Computationally efficient algorithm for optimal battery preconditioning and charging of electric vehicles

Paper in proceeding, 2024

This study addresses the computational challenges in optimizing charge planning strategies for electric vehicles (EVs) on long journeys with low ambient temperature. The scale of the problem, its non-linearity and its mixed-integer nature make the problem intractable in real-time applications. This paper introduces a computationally efficient algorithm whose goal is to provide a good trade-off between computation time and accuracy. We achieved this by designing initial guesses for the optimal solution, facilitating the solver’s task by starting the optimization process near the desired outcome. We also addressed the mixed-integer nature of the original problem by relaxing its binary variables, allowing it to be solved through gradient-based algorithms. By employing initial guesses and relaxing the boolean variables, the average execution time, compared to running the mixed-integer problem without initial guesses, was reduced by about 91.07%, at the cost of an average increase in energy consumption of only about 0.01%.