The Capacitor Bank That Drove Our Diesel Generator Insane

The call came on a Tuesday morning. A plastics factory on the edge of town — new capacitor bank installed the week before, power-factor penalty gone from the utility bill, plant manager proud as anything.

Then the monthly generator test ran. Voltage started at 400. Drifted to 415. Touched 437 and kept climbing. The AVR sat there blinking a fault nobody on site had ever seen. Nothing was wrong with the generator. The brand-new capacitor bank had turned it against itself.

I’ve watched this exact fight play out four or five times now. Every time, the capacitor bank wins the first round, and the generator pays for it. Here’s the thing nobody told that plant manager: a diesel generator and a capacitor bank are built on opposite assumptions about the world.

Your Generator Was Rated 0.8 Lagging. The Capacitors Went Leading.

Look at any generator nameplate. You’ll find a power factor rating — almost always 0.8 lagging. That number isn’t decoration. It means the machine’s excitation system was designed around inductive loads: motors, transformers, the kind of loads that drag current behind voltage.

Capacitors do the opposite. They push current ahead of voltage. Throw enough capacitance at a generator and the power factor swings past unity into leading territory. And that’s where the trouble starts.

A generator’s AVR holds voltage steady by adjusting field current. On a lagging load it pushes more excitation in. On a leading load it backs excitation off. Keep adding capacitors and the AVR keeps backing off — until it hits the floor. Minimum excitation. It can’t go any lower. Past that point the machine has no control authority left, and voltage starts doing whatever the capacitors feel like. Usually up.

I’ve seen a 400-volt system sit at 440, then 455, during a routine test, with the AVR display showing minimum field and a warning light nobody recognized. The generator wasn’t broken. It was out of travel. Like a thermostat already dialed to zero while the room keeps getting hotter.

Technician measuring voltage on a diesel generator control panel with a multimeter during a capacitor bank load test
Technician measuring voltage on a diesel generator control panel with a multimeter during a capacitor bank load test

The Voltage That Wouldn’t Die After Shutdown

Here’s the one that still bothers me. On one job — a cold-storage warehouse, roughly an 800 kVA set — we shut the generator down after a test. Engine stopped. Silence. And the voltmeter on the switchboard was still showing over 300 volts. Climbing, slowly.

Residual magnetism in the rotor plus the capacitor bank still connected across the terminals. The capacitors fed the machine’s own residual field, which made voltage, which fed the capacitors, which made more voltage. A feedback loop with no engine running. The textbooks call it self-excitation. Standing in front of that board, watching numbers climb on a dead machine, “textbook” was not the word in my head.

An old-timer I work with swore the fix was switching the AVR to manual and driving voltage down by hand. I told him that was treating the symptom with a blindfold on. We argued about it over lunch — he still thinks I’m too fussy about interlocks. But here’s what ended the debate: we pulled the capacitor breaker open, and the phantom voltage died in about two seconds. The capacitors were the engine now. That’s not a control problem. That’s a wiring problem.

If your capacitors stay connected to the generator terminals when the set stops — or when the transfer switch moves back to utility — you’ve built a machine that can excite itself with nobody home. Utility engineers lose sleep over islanding for the same reason, just at bigger scale.

Resonance Always Finds the Weakest Capacitor

There’s a second fight, quieter and slower. Capacitors plus the inductance of the whole system — generator windings, transformers, cables — form a tuned circuit. At some harmonic frequency that circuit resonates. When it does, harmonic currents that were harmless get amplified, sometimes several times over.

On generator power this bites harder than on utility, because a generator’s impedance is higher and its waveform already carries more harmonic content than the grid’s. I’ve opened capacitor cabinets on generator-fed sites and found cans bulging like overfilled soda tins, fuses blown for “no reason,” and once a capacitor that had vented its innards across the inside of the enclosure. The maintenance log said “bad batch of capacitors.” It wasn’t a bad batch. It was the fifth harmonic having a party.

Capacitors don’t filter harmonics. Without detuned reactors in series, they invite specific harmonics in and hand them a megaphone. If you haven’t read my piece on how dirty generator power gets, this is the footnote that proves the point — and the excitation system is the part left holding the bag when voltage starts misbehaving.

Technician inspecting power factor correction capacitors with a flashlight inside an electrical cabinet
Technician inspecting power factor correction capacitors with a flashlight inside an electrical cabinet

The Boring Fix That Actually Works

So do you rip the capacitor bank out? No. The utility penalty is real money, and most of the year the plant runs on grid power anyway. The honest fix is making the capacitors and the generator stop meeting each other unprepared:

  • Measure power factor on generator load, not just on the utility bill. The two numbers are often strangers.
  • Size capacitor steps small and switch them. A 200-kvar wall slamming in at once is a hammer; four 50-kvar steps are a conversation.
  • Put detuned reactors in series with every step. It moves the resonance point somewhere harmless.
  • Interlock the capacitor contactor with the generator breaker: generator running means capacitors drop out unless you’ve engineered them to stay. This one line of logic prevents the self-excitation stunt.
  • Aim for about 0.9 lagging on generator, not unity. Unity leaves zero margin, and margin is what keeps the AVR in its comfort zone.

That plastics factory? We interlocked the bank to drop out on generator, added reactors, re-tuned the steps. Next monthly test, voltage sat at 402 like it had never misbehaved. The plant manager’s pride survived. The penalty stayed gone eleven months of the year. Everybody won except the capacitor salesman who’d promised “set and forget.”

Generators are honest machines. They tell you exactly what the load is doing to them — in voltage, in heat, in the way the AVR strains at its limits. Capacitors aren’t the enemy. Unsupervised capacitors are. Treat them like what they are — a powerful correction that needs a chaperone whenever the generator is running — and the two can share a switchboard without drama. If your site is wrestling with this, talk to someone who’s seen the fight before yours starts.

Related news

Contact Us

Contact Us

+86 189-5531-5152

Email:[email protected]

Work:Monday to Friday,9:30-18:30

WhatsApp FaceBook
SHARE
TOP