A feasible operation mode selection via recurrent learning for electric vehicles using renewable energy sources

Electric Vehicles (EVs) can help smart cities to use renewable energy and reduce pollution. This research targets plug-in hybrid electric vehicles (PHEVs), combining battery and ICE power sources; the framework doesn’t directly apply to fully battery-electric vehicles (BEVs), which lack the fuel-based modes considered here. Their modes of operation can be dynamically reconfigured based on changing environmental conditions, population density, vehicle density, battery condition, and availability of renewable energy resources. This research proposes the FOMS (Feasibility-Oriented operating-Mode Selection) framework, which combines recurrent learning, energy-distance feasibility, and renewable-aware mode selection in one decision layer to achieve unified feasibility-driven operation—unlike conventional RL, MPC, or recurrent methods. Unlike conventional RL, MPC, or recurrent approaches that primarily learn policies, perform predictive optimization, or encode temporal dependencies, respectively, FOMS uses RcL to identify feasible switchover conditions and subsequently selects among five defined operating modes — Pure Electric, Eco, Power, Low-Emission, and Hybrid according to remaining energy, coverable distance, renewable-energy availability, and the current operating state. The RcL component is used to determine the viable switchover points in combination with the RcE and RcD components. Upon observing these, the FOMS framework provides an appropriate operating-mode recommendation for improving the energy utilization and maximizing the driving range of EVs. The proposed model is evaluated on 15 vehicles over a 2400 m route and compared against MSMOCO, 2SHOM, and Triple-MPC, improving mode-selection efficiency (the rate of choosing a feasible operating mode) by 10.47 % and reducing switchover failure by 8.172 % on average, with a decision latency of 50.46 ms at 600 m versus 122.64–178.3 ms for the baselines — showing its potential for accurate, low-latency EV operation in smart cities.