System and Method for Distribution Transformer Loading and Health Estimation Using Electric Vehicle Charging Session Voltage Droop Profiling with Cross-Charger Impedance Tomography

Abstract

Disclosed is a system and method that turns every connected home electric vehicle supply equipment (EVSE) unit into a grid sensor. Each charging session begins with a commanded current ramp, typically 0 to 48 A at 240 V for Level 2 charging, which acts as a calibrated load step. The EVSE already measures line voltage and current for billing and safety; the voltage droop observed across the ramp yields the source impedance seen from the charger terminals via a least-squares fit of voltage against current. Because the service-drop impedance of a home is static, changes in the measured impedance over weeks and months isolate drift in the shared upstream path: the distribution transformer and its connections. A fusion service combines timestamped droop profiles from multiple chargers served by the same transformer, solves a cross-charger linear model that separates per-home service drops from the common transformer impedance (impedance tomography), and thereby estimates real-time transformer loading, cumulative thermal aging, and connection health with no utility metering hardware. Statistical change detection flags winding degradation, loose connections, and high-impedance neutrals before failure; split-phase leg asymmetry specifically identifies neutral faults that are a known fire precursor. When estimated loading exceeds safe limits, participating chargers coordinate derating to protect the transformer. The system gives utilities a crowdsourced, zero-hardware-cost view of the exact assets EPRI identifies as the weak point of residential EV charging: the neighborhood service transformer.

Technical Field

This invention relates to electric distribution grid monitoring and electric vehicle charging infrastructure, specifically to estimating distribution transformer loading, thermal aging, and connection health from voltage droop profiles measured by electric vehicle charging equipment during charging sessions, fusing measurements across multiple chargers sharing a transformer, and coordinating protective charging behavior in response.

Background

Residential distribution transformers are the binding constraint on home EV charging growth. A typical US neighborhood transformer is rated 25 to 50 kVA and serves 5 to 10 homes; a single Level 2 charger draws 7.2 to 11.5 kW, a substantial fraction of that rating. EPRI's investigation of EV charging impacts concluded that the first failures appear not on feeders or at substations but at the service transformers stepping voltage down to customer level, and built the HotSpotter tool to help utilities identify the most vulnerable units (EPRI Journal). NREL applied transformer thermal life models to vehicle charging loads and found that four plug-in hybrids charging on a 25-kVA transformer already at 80% peak load produce a 10 to 50% service life decrement, while noting that concentrated neighborhood purchasing patterns create localized overload scenarios leading to tripped thermal relays and outages (NREL 48827; NREL 52006). Subsequent studies confirmed the severity: uncontrolled charging above 7.2 kW can reduce transformer life below its expected lifespan (Energies 2022, 15, 4454), and at just 20% EV penetration with long-range 75 kWh batteries, transformer load exceeds rated capacity in both winter and summer (Energies 2023, 16, 4810).

Utilities today have almost no visibility into these assets. Service transformers generally carry no telemetry; loading is estimated from billing-interval smart meter data at best, and thermal aging is modeled rather than measured. HotSpotter and similar tools simulate vulnerability from adoption forecasts and nameplate data. They do not measure actual loading or detect degradation in the field.

The sensing opportunity sits in millions of garages. Connected EVSE units already sample line voltage and current continuously: open-source and commercial chargers expose charging voltage, current, power, and session energy over local APIs and integrations (OpenEVSE Home Assistant integration), and SAE J1772 lets the EVSE command the vehicle's current draw in fine steps via the pilot signal. Every charging session therefore contains a free, calibrated impedance measurement that nobody is using: the current ramp at session start is a known excitation, and the resulting voltage droop encodes the impedance of everything upstream of the charger. The gap in the art is a complete system that extracts source impedance from routine charging sessions, fuses measurements across co-located chargers to isolate the shared transformer impedance from private service drops, converts the result into loading and thermal-aging estimates, detects connection and neutral degradation from impedance drift and leg asymmetry, and closes the loop with coordinated protective derating. No existing system does this.

Detailed Description

1. Session Droop Capture and Impedance Extraction

At the start of each charging session, the EVSE commands a current ramp through the J1772 pilot signal, typically stepping from 0 to the session target (up to 48 A at 240 V split-phase) over 10 to 60 seconds. During the ramp and the first minutes of steady charging, the EVSE records per-leg line voltage (V_L1, V_L2) and current at 1 to 10 Hz using its existing metering hardware, which in typical connected chargers resolves voltage to 0.1 to 0.5 V and current to 0.1 A.

For each leg, the system fits the linear model V = V0 - Z·I by ordinary least squares over the ramp steps, where V0 is the unloaded voltage and Z is the source impedance seen from the charger terminals on that leg. A 48 A ramp across a typical residential source impedance of 0.05 to 0.3 ohms per leg produces 2.4 to 14 V of droop, far above the metering noise floor; averaging over the ramp's 10 to 60 samples yields impedance estimates with repeatability in the low milliohm range under stable grid conditions. The fit residual also serves as a data-quality gate: sessions with residual variance above threshold (e.g., large concurrent appliance switching during the ramp) are down-weighted rather than trusted. Each session thus produces a compact impedance report: Z_L1, Z_L2, V0, timestamp, ambient temperature from the charger's onboard sensor, and session energy.

2. Service-Drop Calibration and Drift Isolation

The measured impedance decomposes as Z_measured = Z_drop + Z_xfmr + Z_upstream, where Z_drop is the home's service-drop and panel impedance, Z_xfmr is the transformer impedance referred to the secondary, and Z_upstream captures the primary feeder. Z_drop is static for a given installation: conductor gauge and length do not change. The system establishes it at commissioning from the median of the first 30 days of session reports, optionally refined by installer-entered conductor length and gauge. All subsequent analysis operates on the drift component ΔZ(t) = Z_measured(t) - Z_drop, which by construction reflects only the shared upstream path. A slow upward drift in ΔZ indicates transformer winding or connection aging; a step change indicates a new high-resistance connection, a failing tap, or storm damage; both are invisible to the homeowner and the utility today.

3. Cross-Charger Impedance Tomography

A single charger's drift component still mixes transformer impedance with upstream feeder impedance. The fusion service resolves this by combining reports from multiple chargers served by the same transformer. Cluster membership comes from utility interconnection records where available (opt-in), or from unsupervised clustering: chargers whose droop residuals correlate strongly during coincident sessions share upstream impedance and are grouped automatically.

Within a cluster of N chargers with synchronized 1-minute (ΔV_i, ΔI_i) streams, the service fits the linear model ΔV_i = Z_drop,i·ΔI_i + Z_shared·Σ_j ΔI_j + ε_i via recursive least squares over many sessions. Each charger's own current acts on its private drop impedance; the sum of all chargers' currents acts on the shared transformer-plus-upstream impedance. With enough sessions spanning varied coincidence patterns (chargers starting at different times), the regression separates the private and shared terms. The shared term Z_shared, tracked over time, is the transformer's impedance signature. Dividing the observed shared voltage drop by Z_shared yields the total transformer current, and adding a baseline estimate for non-EV load (derived from the nighttime droop floor when no charger is active) gives total transformer loading L(t) as a fraction of nameplate kVA. This is impedance tomography: reconstructing the shared asset's state from edge measurements, with no sensor on the transformer itself.

4. Thermal Aging Estimation

The loading estimate feeds a thermal model following IEEE Std C57.91-2011 Clause 7: top-oil rise and hot-spot rise computed from per-unit loading with the standard exponents, using ONAN time constants parameterized by transformer size class and nameplate data where known (defaulting to conservative class values where not). Ambient temperature comes from the chargers' onboard sensors, cross-checked against weather data. The model integrates the Arrhenius aging acceleration factor over time to estimate cumulative loss-of-life. Unlike HotSpotter-style simulation, which forecasts vulnerability from adoption scenarios, this measures the actual thermal history the transformer experienced. The service reports per-transformer aging rate relative to nameplate expectation and flags units aging faster than 1.5x nominal for utility inspection.

5. Degradation and Fault Detection

Three detectors run on the impedance time series:

Drift detector. An EWMA control chart on Z_shared with limits set from the commissioning-period variance. Sustained drift beyond 3 sigma over a 30-day window indicates winding or connection degradation and generates a utility maintenance ticket with the estimated impedance change in milliohms.

Step detector. A CUSUM test on ΔZ(t) catches abrupt impedance steps from loose or corroded connections, failing splices, or storm damage, distinguishing them from drift by their rise time (under 7 days). Steps above 20 milliohms trigger an expedited inspection flag, since high-resistance connections dissipate I²R heat at the fault point.

Neutral fault detector. For split-phase services, the leg asymmetry metric A = |Z_L1 - Z_L2| / (Z_L1 + Z_L2) is tracked per charger. Real installations have some baseline leg imbalance from unequal conductor runs, so each charger learns its own healthy A baseline during commissioning rather than assuming zero; a deteriorating neutral raises the impedance of both legs asymmetrically as return current finds alternate paths, and produces the leg-to-leg voltage imbalance that destroys 120 V appliances and starts fires. Because the detector uses impedance rather than voltage alone, it catches high-impedance neutrals during the droop transient before steady-state voltage imbalance becomes visible. Confirmed neutral degradation (A exceeding 0.15 with rising trend across two or more sessions) triggers an immediate homeowner alert and utility emergency flag, since open-neutral conditions are an established fire and shock hazard.

6. Active Impedance Spectroscopy via Pilot Dither

Passive ramp data yields impedance at essentially DC ramp rates. For finer diagnostics, the EVSE periodically superimposes a small pseudo-random binary sequence (PRBS) current dither of ±2 A around the session setpoint, at 0.05 to 0.5 Hz, during steady charging. The dither is invisible to the driver (under 500 W variation on a 11.5 kW session, zero net energy change) and stays within the J1772 pilot resolution. Cross-correlating the measured voltage against the known dither sequence extracts the complex impedance Z(f) at the dither frequencies. A rising resistive component with flat frequency response indicates connection degradation; a rising reactive component indicates winding or core changes. This is electrochemical-impedance-spectroscopy thinking applied to grid assets, using the EV as the signal injector and the charger as the analyzer, at zero hardware cost.

7. Coordinated Protective Derating

When the fusion service estimates that a transformer's loading will exceed 100% of nameplate during coincident sessions, it issues coordinated derate commands to participating chargers on that transformer, reducing pilot-advertised current to spread load across the night (valley filling). The coordination respects driver departure-time constraints entered in the charging app: vehicles needing a full charge by morning keep priority, and any homeowner can override derating at any time. This closed loop converts the monitoring system into a protection system, extending transformer life without utility intervention. Utilities receive a dashboard ranking transformers by a composite risk score (loading × aging acceleration × impedance drift × neutral asymmetry), giving them a measured, prioritized upgrade list: a crowdsourced HotSpotter built from field data instead of forecasts.

8. Privacy-Preserving Architecture

Impedance reports contain only electrical features (Z_L1, Z_L2, V0, temperature), timestamps, and a coarse geohash (precision 6, roughly 1.2 km × 0.6 km cells). No vehicle identification number, account identifier, or street address leaves the charger. Cluster membership uses transformer IDs, not household identities, and no cluster statistics are published for groups of fewer than 3 chargers (k-anonymity). Raw voltage and current time series never leave the EVSE; only the fitted impedance parameters are transmitted.

9. Figures Description

• Figure 1: System architecture: home EVSE units on the same distribution transformer recording session droop profiles, impedance reports flowing to the fusion service, and loading/aging/fault outputs delivered to the utility dashboard and homeowner apps.
• Figure 2: Representative session droop plot: per-leg voltage versus current during the commanded ramp, the least-squares fit line, and the extracted source impedance, with a down-weighted session showing appliance-switching noise in the residuals.
• Figure 3: Cross-charger tomography diagram: three homes on one transformer, private service-drop impedances versus the shared transformer impedance, and the recursive-least-squares separation of the two from coincident session data.
• Figure 4: Long-term impedance trend for one transformer cluster showing gradual drift (aging), an abrupt step (loose connection repair event), and a neutral-asymmetry excursion, with detector thresholds marked.
• Figure 5: Coordinated derating timeline: coincident session currents before and after valley-filling derate, estimated transformer loading against nameplate, and the resulting hot-spot temperature reduction.

Claims

1. A system for monitoring an electric distribution transformer, comprising: a plurality of electric vehicle supply equipment (EVSE) units each configured to record line voltage and current during a charging session's commanded current ramp; an impedance extractor that fits voltage against current over the ramp to estimate source impedance per session; and a fusion service that aggregates impedance reports from EVSE units served by the same transformer and estimates transformer loading therefrom.
2. The system of claim 1, wherein the impedance extractor fits a per-leg linear model V = V0 - Z·I by least squares over ramp samples for each split-phase leg, and gates report quality on fit residual variance to reject sessions corrupted by concurrent appliance switching.
3. The system of claim 1, further comprising service-drop calibration that establishes each installation's static service-drop impedance from an initial baseline period and subtracts it from subsequent measurements so that observed impedance drift reflects only the shared upstream path including the transformer.
4. The system of claim 1, wherein the fusion service fits a cross-charger linear model relating each charger's observed voltage droop to its own current through a private impedance term plus the sum of cluster chargers' currents through a shared impedance term, separating per-home service drops from common transformer impedance by recursive least squares over multiple sessions.
5. The system of claim 4, wherein the fusion service divides observed shared voltage drop by the estimated shared impedance to infer total transformer current and combines it with a non-EV baseline load estimate derived from no-charging droop floors to produce transformer loading as a fraction of nameplate rating.
6. The system of claim 1, further comprising a thermal aging estimator that applies a transformer thermal model to the estimated loading and ambient temperature to integrate an aging acceleration factor over time, reporting cumulative loss-of-life relative to nameplate expectation without utility-owned telemetry.
7. The system of claim 1, further comprising a drift detector applying statistical process control to the shared impedance time series to flag winding or connection degradation, and a step-change detector distinguishing abrupt high-resistance connection faults from gradual aging by rise time.
8. The system of claim 1, further comprising a neutral fault detector that tracks split-phase leg impedance asymmetry A = |Z_L1 - Z_L2| / (Z_L1 + Z_L2) per charger and generates a homeowner and utility alert when asymmetry exceeds a threshold with rising trend, detecting high-impedance neutrals before steady-state voltage imbalance appears.
9. The system of claim 1, further comprising active impedance spectroscopy in which the EVSE superimposes a pseudo-random current dither within the charging pilot signal during steady charging and cross-correlates measured voltage against the known dither sequence to extract frequency-dependent complex impedance, distinguishing resistive connection degradation from reactive winding changes.
10. The system of claim 1, further comprising coordinated protective derating in which the fusion service, upon estimating transformer loading above a nameplate-derived threshold during coincident sessions, issues pilot current reduction commands to participating chargers subject to driver departure-time constraints and homeowner override.
11. The system of claim 1, further comprising unsupervised transformer cluster discovery that groups EVSE units by correlation of droop residuals during coincident sessions when utility interconnection records are unavailable.
12. A method for monitoring an electric distribution transformer without transformer-mounted sensors, comprising: recording, at each of a plurality of EVSE units, line voltage and current during the commanded current ramp of charging sessions; estimating per-session source impedance by least-squares fitting voltage against current; calibrating out each installation's static service-drop impedance; fusing impedance reports across EVSE units sharing a transformer with a cross-charger linear model separating private and shared impedance; estimating transformer loading from the shared impedance term; integrating thermal aging from the loading estimate; detecting degradation, step faults, and neutral asymmetry from the impedance time series; and coordinating charge derating when estimated loading exceeds safe limits.

Implementation Notes

Deployable as a firmware update to connected EVSE units plus a cloud fusion service; no new hardware at the charger, the home, or the transformer. Metering requirements: per-leg voltage resolution 0.5 V or better and current resolution 0.1 A or better at 1 Hz or faster during ramps, satisfied by the energy-monitoring hardware already present in typical connected chargers. Compute per session is trivial (least squares over tens of samples); the fusion service's recursive least squares scales linearly in cluster size, and clusters are small (2 to 12 chargers per transformer).

Calibration accuracy: the 30-day commissioning median establishes Z_drop with repeatability limited by grid voltage regulation noise, typically ±5 milliohms; seasonal temperature dependence of conductor resistance (copper ~0.393%/°C) is compensated using the charger's ambient temperature sensor. Upstream feeder impedance changes (feeder reconfiguration, capacitor bank switching) appear as common-mode steps across all clusters on a feeder and are distinguished from transformer-local events by their spatial correlation footprint.

Known limitations: single-charger transformers cannot benefit from cross-charger tomography and rely on single-device drift detection only; the loading estimate inherits uncertainty from the non-EV baseline estimator during periods of heavy non-EV coincident load (e.g., electric heat waves), bounded by reporting loading as a confidence interval rather than a point estimate; and pilot dither is unavailable during sessions where the vehicle ignores small pilot changes, in which case the system falls back to passive ramp data.

Prior Art References

1. EPRI Journal, "Spotting the 'Hotspots'": Service transformers identified as the weak point for residential EV charging; HotSpotter simulation tool for vulnerable-transformer identification
2. NREL 48827, "Application of Distribution Transformer Thermal Life Models to Electrified Vehicle Charging Loads Using Monte-Carlo Method": Four PHEVs on a 25-kVA transformer at 80% peak load produce 10 to 50% service life decrement; concentrated purchasing creates localized overload
3. NREL 52006, "Tool Helps Utilities Assess Readiness for Electric Vehicle Charging": Framework combining vehicle statistics with utility load data to predict transformer aging rates
4. Energies 2022, 15, 4454, "Novel PEV Charging Approaches for Extending Transformer Life": Uncontrolled charging above 7.2 kW can reduce transformer life below expected lifespan; variable-rate strategies compared
5. Energies 2023, 16, 4810, "Impact on Distribution Transformer Life Using Electric Vehicles with Long-Range Battery Capacity": 20% EV penetration with 75 kWh batteries exceeds transformer rated capacity in winter and summer
6. OpenEVSE Home Assistant integration: Connected EVSE exposing charging voltage, current, power, energy, and temperature telemetry over local API
7. IEEE Std C57.91-2011: Guide for Loading Mineral-Oil-Immersed Transformers and Step-Voltage Regulators; Clause 7 thermal model for top-oil and hot-spot temperature rise
8. SAE J1772: Electric vehicle conductive charge coupler; pilot signal current advertisement enabling commanded current ramps and modulation