Harmonic Headroom: Assessing G5/5 Compliance Before You Add Bosch Drives

Few plant upgrades stall for mechanical reasons. The delay usually arrives as a letter from the distribution network operator (DNO) asking for a harmonic assessment before the new connection, or the increased capacity, can be energised. Engineering Recommendation G5 Edition 5 — published in 2022 and in force since 2023 — formalised that conversation. It gives the DNO a structured method and gives plant engineers a reasonably predictable route through it, provided the assessment is scoped sensibly and the evidence is honest.

What G5/5 asks of you in practice

The recommendation works in stages. Broadly, Stage 1 covers installations small enough to connect without formal study. Stage 2 is a desk exercise: you compare the harmonic current your converter load will draw against permitted values tabulated by fault level and converter rating. Stage 3 is the full assessment, where measured or calculated harmonic voltages are compared with the planning levels at the point of common coupling.

The planning level for 400 V networks sits at 5% total harmonic distortion, with individual limits for the low-order odd harmonics. For a six-pulse diode front end — still the most common rectifier in general-purpose drive installations — the 5th and 7th dominate, and the 5th is usually the one that decides whether your application passes. Add a modest amount of existing background distortion on a lightly loaded urban board and the headroom disappears faster than the arithmetic suggests.

Start with the connection agreement, not the spreadsheet

Before any calculation, obtain the current connection agreement and the DNO's stated fault level at your intake. Fault level is the denominator in almost every G5/5 comparison, and engineers routinely use a figure from an old single-line diagram that no longer reflects the network. Reinforcement, reconfiguration and the addition of neighbouring low-carbon loads all move it. Ask the DNO to confirm the value in writing, and ask what they consider the minimum fault level for assessment purposes — G5/5 is generally applied at the lower end of the range, which is the conservative and correct choice.

Record the agreed point of common coupling at the same time. On multi-building sites this is frequently misidentified, which leads to assessments carried out at the wrong busbar entirely.

Choosing measurement points

Where the existing installation already has distortion, a short survey beats any assumption. Use a Class A analyser compliant with BS EN 61000-4-30, connected at the intake for at least seven days, with ten-minute aggregation and harmonics recorded to at least the 25th order. In parallel, log at the busbar that will feed the new drives, so the incremental contribution can be separated from the site's existing background.

Two practical points are often missed. First, log current as well as voltage on the drive feeders, because the harmonic current spectrum is what you will scale to the new load. Second, timestamp against the plant's production schedule. A survey dominated by a shutdown weekend tells you almost nothing about the operating condition the DNO will care about.

The diversity assumption that skews the result

Two opposite errors appear in almost every submission.

The first is arithmetic summation. Engineers add the harmonic currents of every drive at rated output and assume they arrive in phase. In reality, phase angles differ between units and between operating points, so simple addition overstates the total. The result is an oversized filter and an unnecessary capital spend.

The second error is a blanket diversity factor — commonly 0.5 — applied without evidence. Applied to a plant where conveyors, pumps and process drives genuinely run together for most of the shift, that factor understates the load and produces a submission the DNO is entitled to reject.

What replaces both is a load profile. Take the surveyed coincident demand, map it against the names of the drives you intend to install, and apply a diversity factor derived from the process itself: which axes run simultaneously, at what proportion of rated torque, and for how long. For a drive group with motors that are only ever partly loaded, remember that harmonic current emission is not linear with load — it is typically modest at light load, peaks somewhere around half to three-quarters of rated output, and then falls again. Sizing on the worst-case operating point of a representative drive, weighted by a real coincidence factor, is far more defensible than any rule of thumb.

Reactors, passive filters or active correction

Once the shortfall is known, the mitigation choice is usually a balance of capital cost, footprint and future flexibility.

A 4% line reactor on each drive is the cheapest first move and typically takes a useful bite out of the 5th and 7th. It also improves the drive's immunity to supply transients and reduces DC-link capacitor stress, which is a maintenance benefit in its own right.

A tuned passive filter at the busbar is more effective where a single dominant order is the problem, but it is fixed in design, carries a risk of resonance with network changes, and needs protection against capacitor failure.

Active harmonic correction is the flexible answer: it handles multiple orders, adapts as load changes, and can be sized to the measured shortfall rather than the worst conceivable case. It costs more and adds a maintenance item, so it is best applied where a site is expected to grow again.

Where the drives themselves are still being specified, it is worth considering supply modules with integrated line-side filtering or regenerative capability, which reduce emission at source and, on vertical or high-deceleration duty, recover energy rather than dissipating it.

Presenting the case to your DNO

A submission that is accepted first time usually contains the same elements: the confirmed fault level and PCC; a single-line diagram showing existing and proposed converters; the survey data with methodology and instrument calibration certificate; the calculation, clearly showing whether it is Stage 2 or Stage 3; the mitigation selected and the resulting predicted distortion; and a short statement of the assumptions, including diversity. State the assumptions plainly. A DNO will often accept a conservative assumption that is disclosed and will rarely accept an unstated one discovered later.

Decision checklist for 2026 capacity upgrades

• Confirm the fault level and point of common coupling in writing before designing anything.
• Survey the intake for seven days at Class A, including a representative production week.
• Record the harmonic current spectrum of existing drives, not just voltage distortion.
• Derive diversity from the process schedule, and document it.
• Size mitigation on the worst-case operating point, then check it against the full cycle.
• Prefer mitigation that leaves headroom for the next phase of load.
• Keep the assessment on file — it is the baseline for the next upgrade.