Quantum Machine Learning for EV Mobile Charging Prediction

The widespread adoption of Electric Vehicles (EV) is hindered by range anxiety and the limited flexibility of Fixed Charging Points (FCP). There is a pressing need for intelligent, adaptive solutions that ensure timely EV charging while minimising infrastructure costs and grid stress. To address this challenge, we propose to integrate Mobile Charging as a Service (MCaaS) with Machine Learning (ML). We develop asimulation environment modelling 50 EVs operating within an urban area, where GEECharge Mobile Charging Units (MCU) are dynamically dispatched to charge EVs with low-battery charge events. We use a Variational Quantum Neural Network (VQNN) and a LSTM to predict EVs likely to have low-battery charge events. Our VQNN model achieves an accuracy of 88%, outperforming the LSTM baseline at 84%. The VQNN demonstrates a superior performance in detecting critical low-battery charge events, achieving a recall of 50% compared to 0% for the LSTM. MCaaS successfully prevents 81.8% of potential vehicle strandings, with 0 actual strandings recorded on highways during the simulation. The integration of VQNN enhances the proactive nature of MCaaS, thus transforming reactive charging into preventive maintenance. MCaaS has the potential to supply EVs with charge wherever they are and help eliminate the cost of constructing an FCP that requires extensive electrical infrastructure and grid connection, enhancing the green transition and promoting sustainability.