Harmonic-rich office riser: parallel neutrals and a capacitor bank in resonance

Fundamental loading. The riser carries 615 A fundamental. Two 4×185 mm² Cu XLPE on perforated tray are each rated roughly 450 A in air at 30 °C; with four loaded circuits on the tray a grouping factor of about 0.80 applies, giving roughly 360 A per run, 720 A for the pair. At 615 A the pair sits near 85% - tight but not clearly overloaded. The check must use the actual manufacturer table, the true ambient (plant rooms often run at 40 °C, factor about 0.91) and the real spacing.

Harmonic derating. 27% THDi gives 640 A rms, which is what heats the conductor, so the thermal penalty is modest. The third harmonic per phase is 195/3 = 65 A, i.e. 10.6% of fundamental - below the 15% band where IEC 60364-5-52 harmonic reduction factors start to bite, so no further derating is justified. The neutral, however, carries 195 A continuously and must be treated as a loaded conductor in its own right.

Diagnosis. Heat at the lug only, with a clean barrel, points to termination resistance rather than conductor overload. Two parallel neutrals also rarely share equally: differences in route length, lug torque, or a borrowed neutral downstream will push one conductor well above half the total.

Measurements before replacing anything. Clamp each phase and each neutral of both runs individually under full load; a split worse than about 60/40 needs investigation. Thermographic scan of all terminations at full load, insulation resistance, and a DLRO contact-resistance test across each lug. Re-terminate with correct ferrules at the specified torque, then re-scan. Replace cable only if individual conductor currents or DLRO results show the conductor itself is distressed.

Resonance order. With a 1,000 kVA transformer at 5.75% impedance the short-circuit level at the main board is about 17.4 MVA. The parallel resonance of a capacitor bank of Qc on that source sits at h = √(Ssc/Qc) = √(17,400/500) ≈ 5.9. Adding the utility source and the cable shifts the natural resonance downwards, and the measured 5th harmonic rising strongly when the bank is switched in confirms the bank is amplifying it rather than the VFDs being the sole source.

Capacitor duty. At the 5th harmonic the capacitor reactance falls to one fifth, so the harmonic current it absorbs is large; the measured 1.6 × rated current exceeds the 1.3 × continuous overcurrent that IEC 60831 units are built for. The units will fail early, and the chattering contactors are the symptom of a controller seeing a distorted, unstable current and voltage.

Specification. Replace the bank with a detuned arrangement: a 7% series reactor per step tunes the branch to about 189 Hz, below the 5th (250 Hz), so the branch is inductive at the 5th and cannot resonate. Two consequences must be designed in: the capacitor voltage becomes U/(1 − p) = 400/0.93 ≈ 430 V, so specify 440 V (or 480 V) capacitors; and effective output falls to about 93% of nameplate, so roughly 540 kVAR of capacitors is needed for 500 kVAR effective. Reactors must be rated for fundamental plus harmonic current, and each switched step must retain its own detuning.

Alternative. If the VFD harmonic mitigation is being upgraded anyway, a central active filter may be more economic than detuned PFC plus reactors; do not install a passive filter without first confirming the measured impedance scan.

Target. At 400 V the applicable limit is IEEE 519-2014 for systems below 1 kV. Short-circuit current at the board is 17.4 MVA/(√3 × 0.4 kV) ≈ 25 kA, so Isc/IL ≈ 25,000/640 ≈ 39, which falls in the 20–50 band and gives a TDD limit of 8%. The measured 27% is more than three times the limit.

Ranked options.

1. Line reactors or DC-link chokes on each drive. A 4% line reactor or an equivalent DC-link choke typically halves the drive's current distortion. This is the cheapest per drive, needs no maintenance, and also protects the drive from supply transients. Expect overall TDD to fall to perhaps 13–17%; re-measure rather than assume.
2. Central active harmonic filter. The harmonic current is roughly 0.27 × 640 ≈ 170 A, so a 200 A active filter at the main board would remove most of what remains. Higher capital cost but adaptive, and it also cleans up the LED driver distortion, which reactors will not touch.
3. Drive replacement with active front ends or 12-pulse arrangements. Only justifiable at end of drive life; over-specifying here wastes money.

Sequence of work. Make the neutrals safe first (re-termination and verification), then correct the capacitor bank, then install reactors and re-measure over a full week before deciding whether the active filter is still needed. Specify the measurement to IEC 61000-4-7 with a Class A analyser, and record the Isc/IL ratio at the time of the survey because the TDD limit moves with it.