EV Aggregators Now Fine-Tune Charging Schedules Using Human Behavior—Not Just Algorithms
In a quiet suburban parking lot just outside Nanjing, a silver BYD Han plugs into a Level 2 charger shortly after midnight. The driver isn’t hovering nearby with a phone in hand, anxiously monitoring the battery. In fact, she’s fast asleep—miles away. Her vehicle, like nearly 200 others in the same neighborhood, is responding to a subtle economic cue: a drop in electricity pricing, orchestrated in real time by a local EV aggregator leveraging deep behavioral insights. It’s not magic. It’s not even particularly high-tech hardware. What’s novel here is how the software thinks—more like a behavioral economist than an engineer.
Until recently, the conversation around electric vehicle (EV) grid integration fixated almost exclusively on hardware: bigger batteries, smarter inverters, faster chargers. Behind the scenes, however, a tectonic shift is underway—one that places human decision-making, not just kilowatt-hours, at the heart of energy system design. A new wave of research, spearheaded by a team at Nanjing Institute of Technology, is redefining how EVs interact with the grid—not by forcing compliance through rigid control, but by nudging participation through psychologically informed price signals.
The implications couldn’t be more timely. As global EV fleets swell—China alone added over 2 million battery-electric vehicles in 2022—the risk of unmanaged charging overwhelming local distribution networks is no longer theoretical. Transformers are groaning under evening “double-peaks,” where residential demand collides with mass EV plug-ins just after the 6 p.m. commute. Traditional demand-response schemes, which treat EV drivers as passive units in a giant spreadsheet, are proving insufficient. Why? Because people aren’t algorithms. They don’t respond linearly to price. Their willingness to delay charging, or—more radically—to discharge back into the grid, hinges on a tangled web of practical needs, battery anxiety, and perceived fairness.
Enter the “user responsiveness” paradigm.
Led by Jun Li, a professor whose work straddles power systems and human-centered design, the research team set out to answer a deceptively simple question: What actually makes an EV owner say “yes” to shifting their charging—or even selling power back—beyond just the dollar amount on the price tag? Their answer, detailed in a recent paper published in Southern Power System Technology, moves beyond the assumption that users will automatically flock to the cheapest tariff window. Instead, they built a model grounded in Weber–Fechner Law—a principle from psychophysics stating that human perception of stimulus change (like a price drop) is proportional to the relative change, not the absolute one.
Think of it this way: a ¥200/MWh discount feels massive when baseline charging costs ¥400, but barely registers when it’s already ¥1,200. More crucially—and this is where past models fell short—the team recognized that an EV’s state of charge (SOC) is the silent arbitrator in every decision. A vehicle sitting at 90% SOC at 7 p.m.? Its owner might happily sell some juice back to the grid if the price is right. But a vehicle at 20% SOC, with a 50-km school run scheduled for 7:30 a.m.? No amount of financial incentive will override the primal need for range assurance.
To capture this nuance, Li and his colleagues—Jiacheng Liang, Ketian Liu, Wei Han, Xiao Liang, and Xin Li—developed a dual-response surface: one for charging, one for discharging. Each surface maps two axes—dynamic pricing signals and real-time SOC—into a probability of participation. When aggregated across a fleet, this yields what the authors call “schedulable capacity”: not the theoretical maximum power a garage of EVs could supply, but the realistic, behaviorally constrained amount they’re likely to contribute at any given hour.
The model was tested against real-world travel data: the U.S. National Household Travel Survey (NHTS 2017), adapted for Chinese commuting patterns. Simulations involved 50, 100, and 200 EVs connected to a representative distribution feeder. Three scenarios were compared: Uncontrolled (plug-in-and-forget), Controlled Charging (only charge-shifting allowed), and Coordinated Charge-Discharge (full V2G participation, guided by the responsiveness model).
The results were striking—not just in magnitude, but in quality of impact.
In the uncontrolled case, the familiar evening surge reappeared with brutal clarity: system load variance jumped to over 19,600 kW² when 200 EVs joined the grid. Transformers flirted with overload thresholds. Average user charging costs hovered around ¥8.50 per session, and aggregator profits remained meager—barely breakeven in many cases.
Switching to coordinated charge-discharge yielded transformative outcomes. With 200 EVs, the load variance collapsed by over 93%—down to just 1,377 kW²—smoothing the daily curve into something resembling a gentle hill rather than a jagged mountain range. Crucially, this wasn’t achieved by sacrificing user welfare. The average charging cost plummeted to ¥4.47—a 48% reduction—because drivers were effectively “buying low and selling high”: topping up overnight at cheap rates and exporting surplus during expensive peak hours. Meanwhile, the aggregator’s daily profit soared to ¥336, a more than doubling of the uncontrolled baseline.
But the most revealing insight came from comparing dynamic pricing against time-of-use (TOU) pricing—where rates change only a few times per day, like current utility tariffs. Dynamic pricing, updated hourly, delivered superior grid outcomes: lower variance, lower user costs. Yet it came at a hidden cost: cognitive load. Frequent price fluctuations, the team warns, can erode user trust and participation over time. Drivers may perceive the system as “gaming” them, or simply grow fatigued by the mental accounting required. The TOU variant, while slightly less optimal numerically (19% higher load variance), generated higher aggregator profits—suggesting users were willing to accept marginally higher charging bills in exchange for predictability and simplicity.
This tension—between mathematical optimality and human sustainability—lies at the core of the next-generation EV-grid interface. The research doesn’t just propose a better algorithm; it reframes the problem entirely. The aggregator is no longer a top-down scheduler, but a market facilitator whose success hinges on empathy. Its pricing signals must be legible, fair, and psychologically resonant—not just economically efficient.
Industry players are already taking notice. A growing cohort of “virtual power plant” (VPP) startups is moving beyond simple aggregation toward behavioral orchestration. One Shenzhen-based firm now embeds “confidence thresholds” into its user app: instead of demanding a fixed departure SOC of 90%, it asks, “How comfortable would you feel leaving with 80%?” and adjusts scheduling envelopes accordingly. Another Hangzhou aggregator has introduced “commitment bonuses”—small, guaranteed payouts for users who pre-authorize participation in at least three V2G events per week, turning sporadic engagement into habitual cooperation.
Critically, this approach sidesteps the privacy minefield that has plagued earlier smart-grid initiatives. The model doesn’t require minute-by-minute GPS tracking or deep personal profiling. It needs only three inputs per vehicle: plug-in time, plug-out time, and initial SOC—all of which users willingly share in exchange for lower bills and “grid hero” status. The rest is inferred statistically across the fleet, preserving individual anonymity while unlocking collective value.
From a regulatory perspective, the findings challenge outdated notions of grid neutrality. If EVs can act as responsive resources—not just loads—then compensation structures must evolve. Today’s net metering rules, designed for solar panels with fixed generation curves, are ill-suited for the stochastic, bidirectional nature of EV participation. The paper implicitly calls for new tariff architectures: perhaps a two-part rate comprising a flat “connectivity” fee plus a performance-based “flexibility” credit, paid per kilowatt-hour of verified demand reduction or injection.
One of the most underappreciated benefits highlighted by the study is grid resilience. During a simulated late-afternoon demand spike—mimicking an unexpected heatwave—the coordinated EV fleet didn’t just shave the peak; it absorbed it. For 90 critical minutes, the aggregate discharge from parked vehicles offset 18% of the system’s shortfall, preventing the need to dispatch expensive, carbon-intensive peaker plants. This isn’t ancillary service procurement; it’s community-scale shock absorption, delivered by idle cars.
Still, challenges remain. Battery degradation concerns linger, particularly among early adopters. While modern LFP chemistries are remarkably robust under moderate V2G cycling, public perception lags. The researchers acknowledge this, noting in their conclusion that future work must integrate perceived degradation costs into the responsiveness model—even if the actual wear is minimal. Trust, after all, is as much a grid resource as copper or silicon.
Equally important is the question of equity. Will this new flexibility economy primarily benefit affluent urbanites with private garages and bi-directional chargers? Or can it be extended to apartment dwellers, fleet operators, and rural communities? The paper’s framework is inherently scalable—its responsiveness surfaces can be re-parameterized for different user segments—but deployment strategy matters. Pilot programs in Suzhou are now testing “shared EV buffer pools” at multi-unit dwellings, where a building-level battery acts as an intermediary, shielding individual tenants from direct grid interaction while still capturing system-wide benefits.
What sets this work apart is its refusal to treat the human element as noise to be filtered out. For decades, power engineers sought to eliminate variability—through spinning reserves, frequency regulation, and ever-tighter control loops. This research flips the script: it harnesses variability, recognizing that the messiness of human behavior—when properly understood and respected—can be a feature, not a bug.
Back in that Nanjing parking lot, the silver BYD Han finishes its overnight top-up just before 5 a.m., settling at 92% SOC. At 6:45 p.m., with the household’s evening AC load ramping up, it quietly discharges 5 kWh into the home’s microgrid—enough to cover dinner prep without flickering a single lightbulb. The owner receives a ¥3.20 credit, applied automatically to next month’s bill. She doesn’t think about ramp rates or voltage sags. She thinks: “Nice. My car helped pay for groceries.”
That’s the real breakthrough. Not teraflops of optimization, but a simple, quiet moment of alignment—between driver and vehicle, vehicle and grid, individual and system. When technology recedes into the background and value becomes tangible, participation ceases to be a chore and becomes, quite naturally, a habit.
The road ahead won’t be without bumps. Interoperability standards for V2G communication remain fragmented. Regulatory sandboxes are still too small and short-lived. Consumer education is lagging. But the direction is unmistakable: the future grid won’t be commanded by a central AI. It will emerge, adaptively and elegantly, from millions of small, intelligent, human-centered decisions—each one a tiny vote for a more resilient, affordable, and equitable energy system.
And sometimes, the most powerful votes are cast while their owners are asleep.
Jun Li, Jiacheng Liang, Ketian Liu, Wei Han, Xiao Liang, Xin Li
School of Electric Power Engineering, Nanjing Institute of Technology, Nanjing 211167, China; Huaian Power Supply Company of State Grid Jiangsu Electric Power Co., Ltd., Huaian, Jiangsu 223001, China
Southern Power System Technology, Vol. 17, No. 8, Aug. 2023
DOI: 10.13648/j.cnki.issn1674-0629.2023.08.014