The Hidden Costs of Incorrect Active Harmonic Filter Sizing

The Hidden Costs of Incorrect Active Harmonic Filter

Sizing

Got a call last month from a panel builder in Malaysia. His customer — a textile mill running 47 VFD-driven spinning frames — had commissioned a 100A AHF three months earlier. THDi at the PCC dropped from 28% to 19%. Not terrible, but nowhere near what the spec promised. The supplier insisted the unit was working. It was. Problem was, it was undersized by roughly 40%.

I've seen this dozens of times.

The mill's actual harmonic current, logged over a full production cycle, peaked at 167A during ramp-up. The 100A module saturated. The controller hit its limit. The remaining 5th and 7th harmonic current went straight back to the upstream transformer — uncompensated. The customer paid for partial correction and got exactly that.

Twenty-two years in commissioning and field troubleshooting. One thing I know: the gap between datasheet numbers and field measurements almost always traces back to sizing. Sizing that didn't account for real operating conditions.

Why Nameplate Methods Fail

Most people size an AHF by taking the transformer kVA, multiplying by some rule-of-thumb percentage. 25%, a lot of them say. Sounds safe. Until it isn't.

A 2000kVA transformer feeding six-pulse VFDs — the harmonic current follows the drive loading, not the transformer rating. If those drives run at 60% average load with frequent ramp cycles, the harmonic spectrum shifts. Fifth harmonic might hit 33% of fundamental. Seventh, another 12%. These add in RMS, not straight arithmetic.

Then there's the background problem. If the PCC already carries 3% voltage THD from the grid side, your AHF has to eat that first — harmonics it didn't create. I've measured sites where 30% of the filter's capacity went to cleaning up incoming grid distortion before touching the internal loads.

Three traps that rule-of-thumb sizing walks right into:

1. Crest factor of the load current. Drives with DC chokes pull cleaner current than those without. Same kW, very different harmonic profile.

2. Parallel resonance. Existing capacitor banks shift system impedance. At certain frequencies they amplify harmonic current well beyond what the drives generate. The AHF sees the amplified version and has to swallow all of it.

3. Neutral current. Triplen harmonics add arithmetically in the neutral. In a 4-wire system with single-phase loads, neutral current can hit 1.7x phase current. If the AHF is 3-wire, that neutral current is invisible to the CTs — but not to the transformer.

That last row is the one that bites. Utility contract says THDi under 5% at the PCC, load keeps shifting — passive filtering alone almost never holds it.

SVG and AHF: Not the Same Thing

I've seen RFPs that compare SVG and AHF as competing solutions. They're not. Each handles a different problem, at a different point in the system.

SVG compensates fundamental reactive power. It corrects displacement power factor. It holds voltage steady during motor starts and load swings. The control loop locks onto the fundamental voltage phase angle and synthesizes reactive current accordingly.

AHF compensates harmonic current. It doesn't care about power factor at 50 or 60Hz. It cares about 250Hz, 350Hz, 550Hz — every integer multiple of the fundamental where rectifier loads push distortion back into the grid.

The term "hybrid unit" causes confusion. What matters is the architecture. A proper hybrid runs two control loops in parallel — one for reactive, one for harmonic — on a shared DC bus. The cheap approach time-shares a single inverter between tasks. Response time takes a hit. At 45°C ambient, I've watched time-shared units derate by 30% because the IGBT junction couldn't handle the switching frequency swings.

The Thermal Side Spec Sheets Skip

Semiconductor ratings are given at 25°C junction temperature. Real electrical rooms run 40 to 50°C.

IGBT conduction losses increase with temperature. At 45°C ambient, a power module rated 100A at 25°C might only sustain 78 to 82A continuous before junction temperature runs past the limit. Not a defect. Physics.

Three-level topology earns its place here. DC bus voltage splits across two series IGBTs — each switch sees half the voltage stress. dv/dt across the output filter inductor drops. Switching losses run about 25% lower than equivalent two-level designs at the same frequency. The module runs cooler. Derating at high ambient is less punishing.

In Southeast Asia or the Middle East, where 45°C is a normal afternoon, this isn't a nice-to-have. It's the difference between a filter that runs through the hottest shift and one that trips on overtemperature.

My rule now: if the measured ambient at the panel location exceeds 35°C — not the building average, the actual spot — I default to 3-level topology.

No Audit Before Commissioning Is Just Guessing

Cheapest insurance against undersizing: 72 hours of Class-A power quality logging before signing the purchase order.

The logging window must span a full production cycle. For a textile mill, that means startup, ramp, steady state, shutdown. For a welding line, the worst-case weld schedule with maximum simultaneous gun firings. Same equipment, light load versus full throughput — I've watched harmonic current double.

The data tells you straight:

· Actual THDi at each load step

· Which harmonic orders dominate — not all filters handle all orders equally

· Background voltage distortion coming in from the grid

· Reactive swing magnitude and speed

· Neutral current — often the single number that decides 3-wire versus 4-wire AHF

Without this, you're buying on assumptions. The transformer-percentage method gets you a rough direction. A proper audit gets you a solution that holds when production is flat out and the room temperature is 44°C. That's what counts.

FAQ

Q: Can I install an AHF without shutting down the plant?
A: Yes, with planning. CT installation needs the monitored bus de-energized, typically 2 to 4 hours. The AHF itself goes in during a maintenance window. For critical processes that can't stop, install the CTs during the next planned outage, bring the AHF online in bypass mode, then switch to active compensation later.

Q: My load changes a lot with the seasons. How do I size the AHF?
A: Size for the worst-case harmonic current measured across the full operating year, then add 20 to 25% margin. If the seasonal peaks are short, go modular — add power modules later as needed. The control architecture should support parallel operation without master-slave dependency. Each module runs independently off a shared CT signal.

Q: What maintenance does a 3-level AHF need?
A: Annual visual check of DC-link capacitors for bulging or leakage. Clean air filters quarterly on forced-air-cooled units. Capacitor bank replacement every 3 to 5 years, depending on operating hours and ambient temperature. IGBT modules typically exceed 100,000 hours at rated conditions. The control board has no moving parts — no scheduled maintenance beyond firmware updates.

Q: Will an SVG fight my existing capacitor banks?
A: Not if the controller is set up right. The SVG handles fast reactive compensation. The capacitor banks, switched by slower contactors, carry the base reactive load. Set the SVG target power factor at 0.98 to 0.99. Let the capacitor banks switch only during sustained load changes. Don't put the capacitor bank switching threshold and the SVG compensation target in the same zone — they'll hunt.