How to test EV charging load management when controls fail

How do you prove the site cap survives a control failure?

Test EV charging load management with simultaneous charging and a controlled building-load increase, then interrupt the meter, charger-control link, internet connection and controller separately. Measure current on every protected phase with an independent recorder, and verify the approved limits during the fault, fallback and restart. Accept the system only when its measured response meets the adopted NFPA 70 design basis, the installed equipment's instructions and the project's current, timing and recovery criteria.

Begin after EVSE installation and commissioning. Establish normal all-port operation, add the approved building-load step, inject each fault separately and restore it under continuing demand. Then test a credible shared failure, such as loss of the network switch serving both meter and controller. Measure the protected service or feeder; use charger displays and cloud totals as corroboration.

Transport for NSW's TS 04955.6:1.0, effective January 9, 2025, requires an EV load-management default position that does not exceed the nominated limits. Section 8.1 applies to NSW transport assets. Intertek's October 28, 2026 power control system (PCS) certification overview identifies normal operation, load/generation changes and loss of control as test categories. Field acceptance verifies the installation against its specification; certification evaluates the product. S1, S9

Keep five linked records: the service-limit register, charging-state matrix, communications-failure test, meter latency budget and measured cap acceptance record. Stop the applied load increase at the agreed abort threshold, use the prepared EV shedding control when limits are threatened, and preserve the trace. A protective trip fails the cap test.

Which electrical limits must load management enforce?

Under NFPA 70, 2023 NEC 220.70, the maximum ampere setpoint of a qualifying feeder or service energy management system is treated as a continuous load for load calculations. Section 625.42(A) establishes the managed EV equipment load through qualifying control under 750.30. Leviton's authorized 2023 code reprint provides both provisions; a software setting alone does not establish compliance. S2

Record the adopted edition, amendments and approved design basis; publication does not establish adoption. UL's PCS overview and Minnesota DLI's September 2025 review minutes confirm the 2026 section locations. S3, S4

NFPA 70, 2026 NEC 120.7(B), limits the PCS control setting to 80% of the monitored circuit's overcurrent protective device rating and assigns that setting to qualified persons. Under 120.7(C)(1), the controlled-load calculation uses the control setting when only controlled loads are monitored, or their minimum operating current when controlled and noncontrolled loads are monitored together; (C)(2) calculates noncontrolled loads under Article 120 Parts II through VII. Add the controlled and noncontrolled contributions for the circuit calculation; the 2023 continuous-setpoint method is not the complete 2026 method. S16

What belongs in the service-limit register?

Separate equipment ratings, operating ceilings, feeder allowances, branch settings, contractual import limits and billing targets. A 15-minute billing average cannot establish a seconds-scale phase-current maximum; the tariff determines the actual demand-charge interval. The worked register uses the 2023 calculation basis, with ordinary continuous-load treatment supported by DOE's 2022 training and Tesla's 60 A branch/48 A output example. S15, S5

Mark the sensor boundary on the electrical one-line, including EV loads, building loads and generation. Record CT ratio, direction, phase mapping, device IDs, firmware, controlled outlets and unmetered paths. A compliant net-import reading does not establish every internal feeder's loading.

Retain the electrical load calculation and EV feeder sizing checks for conductors, terminations and protection. The design authority approves the load basis, the authority having jurisdiction (AHJ) accepts the installation, and the utility sets contractual import/export conditions. Restrict rating controls to authorized personnel; a driver's app preference is a different setting.

How do different load-management systems fail?

Fixed sharing divides a charger-group allowance; meter-based dynamic control also responds to building load at the protected boundary. Autonomous shedding disconnects charging under its documented conditions, while cloud scheduling manages transactions or scheduled allocations. Identify which layer actually protects each circuit before choosing the fault tests.

Include disconnected secondaries and the still-connected primary in the fallback total. Compare aggregate fallback with the residual building and feeder budgets, using the installed product's supported group size. S6

Why is internet loss not proof of safe fallback?

OCA's Improving Uptime with OCPP, version 1.1 (2024), describes ongoing transactions continuing offline and configurable offline authorization. Test both an existing session and a new start while the WAN is absent. Local electrical limits must remain effective through either authorization outcome. S8

Follow the installed model, region, firmware and listing; UL's Fall 2024 Question Corner locates PCS functionality in the specific listing, instructions and ratings, including communication-based architectures. Retain evidence that the installed controller, meter, sensors and EVSE form the approved combination; an OCPP connection alone does not establish that approval. S13

How should you meter and prepare the test?

Place an independent recorder at each protected boundary and capture every line current, voltage and total real power. Add synchronized EV-feeder measurement to separate charging from building behavior; record neutral current where the design requires it. The submetering and energy monitoring guide covers measurement placement.

Verify current-transformer (CT) ratio, orientation and voltage-phase pairing with a known load change. Tesla's revision 1.5 note requires the CT to match its voltage reference and specifies source-facing orientation; its solar-off commissioning step makes import direction easier to confirm. S5

Fluke's January 2022 specifications for the 1773/1775/1777 support 1-second trends, one-cycle RMS current extrema refreshed each half-cycle, and 200 ms power extrema. Use the 1-second trend plus retained fast current maxima for this procedure, with the ceiling defined on the agreed root-mean-square (RMS) window. An instantaneous waveform peak is a different quantity. Verify recorder/probe range, clipping, calibration status and combined uncertainty. S12

Demonstrate each port's demand in turn within its branch limit, then challenge the aggregate allowance with all required ports requesting. Keep the approved service and feeder controls active throughout. Record requested demand, offered current and actual draw separately, along with battery charge, vehicle settings, ambient temperature and thermal status. Nameplates and handshake-only simulators do not establish actual charging demand.

How do you inject faults without creating an electrical hazard?

Prepare isolated communication break points and load-rated switching controls. OSHA 29 CFR 1910.333 requires de-energizing exposed live parts before work, subject to its stated exceptions, and lockout/tagout and qualified-person verification of absence of voltage. Install or alter CTs, voltage leads and enclosure connections under the site's electrically safe work procedure; include generation, storage and backfeed in the isolation. S17

OSHA 29 CFR 1910.334 requires qualified testing personnel, inspected instruments/accessories rated for the circuit and environment, and load-rated devices for switching under load. Keep required protective functions in service throughout the test; do not create overloads or pull power conductors to simulate lost data. Janitza's current-transformer guidance prohibits opening a current-output CT secondary while primary current flows because hazardous voltage can result. S18, S19

Set the abort threshold with room for uncertainty and further load rise before shedding completes. Tesla's Dynamic Power Management application note, revision 1.5 dated January 27, 2026, verifies normal control by adding panel loads or temporarily lowering the conductor limit and confirming that the charger maintains it. Where permitted, use that lower-cap approach, record temporary settings and verify final settings after restoration. S5

Which charging states belong in the test matrix?

Test one port, all ports requesting, vehicle taper, insufficient headroom, changed priorities and new sessions. Apply a building-load step on each relevant line; record pre-event, maximum and settled currents.

Anvilfield recommends a 60-second baseline, 120 seconds of settled observation and three repetitions of short load-step/fault transitions for this test plan. Extend observation through the installed timers, including delayed recovery and profile expiry. Keep demand available throughout: cars drawing only 20 A cannot challenge the worked 50 A aggregate allowance.

How do you run the communications-failure test?

Establish high charging demand, capture the baseline, isolate one path and timestamp the event. Observe detection, actual current reduction and sustained fallback; restore the path only after its observation window. Attempt a new session and affected-EVSE restart while the fault persists.

Unchanged readings do not prove stale data. Compare sample age, sequence and a known load change; then test fresh-but-wrong values separately through an OEM-supported interface. A frozen value carrying fresh timestamps needs the documented plausibility or sensor-fault response, because a lost-message timer alone cannot detect it.

During meter loss, apply the approved load step while the controller lacks current information. This exposes the interval between the last valid measurement and actual fallback. Keep protection energized when interrupting controller power, and test the shared communications switch where it creates a common failure. Record the exact scope of each fault: a stopped application, frozen process and lost controller supply are separate cases. Unsupported mandatory sensor tests remain incomplete until an approved interface or accepted equivalent is available.

For profile expiry, record the implemented OCPP version and local behavior. OCA's protocol overview distinguishes version-specific smart charging and the incompatibility of OCPP 1.6 and 2.0.1. Expiration of a transaction profile must not release charging above the local protective cap. S14

How much latency and headroom should the test allow?

Anvilfield estimates 30 A of reserve for the worked single 20 A load step, 2 A measurement allowance and 8 A operating margin. A further 3 A/s rise over the selected 7-second fault interval increases reserve to 51 A. These values follow the disturbance envelope; they are project calculations, not standard response-time allowances.

Add worst-case reading age, transport delay, processing time and charger response to form the normal budget. Verify end-to-end response independently of reporting intervals. These are Anvilfield project targets; installed instructions and the accepted specification govern the job.

Reserve must cover the entire credible rise before charging actually reduces, including the failure timeout. Add the fixed step to rise rate multiplied by the longest uncorrected interval, then add uncertainty and margin: reserve ≥ step + rise rate × response time + uncertainty + margin.

The growing-load case requires a target no higher than 109 A under the 160 A ceiling. Leaving the target at 130 A produces an upper comparison of 173 A, which fails the design. Budget persistent fallback separately: every port's fallback plus maximum non-EV current, uncertainty and margin must fit each protected conductor for the whole outage.

The worked design uses a strict RMS ceiling. Where an actual listed system and approved design instead specify a magnitude/duration envelope, record that envelope before testing. A breaker trip curve does not authorize arbitrary cap excursions, and suspending EV charging still leaves the building load to be checked.

Worked example: four chargers on a shared service

The job has a 200 A split-phase service and main OCPD, a selected 160 A operating ceiling, a 130 A normal target and a 125 A EV feeder with a 100 A continuous allowance. Four 48 A ports serve 240 V loads; building load plus EVSE standby is 80 A on L1 and 65 A on L2. Use in-phase unity-power-factor loads and no generation.

Subtract each building-line current from the target, then take the smallest result or the feeder allowance, bounded below by zero: EV allowance = max(0, min(feeder allowance, target − building L1, target − building L2)). Independent conductor measurements govern acceptance with reactive or distorted loads; subtracting RMS magnitudes is not a general method for separating building and EV current.

Method: Anvilfield estimates, dated November 5, 2026, use the stated loads, the 2023 continuous-load basis in S2 and conversion at 240 V and unity power factor. The 30 A reserve bounds all load rise during both normal response and fault timeout; use the larger verified measurement uncertainty if it exceeds the selected 2 A allowance. Reducing the normal target another 10 A reduces available charging by 2.4 kW.

The single step fits the reserve; verify current and timing against the field record below, retaining standby consumption after charging stops.

What changes on three-phase, DCFC, solar or battery sites?

Preserve each charger's phase allocation and each line-to-line port's phase pair. On mixed 120/208 V or 230/400 V sites, test actual charging modes and phase rotation; a balanced kW total can conceal an overloaded phase. The three-phase power guide explains why a scalar balanced-load conversion cannot replace these measurements.

At DC fast charging sites, measure AC input, including conversion losses and auxiliaries. Dispenser DC output does not equal service current. Include sudden loss of solar or battery-storage support: Anvilfield estimates that losing 24 A of support adds the same import step on the affected conductor at unchanged load and voltage. Intertek identifies generation-step testing separately. S9

Apply the same reasoning to installed bidirectional EV export. Check internal feeder and bus boundaries as well as utility import, and give each approved generator or islanded operating mode its own limit register.

What should happen when power or communication returns?

Require valid meter data, correct phase mapping and restored local control before increasing charging. Restore communications with all connected vehicles requesting power, then record whether allocation ramps, queues, staggers or remains stopped under the installed OEM sequence. Verify stored caps through controller reboot, EVSE restart and backend reconnection.

DCC-10's V16 manual distinguishes automatic restoration of station supply after total power loss from capacity-monitoring recovery after load shedding, which also requires the control input to be closed. Verify available capacity with charging included before reconnection. Honor the documented recovery timer. S10

Schneider's EVlink Pro AC FAQ000243727, visibly updated June 23, 2026, distinguishes restart without authentication from authenticated sessions: outages shorter than 30 minutes permit restart, while longer outages require new authentication. Anvilfield proposes separate 5-minute and 35-minute cases for this model; record authorization and current allocation separately. Obtain the manufacturer's rule for an exact 30-minute interruption if that boundary affects the job. S11

Coordinate alternate-source events with automatic transfer switch commissioning. The utility-mode charging setting must not silently replace the approved limit for a smaller alternate source.

Is the controller limiting current, or is something else wrong?

Compare requested current, offered current, actual draw and vehicle acceptance. When offered current is high and draw is low, establish whether vehicle scheduling, charge level or thermal limits explain the difference. Low demand can conceal ineffective control.

Record environmental conditions with the result. DCC-10 V16 specifies operation from -30°C to 45°C, or -22°F to 113°F; use the installed manufacturer's range for other equipment. After correction, restore demand and repeat the cap challenge. S10

What measurements make the site pass or fail?

Anvilfield's strict-ceiling acceptance rule passes a boundary only when its recorded maximum plus verified uncertainty stays within the limit and every required timing, fault and recovery case passes. A recorded maximum above the limit fails; a nominally compliant reading whose uncertainty extends above it is incomplete. At a 160 A ceiling with 2 A uncertainty, 157 A passes the current comparison, 159 A is incomplete and 161 A fails.

Apply the same upper-bound rule to time. Start at the injected event, retain the last valid sample time separately, and stop at actual current reduction into the documented regulation band. An alarm or issued command does not end the response. For the worked step, the new EV allocation is 30 A; define its regulation band independently of the hard service and feeder ceilings. For a selected 30 A upper boundary, verify sustained draw at or below that allocation under continuing demand.

Missing regulation-band criteria, insufficient demand, absent fast maxima or an unperformed mandatory fault leave the affected case incomplete. Justify not-applicable cases by architecture, such as no WAN or generation.

Field acceptance checklist

• Approve edition, limits, monitored loads, controlled ports and model/firmware application.
• Confirm fault criteria, regulation bands, response endpoints and recovery timers.
• Calculate load-rise and aggregate fallback budgets; reduce the target when needed.
• Mark measurement boundaries; verify CT ratio, direction, phase mapping and known-load response.
• Save settings; verify range, RMS window, fast maxima, clock alignment and uncertainty.
• Prepare isolated fault interfaces, abort threshold and EV shedding control.
• Demonstrate sufficient demand; run normal states, phase-specific steps and applicable faults.
• Test new sessions and EVSE restart during faults; repeat short transitions three times.
• Verify recovery, restore settings and remove temporary profiles/network blocks.
• Classify every case, retain evidence, correct failures and repeat affected tests.

Retain raw CSV/event files, recorder and probe IDs, calibration, clock alignment, model/manual/firmware versions, fault times, starting load, per-phase maxima, fallback output, elapsed response, recovery peak, alarms and final settings. Identify the design, installation and operational acceptance roles. OSHA 29 CFR 1910.334(b)(2) requires determining that the circuit can be safely energized before manual re-energization after a protective trip. S18 After failure, preserve the evidence and restrict charging to an independently justified fixed allowance or suspend it; correct the cause and repeat affected normal, fault and recovery cases before returning to managed operation.

How long does testing take, and when is retesting due?

Anvilfield estimates a 6-hour site session with two technicians, plus 3 hours of preparation and 4 hours of analysis/reporting: 19 labor-hours for the defined four-port package. Accessible metering, available vehicles and prepared local fault interfaces support that allowance.

Method: Anvilfield estimates sum the task hours above, with site duration multiplied by crew size, dated November 5, 2026. Add time for simulator integration, long profile-expiry periods, manufacturer recovery timers, additional feeders, unavailable vehicles and repeat tests after repairs. Budget recorder hire, specialty DC load equipment, travel and outage coordination separately.

Anvilfield recommends annual functional retesting as an owner-program starting point, plus targeted tests after changed caps, CTs, control firmware, topology, port count, generation/storage or an unexplained excursion. This is an owner-program recommendation; OEM instructions and contract intervals supply additional requirements.

Workplace, multifamily and fleet parking programs need repeatable acceptance records for commissioning and expansions. Contractors building a recurring EV test territory use [PorchRocket Ground Control parking-lot lists](/lists/parking-lots) to identify properties with visible parking areas for outreach.

EV charging load-management testing FAQs

How do you verify EV charging load-management service limits and failure-mode acceptance?

Measure every protected phase during normal charging, load steps, faults and restart, then compare current and response time with the approved limit register.

What happens to EV power sharing when communication is lost?

The installed model determines fallback. Check the disconnected ports and surviving primary together against the whole feeder allowance; Table B gives model-specific behavior.

How do you verify a charger site load cap?

Challenge the cap with sufficient simultaneous charging demand and record independent service and EV-feeder currents through load changes and failures.

Does EVSE load-management commissioning require real charging load?

A current-cap test needs vehicles or approved test loads drawing enough power to challenge the cap; a handshake-only tester cannot prove it.

Will charging stop if the internet goes down?

Not necessarily; OCA's 2024 uptime paper describes active transactions continuing offline, so local electrical limits must remain effective.

Is a 15-minute demand reading fast enough for a failure test?

A 15-minute average cannot show a seconds-scale excursion; use suitably fast current recording and retained maxima.

Is 80 percent always the correct site cap?

The 2026 NEC 120.7(B) rule applies to the PCS control setting when used for load calculations. Record the adopted edition, monitored circuit and equipment instructions; Table A uses a selected 2023 continuous-load basis.

Can a 6 A fallback still exceed the feeder budget?

Yes; eight 6 A fallback loads on the same conductor total 48 A before other loads are added.

Can chargers restart immediately after an outage?

DCC-10 distinguishes total-power restoration from delayed load-shedding recovery. EVlink Pro AC also conditions session restart on authentication and outage length; test allocation separately from authorization.

How often should EV load-management failure tests be repeated?

Anvilfield recommends annual functional retesting plus targeted tests after changes affecting sensing, control, load or limits, with OEM and contract intervals added.

How much headroom is needed while a load-management controller is unresponsive?

Anvilfield estimates 30 A for the worked single 20 A step, 2 A uncertainty and 8 A margin; a further 3 A/s rise over the 7 s fault budget increases required reserve to 51 A.

Is a reading below the cap enough to pass an EV charging fault test?

Pass only when the recorded maximum plus verified uncertainty fits the cap and all required timing, fault and recovery cases pass; low demand or a borderline reading leaves the affected case incomplete.

Sources, review date and citation

Reviewed November 5, 2026. Sources checked: 18 primary agency, manufacturer, protocol and certification publications, plus the supplemental code-text preview in S16.

1. S1: Transport for NSW, TS 04955.6:1.0, effective 2025-01-09
2. S2: Leviton, 2023 Captain Code NEC Code Changes
3. S3: UL Solutions, Power Control Systems: Advancing Electrification for New and Existing Infrastructure
4. S4: Minnesota DLI, NEC 2026 Adoption Review Committee minutes, 2025-09-25, Attachment A items 9, 11 and 88.
5. S5: Tesla, Wall Connector Application Note: Dynamic Power Management, revision 1.5, 2026-01-27.
6. S6: Wallbox, Static Load Management Activation and User Guide, updated 2024-08-16.
7. S7: Wallbox, Dynamic Load Management in Single-charger Installations, updated 2024-08-16.
8. S8: Open Charge Alliance, Improving Uptime with OCPP, v1.1, copyright 2024
9. S9: Intertek, Preparing for UL 3141 and ULC/ORD-C3141 Certification, 2026-10-28.
10. S10: DCC Electric/RVE/Thermolec, DCC-10 installation manual, Gen 3 V16; drawing dated January 2026.
11. S11: Schneider Electric, FAQ000243727, published 2022-09-20, visible last modified 2026-06-23.
12. S12: Fluke, 1773/1775/1777 Three-Phase Power Quality Analyzer Product Specifications, January 2022 (English)
13. S13: UL Solutions, Fall 2024 UL Question Corner.
14. S14: Open Charge Alliance, Open Charge Point Protocol overview
15. S15: US Department of Energy, FEMP, Electric Vehicle Supply Equipment Infrastructure, Federal Fleet Training, 2022-06-28.
16. S16: NFPA 70, 2026 NEC 120.7, public code-text preview hosted by Electrical License Renewal
17. S17: OSHA, 29 CFR 1910.333, electrical work practices, checked 2026-11-05.
18. S18: OSHA, 29 CFR 1910.334, switching and test-equipment requirements, checked 2026-11-05.
19. S19: Janitza, Operation of current transformers, checked 2026-11-05.

S16 reproduces primary code text on a supplemental host; it is not an NFPA-hosted edition. Tesla revision 1.5 has an inconsistent opening 32 A/60 A example; this guide uses its consistent 60 A/48 A table and test example.

Update log: November 5, 2026, clarified code application, fault timing, safety, demand, standby measurement, field records and test duration.

How to cite this page: Anvilfield, How to test EV charging load management when controls fail, 2026, https://anvilfield.com/field-guides/electrical/ev-charging-load-management-failure-tests/

Related tools

FAQ

People also ask

• What is the difference between Level 1, Level 2, and DC fast charging?
• What is an electrical load calculation?
• What size wire for a 48A charger?
• What is a one-line diagram?
• What is electrical submetering?