Optimizing EV Charging Infrastructure for Valet Services
As the electric vehicle (EV) market surges forward, with China reporting a 43.9% month-on-month increase in new energy vehicle sales in March 2022, the industry faces a critical bottleneck: charging infrastructure. While EV adoption accelerates, owners continue to grapple with long wait times, range anxiety, and the inconvenience of personally managing charging sessions. In response, a new service model—valet charging—is emerging as a promising solution. This innovative approach not only alleviates user stress but also presents complex logistical and economic challenges that demand sophisticated optimization strategies. A recent study by Tong Min and Hu Zhihua from the Logistics Research Center at Shanghai Maritime University, published in Operations Research and Management Science, offers a groundbreaking framework for optimizing the placement of charging stations specifically tailored to valet charging operations.
The valet charging model operates on a simple yet transformative premise: instead of drivers navigating to a charging station, trained staff are dispatched to collect the vehicle, charge it at a nearby facility, and return it to the owner. This service eliminates the need for users to spend hours waiting, significantly reducing charging-related anxiety. However, the profitability and efficiency of such a service hinge on a critical factor—the strategic placement of charging stations. Unlike traditional models where users choose stations based on proximity, valet services introduce a layer of operational complexity. The charging staff must efficiently travel to the customer, transport the vehicle, charge it, and return—all while minimizing travel time, fuel costs, and labor expenses. This creates a dynamic where the optimal network of charging stations is not necessarily the one that minimizes user travel, but rather the one that maximizes operational profit while maintaining service quality.
Tong and Hu’s research addresses this intricate balance by developing a multi-objective, two-stage stochastic mixed-integer programming (TSMIP) model. This advanced mathematical framework is designed to navigate the inherent uncertainties of urban EV charging demand. The model acknowledges that user demand is not a fixed number but a random variable, fluctuating based on time of day, location, and individual behavior. By incorporating this stochastic nature, the model avoids the pitfalls of deterministic planning, which can lead to under- or over-provisioning of infrastructure.
The core of their approach lies in the concept of “valet charging service ratio.” This ratio quantifies the likelihood that a user in a given area will opt for the valet service over self-charging. The decision is influenced by two primary factors: the time cost of charging and the price of the valet service. The researchers model the “time cost” as a random variable, reflecting the fact that different users place different values on their time. For instance, a busy executive may have a much higher time cost than a retiree. This is captured by assuming that the unit time cost for users follows a uniform distribution within a defined range. If the valet service price is less than or equal to the product of a user’s time cost and the total charging time (driving to the station plus waiting), they are more likely to choose the service. This elegant formulation allows the model to predict service adoption rates across a diverse user population.
The optimization problem is structured in two stages. The first stage involves long-term, strategic decisions: where to build charging stations and how many charging points to install at each location. These decisions are made before the exact charging demand for a given period is known. The second stage deals with operational responses: once the actual demand is realized, how should the available charging capacity be allocated to serve users, either through self-charging or valet services? The model seeks to minimize the total cost, which includes the capital and operational expenses of building and maintaining the charging infrastructure, as well as the expected travel costs incurred by users who choose to charge their vehicles themselves.
However, the ultimate goal for a valet charging provider is not just cost minimization but profit maximization. To achieve this, Tong and Hu extend their model into a multi-objective framework. The primary objective becomes maximizing the expected profit from the valet charging business, which is calculated as the difference between the total revenue from service fees and the total cost of providing the service (including staff wages, vehicle depreciation, and administrative overhead). This profit maximization objective is balanced against the goal of minimizing user travel costs. This creates a classic trade-off: a layout that maximizes profit might require charging stations to be placed in more cost-effective or strategically advantageous locations, even if it means users have to travel slightly farther for self-charging.
To solve this complex multi-objective stochastic problem, the researchers employ two powerful computational techniques: the Epsilon-constraint method and the Sample Average Approximation (SAA) method. The Epsilon-constraint method transforms the multi-objective problem into a series of single-objective problems. By setting a constraint on one objective (e.g., user travel cost must not exceed a certain threshold), the model can then focus on maximizing the other objective (e.g., valet profit). By varying this constraint, decision-makers can explore the entire “Pareto frontier”—the set of all possible solutions where improving one objective would worsen the other. This allows for a nuanced understanding of the trade-offs involved.
The SAA method is used to handle the randomness of user demand. Instead of trying to solve the problem for every possible demand scenario—an impossible task—the SAA method uses a large number of randomly generated demand scenarios. The model is solved for this sample of scenarios, and the average outcome is used as an estimate of the true expected value. This approach provides a robust and practical solution that is resilient to the day-to-day fluctuations in real-world demand. The accuracy of the SAA solution improves as the number of sampled scenarios increases, and the researchers provide a rigorous method to verify the quality of the solution by checking for convergence.
The theoretical framework is brought to life through a compelling numerical analysis and a real-world case study in Yaohai District, Hefei, Anhui Province. The numerical experiments are conducted on a 6km x 6km grid with 36 potential locations for both demand and charging stations. The results are both insightful and cautionary. The study finds that when the sole focus is on maximizing valet charging profit, the optimal charging station layout tends to spread out towards the periphery of the region. This dispersion allows the service to capture demand from a wider area and potentially reduce the average distance traveled by charging staff on collection and delivery trips. However, this comes at a significant cost: users who choose to charge their own vehicles must travel much farther to reach the nearest station, dramatically increasing their travel costs and inconvenience.
This finding underscores a critical policy and business dilemma. A purely profit-driven approach to infrastructure development can lead to a suboptimal outcome for the broader community of EV users. It highlights the need for a balanced strategy that considers the interests of both the service provider and the end-user. The research suggests that decision-makers should not aim for the absolute maximum profit but should instead target a point on the Pareto frontier that offers a more equitable balance. For instance, the case study in Yaohai District demonstrates that by accepting a slightly lower profit margin—around 85% of the maximum possible—a charging network can be designed that keeps user travel costs manageable. This “sweet spot” provides a significant boost to the valet service’s profitability while still ensuring that self-charging remains a viable and convenient option for users.
The implications of this research extend far beyond the confines of a single academic paper. For city planners and transportation authorities, it provides a powerful tool for designing a more efficient and user-friendly charging ecosystem. By understanding the trade-offs, they can create regulations or incentives that encourage the development of charging networks that serve the public good. For example, subsidies or tax breaks could be tied to network designs that prioritize accessibility and minimize user travel costs, preventing private companies from creating “profit islands” that neglect underserved areas.
For private companies entering the valet charging market, the study offers a clear roadmap for sustainable growth. It demonstrates that long-term success is not just about maximizing short-term profits but about building a reputation for reliability and convenience. A network that is too spread out might save on operational costs, but it could alienate a large segment of potential customers who find self-charging too inconvenient. By investing in a denser, more accessible network, even at a higher initial cost, companies can build customer loyalty and capture a larger market share in the long run.
The research also opens up several avenues for future exploration. One key area is the optimization of staff scheduling and routing. The current model focuses on station location but assumes a certain level of operational efficiency. A more integrated model could simultaneously optimize the location of stations, the number of charging staff, and their daily routing schedules, creating a truly end-to-end solution. Another critical area is dynamic pricing. The study uses a fixed service price, but in reality, prices could fluctuate based on demand, time of day, or distance. A model that incorporates dynamic pricing could lead to even greater efficiency and profitability.
Furthermore, the model could be enhanced by incorporating more granular data on user behavior. While the use of a uniform distribution for time cost is a good first approximation, future research could use real-world survey data to create more accurate behavioral models. Factors such as user demographics, vehicle type, and typical driving patterns could all be integrated to create a more sophisticated prediction of service adoption.
In conclusion, the work of Tong Min and Hu Zhihua represents a significant advancement in the field of EV infrastructure planning. By moving beyond simplistic models and embracing the complexity of stochastic demand and user choice, they have developed a robust framework that is both theoretically sound and practically applicable. Their findings serve as a vital reminder that the transition to a sustainable transportation future requires more than just technology; it requires intelligent, human-centered planning. As valet charging services evolve from a niche after-sales perk to a standalone industry, the insights from this research will be indispensable for building a charging network that is not only profitable but also equitable, efficient, and truly user-centric.
Tong Min, Hu Zhihua, Logistics Research Center, Shanghai Maritime University, Operations Research and Management Science, doi:10.12005/orms.2024.0252