DC and Bidirectional Measurement in Energy Meters: where the architecture inherited from AC breaks down

For fifty years, the architecture of an energy meter worked on an implicit assumption: alternating current, 50 Hz frequency, unidirectional flow from the grid to the load. The current transformer was the natural sensor, the voltage zero crossing was the natural synchronization reference, and the full-cycle integration window was the natural way to calculate active power. MID Annex MI-003 was written around that architecture.

An OEM developing a DC fast-charging station, a battery energy storage system, a V2G interface or a hybrid inverter for residential photovoltaics today is working in a different physical domain. The operating frequency is zero. Energy flow changes direction dozens of times per day. The transients to be considered are on the microsecond scale. In many cases, the regulatory framework for legal metrology does not yet exist with the same clarity as MI-003.

Adapting a unidirectional AC meter to these applications is not an incremental operation. These are architectural decisions that affect the sensor, ADC, calculation window, energy accumulators, metrological sealing and communication. Skipping any one of them means designing a meter that works in the laboratory and fails in the field.

The four quadrants

In a unidirectional AC system, active power has only one sign: positive, from the grid to the load. Reactive power has two signs, inductive or capacitive, but traditional billing almost always considers only the first quadrant.

In a bidirectional system, active energy can flow in both directions. A photovoltaic system exports energy to the grid, P negative according to the load convention, when it produces, and imports when it consumes. A storage system charges and discharges. An electric vehicle in V2G does both on the same day.

The meter architecture must therefore manage four operating quadrants (+P/+Q, +P/-Q, -P/+Q, -P/-Q), with separate energy accumulators for import and export. This is not a software issue: the metrological accumulators must be sealed independently, because import and export billing follow different tariff regimes and, in the case of storage systems, may involve different commercial parties: utility, aggregator and prosumer.

The sign convention must be documented and immutable after certification. Reversing the sign via firmware on a meter already deployed in the field would invalidate metrological traceability.

Why direct current is a different domain

AC power measurement exploits a property that does not exist in DC: the full-cycle integration window. At 50 Hz, every 20 milliseconds the system closes a natural window and calculates P = (1/T) integral v(t)*i(t) dt over a complete oscillation. Zero-mean noise components cancel out. Instantaneous power becomes active power without ambiguity.

In DC, the window is not defined by physics. It must be defined by the design. The choice of averaging window (10 ms, 100 ms, 1 s) is a trade-off between responsiveness and noise immunity. A window that is too short produces unstable readings; a window that is too long masks transients that, in applications such as DC fast charging or battery state of charge, are part of the useful information.

The second difference concerns the current sensor. As noted in the previous article, current transformers and Rogowski coils do not work in DC: they operate on flux variation. In DC, only shunts survive, with their galvanic isolation and thermal dissipation issues, and Hall-effect sensors, especially closed-loop and fluxgate variants when the required accuracy exceeds 1%.

The third difference concerns isolation. In AC, galvanic isolation comes for free with the CT. In DC, isolation requires an active component: an isolation amplifier, an isolated digital converter, or a fluxgate. Each of these introduces its own offset error, which in DC is not cancelled by full-cycle integration.

Offset error at low currents

A storage system spends most of its operating life in standby or close to its equilibrium point. A DC charging station, between sessions, operates at almost zero current. The same applies to a vehicle connected to a wallbox while waiting for a charging trigger.

In these conditions, the offset error of the measurement front end, typically a fraction of a milliampere for a shunt with amplifier, but up to milliamperes for a low-cost open-loop Hall sensor, becomes the dominant error. On a 600 A nominal system, a 10 mA offset is invisible at full load (0.0017%); on the same system in standby at a real 50 mA, the same offset becomes a 20% error, integrated over weeks of operation.

For legal metrology applications, this translates into two design requirements. The first is an explicit starting current threshold, below which the energy accumulator does not increment. The second is a periodic auto-zero procedure, which requires either a momentary interruption of measurement or a second reference channel.

A meter that does not explicitly specify these two parameters is not qualified for applications where the duty cycle is biased toward low currents.

Synchronization and timestamping

In AC, synchronization between two meters on the same network is implicit: they share the same grid frequency and the same zero crossing. In DC, each meter runs on its own clock. For billing, hourly settlement or flexibility aggregation applications, timestamping becomes a primary metrological function.

The required time accuracy scales with the application. For a domestic user with hourly tariffs, +/-1 second is sufficient. For a 15-minute flexibility market, +/-100 ms is the operational limit. For V2G applications with minute-level resolution, +/-50 ms. For power quality or event measurements, transients included, +/-1 ms with reference to a common master.

The meter’s internal clock must therefore be disciplined, via NTP, via PTP (IEEE 1588) for higher accuracy, or via GPS signal in outdoor installations. The drift of the base clock between one synchronization and the next must remain within the required accuracy window. A standard quartz oscillator at +/-20 ppm drifts by 1.7 seconds per day: for hourly settlement, it must be synchronized at least once per day; for V2G applications, every few hours.

An OEM integrating the meter must specify in the requirements the synchronization source, frequency, and meter behavior in the event of loss of the sync signal.

What this means for an OEM

A specification for a meter intended for a DC or bidirectional application cannot stop at “MID certified, accuracy class 1”. The six technical questions from which the evaluation of a serious supplier should start are:

• Which sensing technology is recommended for the expected current profile (AC, DC, AC+DC, operating range, transient content), and which alternatives have been rejected?
• How is offset error at low currents managed, and what starting current is declared in both flow directions?
• Under which regulatory framework is the product certified for legal metrology (MID MI-003, MI-011, specific national schemes)? Is the cryptographic signature of the measurement on-board?
• How are energy accumulators organized for import and export? Are they separate and independently sealed?
• What is the averaging window for active power calculation in DC, and is it configurable? What is the latency for real-time telemetry applications?
• How is the internal clock disciplined, what drift is guaranteed, and which timestamping standard is supported?

A supplier that answers these questions with the same structure, technical choice, rationale, evaluated alternatives and operating constraints has done the engineering work. A supplier that pushes every question into the application firmware (“we add it custom”) is selling an AC meter reused in a domain for which it was not designed.

In DC and bidirectional applications, the meter architecture is not an implementation detail. It is what determines whether the final OEM product will enter regular billing across Europe or not.