Boosting battery health prediction in electric vehicles via extended quantum optimized model with gaussian process regression

Accurate forecasting of battery state of health is essential for ensuring the reliability of electric vehicles. Conventional Quantum Particle Swarm Optimized Neural Network struggles with capturing non-linear temporal dependencies in battery degradation data, leading to suboptimal prediction accuracy. To address this issue, an innovative model that fuses Extended Quantum Particle Swarm Optimized Neural Network (EQPSONN) and Gaussian Process Regression (GPR) is proposed to boost forecasting accuracy. The model's novelty is threefold. First, two novel statistical health indicators—kurtosis and skewness of battery charging current—are introduced, demonstrating strong correlation with state of health (SOH) and effectively capturing complex degradation patterns. Second, the proposed EQPSONN employs a unique solution generation mechanism that optimizes the exploration–exploitation balance, transforming these indicators into a more informative single extracted feature. Third, this optimized signature is input into a Gaussian Process Regression (GPR) model, which leverages kernel-based modeling to capture intricate nonlinear degradation dynamics, enhancing forecasting accuracy. Simulations from four public NASA Randomized Battery Usage datasets shows that Extended Quantum Particle Swarm Optimized Neural Network with Gaussian Process Regression and Radial Basis Function kernel achieves the highest prediction accuracy of 97.5 % when compared other machine learning models.