One generator is a straightforward thing to specify. Two or more feeding a common bus is a different exercise, and it is where behind-the-meter mining projects most often discover that the electrical scope was underestimated.

Why sites end up with multiple gensets

Gas engines come in discrete sizes. A site whose available fuel supports a load that falls between two standard units ends up with either a machine running at part load, which is inefficient and hard on the engine, or two smaller machines sharing the work.

Multiple units also buy resilience. A single engine taken out for a top-end overhaul stops the whole site. Two engines mean the site runs at reduced capacity during maintenance rather than stopping, which over a multi-year contract is usually worth more than the capital difference.

And gas supply declines. A site specified today for a volume the well will not sustain in two years is better served by units that can be taken offline one at a time as flow falls, rather than one large machine that then runs permanently at part load.

What paralleling actually requires

Connecting two generators to the same bus is not a wiring exercise. Before a breaker closes, four conditions have to match between the incoming machine and the live bus.

• Voltage. The incoming machine has to be at the same voltage as the bus.
• Frequency. Close enough that the two will pull into step rather than fight.
• Phase angle. The waveforms must line up. Closing out of phase is the failure that damages equipment, sometimes severely and sometimes immediately.
• Phase rotation. The sequence has to match. This is set during installation and is the one most likely to be wrong on a hurried commissioning.

This is what synchronising gear does, and it is why paralleling is a controls problem rather than a cabling one. The synchroniser watches the incoming machine, adjusts speed and excitation until the conditions are met, and closes the breaker inside the brief window where they are.

Sharing the load after the breaker closes

Getting machines onto the bus is only the first half. Once they are there, they have to share load in a controlled way, and there are two common approaches.

Droop lets each machine’s speed fall slightly as load rises, which causes units to settle naturally into proportional sharing. It is simple, it needs no communication between controllers, and it is robust. The trade-off is that bus frequency varies a little with load.

Isochronous control holds frequency constant and shares load through communication between the controllers. It gives a steadier bus, at the cost of depending on that communication link working.

For a mining load the practical consideration is that the load is large, constant and switched in big steps. A fleet coming online is not a gentle ramp. Whichever scheme is used has to handle a step change without one machine taking the whole transient.

What goes wrong

• Out-of-phase closure. The expensive one. Usually a commissioning error or a failed synchroniser, and the damage is mechanical as well as electrical.
• Reverse power. One machine ends up being driven by the others as a motor instead of generating. Reverse power protection exists for this and should be tested rather than assumed.
• Load hunting. Two controllers fighting each other, with load oscillating between machines. Normally a governor tuning problem and fixable, but it is hard on both engines until it is.
• Unequal sharing. One unit carrying substantially more than its share, running hot and wearing faster while the other coasts.
• Breaker coordination. A fault should trip the nearest device, not the main. Getting this wrong means a single problem takes the whole site off.

Questions worth asking before the order

• Is the switchgear specified for paralleling, or is it single-machine gear that someone intends to extend later?
• Who owns the controls integration, particularly where engines and switchgear come from different suppliers? This is the gap work falls into.
• Droop or isochronous, and why that choice for this load profile?
• Has reverse power and protection coordination been tested under load, or only energised?
• What happens when one machine is taken out for maintenance? Does the site run derated, or stop?
• Can the controls shed mining load automatically if generation capacity falls, rather than tripping on overload?

That last one matters more than it sounds. A mining load is one of the few industrial loads that can genuinely be curtailed in seconds without damaging anything, which makes it an unusually good partner for a generation set with limited headroom. Controls that use that capability turn a constraint into a buffer.

The practical point

Paralleling sits between the engine supplier and the electrical contractor, which is exactly why it gets orphaned, in the same way wellhead compression does on the fuel side. On projects that run well it is scoped explicitly, with one party responsible for the controls working as a system rather than each part working alone.

Rax Mining deploys natural gas modular data units on sites that frequently run more than one generation unit. If you are planning a multi-genset site and want the electrical scope defined before the equipment is ordered, get in touch.