Voltage-Frequency Co-Regulation For EV-Fleet: A Voltage Augmented Physics-Informed Model Predictive Control for Smart-Logistics Grids

Electrified logistics and future transport increasingly depend on massive, stochastic, bidirectional energy flows between electric-vehicle (EV) fleets and smart grids. Prior physics-informed machine learning (PIML) integrated model predictive control (MPC) formulations demonstrated strong frequency regulation but provided limited guarantees on voltage behavior under logistics-driven reactive-power swings. This paper introduces a voltage-augmented PIML–MPC that jointly predicts active/reactive EV disturbances and coordinates distributed energy resources (DERs) for co-regulation of generator frequency and bus voltages. The PIML surrogate embeds swing and nodal-voltage dynamics and is trained with sparse data via a hybrid data-and-physics loss with curriculum on the physics weight. The ensuing nonlinear MPC (NMPC) minimizes horizon-wise state and input deviations under soft bounds using slack variables and employs a Lyapunov-consistent terminal cost for recursive feasibility and asymptotic stability. On the IEEE 39-bus system with EV clusters at depot-like buses, the proposed controller reduces RMS frequency/voltage deviations versus conventional MPC and a frequency-only PIML baseline, while preserving real-time feasibility. Sensitivity to scaled feet intensity confirms robustness and scalability. The results establish voltage-aware PIML–MPC as a deployable cornerstone for resilient, EV-integrated logistics grids.