Vehicle-to-Grid (V2G) Architecture: ISO 15118-20 Bidirectional Charging, OCPP 2.1 and IEEE 2030.5 for Utilities and Fleets (2026)
Most failed V2G projects did not fail on messaging. The vehicle spoke the protocol and the charger accepted the session. Then the export stopped at the interconnection agreement, because nobody could say which box was the certified grid inverter. Vehicle-to-grid V2G is now a stack of published standards rather than a lab demo. ISO 15118-20 carries bidirectional power transfer between car and charger. OCPP 2.1, released by the Open Charge Alliance in January 2025, adds V2X and DER control between charger and back office. IEEE 2030.5 carries utility commands to the aggregator. The new UL 1741 Supplement SC, approved in May 2026, finally gives AC bidirectional chargers a certification route. This guide treats the stack as one control chain and asks one question at every hop: who owns the setpoint, and who answers to the grid code?
What this covers: the end-to-end V2G reference architecture, the ISO 15118-20 and OCPP 2.1 message flow, AC versus DC V2G and why the inverter’s location decides certification, IEEE 2030.5 and Rule 21, Plug and Charge PKI, worked energy and degradation economics, a Python dispatch scheduler, failure modes, and a decision table.
Context and Background
Vehicle-to-grid is the most demanding member of the V2X family. V2L (vehicle-to-load) powers a tool or a campsite from the car. V2H (vehicle-to-home) and V2B (vehicle-to-building) back up or offset a building behind its meter. V2G exports across the utility meter into the distribution network, or provides grid services such as frequency response or capacity at the request of a market operator, utility or aggregator. Every step along that ladder adds a party, a contract and a safety obligation.
For a decade the only production bidirectional path was CHAdeMO. The Nissan Leaf and Japanese V2H units made DC discharge routine in Japan long before the CCS world caught up. The CCS and NACS ecosystem relied on ISO 15118-2, published in 2014, which handles AC and DC charging, Plug and Charge and smart charging, but not discharge. If you need a refresher on how that generation fits together, our comparison of OCPP 2.0.1, OCPP 1.6 and ISO 15118 covers the unidirectional baseline this post builds on.
Three things changed between 2022 and 2026. ISO published ISO 15118-20 in April 2022, adding bidirectional power transfer (BPT), a Dynamic control mode, mandatory TLS for every identification method, and multiple contract certificates. The Open Charge Alliance shipped OCPP 2.1 in January 2025 with dedicated V2X and DER control functional blocks; IEC later published it as IEC 63584-210:2025. On the grid side, IEEE 1547-2018 made smart-inverter functions and a communications interface mandatory for interconnected DER, and California’s Rule 21 made IEEE 2030.5 with the Common Smart Inverter Profile (CSIP) the default utility protocol. Then, in July 2026, ISO 15118-20 Amendment 1 added AC DER services and the Megawatt Charging System, according to CharIN’s announcement.
The incumbents are therefore no longer missing pieces. The challenge is composition. A V2G session crosses four organisations: the vehicle OEM, the charge point operator (CPO), the aggregator or virtual power plant (VPP), and the distribution utility. Each speaks a different protocol and carries a different liability. Our virtual power plant architecture guide explains the aggregator layer for stationary DER. This post focuses on what makes the mobile battery different: it leaves, it belongs to someone who needs to drive, and its inverter may be in the car or in the wall.
The Vehicle-to-Grid V2G Reference Architecture
A vehicle-to-grid V2G system is a chain of four control loops. The utility or market sends grid needs to an aggregator over IEEE 2030.5 or a market API. The aggregator turns them into per-site limits for the charge point back office (CSMS). The CSMS sends OCPP 2.1 charging profiles to the charger. The charger negotiates power with the car over ISO 15118-20.

Figure 1: Vehicle-to-grid V2G reference architecture. Grid and market signals flow down through the aggregator and the CSMS to the charger, which controls the vehicle over ISO 15118-20. The PKI underpins every TLS hop.
The diagram reads top to bottom. The utility’s distributed energy resource management system (DERMS) exposes an IEEE 2030.5 server that the aggregator polls for DER controls. In parallel, a wholesale market or independent system operator (ISO) sends prices and ancillary-service dispatch. The fleet or telematics system adds the constraint nobody else knows: when each vehicle must leave and with how much charge. The aggregator merges these into limits and setpoints. The CSMS translates them into OCPP 2.1 charging profiles with an operation mode per period. A site energy management system (EMS) can also impose external limits, for example a building’s import cap. Finally the charger controls the car through the ISO 15118-20 BPT services, inside a TLS session whose certificates come from the V2G PKI.
Each hop has a different time constant
The layers are not interchangeable because they run at different speeds. IEEE 2030.5 under CSIP is a polling protocol: DER clients periodically fetch controls from the utility or aggregator server, which trades latency for robustness. OCPP 2.1 is a persistent WebSocket, so the CSMS can push a new charging profile within seconds. ISO 15118-20’s charge loop runs continuously while energy flows, with the car and charger exchanging present voltage, current and targets many times per minute.
That gradient decides where each function belongs. A demand-response event announced an hour ahead fits comfortably in IEEE 2030.5 or OpenADR. A capacity commitment for tomorrow’s evening peak lives in the aggregator’s optimiser. Fast frequency response cannot round-trip through a cloud. OCPP 2.1 acknowledges this directly with two frequency modes. In CentralFrequency, the CSMS keeps sending fresh dynamic profiles to follow frequency curves for services such as FCR or aFRR. In LocalFrequency, the charging station measures grid frequency itself and responds along a configured droop curve. Pick the mode by the service’s response-time requirement, not by architectural taste.
The setpoint owner must be unambiguous
OCPP 2.1 defines eight operation modes for a charging schedule period: Idle, ChargingOnly, CentralSetpoint, ExternalSetpoint, ExternalLimits, CentralFrequency, LocalFrequency and LocalLoadBalancing. They look like a list of features. In practice they answer one question: who is allowed to set the power right now?
ChargingOnly is the initial state for every transaction, with no export permitted. In CentralSetpoint, the CSMS sends a target where positive means charging and negative means discharging, plus hard charge and discharge limits. ExternalSetpoint is the same, except an external entity such as a site EMS or aggregator gateway supplies the target. ExternalLimits lets an external party set only the envelope, which the CSMS can read back. LocalLoadBalancing uses a local smart meter to steer the vehicle’s charge or discharge power. Idle lets the car sleep until a new profile wakes it.
The design rule follows from that list. For every interval, exactly one actor owns the setpoint and every other actor owns, at most, a limit. Ambiguity here is the root of the worst V2G bugs. Two optimisers fight, the car oscillates between import and export, and the site meter records a demand spike nobody planned. Write the ownership matrix down before you write code.
The charger is the trust boundary
The vehicle is a guest on the charger, and the charger is a guest on the grid. That makes the charging station the natural enforcement point for safety and grid-code compliance. It holds the TLS endpoint to the car, the revenue-grade meter, the contactor and the configuration of local grid settings. The Trialog analysis of OCPP 2.1 makes the same point. Charging stations become DER, grid codes must be configurable on the station, and the CPO sits between the station and the utility or aggregator.
In DC V2G this is straightforward, because the inverter is physically inside the charger. In AC V2G the inverter is inside the car, and the charger has to supervise a power converter it does not own. That asymmetry is so important that it gets its own section below.
ISO 15118-20 Bidirectional Charging and the OCPP 2.1 V2X Message Flow
ISO 15118-20 bidirectional charging works by service selection. The car and charger exchange the services they support, and the car selects one. The standard’s service catalogue pairs each energy-transfer type with a BPT variant: AC and AC_BPT, DC and DC_BPT, WPT and WPT_BPT for wireless, plus ACD variants for automated connection devices such as bus pantographs. Selecting DC_BPT or AC_BPT is what makes a session bidirectional.

Figure 2: A DC_BPT session from plug-in to export. ISO 15118-20 messages run between EV and charger, and OCPP 2.1 messages run between charger, CSMS and aggregator.
The sequence starts with a TLS 1.3 session between car and charger. ISO 15118-20 makes TLS mandatory for every identification method, including External Identification Means such as RFID or an app, and requires mutual authentication. Authorization follows, either Plug and Charge with a contract certificate or EIM. The charger forwards the identity to the CSMS with an OCPP Authorize request. Service discovery then reveals DC_BPT, and the car selects it. During charge parameter discovery, the car reports its discharge limits alongside its charge limits.
Where OCPP 2.1 joins the conversation
Once the charger knows the car’s energy needs and departure time, it reports them to the CSMS with NotifyEVChargingNeeds. This is the moment the back office learns how much flexibility the session offers. The CSMS asks the aggregator, or its own optimiser, for a plan. It then sends SetChargingProfile with schedule periods whose operation mode might be ChargingOnly until 17:00, CentralSetpoint at minus 7 kW from 17:00 to 20:00, and ChargingOnly again overnight. The charger translates that profile into ISO 15118-20 terms during ScheduleExchange.
In the energy loop, the charger sends target power to the car and the car reports its state. Throughout, OCPP TransactionEvent messages carry meter values. In a bidirectional session you need both imported and exported energy registers. Settlement, tariff credit and grid-service verification all depend on that export register, so a metering design that only records import is a non-starter.
Dynamic versus Scheduled control mode
ISO 15118-20 introduced a Dynamic control mode next to the Scheduled mode inherited from ISO 15118-2. The difference is who guarantees the driver’s mobility needs. In Scheduled mode, the charger offers schedules and the car picks and executes one, so the car remains responsible for reaching its target state of charge by departure. In Dynamic mode, the car hands that responsibility to the off-board system. It declares its needs and limits, and the charging station sets the power in real time.
For V2G, Dynamic mode is usually the better fit. Grid services need the aggregator to move power at short notice, which Scheduled mode resists because the car owns the plan. The cost is liability. In Dynamic mode, if the aggregator over-discharges and the driver leaves with too little range, the off-board side owns the failure. That has contractual consequences, so fleet operators should require a minimum-SoC-at-departure guarantee in the aggregator agreement.
Scheduled mode still has a place. It suits tariff-driven V2H where the car’s own energy manager optimises against a price schedule, and it tolerates back-office outages because the car carries its plan. ISO 15118-20 also lets schedules be renegotiated in a parallel message stream without opening contactors, which ISO 15118-2 renegotiation forced.
Compatibility facts that bite in procurement
ISO 15118-20 is not backward compatible with ISO 15118-2. The messages were replaced, so chargers must keep both stacks during the transition. A charger that supports Plug and Charge for older cars must also coexist with TLS 1.2 for -2 sessions and TLS 1.3 for -20 sessions. The same contract may need storage in both formats.
OCPP 2.1 is the opposite. It extends OCPP 2.0.1, and the OCA states that application logic built for 2.0.1 continues to work. When a 2.1 endpoint talks to a 2.0.1 endpoint, the 2.0.1 schemas are used, and the version is chosen through the WebSocket subprotocol header. OCPP 1.6 is a different application model and cannot carry the 2.1 V2X features. In practice that means 1.6 hardware either gets a new controller board or stays unidirectional.
AC vs DC Vehicle-to-Grid: The Inverter Decides Everything
The single most important architectural choice in bidirectional EV charging is where the inverter lives. That one fact decides who is certified as the DER, who enforces grid codes, which protocols carry those settings, and roughly what the hardware costs.

Figure 3: DC V2G puts the grid-tie inverter inside the charger. AC V2G puts it in the car, and the charger becomes a supervisor that authorises and polices export.
In DC V2G, the charger contains the bidirectional AC-to-DC converter. The car exposes its DC battery bus through the CCS or CHAdeMO connector, and the charger handles the grid. Certification maps cleanly onto existing practice. The charger is tested as a grid-tied inverter against IEEE 1547-2018 using UL 1741 SB, and grid settings stay inside one product. The cost is the hardware. A DC bidirectional unit carries a full power-conversion stage, so it is substantially more expensive than an AC wallbox at the same power. That is why DC V2G dominates fleet depots, bus yards and early residential V2H products such as the Wallbox Quasar 2, which pairs with the Kia EV9 in the US at up to 12 kW according to launch reporting.
In AC V2G, the vehicle’s onboard charger is bidirectional. The wall unit is close to a conventional AC charger with extra supervision logic. Hardware cost at the site falls sharply, which matters for homes and workplaces. But the grid-tie inverter is now a component of a mobile product that may plug into chargers in different utility territories. Someone must make sure it applies the right grid code for this location, and someone must stop it if it misbehaves.
How the US certification path closed the AC gap
SAE J3072 is the vehicle-side answer. It specifies requirements for the onboard bidirectional inverter, requires conformance to IEEE 1547-2018, and defines how the car declares its grid-support capability to the charger. Until 2026 the charger side had no matching category, so AC V2G deployments in the US typically certified a matched car-and-charger pair as a single DER system. UL 9741 and the UL 1741 Certification Requirement Decision provided that route.
UL 1741 Supplement SC, approved in May 2026, breaks that pairing. It certifies the bidirectional AC charger as a supervisor. The charger authorises discharge based on the vehicle’s J3072 conformance, duplicates a small set of IEEE 1547 responses, and opens its contactor on unauthorised export. Those responses include ride-through, voltage and frequency trips, and enter-service behaviour. The interoperability consequence is the point: any certified vehicle should be able to export through any certified charger, rather than only through the charger it was tested with. UL Solutions launched a certification programme for AC bidirectional chargers in September 2026.
Two open issues remain. First, J3072 allows OEM self-certification, while utilities generally prefer third-party lab results; that debate is unresolved. Second, J3072’s communication options have centred on IEEE 2030.5 and SunSpec Modbus. QualityLogic reports that ISO 15118-20 Amendment 1 is prompting a J3072 update to recognise ISO 15118 as a valid transport. Until that lands, an AC V2G charger may need more than one EV-facing protocol.
Europe and the rest of the world
Outside North America, the inverter question maps onto national grid codes such as EN 50549 rather than IEEE 1547. The architecture is the same, but the certification bodies and settings differ. Renault’s AC V2G offer for the new Renault 5 launched commercially in France in October 2024. It pairs the car’s bidirectional AC onboard charger with the Mobilize PowerBox Verso and an energy contract operated by The Mobility House. Germany initially shipped the same car without V2G because of regulatory delays, which shows that regulation, not hardware, often gates go-live.
CHAdeMO remains the most mature DC bidirectional path, with a dedicated V2H DC guideline. The OCA’s 2026 work on combining OCPP with CHAdeMO extends the OCPP 2.x mapping with a V2G sequence and device-model mapping for bidirectional sessions. If you run a mixed fleet with older CHAdeMO vehicles, the CSMS can still manage them through the same OCPP 2.1 V2X block.
IEEE 2030.5, CSIP and the Utility Interface
IEEE 2030.5 is the utility-facing protocol in most North American V2G designs. It began life as Smart Energy Profile 2.0. It is a RESTful protocol over HTTPS with XML or EXI payloads, a mandatory PKI, and a resource model built around function sets. The 2023 revision added function sets for EV charging and refined DER monitoring.
CSIP narrows IEEE 2030.5 for California Rule 21. It selects a testable subset of function sets and constrains how they are used, without adding capabilities. Since mid-2020, DER interconnecting with PG&E, SCE and SDG&E under Rule 21 has been expected to communicate via IEEE 2030.5 using CSIP. SunSpec Alliance runs the conformance certification. Hawaii’s Rule 14H takes a similar approach, and Australia adapted the profile as CSIP-AUS with dynamic operating envelopes.
The DERControl model in V2G terms
The core of CSIP is a hierarchy: a DERProgram contains time-bound DERControl events and a DefaultDERControl. An event might say “limit export to 3 kW between 17:00 and 21:00”. The default applies when no event is active. Crucially, if communication is lost, the client must revert to the most recently received default settings.
For an aggregator managing vehicles, that model maps neatly onto OCPP 2.1. A DERControl export limit becomes a discharge limit in the charging profile for every affected station. The DefaultDERControl becomes the fallback profile the CSMS installs on each charger. If you do one thing from this section, make the charger’s local default match the utility’s DefaultDERControl, so that the failsafe behaviour is consistent at every layer.
Where OCPP stops and IEEE 2030.5 starts
A recurring architecture mistake is to treat OCPP as the grid interface. IEEE 1547-2018 recognises IEEE 2030.5, IEEE 1815 (DNP3) and SunSpec Modbus as DER communication protocols. OCPP is not on that list. The OCA’s own material on OCPP 2.1 DER control draws the chain as utility, CSMS, charging station and EV. It uses IEC 61850 or IEEE 2030.5 upstream, OCPP 2.1 in the middle and ISO 15118-20 to the car.
OCPP 2.1 does carry the payload a charger needs to behave as a DER. Its DER control block lets the CSMS configure static settings such as fixed power factor setpoints when absorbing or injecting active power, a frequency-watt mode, enter-service-after-trip and ramp rates. The charger can report DER alarms such as over- and under-frequency or current imbalance. That makes OCPP the delivery mechanism for grid settings to a fleet of chargers, while IEEE 2030.5 remains the utility’s contract-grade interface to the aggregator. The two layers complement each other; neither replaces the other.
Plug and Charge PKI: The Security Backbone of V2G
V2G raises the stakes of EV charging security. A compromised unidirectional charger can steal energy or deny service. A compromised bidirectional fleet can inject or withdraw megawatts on command, which is a grid-stability problem. The V2G PKI is therefore not optional plumbing.
ISO 15118 defines several certificate roles. The V2G root CA anchors trust. CPO sub-CAs issue charging-station certificates, mobility operator (MO) sub-CAs issue contract certificates, and OEM provisioning certificates let a car request its contract certificate securely. Hubject, which operates a widely used V2G root, launched an ISO 15118-20-based service in June 2022 using ECDSA on the secp521r1 curve. ISO 15118-20 also adds explicit management of multiple contract certificates, so one car can hold contracts from more than one mobility operator.
OCPP 2.1 closes gaps on the charger side. It supports installing OEM root certificates, longer certificate chains, certificate revocation lists, and multiple contract certificates. Those features matter because a charger in a multi-root world must trust several ecosystems at once. The CSIP PKI is separate again: a three-tier hierarchy with a SunSpec root, a manufacturer CA and a manufacturer-installed CA. Codibly notes that CSIP device certificates have indefinite validity, which simplifies deployment but makes revocation and rotation hard.
The practical lesson is that a V2G site may run three trust domains: ISO 15118 for the car, OCPP security profiles for the back office, and IEEE 2030.5 for the utility. Each has its own lifecycle. Our guide to ACME and short-lived certificate automation covers the automation patterns for the parts you control. The handshake cost of TLS 1.3 on constrained devices is worth reading if your charger controller is an embedded MCU.
Worked Economics: Energy, Losses and Battery Wear
Every V2G business case reduces to one inequality. For each kilowatt-hour you cycle, the revenue spread must exceed the conversion losses plus the battery wear you cause. The numbers below are illustrative, not quotes. Substitute your own tariffs, pack costs and measured efficiencies.
Illustrative residential or light-fleet case
Take a car with a 75 kWh usable pack and an 11 kW bidirectional charger. The driver needs 60% state of charge at 07:30 and plugs in at 18:00 with 80%. The flexible window above the reserve is 20% of 75 kWh, which is 15 kWh. Exporting that 15 kWh during a three-hour evening peak averages 5 kW, well inside the charger rating.
Now the wear and loss side, again illustrative. Assume a replacement cost of $150 per kWh for the pack and a cycle life of 3,000 equivalent full cycles to end of life. Lifetime throughput is 75 × 3,000 = 225,000 kWh, and the pack costs 75 × 150 = $11,250. The marginal wear cost is therefore about $0.05 per kWh cycled. Assume 90% one-way conversion efficiency in each direction, giving roughly 81% round-trip. To deliver 15 kWh to the grid, you draw about 16.7 kWh from the battery. Recharging those 16.7 kWh costs about 18.5 kWh from the grid.
With an illustrative off-peak price of $0.12 per kWh and an export value of $0.40 per kWh, the session earns 15 × 0.40 = $6.00. Recharging costs 18.5 × 0.12 ≈ $2.22, and wear on 16.7 kWh is about $0.84. Net value is roughly $2.94 per event. Across 150 qualifying evenings a year, that is about $440. Halve the spread and the value falls below $100, before any aggregator fee. The arithmetic shows why residential V2G pencils out only where export value is high: capacity payments, critical-peak events, or avoided backup-generator costs.
Why degradation is not a single number
The $0.05 per kWh figure is a linear simplification. Real lithium-ion ageing depends on depth of discharge, average state of charge, temperature, C-rate and calendar time. Some studies found extra cycling harmful, for example Dubarry and colleagues’ 2017 analysis of bidirectional charging impact in the Journal of Power Sources. Others have argued that smart V2G which keeps cells away from high state of charge can partly offset calendar ageing. Both can be true. A shallow evening discharge from 80% to 60% on a temperate day costs far less than a deep discharge in summer heat.
For a fleet, that argues for a degradation model per vehicle rather than a flat cost. It also argues for using the vehicle’s own state-of-health estimate. Our guides to state-of-charge estimation with coulomb counting and extended Kalman filters and BMS architecture for SoC and SoH estimation explain why the SoC the car reports carries an error bar. Your reserve floor should include that error.
School buses: the fleet case that works
School buses are the strongest early fleet case because their duty cycle is predictable. They sit idle for long midday and evening periods and for the entire summer, which coincides with peak grid stress in many regions. Zum’s Oakland Unified deployment put 74 electric school buses with bidirectional chargers into service for the 2024-2025 school year, with PG&E as utility partner. Zum projects about 2.1 GWh returned to the grid annually. That is a company projection, not a measured result. Spread across 74 buses, it implies roughly 28 MWh per bus per year, or about 78 kWh per bus per day if averaged over a full year. A school bus pack is large, so that is plausible mainly because summer availability is high.
The fleet lesson generalises. V2G value scales with hours plugged in, predictability of departure, and pack size relative to daily driving. Delivery vans that return to depot at 18:00 and leave at 06:00 score well. Ride-hail vehicles score badly.
A Python Sketch of a V2G Dispatch Scheduler
The scheduler sits in the aggregator or CSMS. For each plugged-in session it decides, per interval, whether to charge, discharge or hold. The sketch below is deliberately simple. It is a greedy price-spread planner with a reserve floor, a wear cost, charger and site limits, and a guaranteed departure SoC. Production systems use linear or mixed-integer programming across the whole fleet, but the constraints are the same.

Figure 4: Dispatch logic for one session. The reserve floor and departure target are hard constraints, the spread-versus-wear test is economic, and the failsafe default applies when communications fail.
from dataclasses import dataclass
@dataclass
class Session:
capacity_kwh: float # usable pack energy
soc: float # current SoC, 0..1
soc_departure: float # required SoC at departure, 0..1
soc_error: float # BMS estimation margin, e.g. 0.03
p_charge_kw: float # max charge power at this EVSE
p_discharge_kw: float # max discharge power (positive number)
wear_usd_per_kwh: float # marginal degradation cost per kWh cycled
eta: float = 0.9 # one-way conversion efficiency (illustrative)
def plan(session, prices_import, prices_export, dt_h, site_export_cap_kw):
"""Return per-interval setpoints in kW (+ charge, - discharge)."""
n = len(prices_import)
floor = min(1.0, session.soc_departure + session.soc_error)
energy = session.soc * session.capacity_kwh
setpoints = [0.0] * n
# 1. Discharge in the best export intervals while the spread beats wear + losses.
cheapest_recharge = min(prices_import) / (session.eta ** 2)
order = sorted(range(n), key=lambda i: prices_export[i], reverse=True)
for i in order:
margin = prices_export[i] - cheapest_recharge - session.wear_usd_per_kwh
if margin <= 0:
break
p = min(session.p_discharge_kw, site_export_cap_kw)
e_out = p * dt_h / session.eta # energy drawn from the pack
if energy - e_out < floor * session.capacity_kwh:
continue
setpoints[i] = -p
energy -= e_out
# 2. Charge in the cheapest intervals until departure SoC is met.
target = floor * session.capacity_kwh
for i in sorted(range(n), key=lambda i: prices_import[i]):
if energy >= target:
break
if setpoints[i] < 0:
continue
p = session.p_charge_kw
setpoints[i] = p
energy += p * dt_h * session.eta
# 3. Feasibility check: if the departure target is still unmet, refuse to export.
if energy < target:
return [session.p_charge_kw if s <= 0 else s for s in setpoints]
return setpoints
The sketch has three deliberate weaknesses. It ignores time order: discharging in hour 2 assumes the car had enough energy then, which a real planner must simulate step by step. It treats wear as linear. And it plans one session at a time, while the site export cap is shared across all sessions. Fixing all three turns the problem into a small linear program. The objective maximises spread minus wear, with SoC dynamics per interval and coupling constraints for the site cap. The structure of Figure 4 survives that upgrade intact.
The step that must never be skipped is the last one. If the departure target cannot be met, the planner abandons export rather than gambling on a later recharge. Drivers forgive a missed grid payment. They do not forgive a bus that cannot complete its route.
Trade-offs, Gotchas, and What Goes Wrong
Interoperability is a test-matrix problem. A protocol logo on the datasheet guarantees little. Different cars implement different subsets of ISO 15118-20, and BPT support in particular is uneven. Budget for plugfest-style testing of every vehicle model against every charger firmware, and retest after each over-the-air update on either side.
Warranty language can kill the business case. Several OEMs restrict or exclude battery warranty coverage for energy exported beyond approved programmes, and the terms vary by model and market. Some programmes cap annual export energy. Read the warranty before signing a capacity contract, and encode any OEM throughput cap as a hard constraint in the scheduler.
Metering and settlement get messy. Export must be measured with a meter the utility or market accepts. A charger meter may not qualify, and a building meter mixes the vehicle with every other load. Dual-register metering and clear baseline methodology are prerequisites, not refinements.
Comms loss behaviour must be consistent. IEEE 2030.5 clients revert to DefaultDERControl. OCPP chargers fall back to their local profile or a default. The car might continue its Scheduled plan. If those three defaults disagree, a WAN outage produces unexpected export. Align them and test by pulling the network cable.
Cybersecurity scales with aggregation. A V2G aggregator that can command thousands of vehicles is a grid-relevant control system. Treat its credentials like utility SCADA credentials: hardware-backed keys, least-privilege CSMS roles, signed firmware on chargers, and rate limits on fleet-wide setpoint changes. OCPP 2.1’s minimum of TLS 1.2 should be configured with modern cipher suites only.
Grid codes are location-specific. An AC V2G car that drives across a utility boundary may need different trip settings at its next charger. That is precisely the problem AC DER services and UL 1741 SC address, but the process for delivering settings to the car is still maturing. Until it does, restrict AC export to sites where the settings are known and verified.
Two optimisers in one building. A site EMS doing peak shaving and an aggregator selling capacity can issue conflicting commands to the same charger. Use OCPP 2.1 ExternalLimits for the EMS and CentralSetpoint for the aggregator, so the EMS can only constrain, never command.
Decision Table: Choosing a V2G Architecture
| Scenario | Recommended path | Key standards | Main risk |
|---|---|---|---|
| Bus or truck depot, predictable dwell | DC V2G, Dynamic control | ISO 15118-20 DC_BPT, OCPP 2.1 CentralSetpoint, IEEE 2030.5 upstream | Departure-SoC liability |
| Home V2H with backup | DC wall unit today, AC as certified pairs appear | ISO 15118-20 or CHAdeMO, UL 1741 SB | Warranty terms, islanding settings |
| Home V2G export in the US | AC V2G with certified car and charger | SAE J3072, UL 1741 SC, IEEE 2030.5 CSIP | Immature ecosystem, settings delivery |
| Workplace or campus | AC charging with site EMS, selective DC V2G | OCPP 2.1 ExternalLimits, LocalLoadBalancing | Two-optimiser conflicts |
| Frequency services | Local frequency response at charger | OCPP 2.1 LocalFrequency, droop curves | Market prequalification, metering |
| Mixed legacy fleet | OCPP 2.1 CSMS managing CCS and CHAdeMO | OCPP 2.1, CHAdeMO V2H guideline | OCPP 1.6 hardware cannot be upgraded |
Practical Recommendations
Start with the inverter question, because it determines your certification path, hardware budget and protocol stack. For fleets and depots in 2026, DC V2G remains the lower-risk option. The charger is the certified DER, settings live in one product, and ISO 15118-20 DC_BPT is the path most vehicles support first. AC V2G is the long-term volume play for homes and workplaces. Now that UL 1741 SC exists, it is worth piloting, but expect a test matrix of vehicle and charger pairs for at least another product cycle.
Next, draw the setpoint-ownership matrix. For each OCPP 2.1 operation mode you plan to use, write down who sets power, who sets limits, and what happens when each party goes silent. Choose ISO 15118-20 Dynamic mode where the aggregator must respond quickly. If you choose it, accept the departure-SoC liability contractually.
Finally, treat the PKI and metering as first-class design work. Map every trust domain, every certificate lifecycle and every meter register before the pilot, not after the first settlement dispute. For a wider view of the vehicle data pipeline that feeds the scheduler, see our IoT electric vehicle architecture for battery telematics.
Checklist before go-live:
- Confirm ISO 15118-20 BPT and OCPP 2.1 support on charger firmware, not just hardware.
- Certify the correct box: UL 1741 SB for DC chargers, SAE J3072 plus UL 1741 SC for AC pairs.
- Align comms-loss defaults across IEEE 2030.5, OCPP and the vehicle.
- Encode departure SoC, BMS error margin and OEM warranty caps as hard scheduler constraints.
- Verify that import and export meter registers meet settlement requirements.
- Run a vehicle-by-charger interoperability matrix and repeat it after every OTA update.
Frequently Asked Questions
What is vehicle-to-grid (V2G) and how is it different from V2H?
Vehicle-to-grid is bidirectional EV charging in which the car exports energy across the utility meter, or provides grid services at the request of a utility, market or aggregator. Vehicle-to-home (V2H) stays behind the meter. The car powers or offsets the household’s own load, often as backup during an outage. V2G therefore needs an interconnection agreement, a grid-code-compliant inverter and a utility-accepted meter, while V2H mainly needs safe islanding and transfer switching. The hardware can be identical; the contracts and certifications differ.
Does ISO 15118-20 support bidirectional charging?
Yes. ISO 15118-20, published in April 2022, adds bidirectional power transfer services: AC_BPT, DC_BPT and WPT_BPT for wireless, plus versions for automated connection devices. It also adds a Dynamic control mode, in which the charging station controls power while the car declares its needs and limits. TLS becomes mandatory for every identification method. ISO 15118-20 is not backward compatible with ISO 15118-2, so chargers must support both during the transition. Amendment 1, published in July 2026, adds AC DER services and the Megawatt Charging System.
What does OCPP 2.1 add for V2X?
OCPP 2.1, released by the Open Charge Alliance in January 2025, adds a bidirectional power transfer functional block and a DER control functional block on top of OCPP 2.0.1. Charging schedule periods gain an operation mode: Idle, ChargingOnly, CentralSetpoint, ExternalSetpoint, ExternalLimits, CentralFrequency, LocalFrequency or LocalLoadBalancing. The DER block configures settings such as power factor, frequency-watt and ramp rates, and lets stations report DER alarms. It stays compatible with OCPP 2.0.1 logic and was later published as IEC 63584-210:2025.
Is AC or DC V2G better?
Neither wins outright. DC V2G puts the inverter in the charger, which simplifies certification and grid-code enforcement but raises charger cost. AC V2G puts the inverter in the car, making the wall unit cheaper. However, the vehicle must comply with SAE J3072 and the charger needs UL 1741 SC certification in the US. In 2026, DC is the safer choice for fleets and depots. AC is the likely long-term route for homes and workplaces as certified vehicle and charger pairs multiply.
Does V2G damage the EV battery?
Extra cycling adds wear, but the size of the effect depends on how the battery is used. Deep discharges, high state of charge, high temperature and high currents accelerate ageing. Shallow evening discharges at moderate temperature cost far less. Some research suggests that smart V2G which avoids parking at high SoC can partly offset calendar ageing. Treat wear as a cost per kilowatt-hour in the scheduler, respect any manufacturer export cap, and read the battery warranty before enrolling a vehicle.
Do I need IEEE 2030.5 if I already use OCPP?
In most North American utility programmes, yes. IEEE 1547-2018 recognises IEEE 2030.5, IEEE 1815 (DNP3) and SunSpec Modbus as DER communication protocols, and OCPP is not among them. California’s Rule 21 relies on IEEE 2030.5 with CSIP. The usual pattern is IEEE 2030.5 between utility and aggregator, OCPP 2.1 between aggregator or CSMS and chargers, and ISO 15118-20 between charger and car. Each protocol covers a different hop and a different liability.
Further Reading
- OCPP 2.0.1 vs OCPP 1.6 vs ISO 15118: EV charging architecture covers the unidirectional baseline that OCPP 2.1 extends.
- Battery management system SoC estimation with coulomb counting, EKF and OCV explains the SoC error bar your reserve floor must absorb.
- Virtual power plant DER aggregation with OpenADR 3, IEEE 2030.5 and IEC 61850 is the aggregator layer for stationary and mobile DER.
- Sodium-ion vs LFP batteries for grid storage gives the stationary-storage comparison point for V2G economics.
- Connected vehicle V2X reference architecture covers the vehicle communications side of the V2X family.
References
- Open Charge Alliance, OCPP 2.1 is now available, 23 January 2025.
- Open Charge Alliance, What’s new in OCPP 2.1, overview of V2X operation modes and DER control.
- CharIN, ISO 15118-20 Amendment 1 published: AC DER service, MCS service and improved security concept, July 2026.
- Trialog, ISO 15118-20: not only a further step towards bidirectional charging.
- QualityLogic, UL 1741 SC: what it takes to certify the bidirectional charger behind AC V2G and Implementing interoperable V2G AC.
- SunSpec Alliance, IEEE 2030.5 / CSIP certification.
- Hubject, Next generation security services based on ISO 15118-20, June 2022.
- Zum, Nation’s first 100% electrified bidirectional V2G school bus fleet in Oakland.
By Riju — about
