model and scaler are serialized and loaded at runtime
4. A Flask API exposes a /predict endpoint for inference
5. Forecast results are returned as JSON and rendered using Chart.js

Motivation

Battery temperature is a key indicator of:

• Thermal degradation
• Capacity fade
• Operational safety

Accurate temperature forecasting supports early-stage analysis of battery health and thermal behavior under repeated usage cycles.

Technology Stack

• Backend: Python, Flask
• ML: Scikit-learn (Linear Regression), Pandas, NumPy
• Visualization: HTML, JavaScript, Chart.js
• Data Processing: MinMaxScaler

Scope & Limitations

• The model assumes simplified operating conditions
• Results are indicative and not intended for real-time production deployment
• Designed primarily for learning, experimentation, and system prototyping

Future extensions include multivariate forecasting and recurrent neural network–based models (LSTM/GRU).

References

• NASA Ames Prognostics Data Repository – B0005 Battery Aging Dataset
  https://www.nasa.gov/content/prognostics-center-of-excellence-data-set-repository

Note

• This implementation serves as a simplified demonstration of the concepts presented in the research paper.
• The project focuses on illustrating the core idea of battery temperature forecasting, while the paper discusses the complete system design and provides a detailed analysis of why LSTM-based models are more suitable for capturing long-term temporal dependencies in battery behavior.

⭐ Contributions and improvements are welcome.