REFCL management of harmonics - Energy - Energy Web viewAll electricity distribution networks suffer from distortion of the pure 50Hz sinusoidal voltage waveform caused by non-linear
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
P a g e | 79
11REFCL management of harmonicsAll electricity distribution networks suffer from distortion of the pure 50Hz sinusoidal voltage waveform caused by non-linear customer loads. These distortions are described in terms of mathematically equivalent harmonic voltages and currents. Network harmonic voltages can cause harmonic currents to flow in earth faults, so they are therefore relevant to powerline fire risk.
One REFCL technology, the GFN, has the capability to inject a synthesised voltage waveform to cancel network harmonic voltages and reduce harmonic currents flowing in faults. This capability was explored in Kilmore Tranche 4 tests. The tests confirmed the capability was effective but revealed that further GFN development would be required to realise the potential benefits of this feature in addressing fire risk from network earth faults.
11.1 Summary of findings and recommendations
The effects of network harmonics on fire risk in a REFCL-protected network were explored in the KMS test program. The following conclusions were drawn:
1. In some circumstances, harmonic currents generated by customer loads can flow in a network earth fault and they may have the potential to increase fire risk.
2. It is likely the contribution of harmonic currents to fire risk is second order. 3. Tests demonstrated that:
a. Harmonics do not significantly increase fire risk from GFN diagnostic tests to confirm the presence of a fault and identify the faulted powerline.
b. GFN measurement of low-order network voltage harmonics is consistent with that performed by other network voltage monitoring devices;
c. The GFN has the capability to effectively reduce low-order harmonic components of earth fault current;
d. Further product development would be required for the GFN harmonic compensation capability to provide its full potential fire risk benefits; and
e. The GFN can itself produce some harmonic fault current when it displaces network voltages – this is likely to be a property inherent in all ASC-based REFCLs.
It is recommended that industry work with the GFN manufacturer to further develop the GFN harmonic compensation capability to reduce fire risk in network earth faults.
11.2 Fire risk from harmonics in network earth faults
Many electrical devices have non-linear voltage-current characteristics. When 50Hz network voltages supply customer loads that include such devices, the network load current contains harmonics, i.e. components at frequencies that are multiples of 50Hz. When harmonic currents flow in the series impedances of powerlines, they create harmonics in network voltages. A typical frequency spectrum of a KMS21 test network voltage is shown in Figure 1 – harmonic voltages of between one and 30 volts extend in frequency up to 1600Hz.
In a single phase to earth fault, harmonic network voltages produce harmonic components of fault current that increase the total rms fault current. If a REFCL is used to cancel the 50Hz component of fault current, the harmonic currents will remain largely unaffected and their presence can significantly increase the magnitude of residual fault current and thereby potentially increase fire risk from ground ignition.
This is illustrated by Figure 2 which shows the 50Hz voltage on the faulted conductor has been reduced by the GFN from its pre-fault value of 12,700 volts to about 65 volts. However, the post-fault harmonic voltage magnitudes remain of the same order as before the fault, i.e. between one 1 volt
Ignition tests indicate that total rms of fault current is likely to be the most meaningful measure of the environmental energy release that drives fire risk. Harmonic components increase the total rms
value of fault current in accordance with a root-sum-of-squares formula:
irms=√∑n=0
n=∞
(in )2This means that an additional harmonic component equal in size to the 50Hz
component would increase the total rms fault current by 40%, not 100%.
The fire risk contribution of harmonic components of fault current is therefore only likely to be significant if the harmonics are of a similar order of magnitude to the 50Hz component.
11.3 Increased ground ignition fire risk from harmonics
High impedance faults are realistic representations of known fire causes, e.g. a live high voltage wire falling into dry grass on dry ground. When the fault impedance is high, the ignition mechanism is ground ignition rather than bounce ignition.
In a high impedance fault, both the 50Hz and harmonic components of fault current are reduced to low levels by the fault impedance. The 50Hz component is reduced more because of the high impedance presented by the resonant neutral-earth connection (the parallel combination of the ASC coil and the combined capacitance of the network and the ASC tuning capacitors), as well as by the RCC compensation to cancel the 50Hz component of voltage on the faulted phase. This is illustrated by Figure 3 for a 40,000 Ohm earth fault. The fifth harmonic of the fault current is 1.2mA which is consistent with the pre-fault value of fifth harmonic voltage (49 volts) divided by the fault resistance (40kΩ). The 50Hz component of fault current is 2.5mA which less than one-hundredth of the 320mA that the 40kΩ fault resistance alone would produce with an applied 50Hz voltage of 12,700 volts.
Figure 3: KMS Test 634 (40,000Ω fault) fault current with RCC compensation
In ‘wire on ground’ ignition tests with the GFN in service, once the GFN detected and compensated for the fault, no ground ignitions occurred. The tests outlined in Section Error: Reference source not found above confirm that fault current of 0.5 amps is required for ground ignition in a worst-case earth fault. For a relatively high total harmonic voltage of, say, 500 volts to generate this level of harmonic current the fault impedance would have to be not more than 1,000 Ohms.
Even in this worst-case situation, there are two considerations that indicate the fire risk from harmonics would at most be second order:
1. It is difficult to postulate realistic circumstances in which just three metres of conductor on the ground would have a fault resistance as low as 1000 Ohms without assuming higher levels of soil and fuel moisture content than would be consistent with worst-case fire weather and fuel conditions.
2. Detection of such a fault would be relatively fast, both in the initial instance and in the GFN fault-confirmation test, so the harmonic current flow would last at most about five seconds - the period between the fault and the first steps of the fault-confirmation test. A fault current of 0.5 amps generally takes longer than this to produce ground ignition – often tens of seconds.
Whilst there is no guarantee that a particular combination of factors capable of starting a fire from harmonic currents can never occur, the probability of this combination is considered to be relatively low, i.e. fire risk from harmonic currents is likely to be a second order risk.
11.4 Fire risk from harmonics in diagnostic tests
In high-impedance faults, the key factor in fire risk is the current produced in diagnostic tests such as the fault-confirmation test. The GFN performs such tests by injecting a 50Hz voltage. This leaves
harmonic components of fault current largely unaltered as can be seen from Figure 4 - the fault-confirmation test increased the 50Hz current by twenty-five times from 2.5mA to around 65mA (shortly afterwards the FCT re-detected the fault at a current of 132mA), but the fifth harmonic current remained unchanged at 1.2mA. It can be concluded that harmonic currents do not significantly increase fire risk in diagnostic tests.
Figure 4: KMS Test 634 (40,000Ω fault) fault current during GFN fault confirmation test (Stage 1, Step2)
This leads to the conclusion that in high impedance faults, harmonics are unlikely to contribute significant fire risk from ground ignition.
11.5 Increased bounce ignition fire risk from harmonics
The ignition mechanism in low impedance faults tends to be bounce ignition. The REFCL response required to prevent fires from this cause is fast reduction in the voltage on the faulted conductor to less than 1,900 volts in 85 milliseconds. For a relatively high level of network harmonics of perhaps 500 volts, the root-sum-of-squares addition means the total voltage at 85 milliseconds would only be increased by 3.4% to 1,960 volts compared to the 1,900 volts 50Hz component. The effect of this increase on bounce ignition fire risk would likely be immaterial.
11.6 Measurement of harmonics by the GFN
The GFN samples the network voltage waveforms with 16-bit precision at a rate of 100 samples per 50Hz cycle. To measure harmonics the GFN applies a discrete Fourier transform to groups of 100 samples. GFN documentation does not identify the window function (if any) applied prior to the transform calculation to minimise spectral leakage.
Theory predicts this arrangement should be capable of measuring harmonics up to about 2 kHz, i.e. to the 40th harmonic of the 50Hz fundamental. However, the voltage waveforms seen by the GFN are produced by an electromagnetic voltage transformer on the substation’s 22kV busbars. The limited bandwidth of this type of transformer may affect measurement accuracy at higher harmonic frequencies.
A manually synchronised spot check was performed to compare measurements on two selected low-order harmonics supplied by three devices:
1. The GFN supplied by the substation 22kV bus voltage transformer;2. An Elspec supply quality meter in the zone substation also supplied by the substation 22kV
bus voltage transformer; and3. The Gen3i data acquisition system at the test site supplied by a wide-band capacitive voltage
divider.
The three sets of readings are shown in Table 1.
Table 1: comparison of harmonic measurements - red phase voltage 8th October 2015 at about 1:20pm
Within the limitation imposed by the single significant figure display of the GFN, its displayed readings were consistent with the measurements taken by the other two devices. This spot check also indicated that the limited bandwidth of the busbar voltage transformer was not a material factor in accurate measurement of harmonics up to 350Hz3.
11.7 Compensation of harmonics by the GFN
The GFN has the capability to include low-order harmonic components into its RCC-injected compensation voltage to cancel corresponding harmonic components of fault current. This capability was tested in the Kilmore test program. Test 723 best illustrated its effectiveness as shown in the three fault current frequency spectra set out in Figure 5, Figure 6 and Figure 7 together with the fault current waveforms associated with each.
With no harmonic compensation, Figure 5 shows that the third harmonic and fifth harmonic components of the fault current are of the same order of size as the 50Hz component. Those two components increased the total rms fault current by 94% above the 50Hz value.
After manipulation of the GFN harmonic compensation settings, Figure 6 shows the third harmonic component of fault current has been reduced by 90%. However in the 23 second period since the first record, the fifth harmonic component of fault current has increased by 37% to 0.14 amps.
After further manipulation of the GFN harmonic compensation settings, Figure 7 shows the fifth harmonic component of fault current has been reduced by 70%. At this point, the third and fifth harmonic components of fault current were increasing the total rms fault current by just 6% above the 0.117 amps 50Hz value.
Though Test 723 demonstrates an effective capability to compensate low-order harmonic components of fault current, this was done by manual manipulation of GFN settings. To reduce fire risk by minimising fault current caused by network harmonics, the GFN must include a dynamic capability to measure and cancel the ever-changing levels of low-order harmonics on the network.
2 The GFN readings shown are those available on the front panel and only had one significant figure displayed. The GFN only displays the 3rd, 5th, 7th and 9th harmonics though its measurement system is theoretically capable of measuring harmonics up to the 40th.3 The Gen3i readings used a wide-band capacitive transducer, whereas the Elspec readings used the substation busbar electromagnetic voltage transformer.
It is recommended that industry work with the GFN manufacturer to enhance the harmonic compensation capability of the GFN to minimise fire risk from network harmonics.
Figure 5: KMS Test 723 (400Ω fault) fault current with 50Hz compensation only
The response of a GFN to an earth fault, i.e. voltage injection by the RCC to cancel the voltage on the faulted phase, can itself produce harmonic components of fault current.
11.8.1 How a non-linear ASC coil generates harmonic earth fault currents
The mechanism is as follows:
1. Harmonics in the magnetising current drawn by the ASC coil
When the GFN compensates the 50Hz component of earth fault current, it effectively applies a synthesised pure 50Hz sinusoid voltage across an iron-cored inductor (the ASC coil) that is in resonance with the capacitance of the network in parallel with the capacitors in the ASC unit itself. The iron core of the coil is not linear and it will draw some harmonic currents from the pure 50Hz source voltage applied by the RCC. The harmonic current components generated by the coil’s non-linear characteristics will flow in both the coil and the RCC inverter.
2. Harmonic voltage produced on the neutral of the transformer and on all network conductors
If the RCC were an ideal voltage source, the harmonic currents flowing between the RCC and the ASC coil would not affect network voltages. However, the non-zero source impedance of the RCC4 means they produce harmonic components of neutral voltage. The resonant combination of the coil and the combined network plus ASC capacitance will have very high impedance at 50Hz but lower impedance at frequencies other than 50Hz – the actual value will depend on the total system (network plus ASC) damping. This impedance forms a voltage divider with the RCC source impedance to produce a harmonic voltage across the ASC coil, i.e. at the neutral point of the transformer. For lower harmonic frequencies, this harmonic voltage will appear between every powerline conductor in the whole network and earth.
3. Harmonic current flow in an earth fault
When a single phase earth fault is present, the harmonic voltage on the faulted conductor will produce a harmonic current in the circuit loop comprising:
The substation transformer winding between the neutral-ASC coil connection and the faulted phase of the substation 22kV busbars;
The faulted phase conductor from the substation 22kV busbars to the fault location; Through the fault to earth; and Through the earth back to the ASC coil-earth connection at the substation.
11.8.2 Factors that determine the magnitude of harmonic fault current
Two main factors will influence the magnitude of the harmonic component of fault current caused by the GFN response to the fault:
1. Harmonic neutral voltage: The magnitude of the harmonic voltage across the ASC coil is likely to be close to the magnitude of harmonic magnetising current necessary to support the 50Hz voltage applied across the coil by the RCC multiplied by the effective source impedance of the RCC. Other factors are likely to be second order.
4 This impedance is due to the length of 345 volt cables between the RCC inverter and the ASC coil as well as the leakage flux between ASC windings. The 37:1 turns ratio of the ASC multiplies the impedances of these elements by 1350 times when they are seen from the 22kV network.
2. Earth fault loop impedance: two impedances will determine the magnitude of the harmonic component of fault current produced by the harmonic voltage across the coil:
a. Transformer winding impedance: the substation transformer will offer some impedance to the harmonic current flow. Transformers can have complex internal magnetic configurations which may influence this impedance. For example, most large substation transformers (such as the one at KMS) have three limbs to the internal iron core. The internal distribution of magnetic flux this produces means that lower impedance may be offered to the flow of triplen5 harmonic currents on the network and higher impedance to harmonics at other frequencies. The detailed modelling that would be required to confirm this was beyond the scope of this project. However, the test results did indicate a qualitative difference between the behaviour of triplen harmonics and that of other harmonics in the fault current.
b. Fault resistance: the fault resistance impedes current flow equally at all frequencies. A high-impedance fault will greatly reduce harmonic current flow, whereas a low-impedance fault can produce high harmonic currents.
In summary, the non-linearity of the ASC coil can create some harmonic fault current even if there are no harmonics in pre-fault network voltages.
11.8.3 Confirmation in simulations
Simulations of the response of a GFN with a non-linear ASC iron core were used to confirm the mechanism for generation of harmonic fault currents by a GFN. These were not exact simulations but served to verify the concept. The model used is shown in Figure 8.
Figure 8: conceptual model of GFN response to an earth fault with non-linear ASC coil
The non-linearity is introduced by the Zener diodes D5 and D6 (±15,000 volts withstand) and D7 and D8 (±10,000 volts withstand). The source impedance of the RCC is modelled by R14 and L4 - (50Ω+j50Ω), equivalent to about 50 milliohms on the low voltage side of the ASC.
The concept simulation results are set out in Table 2.
5 Any odd harmonic number divisible by three: 3rd, 9th, 15th, 21st, etc.
The results in Table 2 confirm that ASC non-linearity combined with finite (i.e. non-zero) RCC source impedance will produce harmonic fault currents. The fault current waveform produced in Simulation 2 (non-linear ASC with finite RCC source impedance) is shown in Figure 9.
Figure 9: fault current result in simulation of non-linear ASC with finite RCC source impedance
It was concluded that the conceptual mechanism described in Section 11.8.1 above is a probable cause of some harmonic current components in network earth faults when a GFN is present.
11.8.4 Confirmation in test results
This phenomenon was observed in the KMS test results. Ten tests were analysed for harmonic components of voltages before and after fault detection and RCC response. Pre- and post-fault measurements were only a few seconds apart. However, harmonic levels vary continuously on the network due to changes in customer loads so even this short interval may have involved changes due to factors outside the experiment. To find underlying patterns, averaging over a group of tests was used.
Seven of the tests had a 40,000 Ohm fault resistance and three had a 400 Ohm fault resistance. The average harmonic levels across all ten tests are shown in Table 3 to Table 5.
1. RCC action was associated with increases in triplen harmonic voltages: The triplen voltage harmonics increase in the post-fault period when there is 12.7kV of 50Hz voltage across the ASC coil. The increase in the third harmonic (150Hz) was particularly dramatic – in tests it increased by a factor of five times. The ninth harmonic (450Hz) approximately doubled.
2. Non-triplen harmonic voltages seemed uncorrelated with RCC action: The effect of the fault and its compensation on non-triplen voltage harmonics was mixed: the change in the fifth was within the normal test-to-test variation, the seventh was approximately halved on the faulted phase but remained unaltered on the un-faulted phases, the eleventh appeared slightly diminished on the faulted phase but unchanged on the others, and the thirteenth was affected hardly at all.
Because third harmonic currents are zero-sequence, they cannot be generated by customer loads except in a network of unbalanced impedances – and even in that situation, they are usually small. The large increase in the third harmonic seen in KMS tests when a fault is applied is almost certainly due to the non-linearity of the magnetising curve of the ASC coil.
11.8.5 Relevance to fire risk
This phenomenon may tend to increase potential fire risk as it can drive a significant 150Hz component of fault current. The ratio of fundamental to harmonic components of fault current in the ten tests is shown in Table 6, grouped and averaged by fault resistance.
Table 6: harmonic components of residual fault current averaged over ten tests
The third harmonic component of fault current was about 85% of the 50Hz current. This increased the total rms value of fault current by about 30% above the 50Hz value. If the third harmonic component of fault current was solely that driven by network voltages without the effect of the ASC coil (i.e. five times less), the total rms value of fault current would only have been increased by 1-2% above the 50Hz value.
It is recommended that the GFN manufacturer consider the addition of triplen harmonic compensation to cancel the harmonic current generated by the ASC coil and minimise fire risk from triplen harmonic fault currents. Section 11.7 above indicates the GFN has the capability to do this with appropriate settings.
12REFCL management of two-phases-to-ground earth faultsNetwork faults that convey current from two phase conductors to earth arise most commonly from two causes: fallen trees on powerlines and a fallen conductor which remains connected to the powerline on the ‘downstream’ side of the break (called a back-fed fault). Both of these faults involve potential fire risk. Tests were performed at the Kilmore South test site to better understand the capability of a REFCL (specifically the GFN) to manage this risk.
12.1 Summary of findings and recommendations
The test results indicate that the current GFN design will significantly reduce fire risk in two-phases-to-ground earth faults. However, the test results confirmed that the GFN product is not yet optimised for minimum fire risk in these types of faults, though there appears to be clear potential to achieve this outcome.
It is recommended that the GFN manufacturer be encouraged to enhance the GFN design to apply effective compensation to reduce fault current in all types of earth fault and to apply the same fault-confirmation test (proven to have low fire risk) to detect sustained faults and allow fast powerline disconnection to reduce fire risk.
12.2 Relevance to fire risk
Back-fed and other two-phases-to-ground earth faults are relevant to fire risk:
12.2.1 Back-fed faults (wire down, ‘downstream’ end remains connected)
Back-fed faults are ‘wire down’ earth faults where the break in the conductor leaves the fallen wire on the ground connected to the downstream side of the powerline, i.e. connected at the end furthest from the source substation. It is not energised directly from the substation but can convey customer load current (from customers downstream of the break) into the earth. This current can be sufficient to start a fire. Back-fed faults are notoriously difficult to detect with traditional powerline protection systems. They may remain undetected for some time which further increases the fire risk.
Though back-fed faults are not as common as normal earth faults, they are not rare. Their occurrence is determined by the random location of conductor breakages under the stress of high wind conditions.
The 2014 Theoretical Study investigated back-fed faults using the model shown in Figure 10.
Figure 10: theoretical model of back-fed earth fault on a resonant earthed network
The Study concluded:
The response of a REFCL-protected network to a back-fed fault is the same as to a direct-fed earth fault with some adjustment to the parameters: the effective fault resistance is the actual fault resistance plus half the phase-to-phase resistance that represents the customer load downstream of the conductor breakage.
The neutral voltage displacement may be as little as half that which would occur if the fault was not back-fed. This means that detection of back-fed faults may only be half as sensitive as normal faults having the same fault resistance.
Fault detection sensitivity will depend on the customer load downstream of the conductor breakage. To detect a one amp back-fed earth fault, the customer load downstream of the breakage would have to be at least 200kW and the REFCL fault detection threshold would have to be a neutral voltage displacement equal to not more than 10% of nominal voltage.
Though not common, back-fed faults present a clear fire risk when they occur in high fire risk conditions, so the response of a REFCL to this type of fault is of relevance to fire risk reduction.
A back-fed fault is simply a special case of a two-phases-to-ground earth fault where the fault current flows from the two unbroken phase conductors through the customer loads to the fallen wire and thence to earth. Analysis of back-fed faults often relies on an assumption that customer load located downstream of the fault location is balanced (equal load on each of the three phases). However whilst load balance is an objective of network planners, it is not always achieved in practice. With unbalanced load, a back-fed earth fault is equivalent to the more general case of a two-phases-to-ground earth fault – which may be balanced or unbalanced depending on circumstances.
12.2.2 Two-phases-to-ground earth faults (‘tree in powerline’ faults)
One of the most common causes of a two-phases-to-ground earth fault is trees near powerlines. When a tree falls across or into a powerline, there are a number of possible outcomes, each of which can present a different form of earth fault to the high-voltage network:
1. Fallen tree branch resting on one powerline conductor: this is an instance of a ‘branch touching wire’ earth fault. The fire risk from such faults can be greatly reduced by a REFCL as outlined in Section 13 below.
2. Fallen tree resting on two or more powerline conductors: this can present as a combination of a ‘branch touching wire’ earth fault plus a ‘branch across wires’ fault or even a direct
phase-to-phase fault if the two phase conductors are brought into contact with each other by the weight of a tree branch.
3. Fallen tree entangled in powerline conductors with multiple current paths to earth: a common example of this is when a fallen tree brings the conductors low (though still not touching ground) and tree branches come to rest on the ground on the far side of the powerline. It can present as a two-phases-to-ground earth fault or a three-phases-to-ground earth fault. Again, if any two conductors touch, a direct phase-to-phase fault will occur.
4. Fallen tree and powerline conductors on ground: This usually occurs if poles break or fall or conductors break under the weight of a tree. Again, it can present as a ‘wire down’ earth fault, a two-phases-to-ground earth fault or a three-phases-to-ground earth fault and if two conductors touch, a direct phase-to-phase fault.
A REFCL can manage many of the fire risks in this complex set of scenarios while some others may remain relatively unaffected by its presence. For example, direct phase-to-phase faults (conductor clash or flashover) will normally be addressed by fast overcurrent protection systems6 - the REFCL cannot respond to them. Phase-to-phase vegetation faults (‘branch across wires’) will also be undetected by a REFCL. However, unless the branch concerned is fully detached from the tree and suspended above ground, this type of fault may be accompanied by an earth fault that is detectable by a REFCL. Both two-phases-to-ground and three-phases-to-ground earth faults create earth current that is detectable by a REFCL, either as neutral current (by an SSFCL) or as neutral voltage displacement (by a GFN or ASC).
Whilst the set of tree fault scenarios listed above is complex, it presents only a limited range of fault types that create earth current: they are single-phase-to-ground earth faults or two-phases-to-ground earth faults, combined with various permutations of phase-to-phase faults. For example, a three-phases-to-ground earth fault can be accurately modelled as a two-phases-to-ground earth fault plus a set of three phase-to-phase faults. Hence the way in which a REFCL responds to two-phases-to-ground earth faults has an important bearing on fire risk from all fault types involving trees, as well as from back-fed faults whether balanced or unbalanced.
If a REFCL detects a two-phases-to-ground earth fault and responds by quickly disconnecting the powerline7, fire risk from many complex tree faults as well as back-fed earth faults may be reduced.
The Kilmore South test program included a series of tests of resistive two-phases-to-ground earth faults with the GFN in service, including balanced faults equivalent to back-fed earth faults with balanced downstream customer load. The other two REFCL configurations were not tested with these types of faults as they were not yet compliant with the draft REFCL performance standard.
12.3 GFN treatment of two-phases-to-ground earth faults
The GFN (and other fault detection schemes in resonant earthed systems) detects the presence of a fault by the magnitude of the neutral voltage and deduces the type of fault from the phase angle of the neutral voltage:
1. If the neutral voltage closely aligns with one of the network phase voltages, this indicates the fault is a single-phase-to-ground earth fault on that phase.
6 Disconnection times, potential ignition outcomes and fire risks of such events are beyond the scope of this report.7 The 2015 Vegetation Conduction Ignition test program found that interruption of supply within five to ten seconds might substantially reduce fire risk from ‘branch across wires’ faults. (Refer Vegetation Conduction Ignition project report Section 8.6, page 62). This speed of response is within REFCL capability.
2. If the neutral voltage does not align with any single network phase voltage, this indicates that the fault is a two-phases-to-ground earth fault involving the two phases that have voltages ‘to either side’ of the neutral voltage.
To allow for measurement errors and various factors that can disturb the measured phase of the neutral voltage, ‘phase angle zones’ are defined. The GFN has six zones, each of approximately8 60 degrees width: three centred on the three phase voltages; and three centred on the mid-points between each pair of phase voltages. Depending on which zone the neutral voltage falls within, the two associated GFN responses are:
1. Phase zone: the GFN treats the fault as a standard single-phase-to-ground earth fault, i.e. it uses RCC compensation to reduce the voltage on that phase close to zero; uses the FCT to test for the fault after a few seconds; and then either disconnects the powerline9 if the fault is still present, or removes the compensation if the fault has disappeared.
2. Inter-phase zone: the GFN treats the fault as a back-fed fault; it applies no compensation and immediately disconnects the powerline (once it has been identified – this can take some seconds).
The ‘inter-phase zone’ response has been developed to suit conditions in other countries that use the GFN product. The rationale of the immediate disconnection response is that such a fault will always involve a live wire on the ground with unacceptable public safety risk.
12.4 GFN test results: two-phases-to-ground earth faults
The GFN was tested using bolted resistive fault tests with different combinations of available high-voltage resistors to simulate a two-phases-to-ground earth fault. The configuration is shown in Figure11 and the resistance values in each test configuration are listed in Table 7.
The test conditions included two relatively unrealistic features:
1. The fault impedances were relatively high compared to many real earth faults of this type. This was partly compensated by increased fault detection sensitivity in Test 728 and later tests; and
2. To protect the high-voltage resistors from thermal overload, fault current duration was limited to less than one second in tests with the 15,200 Ohm common resistor and less than 700 milliseconds in tests without it, i.e. fault duration was too short for the GFN to find a sustained fault using its FCT.
8 With standard GFN settings, the phase zones are 66 degrees wide and the inter-phase zones 54 degrees wide. The relative widths of the two classes of zones can readily be modified by changing a setting.9 This is subject to the operating policy of the network owner that applies at the time of fault occurrence. Disconnection is the appropriate policy at times of extreme fire risk.
Table 7: resistance values in KMS two-phases-to-ground earth fault tests
Test Phase 1 R1 (Ω) Phase 2 R2 (Ω) R_common (Ω)727 Blue 8,350 White 16,700 15,200728 Blue 8,350 White 16,700 15,200729 White 8,350 Blue 16,700 15,200730 White 16,700 Blue 16,700 15,200731 White 16,700 Blue 16,700 0732 Red 16,700 Blue 16,700 0733 Red 16,700 Blue 16,700 0734 Red 16,700 White 16,700 0735 Blue 16,700 White 16,700 0736 Blue 16,700 White 8,350 0737 White 16,700 Blue 8,350 0
Eleven tests were performed and the GFN response recorded. These are listed in Table 8.
Table 8: GFN responses to two-phases-to-ground earth faults
Test GFN response Max fault current727 Fault not detected (threshold 0.5A) 0.34A728 White phase compensated, no FCT (threshold set to 0.2A) 0.35A729 Blue phase compensated, feeder not identified, FCT 0.37A730 White phase compensated, feeder not identified, FCT 0.27A731 White phase compensated, feeder identified, FCT 0.75A732 Blue phase compensated, feeder not identified, FCT 0.76A733 Detected, no compensation, feeder not identified, no FCT 0.76A734 Late compensation of White phase, no FCT 0.69A735 Late compensation of Blue phase, FCT 0.74A736 Reverse earth fault detected, no compensation >0.92A737 Blue phase compensated, FCT 1.33A
The conclusions drawn from the test results were:
1. The GFN detected the two-phases-to-ground earth faults in all tests where the earth current exceeded the fault detection sensitivity setting.
2. In those tests where the GFN applied voltage compensation using the RCC, it was always applied to reduce the voltage on one of the two phases involved in the fault.
3. ASC tuning mismatch and network capacitive imbalance can cause the phase of the neutral displacement produced in low-current two-phases-to-ground earth faults to vary somewhat from that expected from theory and this can influence determination of fault type.
4. The GFN response was not always consistent, with different responses recorded in tests having identical conditions. The observed response was not always as expected based on existing knowledge of the GFN design.
5. The GFN response to two-phases-to-ground earth faults is complex and somewhat uncertain due to multiple internal processes running in parallel within the GFN firmware.
6. The demonstrated GFN performance in two-phases-to-ground earth faults is not optimised for lowest fire risk, though the GFN appears to have the capability to deliver this.
The result of each test was discussed and analysed in some detail with the GFN manufacturer on site. It was recognised that the GFN design had not yet been optimised for low fire risk in this class of earth faults. These discussions tended to confirm the potential for this to be done.
12.5 Fire risk reduction with the current GFN design
Although the test results demonstrated the GFN is not yet optimised to reduce fire risk in this class of faults, it is clear that its current design would nevertheless achieve some reduction in fire risk.
12.5.1 Back-fed faults
There are reasonable grounds to believe that a REFCL-protected network will not have fires resulting from back-fed faults:
For a conductor breakage to create a back-fed earth fault without creating a simultaneous direct-fed earth fault, the break must be at the end of the span closer to the source substation, so in the worst case the length of conductor on the ground would be not less than 16 metres by the logic of Section Error: Reference source not found on page Error: Reference source not found. The current into the earth required to start a fire would be 2.4 amps based on the ignition threshold of 0.15 amps per metre.
A REFCL set to detect a 0.5 amp direct-fed earth fault would detect a back-fed earth fault involving one amp of current into the earth provided the downstream customer load was high enough – more than about 200 kilowatts.
At times of high fire risk, ‘downstream’ customer load is likely to be higher (due to air conditioning, refrigeration and water pumping) rather than lower. Most faults on a 22kV powerline would have more than 200 kilowatts of downstream load at these times.
Simulations of a back-fed fault on the larger Frankston South network demonstrate that there is no combination of downstream load and fault resistance that results in an undetectable back-fed fault capable of causing a fire with 16 metres of conductor on the ground10. Lower levels of downstream load make the fault hard to detect, but they also reduce the fault current below the 2.4 amp ignition threshold.
The simulations also indicate that downstream load balance does not materially alter this conclusion: severely imbalanced load makes the fault detection faster while producing a higher current to earth for only the first few tens of milliseconds – insufficient time for a ground ignition.
The GFN with its FCT could enable disconnection of a powerline on which a back-fed fault has occurred before a fire could start. Disconnection as soon as the fault is confirmed by the FCT to be still present is likely to be required for public safety as well as fire risk reduction.
12.5.2 Tree faults
There are reasonable grounds to believe that a REFCL protected network will have 50% to 90% less fire risk from tree faults compared to a non-REFCL network:
All tree faults involve some earth fault current that is potentially detectable by a REFCL11. Trees that fall on powerlines usually have a trunk diameter greater than 100mm at the point
of contact with the powerline. The 2015 Vegetation Conduction Ignition test program demonstrated that larger (80-90mm)
diameter branches generate an initial earth fault current of more than one amp and that this initial current will increase as the square of the diameter. This implies that tree faults that involve more than one phase of the powerline will normally produce earth fault current in
10 There are combinations that create other public safety risks, e.g. electrocution if someone approaches or touches the fallen wire. 11 The only vegetation fault type that cannot be detected by a REFCL is a pure ‘branch across wires’ fault, i.e. where the branch is completely detached from the tree and suspended wholly above the ground.
excess of one amp, i.e. in excess of the REFCL fault detection sensitivity, so they will be detected.
Vegetation faults take seconds or tens of seconds to produce a ground-level fire. The 2015 Vegetation Conduction Ignition test program indicated that interruption of supply to the fault within ten seconds would reduce fire risk by the order of 50%. Interruption within five seconds would cut fire risk by perhaps 90%.
The GFN’s FCT normally detects a sustained one amp fault in the first few steps, i.e. about five or six seconds after initial fault detection.
The GFN with its FCT could enable disconnection within five to ten seconds of a powerline on which a tree fault has occurred, making a 50% to 90%12 reduction of fire risk possible.
12.6 Approaches for lower fire risk in two-phases-to-ground earth faults
Although the current GFN design is likely to provide fire risk reduction in two-phases-to-ground earth faults, the tests revealed that under extreme conditions (high fault impedances) it is not optimised for consistent delivery of maximum fire risk benefits. Further product development could achieve this.
Provided the magnitude and phase angle of RCC compensation is carefully chosen, fault current to earth in any type of earth fault can be cancelled, whether it is a single-phase-to-ground fault or a two-phases-to-ground fault, balanced or unbalanced.
In two-phases-to-ground earth faults the optimum response may be:
1. Use the RCC to apply the compensation required to bring the earth fault current to a very low value. The measurements required to correctly calculate the magnitude and angle of the required compensation are already gathered by the GFN but the calculation of the compensation appears constrained somewhat by the currently imposed dichotomy in the GFN’s internal logic processes between normal earth faults and reverse (back-fed) earth faults.
2. Once the correct compensation has been applied to reduce the earth fault current to a level close to zero, use the FCT to test for a sustained fault and identify the faulted powerline. The Kilmore South tests have confirmed the FCT can be performed with minimal additional fire risk.
3. If the fault is confirmed to be a sustained fault, apply the network owner’s policy for such faults (in high fire risk conditions, immediately disconnect the powerline).
It is recommended that the GFN response to two-phases-to-ground earth faults be re-engineered to provide this response without reliance on any assumptions about whether such faults are balanced or not. This will reduce fire risk in both back-fed faults (usually but not always reasonably balanced) and tree faults (often grossly unbalanced). It will also enhance public safety in ‘wire down’ faults.
12 The FCT starts about four seconds after the fault is detected and is complete about six seconds later.
13REFCL management of vegetation earth faultsThis section describes ‘branch touching wire’ vegetation earth fault tests staged at Kilmore South with a REFCL in service. The tests at Kilmore South validated the simulation of REFCL action used in the Vegetation Conduction Ignition test program completed earlier in 2015 at Springvale13 and hence confirmed the Springvale test results can be applied to REFCL-protected networks.
This rigorous confirmation allows the conclusion to be reliably drawn that REFCLs can cut fire risk from ‘branch touching wire’ earth faults by a factor of twelve, from 100% to 8%, when averaged across all of the eleven species tested at Springvale. The Kilmore tests also demonstrated that this benefit will be delivered even when a ‘soft’ fault-confirmation test is used to confirm the fault is sustained and identify the 22kV powerline upon which it is located.
13.1 Summary of findings
A total of 61 valid vegetation ignition tests carried out at Kilmore South showed that:
1. The Kilmore South test results, when combined with those from the Springvale tests, show REFCL network protection can cut fire risk from ‘branch touching wire’ earth faults averaged over a wide range of species, from 100% to 8%.
1. A GFN can detect current in a ‘branch touching wire’ earth fault at a level of 0.5 amps and upon detection, reduce current through the vegetation to a level close to zero.
2. Vegetation earth fault fire probability is not materially affected by diagnostic tests that comply with the REFCL performance standard.
13.2 The Kilmore vegetation earth fault tests
The 2015 REFCL Technologies test program tests at Kilmore included 61 valid ignition tests on sample branches of Willow (Salix Sp.). The 61 valid tests included 19 that produced fires. The fire probability from statistical analysis of the Kilmore tests is shown in comparison with 0.5 amp artificially current-limited tests of Willow at Springvale in Figure 12.
Figure 12: comparison of Kilmore and Springvale tests of Willow branches
0
20
40
60
80
100
Kilmore Springvale
Fire
pro
babi
lity
(%)
13 Over 1,000 tests carried out at Springvale earlier in 2015 on a wide range of vegetation species showed that: if it could be confirmed that a REFCL can detect current in ‘branch touching wire’ earth faults at a level of 0.5 amps and upon detection interrupt that current, then the probability of fire from such faults would be reduced by a factor of around ten to 8% (less than 13% with 95% confidence); and secondly, for tests limited to 0.5 amps current, Willow was the worst case species with a fire probability of 36%. The 31% fire probability of Willow in ‘branch touching wire’ earth fault tests done at Kilmore with a GFN in service is, within the limits of statistical certainty, the same as the 36% Springvale result with an NER in service and test current artificially limited to 0.5 amps.
The Kilmore ‘branch touching wire’ tests drew extensively on concepts and designs developed in the 2015 Vegetation Conduction Ignition test program, including:
Defined worst case conditions for ignition (45°C air, turbulent 10-15 kph airflow, samples conditioned for approximately 8-24 hours at 45°C and kept in 2°C storage at all other times)
Test rig (two parallel 19/3.25AAC conductors 1.5m apart: an earthed one supporting the thicker end of the sample and the other energised at 12.7kV supporting the thin end)
Fire start criteria (prior to any flashover, small embers must reach the floor of the rig at 350°C or higher, or larger ones at 250°C or higher)
Sample diameter (thick end) around 25-30mm and sample length about 2.5 metres.
The Kilmore experiments varied from those at Springvale in some minor respects:
There was 400 Ohm series resistance in the high voltage supply, instead of 200 Ohms. Given this was only a 1.5% variation in total current path resistance, it was considered to have negligible influence on the results
Only Willow (Salix Sp.) tests were analysed as it was not possible to obtain enough suitable samples of the other identified worst-case species, Desert Ash (Fraxinus Angustifolia)
The vegetation samples were harvested in October, i.e. mid-spring, rather than February and March (late in a relatively cool summer).
The small size of the KMS network meant the GFN was extremely sensitive. To mimic the 0.5 amp fault detection sensitivity typical of larger networks, it was desensitised - the fault detection threshold was increased from 23 volts (2,650 volts neutral displacement) to 40 volts (4,620 volts neutral displacement).
The current immediately prior to detection of vegetation faults by the GFN was recorded and is shown in Figure 13. The average fault detection sensitivity demonstrated in the 59 valid tests in which fault detection took place was 0.40 amps.
Figure 13: fault current immediately prior to fault detection by the GFN
0
2
4
6
8
10
12
14
0.1 0.2 0.3 0.4 0.5 0.6
Coun
t (n
= 59
)
Current at which fault was detected (Arms) The moisture content and diameter (thicker end) of samples were also recorded. The values are shown in Figure 14.
Figure 14: Salix Sp. vegetation sample moisture content and diameter
0
10
20
30
40
50
60
70
20 25 30 35 40 45
Coun
t (n
= 84
)
Moisture (%)
Sample moisture
0
5
10
15
20
25
30
35
10 15 20 25 30 35 40 45
Coun
t (n
= 84
)
Diameter (mm)
Sample diameter
The sample moisture levels and diameters were close to the recorded average values for Salix samples in the Springvale tests (39% moisture and 26mm diameter). It was concluded that the Kilmore tests were an adequate replication of the Springvale tests to allow direct application of fire probability results from the Springvale tests to REFCL protected networks.
13.3 The ‘flarc’ – a qualitative difference with REFCL protection?
The Springvale tests revealed all vegetation faults progress through four stages. The Kilmore tests showed the first three stages were unchanged by the presence of the GFN. However, the fourth was qualitatively different for reasons not yet understood. At Springvale, the fourth stage was a flashover – flame being an excellent conductor, once a continuous line of flame was established from one wire to the other, an arc instantly formed and the current immediately increased to the 64 amp limit set by the series resistor in the rig’s high-voltage supply. The test ended in a flash of light.
In the Kilmore tests, nearly every time a continuous line of flame formed between the two wires supporting the branch, the test sample settled into a relatively steady state condition that became known among the test team as a ‘flarc’- a combined flame and arc but not a flashover. In a flarc, the current was rich in harmonics and remained relatively steady for many seconds at a level generally below 0.5 amps. Some data recorded during a typical flarc event is shown in Figure 15.
Figure 15: KMS Test 749 Salix sample - extended flarc from 122 seconds14
14 The same chart processing tools were used in both ‘wire down’ and vegetation ignition tests. The vertical axis labelled ‘Soil Current’ here shows the vegetation current in the latter tests.
It can be seen from Figure 15 that during the 12 second period of the flarc, the voltage across the burning branch was around 9,000 volts (the voltage waveform was an almost pure 50Hz sinusoid) but the current remained between 0.4 and 0.5 amps with a very non-sinusoidal waveform suggestive of a bi-directional voltage limiting mechanism (thresholds at plus/minus some kilovolts) in series with a high source impedance (of the order of ten kilo-ohms).
In only a few tests did a flarc result in a form of flashover, e.g. Test 772, but the flashover current was well below the level allowed by the 400 Ohm current-limiting resistor in the high-voltage supply to the test rig15.
Identification of the physical mechanism behind the flarc phenomenon was beyond the scope of this project. The relevance of flarcs for fire risk was also not clear. Flashovers are unrealistic features of ‘branch touching wire’ faults. If they are realistic for other types of earth faults (e.g. wire into low vegetation), the production of flarcs in a REFCL-protected network may have some bearing on fire risk. Ignition tests on such fault types were beyond the scope of this project.
The flarc phenomenon is recorded here for possible future reference as a strikingly different and unexplained outcome of the KMS vegetation ignition tests.
15 The current limit was 32 amps, but the few flashovers that did occur appeared to only reach 5-10 amps.
14REFCL technologies test program designThe 2015 REFCL Technologies test program used a custom-built test facility located at Kilmore South about 50 kilometres north of Melbourne. Site commissioning, proof-of-concept tests and the first three tranches of the main test program comprised 538 tests done over 26 test days in May and June. A further 297 tests were done in Tranche 4 over 20 test days in September and October. The design elements of the test facility and test program are described in the following sections.
14.1 Use of 2014 REFCL Trial design
A detailed description of the experiment design and test site concept is set out in the 2014 REFCL Trial report. The key elements carried over into the KMS tests included:
1. The underlying principles of realism, direct comparison, and worst case conditions2. Selection of conductor type, conductor impact speed and bounce height3. Selection of soil, fuel and soil-fuel bed geometry4. Selection of worst case fire weather conditions and fuel moisture content5. Overall site concept including safety architecture6. Test procedures and site operating procedures7. The ‘wire down’ test rig8. The data acquisition system and transducers
Changes made to the Frankston design were primarily matters of detail. They included:
1. Use of insulated shipping containers for the test rig and soil-bed conditioning spaces – this reduced heater load and allowed removal of the air diffuser around the ignition test space.
2. Use of isolated LV supply to the Control Hut instead of a separate generator3. Use of normal LV supply to the test rig and HV yard equipment instead of a generator4. Overhead HV supply to the test rig instead of short lengths of underground cable5. Omission of the ‘sandpit’ parallel current path in the test rig container6. Use of a GoPro camera in the test ignition space as the recorder of fire results7. Use of a wideband 200.0mΩ coaxial shunt to measure soil current.
The test site is shown in Figure 16, Figure 17 and Figure 18.
Figure 16: Kilmore South test site prior to construction
Figure 17: the Kilmore South test facility under construction
The test facility was commissioned on 11th May 2015. Figure 18 shows its condition during test days.
Figure 18: the Kilmore South test facility as commissioned
14.2 Use of 2015 Vegetation Conduction Ignition test program design
A detailed description of the experiment design and test site concept is set out in the 2015 Vegetation Conduction Ignition Test Program report. The key elements carried over into the KMS tests included:
1. Test procedures and site operating procedures for vegetation earth fault simulation2. The vegetation conduction ignition test rig
Significant changes made to the Springvale design included:
1. Use of 400 Ohm resistor in the test rig high-voltage supply instead of 200 Ohms2. Omission of the automated system to terminate a test when a set current limit was reached3. Omission of custom-designed transducers to capture high frequency fault signatures4. Modified test procedure to include real-time monitoring of infrared camera vision and to
manually terminate a test if fire start criteria were met.
14.3 The test network – feeder KMS21
High voltage supply to the test site was from feeder KMS21 which comprised the total test network. Prior to commencement of the test program the following works were carried out on the test network:
1. Network hardening: all surge diverters were disconnected. As part of GFN commissioning, each phase of the network was ‘soaked’ for 20-40 minutes at 22kV using the RCC to control
neutral displacement voltage. This approach appeared to be successful as there were no cross-country faults experienced during the test program.
2. Network balancing: the residual current (with NER in service) was reduced from 0.5 amps to 0.19 amps by re-phasing two-wire spur take-off points. For Tranche 4, further changes reduced this to 0.12 amps.
14.3.1 Network capacitive balance
With the NER in service, the residual voltage16 at the test facility was typically less than fifteen volts as shown in Figure 19.
Figure 19: Residual voltage and phase voltages at the test facility with NER in service (Tests 58 and 362)
0.831
7.619
0.381
2.469
0.123
0.653
0.383
0.165
0.5240.522
0.9610.776
0.477
0.9300.633
0.3300.407
0.9591.102
0.245
0.851
0.100
1.000
10.000
100.000
0 200 400 600 800 1000 1200
Resi
dual
Vol
tage
(V)
Frequency (Hz)
Test 58: 19 May 12:53pm
1.540
13.043
0.735
2.340
0.479
0.187
0.680
0.245
0.465
0.9520.674
0.257
0.8650.717
0.345
0.7500.8370.977
0.585
0.243
0.749
0.100
1.000
10.000
100.000
0 200 400 600 800 1000 1200
Resi
dual
Vol
tage
(V)
Frequency (Hz)
Test 362: 17 June 4:20pm
The neutral voltage measured at the zone substation was 1.6 volts, consistent with the measured neutral current of 0.19 amps and the nominal 8Ω value of the NER, so the 7-13 volt residual voltage recorded at the test facility was mainly due to unbalanced powerline voltage drop caused by customer load current flowing in the series impedance of the phase conductors between the zone substation and the test facility. This could have caused unbalanced capacitive currents from each phase conductor to ground. Measurement error may have made a small (perhaps one volt) contribution to the recorded residual.
With the GFN in service, the recorded phase to earth voltages and calculated residual voltage at the test facility (Test 45 pre-fault, Monday 18th May 2015, 3:46pm) are shown in Table 9.
Table 9: test network phase and residual voltages early in the test program
The large residual voltage was caused by network capacitive imbalance. It equalled the product of the imbalance current and the network damping resistance. The value shown in Table 9 indicates the damping resistance of the test network was about 12 kilo-Ohms since the capacitive imbalance current was 0.190 amps. This quantity can only be measured once some form of high-impedance earthing, such as the GFN, has been established. The residual was closely aligned with Blue phase volts17 as shown in Figure 20.
16 The residual voltage at the test facility is the instantaneous average of the three phase-to-earth voltages. It is a notional neutral displacement (there is no 22kV network neutral available except at the zone substation). 17 On the HMI screen of the GFN, neutral displacement is indicated by a dot indicating the voltage of the earth with respect to the centre of the three-phase set of voltages, i.e. it is the opposite of the calculated residual voltage which is the voltage of the centre of the three-phase voltage set with respect to earth.
Later in the test program, the GFN’s U0 Injector was used to reduce standing neutral displacement. This device comprises a capacitor connected to the 240 volts supply to the RCC cubicle and the low-voltage winding of the GFN’s ASC. The capacitor can be any combination of two 84µF capacitors, i.e. one half (with the two in series), one capacitor, or twice that value (the two capacitors in parallel). With one capacitor, the U0 Injector adds 6.3 amps of current into the low-voltage ASC winding which is equivalent to 0.22 amps on the high-voltage network.
The phase of the injection is set by the phase of the 240 volts supply, i.e. there are three options separated by 120 degrees. A second U0 injector provided more options for the Tranche 4 tests.
Figure 20: Test facility voltages 18th May 2015 (Test 45 pre-fault)
-6000
-4000
-2000
0
2000
4000
6000
-25000
-20000
-15000
-10000
-5000
0
5000
10000
15000
20000
25000
0 0.004 0.008 0.012 0.016 0.02
Resi
daul
Vol
tage
(V)
Phas
e-to
-ear
th vo
ltage
s (V)
Time (seconds)
Red V White V Blue V Residual
The U0 Injector can only be used to improve fault detection sensitivity. It may not improve fault-location or fault-confirmation sensitivity as the measured powerline currents still remain unbalanced. After the U0 Injector was connected and adjusted, the residual voltage at the test facility was reduced as shown in Table 10. Further reductions were achieved for Tranche 4 by additional network balancing and adjustment of dual U0 injectors.
Table 10: test network phase and residual voltages after application of U0 injector
Figure 21: phase and residual voltages in Tests 101 and 518
-6000
-4000
-2000
0
2000
4000
6000
-25000
-20000
-15000
-10000
-5000
0
5000
10000
15000
20000
25000
0 0.004 0.008 0.012 0.016 0.02
Resi
daul
Vol
tage
(V)
Phas
e-to
-ear
th vo
ltage
s (V)
Time (seconds)
Test 101
Red V White V Blue V Residual
-6000
-4000
-2000
0
2000
4000
6000
-25000
-20000
-15000
-10000
-5000
0
5000
10000
15000
20000
25000
0 0.004 0.008 0.012 0.016 0.02
Resi
daul
Vol
tage
(V)
Phas
e-to
-ear
th vo
ltage
s (V)
Time (seconds)
Test 518
Red V White V Blue V Residual
The use of the U0 Injector was sufficient to allow fault detection sensitivity of better than 0.5 amps, though the residual imbalance meant detection sensitivity was different on the different phases of the network. It also limited the sensitivity threshold that could be applied in tests of the fault-confirmation/powerline-identification test of the GFN.
14.3.2 Network voltage harmonics
Harmonic levels on the KMS21 network were low compared to levels on other rural networks in Victoria. The values shown are compared with spot check readings on other Victorian substations in Table 11. The KMS measurements were taken late afternoon on Mondays in May and October, whereas the measurements at the four other substations were taken at 5:30 pm on a Tuesday in late June.
Table 11: spot check comparison of network voltage harmonics
The KMS21 feeder load current profile is shown in Figure 23. The 2014 Theoretical modelling study identified average peak powerline load in Victorian rural networks as 160-240 amps. The KMS21 peak load was only 5-10% of this.
Figure 23: KMS21 feeder phase currents prior to start of test program
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
1 49 97 145 193 241 289 337 385
Curr
ent (
amps
)
Time (30 minute intervals)
KMS21 phase currents 22 April to 1 May 2015
Red White Blue
0
20
40
60
80
100
120
140
160
0 1 2 3 4 5 6 7 8 9 10
Coun
t (N
= 4
60)
Phase current (A)
Red White Blue Figure 24: KMS21 load current in last week of Tranche 3 tests
0
2
4
6
8
10
12
14
16
1 11 21 31 41 51 61 71 81 91 101
111
121
131
141
151
161
171
181
191
201
211
221
231
241
251
261
271
281
291
301
311
321
331
341
351
361
371
381
391
401
411
421
431
441
451
461
471
481
491
501
511
521
531
KMS2
1 ph
ase
curr
ent (
A)
Time (intervals of 10 Minutes)
KMS21 phase currents 0600 22nd - 1830 25th June 2015
The following tables summarise the tests and results, except for lo-sag covered conductor tests. An audited spreadsheet of all tests (Test log BAS edit 26-10-15.xlsx) is the master record for this data.
15.1.1 Valid ignition tests
All ignition tests were audited against a number of criteria to verify the result could reliably be regarded as valid.
695 8/10/2015 9:25:50 N Y 135910 0.03 0.57696 8/10/2015 9:34:29 N Y 130980 0.04 0.32697 8/10/2015 9:42:10 Y Y 119480 0.06 0.53698 8/10/2015 9:50:35 Y Y 90380 0.03 0.6699 8/10/2015 9:57:55 Y Y 105870 0.03 0.74700 8/10/2015 10:06:24 Y Y 76240 0.04 0.77701 8/10/2015 10:13:22 Y Y 96030 0.04 0.54702 8/10/2015 10:19:53 N Y 44520 0.06 0.5703 8/10/2015 10:27:55 Y Y 117860 0.03 0.64764 12/10/2015 13:25:53 N Y 134350 0.03 0.52
132 28/05/2015 13:44 Blue 7.2 80 N 0133 28/05/2015 14:00 Blue 6.4 73 N 0143 28/05/2015 16:20 Blue 5.3 82 N 0384 19/06/2015 9:28:00 White 6.2 77 2.5 N 0385 19/06/2015 9:34:56 White 6.1 159 4.9 Y 80386 19/06/2015 9:39:57 White 5.7 161 4.5 Y 80387 19/06/2015 9:47:10 White 6.4 159 6.0 Y 80388 19/06/2015 9:52:59 White 6.7 156 7.0 Y 80391 19/06/2015 11:10:39 White 7.0 141 6.1 Y 60392 19/06/2015 11:39:59 White 7.2 132 6.1 Y 60393 19/06/2015 11:43:06 White 7.1 135 4.8 Y 60394 19/06/2015 11:48:31 White 6.9 135 6.0 N 60395 19/06/2015 11:56:44 White 7.3 139 6.3 Y 60396 19/06/2015 12:52:39 White 7.3 120 5.6 Y 40397 19/06/2015 12:57:51 White 7.6 116 5.4 Y 40398 19/06/2015 13:03:51 White 7.1 121 5.4 Y 40399 19/06/2015 13:08:10 White 7.2 120 5.5 Y 40400 19/06/2015 13:12:13 White 7.0 124 5.6 Y 40401 19/06/2015 13:22:17 White 7.3 95 4.5 Y 20402 19/06/2015 13:28:04 White 7.2 97 4.7 N 20403 19/06/2015 13:33:01 White 7.8 97 5.6 Y 20404 19/06/2015 13:38:02 White 7.6 91 4.9 N 20405 19/06/2015 13:43:24 White 7.6 98 5.3 N 20406 19/06/2015 13:48:21 White 7.6 102 5.6 N 20407 19/06/2015 13:56:12 White 7.1 89 4.2 N 10408 19/06/2015 14:02:10 White 7.6 91 5.0 Y 10409 19/06/2015 14:07:52 White 7.3 89 4.5 N 10410 19/06/2015 14:12:15 White 7.4 89 4.2 Y 10411 19/06/2015 14:18:13 White 7.5 93 4.8 Y 10412 19/06/2015 14:23:23 White 7.4 82 3.9 N 10414 22/06/2015 12:39:18 White 7.0 73 3.1 N 0415 22/06/2015 14:06:00 White 6.7 80 3.0 N 0416 22/06/2015 14:15:41 White 6.9 83 3.8 N 0417 22/06/2015 14:22:19 White 6.8 84 3.6 N 0418 22/06/2015 14:28:02 White 6.8 74 3.2 N 0419 22/06/2015 14:44:12 White 6.8 78 3.4 N 0420 22/06/2015 14:52:16 White 7.2 77 3.5 N 0421 22/06/2015 14:58:56 White 7.3 77 3.7 N 0422 22/06/2015 15:07:50 White 7.0 85 3.6 Y 0423 22/06/2015 15:14:58 White 7.3 77 3.6 N 0424 22/06/2015 15:25:24 White 6.8 93 3.5 Y 10425 22/06/2015 15:34:51 White 7.0 88 3.8 N 10426 22/06/2015 15:42:29 White 7.6 86 4.1 N 10427 22/06/2015 15:47:30 White 7.1 90 3.6 Y 10428 22/06/2015 15:54:21 White 7.2 92 4.2 N 10
661 7/10/2015 9:58:16 White N 39430 0.06 0.45 Salix662 7/10/2015 10:09:10 White N 109930 0.04 0.37 Salix663 7/10/2015 10:20:30 White N 38240 0.05 0.22 Salix664 7/10/2015 10:39:28 White N 100470 0.03 0.27 Salix665 7/10/2015 10:54:35 White N 40360 0.04 0.29 Salix666 7/10/2015 11:09:00 White N 130320 0.04 0.54 Salix667 7/10/2015 11:42:00 White N 40920 0.06 0.3 Salix668 7/10/2015 11:54:22 White N 139060 0.04 0.34 Salix669 7/10/2015 12:06:04 White Y 161190 0.07 0.36 Salix670 7/10/2015 12:18:20 White Y 137980 0.05 0.36 Salix671 7/10/2015 13:38:59 Blue Y 114640 0.04 0.36 Salix672 7/10/2015 13:47:52 Blue N 106200 0.04 0.42 Salix674 7/10/2015 14:02:28 Blue Y 140700 0.05 0.39 Salix675 7/10/2015 14:11:24 Blue Y 113580 0.04 0.51 Salix676 7/10/2015 14:25:46 Blue N 39500 0.07 0.3 Salix677 7/10/2015 14:34:58 Blue N 32460 0.09 0.33 Salix678 7/10/2015 14:49:35 Blue N 170810 0.07 0.48 Salix679 7/10/2015 15:00:57 Blue Y 158460 0.1 0.4 Salix680 7/10/2015 15:10:52 Blue Y 149620 0.09 0.44 Salix681 7/10/2015 15:22:02 Blue N 137470 0.09 0.39 Salix682 7/10/2015 15:31:19 Blue N 24410 0.21 0.39 Salix683 7/10/2015 15:47:12 Red N 26980 0.14 0.34 Salix684 7/10/2015 15:55:52 Red N 55790 0.07 0.32 Salix685 7/10/2015 16:02:42 Red Y 136990 0.03 0.48 Salix686 7/10/2015 16:17:59 Red N 107430 0.02 0.39 Salix688 7/10/2015 16:37:25 Red N 134240 0.05 0.41 Salix689 7/10/2015 16:47:13 Red N 32230 0.06 0.34 Salix690 7/10/2015 16:54:15 Red N 26640 0.13 0.38 Salix691 7/10/2015 17:03:35 Blue Y 185430 0.1 0.45 Salix693 7/10/2015 17:22:37 Blue N 22500 0.14 0.39 Salix694 7/10/2015 17:28:23 Blue N 40390 0.13 0.41 Salix705 8/10/2015 11:13:51 White Y 138290 0.06 0.33 Salix706 8/10/2015 11:21:37 White Y 200000 0.05 0.3 Salix707 8/10/2015 11:34:16 White Y 125950 0.03 0.46 Salix708 8/10/2015 11:43:25 White N 154810 0.04 0.33 Salix709 8/10/2015 11:51:12 White Y 123530 0.03 0.13 Salix711 8/10/2015 13:29:39 White N 210850 0.05 0.3 Salix712 8/10/2015 13:41:23 White Y 217240 0.03 0.32 Salix713 8/10/2015 13:52:53 White N 123990 0.03 0.34 Salix714 8/10/2015 14:01:22 White N 116100 0.04 0.31 Salix715 8/10/2015 14:14:22 Blue N 134750 0.05 0.51 Salix
716 8/10/2015 14:25:16 Blue Y 136520 0.05 0.37 Salix718 8/10/2015 14:45:31 Blue N 118180 0.04 0.46 Salix719 8/10/2015 14:55:22 Blue Y 169070 0.06 0.57 Salix742 9/10/2015 11:53:06 White N 157200 0.03 0.44 Salix743 9/10/2015 12:00:53 White N 40270 0.07 0.34 Salix744 9/10/2015 12:06:29 White Y 130640 0.05 0.47 Salix745 9/10/2015 12:16:51 Blue Y 161120 0.06 0.51 Salix746 9/10/2015 12:35:18 Blue Y 130960 0.05 0.49 Salix747 9/10/2015 12:43:10 Blue N 120400 0.03 0.58 Salix748 9/10/2015 12:50:49 Blue N 160140 0.06 0.48 Salix749 9/10/2015 13:00:40 Blue Y 133490 0.04 0.58 Salix750 9/10/2015 13:10:28 Blue N 128780 0.05 0.4 Salix759 12/10/2015 12:16:59 Blue N 160570 0.05 0.48 Salix761 12/10/2015 12:36:15 Blue N 52890 0.08 0.4 Salix766 12/10/2015 14:50:05 Blue Y 111830 0.02 0.45 Salix767 12/10/2015 14:57:37 Blue N 53320 0.07 0.36 Salix768 12/10/2015 15:05:28 Blue N 47370 0.07 0.36 Salix769 12/10/2015 15:13:52 Blue N 128340 0.02 0.44 Salix770 12/10/2015 15:22:12 Blue N 121080 0.02 0.48 Salix771 12/10/2015 15:29:37 Blue N 117110 0.02 0.53 Salix772 12/10/2015 15:37:48 Blue N 105450 0.03 0.35 Salix788 13/10/2015 12:20:20 White N 64740 0.11 0.36 Salix790 13/10/2015 12:57:02 White N 127970 0.03 0.31 Salix791 13/10/2015 13:05:11 White Y 214170 0.07 0.47 Salix792 13/10/2015 14:19:49 White N 60510 0.05 0.28 Salix
751 9/10/2015 13:49:47 Blue N 51030 0.09 0.43752 9/10/2015 13:56:02 Blue N 29810 0.15 0.43753 9/10/2015 14:03:24 Blue N 19870 0.24 0.4754 9/10/2015 14:10:23 Blue N 21470 0.14 0.41755 9/10/2015 14:17:27 Blue N 23180 0.15 0.4756 9/10/2015 14:23:45 Blue N 48710 0.06 0.37760 12/10/2015 12:27:48 Blue N 44350 0.11 0.41762 12/10/2015 12:49:49 Blue N 59480 0.09 0.39765 12/10/2015 14:53:54 Blue N 47820 0.12 0.41
The following tests were performed with the test rig shorted out by a length of flexible welding cable. They were performed to for setup purposes and to explore and measure the performance of each REFCL type.
The following tests were performed for the purposes of setup or they were ruled invalid for some reason and excluded from the sets of ignition results used in some analyses.
Table 21: Setup and invalid tests – earlier versions of GFN firmware (up to Test 292)
Test No. Date Time Test Type Protection Compensation ResistanceSupply Phase
Table 22: set-up and invalid tests - other reasons (all ignition tests on white phase)
Test No. Date Time Protection Compensation Resistance Invalid reason17 11/05/2015 18:38 NER - 100 No grass18 11/05/2015 18:44 NER - 100 No grass19 11/05/2015 19:20 NER - 100 No grass20 11/05/2015 19:22 NER - 100 No G3i record21 11/05/2015 19:27 NER - 100 G3i noise trigger22 14/05/2015 13:21 NER - 100 Same bed23 14/05/2015 13:40 NER - 100 Short test32 14/05/2015 14:48 NER - 100 No G3i record49 19/05/2015 11:00 NER - 100 Arm sitting high50 19/05/2015 11:21 NER - 100 Arm sitting high51 19/05/2015 11:29 NER - 100 Arm sitting high52 19/05/2015 11:40 NER - 100 Arm sitting high56 19/05/2015 12:40 NER - 100 Arm sitting high57 19/05/2015 12:47 NER - 100 Arm sitting high58 19/05/2015 12:53 NER - 100 Arm sitting high
Test No. Date Time Protection Compensation Resistance Invalid reason216 3/06/2015 17:16 ASC FPE 100 Tripped power no record297 15/06/2015 17:28:40 GFN3 RCC 100 Not detected on bounce299 16/06/2015 8:59:44 GFN3 RCC 100 Not detected in FCT317 16/06/2015 17:06:30 GFN3 RCC 100 GFN malfunction339 17/06/2015 12:41:39 GFN3 RCC 100 Not detected on bounce383 19/06/2015 9:19:00 SSFCL FPE 100 No FPE operation443 23/06/2015 9:36:58 GFN3 RCC 100 No RCC operation465 23/06/2015 15:47:26 GFN3 RCC 100 Odd GFN behaviour489 24/06/2015 13:40:00 GFN3 RCC 100 Same bed490 24/06/2015 13:50:57 GFN3 RCC 100 Same bed491 24/06/2015 13:56:35 GFN3 RCC 100 Same bed492 24/06/2015 14:00:20 GFN3 RCC 100 Same bed508 24/06/2015 16:00:48 GFN3 RCC 100 Same bed509 24/06/2015 16:04:53 GFN3 RCC 100 Same bed516 24/06/2015 16:51:08 GFN3 RCC 100 GFN malfunction532 25/06/2015 10:38:43 GFN3 RCC 100 Arm sitting high536 25/06/2015 12:51:53 NER - 3300 Duration too long537 25/06/2015 13:34:23 NER - 100 Arced to bow538 25/06/2015 13:52:49 NER - 100 Fulgurite
Table 23: Tranche 4 setup and invalid tests
Test No. Date Time Test Type ProtectionCompensation
DAQ Notes659 7/10/2015 9:37:00 Vegetation GFN4 RCC Willow branch. Phase to earth. Not valid test.660 7/10/2015 9:42:30 Vegetation GFN4 RCC Not a valid test.673 7/10/2015 13:55:35 Vegetation GFN4 RCC Branch fell off. Invalid test.687 7/10/2015 16:28:57 Vegetation GFN4 RCC Sample was tied. Sample fell offf. Ivalid test.692 7/10/2015 17:13:58 Vegetation GFN4 RCC Branch fell off. Invalid test.704 8/10/2015 10:37:11 Vegetation NER NIL Sample broke in half. Not a valid test.710 8/10/2015 12:04:20 Vegetation GFN4 RCC ACR tripped. No GEN3i data. Not a valid test.717 8/10/2015 14:37:04 Vegetation GFN4 RCC Sample fell off. Invalid test.741 9/10/2015 11:44:00 Vegetation GFN4 RCC Willow sample. Invalid test because GFN detection sensitivity was 15V, instead of 40V.763 12/10/2015 13:17:04 Vegetation NER NIL Desert ash branch. Same branch as test 762. Not a valid test. Current span = 5A. Flashover.773 12/10/2015 15:50:17 Vegetation GFN4 RCC GFN didn't perform FCT and released fault. GFN malfunction? 248° ember.774 12/10/2015 17:07:43 Ignition GFN4 RCC Not a valid test. Setting up ignition tests.775 12/10/2015 17:11:44 Ignition GFN4 RCC Not a valid test.776 12/10/2015 17:14:59 Ignition GFN4 RCC Not a valid test. Arm fall time = 12.579s.777 12/10/2015 17:18:00 Ignition GFN4 RCC Not a valid test. Arm fall time = 12.579s.778 12/10/2015 17:28:54 Ignition NER NIL Not a valid test.780 13/10/2015 10:06:01 Ignition GFN4 RCC Same bed as 779. Invalid test. GFN didn't find fault.789 13/10/2015 12:26:28 Vegetation GFN4 RCC GoPro video says 773. IR camera span was on <120°. Willow branch. Nearly flashover at ~1.14A.794 13/10/2015 15:20:58 Ignition GFN4 RCC Arm lifted at 62s mark. No fire untill that point.796 13/10/2015 16:02:58 Ignition GFN4 RCC RCC miscalibrated by 65V. GFN detection sensitivity set to 15V.797 13/10/2015 16:06:54 Ignition GFN4 RCC Same bed. Not a valid test. RCC miscalibrated by 130V.816 14/10/2015 11:35:02 Ignition NER NIL Wet soil bed. Demo test. Not valid.817 14/10/2015 11:51:00 Ignition GFN4 RCC Used soil bed. Demo test. Not valid.824 15/10/2015 16:25:24 Ignition NER NIL Arm timing test. Arm fall time =12579ms. Dry bed825 15/10/2015 16:30:13 Ignition NER NIL New bed. 100A current span.826 15/10/2015 16:39:07 Ignition GFN4 RCC 100A current span. GFN setting was still at elevated residual.827 15/10/2015 16:49:47 Ignition GFN4 RCC828 15/10/2015 17:06:13 Vegetation GFN4 RCC Rehearse vegetation test829 16/10/2015 8:46:36 Ignition NER NIL Practice test 1830 16/10/2015 8:54:10 Ignition GFN4 RCC Practice test 2831 16/10/2015 9:07:20 Vegetation GFN4 RCC Practice test 3832 16/10/2015 10:29:32 Ignition NER NIL Test for media 1833 16/10/2015 10:35:25 Ignition GFN4 RCC Test for media 2834 16/10/2015 11:42:39 Ignition NER NIL Repeat test 1 for media835 16/10/2015 11:47:56 Ignition GFN4 RCC Repeat test 2 for media