Degradation-Aware Model Predictive Control for Battery Swapping Stations under Energy Arbitrage

TL;DR

This work tackles the profitability-degradation trade-off in battery swapping stations (BSS) by introducing BSS-MPC, a degradation-aware Model Predictive Control framework designed for real-time operation under energy market arbitrage. It combines a high-fidelity electrochemical aging model with a computationally efficient Kriging surrogate, formulating a finite-horizon mixed-integer optimal control problem that jointly optimizes swapping logistics and grid energy transactions. The approach is implemented as a receding-horizon MPC solved via JuMP/DICOPT, with a 24-hour horizon and Big-M linearization to manage discrete swap decisions. Experiments on a simulated community (200 batteries, 21 in station) show BSS-MPC outperforms rule-based and low-fidelity baselines in profit, capacity fade reduction, and strict SOC constraint satisfaction, while maintaining near real-time computational feasibility. The results demonstrate the practical viability of degradation-aware optimization for grid-integrated BSS and highlight potential for scalable deployment with faster surrogates and stochastic extensions.

Abstract

Battery swapping stations (BSS) offer a fast and scalable alternative to conventional electric vehicle (EV) charging, gaining growing policy support worldwide. However, existing BSS control strategies typically rely on heuristics or low-fidelity degradation models, limiting profitability and service level. This paper proposes BSS-MPC: a real-time, degradation-aware Model Predictive Control (MPC) framework for BSS operations to trade off economic incentives from energy market arbitrage and long-term battery degradation effects. BSS-MPC integrates a high-fidelity, physics informed battery aging model that accurately predicts the degradation level and the remaining capacity of battery packs. The resulting multiscale optimization-jointly considering energy arbitrage, swapping logistics, and battery health-is formulated as a mixed-integer optimal control problem and solved with tailored algorithms. Simulation results show that BSS-MPC outperforms rule-based and low-fidelity baselines, achieving lower energy cost, reduced capacity fade, and strict satisfaction of EV swapping demands.

Figures (8)

• Figure 1: Schematic of BSS management with energy market arbitrage.
• Figure 2: Histogram of Kriging fitting for $C^{\text{avg}}_p$, $C^{\text{avg}}_n$, $\delta_{\text{SEI}}$, and $c_f$
• Figure 3: Temporal heatmaps of swap demand and electricity price: swap demand is concentrated during daytime and minimal overnight. Electricity prices exhibit a diurnal pattern with late-afternoon peaks, occasional spikes, and a mid-year volatility patch.
• Figure 4: Within-horizon scheduling: optimized battery operation (MPC v.s. baselines) with grid price and swap demand
• Figure 5: Plan vs. Realization (rolling horizon): charging power, swap decisions, and SOC over a day.