Battery Health Prediction using Neural Networks

Overview

This project demonstrates how to use a feedforward neural network (ANN) to predict battery health scores based on several measurable parameters such as temperature, charge cycles, and efficiency. The goal is to model a relationship between input parameters and the resulting battery health score using TensorFlow and Keras.

Libraries Used in Python

1.NumPy and Pandas – for numerical data handling
2.Matplotlib – for plotting graphs
3.Scikit-learn – for preprocessing and evaluation
4.TensorFlow / Keras – for building and training the neural network

How It Works

Dataset Preparation
A small dataset is manually defined for training and testing.
Each row represents a battery sample with input parameters:
Temperature / Age
Charge cycles
Internal resistance
Efficiency
Target: Health Score
Feature Scaling
Input data is normalized using MinMaxScaler to improve model performance.

Model Architecture

4 input features
2 hidden layers:
16 neurons (ReLU activation)
8 neurons (ReLU activation)
1 output neuron (linear activation)

Training

Optimizer: adam
Loss function: Mean Squared Error
100 epochs of training

Evaluation

Model is evaluated using R² Score.
Visualization
Training loss curve
True vs Predicted battery health comparison