Optimizing Electric Vehicle Consumption Models Using Genetic Algorithms: Real-World Case Study

Electric vehicles (EVs) can potentially reduce emissions and dependence on fossil fuels. However, accurately predicting EV energy consumption remains a major challenge due to driver behavior, seasonal weather variations, and auxiliary loads. The scope is to take a validated EV consumption model and measure the improvement from known parameters to the optimal one. This study proposes a refined EV energy consumption framework where physics-based and data-driven models are re-parameterized using a genetic algorithm, obtaining significant improvements in predictive performance using an extensive real-world dataset. By comparing the proposed optimization against naive parameter settings in multiple driving scenarios, we demonstrate measurable reductions in MAE and RMSE. The proposed genetic algorithm is robust in exploring broad parameter spaces and optimized for this use case. Our findings underscore that evolutionary optimization strategies can quickly obtain near-optimal solutions faster than Nelder-Mead in this context.