Anti-islanding protection of photovoltaic inverters

One of the mandatory requirements for photovoltaic inverters is to be equipped with protection against unintentional islanding.

Unintentional islanding refers to the improper operation of photovoltaic generators in the absence of the electrical network provided by the local utility, as shown in figure 1.

Unintentional islanding, in general, has a small number of occurrences. Even so, inverters connected to the electrical grid need to be prepared to deal with this problem.

As the energy in the isolated area is not maintained or regulated by the concessionaire, the voltage and current can vary in such a way as to damage the most sensitive loads of consumers and equipment on the distribution network itself.

Furthermore, the presence of the photovoltaic generator connected to a disconnected distribution network could put people at risk.

The time limit for detection and disconnection due to islanding is up to 2 seconds, defined in the ABNT NBR 16149 standard – Photovoltaic systems – Characteristics of the connection interface with the electrical distribution network.

Islanding detection

Islanding detection is done by the control system built into the inverter. Islanding detection methods can be classified as active and passive. Passive methods are based on observing the state of the electrical network, checking whether the system is islanded based on measuring the voltage and frequency of the network.

The electrical grid may suffer disturbances when the system is islanded. These disturbances depend on the relationship between the active and reactive powers that the inverter injects and that the load uses. Figure 2 shows the equivalent electrical circuit of a photovoltaic system at the time of islanding.

During normal system operation, the load consumes the sum of the power injected by the inverter (P,Qinverter) and the power supplied by the distribution network (if applicable and represented by P,Qnetwork). In other words, in normal operation: P,Qload = P,Qinverter + P,Qgrid.

At the moment the energy supply from the distribution network stops, the inverter is still supplying the local loads, represented by the RLC set. When there is no longer any influence from the utility network (P,Qgrid = 0, system disconnected) the system voltage will vary according to the relationship between the active power supplied by the inverter (Pinverter) and the active power that the resistive part of the system consumes ( Pcharge).

In the event of islanding, the frequency of the current injected into the islanded system will also vary, respecting the relationship between the reactive power supplied by the inverter (Qinverter) and the reactive power consumed by the equivalent capacitance and inductance of the load (Qload).

As the inverter will very rarely be found feeding a network in which both the active and reactive power generated by them are perfectly matched to the load supplied, passive methods are usually sufficient to detect islanding.

Even so, the standard for photovoltaic inverters requires the existence of active methods, which are theoretically capable of detecting any islanding situation, even if there is a perfect match between inverter generation and local consumption (in which case islanding would not be noticed). by passive methods).

Table 1 – Relationship between active and reactive power at the time of islanding and its electrical effect on the system

If the power injected by the inverter is exactly the same as that consumed by the load, it is possible for the system to continue operating in an islanded manner. Figure 3 shows an experiment carried out in the laboratory with an inverter in which it was possible to turn off the islanding protection. This is not possible in commercial inverters, but only in inverters developed in the laboratory.

The load was intentionally tuned so that the active and reactive powers that the inverter injects were equal to that consumed. In other words, this is the case in which the inverter does not notice the absence of the electrical distribution network and continues to supply the local load in an islanded manner, as no electrical changes are produced in the network after the islanding occurs.

Only active detection methods (which in this case were turned off) are capable of detecting islanding in this very particular situation. In this test, the consumer loads are represented by an RLC load (resistor, inductor and capacitor in parallel) capable of absorbing all the power supplied by four photovoltaic inverters (all of them with their islanding detection systems turned off).

It can be observed that, as the injected and consumed powers are identical, there was no variation in the voltage and frequency of the island, which remained at around 60 Hz. Without an activated anti-islanding system, this situation could last for a long time. , which is prohibited by regulations.

In figure 3, the upper graph (red) shows the voltage at the connection point of the inverters with the electrical installation. The bottom graph (blue) shows the electrical current consumed by the load. Islanding (disconnection from the distribution network) occurred at the time highlighted in the graph.

We observed that no change occurred in the frequency or amplitude of the electrical voltage and the inverters continued to supply the load in an islanded manner. To ensure that the inverter does not operate in isolation under any circumstances, it is necessary to use active detection methods.

Active methods use direct disturbances at the connection point to destabilize some of the network parameters when there is islanding. Measuring the effect of this disturbance is used as a criterion for the inverter to detect islanding or not.

Each inverter uses some combination of these two approaches (active and passive), usually robust enough to avoid undue injection of power in times of main grid failure.

Non-detection zone

To prevent the inverter from turning off due to any variation in voltage and frequency on the utility network, standard NBR 16149 establishes a range of voltage and frequency variations in which the inverter can continue operating. These operating ranges can be described through the non-detection zone (NDZ) highlighted in the graph below.

The rectangular zone in figure 4 represents the variation ranges that the inverter must support without activating the anti-islanding protection. The variation in active power is represented by the limits UV (undervoltage – undervoltage) and OV (overvoltage – overvoltage). The variation in reactive power is represented by the limits UF (under frequency) and OF (over frequency). To guarantee system safety, the inverter's NDZ must respect the variation limits defined by the NBR 16149 standard.

Conclusion

However, the more inverters there are, new problems appear. One of them consists of mutual interference between inverters so that one piece of equipment undermines the protection measures of other inverters [2]. New studies are being carried out to mitigate such interference to minimize the probability of unintentional islanding occurring in the presence of multiple inverters connected in the same area.

Reference

• MVG dos Reis, “Study and implementation of islanding detection strategies in inverters for distributed generation photovoltaic systems” – master’s thesis, State University of Campinas (UNICAMP), 2016