Battery Degradation Model

Degradation models fall into two primary categories:

• Empirical Models: Fit mathematical curves (e.g., exponential or power-law) to observed capacity fade data without modeling internal physics. They are computationally cheap but may fail outside tested conditions.
• Physics-Based Models: Simulate electrochemical side reactions like Solid Electrolyte Interphase (SEI) growth and lithium plating. These are more generalizable but require detailed parameterization of cell chemistry.

A robust model must account for multiple interacting degradation accelerants:

• High State of Charge (SoC): Calendar aging accelerates exponentially above ~80% SoC.
• Temperature: Arrhenius kinetics dictate that degradation rates roughly double for every 10°C rise.
• C-Rate: High charge/discharge currents induce mechanical stress and lithium plating.
• Depth of Discharge (DoD): Larger DoD swings per cycle generally reduce total cycle life.

Degradation manifests in two distinct ways:

• Capacity Fade: Loss of available ampere-hours, reducing operational range. Primarily caused by loss of cyclable lithium and active material.
• Power Fade: Increase in internal resistance, limiting peak current delivery. Critical for high-demand maneuvers like lifting or acceleration. A model must track both to prevent task failure.

Modern fleets deploy Gaussian Process Regression or Long Short-Term Memory (LSTM) networks trained on telemetry streams. These models:

• Learn complex, non-linear interactions between stress factors without explicit electrochemical equations.
• Enable Remaining Useful Life (RUL) predictions with uncertainty quantification.
• Require careful handling of covariate shift when deployed on new cell chemistries.

The degradation model acts as a cost function input to the Charge Scheduling Algorithm. By assigning a virtual dollar cost to each percentage point of capacity fade, the orchestrator can:

• Trade off immediate throughput for long-term battery longevity.
• Dynamically cap maximum SoC during low-demand periods to slow calendar aging.
• Sequence tasks to avoid deep discharges immediately before high-power operations.

The primary output of a degradation model is the State of Health (SoH) metric. Estimation techniques include:

• Coulomb Counting: Integrating current over a full cycle to measure actual capacity.
• Differential Voltage Analysis (DVA): Identifying peaks in dV/dQ curves that shift with degradation.
• Electrochemical Impedance Spectroscopy (EIS): Measuring internal resistance growth, often used during scheduled maintenance windows.