The IEC 60909 standard is well established, and used across most of industry to calculate the maximum fault currents that can occur on a system. The primary use of the standard is to select switchgear fault ratings. In recent years, problems have been encountered using the standard when calculating fault levels on networks with a large penetration of Converter Fed Generation (CFG) and low Grid infeed fault levels.
Background
Specific problems have been identified on sites with a large amount of Grid Following Inverters (GFL), on a very weak DNO networks, as the calculated results appear significantly larger than would be expected, and exceed the apparent let through of upstream transformers. With the increased use of Grid Forming Inverters (GFM), it has also been identified that these are not defined in the IEC 60909 standard, and some interpretation is therefore necessary on how to model them.
The IEC 60909 standard does not explicitly address CFG in detail, and only advises that it be considered as a constant current source, and that the standard is only valid if there is a small percentage of CFG on the system. It should be noted that IEC 60909 is a relatively “old” standard, that was last updated in 2016; it is planned for updated in 2026, but this is likely to be delayed, and problems are being encountered now and so it is not practical to wait for an updated standard to arrive.
Fault analysis standards are inherently conservative and take a number of worse case assumptions to ensure that any specified equipment is able to meet the required fault duty, therefore deviating from formal fault analysis standards must be done with extreme caution.
There are three principle issues with networks with a large amount of CFG.
- The constant current output of GFL can cause retained voltages on the network to increase beyond a ‘reasonable’ value (i.e. Un = 1.1pu), which increases the apparent fault level on the network
- The IEC method assumes that the GFL outputs fault current in all scenarios. In reality all GFLs are design to only output fault current, should a network voltage dip occur where the GFL terminal is Un < 0.9 (or similar).
- GFM technology is designed to mimic synchronous machine response through various control loops. Their response is similar, but not identical to, a synchronous machine; the maximum fault contribution is typically much lower at 2.0pu, and maintained constantly, rather than as a decaying component moving through sub-transient and transient response and the response is triggered partly by the retained voltage at the local busbar.
It is important to note that most simulation software packages calculate fault levels at all busses in what appear to be a ‘simultaneous’ basis. However, what actually happens is the software calculates a fault at each location in sequence and then plots / tabulates the results as a single set of data. This can mask the problems that may be occurring with the calculation process, and it is therefore recommended to carry out an analysis of a single fault in a remote location in order to understand the issue, should results be of concern.
Lastly, it should also be noted that this explanation is focused on bolted, 3phase faults for simplicity, but can also apply to unbalanced faults.
Constant Current Output of Grid Following Inverters
When there is a large presence of GFL on a network, and particularly in weak systems (low grid fault level), the problem that can occur is the constant current output of the GFL can displace the retained voltage on the unfaulted busbars. In strong systems, and networks with only a small value of GFL this impact is negligible and not observed, however in weak networks, or networks with high GFL penetration this can be observed more closely. In some cases a network with a low fault level can show a higher fault level on a local busbar, as the
Consider a simple test network as shown below. The model is based on a 33 kV distribution systems, with three identical 2.5 MVA, Dyn11, Z=7.0% transformers, supplying a local 400V switchboard, interconnected via 3c185mm2 aluminum cable, each 10 km long (chosen to highlight the effect). The cable type and lengths are deliberately selected to show differing retained voltages on the HV systems. The grid infeed is set with a nominal fault current of 5kA X/R ratio of 10, through all studies. A 100 MVA GFL and 100 MVA GFM are added to the HV1 busbar, to show the model behaviour.
With the GFL in service, if a fault is applied at LV3 , the bus voltage at LV3 will collapse to 0, but across the rest of the network will remain higher, depending on the system strength and impedances between the source, and the busbars, and fault. Considering the plot below the retained voltages at HV1, HV2 and HV3 are 1.3pu, 1.25pu and 1.2pu respectively, due to the constant current output of the CFG. Such voltages are not credible, and are a result of the calculation method adopted by IEC 60909.
Figure 1: IEC 60909 Base Calculation with full CFG and high retained Voltages
Voltage Controlled Output of Grid Following Inverters
A second and equally important point is that the output of a GFL during a fault is set by the control logic within the GFL. Typically the GFL fault output is triggered by a threshold setting Vth, which is usually set to 0.9pu. Therefore during a fault on a remote busbar the retained voltage may not fall enough to actually trigger a response from the GFL, whilst the IEC approach assume that it will always contributed, this can therefore lead to unrealistic values in the output. The formulation within DIGSILENT for the Complete Method is shown below
An example of a typical control logic is shown below (from the WECC ‘REEC_C’ model), with the reactive current injection loop shown in the red box. Whilst this is a typical, most CFG follow a very similar principle.
Grid Following Inverter Method Solutions
The IEC 60909 method can therefore be seen as inaccurate for GFL dominated networks, particularly with a low fault level and will potentially results in unrealistic calculated values. There are two potential solutions that can be adopted to address this issue. Our preferred solution is to follow a modified IEC 60909 method (solution 1 below).
- Modified IEC 60909 method: Carry out a fault assessment of a single busbar of interest. Use the calculation to determine the retained voltages at busbars in the system. On any remote busbars with other CFG connected, if the retained voltage is >0.9pu temporarily set the CFG out of service. Recalculate the fault level on the same busbar to obtain a more accurate fault value.
- ENA G74 (DigSILENT Complete method): The ENA G74 method is intended more for DNOs & TSOs, and is less widely accepted as it is UK specific and contains a number of complications above IEC 60909. However, the G74 method has provision to account to some extent for this phenomena, as the CFG output current is calculated as a function of a the CFG reactive fault current gain and retained voltage using the formula shown earlier (If = K . ∆u . In), this allows a more accurate output (although it is not totally correct ).
- RMS Based method: RMS time-based calculations can be used to accurately predict fault levels. These methods are not generally used, as they are easy to get wrong and do not consider worst case scenarios, that can realistically occur. They also require a reasonably accurate dynamic model of the CFG, which may not always be readily available .
A comparison of the results for the different analysis methods are shown below. In the IEC case the c-factor is set to 1.1 and in the G74 and RMS cases the pre-fault voltage is set to 1.1 pu. It can be seen that the ENA G74 method, Modified IEC 60909 method and RMS method all give similar results.
| Scenario | HV1 Retained Voltage (pu) | LV3 Fault Level (kA) |
| IEC 60909 Method | 1.44 | 71.3 |
| ENA G74 Method | 1.06 | 50.6 |
| IEC 60909 Modified Method | 0.93 | 45.9 |
| RMS Calculation | 0.98 | 47.9 |
Grid Forming Inverters
Grid Forming Inverters (GFM) are fundamentally different to GFL technology, and are not captured in any formal fault level analysis standard. Furthermore, the technology is evolving rapidly and there are few agreed performance criteria, although work is being done on this area by groups such as NERC, UNIFI and WECC. As the technology is fundamentally different to GFL, it is not correct to model them as a constant current source in the same way as GFM.
During faults a GFM fault contribution is set by the internal voltage compared to the external reference voltage, so in a similar manner the fault contribution is determined by the retained voltage on the unfaulted bus; this is fundamentally different to a synchronous machine, where the fault contribution is set by the machine electromagnetic properties and field.
At present, Aurora believe that the best option available is to model them as a type of equivalent synchronous machine using the IEC 60909 method, with the sub-transient fault contribution as advised by the manufacturer (typically 2.0pu), with an X/R ratio of 1. However, it should be noted that this is not fully correct and can overestimate the fault contribution and does not correctly consider the time based decay. On sensitive sites, an RMS based fault current analysis may be the only way to fully understand the fault behaviour, however such an approach must be handled with care to avoid under-estimating fault levels.
Summary
In summary, the IEC 60909 method can give very misleading results for short circuit analysis in weak grids dominated by GFL type inverters. The ENA G74 method, can potentially give more accurate results, but also suffers from a drawback as it still assumes a GFL will contribute regardless of the network retained voltage, and this can cause issues on large systems. GFM technology presents further complications, but the use of an equivalent synchronous machine is currently considered the most credible approach for modelling.
At present Aurora consider that for GFL technology, the modified IEC 60909 method remains the best approach for accurately calculating maximum fault levels in CFG dominated networks, and for GFM technology modelling as an equivalent synchronous machine is the most appropriate method.
The use of RMS modelling will always give a more accurate result, but requires more input information, and it is easy to make incorrect assumptions leading to an overly optimistic result (low fault level). Therefore RMS calculations should be used to support and validate the IEC 60909 results, rather than to replace them.