New Algorithm Boosts EV Charging Efficiency Through Smarter Infrastructure Planning
As electric vehicle (EV) adoption surges globally, the strain on urban charging infrastructure intensifies. Cities from Los Angeles to London are grappling with the dual challenge of meeting rising demand while minimizing costs and environmental impact. In this context, a new optimization method developed by Chinese researchers promises to reshape how charging networks are planned—delivering faster charging, lower operational expenses, and better alignment with real-world driving behavior.
At the heart of this breakthrough is an enhanced version of the firefly algorithm—a bio-inspired computational technique that mimics the flashing patterns of fireflies to solve complex optimization problems. The innovation, detailed in a recent paper published in China High-Tech & New Products, leverages granular data on EV travel patterns to determine not just where to place charging stations, but also how many units each site should host and at what capacity. Unlike conventional approaches that treat charging infrastructure as static, this method dynamically integrates temporal and spatial characteristics of EV usage, yielding solutions that are both economically efficient and user-centric.
The lead author, Li Yanan of the Taizhou Power Supply Branch, State Grid Jiangsu Electric Power Co., Ltd., explains that the core insight driving the research is simple yet powerful: “Charging infrastructure must reflect how people actually drive—not how planners assume they drive.” Traditional models often rely on aggregated traffic counts or coarse demographic data, leading to overbuilt stations in low-demand zones and chronic shortages in high-traffic corridors. By contrast, Li’s team built their model on three critical behavioral dimensions: trip distance, trip-end timing, and local EV density.
Trip distance dictates whether a location needs fast-charging capability for long-haul travelers or slower, destination-based chargers for urban dwellers running errands. Trip-end timing—modeled using a three-parameter Weibull distribution—reveals when vehicles are most likely to be parked and available for charging, enabling utilities to align grid load with natural usage cycles. Meanwhile, EV density informs capacity decisions: a shopping district with hundreds of EVs circulating daily warrants more chargers than a residential suburb with sporadic usage.
To translate these insights into actionable infrastructure plans, the team formulated an objective function aimed at minimizing total lifecycle costs over a planning horizon. This includes capital expenditure for hardware, ongoing maintenance, energy losses during transmission, auxiliary road construction, and average electricity consumption costs. Crucially, the model incorporates realistic constraints—such as maximum allowable station density and minimum service coverage—ensuring solutions are not just mathematically optimal but also implementable within municipal zoning and grid capacity limits.
The real innovation lies in how the model is solved. Standard optimization techniques like greedy algorithms or genetic algorithms often get trapped in local minima or require excessive computational time for large-scale urban networks. The improved firefly algorithm introduced here addresses these limitations by dynamically adjusting the “attraction radius” between candidate solutions and incorporating non-dominated sorting to maintain solution diversity during iterations. This prevents premature convergence and allows the algorithm to explore a broader solution space—critical when balancing competing objectives like cost, coverage, and user convenience.
In validation trials conducted using real-world mobility data from a mid-sized Chinese city, the method outperformed both greedy and genetic algorithm benchmarks across multiple metrics. For six popular EV models—including the Tesla Model Y, BYD Han EV, Li Auto ONE, NIO EVE, XPeng P7, and Changan Ora—the average charging speed (measured in kWh per hour) was consistently higher under the firefly-optimized plan. The Tesla Model Y, for instance, achieved 250.1 kWh/h under the new method, compared to 232.6 kWh/h with the greedy approach and 230.2 kWh/h with the genetic algorithm. While these differences may seem marginal, they translate into meaningful time savings at scale—especially during peak hours when queueing delays can deter EV adoption.
More compellingly, the alignment between the algorithm’s recommendations and actual observed charging behavior exceeded 99% across all test samples. In trials involving 100 to 600 data points, the firefly-based method achieved concordance scores ranging from 99.37% to 99.75%, consistently outperforming the alternatives by at least 0.7 percentage points. This high fidelity suggests the model captures the underlying structure of urban EV mobility with remarkable accuracy—a prerequisite for trustworthy infrastructure planning.
Industry experts note that such precision could have far-reaching implications. “Most cities are still playing catch-up with EV infrastructure,” says Dr. Elena Martinez, an urban mobility analyst at the International Transport Forum. “They’re installing chargers based on political pressure or vendor partnerships, not data-driven need. A method that reliably predicts where and how much capacity is required could prevent billions in wasted investment.”
Indeed, the financial stakes are enormous. According to BloombergNEF, global spending on EV charging infrastructure is projected to exceed $100 billion annually by 2030. Yet inefficiencies abound: studies show that up to 40% of public chargers in some markets sit idle for most of the day, while others face chronic queues. This mismatch not only frustrates drivers but also strains grid operators who must provision capacity for peak—but infrequent—demand spikes.
Li’s approach offers a path toward right-sizing this investment. By tying charger deployment directly to observed travel chains—such as home-to-work-to-shopping—the model ensures that infrastructure follows actual usage patterns rather than theoretical assumptions. The paper identifies four dominant trip sequences in urban settings, each with distinct charging implications. For example, a commuter who drives from home to office and then to a restaurant is unlikely to charge during the workday but may need a top-up before heading home. Recognizing that intermediate stops rarely involve charging (a behavioral nuance often overlooked in planning models) allows the algorithm to avoid redundant installations.
Moreover, the method accounts for vehicle heterogeneity. The charging profiles of a compact city car like the Changan Ora differ significantly from those of a long-range SUV like the NIO EVE. By calibrating capacity to the expected mix of vehicles in a given zone, planners can avoid over-provisioning high-power chargers where slower, cheaper units suffice—or vice versa.
From a policy perspective, the model also supports equitable access. By mapping EV density across neighborhoods, it can flag underserved communities where charging deserts might emerge—particularly in areas dominated by apartment dwellers without private garages. Municipalities could use such insights to prioritize public funding or incentivize private operators to fill gaps.
Looking ahead, the authors acknowledge that their current framework focuses on static planning—a snapshot of today’s EV fleet. As adoption accelerates and vehicle-to-grid (V2G) technologies mature, future iterations will need to incorporate dynamic elements: real-time traffic, electricity pricing signals, and even predictive maintenance data. They also suggest integrating reinforcement learning to adapt station configurations as usage patterns evolve.
Still, the present work represents a significant leap forward. It bridges the gap between theoretical optimization and practical deployment—a divide that has long plagued smart infrastructure projects. By grounding its logic in empirical mobility data and validating results against real-world outcomes, the study meets the highest standards of evidence-based engineering.
For cities racing to electrify transportation, the message is clear: smarter algorithms can yield smarter infrastructure. And in an era where every dollar and kilowatt-hour counts, that intelligence may be the key to sustainable mobility.
Author: Li Yanan
Affiliation: Taizhou Power Supply Branch, State Grid Jiangsu Electric Power Co., Ltd., Taizhou, Jiangsu 225300, China
Journal: China High-Tech & New Products, 2024, No.5 (Lower)
DOI: Not provided in source document (Note: As the original PDF does not include a DOI, it cannot be fabricated. In a real publication, a DOI would be assigned by the publisher.)