Capacitor Bank Failures and Harmonic Resonance on a VFD-Loaded 630 kVA Supply
A factory is fed by a dedicated 630 kVA, 11/0.433 kV Dyn11 transformer with 5.5% impedance. The main LV board supplies six 75 kW 6-pulse variable-frequency drives on pump and fan duty, plus about 300 kVA of single-phase load consisting of LED lighting, switch-mode chargers and office IT, on the same board. Power factor correction of 400 kvar in fixed steps is connected at the main board to avoid a utility penalty. In the last quarter two capacitor steps failed with ruptured fuses and bulged cans. Power quality logging at the main board shows phase current THD 34%, with the 5th at 26%, the 7th at 12% and the 11th at 6%. Voltage THD is 7.8% and the power factor is 0.93 leading at times. The measured neutral current on a 400 A four-core feeder is 165 A while its phase currents are 210 A. The transformer is running hot, with winding temperature about 12 K above expectation for its load.
1. Calculate the parallel resonance between the transformer and the capacitor bank and explain why the capacitors are failing.
Read the worked answer
Short-circuit capacity of the transformer. Ssc = 630 kVA / 0.055 = 11,455 kVA.
Resonant order. h = sqrt(Ssc / Qc) = sqrt(11,455 / 400) = sqrt(28.6) = 5.35.
The parallel resonance sits between the 5th and the 6th harmonic, very close to the 5th, which is the dominant harmonic at 26% of fundamental. The system is therefore operating almost on top of its resonance, where the capacitor bank and transformer form a high-impedance tank and harmonic currents circulate with large magnification.
Why the capacitors fail.
- Capacitive reactance falls with frequency (Xc = 1 / 2 pi f C), so the capacitor draws harmonic current in proportion to harmonic order. The 5th harmonic current through the bank can be several times the fundamental component of that order.
- The dielectric and the internal fuses see a combination of raised RMS current and peak voltage distortion. Overheating, bulging and fuse rupture follow, and the weakest steps fail first.
- With steps switching in and out, the resonant order shifts. At part load with fewer steps in service the resonance moves towards the 7th, so the plant passes through different amplification points on a daily basis.
- Leading power factor at light load confirms over-correction: too many kvar remain connected when the drives are unloaded, which raises the voltage and increases dielectric stress.
Other contributing factors. Voltage THD of 7.8% already exceeds the 5% planning level usually applied at LV, so the capacitors are being asked to operate in a distorted environment they were not specified for.
2. Design the corrective solution. Check the capacitor voltage rating and effective output with a detuned reactor, and state what else you would change.
Read the worked answer
Detuned bank. Fit a series reactor with a 7% tuning factor (tuned to about 189 Hz, h = 1 / sqrt(0.07) = 3.78). The series resonance at 3.78 is below the 5th, so the bank is capacitive at 50 Hz but inductive at the 5th and above, and the parallel resonance with the transformer is pushed below the 5th where there is little harmonic current. A 5.67% reactor (h = 4.2) is an alternative but leaves less margin below the 5th; 7% is the common choice for 6-pulse drive loads.
Voltage across the capacitors. With the reactor in series, the capacitor voltage rises: Uc = U / (1 - p) = 433 / (1 - 0.07) = 466 V. The capacitors must therefore be rated at least 480 V, and preferably 525 V, not 440 V. This is the most common design error in retrofitted detuned banks and explains a second wave of failures when the reactor is added to existing 440 V cans.
Effective output. The reactor consumes reactive power, so the effective bank output is about (1 - p) = 93% of the nominal kvar. To deliver the 400 kvar actually needed, specify roughly 430 kvar nominal.
Control and specification changes.
- Switch to automatic step control with a target displacement power factor of about 0.95 lagging, and a deadband, so that over-correction and leading power factor at light load are eliminated. Never correct beyond the transformer's no-load reactive demand.
- Verify the detuned bank's own duty with the 5th and 7th harmonic currents at the point of connection; the reactor current rating must include them.
- Add 4% line reactors or DC-link chokes on the 6-pulse drives to cut the current distortion at source. This is usually the cheapest kVA of reduction.
- Consider an active harmonic filter for the residual distortion if voltage THD cannot be brought below 5% by passive means.
3. Explain the 165 A neutral current and specify the feeder and transformer measures needed.
Read the worked answer
Why the neutral carries 165 A. The drives are the source of the 5th, 7th, 11th and 13th harmonics, but those are positive- and negative-sequence and do not add in the neutral. Triplen harmonics (3rd, 9th, 15th) are zero-sequence and add arithmetically in the neutral. The 300 kVA single-phase load of LED drivers, switch-mode chargers and IT power supplies is a rich source of 3rd harmonic, and in a four-wire system those currents sum rather than cancel. A neutral current of 165 A against phase currents of 210 A is roughly 79% of phase current, which is the normal fingerprint of a heavily triplen-loaded lighting and IT board.
Feeder measures.
- A four-core cable with a reduced neutral is not acceptable here; specify a full-size neutral or a five-core cable with the neutral sized for at least the phase current, and check the installed cable's neutral cross-section before increasing load.
- Split the single-phase IT and lighting load across separate distribution boards, and consider a dedicated transformer for the IT load so that triplen currents do not circulate through the main board.
- Check for neutral overload protection: phase-only protection will not detect a neutral at 165 A.
Transformer measures.
- The 12 K excess winding temperature is consistent with harmonic loss in the windings and stray loss in the tank. Quantify it with a harmonic loss factor and either derate the transformer by the appropriate factor or replace it with a K-rated unit.
- Check the neutral busbar and the transformer neutral connection temperature; a Dyn11 transformer carries the full zero-sequence current in the LV neutral.
- Verify the transformer's thermal protection settings against the derated capacity, and re-check the voltage THD at the LV terminals after the detuned bank and drive reactors are fitted, with the target being 5% or less.