Generator synchronisation and paralleling
What this page is for: understanding what your multi-set system is doing, reading the alarms it raises correctly, and knowing what to describe when you call someone.
This is an explanation, not a procedure — and it must not be used as one
Paralleling switchgear brings together the two most dangerous things on a generating installation: stored rotational energy and switchable fault current. Getting a closure wrong is measured in wrecked couplings, destroyed breakers and injured people. Commissioning, adjusting or fault-finding on a paralleling system is work for a competent engineer with the right instruments and the manufacturer's data, on a system that has been made safe.
Which one is yours?
Why sets are run in parallel at all
Almost every multi-set installation in the UAE exists for one of four reasons, and knowing which tells you a good deal about how it will behave.
Capacity beyond one machine
A load larger than any single set you would want to own, buy or move. Several medium sets rather than one very large one.
Redundancy — N+1
A hospital, a data centre or a critical process needs the power to survive a generator failure, not only a mains failure. One spare set's worth of capacity, so any one machine can be down without losing cover.
Maintenance without an outage
With a spare set on the bus, any machine can be taken off, serviced, tested and returned while the site stays covered. This is why paralleling and a maintenance regime tend to arrive together.
Loading efficiency
A site whose load varies widely runs sets on and off the bus to keep the running machines properly loaded, rather than running one large set at a small fraction of its rating and suffering everything light-load running does to a diesel.
Closing onto a live bus
Four conditions, plus one that catches people out
Two alternating supplies can only be connected when they match. Every synchronising system in existence exists to establish these.
- Phase sequence — the same rotationThe order in which the three phases peak must be identical. Fixed by the wiring, not the controls — a commissioning check, proved once when the system is built or after any work that could have changed it. Closing onto a bus with reversed rotation is the worst of all failure modes.
- Voltage magnitude — the same voltageWithin a small tolerance. A mismatch produces a surge of reactive current between the machines at closure, and a step in bus voltage.
- Frequency — the same speedWithin a small tolerance, conventionally with the incoming machine running very slightly fast, so on closing it takes up a little load rather than being motored by the bus.
- Phase angle — the same point in the cycleThe waveforms must be aligned, or nearly so, when the contacts make. The condition that changes fastest, and the one that decides whether a closure is smooth or violent.
- The breaker's own closing timeA breaker takes a finite time to close after being commanded, so a synchroniser issues the command slightly early, timed so the contacts meet at the aligned instant. One never set up for the breaker it operates produces closures consistently out by the same margin — a controller setting, not a machine fault.
Why nobody should experiment with this
What happens when a set closes out of synchronism
The system pulls the machines into step, instantly. The energy comes from the rotating masses of the engines and rotors, and arrives as a mechanical shock through couplings, crankshafts, alternator shafts and foundations, with a very large current transient through the windings, cables and switchgear.
Results range from a bang, a lurch and a tripped breaker through to sheared coupling bolts, damaged crankshafts and bearings, distorted windings, and switchgear that must be replaced rather than reset. The damage is frequently not visible, which is worse — a set that appears fine may have taken a torsional load its coupling will remember.
If a set has closed out of synchronism, it is not a set to reset and carry on with. Stop, and have it inspected: the engine coupling, the alternator bearings and windings, and the switchgear that made the closure.
- The check synchronising relay
An independent device in the breaker's close circuit that physically prevents a close command reaching the breaker unless the four conditions are within its settings — a second opinion, wired in series. It is why a synchroniser failure does not automatically become a wrecked machine.
- Defeated check sync is a serious finding
Bypassed, wrongly set or removed during someone's troubleshooting is among the more serious things we find on installed switchgear. It should be function-tested periodically, with settings matching the installation rather than the factory default.
- Dead bus closing is a separate case
The first set onto a bus that is not live has nothing to synchronise with, so it closes on different logic — ensuring only one machine ever takes a dead bus, and none takes a bus that is live but undetected. A persistent "bus not live" alarm is usually voltage sensing, not the generators.
Two independent problems
Sharing the load
Once several sets are on a bus they run at exactly the same frequency and voltage by definition — they are electrically locked together. So sharing is not about speed and voltage as such. It is about how much of the total each machine contributes, and it splits in two.
Real power (kW) — the engines' job
Governors, droop and isochronous
- Droop: a set's speed reference falls slightly as load rises, so droop machines settle at a shared frequency, each contributing to its own characteristic. Simple, stable, no communication needed — but system frequency varies with total load
- Isochronous: constant frequency at all loads. One machine alone is straightforward; several will fight over which sets the frequency unless tied by a load sharing system — dedicated sharing lines or a data link between controllers
- Isochronous load sharing is what most modern installations use, and it depends entirely on that link working
- The classic failure — one set hogging the load while another sits near zero, or one motored while another is overloaded — is very often a broken or wrongly terminated sharing line, or a controller with the wrong parameters after a replacement. A control system diagnosis, not an engine fault
Reactive power (kVAr) — the alternators' job
AVRs, quadrature droop and cross-current
- Quite separate from kW sharing. Two sets can share kW perfectly and still fight over kVAr, one alternator heavily over-excited and running hot while the other is under-excited
- Quadrature droop: each AVR reduces its voltage reference slightly as it takes reactive load, so the machines settle into a stable split — as speed droop does for kW
- Cross-current compensation: the AVRs are linked so reactive load is shared deliberately
- The symptom is distinctive and frequently misread: circulating current between the sets, one alternator noticeably hotter, and current readings that do not match the kW readings. A set can be at its thermal limit on a site whose total load looks modest — see voltage problems
A generator has stopped generating and is being driven as a motor by the other machines on the bus — its engine no longer producing enough power to carry even itself. It happens when a set loses fuel, when a governor fails or drifts low, when a set is unloaded too fast during a shutdown sequence, or when load sharing has gone badly wrong. It is dangerous mechanically, so every paralleling system carries reverse power protection that trips the set off the bus. The trip is therefore usually a symptom of a fuel, governor or load sharing fault, and resetting it without finding that fault repeats the event. It is also the alarm most often reported to us as "one of the sets keeps tripping off for no reason".
A different order of undertaking
Paralleling with the mains
- It requires the electricity authority's agreement. Connecting generation in parallel with the network — continuously, for peak lopping, or for the few seconds of a closed-transition return — is subject to that authority's requirements and the protection they specify. A formal process, not a controller setting.
- The protection requirement is different. Parallel operation requires protection against islanding — where the network supply is lost but the generator keeps energising a section of it — along with the voltage, frequency and rate-of-change protection the authority stipulates. It protects the network and the people working on it, and it is not a settings menu to be adjusted locally.
- Open versus closed transition. Most standby installations use open transition: the load leaves one source before joining the other, so the supplies are never connected and no synchronising with the network occurs, at the cost of a brief interruption. Closed transition synchronises with the mains, closes both for a very short overlap and then opens the mains, so the load never sees a break — genuinely paralleling with the network. Which one you have determines whether the return to mains involves synchronising at all — see AMF and ATS panel repair.
Reported versus actual
The faults we are actually called to
| What is reported | What it usually turns out to be |
|---|---|
| Set runs up but never closes; "failed to close" | Synchroniser cannot bring the conditions into the window, breaker close circuit or auxiliary contacts, or the check sync relay correctly refusing |
| "Bus not live" or dead bus alarm that will not clear | Bus voltage sensing — fuses, links, transformers, wiring — rather than the generators |
| One set takes all the load, another sits near zero | Load sharing line broken or wrongly terminated, or governor parameters wrong after a controller replacement |
| Sets share kW correctly but one runs hot | Reactive sharing — quadrature droop or cross-current compensation misconfigured, or an AVR fault |
| Reverse power trips on one set | Fuel restriction, governor drift, or a shutdown sequence unloading too fast |
| Frequency drifts with total site load | Droop mode operating as designed. Whether that is acceptable is a design question, not a fault |
| Load acceptance poor when a set joins the bus | Ramp settings, load acceptance capability, or a set that cannot make its rating — see under-frequency shutdown |
| System has never been tested end to end since commissioning | The commonest finding of all, and the reason the first real mains failure is where problems appear |
The honest description
What we do on multi-set installations
We work on the machines and the controls that sit on them: engine and governor faults that show up as sharing or reverse power problems, alternator and AVR faults that show up as reactive sharing problems, controller and panel work, sensing and wiring faults, and the servicing and overhaul of the individual sets — with the heavy engine machining done in our own workshop rather than sub-contracted, so a set can come off a bus and go back on without a long wait for components.
We also load test individual sets and whole systems, which on a paralleling installation is the only way to find out what actually happens rather than what the design intended. Load bank testing puts real, stepped load onto the bus and records how it is shared, how frequency and voltage behave through each step, and whether every machine carries its share up to rating.
Where an installation needs switchgear design, utility approval or protection settings determined by the electricity authority, that is a different discipline and we will say so rather than take the work. What we will not do — and what nobody should — is adjust protection settings on live paralleling switchgear to make an alarm go away.
What to have ready when you call
How many sets, their ratings, and how many are normally on the bus. The controller make and model on each set and on the synchronising panel. Which set raises the alarm, and whether it is always the same one. The exact alarm text, and a photograph of the screen and event log on every set, not only the one complaining — the log on the healthy machines usually holds the answer. Whether the sets share kW, kVAr, both or neither. Whether the system is droop or isochronous, if you know. And what changed — a controller replaced, a set serviced, a machine added, a cable disturbed, a contractor on site.
Common questions
Paralleling, answered
Why will my generator not synchronise with the bus?
One of the four conditions is not being met inside the synchroniser's window. On a system that used to work, the usual causes are voltage or speed control that has drifted, a sensing fault so the synchroniser cannot see one side, breaker auxiliary contacts or a close circuit fault, or a setting changed during other work. A check sync relay refusing the close is also doing its job — the fault is upstream of it.
What is the difference between droop and isochronous control?
In droop, the speed reference falls slightly as load rises, so frequency varies with total load and machines share naturally without communicating. In isochronous, frequency is constant at all loads, which needs a load sharing link so the sets do not fight over which one sets the frequency. Most modern installations run isochronous with load sharing; droop is simpler and more tolerant of a broken link.
Why is one generator taking all the load while the other takes none?
Almost always the load sharing system rather than an engine fault: a broken sharing line, a controller configured differently after a replacement, or governor parameters that do not match. If each engine is fine alone and they misbehave together, the fault is in the sharing.
What causes a reverse power trip?
The set has stopped producing power and is being motored by the others. Behind it is normally a fuel restriction, a governor fault or drift, or a load sharing problem that has driven that machine's output down. The trip is protection working correctly, and resetting it without finding the cause repeats the event.
Can I add a second generator in parallel with my existing one?
Sometimes, but it is a design question rather than a purchase. Ratings, alternator characteristics, governor and AVR types, controller compatibility, switchgear, protection and earthing all have to work together, and two sets that work well alone are not automatically parallelable. Have it assessed at design stage — see installation and commissioning.
Do you set up and commission paralleling systems?
We work on the sets and the controls — engines, alternators, governors, AVRs, controllers, sensing and panel faults — and we load test systems to prove what they actually do. Where a job requires switchgear design, protection settings determined by the electricity authority, or utility approval for parallel operation, that sits with switchgear and design specialists and we say so rather than take it on.
Talk to us about a multi-set installation
Send the number of sets, their ratings, the controller types and what the system is feeding. On paralleling faults the event logs from all the sets, read side by side, are frequently the whole diagnosis.