Harmonic Studies and Filter Design for Transmission-Connected Solar and BESS

A harmonic study for a transmission-connected solar farm or battery predicts the harmonic voltage distortion the plant will cause at its connection point, across every network condition the transmission network service provider (TNSP) considers credible, and compares it with the plant's allocated emission limits. When the plant exceeds its allocation, the study becomes a filter design exercise. Done well, it produces a filter that is buildable, robust to component tolerance, and consistent with the plant's reactive power design.

Our article on AS/NZS 61000.3.6 emission allocation covers how limits are allocated. This article is about the study and filter design that follow.

The Inputs

InputSourceWatch out for
Emission limits per harmonic orderNSP allocation for the connection pointLetters can have a validity period; confirm the current issue
Network harmonic impedanceNSP impedance polygons or loci per orderPolygons are reissued when the network changes
Background distortionMeasurement, or an NSP-specified assumptionOften already above the plant's allocation at some orders
Inverter harmonic emissionsOEM data, as current sources or Norton equivalentsState the base: amperes or percent of rated current
Transformer dataManufacturer data at the principal tapExcitation harmonics are a separate input
Collector cables and filtersDesign dataPlus or minus tolerance on capacitance and length

How the Study Is Built

The plant is modelled in detail, inverters, transformers, collector cables, auxiliary transformers and any filters, and connected to the network represented by its harmonic impedance at each order.

Impedance polygons and percentiles

TNSPs usually describe the network impedance at each harmonic order as a polygon or locus in the R-X plane, covering the range of credible network conditions. The older approach reported the worst result over the polygon vertices. A more recent approach, aligned with the 2012 revision of the IEC technical report, sweeps a large number of equidistant points inside each polygon, treats them as equally likely, and reports the 95th percentile result. The two methods can give materially different answers. Confirm which the NSP wants before running the full study.

Summation

Within the plant, harmonic currents from similar inverters are generally summed arithmetically (a summation exponent of 1), because identical units with identical control do not cancel reliably. Between the plant and the background, the summation law exponents from the standard are normally used, varying by order. Some DNSPs specify arithmetic summation for all orders on smaller connections, which is conservative.

What is compared with what

Several distinct checks are often made:

  • Plant emission at the connection point, with background set to zero, against the allocated emission limits
  • Total distortion, combining the plant's emission and the amplified background, against planning levels from AS/NZS 61000.3.6, not against the access standard tables, which serve a different purpose
  • Distortion at the inverter terminals, against the compatibility levels the equipment is designed for
  • Amplification of background, with the plant's passive elements connected but sources off

The basis of the study has to match how compliance will be measured on site. Most NSPs measure before and after connection and compare, so the study's definition of emission should align with that.

Amplification: The Hidden Problem

A plant's cables, filters and transformers form a resonant circuit with the network impedance. At some harmonic orders this circuit amplifies background distortion that already exists, even if the plant itself emits very little at that order.

Some NSPs set limits on the amplification factor, the ratio of background distortion with and without the plant. Low-order harmonics such as the 5th, 7th, 11th and 13th often have the tightest limits, because background is highest there.

The worst case for amplification is frequently night time or commissioning: collector cables and filters energised, inverters off. Inverters provide some damping when they are running. With them off, the cable capacitance resonates freely. For solar plant this is every night. For a battery, idle periods matter similarly. Always include this case.

High amplification can also come mainly from the main transformer impedance resonating with a capacitive transmission network. Filters on the medium voltage side cannot fix that, and high-voltage filters are costly. When stuck, it is worth asking the NSP which polygon points represent short-term contingencies and which represent long-term conditions, and checking measured background, because amplified low background may be insignificant.

Transformer Excitation Harmonics

A frequent reason for rejection is leaving out the harmonics produced by transformer magnetising current. Both main and unit transformers contribute, especially at elevated voltage.

A practical method:

  1. Simulate the unloaded transformer in an EMT tool at the highest expected operating voltage, using the manufacturer's excitation curve.
  2. Take the harmonic content of the magnetising current.
  3. Inject it as a current source at the transformer terminal in the harmonic model.

OEM inverter emission data referred to medium voltage is usually measured at low voltage and calculated upwards, so it does not include unit transformer excitation. Where manufacturer curves are not yet available, an assumed excitation harmonic level can be used as a sensitivity, and that assumption then becomes a procurement specification for the transformer.

Triplen harmonics need care. Inverter OEMs may state that triplens are zero sequence and blocked by delta windings. Reviewers commonly dispute treating them as fully blocked, because unbalance in components and in the fundamental voltage produces triplens that are not purely zero sequence.

Scenarios to Cover

Harmonic Study Scenarios

  • All collector feeders in service, and each feeder out
  • Inverters off with reticulation energised (night or commissioning)
  • For a BESS: charging, discharging and idle
  • Cable capacitance and length varied by plus or minus 10 percent
  • Transformer impedance varied by plus or minus 10 percent
  • Filter component tolerances and ageing
  • Each harmonic impedance polygon from the NSP's current issue
  • Emission levels at the power ranges given in OEM data, where installed inverter capacity exceeds export

Filter Design

When the plant exceeds its allocation, the typical solutions are double-tuned or single-tuned filters at medium voltage, C-type filters for damped low-order performance, and sometimes shunt reactors in parallel to manage the net reactive power.

Filters and reactive power are one problem

A filter is a capacitor at fundamental frequency. Its reactive power changes the plant's reactive capability, the voltage control behaviour and the steady-state voltage results. Changing filter size can invalidate the S5.2.5.1, S5.2.5.4 and S5.2.5.12 studies. Designers sometimes fix the filter's reactive power to avoid re-running those studies, and end up with an over-complex or unbuildable filter. It is better to treat reactive power capability and filtering as one design from the start.

Design for tolerance

Filters tuned very sharply, or tuned to high frequencies with small undamped elements, look good in an ideal model and fail in reality, because inductance and capacitance drift with manufacturing tolerance, temperature and ageing. The robust rule is either:

  • the filter is needed, so it is tuned or detuned to remain effective across component tolerance and ageing, or
  • the filter is not needed, so reactive plant is detuned below the lowest significant harmonic to avoid creating a resonance

At the connection application stage, a filter design is proof that a solution exists. Expect it to be refined at detailed design.

Component ratings

  • Capacitors are rated for the arithmetic sum of fundamental and harmonic voltages, under IEC 60871-1
  • Reactors and resistors are rated for the RMS sum of fundamental and harmonic currents, under IEC 60076-6
  • Include background distortion above planning level and fundamental overvoltage in the rating cases

Filter protection needs unbalance detection sensitive enough to see a single internal capacitor fuse operation, plus overcurrent, overvoltage and earth fault protection, and a circuit breaker verified for capacitive switching duty.

The Review Process

Many TNSPs run harmonic review in stages. The applicant first submits the model and results for a sample impedance point, so the NSP can confirm method and data. The full study follows. The NSP then runs its own check in its own tools and compares results. Provide tables to three or four decimal places so differences can be traced. Data queries are common: the per-unit base of inverter currents, transformer impedances that differ between model and report, and skin effect that is claimed but not modelled.

After energisation, measured connection point voltages and currents over several weeks can be used to refine the harmonic model and confirm the study basis.

Get Your Harmonic Design Right

ElectraGlobe carries out transmission harmonic studies, amplification assessments and filter designs for solar and BESS projects, and works through staged NSP review. See our power quality analysis service, or contact us.

Frequently Asked Questions

Why do solar farms and batteries need harmonic filters?

Inverters emit small harmonic currents, but the plant's cables and transformers can resonate with the network and amplify distortion at particular harmonic orders. On weak or capacitive transmission networks, the result can exceed the plant's emission allocation or amplify background distortion beyond what the NSP allows.

What is a harmonic impedance polygon?

It is a region in the resistance-reactance plane, given by the NSP for each harmonic order, describing the range of network impedance the plant may see under credible network conditions. The study tests the plant across the whole region.

Why is night time the worst case for harmonic amplification?

At night the collector cables and filters remain energised but the inverters are off, so the damping the inverters normally provide is absent. The cable capacitance can then resonate with the network more strongly, amplifying background distortion.

Are transformer magnetising harmonics included in the study?

They should be. Both main and unit transformers produce harmonics from their magnetising current, especially at elevated voltage. Reviewers commonly reject studies that leave them out.

Power Quality Analysis

Harmonics, filter design, flicker, voltage unbalance, and AFI studies.

Learn more