Origine des harmoniques à une entrée électrique

Introduction

Variateurs de fréquence (VFD) based on the 6-pulse rectifier topology are among the most widely deployed power conversion devices in industrial applications. Their inherent non-linear input characteristic makes them a significant source of harmonic distortion on electrical distribution systems. While the theoretical harmonic spectrum of a 6-pulse rectifier is well established and commonly described by the \(1/n\) amplitude model [1], practical measurements consistently reveal meaningful deviations from this ideal behaviour — deviations that carry real consequences for system design, filter sizing, and compliance with harmonic standards such as IEEE 519 [2] et CEI 61000-3-6 [3].

Cet article présente une analyse comparative des spectres harmoniques théoriques et pratiques d'un VFD à 6 impulsions, examiner les ordres harmoniques 3 à travers 50 dans quatre conditions de charge (25%, 50%, 75% et 100% de charge nominale). Magnitude harmonique, angle de phase, et la séquence sont discutées, et les limites du modèle d'injection de courant idéal sont examinées à la lumière du comportement réel du système.

01 Contexte théorique

Un redresseur à 6 impulsions se compose d'une diode pleine onde triphasée ou d'un pont à thyristors produisant six impulsions de courant par cycle fondamental.. Dans des conditions idéales : une alimentation triphasée parfaitement équilibrée, une charge DC purement inductive produisant un courant DC parfaitement lisse, et dispositifs de commutation idéaux - la forme d'onde du courant alternatif est une onde quasi carrée dont la décomposition de Fourier ne contient que des ordres harmoniques spécifiques [1].

Harmoniques caractéristiques

These characteristic harmonics follow the relationship:

This yields harmonic orders 5, 7, 11, 13, 17, 19, 23, 25 et ainsi de suite. The theoretical amplitude of each characteristic harmonic relative to the fundamental is given by:

Où \(I_h\) is the RMS magnitude of the \(h\)-th harmonic current, \(I_1\) is the RMS magnitude of the fundamental current, et \(h\) is the harmonic order. This gives a 5th harmonic of 20% des droits fondamentaux, a 7th of 14.3%, an 11th of 9.1%, et ainsi de suite [1][4].

The total harmonic distortion under the ideal model is:

For a 6-pulse rectifier with purely inductive DC load this converges to approximately 28.6% [4].

Under this ideal model, all even harmonics and all triplen harmonics (3e, 9e, 15e, 21st…) are theoretically absent from the line currents. Triplen harmonics are zero-sequence — all three phases carry them with identical phase angles — and in a balanced three-phase system they cannot circulate in the line conductors. Even harmonics are suppressed by the half-wave symmetry of the rectifier waveform:

Harmonic sequence

The characteristic harmonics follow a defined sequence pattern with direct implications for rotating machinery and power system behaviour:

Negative-sequence harmonics rotate in opposition to the fundamental, producing reverse torque effects in induction motors and contributing to rotor heating. Positive-sequence harmonics rotate in the same direction as the fundamental [4][5].

02 Practical Harmonic Spectra — Deviations from the Ideal Model

En pratique, the conditions required by the ideal \(1/n\) model are never fully met. The most significant departure from idealised behaviour in a modern VFD is the replacement of the inductive DC load assumption with a large electrolytic capacitor on the DC bus. Rather than drawing a smooth continuous DC current, a capacitor-fed rectifier draws current only during the intervals when the instantaneous supply voltage exceeds the DC bus voltage, producing narrow, high-amplitude current pulses [6].

Figure 1 — Waveform comparison

The Fourier decomposition of the peaked waveform reveals two systematic deviations from the \(1/n\) model. À lower harmonic orders (5th and 7th), practical magnitudes exceed or approach ideal values, driven by the narrow pulse shape concentrating energy in lower-frequency components. À higher harmonic orders (17th and above), the opposite dominates — AC-side inductance and finite pulse rise time attenuate these components more rapidly than \(1/n\) predicts. The crossover typically occurs between the 11th and 13th harmonic [4][6].

This behaviour is expressed by introducing a correction factor \(k_h\) to the ideal model:

Où \(k_h > 1\) for lower-order harmonics, \(k_h < 1\) for higher-order harmonics, and \(k_h \approx 1\) near the 11th–13th. The value of \(k_h\) varies with load level, DC bus capacitance, and AC-side impedance [7].

Phase angle of harmonic currents also shifts with load, reflecting the changing commutation overlap angle \(\mu\) governed by:

Où \(\omega\) is the angular frequency, \(L_s\) is the AC-side inductance per phase, \(I_d\) is the DC load current, et \(V_{LL}\) is the line-to-line supply voltage [5][8].

Figure 2 — Harmonic spectrum: ideal vs practical at 100% charge

Table 1 — Harmonic magnitude and phase: h3 to h50 across all load conditions

The table below covers harmonic orders 3 à travers 50 at four load levels, showing both magnitude (% des droits fondamentaux) and phase angle (°) for each. Characteristic harmonics are highlighted. Values are practical estimates based on published drive measurements — see Section 2 for methodology.

Table 2 — Idéal vs pratique à 100% charge: h3 à h50

Ce tableau compare l'idéal 1/n l'amplitude du modèle par rapport à la valeur pratique estimée à 100% charge pour chaque ordre harmonique, avec classification de séquence et différence signée. Les harmoniques où les valeurs pratiques dépassent l'idéal sont marquées ▲; ceux qui roulent plus vite que l'idéal sont marqués ▼.

03 Interaction du système et limites du modèle source actuel

L'analyse harmonique dans les systèmes électriques s'appuie traditionnellement sur le modèle d'injection de source actuel., dans lequel chaque charge non linéaire est représentée comme une source de courant idéale injectant des courants harmoniques fixes dans le réseau au point de couplage commun (PCC). Ce modèle sous-tend la méthodologie d'évaluation harmonique des deux IEEE 519 [2] et CEI 61000-3-6 [3]. Cependant, the current source model is a significant simplification of the actual behaviour of a 6-pulse VFD.

Figure 3 — Norton equivalent of a 6-pulse VFD on a distribution network

Supply impedance dependence

A true current source is independent of the impedance of the network into which it injects. A 6-pulse drive is not. Une 3% line reactor typically reduces 5th harmonic current from approximately 18% à 12% of fundamental at full load [6][7]. The Norton equivalent formulation captures this dependency:

Résonance

Parallel resonance between capacitor banks and supply inductance creates high-impedance nodes at specific harmonic frequencies. The resonant frequency is:

Où \(S_{sc}\) is the short-circuit power at the PCC and \(Q_c\) is the reactive power of the capacitor bank [9].

Multiple drive interaction

Arithmetic addition of individual drive harmonic spectra consistently overestimates actual distortion at the PCC [2][3]. CEI 61000-3-6 addresses this through a summation law:

Table 3 — IEC 61000-3-6 summation exponent α by harmonic order

In systems dominated by a single drive type, arithmetic summation (\(\alpha = 1\)) may be more representative than \(\alpha = 1.4\) for characteristic orders. Engineering judgement and where possible actual measurement remain essential [2][3].

04 Practical Implications and Mitigation

Transformer and cable sizing

Harmonic currents increase the RMS line current above the fundamental value:

Transformers supplying non-linear loads must be evaluated using the K-factor:

A typical 6-pulse drive installation without mitigation may present a K-factor of 4 à 8 depending on load level and system impedance [6][9].

Neutral conductor loading

Triplen harmonics are zero-sequence and circulate freely in the neutral conductor of four-wire systems. Installations combining VFDs with single-phase non-linear loads can produce significant neutral currents at the 3rd and 9th harmonic. The neutral conductor must be sized accordingly [9].

Motor and connected load considerations

Negative-sequence harmonics — the 5th, 11e, 17th and higher following the \(6k-1\) pattern — produce counter-rotating magnetic fields in the air gap, generating braking torque and elevated rotor temperature. Inverter-rated motors conforming to NEMA MG1 Part 31 or IEC 60034-25 incorporate design features that improve tolerance to harmonic content and are the recommended choice for all VFD applications. A detailed treatment of motor harmonic impedance, rotor loss mechanisms, and derating methodology is reserved for a subsequent article in this series.

Mitigation strategies

The tuning frequency of a passive filter is deliberately set below the target harmonic to avoid series resonance:

05 Measurement Considerations and Interpretation of Field Results

Instrument requirements

Harmonic measurement requires a power quality analyser capable of resolving individual harmonic components to at least the 50th order, implementing a synchronised DFT with a rectangular window of exactly 10 cycles (200 ms at 50 Hz) as specified by IEC 61000-4-7 [10]. Rogowski coils are generally preferred for harmonic work above the 25th order due to their superior frequency response and absence of core saturation.

Measurement point selection

For compliance assessment against IEEE 519 [2] or IEC 61000-3-6 [3], measurement must be performed at the PCC as defined in those standards. Recording simultaneously at the drive input and the PCC provides direct information about the harmonic impedance of the intervening network — valuable for resonance risk assessment.

Operating condition during measurement

CEI 61000-3-6 recommends that harmonic assessment be based on the 95th percentile of measured values over a representative observation period — typically one week [3]. Where continuous monitoring is not practical, measurements should be taken at a minimum of three load points spanning the expected operating range.

Interharmoniques

Modern VFDs may generate interharmonic currents — components at non-integer multiples of the fundamental — particularly during speed ramps and transient operating conditions. CEI 61000-4-7 defines the measurement methodology using sub-group analysis with a 200 ms window [10]. Their presence should be noted as they can contribute to flicker, ripple control interference, and sub-synchronous torque oscillations.

Emission studies and compliance with utility limits

Most utilities will not accept field measurements alone as the basis for a connection approval or compliance demonstration. A formal harmonic impact study, conducted in accordance with the utility’s accepted methodology and submitted prior to commissioning, is the standard requirement in the majority of jurisdictions [2][3]. The utility’s need to assess cumulative impact on all customers connected to the same network is fundamental to the IEC 61000-3-6 framework, which allocates emission limits based on the agreed power of the installation relative to the short-circuit capacity of the network [3].

Conclusion

The ideal \(1/n\) amplitude model systematically misrepresents the harmonic spectrum of a modern capacitor-fed 6-pulse drive. Lower-order characteristic harmonics are more load-sensitive than the model predicts; higher-order harmonics roll off faster. The crossover occurs near the 11th–13th harmonic. THD varies from approximately 22% at full load to 45% or more at 25% load — a range that spans the boundary between compliant and non-compliant for many utility connection agreements.

The representation of a 6-pulse drive as an ideal harmonic current source breaks down in the presence of supply impedance variation, background voltage distortion, network resonance, and multi-drive interaction. The Norton equivalent provides a more faithful description, and the frequency dependence of both \(Z_h(h)\) et \(Z_\text{système}(h)\) must be accounted for in any rigorous analysis.

For compliance studies submitted to electrical utilities, field measurement alone is unlikely to be accepted. A formal harmonic impact study is the standard requirement. High-fidelity simulation tools produce spectra significantly closer to measured field values, reducing the risk of unnecessary mitigation measures and over-designed filter solutions. The three-stage approach — theoretical screening, high-fidelity simulation, and post-commissioning measurement — provides a proportionate and technically defensible framework across the full project lifecycle.

Références

1. Mohan, N., Undeland, T.M., Robbins, W.P., Power Electronics: Convertisseurs, Applications et conception, 33e éd., John Wiley & Sons, 2003.
2. IEEE Std 519-2022, Norme IEEE pour le contrôle des harmoniques dans les systèmes d'alimentation électrique, IEEE, 2022.
3. CEI 61000-3-6:2008, Electromagnetic Compatibility — Limits — Assessment of Emission Limits for the Connection of Distorting Installations to MV, HV and EHV Power Systems, CEI, 2008.
4. Arrillaga, J., Watson, N.R., Harmoniques Power System, 2nd ed., John Wiley & Sons, 2003.
5. Audacieux, JE., Vautour, S.A., Le manuel de la machine à induction, CRC Press, 2002.
6. Skibinski, G., Ecclésiastique, R., Schlegel, D., “EMI emissions of modern PWM AC drives,” Magazine des applications industrielles de l'IEEE, vol. 5, pas. 6, pp. 47–81, 1999.
7. Rockwell Automation, Harmonics and IEEE 519, Application Guide DRIVES-AP001A, 2013.
8. Moreira, J.C., Lipo, T.A., “Modeling of saturated AC machines including air-gap flux harmonic components,” IEEE Transactions on Industry Applications, vol. 28, pas. 2, pp. 343–349, 1992.
9. Dugan, R.C., McGranaghan, M.F., Santoso, S., Beaty, H.W., Electrical Power Quality Systems, 33e éd., McGraw-Hill, 2012.
10. CEI 61000-4-7:2002+A1:2008, Electromagnetic Compatibility — Testing and Measurement Techniques — General Guide on Harmonics and Interharmonics Measurements and Instrumentation, CEI, 2008.
11. CEI 61000-4-30:2015, Electromagnetic Compatibility — Testing and Measurement Techniques — Power Quality Measurement Methods, CEI, 2015.