Why kW and kVAR Sharing Is Critical in Genset Synchronization

When multiple gensets operate in parallel, successful synchronization is not only about matching voltage, frequency and phase angle. The real challenge is maintaining stable and balanced kW and kVAR load sharing.

I make that distinction early with almost every customer I work with, because the two things get collapsed into one. Synchronising is a condition that has to be true for a single instant — the instant the breaker closes. Load sharing is a condition that has to stay true for every hour afterwards, through every load step, every set start and stop, and every setpoint that quietly drifts between one service visit and the next. A plant can pass the first test perfectly and fail the second one for years without anybody noticing.

What makes it manageable is that the sharing problem separates cleanly into two independent control loops. Most of the confusion I see on site comes from treating them as one.

On a genset running alone, the governor's job is intuitive: more fuel, more speed. That intuition breaks the moment the set is paralleled onto a bus.

Once synchronised, the machine's speed is locked to bus frequency. The other sets on the bus — and the grid, in a mains-parallel system — hold it there. Adding fuel can no longer accelerate the rotor. The energy has to go somewhere, and where it goes is real power output. The governor has quietly changed jobs, from a speed controller into a kW controller.

That is the whole basis of kW sharing:

• It determines how the active load is distributed between gensets.
• It is mainly controlled through the engine governor and speed control.
• Done properly, it prevents one genset from being overloaded while another remains lightly loaded.
• Done poorly, it leads to reverse power, overload trips and unstable operation.

Two schemes deliver it in practice. Speed droop gives each governor a setpoint that falls slightly — conventionally a few percent from no load to full load — as the machine takes kW. Since every set on the bus has to settle at the same frequency, each one ends up at the load point where its own droop line crosses that frequency. Matched droop settings and matched no-load setpoints give proportional sharing. Mismatched ones do not, and the machine with the higher effective setpoint takes the larger share whether or not that is what anyone intended.

Isochronous load sharing removes the frequency sag altogether. The controllers exchange load information — over analogue load-share lines on older plant, over a CAN link on modern controllers — and adjust fuel so that every set carries its rated proportion while bus frequency stays flat. On most installations it is the better answer, and it depends completely on that communication path being intact, correctly wired and correctly terminated.

Of the consequences of poor kW sharing, reverse power is the one I would single out, because it is the one most often misdiagnosed at site level.

If a set's governor is calling for too little fuel relative to the others, its share of kW falls — and it does not stop at zero. Past zero, the alternator stops generating and starts being driven as a motor by the rest of the bus, dragging its own engine around with it. Reverse power protection, ANSI function 32, exists precisely to catch that condition before the engine suffers for it.

The trip is genuine. The fault, usually, is not in the relay. It is a governor setpoint that has drifted, a droop percentage that does not match its twin, or an unloading ramp in the stop sequence that pushes the set into reverse power before its breaker opens. Investigating the protection instead of the sharing is one of the most reliable ways I know to spend a week and change nothing.

The mirror image is just as damaging and far quieter. The set with the higher effective setpoint absorbs a disproportionate share of every load block, runs hotter, burns more fuel, and accumulates wear that its hour meter alone will never explain. Where nobody has compared the two kW readings under load, this can continue for years. It is the control-side counterpart of a problem GensetPedia has covered on the operating side — having two gensets installed does not mean both should be running.

Reactive power gets far less attention than kW, largely because it does no useful work. It magnetises motors and transformers and moves back and forth without being consumed. But it flows through the same windings and the same cables as real power, and it has to be shared just as deliberately.

The excitation loop mirrors the governor loop almost exactly. On an isolated alternator, more excitation means more terminal voltage. Paralleled, terminal voltage is the bus voltage, and every machine on the bus is holding it. Raising excitation on one machine can no longer lift bus voltage by itself — instead, that machine takes a bigger share of the reactive load. The AVR has changed jobs in the same way the governor did, from a voltage controller into a kVAR controller.

The reactive side, stated the same way:

• It determines how reactive load is distributed between gensets.
• It is mainly controlled through the AVR and excitation system.
• Done properly, it maintains voltage stability and prevents excessive circulating reactive current.
• Done poorly, one generator takes excessive reactive load while another operates at low or negative kVAR.

That last point describes the characteristic failure, and it is worth picturing. One machine is over-excited and heavily lagging. Another is under-excited, possibly leading — importing reactive power from its neighbour rather than supplying any. The reactive current circulating between them serves no load at all. It heats both alternators, consumes capacity that could have gone to real work, and leaves voltage regulation twitchy through load steps. In severe cases the under-excited machine approaches its stability limit, and loss-of-excitation protection, ANSI 40, has something to say about it.

The two schemes here parallel the governor's. Reactive droop compensation — often called voltage droop, or referred to by the paralleling CT it needs — feeds each AVR a signal proportional to its own reactive current, so its voltage setpoint falls slightly as it takes kVAR, exactly as speed droop works for kW. Cross-current compensation interconnects the AVRs so that only the difference in reactive current between machines acts on excitation, sharing kVAR without any droop in bus voltage at all.

For stable parallel operation, this is the relationship I would ask anyone to hold in mind:

Governor → kW Sharing   → Frequency Stability
AVR      → kVAR Sharing → Voltage Stability

Frequency is a real-power phenomenon. Voltage is a reactive-power phenomenon. When a paralleled plant hunts on frequency, the cause sits on the governor side of that table. When it hunts on voltage, it sits on the AVR side. Knowing which column a symptom belongs to eliminates half the possible causes before anyone opens a panel door, and it is the single most useful habit I can pass on to an engineer new to parallel systems.

Six things have to be correct for stable parallel operation, and each of them is a place I have seen commissioning go wrong:

• Correct controller configuration. Rating, CT ratio, droop mode and load-share scheme have to match the machine they are configured on — not the machine the settings were copied from.
• Proper governor response. Sharing that is correct in steady state can still be unstable through a load step if the response is too slow or too aggressive.
• AVR stability and droop settings. The same applies on the reactive side, with the added trap that a poor droop setting can look perfectly fine at no load and diverge only once reactive load appears.
• Accurate CT/PT polarity and sensing. This is the classic one. Droop compensation depends on the sign of the reactive current signal. Reverse that CT and the correction pushes sharing apart instead of pulling it together — the machine already taking too much kVAR takes more. It is a wiring error that presents itself as a control instability, which is why it costs so much time.
• Correct generator ratings and operating parameters. Sets of different ratings must share in proportion to their ratings; scaling that wrongly guarantees one machine reaches its limit first.
• Proper load-sharing communication. Analogue load-share lines and CAN links both need correct wiring, screening and termination. A degraded link does not always fail cleanly — it can share badly rather than not at all, which is the harder failure to spot.

None of these are exotic. They are the things that get assumed rather than verified, particularly when a second set is added years after the first and the commissioning of the original machine is nobody's living memory.

Synchronizing is only the beginning. Stable kW and kVAR sharing is what makes multiple gensets operate safely and reliably in parallel.

A paralleling relay confirms one instant. Load sharing has to hold for every hour after it. So if your site runs sets in parallel, the two most informative numbers on the controller are not on the synchroscope — they are the kW and kVAR readings on each machine, under load, compared against each other. If they do not sit in proportion to the machines' ratings, the plant is already telling you which of the two loops to look at first.

Good engineering here is not really about the sophistication of the controller. It is about verifying the handful of things that decide whether two machines cooperate or quietly fight each other.

Raji Babu Naidu J leads Application Engineering for the MENA region at Kirloskar and has 17+ years of experience in diesel engines, generator sets and industrial power systems across India, the GCC and international markets, including hands-on work with DSE, ComAp, DEIF, Woodward and GAC control systems. This article is contributed to GensetPedia in a personal capacity; the views expressed are the author's own and not those of any employer. Load-sharing, droop and protection settings are specific to your machines, controllers and installation — confirm any change with your genset manufacturer, panel supplier or commissioning engineer before altering a commissioned setup.