Abstract
Although electric vehicles are considered a viable solution to reduce greenhouse gas emissions, their uncoordinated charging could have adverse effects on power system operation. Nevertheless, the task of optimal electric vehicle charging scales unfavorably with the fleet size and the number of control periods, especially when distribution grid limitations are enforced. To this end, vehicle charging is first tackled using the recently revived Frank-Wolfe method. The novel decentralized charging protocol has minimal computational requirements from vehicle controllers, enjoys provable acceleration over existing alternatives, enhances the security of the pricing mechanism against data attacks, and protects user privacy. To comply with voltage limits, a network-constrained EV charging problem is subsequently formulated. Leveraging a linearized model for unbalanced distribution grids, the goal is to minimize the power supply cost while respecting critical voltage regulation and substation capacity limitations. Optimizing variables across grid nodes is accomplished by exchanging information only between neighboring buses via the alternating direction method of multipliers. Numerical tests corroborate the optimality and efficiency of the novel schemes.
Introduction
The rapid adoption of electric vehicles (EVs) presents a significant challenge to existing power grids. Uncoordinated charging—where vehicles begin charging simultaneously upon arrival—can lead to transformer overloads and severe voltage drops. While optimal charging strategies aim to "fill the valleys" of low demand, traditional centralized optimization becomes computationally intractable as fleet sizes grow. Furthermore, centralized approaches raise concerns regarding user privacy and the security of pricing signals against malicious data attacks.
Methodology
This research introduces two decentralized protocols to address these challenges across different operational scenarios.
Scalable Charging via Frank-Wolfe
For scenarios focused on basic valley-filling without explicit grid constraints, the authors employ the Frank-Wolfe method (conditional gradient algorithm). Unlike standard Projected Gradient Descent (PGD) methods that require expensive projection steps for each vehicle, the Frank-Wolfe approach simplifies the problem into a sorting task.
As illustrated in Figure 1, the control center broadcasts a ranking of time slots based on cost. EV controllers then greedily select the most favorable slots according to their individual needs. This iterative process ensures privacy, as the central controller only observes the aggregated load rather than individual charging profiles.
Network-Constrained Charging via ADMM
In cases where grid physical limits (voltage levels and transformer capacities) are critical, a more robust protocol is required. The authors utilize a linearized model for unbalanced distribution grids and apply the Alternating Direction Method of Multipliers (ADMM). This technique decomposes the global optimization problem, allowing each grid node (bus) to reach a consensus by communicating only with its immediate neighbors.
Experimental Evaluation
The protocols were validated using the IEEE 123-bus feeder benchmark, combined with real-world residential load data from Southern California Edison (2014) and travel statistics from the National Household Travel Survey (NHTS) to simulate 59 EVs.
Key Results
The Frank-Wolfe-based algorithm (Algorithm 1) demonstrated a remarkable 100x speedup, converging in just 0.78 seconds compared to 734.74 seconds for the PGD method. Crucially, the update time per vehicle remained nearly constant regardless of the number of time slots, a significant improvement over traditional linear-scaling methods.
As shown in Figure 3, the proposed algorithms successfully achieve optimal valley-filling, maintaining grid stability and ensuring all voltages remain within safe operating limits even under high EV penetration.
Conclusion and Future Work
The proposed decentralized protocols offer a scalable, secure, and privacy-preserving solution for EV integration. By reducing complex optimizations to simple sorting and local communication, these methods pave the way for real-time grid management. Future research will focus on extending these models to non-linear AC power flow equations and implementing asynchronous updates to better reflect real-world network dynamics.
Summary updated