Evaluating isothermal degradation error in lithium-ion batteries under WLTP drive cycles

Abstract

Physics-based battery management systems frequently assume isothermal cell operation, eliminating the thermal sub-model to reduce computational cost. The error introduced by this assumption in state-of-health prediction depends on the activation energy of the solid electrolyte interphase kinetic rate constant. In widely used parameterisations this energy is set to zero, decoupling temperature from degradation kinetics and concealing the true thermal sensitivity. A calibrated multi-mechanism Single Particle Model with Electrolyte is developed for nickel manganese cobalt oxide and lithium iron phosphate cells. A two-stage sequential procedure calibrates the activation energy from temperature-dependent fade rate ratios, then jointly optimises the solid electrolyte interphase rate constant and loss of active material coefficient against a longitudinal state-of-health trajectory, achieving a root mean square error of 1.45 pp. — 6.3-fold improvement over the uncalibrated default. A parametric sweep of 24 conditions spanning both chemistries, ambient temperatures of 10, 25, and 40 °C, and charge rates of 1C and 2C shows that the isothermal assumption introduces state-of-health errors of 1.4–8.2 pp. (for nickel manganese cobalt oxide) over 500 Worldwide Harmonised Light Vehicle Test Procedure cycles with the calibrated activation energy of 78.5 kJ mol−1, compared with a maximum of 0.20 pp. at zero ac-tivation energy. For nickel manganese cobalt oxide cells, solid electrolyte interphase growth accounts for >99% of capacity loss. For lithium iron phosphate cells, solid electrolyte interphase and loss of active material contribute approximately 62% and 38% respectively, due to the larger depth of discharge per cycle. A three-layer parameter uncertainty analysis one-at-a-time sensitivity, an activation energy sweep across 62 – 95 kJ mol−1, and an analytical Monte Carlo, confirms that the isothermal SOH error exceeds 0.5pp. across the full physically plausible parameter space and that activation energy is the dominant source of uncertainty.