Skip to content

Battery Degradation Theory

This document provides background on battery degradation modeling.

Overview

Battery degradation refers to the gradual loss of capacity and performance over time and use. Understanding degradation mechanisms is crucial for:

  • Predicting remaining useful life
  • Optimizing battery usage
  • Designing better batteries

Degradation Mechanisms

Loss of Lithium Inventory (LLI)

Description: Loss of active lithium ions that can participate in charge/discharge cycles.

Causes:

  • SEI (Solid Electrolyte Interphase) growth
  • Lithium plating
  • Side reactions

Indicators:

  • Capacity fade
  • Voltage curve shifts (in ICA analysis)

Loss of Active Material (LAM)

Description: Loss of active electrode material (anode or cathode).

Causes:

  • Particle cracking
  • Electrode delamination
  • Material dissolution

Indicators:

  • Capacity fade
  • Peak height changes in ICA

Impedance Rise

Description: Increase in internal resistance.

Causes:

  • SEI growth
  • Contact loss
  • Electrolyte degradation

Indicators:

  • Voltage drop under load
  • Peak width changes in ICA

State of Health (SOH)

SOH is a key metric for battery health, defined as the ratio of current capacity to the nominal initial capacity:

\[ \text{SOH} = \frac{Q_{\text{current}}}{Q_{\text{nominal}}} \]
  • SOH = 1.0: New battery (100% capacity)
  • SOH = 0.8: 80% capacity remaining
  • SOH < 0.8: Often considered end-of-life

Factors Affecting Degradation

Temperature

  • High temperature: Accelerates degradation
  • Low temperature: Can cause lithium plating
  • Optimal: Moderate temperatures (20-30°C)

Charge/Discharge Rate (C-rate)

  • High C-rate: Increases degradation
  • Low C-rate: Slower degradation
  • Optimal: Low to moderate C-rates

Depth of Discharge (DOD)

  • Deep discharge: Increases degradation
  • Shallow discharge: Reduces degradation
  • Optimal: Moderate DOD (20-80%)

Cycle Count

  • More cycles = more degradation
  • Degradation rate may change over time

Modeling Approaches

Empirical Models

  • Arrhenius equation: Temperature dependence
  • Power law: Cycle count dependence
  • Linear models: Simple capacity fade

Physics-Based Models

  • P2D model: Pseudo-2D electrochemical model
  • Equivalent circuit models: Electrical circuit analogs
  • Degradation mechanism models: Explicit mechanism modeling

Data-Driven Models

  • Machine learning: Learn from data
  • Neural networks: Flexible function approximation
  • Gradient boosting: Tree-based models

Feature Engineering

Summary Statistics

  • Cumulative throughput
  • Resistance measurements
  • Temperature statistics

Incremental Capacity Analysis (ICA)

  • Peak positions (LLI indicator)
  • Peak heights (LAM indicator)
  • Peak widths (impedance indicator)

Time-Series Features

  • Degradation trajectories
  • Temporal patterns
  • Sequence modeling

Evaluation Metrics

Regression Metrics

  • RMSE: Root mean squared error
  • MAE: Mean absolute error
  • MAPE: Mean absolute percentage error
  • R²: Coefficient of determination

Domain-Specific Metrics

  • Capacity retention: Percentage of initial capacity
  • Cycle life: Number of cycles to end-of-life
  • Energy efficiency: Energy in / energy out

Research Challenges

  1. Generalization: Models trained on one condition may not generalize
  2. Interpretability: Understanding model predictions
  3. Uncertainty: Quantifying prediction uncertainty
  4. Data scarcity: Limited degradation data
  5. Multi-mechanism: Multiple degradation mechanisms interact

References

  • Oxford Battery Intelligence Lab: LG M50T dataset
  • Battery degradation literature
  • Electrochemistry textbooks

Next Steps