A battery system can report state of charge with confidence and still get its grid export wrong by several kilowatts. The BMS is usually fine. The trouble sits on the AC side, in a current sensor at the inverter output or the grid connection that was sized for the wrong range or mounted backwards.
These sensors get far less design attention than the cells. They shouldn’t. Protection trips and export limits both start from their readings.
Where does AC current actually flow in a storage system?
AC current turns up in three places:
-
Charger or rectifier input. A CT here usually feeds input metering and overcurrent protection. The converter’s fast control loops usually have their own shunts or Hall sensors.
-
PCS or inverter output, ahead of any step-up transformer. External CTs serve metering and protection relays; internal control loops typically use Hall sensors.
-
The point of interconnection. A single current transformer (CT) feeds the site meter or energy management controller, and if it reads wrong, export limiting runs on bad data.
Why a CT has no place on the DC bus
If CTs are cheap, passive and isolated, why not put one on the pack output? Because a CT works by induction, and only a changing current induces anything in the secondary. Worse, a DC component pushes the core toward saturation and distorts any AC riding on top.
Inside the pack, shunts and Hall-effect sensors do this job. A shunt, a precision low-value resistor, is accurate and simple but sits at pack potential and needs an isolated amplifier. A Hall sensor reads the magnetic field around the conductor and is isolated by design. The BMS and PCM page on large-battery.com covers how the pack side uses those readings.
On the AC side, that same isolation is the CT’s quiet advantage. Its secondary has no electrical connection to the primary conductor, so the meter or relay works at a much lower, isolated voltage. The secondary circuit is normally grounded at one point, so it cannot float to a coupled voltage and has a fault path if insulation fails. And the insulation rating must cover the conductor: a 600 V part does not belong on a 690 V inverter output, however well the window fits.
CT, Rogowski coil or Hall sensor: which one goes where?
A Rogowski coil is an air-core winding, usually flexible, whose output tracks how fast the current changes; an integrator turns that into a current reading. A closed-loop Hall sensor handles AC and DC alike, and that matters on the AC side too. Grid interconnection rules limit how much DC an inverter may inject, and a conventional CT is blind to it.
|
Sensor |
Measures |
Strengths |
Watch out for |
Typical use |
|
Current transformer |
AC only |
Passive, isolated, well-defined accuracy classes |
DC offset saturates it; open-secondary hazard |
Metering and protection at the grid connection and PCS output |
|
Rogowski coil |
AC only |
No saturation; fits around large busbars |
Needs an integrator; weaker at low currents |
Retrofits, high-current busbars, harmonic measurement |
|
Hall sensor (closed-loop) |
AC and DC |
Measures both, isolated output |
Needs a power supply; offset and temperature drift |
DC bus, converter control loops, DC injection checks |
In practice, many storage systems end up using all three, each in the spot where its weak points matter least.
Four lines on a CT datasheet worth reading twice
Ratio
A 400/5 CT delivers 5 A on the secondary at 400 A primary. How big should the primary be? Closer to real operating current than many first drafts assume. A 250 kW PCS on a 480 V three-phase connection draws about 300 A per phase at full load, so 400/5 fits. A 600/5 part chosen "for margin" sees the same system idling at 30 kW, roughly 36 A, as 6% of its rating.
Accuracy class
That low end is where accuracy classes part ways. Class 0.5 limits ratio error to ±0.5% at rated current but allows about ±1.5% near 5% of rating; class 0.5S holds roughly ±0.75% there. On a site that idles every night, the difference ends up in the energy accounts.
5P10, a protection class, means composite error stays within 5% up to ten times rated current. A metering CT is meant to saturate during a fault, sparing the meter, while a protection CT must stay linear for the relay.
Burden
A tenth of an ohm of wire on a 5 A secondary consumes 2.5 VA by itself. On a 5 VA CT, that is half the rating gone. Easy to miss.
That wire is part of the burden, the secondary circuit’s total load in VA, and accuracy holds only within the rated figure. On a 1 A secondary the same wire costs 0.1 VA, one reason 1 A CTs are common on long runs.
Form factor
Split-core CTs clamp around an existing cable, which makes retrofits far easier, though in many designs they give up a little accuracy. The window is sized for the insulated conductor, not the bare copper.
The open-secondary rule
Never leave the secondary of a conventional CT open while the primary is energized. With no path for current, the core saturates and the secondary terminals can develop a dangerously high voltage, enough to injure a technician or break down insulation. Shorting terminal blocks let a meter be swapped safely.
The CT that was installed backwards
During commissioning, the site meter shows export while the batteries are plainly charging from the grid. Nothing is broken. The CT went on with P1 and P2 reversed, so the meter reads import as export. Export limiting then works against the site.
The fix takes minutes. Finding it can take a day, especially when one CT of three is reversed and the totals look odd rather than wrong. Checking every phase before energizing, with P1 toward the source and S1 and S2 on the matching meter terminals, is the cheapest test on the site.
What samples tell you that a datasheet won’t
Plenty of OEMs buy from a current transformer manufacturer in China, and the datasheets rarely tell suppliers apart. The differences show up in the parts and the paperwork:
-
Samples tested in the buyer’s own setup. Check ratio and phase error at several points, including the low end where the system idles, with the real meter and lead lengths connected. A part that meets its class on the supplier’s bench can fall outside it with a different burden.
-
Test reports for the exact model. Routine test records per batch, plus type tests from an independent lab. A report for a similar model does not automatically cover the delivered part.
-
UL recognition for North American projects, checked in UL’s public database rather than taken from a logo on a PDF.
-
Lead times. Standard ratios often ship from stock; custom windows or ratios can add weeks. Ask for production lead time, not sample lead time.
Working with a current transformer manufacturer in China is routine for many OEMs and integrators, and capable suppliers expect these questions. A vague answer is a signal to pause the order.
Worth the afternoon
Nobody schedules a design review around a CT. Yet the one at the grid connection decides whether the site controller thinks it is importing or exporting, and the one at the charger input decides when protection trips. An afternoon spent on ratio, burden and a bench test of samples costs far less than chasing a nuisance trip on a site three hours away.
FAQ
Can a CT measure current inside the battery pack? No. A CT responds only to changing current, so it cannot read pack DC. Shunts and Hall sensors handle that side.
Is a 1 A or a 5 A secondary better? Neither by default. A 1 A secondary tolerates long lead runs far better; a 5 A secondary suits short runs and many existing meters.
Does a split-core CT give up accuracy? In many designs, slightly. For retrofit metering the trade is usually acceptable, provided the stated accuracy class still fits the application.
