Troubleshooting 2.3 kV Motor Capacitor Bank Failures

At the first start of a 625 HP, 2.3 kV, 890 rpm induction motor, the bus suffered a large undervoltage, the phase A and B fuses opened, and another energized motor lost the same two phases. Both 160 kVAR, 2.3 kV capacitor banks were then found internally shorted between phases A and B with expanded tanks. That pattern requires a common-bus investigation: an isolated manufacturing defect does not explain simultaneous operation of protection on two parallel motor circuits.

Failure mechanism and branch logic

An expanded capacitor tank indicates a destructive internal dielectric failure that produced heat, pressure, and fault current. The phase A-to-B shorts identify the failed electrical path, but they do not by themselves identify which bank failed first.

Three mechanisms fit the event sequence. First, failure of the newly energized bank could have imposed a bus transient or voltage unbalance that damaged the existing bank. Second, adding 160 kVAR could have shifted a system resonance close to a harmonic already present on the 2.3 kV bus. Third, capacitor energization could have created an oscillatory switching transient, particularly with another bank already energized. Motor starting current could deepen the simultaneous voltage sag without being the initiating cause.

Follow the checks in order. Check 1 establishes whether the event originated inside one bank or across the bus. Check 2 determines whether connection and protection allowed one failure to involve the other circuit. Checks 3 and 4 test resonance and harmonic excitation. Check 5 addresses capacitor sizing and switching state.

Check 1 — common-bus event versus isolated defect

Quarantine both expanded banks and do not re-energize them. Record the single-line arrangement, transformer connections, bus grounding method, motor contactor positions, capacitor switching devices, and which banks were connected immediately before the event. The term common-bus event here means a voltage or current disturbance visible at both motor feeders, not merely two components failing near the same time.

1. Review protection records for the relative opening order of phase A and B fuses on both feeders. If one feeder began conducting fault current before the other, treat that circuit as the likely initiating branch and continue to Check 2. If the records cannot resolve the order, proceed with the common-bus measurements in Check 3.
2. Test each isolated capacitor bank for phase-to-phase resistance, phase capacitance, and insulation to the tank using the capacitor manufacturer's test method. Expect no low-resistance phase-to-phase path in a serviceable bank and comparable phase capacitance. The reported A-to-B short and tank expansion require replacement or manufacturer-directed failure analysis, not another trial energization.
3. Inspect switching contacts, cable terminations, fuse clips, and bus connections for arc marks or tracking. A fault external to a tank changes the initiating-event branch; internal damage in both tanks keeps transient and resonance mechanisms active.
4. Compare motor-start data with the event. A normal start should show a voltage sag correlated with motor current and then recovery. A superimposed high-frequency oscillation, abrupt phase collapse, or rising capacitor current points to capacitor switching or failure.

Check 2 — bank connection and protection

The capacitor banks were connected in parallel with their motors and lacked individual capacitor fuses. Motor feeder fuses therefore may have been the only devices available to interrupt capacitor fault current. That arrangement can allow a capacitor fault to persist until upstream or motor-circuit protection clears it, increasing fault energy and exposing the common bus.

Document whether each bank is switched by the motor contactor, by a separate contactor, or remains connected after the motor disconnects. Also identify the capacitor topology. Medium-voltage banks are often arranged as ungrounded wye, but the actual connection must come from drawings and terminal inspection. In an ungrounded wye bank containing multiple units per phase, operation of one unit fuse can shift the neutral and raise voltage on the remaining units. Use ANSI/IEEE 37.99 as a document to check when evaluating bank protection and voltage unbalance; it is not a substitute for the capacitor manufacturer's permitted unit voltage and fuse-coordination data.

Check each phase for dedicated capacitor overcurrent protection. Select the device type and rating from rated capacitor current, permissible tolerance, switching inrush, harmonic current, available fault current, and the bank manufacturer's instructions. Protection must tolerate normal energization yet clear an internal fault before the tank absorbs destructive energy. Apply the electrical code governing the installation; the source information identifies individual phase protection as a United States code consideration but does not provide a jurisdiction or clause.

Check 3 — resonant harmonic order

Parallel resonance occurs when the effective inductance of the source and the connected capacitance exchange energy near a network natural frequency. At resonance, a modest harmonic voltage or current can produce much larger capacitor current and bus-voltage distortion. Adding the second bank changes total capacitance and lowers the resonant frequency, so a system that operated for six months can become unstable after one switching change.

Use the following screening calculation at the capacitor point of connection:

h ≈ √(Ssc / Qc)

Here, h is the approximate harmonic order, Ssc is three-phase short-circuit strength in MVA at the bus, and Qc is the total connected capacitor rating in Mvar. One bank is 0.160 Mvar. If both 160 kVAR banks are simultaneously connected to the same effective bus, use 0.320 Mvar for that operating state. Calculate both cases because switching one step changes the resonance.

Obtain Ssc from a short-circuit study or calculate it from source and transformer impedance. Do not substitute transformer nameplate MVA for bus short-circuit MVA. A result near an integer harmonic calls for a frequency-domain network study. A value such as h = 5.1 deserves close examination around the fifth harmonic; a value near 49 is far less likely to couple strongly to common low-order converter harmonics. This square-root equation is a screening tool. Cables, busbars, transformer branches, multiple capacitor locations, and motors create additional resonant modes, potentially from roughly 5 to 50 times line frequency.

Check 4 — harmonic sources and bus distortion

Low-voltage electronic converters remain relevant even when supplied at 460 V downstream of transformers. Their harmonic currents flow through transformer leakage impedance and the upstream network. The transformer buffers high-frequency components but does not isolate the 2.3 kV bus from low-order harmonics. If a network mode lies near one of those harmonics, the new capacitance can amplify bus voltage and capacitor current.

1. Inventory adjustable-frequency drives, rectifiers, and other converter loads supplied through every transformer connected to the 2.3 kV system. Record operating combinations at the time of failure.
2. Measure three-phase bus voltage distortion and individual capacitor currents with all banks disconnected, with each bank alone, and with both banks connected only after the installation is safe for controlled testing. Expect a stable spectrum without a sharp rise at one harmonic as capacitance is added.
3. Capture harmonic magnitudes and phase angles at the capacitor bus rather than relying only on measurements at 460 V. A large increase in one voltage harmonic or disproportionate capacitor current identifies a resonant branch.
4. Compare converter load with system loading. More than 20% converter or other harmonic-producing load is a rough screening threshold against uncontrolled local correction, not a design limit or standards allowance. Define the percentage base explicitly in the harmonic study.

If resonance is confirmed, adding replacement capacitors of the same rating without changing the network recreates the failure mechanism. Evaluate centralized correction using capacitors with series reactors. A commonly cited arrangement tunes the branch slightly below the fifth harmonic so it does not form an uncontrolled parallel resonance at that harmonic. Calculate the reactor and capacitor duties from the measured spectrum and system model; no reactor percentage or tuning frequency can be selected from the capacitor rating alone.

Check 5 — correction level and switching state

Verify why 160 kVAR was selected for each motor. Use measured motor reactive power at full load, partial load, and no load rather than horsepower alone. The 625 HP and 450 HP ratings do not define required correction because efficiency, displacement power factor, and actual loading are also required.

At fixed capacitance and frequency, capacitor current follows I = ωCV, while reactive output follows Q = ωCV². Elevated terminal voltage therefore raises both capacitor current and kVAR. An overcorrected induction-motor circuit can experience excessive voltage at light load even while still connected to the source. A practical rule of thumb is to limit correction to approximately 0.97 power factor at no load, but the motor and capacitor voltage limits remain controlling.

A separate hazard exists if the running motor is disconnected from the line but remains connected to its capacitor. The rotating machine and capacitance can sustain a self-excited voltage that may exceed the normal terminal voltage. Verify contactor sequencing so the capacitor cannot remain connected to an isolated, coasting motor unless the system was specifically engineered for that state.

Capacitor switching also produces an oscillatory voltage transient. Back-to-back switching—energizing one bank while another is already charged on the same bus—can create high-frequency inrush that exceeds ordinary steady-state capacitor current. Capture peak current, oscillation frequency, damping, and phase voltage during switching; an RMS-only meter can miss the initiating transient.

Resolving procedure and controlled verification

1. Remove both damaged banks from service and obtain the manufacturer's examination of the internal A-to-B failures where practical. Replace expanded tanks rather than attempting to prove them serviceable by energization.
2. Reconstruct the event sequence from protection records and switching states. Identify which motor, capacitor, and upstream devices were closed before each fuse operation.
3. Calculate the approximate harmonic order for 0.160 Mvar and 0.320 Mvar, then build a frequency scan when either result approaches a significant measured harmonic.
4. Measure 2.3 kV bus voltage distortion and converter-related harmonic current under representative load combinations. Include the 460 V converter loads through their transformer models.
5. Recalculate correction from measured motor reactive power at no load and normal operating load. Reduce or stage the bank if no-load power factor would exceed the approximately 0.97 screening target or terminal voltage rises beyond equipment limits.
6. Engineer the mitigation identified by the study: revised bank size or staging, controlled switching, a centralized reactor-capacitor arrangement, or another modeled filter solution.
7. Add coordinated overcurrent protection for every capacitor phase and verify the actual bank connection, grounding arrangement, contactor sequence, interrupting rating, and unbalance consequences.
8. Commission one operating state at a time while recording all three phase voltages, motor currents, capacitor currents, harmonic spectrum, and protection status.

Check 1: With no capacitor connected, expect balanced 2.3 kV bus voltage and a repeatable motor-start sag that recovers after acceleration. Check 2: With each approved bank step connected separately, expect phase currents to remain balanced and no sharp amplification at a harmonic. Check 3: During switching, expect transient peak and damping to remain within the selected capacitor, switch, and protection ratings. Check 4: At no load and normal load, expect power factor, terminal voltage, and capacitor current to remain within the design limits. Check 5: During a controlled motor trip, expect the capacitor to disconnect in the engineered sequence and no sustained voltage on an isolated coasting motor.

Frequently asked questions