Joint Price and Power MPC for Peak Power Reduction at Workplace EV Charging Stations

TL;DR

The paper tackles peak-demand charges for workplace EV charging by introducing a forecast-enabled model predictive control framework that jointly optimizes prices and charging power to shape load. It combines a discrete-choice pricing model with convex reformulations and a softplus penalty to anticipate and curb peaks, and it integrates time-series forecasts to inform control actions over a horizon. A Monte Carlo study on real-world session data demonstrates that the MPC approach reduces demand charges by about 15% and total costs by roughly 4.5%, outperforming a state-of-the-art benchmark and non-MPC variants. The work suggests practical gains for operators and highlights the importance of forecast quality and scalability to larger networks, with future directions including data-driven retraining and deployment across multiple sites.

Abstract

Demand charge often constitutes a significant portion of electricity costs for commercial electric vehicle (EV) charging station operators. This paper explores control methods to reduce peak power consumption at workplace EV charging stations in a joint price and power optimization framework. We optimize a menu of price options to incentivize users to select controllable charging service. Using this framework, we propose a model predictive control approach to reduce both demand charge and overall operator costs. Through a Monte Carlo simulation, we find that our algorithm outperforms a state-of-the-art benchmark optimization strategy and can significantly reduce station operator costs.

Figures (3)

• Figure 1: The softplus function adds a penalty when the peak of the optimized station power is approaching the running peak (as $x \xrightarrow{}0^{-}$), and converges to the original penalty as the station power exceeds the running peak (as $x \xrightarrow{}+\infty$).
• Figure 2: Example of control actions for benchmark (left) and MPC (right) on September 6th, 2023. For these two examples, the users all arrived at the same time, with the same requirements and all chose SCHEDULED. Each color represents a single user's power profile. The MPC algorithm shows better repartition of the load throughout the off peak hours and achieves a lower peak (24.57 kW) compared to the benchmark solution (28.64 kW), while also reducing the energy consumed during peak TOU hours.
• Figure 3: Example of pricing strategy for benchmark (top) and MPC (bottom) solutions, overlaid with the average station power profile.