Gas Stations Reborn: A Multi-Stage Blueprint for Hybrid Fuel-and-Charge Hubs
Amid the thunderous hum of a fast-charging bay and the familiar hiss of a fuel nozzle refilling a tank, a new kind of energy station is taking shape—one that doesn’t choose sides in the great mobility transition, but instead bridges them. As electric vehicle (EV) adoption surges across global cities, legacy infrastructure once devoted entirely to gasoline is quietly being reinvented—not scrapped, not abandoned, but repurposed with surgical precision. In Beijing, Shanghai, and soon in dozens of metropolitan centers worldwide, traditional gas stations are evolving into integrated fueling and fast charging stations (IFFSs): dual-purpose hubs offering both high-speed EV charging and conventional refueling, buffered by on-site battery storage to manage grid stress and maximize uptime.
This isn’t retrofitting as an afterthought. It’s strategic infrastructure evolution—deliberate, data-driven, and economically calibrated for the long haul. A groundbreaking study published this year in Proceedings of the CSEE lays out a rigorous, multi-stage planning framework that treats the transition from gasoline to electricity not as a binary flip, but as a nuanced, decades-long migration. At its core: the recognition that drivers aren’t automatons executing centrally optimized routing—they’re rational agents making real-time trade-offs between time, cost, and convenience. Ignoring that behavioral reality, the authors argue, produces elegant models that fail on the pavement.
Back in 2017, when France and the UK first announced bans on new internal combustion engine (ICE) vehicles by 2040, many assumed the future would simply erase the gas station. Yet nearly a decade later, reality has proven more complex. Fleet turnover is slower than policy timelines; rural adoption lags urban electrification; and hybrid vehicles—which still require refueling—remain a dominant transitional technology. Meanwhile, oil majors like Sinopec have publicly committed to building 5,000 EV charging and battery-swap stations by 2025—not by greenfield development, but by transforming their existing retail footprint. This pivot reflects a deeper truth: land in dense urban centers is scarce, expensive, and often zoned with decades-old permits. Demolishing a functioning station to rebuild a charging-only facility makes little economic or environmental sense. Far smarter to adapt in place.
But adaptation demands more than bolting chargers onto a forecourt. The challenge is operational, financial, and profoundly spatial: Which stations should be upgraded first? How many chargers should coexist with how many pumps—and for how long? When is the right moment to retire a fuel dispenser and reclaim that square footage for power electronics or battery stacks? And critically: how do you size everything so that no driver waits more than 30 minutes in line, peak or off-peak, without overspending on idle capacity?
Enter the work of Xiaoyu Duan, Zechun Hu, Yonghua Song, and Jianzhou Feng—a team straddling Tsinghua University and the University of Macau’s State Key Lab for Smart City IoT. Their model sidesteps two common pitfalls in infrastructure planning: first, the assumption that demand distributes itself according to socially optimal routing (e.g., minimizing total citywide travel time); and second, the reliance on fine-grained traffic network simulations that become brittle over decade-scale horizons. Instead, they propose a rasterized—that is, gridded—urban model, where demand originates from discrete cells (e.g., 2 km × 2 km zones), and stations compete to attract rational users from neighboring cells based on a composite “cost”: drive time (estimated via Manhattan distance and average speed), expected queue delay, service duration, and energy price.
This user-equilibrium approach isn’t new in transport theory—but its application to hybrid energy infrastructure is. Here, every driver calculates their own optimal stop. If Station A adds two extra 200 kW chargers, its queue time shrinks, its “cost” drops, and demand from nearby cells naturally shifts toward it—without any central dispatcher issuing commands. That self-organizing behavior is baked directly into the planning loop: the model iteratively solves for demand allocation given current station capacities, then re-optimizes capacities given the new demand map, converging on a stable equilibrium via Brouwer’s Fixed Point Theorem—a mathematical guarantee that at least one such stable configuration exists.
In practice, this means planners no longer guess how users will behave. They simulate it. And the results are revealing. In a simulated Beijing core zone—300 grid cells served by 15 large legacy stations—the model predicts a non-linear, accelerating shift: by Stage 4 (20 years out, with 50% EV penetration), total fueling demand drops to a fraction of its initial level, but the distribution of remaining demand becomes highly concentrated—not because people drive more, but because rational users flock to the few remaining high-capacity fuel bays that still offer acceptable wait times. Conversely, charging demand spreads more evenly, yet hotspots emerge around high-density commercial and transit nodes, where even minor queue spikes trigger visible demand leakage to neighboring stations.
Crucially, the model reveals that timing matters more than volume. The largest investment in chargers doesn’t appear in the final stage—when EVs dominate—but in Stage 1, when the first wave of commercial fleet operators (e.g., ride-hailing, delivery EVs) begin adopting fast-charging at scale. Early movers need infrastructure now, not in a decade. Delaying Stage 1 build-out doesn’t save money; it cedes market share to competitors and depresses utilization in later stages.
Equally telling is the role of on-site energy storage. The study treats battery systems not as optional add-ons, but as essential grid-smoothing devices. With 200 kW DC fast chargers drawing megawatts during peak hours, a single station can strain local transformers—especially if multiple vehicles plug in simultaneously. By installing storage (e.g., 500 kWh lithium-ion systems, costing ~$19,500 per kWh in present-value terms), operators can “shave” peak demand, buying power slowly off-peak and discharging rapidly during charging surges. This avoids costly grid upgrades, reduces electricity bills under time-of-use tariffs, and—critically—improves service reliability during outages. The model shows that in high-utilization locations, storage isn’t a luxury; it’s a prerequisite for economic viability.
But perhaps the most pragmatic insight lies in how the model handles decommissioning. Most academic studies treat infrastructure expansion and contraction as independent processes. This one doesn’t. It explicitly couples them through shared resource constraints: land area, electrical substation capacity, capital budgets. And here, human behavior re-enters the equation—not just of drivers, but of managers. If tearing out a fuel dispenser incurs net costs (equipment removal, hazardous material handling, lost short-term revenue), operators will delay retirement until absolutely forced—typically when space or power is needed for chargers. Yet if decommissioning saves money (e.g., avoiding annual maintenance on idle pumps), rational firms will shed capacity early, even before demand fully evaporates. The model accommodates both scenarios by decoupling the optimization: when fuel removal is costly, it prioritizes charger rollout first and only removes pumps when resource bottlenecks appear; when it’s profitable, it front-loads pump retirement to liberate resources.
The implications are strategic. A city planning agency using socially optimal demand allocation—where users are assumed to distribute themselves to minimize total system cost—might build fewer, larger charging hubs, relying on centralized routing apps to guide traffic. But in reality, drivers pick the perceived fastest option, not the globally optimal one. The study’s simulations show that plans based on the former overestimate utilization at mid-tier stations and underestimate congestion at prime locations—leading to 6–9% higher actual operating costs than projected. In contrast, user-equilibrium-based designs, though slightly more expensive in planning-stage estimates, deliver lower real-world costs because they match how people actually behave.
Real-world validation is emerging. In Shenzhen, China’s EV capital, PetroChina has converted over 120 stations to hybrid IFFS format since 2021, typically keeping 4–6 fuel lanes while adding 8–12 fast chargers and 1–2 MWh of storage. Early telemetry shows near-identical utilization curves to the Tsinghua model: initial Stage 1 charger uptake dominated by taxi and logistics fleets; Stage 2 marked by rising private EV adoption during off-peak hours; and Stage 3 (now underway) seeing fuel demand collapse outside of long-haul trucking corridors. Wait times remain under 25 minutes across 92% of operating hours—within the 30-minute threshold deemed acceptable for urban refueling by industry consensus.
Yet challenges remain. The current model treats fuel and electric queues as independent, though in practice, a clogged fuel island can spill into charging bays, especially at constrained sites. Future iterations may integrate cross-modal queuing dynamics. Similarly, the assumption of continuous capacity variables—e.g., “6.73 chargers”—requires integer-rounding heuristics for real deployment; the research team is now exploring mixed-integer extensions. And while the raster approach sacrifices street-level topography, it gains robustness over 20-year planning horizons where road networks, zoning, and even city boundaries may shift unpredictably.
Perhaps the most profound shift the IFFS model enables is cultural: it moves the narrative from replacement to coexistence. For decades, energy transitions were framed as zero-sum contests—coal vs. gas, oil vs. electrons. But infrastructure doesn’t turn over overnight. The average U.S. gas station is over 30 years old; many in Europe and Asia are older. Their physical assets—underground tanks (once remediated), canopy structures, utility connections, and, most importantly, real estate—represent sunk capital that can be productively redeployed. An IFFS isn’t a compromise; it’s an acknowledgment that the EV revolution will unfold unevenly, across geographies, vehicle segments, and user demographics.
From a policy standpoint, the study offers a template for adaptive regulation. Instead of mandating fixed charger-to-pump ratios, cities could require stations to submit dynamic transition plans—updated every five years—based on local EV penetration, queue performance, and land-use efficiency. Regulators could then grant expedited permitting for upgrades that demonstrably improve demand coverage or reduce grid strain. Utilities, meanwhile, could offer demand-charge credits for stations that deploy storage and participate in grid-support programs.
The business case is equally compelling. According to the team’s NPV analysis, a well-executed multi-stage IFFS rollout can achieve positive returns even under conservative assumptions—especially when factoring in avoided costs: no land acquisition, reduced permitting delays, and leveraged brand equity. Oil retailers aren’t being forced into obsolescence; they’re being handed a playbook for relevance. As one Sinopec executive recently remarked: “We don’t sell gasoline. We sell energy convenience. The nozzle just changed shape.”
Looking ahead, the IFFS concept may expand beyond cars. Heavy-duty trucking corridors are already testing megawatt-scale charging alongside hydrogen refueling. Urban micro-mobility hubs could layer in e-bike and e-scooter swap stations. And as vehicle-to-grid (V2G) technology matures, tomorrow’s IFFS might not just draw power from the grid—but stabilize it, using parked EVs as distributed storage assets. The station becomes less a point of consumption and more a node in a responsive, bidirectional energy web.
None of this happens automatically. It requires planning that respects both engineering constraints and human agency—that anticipates not just what vehicles will dominate roads in 2040, but how drivers will decide where to stop along the way. The work by Duan, Hu, Song, and Feng doesn’t predict the future. It equips operators to navigate it—stage by stage, charger by charger, queue minute by queue minute—without ever losing sight of the person behind the wheel.
In an era of radical disruption, the most resilient infrastructures aren’t those that leap boldly into the unknown, but those that evolve—intelligently, incrementally, and without leaving anyone stranded on the side of the road.
—
*Xiaoyu Duan¹, Zechun Hu¹, Yonghua Song², Jianzhou Feng¹*
¹Department of Electrical Engineering, Tsinghua University, Beijing 100084, China
²State Key Laboratory of Internet of Things for Smart City, University of Macau, Taipa 999078, Macau SAR, China
Proceedings of the CSEE*
DOI: 10.13334/j.0258-8013.pcsee.221514