What Correcting Power Factor Actually Changes

Why this matters

A shop gets told its power factor is poor, a bank of capacitors goes in, and six weeks later the same motor is still tripping its overload and the same panel is still running warm. Nobody lied. Power factor correction did exactly what it does, which is unload one specific segment of conductor and nothing else. If you cannot say which segment before the capacitors are ordered, you cannot predict whether anything the customer complained about will improve, and you may hand them a new problem they did not have.

Before anything is opened

Capacitors hold charge after the supply is removed, so this is a stored-energy job before it is a diagnostic one. Work de-energized: 29 CFR 1910.333(a)(1) permits energized work only where the employer can demonstrate that de-energizing introduces additional or increased hazards or is infeasible due to equipment design or operational limitations. Open and lock the disconnect under 29 CFR 1910.333(b)(2), which also requires that stored electric energy which might endanger personnel be released and capacitors discharged. Note the fork: 29 CFR 1910.147 excludes work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C), so the electrical procedure is 1910.333(b)(2), with 29 CFR 1926.417 as the construction counterpart. Wait the manufacturer's stated discharge time, then prove it with a meter across every terminal pair and from each terminal to ground, using live-dead-live per NFPA 70E-2021, 120.5, in the edition your employer's electrical safety program adopts. The discharge resistor inside the can is the reason the wait works, and a resistor that has failed open leaves the can charged with nothing on the outside to show it, which is why the meter reading and not the clock is the proof.

Where the reactive current actually goes

A capacitor connected across the supply supplies magnetizing (reactive) current locally. The motor still draws the same reactive current it always did; it just draws it from the capacitor a few feet away instead of from the transformer a few hundred feet away. That is the whole mechanism, and everything else follows from it.

   utility panel with capacitor motor
     |                      |                       |
     +======================+=======================+
     |                      |                       |
     | real and reactive    |   real and reactive   |
     | current both flow    |   current unchanged   |
     | here, so total       |   on this segment     |
     | current falls        |                       |
                     capacitor supplies
                     the reactive part

The conductors carrying less current are the ones between the capacitor and the source. Move the capacitor to the motor terminals and the unloaded stretch grows to include the feeder; put it at the service entrance and the unloaded stretch is the utility side only. That placement choice is the single decision that determines what improves.

Four things correction does not change

The current on the load side of the capacitor. Conductors, starter contacts, overload relay and motor leads between the capacitor and the machine carry what they always carried. A contactor pitting from load current pits at the same rate.

The real power the load consumes. Real power is set by the mechanical work plus losses. Correction changes the current that delivers it, not the amount.

Inrush at start. Locked-rotor current is a low-impedance condition in the motor, not a power factor condition on the feeder, and a standing capacitor bank at the panel does not meaningfully change it.

Harmonic current. Harmonic current is not 60 Hz reactive current, and a capacitor sized for reactive compensation does not cancel it. It gets worse, for the reason in the next section.

That is four items, and none of them is the thing most customers are complaining about when the word "power factor" first comes up.

What it does change, on the segment that sees it

Lower current on the upstream segment means lower I squared R heating in those conductors and lower voltage drop across them, both scaling with the square and the first power of current respectively. That recovers real headroom: a transformer loaded to its nameplate by apparent current can carry more real load after correction without any change to the transformer. It also changes what the utility meters, which is a separate question with its own economics. This article does not price it; see the sibling reference on what power factor actually costs a shop, which owns that gate.

Four behaviours the bank brings with it

Parallel resonance with the source. The capacitor and the source inductance form a parallel resonant circuit. For an idealized single source inductance and a single capacitor bank, with load damping neglected, the resonant harmonic order is approximately the square root of the ratio of short-circuit capacity at that bus to the capacitor's reactive rating, both expressed in the same units. That estimate tells you which harmonic order is nearby, not an exact frequency, because real load resistance damps and shifts it.

The energizing transient. Switching an uncharged capacitor onto a live bus produces an oscillatory overvoltage. In the ideal lossless single-bank case, switched at the voltage peak with no pre-insertion resistor, the theoretical ceiling at the capacitor is about 2 per unit. Real installations land under that, but it is enough to trip a drive on DC bus overvoltage every time the bank steps in.

Self-excitation. Capacitors left connected to a motor after the motor is disconnected from the line can excite it as a generator while it coasts. The guard is to keep the capacitor's reactive rating below the motor's no-load magnetizing requirement, and that value belongs to the motor manufacturer's data, not to a rule of thumb.

Leading power factor and terminal overvoltage at light load. A fixed bank sized for the busy shift is oversized for the night shift, and overcompensation raises terminal voltage rather than lowering it. This is the argument for automatic staged banks in a plant whose load swings.

Worked example: the correction that made the problem worse

A three-phase panel feeds a mix of motors and several six-pulse rectifier loads. Short-circuit capacity at that bus is 25 times the reactive rating of the capacitor bank the supplier quoted. Using the relationship above and its stated idealization, the resonant harmonic order estimates as the square root of 25, which is 5.

Five is exactly where a six-pulse rectifier puts its largest characteristic harmonic. So the bank does not sit quietly beside the harmonic source; it parks a parallel resonance on top of it, and the harmonic current circulating between capacitor and source is amplified relative to what the rectifiers actually inject.

What shows up in the field: capacitor fuses that clear on no obvious event, a transformer that runs hotter after the correction than before it, and a measured true-RMS current on the transformer secondary that has not fallen by anything like the amount the supplier's arithmetic predicted. The 60 Hz reactive current genuinely did fall. The harmonic current rose enough to eat the difference.

Now change one input and watch the answer flip. If the short-circuit capacity at that bus were 12 times the bank rating rather than 25 times, the estimated order is the square root of 12, about 3.5, which sits between the third and fifth and clear of the largest six-pulse orders. Same bank, same plant, different stiffness of supply, and the resonance no longer lines up with anything the load produces. This is why a bank that has run for a decade at one facility misbehaves at another with the same connected load: the number that changed is upstream of both of them.

The fix is not a smaller bank chosen by feel. A series reactor ahead of the capacitor moves the system's parallel resonance down below the lowest significant harmonic order present, and the tuning order is a property of the reactor and capacitor chosen together, so the supplier who sizes the pair owns that number. What flips the recommendation entirely: if the plant has no significant rectifier, drive or electronic ballast load, the harmonic spectrum is clean, there is nothing for a resonance to amplify, and plain capacitors are the correct and cheaper answer.

The failure mode to catch: the supplier's proposal arrives with a before and after current figure and no statement of short-circuit capacity at the point of connection. Without that number the resonance order cannot be estimated at all, and the proposal is arithmetic about 60 Hz applied to a bus that is not only carrying 60 Hz.

Who owns which number

Three different parties own the three numbers this decision needs, and taking any of them from the wrong source is how the estimate goes wrong.

• Short-circuit capacity at the point of connection belongs to the utility or to an engineering study of the site's transformer and service conductors. It is not on any nameplate in the building.
• The motor's no-load magnetizing requirement, which caps the capacitor rating for any bank switched with the motor, belongs to the motor manufacturer's data for that frame and speed.
• Conductor ampacity and overcurrent protection for the capacitor circuit belong to the NEC in the edition your authority having jurisdiction has adopted, with capacitor provisions in Article 460, and to the equipment manufacturer's instructions, which NEC 110.3(B) makes enforceable in that same adopted edition.

How to verify you got this right

Measure at the point that should have changed and at a point that should not. On the source side of the capacitor connection, true-RMS current should fall and the ratio of real to apparent power should rise. On the load side of the connection, current should read within measurement noise of what it read before, and if it does not, either the capacitor is not where you think it is or something else changed at the same time.

Then check what the correction did to the rest of the spectrum. If your instrument reports total harmonic distortion of current or the individual harmonic orders, record them before energizing the bank and again after, on the same feeder, at a comparable load. A correction that lowers fundamental current and raises fifth-harmonic current on the same conductor is the resonance case above declaring itself early, while the capacitor fuses are still intact.

Do not compare a post-correction reading taken on a light shift against a pre-correction reading taken on a busy one. If the two readings cannot be taken at comparable load, say in the record that the comparator is uncorrected rather than reporting the difference as an improvement.

References

• 29 CFR 1910.333, electrical safety-related work practices, including the de-energized-work gate at (a)(1) and the lockout and stored-energy release requirements at (b)(2); 29 CFR 1926.417 for the construction counterpart
• NFPA 70E-2021, 120.5, process for establishing an electrically safe work condition, in the edition adopted by your employer's electrical safety program
• NEC Article 460 (capacitors) and 110.3(B) (installation per listing and instructions), in the edition adopted by your authority having jurisdiction
• Motor and capacitor manufacturer data for no-load magnetizing requirement and permissible switched capacitor rating
• See related: What Power Factor Actually Costs a Shop; What Power Factor Does to a Current Reading