Einführung
Power factor correction capacitors are among the most widely installed pieces of electrical equipment in industrial and commercial facilities. Their purpose is straightforward — compensate for the reactive power drawn by inductive loads, reduce current in distribution cables and transformers, and avoid the financial penalties that utilities impose on facilities with poor power factor. For decades, in a world dominated by linear loads such as motors, Transformatoren, and lighting, they performed this role reliably and cost-effectively.
The widespread adoption of variable speed drives, switch-mode power supplies, and other non-linear loads has fundamentally changed this picture. In a plant where a significant proportion of the load is non-linear, the installation of power factor correction capacitors without accounting for harmonic distortion is only worse than ineffective — it is actively dangerous. Capacitors that were correctly specified, installed, and operating without issue for years can begin to fail repeatedly and unexpectedly once non-linear loads are introduced or expanded. Fuses blow for no apparent reason. Capacitor cases bulge or rupture. Transformers run hot. Protection relays trip on overcurrent with no fault on the load side. The root cause in most cases is the same: harmonic resonance.
This article explains why power factor capacitors behave the way they do in harmonic environments, what resonance is and how to calculate the conditions under which it occurs, what the field symptoms of resonance look like, and what the engineering solutions are — from detuned capacitor banks through passive harmonic filters to active harmonic filters. A practical selection guide is provided to help engineers choose the right approach for their specific installation.
A note on scope: the question of true power factor versus displacement power factor — and the optimal combination of passive and active filtering to achieve both harmonic correction and unity power factor — is a subject of sufficient depth to warrant its own dedicated treatment and will be addressed in a subsequent article in this series.
01 Power Factor Correction Fundamentals
Power factor is a measure of how effectively electrical power is being converted into useful work — the ratio of active power \(P\) (Watt) Scheinleistung \(S\) (volt-amperes):
\[PF = \frac{P}{S} = \frac{P}{V \cdot I}\]
A power factor of 1.0 means all current drawn from the supply contributes to useful work. A power factor below unity means some portion of the current is circulating between source and load without performing work, increasing losses in cables, Transformatoren, and switchgear without contributing to production.
Displacement power factor
In a purely sinusoidal system with linear loads, power factor degradation has a single cause: the phase displacement between voltage and current produced by inductive loads. The displacement power factor is:
\[DPF = \cos\phi\]
This is the power factor that traditional electromechanical meters measure, and the quantity that most utility tariff structures have historically used for power factor penalties. Capacitor banks correct displacement power factor by supplying the reactive current the inductive load requires locally. The reactive power required is:
\[Q_C = P \left(\tan\phi_1 – \tan\phi_2\right)\]
wo \(P\) is the Durchschnitt active power over the measurement period — not the instantaneous peak — to avoid oversizing the capacitor bank.
Modern 6-pulse variable speed drives use a diode bridge front end with no firing angle. The fundamental current is very nearly in phase with the voltage — displacement power factor is typically 0.95 zu 0.98 and is not a significant concern. The power quality problem has shifted entirely to the distortion component caused by harmonic currents, which capacitors cannot correct and which the classical displacement factor concept did not address. What IEEE 519 and IEC standards now call displacement power factor — \(\cos\phi_1\), the phase angle between fundamental voltage and fundamental current — is in practice almost unity for modern drive installations. The penalty on the utility bill, and the real engineering challenge, comes from distortion power factor alone.
Distortion power factor and true power factor
In a system containing non-linear loads, the current waveform contains harmonic components at integer multiples of the fundamental. These harmonic currents contribute to the RMS value of total current but carry no net active power at the fundamental frequency. The true power factor of a non-linear load is:
\[PF_{Stimmt} = DPF \times \dfrac{1}{\sqrt{1 + THD_I^{\,2}}}\]
A 6-pulse variable frequency drive at full load with \(THD_I = 35\%\) has a distortion factor of approximately 0.944. Even with displacement power factor corrected to unity by a capacitor bank, the true power factor will not exceed 0.944. A facility with a large population of drives may install capacitor banks in good faith to address a utility penalty, only to find the penalty persists because the utility meter measures true power factor.
Where capacitors are installed
Power factor correction capacitors are installed at one of three levels. At the individual equipment level, capacitors are connected directly at motor terminals, providing precise correction but multiplying the number of potential resonant circuits. At the group or busbar level — the most common industrial arrangement — a single fixed or automatically switched bank corrects the reactive demand of a group of loads. At the main service entrance level, a single large bank corrects the entire facility at the point of supply — simplest to install but concentrating the full resonance risk in one location.
Six-step assessment methodology
Before specifying any power factor correction equipment in a facility with non-linear loads, the following structured assessment should be performed.
Schritt 1 — Determine the utility penalty threshold. Identify the minimum acceptable power factor from the utility tariff — typically 0.90 oder 0.95 depending on jurisdiction.
Schritt 2 — Measure existing power factor. Measure \(P\) (kW) und \(Q\) (links) at the billing meter over a representative period — ideally one full week covering all operating modes. A single snapshot is insufficient.
Schritt 3 — Calculate the required capacitor rating Verwendung \(Q_C = P(\tan\phi_1 – \tan\phi_2)\). For automatic banks add a 10–15% margin for load growth.
Schritt 4 — Assess the need for a harmonic study. There are no universally standardised percentage thresholds that mandate a harmonic study. The technically defensible triggers, consistent with IEC 61642:2020 [4] and IEEE 519-2022 [1], sind: measured \(THD_V\) exceeding 5%, measured \(THD_I\) exceeding 15%, unexplained capacitor failures or fuse operations, or significant and growing non-linear load. As a practical screening guide — not a normative requirement — the following table reflects historical incident frequency in industrial installations [10][13]:
| Non-linear load proportion | Indicative recommendation |
|---|---|
| < 15% of total kVA | Standard capacitor bank may be acceptable — resonance check required |
| 15% - 30% | Detuned capacitor bank recommended |
| 30% - 50% | Detuned bank mandatory — passive harmonic filter may be required |
| > 50% | Full harmonic study required before any capacitor installation |
Schritt 5 — Resonance check. A simplified preliminary check uses only the transformer rating:
\[h_r \approx \sqrt{\dfrac{S_T}{Q_C}}\]
A rigorous assessment requires the short-circuit power \(S_{sc}\) at the point of common coupling:
\[h_r = \sqrt{\dfrac{S_{sc}}{Q_C}}\]
The simplified method overestimates \(h_r\) and is non-conservative — it is acceptable for first screening only. Wenn \(h_r\) falls within 10% of a characteristic harmonic order (5th, 7th, 11th, 13th) the bank design must be modified. Abschnitt 3 develops this calculation with a full worked example.
For larger installations requiring greater precision — particularly at medium voltage or where significant non-linear load is concentrated at a single point of connection — the design engineer should formally request from the utility not just the short-circuit level but the network impedance as a function of frequency. This harmonic impedance spectrum, sometimes provided as R and X values at each harmonic order, accounts for resonance conditions within the utility network itself that a single short-circuit MVA figure cannot reveal. IEC 61000-3-6 [5] provides a framework for this type of emission and impedance assessment at the point of common coupling.
Schritt 6 — Verify utility metering basis. Confirm whether the utility penalises on displacement PF or true PF. If true PF and \(THD_I\) exceeds approximately 15%, a capacitor bank alone will not eliminate the penalty. This should be verified against both IEC 60831-1 [2] und IEEE Std 18-2012 [3].
02 How Harmonics Interact with Capacitors
The impedance of a capacitor is inversely proportional to frequency:
\[Z_C = \frac{1}{j\omega C} = \frac{1}{j \cdot 2\pi f \cdot C}\]
At the 5th harmonic — 250 Hz — the capacitor impedance is one fifth of its fundamental value. At the 7th harmonic it falls to one seventh. Capacitors actively attract harmonic currents: in a network where harmonic currents circulate, the capacitor bank represents the lowest impedance path at harmonic frequencies. The harmonic current flowing into the bank is:
\[I_{C,h} = I_h \cdot \frac{Z_{System,h}}{Z_{System,h} + Z_{C,h}}\]
Als \(Z_{C,h}\) decreases with increasing harmonic order, the proportion of harmonic current flowing into the capacitor increases.
Thermal consequences
Additional harmonic current flowing through the capacitor produces losses not accounted for in the original specification. Capacitor losses at harmonic frequencies are governed by the dissipation factor \(\tan\delta\), which increases with frequency. The total losses are:
\[P_{Verlust} = \sum_{h=1}^{n} I_{C,h}^2 \cdot \frac{\tan\delta_h}{\omega_h C}\]
IEC 60831-1 [2] und IEEE Std 18-2012 [3] both specify a maximum continuous RMS current of 1.8 p.u. of rated current when the combined effects of voltage harmonics, capacitance tolerance, and operating voltage are taken into account. In installations with significant harmonic distortion this limit is frequently exceeded without any indication from conventional metering, which measures only fundamental current.
Dielectric ageing
The dominant ageing mechanism in modern metallised polypropylene film capacitors is thermal rather than electrical. The relationship between operating temperature and service life follows the Arrhenius model [7]: every 10°C rise in sustained operating temperature above the rated value approximately halves the expected service life. Harmonic currents elevate internal losses and therefore operating temperature, accelerating ageing at a rate not predictable from nameplate data alone.
This explains a field observation frequently reported but rarely understood: a capacitor bank that has operated without problems for years begins to fail after installation of new variable frequency drives, even though the fundamental reactive demand has not changed and the bank appears correctly sized by conventional criteria. The nameplate rating is met at the fundamental — but harmonic currents have elevated the internal temperature beyond the rated thermal envelope.
Voltage stress on the dielectric is a secondary ageing mechanism, more relevant to older impregnated paper or paper-film capacitors that lack the self-healing capability of metallised film technology. For modern metallised film capacitors, sustained elevated temperature is the primary life-limiting factor.
03 Parallel Resonance — The Core Problem
When a capacitor bank is connected to a distribution system, it forms a parallel resonant circuit with the inductive impedance of the network. This resonant circuit has a natural frequency at which its impedance becomes theoretically infinite — in practice, very high — and at which even small harmonic currents can produce large harmonic voltages and large circulating currents between the capacitor and the inductive elements of the network.
The parallel resonant frequency, expressed as harmonic order, ist:
\[h_r = \sqrt{\dfrac{S_{sc}}{Q_C}}\]
wo \(S_{sc}\) is the short-circuit power at the point of capacitor connection in kVA and \(Q_C\) is the capacitor bank rating in kVAr. The simplified form using only transformer rating \(S_T\) is acceptable for preliminary screening only — it overestimates \(h_r\) and is non-conservative.
Abbildung 1 — Impedance vs frequency: interactive resonance explorer
What happens at resonance
At the resonant harmonic order \(h_r\), the parallel impedance reaches a maximum. The impedance at resonance is limited only by resistive damping — the resistance of transformer windings, Kabel, and other resistive elements. In a typical industrial distribution system this damping is small, and the impedance at resonance can be 20 zu 50 times higher than the off-resonance impedance at the same frequency. The amplification factor is approximately:
\[A_h = \frac{X_{Die,h} \cdot X_{C,h}}{R \cdot |X_{Die,h} – X_{C,h}|}\]
At resonance \(X_{Die,h} = X_{C,h}\) and the denominator approaches zero — amplification is limited only by circuit resistance \(R\). In practice amplification factors of 10 zu 30 are not unusual in lightly damped industrial networks [8][9].
The apparent paradox of parallel resonance
The behaviour of a parallel resonant circuit is counterintuitive and deserves careful explanation. A field engineer looking at a capacitor bank connected to a busbar alongside a transformer might reasonably expect the capacitor to simply absorb harmonic currents — after all, its impedance decreases with frequency, making it a natural harmonic sink. This reasoning is correct away from resonance. What is not immediately obvious is what happens when the parallel combination of transformer inductance and capacitor bank is excited at its natural resonant frequency.
At resonance, the parallel LC circuit presents very high impedance to the harmonic current source — in this case the variable speed drive. The drive, acting as a current source, injects a relatively small harmonic current into the bus. This small current, jedoch, is sufficient to excite the LC tank into oscillation. Energy begins to circulate back and forth between the inductance and the capacitance at the resonant frequency — the inductor charges the capacitor, the capacitor discharges through the inductor, and the cycle repeats. The drive does not need to supply this circulating energy — it only needs to overcome the resistive losses in the circuit to sustain the oscillation.
From the outside — from the drive’s perspective — the parallel combination looks like a very high impedance. Very little current appears to enter the loop. But inside the loop, between the capacitor and the transformer inductance, the circulating current is \(Q_T\) times larger than the harmonic current injected by the drive. For a typical industrial transformer with \(Q_T\) = 30 zu 50, a drive injecting 4% of rated current as 7th harmonic can produce a circulating current of 1.2 zu 2.0 p.u. inside the LC loop — sufficient to exceed the IEC 60831-1 continuous current limit of 1.8 p.u. and operate the capacitor fuses. The capacitor is overloaded not because the drive forces large current into it directly, but because it is part of an oscillating circuit whose internal currents greatly exceed anything visible from outside the loop.
The network attraction effect
A resonant condition within a facility does not only amplify harmonics generated by local loads. The resonant circuit presents a low-impedance path — at and near the resonant frequency — that is visible from the utility network. Harmonic currents generated by other customers connected to the same distribution feeder will flow preferentially toward this low-impedance node. The facility’s capacitor bank effectively becomes a harmonic sink for the wider network, absorbing harmonic energy from sources it has no knowledge of and no control over [9][10].
This explains cases where harmonic problems at a facility cannot be fully accounted for by the harmonic sources within that facility — measured harmonic currents at the capacitor bank exceed what the facility’s own non-linear loads could plausibly generate.
Practical example
Consider a facility with: 1000 kVA transformer, 6% Impedanz; 150 MVA utility short-circuit at 11 kV; 200 kVAr capacitor bank; six 6-pulse VFDs totalling 300 kW.
Transformer short-circuit contribution:
\[S_{sc,T} = \frac{S_T}{Z_T\%} = \frac{1000}{0.06} = 16{,}667 \Text{ kVA}\]
With the 150 MVA utility bus (strong network), transformer impedance dominates: \(S_{sc} \ca. 16{,}667\) kVA. Resonant orders:
\[200 \Text{ links}: h_r = \sqrt{\dfrac{16{,}667}{200}} = 9.1 \quad \text{(safe — between h7 and h11)}\]
\[400 \Text{ links}: h_r = \sqrt{\dfrac{16{,}667}{400}} = 6.5 \quad \text{(caution — close to h7)}\]
\[500 \Text{ links}: h_r = \sqrt{\dfrac{16{,}667}{500}} = 5.8 \quad \text{(danger — within 16\% of h5)}\]
A system safe at 200 kVAr becomes dangerous at 500 kVAr — the resonance shifts with bank size.
For a weaker utility network (20 MVA at 11 kV), \(S_{sc,combined} \ca. 9{,}091\) kVA:
\[200 \Text{ links}: h_r = \sqrt{\dfrac{9{,}091}{200}} = 6.7 \quad \text{(now close to h7)}\]
\[400 \Text{ links}: h_r = \sqrt{\dfrac{9{,}091}{400}} = 4.8 \quad \text{(below h5 — full danger zone)}\]
04 Failure Modes and Field Symptoms
The interaction between power factor correction capacitors and harmonic currents manifests in field symptoms that are frequently misdiagnosed because their root cause — harmonic resonance or harmonic overloading — is not visible to conventional instrumentation.
Capacitor fuse operations
The most common visible symptom of harmonic overloading is repeated operation of capacitor fuse elements. Fuse operations that recur after replacement, occur without identifiable load fault, or happen preferentially at certain times of day are a strong indicator of harmonic overcurrent. A harmonic-related operation leaves the capacitor unit physically undamaged and recurs after replacement because the harmonic condition that caused it has not been addressed. Fuse ratings must account for total RMS current including harmonic components per IEC 60831-1 [2] und IEEE Std 18-2012 [3].
Capacitor case bulging and rupture
Physical deformation of capacitor cases indicates internal pressure buildup caused by excessive internal heating. In a harmonic environment this failure mode is associated with sustained thermal overload. Case rupture is a serious safety event — a bank experiencing repeated case deformations should be taken out of service immediately pending a harmonic assessment.
Nuisance tripping of overcurrent protection
Overcurrent relays and circuit breakers may trip repeatedly without apparent load fault. A parallel resonance condition generates large circulating currents between the capacitor bank and the transformer that flow through the protection equipment even when load current is normal. Distinguishing between resonance-related and switching-transient-related trips requires power quality measurement at the time of the event. Resonance produces sustained elevated current at a specific harmonic frequency; switching transients produce a short-duration high-frequency oscillation at the moment of switching [9][10].
Transformer overheating
Unexplained transformer overheating in the absence of apparent overload is a classic symptom of harmonic circulating currents. Resonance drives large harmonic currents through the transformer secondary windings in a closed loop with the capacitor bank, producing additional copper losses and elevated core losses at harmonic frequencies. Transformer harmonic loading is quantified by the K-factor — a transformer whose K-factor rating is exceeded will run at elevated temperature even when fundamental load current is within rated limits.
Neutral conductor overloading
In four-wire installations with a mix of three-phase and single-phase non-linear loads, triplen harmonics (3rd, 9th, 15th) are zero-sequence in nature and add arithmetically in the neutral conductor rather than cancelling. This can cause neutral conductor overheating often misattributed to load unbalance rather than harmonics. The presence of significant triplen harmonic content changes the harmonic spectrum seen by the capacitor bank and may require a detuning factor of p = 14% rather than the standard p = 7% [4].
Harmonic voltage distortion and equipment interference
Elevated voltage harmonic distortion at the busbar supplying the capacitor bank is a direct indicator of resonance amplification. A characteristic signature of capacitor-related resonance is a harmonic voltage spectrum with a pronounced peak at one specific harmonic order — disproportionately large relative to the harmonic current injected by non-linear loads. This distortion can also cause malfunction of sensitive electronic equipment — PLCs, drive control boards, metering, and communications systems.
Measurement approach for diagnosis
When any of the above symptoms are observed, the diagnostic sequence should follow the measurement methodology of IEC 61000-4-30 Klasse A [6]: simultaneous measurement of voltage and current harmonics at the capacitor bank connection point and at the transformer secondary busbar, over a period of at least 24 hours covering all operating modes, capturing individual harmonic components to at least the 50th order with phase angle information.
05 Detuned Capacitor Banks
A detuned capacitor bank prevents resonance by connecting a series reactor with each capacitor unit, shifting the resonant frequency of the reactor-capacitor combination to a point below the lowest characteristic harmonic of concern. A series reactor connected in series with a capacitor forms a series resonant circuit. Below this series resonant frequency the combination behaves capacitively. Above it the combination behaves inductively, presenting increasing impedance to harmonic currents.
The series resonant frequency is expressed as a tuning factor \(p\):
\[p = \left(\frak{f_r}{f_1}\Recht)^2 = \frac{X_L}{X_C} \times 100\% \qquad h_r = \dfrac{1}{\sqrt{p}}\]
Standard tuning factors
IEC 61642:2020 [4] recognises several standard tuning factors:
| Tuning factor p | Series resonant frequency (50 Hz) | Harmonic um | Typical application |
|---|---|---|---|
| 14% | 133 Hz | h = 2.68 | Networks with significant 3rd harmonic content |
| 7% | 189 Hz | h = 3.78 | Standard industrial networks — protection against 5th and above |
| 5.67% | 210 Hz | h = 4.20 | Networks where 3rd harmonic is absent |
| 3.8% | 256 Hz | h = 5.13 | Not recommended — too close to 5th harmonic |
The most widely used tuning factor in European industrial practice is p = 7%, placing the series resonant frequency at 189 Hz — safely below the 5th harmonic at 250 Hz with sufficient margin for component tolerances [4].
Effect on reactive power output
The series reactor reduces net reactive power output and elevates the voltage across the capacitor:
\[Q_{net} = Q_C \times (1 – p) \qquad V_C = V_{supply} \times \frac{1}{1-p}\]
For a 200 kVAr capacitor with p = 7%: \(Q_{net} = 186\) links, \(V_C = 430\) IN. Standard detuned units are manufactured with elevated voltage ratings — typically 440 Oder 480 V for use on 400 V networks [2][4].
Switching transient advantage
When a detuned step is energised, the series reactor limits inrush current — significantly reducing switching transients compared to an undetuned bank. Automatic power factor controllers must switch complete reactor-capacitor units. Switching a capacitor without its associated reactor creates an unprotected capacitor directly on the network [4][13].
What detuning does and does not achieve
Component ratings depend on the accuracy of reactor and capacitor values. Capacitance tolerance under IEC 60831-1 [2] is ±5% for individual units. Inductance tolerance under IEC 60076-6 [15] is typically ±3%. This is why a tuning factor of 3.8% is not recommended — manufacturing tolerances could shift the actual resonant frequency above 250 Hz, directly at the 5th harmonic.
06 Passive Harmonic Filters
This continuum is explicitly recognised in both IEEE Std 1531-2003 [14] und IEC 61642:2020 [4]. A capacitor bank designed for power factor correction alone will function as an unintended harmonic filter — absorbing harmonic currents it was never rated to carry, overheating, and failing. A passive harmonic filter that is correctly designed will simultaneously correct displacement power factor and filter harmonics. The design question is therefore never “do I need power factor correction or a harmonic filter?” It is always “how do I design the reactive compensation for this installation to safely and correctly perform both functions in the harmonic environment that actually exists?”
A shunt passive harmonic filter consists of a series-connected reactor and capacitor tuned to present minimum impedance — series resonance — at the target harmonic frequency. It is connected in parallel with the load so that harmonic currents flow preferentially through the low-impedance filter path rather than back into the supply network.
The series resonant frequency is:
\[h_{tuned} = \dfrac{1}{2\pi f_1 \sqrt{LC}}\]
In practice the filter is deliberately tuned slightly below the target harmonic order — typically at 4.7 statt 5.0 for a 5th harmonic filter. This detuning margin prevents a new parallel resonance from coinciding with the target harmonic: a filter tuned at 4.7 presents capacitive impedance at h = 5.0, which combined with the network inductance creates a parallel resonance unten h = 5.0 rather than at it, keeping the dangerous resonance away from the characteristic harmonic [14]. The tuning margin is:
\[f_{tuned} \ca. 0.94 \times h_{target} \times f_1\]
Quality factor and reactive power
The effectiveness of the filter depends on the quality factor Q:
\[Q = \frac{X_L}{R} = \frac{\omega_{tuned} Die}{R}\]
A higher Q factor means lower filter resistance and better harmonic attenuation, but a sharper tuning characteristic — more sensitive to component tolerances and ageing. Practical Q factors range from approximately 30 zu 100 [9][14]. The fundamental reactive power contribution of the filter is:
\[Q_{filter} = \frac{V^2 \cdot \omega_1 C}{1 – \links(\frak{f_1}{f_{tuned}}\Recht)^2} \ca. 1.047 \times V^2 \cdot \omega_1 C\]
Filter types
Ein single-tuned filter — one reactor-capacitor branch tuned to one harmonic frequency — is the simplest and most common configuration. A complete installation for a 6-pulse drive system typically requires at least two branches: one near the 5th harmonic and one near the 7th. Each branch must be designed accounting for interaction between branches — the 5th harmonic filter affects the impedance seen by the 7th harmonic and vice versa. A combined design approach using network simulation software is required [9][10][14].
Ein double-tuned filter provides attenuation at two harmonic frequencies using a single four-element circuit. More common at medium and high voltage where the cost of multiple switching devices is significant.
Ein C-type filter minimises fundamental frequency losses by placing a capacitor in series with the reactor such that the reactor-series capacitor combination resonates at the fundamental frequency, effectively bypassing the reactor at 50 Hz while retaining its impedance at harmonic frequencies. More commonly found in large arc furnace compensation systems and HVDC converter stations [9][14].
Abbildung 4 — Four technologies: frequency response comparison
Limitations of passive harmonic filters
Passive filters are effective and economical for stable harmonic environments dominated by characteristic harmonics of 6-pulse rectifier loads. Their principal limitations are: performance is load-dependent; performance changes with network impedance; they can create new resonance conditions at frequencies slightly below each tuning point; they provide no attenuation for non-characteristic harmonics or interharmonics; and they cannot independently optimise reactive power correction and harmonic filtering. These limitations explain why the combination of passive and active filtering offers performance advantages that neither technology achieves alone [11][12].
07 Aktive Oberschwingungsfilter
An active harmonic filter measures the harmonic content of the load current in real time and injects equal and opposite harmonic currents into the network, cancelling harmonics at the point of connection by superposition. It operates as a controlled current source:
\[I_{supply} = I_{einlegen} + I_{AHF}\]
A current transformer or Rogowski coil measures the total load current. A digital signal processor identifies the magnitude and phase angle of each harmonic component. A pulse-width modulated voltage source inverter — built around IGBTs — injects the compensating current [11][12]:
\[I_{AHF} = -\sum_{h=2}^{n} I_{h}\]
Modern active harmonic filters compensate harmonics to the 50th order with residual THD below 5% at rated load.
Abbildung 6 — Active filter operating principle: waveform cancellation
Simultaneous reactive power compensation
Most modern active harmonic filter designs also inject a fundamental frequency reactive current component, acting as a static VAR compensator. In installations requiring both significant reactive power correction and substantial harmonic mitigation, the combined demand may exceed the capacity of a single active filter unit — in which case the combination of a passive filter for bulk reactive power and harmonic attenuation with an active filter for residual correction becomes the optimal solution, developed in the subsequent article in this series [11][12][13].
Advantages over passive filters
The active harmonic filter adapts automatically to changes in the harmonic spectrum, creates no resonance risk, compensates non-characteristic harmonics and interharmonics simultaneously, provides precise control of compensation level, and is largely independent of network impedance changes [11][12].
Begrenztheit
Active harmonic filters are rated in amperes of harmonic current, not kVAr — in a facility with large absolute harmonic currents the required rating and capital cost can be significant. Performance degrades at harmonic orders approaching the control bandwidth limit (typically effective to the 50th harmonic at 50 Hz). They require stable network voltage — most modern units tolerate THDIN up to 10–15% at the connection point [11][12]. They introduce switching frequency harmonic components into the network, typically attenuated by an output LCL filter.
Placement relative to passive elements
In installations where both passive and active filters are present, the active filter should be connected at the same busbar as the passive filter, on the source side of the passive filter branches. This allows the active filter to cancel residual harmonic currents that the passive filter does not fully absorb, and to eliminate the risk of parallel resonance between the passive filter branches and the network impedance [11][12][13].
08 Selection Guide — Choosing the Right Solution
Primary selection criteria
The selection process is driven by five questions: (1) What is the objective — PF correction, harmonische Schadensbegrenzung, or both? (2) What is the harmonic environment — measured THDIN and THDIch with individual harmonic spectrum per IEC 61000-4-30 Klasse A [6]? (3) Is the harmonic load fixed or variable? (4) What are the utility requirements — applicable standard, PCC definition, and metering basis [1][5]? (5) What is the short-circuit level at the point of connection — required for resonance calculation and formal assessment [4][5]?
Abbildung 7 — Selection decision flowchart
Technology comparison summary
| Characteristic | Standard capacitor | Detuned bank | Passive filter | Active filter |
|---|---|---|---|---|
| Reactive power correction | Yes | Yes (−p%) | Yes | Yes (limited) |
| Harmonic filtering | Nicht | Nicht | Yes (tuned orders) | Yes (full spectrum) |
| Resonance risk | Hoch | Eliminated | New parallel resonance possible | None |
| Load variability tolerance | N/A | N/A | Poor | Ausgezeichnet |
| Effective harmonic orders | N/A | N/A | Fixed (designed) | 2nd to 50th |
| Standards compliance path | Displacement PF only | Displacement PF only | IEEE 519 / IEC 61000-3-6 possible | IEEE 519 / IEC 61000-3-6 |
| Relative capital cost | Niedrig | Low–Medium | Medium | Hoch |
| Applicable standard | IEC 60831-1 [2] | IEC 61642 [4] | IEEE 1531 [14] | IEEE 519 [1] |
Practical example
A food processing plant: 1600 kVA transformer, 6% Impedanz; 200 MVA utility at 11 kV; 400 kVAr undetuned capacitor bank; twelve 6-pulse VFDs totalling 500 kW (approximately 40% of total kVA); measured THDIch = 32%, THDIN = 7.8%; IEEE 519-2022 compliance required; symptoms: repeated capacitor fuse operations, transformer running 15°C above normal.
Resonance check: \(S_{sc} \ca. 1600/0.06 = 26{,}667\) kVA; \(h_r = \sqrt{26{,}667\,/\,400} = 8.2\) — not at a characteristic order, but THDIN von 7.8% and transformer overheating are consistent with near-resonance amplification. The existing undetuned bank must be replaced or detuned.
Applying the decision flowchart: non-linear load proportion 40% → detuned bank mandatory; IEEE 519 compliance required; load profile variable (VFDs at varying speed) → active filter preferred.
Empfehlung: Option D — detuned bank (p = 7%) for reactive power correction combined with an active harmonic filter for harmonic mitigation. The variable load profile and utility compliance requirement make an active filter the preferred technology; the detuned bank handles reactive correction economically and safely without harmonic risk.
Economic considerations
The capital cost of harmonic mitigation varies significantly. Passive filters have lower capital cost but may require periodic retuning as components age. Active filters have higher capital cost but adapt automatically to load changes. The increasing availability of real-time power quality monitoring — both as utility-provided services and from independent monitoring providers — changes the economics of continuous compliance verification, making it increasingly feasible to verify that the installed solution continues to perform as designed as the load profile evolves [10][13]. In many industrial installations the cost of a single transformer replacement or production interruption caused by harmonic-related failure exceeds the capital cost of a properly specified active harmonic filter.
Abschluss
Power factor correction capacitors and harmonic distortion are not independent subjects that can be addressed sequentially — they are deeply coupled, and decisions made about one directly determine the consequences of the other. In any electrical installation where non-linear loads represent a meaningful proportion of total demand, power factor correction cannot be specified independently of harmonic mitigation.
The progression from standard capacitor banks through detuned banks, passive filters, and active filters represents increasing capability at increasing cost and complexity. The correct point on this progression depends on the harmonic environment, the load variability, the utility requirements, and the economic context — not on a fixed rule based on drive horsepower ratings or arbitrary technology preferences.
A detuned capacitor bank is a protective measure, not a mitigation measure. Passive harmonic filters are effective and economical for stable harmonic environments dominated by characteristic harmonics of 6-pulse rectifier loads. Active harmonic filters eliminate resonance risk and adapt to variable harmonic spectra. The combination of a detuned capacitor bank for reactive power correction and an active harmonic filter for harmonic mitigation represents the optimal solution for many modern industrial installations — developed in detail in the next article in this series.
The role of measurement cannot be overstated. The harmonic environment of an industrial facility is not static. Periodic power quality monitoring, consistent with IEC 61000-4-30 [6], is the only reliable way to ensure that the installed mitigation solution continues to perform as designed throughout the life of the installation.
Referenzen
- IEEE Std 519-2022, IEEE Standard for Harmonic Control in Electric Power Systems, IEEE, 2022.
- IEC 60831-1:2014, Shunt power capacitors of the self-healing type for a.c. systems having a rated voltage up to and including 1 000 V — Part 1: Allgemeine, IEC, 2014.
- IEEE Std 18-2012, IEEE-Standard für Shunt Leistungskondensatoren, IEEE, 2012.
- IEC 61642:2020, Industrial networks — Guide for application of capacitors and harmonic filters, IEC, 2020.
- IEC 61000-3-6:2008, Electromagnetic Compatibility — Limits — Assessment of emission limits for the connection of distorting installations to MV, Hoch-und Höchstleistungssysteme, IEC, 2008.
- IEC 61000-4-30:2015, Electromagnetic Compatibility — Testing and measurement techniques — Power quality measurement methods, IEC, 2015.
- IEC 60216 Serie, Electrical insulating materials — Thermal endurance properties, IEC.
- Girgis, A.A., Fallon, C.M., Catoe, R.C., Rubino, C.P., “Harmonics and transient overvoltages due to capacitor switching,” IEEE Transactions on Industry Applications, Flug. 28, KEIN. 1, pp. 196-204, 1992.
- Arrillaga, J., Watson, N.R., Power System Harmonics, 2nd Ed., John Wiley & Sons, 2003.
- Dugan, R.C., McGranaghan, M.F., Santoso, S., Beaty, H.W., Electrical Power Systems Quality, 3rd ed., McGraw-Hill, 2012.
- Singh, B., Al-Haddad, K., Chandra, A., “A review of active filters for power quality improvement,” IEEE Transactions on Industrial Electronics, Flug. 46, KEIN. 5, pp. 960-971, 1999.
- Akagi, H., “Aktive Oberwellenfilter,” Proceedings of the IEEE, Flug. 93, KEIN. 12, pp. 2128-2141, 2005.
- ABB Technical Application Paper No. 8, Power Factor Correction and Harmonic Filtering in Electrical Plants, ABB SACE, 2008.
- IEEE Std 1531-2003, IEEE Guide for Application and Specification of Harmonic Filters, IEEE, 2003.
- IEC 60076-6:2007, Power transformers — Part 6: Reactors, IEC, 2007.
- IEC 61000-3-4:1998, Electromagnetic Compatibility — Limits — Limitation of emission of harmonic currents in low-voltage power supply systems for equipment with rated current greater than 16 Ein, IEC, 1998.
Mit KI-Unterstützung erstellte und vom Autor validierte Inhalte anhand von 30 Jahrelange Erfahrung im Bereich Power Quality.
© 2026 Denis Ruest — International Power Quality Discussion Forum (IPQDF). Reproduction permitted for non-commercial educational purposes with full attribution to the author and a link to the original article at ipqdf.com.