FIELD OF THE INVENTION

The present invention relates to energy distribution, and particularly to detecting islanding conditions in an electrical network. BACKGROUND INFORMATION

Lately, distributed generation (DG) based on renewable energy resources has shown a significant growth facilitated by policy makers, global concerns about climate change, availability of affordable energy shortage technologies, interest in clean energy production, etc. Energy suppliers using power plants based on fossil fuel (coal, natural gas, etc.) are also investing in an extension of energy generation portfolio by renewable alternatives such as wind turbines and photovoltaic systems.

However, there can be several requirements to be met before such systems can be connected to the utility grid. These requirements are typically published by standardizing institutions, such as IEC and IEEE, but also by local regulating authorities. One of the requirements, mandatory in many parts of the world, is that distributed generators, especially those connected to low voltage distribution grids, should be able to detect islanding conditions.

Islanding refers to a condition of a distributed generator, where the generator continues powering a part of a distribution network even though power from an electric utility is no longer present. Figures 1 a and 1 b show a difference between a grid-connected mode and an islanding mode.

In Figure 1 a, a switching device 1 , e.g. circuit breaker or fuse, is closed, and distributed generators operate in grid-connected mode.

In Figure 1 b, the switching device 1 is opened and the lower part 2 of the network is no longer connected to the main grid. If the power generated by the distributed generators closely matches the power required by the load, the network can continue operation in islanding mode. If the powers do not match closely, under/over voltage and under/over frequency relays of the dis- tributed generators typically stop the power generation. Therefore, the probability of having islanding conditions can be very small.

However, unintentional islanding can be dangerous to utility workers, who may not realize that the particular part of the network on which they are working is still powered even though there is no power from the main grid. Also, islanding can lead to damages to customer equipment, especially in situations of re-closing into an island. For these reasons, distributed generators may have to be able to detect islanding and immediately stop power production.

There has been a lot of interest in micro-grids, i.e. distribution grids that can operate in controllable, intentional islanding conditions, decoupled from the main grid. In such grids, islanding detection can also be an important issue. Detection of islanding conditions may be required in order to switch the control modes of distributed generators from power injection to voltage and frequency control during disconnection and vice versa during reconnection to the main grid.

The need for reliable islanding detection combined with development of distributed power generation industry has led to an intensive research and development of methods for identification of islanding conditions. These methods can be categorized in three main groups: passive, active, and com- munication-based methods.

Passive methods typically monitor one or more grid variables and, on the basis of deviation of the variables from allowed thresholds, a decision of disconnecting (detection of islanding) can be made. A passive method may look for an abnormal change in, for instance, frequency, voltage or phase an- gle but also in some particular harmonics or the total harmonic distortion (THD). If the monitoring algorithm detects large or sudden changes of these variables, the method may determine islanding conditions to be present. The islanding conditions may be determined on the basis of a combination of passive methods and multi-criteria decision making.

Passive methods may, for instance, be implemented by an algorithm within the controller of a distributed generator or in a dedicated external device. Passive methods are typically easy to implement and are quite effective in majority of situations that may occur in the grid.

However, non-detection zones (NDZ) of the passive methods are quite large. In situations where the power absorbed by the load closely matches the power generated by a distributed generator, the variations in voltage, frequency or phase angle may be lower than those specified in the standard because the network remains balanced even though the connection with the main grid has been lost. In such a case, the distributed generator may not trip when an island has been formed. Passive methods are generally considered as an insufficient means for anti-islanding protection. Active methods appeared as result of a need to minimize the non- detection zone of islanding detection methods. Active methods deliberately disturb the grid and, on the basis of the grid response to that disturbance (variation of grid electrical quantities), decide whether or not islanding has oc- curred.

For instance, disturbances in terms of shifts from normal operating values to grid voltage magnitude, frequency, or phase angle can be added by a distributed generator and, in case of grid connected situation, these disturbances should be corrected by the grid through the voltage and frequency con- trol.

However, if a voltage magnitude, frequency or phase angle follow the shift introduced by the distributed generator, it may be determined that the grid has been disconnected, hence an island has been formed.

An active method may, for instance, be implemented using a posi- tive feedback in a controller of a distributed generator. The controller tries to alter grid variables, such as frequency, phase, or voltage magnitude, in order to obtain, for instance, a frequency jump or phase jump, or a frequency bias. If the grid follows the changes generated by the distributed generator, the grid voltage will exceed imposed operating ranges and result in detection of island- ing conditions. For instance, if the grid frequency follows the inverter current, an island has formed and the distributed generator should disconnect. Alternatively, a positive voltage feedback altering voltage magnitude at the point of common coupling may also be formed.

Alternatively, non-characteristic harmonics may be injected by the distributed generator and the grid response may be registered. Injection of non-characteristic inter-harmonic current can be used to derive the grid impedance at that particular frequency.

Impedance detection may be used as another active method approach. This approach is promoted by the requirements in the German stan- dard. A current spike may, for instance, be periodically injected at the point of common coupling by a grid tied power converter. The grid impedance value is determined using Fourier transform, on the basis of the voltage response to this disturbance. A phase angle signal used to generate the reference for current controller can also be slightly altered to be able to estimate the grid im- pedance on the basis of a grid reaction to the generated current. A high frequency signal may also be injected at zero crossing in order to determine the value of grid impedance, and active and reactive power oscillations may be used in to identify the value of grid impedance.

Although active methods may result in a more reliable islanding identification, they may also distort the delivered power by detecting the island- ing conditions. Power system disturbance may be at an unacceptable level when more than one distributed generators are connected on a same feeder.

Because synchronization in respect to the inter-harmonic injection is possible, each distributed generator may inject a unique inter-harmonic in order to enable connection of more than one similar generation units on the same grid. However, only a finite number of distributed generators can be connected on the same feeder. The number of distributed generators is directly depending on the standard demands regarding disconnection time and the number of injections necessary to obtain accurate impedance identification.

Large number of distributed generators using active methods may not only decrease the quality of power in the grid but also increase the non- detection zone of all active methods.

Communication-based methods make use of a communication means (owned, for instance, by a distribution system operator) which signalises operational states of the switching equipment to the distributed genera- tors.

A power line may be used as a carrier for communication between, for instance, a distributed generator and a utility grid. A continuous signal is transmitted by utility network via the power line. A receiver may be placed inside the distributed generator in order to detect a loss of this signal and, hence, islanding.

When a utility re-closer is equipped with a transmitter which communicates with DG when opens, a signal may, alternatively, be produced on disconnect.

Yet another approach is a SCADA (Supervisory Control And Data Acquisition) based method. For instance, voltage sensors may be placed at the location where a distributed generator is connected and those sensors may be integrated in the SCADA system. The SCADA system may then monitor for islanding conditions and alarm the distributed generators to disconnect in case of islanding.

A disadvantage of the communication based methods is that they typically require involvement from utility providers in implementation of island- ing detection schemes, thus, making the methods less favourite for practical implementation. Because implementing communication adds also costs to both the distributed generator and the grid infrastructure, these methods are not commonly used today.

Another disadvantage of a SCADA based approach can be a relatively slow response time. Typically, a response time for a SCADA system is around 5 to 10 seconds in an event of islanding conditions. This is far behind the typically requested disconnection time imposed by regulations, i.e. 2 seconds.

The conventional electrical power network systems may have difficulties coping with increasing demand for power and need to reduce carbon dioxide emissions. Therefore, a new form of electrical power network systems that can handle these challenges in a sustainable, reliable and economic way is emerging. These networks may, for instance, utilize the same basic electric infrastructure as today, but will also draw on advanced monitoring, control and communications technology.

The result, for instance, can be a smart grid that is largely automated, applying greater intelligence to operate, monitor and even heal itself. The smart grid can be more flexible, more reliable and better able to serve the needs of a digital economy.

Availability of bidirectional communication in smart grid infrastructures combined with the power quality issues brought by active methods, especially when a large number of distributed generators are installed in the grid, may favour the communication based methods for islanding detection.

On the other hand, implementation of communication based methods may depend on co-operation with utility providers. The utility providers may be unwilling to adopt a centralized protection scheme as it could require significant investments from their part.

Additionally, in a smart grid, power may have more than one path to flow on, and the direction of the power may change. This may have a negative influence on the communication based methods listed above. For instance, if power lines are used as carriers for signals, multiple paths for power may cause signals to interfere and cause nuisance tripping of some of the distributed generators. The signals can be set with different frequencies but it may be necessary to ensure that all of the distributed generators located on the feeder listen to all of the used frequencies. Multiple paths for power may also have an effect on methods where a signal is produced on disconnect. In a conventional radial network, once a switching device opens, it may send out a trip signal to all generators in the network. However, when multiple flow paths exist, the islanding conditions may depend on the direction of power in the paths. The switch may have to send out the trip signal in either direction. This can cause undesired disconnection of generators from a properly working part of the network.

Figure 2 illustrates an example of a situation where a network of power electrical units is supplied by a main grid through two substations 1 and 2. The substations 1 and 2 are interconnected through switches 5 and 6.

If switch 3 of the first substation 1 is open and switch 4 of the second substation 4 is closed, the power is supplied to the power electrical units from the second substation 4 and the local generators in the network. When switch 5 opens, it should send a trip signal only to generators located in the direction of first substation 1 .

However, if the switch 4 of the first substation 1 and the switch 6 are closed and a switch 4 of the second substation 2 is open, switch 5 should send a trip signal only to generators located in the direction of the second substation 2 at disconnect. Thus, it may be very difficult for switch 5 to decide to which generators to send the trip signal.

In smart grids, a much higher sampling rate may be necessary in order for a SCADA approach to become efficient and compliant with the standards. Large sizes of distribution networks may require more space in the SCADA database and faster data processing. Moreover, due to typically large number of customers of Distribution System Operators, data sent to SCADA system is generally aggregated at secondary substation level. Thus, concrete information about a particular customer, for instance, a distributed generator, may be non-existent. The situation may change with the installation of smart meters all over the distribution grid. However, the installation of smart meters may challenge the communication bandwidth and speed and data storage if all data is stored at control centre level.

BRIEF DISCLOSURE

An object of the present disclosure is to alleviate the above disadvantages. The object of the invention is achieved by an automation and control system and by a method which are characterized by what is stated in the inde- pendent claims. The preferred embodiments of the invention are disclosed in the dependent claims.

The present disclosure proposes a new method for detecting islanding conditions in an electric power network supplied by a main grid. The elec- trical power network is divided into a plurality of sub-networks, wherein the sub-networks comprise at least one power electrical unit. The sub-networks are separable from each other and a main grid supplying the network by switching devices.

Islanding conditions within some or all of the sub-networks may be determined in a decentralised manner. Topological information of the subnetworks may first be determined. Islanding conditions in the sub-networks may then be determined on the basis of the topological information. Separate detecting means may be used for each sub-network. Finally, a signal resulting in disconnection or control mode change may be sent to the power electrical units of an island in the sub-network, on the basis of the islanding conditions determined by using the detecting means.

Power electrical units located in-between neighbouring switching devices may form anti-islanding groups (AIG). The power electrical units of an AIG may share disconnect or mode-change signals from the islanding condi- tion detecting means, which reduces the response time following detection of islanding conditions.

Decentralized approach in detecting islanding conditions can reduce response time since it reduces complexity in topology used in detecting islanding conditions. Detecting means assigned to a small part of the network can operate in almost real time, thus, meeting the requirements for disconnecting within two seconds imposed by typical standards.

The disclosed method fits well to a smart grid. A smart grid typically uses a communication infrastructure, e.g. for metering, feeder automation, etc, which can be utilised in detecting islanding conditions. The disclosed method can be implemented without significant investments in communications infrastructure. Thus, the disclosed methods can be cost competitive with the active methods without their weaknesses.

According to the invention, an automation and control system of a low or medium voltage electric power network comprises a communication system connecting a substation control device located at, or assigned to, a substation of the network to switching devices and to Distributed Generation (DG) units, e.g. photovoltaic panels, wind turbines, or charged batteries, of the network. The system detects islands in the power network, an island comprising loads and DG units that are mutually interconnected on a same voltage level of the power network and disconnected from a main grid supplying the power network via the substation. The substation control device is adapted to:

- receive topological information, i.e. a dynamic switch status of switching devices in downstream feeders and inter-substation connection lines, on behalf of a sub-network of the power network,

- identify, from the topological information, an island within the sub network,

- send signals to all the DG units of the identified island, independently of whether an individual DG is connected or not, the signals either disconnecting the DG unit, or changing a control mode of the DG units from power injection to voltage and frequency control in case the identified island is permitted.

The automation and control system is further adapted to receive the topological information from the switching devices, to identify the island, and to send the signals to the DG units such that the latter are received within a delay of less than two seconds following an initiating open/close action of a switching device.

In the automation and control system, DG units located in-between two or more neighbouring switching devices form Anti Islanding Groups (AIG) as the atomic building blocks of all conceivable islands. The substation control device is adapted to send a single signal to the AIG on behalf of all its constituting DG units.

BRIEF DESCRIPTION OF THE DRAWINGS

In the following the invention will be described in greater detail by means of preferred embodiments with reference to the attached drawings, in which

Figures 1 a and 1 b illustrate a difference between a grid-connected mode and an islanding mode;

Figure 2 illustrates an example of a situation where two substations are interconnected through switches; and

Figure 3 illustrates an exemplary arrangement comprising appara- tuses and for detecting islanding conditions in an electric power network. DETAILED DISCLOSURE

The present disclosure proposes a method for detecting islanding conditions in an electric power network, wherein the electrical power network comprises a plurality of sub-networks. The sub-networks are separable from each other and a main grid supplying the network by switching devices. Each sub-network comprises at least one power electrical unit. A power electrical unit may, for instance, be a distributed generation unit (DG) such as a photovoltaic panel, a wind turbine, or an energy storage device such as a battery. The sub-networks may, for instance, be a low voltage (LV) grid supplied by the main grid via a substation, with the main grid being a medium voltage (MV) grid and the substation comprising a MV/LV transformer. Alternatively, the subnetworks may be at a medium voltage level above 10 kV, with the connected DG units including suitable step-up transformers.

The method comprises determining topological information for at least one sub-network of interest. Determining the topological information may, for instance, be performed by determining an operational state of at least one switching device. Further, the topological information may disclose details of the structure of the electrical network and/or sub-network in question. This structure may, for instance, be represented by a truth table. The truth table may tell which combinations of switching device operational states produce islanding conditions within the particular sub-network.

When the topological information is known, islanding conditions in the at least one sub-network of interest is detected on the basis of the topological information by using separate detecting means for each sub-network of interest. In other words, islanding conditions of each sub-network of interest are calculated locally, independently from each other. Islanding conditions of some or all of the sub-networks may be determined in a decentralised manner. Because the method is localized to only a part of the network, it can operate practically in real time, thus meeting the requirements for disconnect imposed by standards. Thus, determining the islanding conditions can be performed within two seconds from a topology change, as regulated in typical standards.

The detecting means may, for instance, be implemented on network substation computers. The substation computers may, for instance, get the network topology from an upper level e.g. distribution control centre. The net- work substation computers may then monitor the status of the switching device along the feeder and the interconnecting lines with the neighbouring substa- tions and, on a disconnect signal from a switching device, perform logical operations for deciding to which distributed generators it shall send a disconnect signal in case of islanding.

The substation computers may have also other functions, for exam- pie, feeder reconfiguration, restoration, protection and control, optimization of energy use. Furthermore, in a smart grid, the disclosed method may be able to use the communication infrastructure deployed in the smart grid for, e.g. metering, feeder automation, etc. Thus, the cost of implementing the disclosed method can be reduced.

The substation computers are just one example of the detecting means. Any means capable of calculating the islanding conditions for a subnetwork on the basis of the topological information may be used.

Finally, a disconnect signal may be sent sending to at least one power electrical unit in at least one sub-network of interest on the basis of the islanding conditions determined by using the detecting means. On the basis of the signal, the power electrical unit disconnects from the network.

The power electrical units may form anti-islanding groups (AIG). In other words, at least one sub-network of interest comprises at least two power electrical units interconnected on the same LV level forming an anti-islanding group. The same disconnect signal is sent to all electrical units in the anti- islanding group when islanding conditions are detected.

These groups may, for instance, comprise all distributed generators located between two consecutive switching devices in case of radial feeders. In case of a meshed grid with multiple paths for power, an AIG could be de- fined between two or more switches with a direct connection, i.e. without any further subdividing switch. A goal of an anti-islanding group is to speed up the processing time in detecting islanding. For instance, once a part of the feeder is disconnected due to opening of switches, a group of distributed generators connected to that part of the network (modelled as one AIG) may receive a common disconnect signal from a substation computer.

Figure 3 illustrates an exemplary arrangement comprising apparatuses 10 and 1 1 for detecting islanding conditions in an electric power network 12. The network 12 may, for instance, be a low voltage (LV) network. In Figure 3 network 12 comprises two sub-networks 13 and 14. The sub-networks 13 and 14 are supplied by the main grid 15 via substations 16 and 17. The main grid may, for instance, be a medium voltage (MV) grid, and the substations 16 and 17 may comprise MV/LV transformers. The amount of sub-networks is not limited to two. It may comprise any plurality of sub-networks. A communication system with a communication network (broken line) enables exchange of messages between the apparatuses 10, 1 1 and devices of the sub-networks.

The sub-networks 13 and 14 are separable from each other and a main grid 15 supplying the network by switching devices 18, 19, and 20. The sub-networks 13 and 14 comprise at least one power electrical unit. A power electrical unit may, for instance, be a load or a distributed generator. In Figure 3, the power electrical units are mutually interconnected on the same LV level.

In Figure 3, the sub-networks 13 and 14 can be further divided into groups separable by switching devices. The first sub-network 13 is divided into three groups 21 , 22, and 23 separable by switching devices 24 and 25. The second sub-network 14 comprises group 26.

An apparatus according to the present disclosure comprises means for determining topological information of at least one sub-network of interest. In Figure 3, means for determining the topological information are implemented in substation computers 10 and 1 1 . The substation computers are part of the substations 16 and 17, respectively. The substation computers 10 and 1 1 may, for instance, receive information on the network structure from an upper level e.g. a distribution control centre 27. Further, the substation computers 10 and 1 1 receive information of the operational status of the switching devices of part of the electrical power network 12. The substation computers 10 and 1 1 may have also other functions relating to the network 12.

A required maximum response time may impose some limits on the size of the network as islanding conditions should typically be detected within two seconds of a topology change. Therefore, a decentralised approach allows reduction in response time. An apparatus according to the present disclosure comprises separate detecting means for each sub-network of interest to determine islanding conditions in the sub-networks of interest on the basis of the topological information. Detecting islanding conditions of each sub-network of interest is localized to a small part of the network. Thus, the detecting means can operate practically in real time meeting the requirements of the standards.

In Figure 3, the substation computers 10 and 1 1 operate as the detecting means and perform the logical operations for deciding to which power electronic units they shall send the disconnect signal in case of islanding. The topological information may, for instance, represented by a truth table. Table 1 illustrates an exemplary truth table for islanding conditions of power electrical units in group 23. In Table 1 , the islanding conditions are determined on the basis of operational status of switches 18, 19, 20, and 25. represents a closed switch and Ό' represents an open switch. The switch 24 does not have an effect on the islanding conditions of the group 23.

Table 1

In Table 1 , islanding conditions can be detected correctly regardless of the direction of power in the power lines of the network 12.

To react to detected islanding conditions, an apparatus according to the present disclosure further comprises means for sending, on the basis of the islanding conditions determined by using the detecting means, a disconnect signal to at least one power electrical unit in at least one sub-network of interest. In Figure 3, the substation computers 10 and 1 1 send disconnect signals to the power electrical units.

Power electrical units interconnected on the same LV level may also form anti-islanding groups (AIG). In case of radial feeders, these anti-islanding groups may, for instance, comprise all the distributed generators located be- tween two consecutive switching devices. In case of a grid with multiple paths for power, an AIG can be defined between the switches with direct connection without a possibility of interrupting between any combination of two switches.

Once islanding conditions are detected, anti-islanding group of electrical units receive a shared disconnect signal from the substation computer. In Figure 3, groups 21 , 22, 23 and 26 may, for instance, form anti-islanding groups. A saving in processing time may be achieved due to the fact that distributed generators are structured in groups. A shared disconnect signal can be sent to all electrical units in the anti-islanding group. The electrical units of the group may receive the signal regardless of whether an individual electrical unit is actually connected or not.