1. Linear Regression

What it does: Assumes constant degradation per kilometer. Draws a straight line forcing the start at 100% (new battery).

When useful: Provides the worst-case scenario. Useful for conservative planning and warranty limits.

2. Square Root (√km)

What it does: Models intense degradation at the beginning that slows down with use. Reflects SEI layer growth.

Scientific basis: Fick's Law (ionic diffusion) predicts that SEI layer growth follows a √t relationship.

3. Power Law

What it does: Adjusts a variable exponent b. When b ≈ 0.5, behavior matches the √km model (diffusion-limited SEI).

Scientific basis: NREL (National Renewable Energy Laboratory) identifies b ≈ 0.5 as typical value for calendar aging in LFP.

What is R² (R-squared)?

The coefficient of determination R² measures what proportion of data variability the model explains. Ranges from 0% to 100%.

• R² > 80% Excellent fit, very reliable prediction
• R² 50-80% Moderate fit, useful prediction
• R² < 50% Poor fit, indicative prediction

The model with highest R² is marked as "Best model".

Confidence Score (0-100)

Indicates how reliable the predictions are based on three factors:

• Records (max 30 pts): More samples = greater statistical precision
• Km coverage (max 40 pts): Data at higher mileage = better extrapolation
• R² maximum (max 30 pts): Better fit of the dominant model