For decades, utility distribution planning has been built around a predictable daily load shape: a morning ramp, a midday plateau, an afternoon peak, an evening decline. The curves differed by climate and region but followed recognizable patterns that decades of AMI interval data and load research had characterized well enough to plan around.
EV adoption at scale is breaking those curves. In service territories with meaningful EV penetration, the 6 PM to 9 PM window now carries a second peak that superimposes on the declining edge of the traditional afternoon peak. In some suburban distribution circuits, this is already the dominant stress event of the day. The planning assumptions baked into distribution infrastructure investments made five years ago did not include this load shape.
The mechanics of the evening EV peak
The basic driver is simple. Most EV owners charge at home using Level 2 EVSE (240V, typically 7.2 to 11.5 kW draw). They return from work between 5 and 7 PM and plug in immediately. In the absence of any managed charging signal, the vehicle charges at full rated power from plug-in until the battery reaches a programmed state, which for many drivers is the default full charge setting.
On a residential feeder with 300 homes, adding 50 EVs that each draw 7.2 kW simultaneously in the 5 PM to 8 PM window is 360 kW of new load. The same feeder's traditional residential evening peak might be 900 kW. This is a 40% increase in feeder peak demand from EV charging alone, concentrated in the two hours when the feeder is already under stress from HVAC and appliance loads. The transformer serving that section of feeder was sized for the original peak, not this one.
What makes this particularly challenging from a DER orchestration perspective is that EVs behave differently from the loads and resources that existing demand response programs were built around. A commercial HVAC system enrolled in a demand response program can be pre-cooled and then cycled to reduce load during a demand event. The flexibility envelope is defined by building thermal mass and temperature setpoint ranges: slow to respond, wide to influence.
An EV battery sitting at 20% SoC at 6:15 PM has no thermal buffer and limited flexibility. The owner needs that vehicle charged by 7 AM tomorrow. The flexibility window for charging delay is narrower and more condition-dependent than traditional DR asset flexibility. Applying the same dispatch logic designed for commercial loads to a residential EV fleet will either fail to shift load meaningfully or generate customer complaints and opt-outs that erode program participation over time.
What managed charging programs actually need to track
The data visibility requirements for a well-run managed EV charging program are more demanding than for a traditional demand response program, and most existing DERMS platforms were not designed with those requirements in mind.
For each enrolled EV, the program needs current state-of-charge at check-in time (when the vehicle plugs in), the customer's stated departure time or departure-time inference from historical plug-in patterns, and the current feeder load on the circuit serving that address. From those three inputs, the dispatch logic can calculate the latest charging start time that still achieves a full charge by departure, and schedule charging across enrolled vehicles on that feeder to spread the load over a wider time window.
The challenge is that most EVSE APIs expose plug-in state and current draw, but state-of-charge data availability varies by vehicle-to-EVSE communication protocol. The OCPP (Open Charge Point Protocol) standard used by most commercial EVSE does not include vehicle SoC in its basic telemetry; the vehicle SoC comes from the vehicle manufacturer's telematics API, which is a separate data source with its own authentication requirements, data update cadence, and coverage gaps for older vehicles without telematics subscriptions.
Without reliable SoC data, the managed charging program cannot calculate the actual flexibility window for each vehicle. It defaults to conservative assumptions (short delay window, early charging start to avoid risk of undercharging) that reduce the achievable load shift. The program that looks good on paper because it enrolled 500 EVs may achieve a fraction of its projected peak reduction because the dispatch logic is working with incomplete state information.
How EVs interact with behind-the-meter solar and batteries
For customers who have rooftop solar, a paired battery, and an EV, the interaction between these three assets adds another layer of complexity. In a well-optimized home energy system, the battery might absorb excess solar generation during the afternoon, then discharge to power EV charging in the evening, reducing grid import to near zero during the peak window. From the feeder perspective, this customer is a zero-draw asset during the evening peak if the battery was charged to sufficient capacity during the day.
In an uncoordinated scenario, the same customer may have an EVSE that pulls directly from the grid because the home energy management system doesn't have communication with the EVSE. The battery is full, the solar is no longer generating, and the EVSE is drawing 7.2 kW from the grid because no orchestration layer is coordinating the three assets to use local stored energy first.
From the utility's DER program perspective, this customer may be enrolled in both a solar export program and a battery demand response program, but the EV is enrolled in neither, or is enrolled in a separate managed charging program that doesn't know about the battery state. The three programs operate independently, and the optimal dispatch outcome that would have used stored solar energy to charge the EV while managing feeder load never gets executed because no system has the full picture.
The gap between enrollment and operational readiness
Most utility EV programs currently function at the enrollment layer, not the dispatch layer. The utility knows which customers are enrolled in managed charging. It can send a demand response signal through an OpenADR 2.0b event notification that instructs enrolled EVSE to delay charging start. But the precision of that instruction is low: "delay charging by 2 hours" applies uniformly to all enrolled customers regardless of whether their vehicle is at 15% SoC or 85% SoC, and regardless of whether the feeder they are on is under stress or running well within capacity.
The gap between this blunt instrument and feeder-level dispatch optimization is exactly the kind of problem that benefits from a DER orchestration layer that maintains current state for each enrolled asset. Not another portal to log into; a platform that continuously receives EVSE telemetry, pulls vehicle telematics where available, knows the current feeder load from SCADA, and translates that information into per-customer dispatch signals that reflect actual flexibility rather than assumed flexibility.
We are not saying the enrollment-layer programs are failing; they provide real value in their current form. The question is whether the value ceiling of those programs is high enough given the pace at which EV load is growing. In service territories where EV penetration is already reshaping peak load curves, a 2-hour uniform delay program may be insufficient within a few years. Building the operational infrastructure to do better while the EV fleet is still small is less disruptive than trying to retrofit it when the fleet has grown to the point where the feeder stress is already visible in reliability statistics.
What planning and operations teams should be tracking now
Distribution planning teams in higher-EV-penetration service territories have an opportunity to instrument their circuits before the load shape problem becomes acute. AMI interval data already shows the emerging evening peak signature in any circuit where EV adoption has reached 10 to 15 percent penetration. Identifying those circuits, characterizing the load growth trajectory, and mapping the enrolled DER assets (solar, storage, EVSE where available) on those circuits now provides the baseline for a feeder-level demand response strategy before the transformer or secondary conductor loadings hit threshold.
The operational side requires knowing, in near real-time, which enrolled EVs on a given feeder are plugged in and what their charge state is, so that an evening peak pre-positioning action can target the right customers rather than all enrolled customers uniformly. That is a data infrastructure investment, and it is the kind that takes time to deploy and validate. The utilities that start building it while they have operational headroom will be better positioned to manage the load shape problem they are going to face over the next three to five years.