dQ/dV Peak Shift Detection on Oxford Dataset
Processing 7,000+ charge-discharge cycles to locate the knee point, the moment linear aging gives way to accelerated capacity fade, using only voltage and current logs.
7,000+
Cycles Analyzed
Full lifecycle analysis from raw MATLAB logs.
3 Modes
Degradation ID
LLI, LAM-PE, LAM-NE classified automatically.
Ideal Deployment Profile
Reference study summary
Standard BMS reporting still suggests a healthy cell while lithium plating is already developing beneath the surface, invisible to coulomb counting.
dQ/dV analysis reveals a progressive peak shift across cycles and marks the point where degradation stops looking routine and starts accelerating.
Operators receive a cycle-specific degradation fingerprint they can act on before the safety and warranty picture worsens.
Methods used in this study
Savitzky-Golay noise filtering
Per-cycle dQ/dV peak extraction
Baseline comparison (Cycle 10 reference)
Degradation mode classification (LLI/LAM)
What made this dataset hard
to review well
Lithium plating invisible to standard BMS coulomb counting
No lab hardware available, so the review had to work from raw voltage and current CSV files
Need to separate degradation mode (LLI vs LAM) automatically at fleet scale
How the review
was carried out
Review step
Gaussian noise removal + cubic spline interpolation for low-res data
Review step
Per-cycle dQ/dV electrochemical signature generation
Review step
Peak shift direction mapping to LLI, LAM-PE, or LAM-NE degradation modes
What this validation
confirmed
This public reference study confirmed the following signals and decision points in the dataset.
Turn this reference study into a scoped review
If the pattern looks familiar, start with a fixed-scope Verify review. That establishes the asset baseline before anyone decides whether Horizon belongs in the operating stack.