Power quality is the other half of the connection

A connection agreement says what a site may export. Power quality decides what it actually manages to. Voltage, power factor and harmonic content are why a healthy array can produce less than a clear sky allows.

Most conversations about connecting an array stop at one number: how much the network will allow the site to export. It is the right question and it is not the whole question.

Three quantities sit behind that. Voltage, which has to stay inside a band at the point of supply. Power factor, which describes how much of the current delivered is doing useful work. And harmonic content, which describes how far the current waveform departs from a clean sine.

A connection is an agreement about behaviour as well as quantity. The equipment has to hold the voltage inside limits, respond in defined ways when the network moves, and not degrade the waveform for everyone else on the same street.

Those requirements are not optional extras and they are not negotiated site by site. They are written into the inverter standard, and compliant equipment implements them whether anybody on site is aware of it or not¹.

The reason to understand them is practical. Several of them reduce output, legitimately and by design, and a site that does not know they exist will read the reduction as a fault.

That single misreading is behind many of the performance questions raised on commercial arrays, and it is avoidable with one conversation at commissioning.

To export, an inverter has to push its terminal voltage slightly above the voltage of the network it is feeding. Current flows from higher to lower, and that difference is the mechanism.

The consequence is that the voltage at the inverter is always a little higher than at the point of supply, and it rises further the harder the array works. Add a network that is already sitting near the top of its allowed band on a sunny afternoon and the headroom disappears.

The inverter then does what the standard requires. It reduces its own output to hold the voltage down, and if the voltage continues to rise it disconnects¹.

Every step above is the equipment working correctly. None of it is a defect, and none of it is recoverable by replacing the inverter with a different brand.

What is recoverable is the part of the voltage rise that belongs to the installation rather than to the network, and that is mostly cabling.

Two of the response modes do most of the work, and the names are more forbidding than the behaviour.

Volt-var response adjusts reactive power. As voltage rises the inverter absorbs reactive power, which pulls the local voltage down without giving up any real output. It is the gentle intervention and it happens first.

Volt-watt response adjusts real power. As voltage keeps rising the inverter reduces what it is actually generating, which works reliably and costs kilowatt hours.

There is a hierarchy in that, and it is deliberate. Reactive power response is set to act before real power is reduced, so the first line of defence costs the owner nothing in output.

Both are required of compliant equipment, and both are configured with settings that the installer applies at commissioning to suit the network the site sits on¹.

This is why the settings belong in the handover pack. They describe the commissioned state, and a year later they are what a performance question gets compared against.

A site that cannot produce its own inverter settings cannot tell the difference between an inverter behaving as configured and an inverter misconfigured, which is the first question any investigation asks.

It is worth noting what these modes are for. They exist so that a street full of inverters can share a network without each one tripping the others offline, and they work. The cost is that individual sites occasionally produce less than their own irradiance would allow.

Curtailment has a signature, and it is distinctive enough to recognise without instruments.

It appears as a flat top on the generation curve on the clearest days, at the sunniest part of the day, when the array should be at its most productive. Cloud produces a ragged profile. Curtailment produces a ceiling.

The distinction matters because the two have opposite remedies. A ragged profile on an overcast week is weather and there is nothing to fix. A flat ceiling on a clear day is a design or a network condition and there usually is.

It is also often worse on weekends and holidays, because the site itself is consuming less, so more of the generation is being pushed out into a network that is already lightly loaded and sitting high.

A second signature worth knowing is per-inverter variation. Where one inverter on a multi-inverter site curtails and the others do not, the difference is usually its cable run rather than its hardware.

None of this requires expensive diagnostics. It requires generation data at interval resolution, which is the same monitoring a performance warranty claim needs anyway.

Of the voltage rise between the point of supply and the inverter terminals, the share created by the installation's own cabling is the share the design can do something about.

It is a function of conductor size, run length and current. A long run in an undersized conductor produces a larger rise at full output, which brings the inverter closer to the point where it starts backing off.

The arithmetic is ordinary electrical design, and it is done before anything is ordered. What varies is whether the calculation was run against the full output of the array or against a nominal figure.

That makes cable sizing a generation decision, not only a compliance one. Sizing to the minimum acceptable conductor costs less on the day and costs output on every clear afternoon for the life of the system.

Inverter placement is the same decision from the other direction. Putting the inverter near the switchboard rather than near the array shortens the alternating current run, which is the run that matters for this.

Where an existing array curtails, this is the first thing to measure and the most likely thing to be fixable. Where an array is being designed, it is a line in the specification.

The network's own voltage is the other half of the sum, and it is outside the site's control. Where it is persistently high, that is a conversation with the network rather than with the installer².

The other half of power quality has nothing to do with the array. It is what the site's own loads do to the supply.

Modern commercial sites are full of switching electronics: variable speed drives, electronic lighting drivers, uninterruptible supplies, inverters and vehicle chargers. Each draws current in pulses rather than smoothly, and the sum of those pulses is harmonic distortion on the board.

Two consequences follow. Transformers and neutral conductors run hotter than the real load would suggest, and capacitors installed for power factor correction can be stressed by harmonic currents they were not chosen for.

That is why power factor correction on a site with significant electronic load is a measured exercise rather than a catalogue selection. The right answer may be an active unit or a filtered one, and it depends on what the measurement finds.

It also connects directly to the invoice. Where the site is billed on kilovolt amperes, poor power factor is being paid for directly: see "Demand charges are a different bill".

There is a design consequence too. Where a site is adding vehicle charging or a new set of drives alongside an array, the power quality question belongs in the same piece of work rather than in a later one, because both are connecting to the same board.

The whole of this subject reduces to a handful of requests, all of which are ordinary and none of which are expensive.

At design, ask what alternating current cable size was selected and what voltage rise it was calculated to produce at full output. Ask where the inverters will sit and why.

At commissioning, ask for the inverter settings as applied, including the response mode configuration and any export limit, recorded in the handover pack rather than described verbally.

Ask for generation monitoring at interval resolution, per inverter where the site has more than one. It is the only way curtailment is ever identified.

Ask, as well, what the network's measured voltage was at the point of supply during commissioning. It is the baseline every later voltage question is compared against, and it takes a minute to record on the day.

And where the site has significant electronic load, ask for a power quality measurement before any correction equipment is specified, not after.

Doing those five turns the most opaque part of a commercial system into something that can be checked. Which is the point: a connection that behaves as designed is a connection somebody can prove is behaving as designed.