DEVICE FOR DETERMINING A FAILURE PROBABILITY VALUE FOR A TRANSFORMER COMPONENT AND A SYSTEM WITH SUCH A DEVICE

Description

[0001] Transformers are generally known from the prior art. These are transformers that can convert an electrical input voltage into an electrical output voltage. Usually, the input voltage is an alternating current (AC) voltage and the output voltage is an alternating current (AC) voltage. In a very simple design, the transformer has a primary coil and a secondary coil coupled to each other via a common core. The coils and / or the core can be at least partially submerged in a tank containing a cooling fluid to cool the coils. The primary coil, the secondary coil, the core, the cooling fluid, and the tank can each be a transformer component. The transformer can also include other transformer components.

[0002] In practice, transformers are often monitored using a measuring system. This system comprises several sensors designed to detect the physical and / or chemical properties of the transformer. A measuring system is frequently employed when the transformer is a so-called power transformer. This is the case, for example, when the electrical power transferred from the primary to the secondary winding exceeds 10 kilowatts. During operation, the two windings of a power transformer generate a considerable amount of heat, which is at least partially absorbed by the transformer's coolant. The coolant may be circulated within a cooling circuit of the power transformer, which may include a heat sink that dissipates the heat transferred to the coolant into the surrounding environment and / or another medium.The measuring system can therefore, for example, include a sensor designed to detect the temperature of the power transformer's coolant. A fan can be mounted on the cooler to achieve the desired cooling capacity. The electrical current flowing through the fan's motor can be detected, for example, by another sensor in the measuring system. An oil, which can also be described as insulating oil, is often used as the coolant. The oil's insulating properties prevent electrical short circuits within or between the coils. The measuring system can include another sensor designed to directly or indirectly detect the water content in the oil. For example, the sensor can be in direct contact with the oil to directly measure the water content.However, it is also possible that a sensor is located in the tank of the power transformer, this sensor being designed to detect a gas concentration containing hydrocarbon chains. A gas concentration containing hydrocarbon chains detected by the sensor that exceeds a predetermined threshold can indicate a certain minimum water content in the oil. Another sensor in the measuring system can, for example, be designed to detect a mechanical winding tension force between the turns of one of the coils. This sensor can, for example, be designed to detect mechanical pressure or mechanical force, being arranged on the primary or secondary coil of the power transformer in such a way as to detect the mechanical tension force of the respective coil.Another sensor of the measuring system can be arranged and / or configured to detect the current through the primary or secondary coil. Furthermore, the measuring system can include a voltage sensor configured to detect the electrical voltage applied to the primary or secondary coil. Another sensor of the measuring system can, for example, be configured to detect the condition of the transformer bushings, whereby the sensor is configured, for example, to determine the capacitance and / or loss factor of the transformer bushings. Another sensor of the measuring system can, for example, be configured to detect vibroacoustic signals. This sensor can, for example, be configured to detect acoustic signals generated when a tap changer of the power transformer is operated.The measuring system can be configured to generate a status signal representing several different status variables. The measuring system can be configured to determine at least one of the status variables based on the sensor signal from exactly one of the sensors, and / or one of the status variables based on multiple sensor signals from multiple sensors. Each sensor is arranged on the power transformer to detect a physical and / or chemical property of the power transformer. The physical and / or chemical properties can thus be corresponding physical and / or chemical properties of components of the power transformer. The components of the power transformer are also referred to as transformer components. For example, the transformer's coolant can be a component of the power transformer. However, the voltage or current at the first or second terminal can also be a component.The second coil of the power transformer can each be a physical property of the power transformer. Each status variable determined by the measuring system represents a status state of the property of the power transformer detected by the at least one associated sensor, either as a normal state (N) or as an abnormal state. For example, if a status variable is determined based on the sensor signal of the sensor for detecting the temperature of the coolant, a normal state can be determined by the measuring system for the transformer component "coolant," particularly if the temperature of the coolant detected by the sensor is lower than a reference value. Otherwise, the measuring system can determine an abnormal state for the status variable.A functional relationship therefore exists between each status variable and at least one of the transformer components whose properties are directly or indirectly detected by means of the at least one associated sensor. It is therefore also said that each status variable is coupled to at least one of the transformer components via a functional relationship.

[0003] The status variables determined by the measurement system can represent the status states of the power transformer, either as a normal or anomalous state. If a status variable relates to a single transformer component, its state can directly indicate whether that component is in a normal or anomalous state. However, if a status variable is determined based on multiple sensor signals from several sensors, there is often no direct, unambiguous relationship between the status state represented by that variable and a single transformer component. Rather, a functional relationship can exist between the status state and multiple transformer components. Furthermore, a single transformer component can influence multiple sensor signals.A transformer component in an abnormal state can cause the measuring system to determine several status variables in such a way that each status variable represents a status state as an abnormal state.

[0004] When several status variables represent an abnormal state, it is not always trivial and easy to determine in practice whether and which of the transformer components are in an abnormal state, or whether only a single transformer component in an abnormal state is the cause of several status variables representing an abnormal state.

[0005] In this context, CN 111 272 222 A describes a method for diagnosing transformer faults based on a set of characteristic values.

[0006] CN 104 007 343 B shows a method for dynamic fault diagnosis for a transformer based on a Bayesian network.

[0007] CN 102 779 230 B discloses a method for condition analysis and for creating maintenance recommendations for power transformers using a Markov model.

[0008] JP 2007 328522 A describes a method for determining the probability of failure of a device based on fault logs obtained after the device has been put into operation.

[0009] US 2014 / 222355 A1 relates to the prediction and diagnosis of conditions of energy transmission and distribution equipment, such as transformers, and in particular to a statistical approach to the diagnosis, detection and identification of equipment faults, as well as the overall health of equipment.

[0010] The invention is therefore based on the objective of proposing a device and a system that allows a particularly simple evaluation of whether and which transformer component(s) of a power transformer is / are defective. According to a first aspect of the invention, this objective is achieved by a device with the features of claim 1. Thus, a device for determining a probability of failure value for a transformer component is provided. The device comprises a signal input interface, a processor unit, and a storage unit.The signal input interface is configured for direct or indirect coupling with a measurement system for a power transformer, wherein the measurement system has several sensors, each configured to detect a physical and / or chemical property of the power transformer, wherein the power transformer has several transformer components, and wherein each sensor is coupled to at least one of the transformer components of the power transformer via a corresponding direct or indirect connection, such that the property of the power transformer detected by the respective sensor is influenced by at least one of the transformer components.Furthermore, the signal input interface is designed to receive a status signal from the measuring system, wherein the status signal represents several different status variables, with each status variable being assigned at least one of the sensors, so that each status variable is in a functional relationship with at least one of the transformer components, with each status variable representing a status state of the property of the power transformer detected by the at least one assigned sensor as a normal state or as an abnormal state.Furthermore, a data record is stored by the storage unit, wherein the data record has for each status variable an associated first probability value that the respective status variable represents an incorrect state, wherein the data record has for each transformer component an associated second probability value that a fault has occurred in the respective transformer component, and wherein the data record has for each of the functional relationships from one of the transformer components to one of the status variables an associated third probability value that a fault in the respective transformer component causes an abnormal state of the respective status variable.The processor unit is configured to determine for each transformer component an associated fault probability value based on a Bayesian network representing a directed graph from the transformer components to the status variables based on the first, second, and third probability values, as well as based on the status states represented by the status variables, that the respective transformer component is in a fault state.

[0011] With regard to the power transformer and the measuring system, reference is preferably made to the explanations mentioned at the outset in an analogous manner. The measuring system and the power transformer are preferably not parts of the device. Rather, it is preferably provided that the signal input interface is configured to receive the status signal of the measuring system. The status signal represents several different status variables, each representing a status state of the property of the power transformer detected by the at least one associated sensor of the measuring system, either as a normal state or as an abnormal state. Preferably, the status state is either the normal state or the abnormal state.The device uses the received status signal to provide several different status values, each representing an abnormal or normal state for a corresponding property of the power transformer. However, determining which transformer components might actually be faulty requires considerable experience to interpret these status values. To obtain a repeatable and quantifiable indication of a potential transformer component failure, the device incorporates a processor unit and a memory unit.

[0012] The storage unit stores a data record that includes an associated first probability value for each status variable. This first probability value indicates a probability, preferably independent of the actual state represented by the status variable, that the state represented by the respective status variable—i.e., normal state or abnormal state—is an incorrect state. For example, if a status variable represents a normal state, there is a certain probability, corresponding to the first probability value, that the state represented by the status variable is not actually a normal state, but rather an abnormal state.For example, if the first probability value is 0.05, there is a five percent probability that the status state represented by the respective status variable is false.

[0013] The data set stored by the memory unit also includes a corresponding second probability value for each transformer component of the power transformer. This second probability value indicates the likelihood that a fault might have occurred in the respective transformer component. Statistical surveys of transformers and / or practical experience can reveal, for example, that some transformer components fail more frequently than others within a power transformer. From these statistical surveys and / or experience, the statistical probabilities for failure of the transformer components can be derived. Based on these statistical probabilities of the transformer components, the second probability values ​​can therefore be predetermined.For example, a corresponding second probability value can be predetermined for the primary winding of a power transformer, indicating the probability that the primary winding of the power transformer has a fault. Similarly, a corresponding second probability value can be predetermined for a predetermined number of transformer components of a power transformer.

[0014] The data set stored by the storage unit also includes, for each of the functional relationships between a transformer component and a status variable, a corresponding third probability value. This third probability value therefore indicates the probability that a fault in the respective transformer component will cause an abnormal state in the status variable to which it is functionally related. If a status variable is assigned to exactly one sensor that can detect a physical and / or chemical property of exactly one transformer component, then the third probability can represent whether a fault in this transformer component will lead to an abnormal state in the status variable determined by the sensor.However, it is also possible that a transformer component has a direct or indirect relationship to multiple sensors, meaning that a fault in the respective transformer component can affect several status variables simultaneously. Whether this occurs statistically for the multiple sensors and corresponding status variables is represented by the associated third probability values.

[0015] From the preceding explanations, it can be inferred, at least indirectly, that the sensors, and therefore the status variables, do not always allow for a direct conclusion regarding the state of exactly one transformer component that might be faulty. Rather, it is usually the case that if one or more of the status variables represent an abnormal state, then several transformer components could be potentially faulty. Functionally, therefore, there is a directed relationship from the transformer components to the sensors and thus also to the status variables, but no one-to-one relationship from the status variables to the transformer components.To nevertheless be able to determine which and whether any of the transformer components are faulty, the device's processor unit is configured to calculate a corresponding fault probability value for each transformer component based on a Bayesian network and on the status states represented by the status variables, indicating whether the respective transformer component is in a fault state. Each fault probability value preferably indicates the probability (fault probability) that the respective transformer component is in a fault state. The fault state can be understood as an abnormal state of the transformer component. For example, if a fault probability value of 0.001 is determined for a transformer component, there is a 0.1% probability that the respective transformer component is faulty.This is relatively small and suggests that the respective transformer component is not faulty. However, if a corresponding fault probability value of 0.7 is determined for another transformer component using the processor unit, this indicates that the respective transformer component is faulty with a probability of 70%. The determination of the fault probability values ​​is based on the Bayesian network and the status variables provided to the device by the measuring system via the status signal. Bayesian networks are generally known from the prior art. Therefore, no further or more detailed explanation is necessary here.In the Bayesian network, the transformer components form a first group of nodes and the status variable a second group of nodes, with exclusively directed, functional relationships existing from each transformer component to at least one of the status variables.

[0016] The processor unit can be configured to calculate error probability values ​​by evaluating the Bayesian network and considering the actual state of the system, for example, using one of the following algorithms: the sampling algorithm or the variable elimination algorithm. Both algorithms are known from the prior art. Other algorithms can also be used. The processor unit can be configured to execute either of these two algorithms or other algorithms. However, it may also be sufficient for the processor unit to be configured to execute only one of the two aforementioned algorithms to determine the error probability values.

[0017] The Bayesian network used to determine the error probability values ​​is represented by a directed graph from the transformer components to the state variables. The dataset includes a first probability value for each transformer component, a second probability value for each state variable, and a third probability value for each directed relationship between a transformer component and a state variable. By representing the graph as a directed graph from the transformer components to the state variables, the Bayesian network accurately reflects the actual statistical relationships between the power transformer and the measurement system.Based on the Bayesian network, however, it is possible, taking into account the actual status variables, to determine which fault probability values ​​and associated fault probabilities exist for the transformer components, indicating that the respective transformer component is in a fault state or in an abnormal state.

[0018] The device therefore offers the advantage that, despite the directed, functional relationship between the transformer components and the status variables, quantitative error probability values ​​for the transformer components can be derived, which allow a statement about which of the transformer components, for example, has the highest probability of a failure.

[0019] An advantageous embodiment of the device is characterized in that the processor unit is configured to generate a multitude of patterns based on the Bayesian network, which represents a directed graph from the transformer components to the status variables based on the first, second, and third probability values. Each pattern represents a corresponding pattern state as either a normal or abnormal state for each status variable and for each transformer component. The processor unit is configured to determine a reference group of patterns from the multitude of patterns such that, for each pattern in the reference group, the corresponding pattern states for the status variables correspond to the status states of the status variables in the status signal.The processor unit is also configured to determine for each transformer component an associated probability of failure based on the pattern states for the transformer components from the patterns of the reference group, indicating that the respective transformer component is in an abnormal state.

[0020] The aforementioned design of the device is based on the fundamental principle that a large number of patterns are first generated, and then, based on the actual status states of the status variables, it is determined which of the previously generated patterns are assigned to a reference group. The patterns in the reference group thus exhibit pattern states for the status variables that correspond to the actual status states represented by the status signals that the device receives from the measuring system. In other words, the processor unit determines, based on the actual status states, which of the patterns exhibit pattern states for the corresponding status variables that correspond to the actual status states. The patterns determined in this way constitute the patterns of the reference group.However, when determining the patterns for the reference group, the pattern states for the transformer components are not considered. Rather, these are used in a further step to determine the failure probability values ​​of the transformer components.

[0021] As previously explained, the status states of transformer components generally do not allow for direct inferences about their states. However, each pattern from the reference group also has a corresponding pattern state for each transformer component. Due to the large number of patterns in the reference group, this results in a corresponding number of pattern states for each transformer component. The distribution of these patterns allows for the derivation of a probability that the respective transformer component is in an abnormal state. Therefore, based on the patterns in the reference group, the processor unit can determine a fault probability value for each transformer component, representing the probability that the respective transformer component is in an abnormal state.

[0022] The processor unit is designed to generate the multitude of patterns, with each pattern representing a corresponding pattern state as either a normal or abnormal state for each status variable and each transformer component. The data set stored by the memory unit includes a corresponding first probability value for each status variable. Generating the multitude of patterns, for example, creates a corresponding multitude of pattern states for each status variable. However, this generation of pattern states is not arbitrary, but rather takes into account the first error probability value, so that the pattern states for the status variables are generated with a probability corresponding to the first probability value as an abnormal state.If the first probability value is, for example, 0.2, then the multitude of pattern states for the corresponding status variable are generated such that 20% of the pattern states represent an anomalous state, and the remaining 80% represent a normal state. The generation of the multitude of pattern states for the other status variables is analogous. The same applies to the pattern states for the transformer components. For each transformer component, a corresponding second probability value is represented by the data set. Each pattern represents a corresponding pattern state for each transformer component, either as a normal state or as an anomalous state.This generation of pattern states for the transformer components is not arbitrary, but rather takes into account the second error probability value. This means that the pattern states for the transformer components are generated as abnormal states with a probability corresponding to this second probability value. For example, if the second probability value is 0.15, the processor unit generates the multitude of patterns such that 15% of the pattern states for the respective transformer component represent an abnormal state, and the remaining 85% represent a normal state. Furthermore, it should be noted that the multitude of patterns are generated randomly, taking the error probability values ​​into account. The processor unit is designed accordingly.The greater the number of patterns, the more accurately a corresponding error probability value can be determined for each transformer component, indicating the probability that the respective transformer component is in an abnormal state.

[0023] Another advantageous embodiment of the device is characterized in that the device has an output interface, wherein the device is configured to generate an output signal that represents at least the largest of the error probability values ​​and the associated transformer component, and wherein the device is configured to transmit the output signal via the signal output interface.

[0024] As explained in the preceding description, a corresponding fault probability value is determined for each transformer component. If multiple transformer components are present, a corresponding number of fault probability values ​​are also determined. These fault probability values ​​can differ. However, the highest fault probability value is the most relevant, as it indicates which of the transformer components is most likely to be in an abnormal state. The processor unit is therefore preferably configured to determine the highest fault probability value and the corresponding transformer component from the majority of fault probability values. Furthermore, the device, and in particular the associated processor unit, is configured to generate an output signal that represents the highest fault probability value and indicates the corresponding transformer component.For example, the output signal could represent a transformer component with the digits "01" and the corresponding error probability value as "0.7". However, it is also possible for the output signal to represent the respective transformer component in plain text. For example, the output signal could represent the transformer component "coolant" and the corresponding error probability value as "0.7". The output signal can be transmitted via the device's signal output interface. The device can be designed accordingly for this purpose. The output signal can, for example, be transmitted to a display unit to visually represent the error probability value associated with the transformer component.

[0025] Another advantageous embodiment of the device is characterized by the fact that the device is configured to generate the output signal such that the output signal represents at least the two or three highest failure probability values ​​and the associated transformer components. In practice, the transformer component with the highest failure probability value is often of most interest. However, the corresponding failure probability value still represents a probability, so the associated transformer component is not necessarily faulty. Therefore, in practice, transformer components for which high failure probability values ​​have also been determined can also be of interest. The processor unit of the device can therefore be configured to determine the two or three highest failure probability values.Furthermore, the device, and in particular the associated processor unit, can be configured to generate the output signal such that the output signal represents the previously determined two or three highest probability values ​​of failure and at least indicates the corresponding transformer components. The preceding explanations regarding the highest probability value of failure and the output signal apply analogously to the two or three highest probability values ​​of failure and the corresponding transformer components. However, with the output signal representing the two or three highest probability values ​​of failure and the corresponding transformer components, the person skilled in the art also obtains information about which other transformer components may also be faulty, particularly if the transformer component for which the highest probability value of failure was determined is in fact not faulty.The expert can therefore continue his investigation and, for example, also examine the transformer component for which the second highest probability of failure was determined.

[0026] Another advantageous embodiment of the device is characterized by the fact that the status variables represented by the status signal are ternary status variables, each representing exactly one status state from the following group of possible status states: normal state, abnormal state, or non-installation state. This prevents the status state from representing anything other than the normal state, the abnormal state, or the non-installation state. Alternatively, the group of possible states can be limited to two states: normal state and abnormal state. In this case, the status state can represent either the normal state or the abnormal state, but no other state.

[0027] A further advantageous embodiment of the device is characterized in that the processor unit is configured to generate a Bayesian network based on the data set. The Bayesian network is preferably a directed acyclic graph in which probability values ​​are assigned to each node of the graph, and the dependencies on the nodes are also represented by probability values. In the present case, the Bayesian network represents a first group of nodes consisting of status variables, each of which is assigned a corresponding first probability value. The second group of nodes of the Bayesian network is represented by the transformer components of the power transformer, with each of these nodes being assigned a corresponding second probability value. The dependencies of the status variables on the transformer components are represented by corresponding third probability values.The first, second, and third probability values ​​are present in the data set, which is stored in the device's memory unit. Therefore, the processor unit can, particularly based on the preceding explanations, generate the Bayesian network from the probability values ​​in the data set.

[0028] Another embodiment of the device is characterized by the Bayesian network being configured as a bipartite graph. This graph represents a first group of nodes, a second group of nodes, and functional relationships between the first and second groups of nodes. The first group of nodes is defined by the first probability values, the second group by the second probability values, and the functional relationships by the third probability values. Thus, in this embodiment, the Bayesian network is used as a special type of Bayesian network in the form of a bipartite graph. This ensures that only functional relationships exist between the transformer components and the status variables, and that there is no inverse, dependent relationship.In other words, the bipartite graph is designed such that the state variables depend only on the properties of the transformer components, but the properties of the transformer components do not depend on the states of the state variables. Relationships other than those mentioned above are preferably excluded. This allows all relationships between the transformer components and the state variables to be represented as a two-dimensional table, the entries of which are preferably determined by expert knowledge and are easily configurable.

[0029] Another advantageous embodiment of the device is characterized in that the processor unit is configured to generate the patterns using a likelihood weighting algorithm. In this process, the processor unit preferably considers the first, second, and third probability values ​​of the data set.

[0030] A further advantageous embodiment of the device is characterized in that the processor unit is configured to generate the output signal such that the output signal represents the associated error probability values ​​and transformer components as a first image in which the error probability values ​​are graphically represented and the transformer components are indexed and / or labeled. Such an image can, for example, represent the error probability values ​​as a bar graph, with the respective transformer component being indexed and / or labeled for each bar graph. The respective transformer component can, for example, be indexed by a number or labeled as text, such as "coolant." The probability values ​​can be represented quantitatively and / or qualitatively in the image.A qualitative representation of a probability value can be, for example, represented by the color and / or the length of a bar. A quantitative representation of a probability value can be, for example, a numerical graphical representation of the probability value. The device can use the output signal, for example, to display a corresponding image via a display unit of the device. However, it can also be provided that the output signal is transmitted via the output interface. For this purpose, the processor unit and, in particular, the associated output interface can be designed and / or configured.

[0031] According to a second aspect of the invention, the aforementioned problem is solved by a system with the features of claim 10. The system comprises a device and a display unit. The device is preferably configured according to the first aspect of the invention and / or one of the associated advantageous embodiments. The display unit has a screen. The signal interface of the device is coupled to the display unit via a direct or indirect first signal connection to transmit the output signal to the display unit. The display unit is also configured to display a second image on the screen, based on the output signal, which indicates the transformer components and associated error probability values. The aforementioned second image does not require the prior existence of a first image.Furthermore, reference is made in an analogous manner to the advantageous embodiments, preferred features, effects and / or advantages as explained in connection with the device according to the first aspect and / or one of the advantageous, associated embodiments for the system according to the second aspect of the invention.

[0032] The system comprises the device and the display unit with the screen. The screen can display a second image, which graphically represents the at least one error probability value represented by the output signal and the associated transformer components. This representation can include a qualitative and / or quantitative representation of the at least one error probability value and a quantitative and / or qualitative representation of the at least one associated transformer component. For example, if the output signal represents the highest error probability value and the associated transformer component, the second image can graphically represent the highest error probability value and the associated transformer component accordingly.The display unit can be physically separate from the device, so that the first signal connection also extends between the device and the display unit. This first signal connection can be a wired or wireless signal connection. However, it is also possible for the display unit to form an integral part of the device, so that the system consists of the device with the integrated display unit. The system may also include other components.

[0033] A further advantageous embodiment of the system is characterized by a display unit configured such that the second image shows the transformer components represented by the output signal in list form and qualitatively displays the associated fault probability values ​​using different colors and / or bars. The fault probability values ​​can also be displayed quantitatively by the display unit. Furthermore, it is preferably provided that the display unit is configured to display the fault probability values ​​in list form, in order of their magnitude. This allows the viewer to quickly gain an overview of which of the transformer components might be in an abnormal state and which of these possible transformer components have the highest probability of being in the corresponding abnormal state.

[0034] A further advantageous embodiment of the system is characterized in that the system includes a measuring system for the power transformer, wherein the measuring system is coupled to the signal input interface of the device in order to transmit the status signal to the signal input interface, and wherein the measuring system comprises several sensors, each configured to detect the physical and / or chemical properties of the power transformer. The power transformer comprises several transformer components. Each sensor is coupled to at least one of the transformer components of the power transformer via a corresponding direct or indirect connection, such that the property of the power transformer detected by the respective sensor is influenced by at least one of the transformer components.The explanations given above apply analogously to the measuring system, the sensors, and the power transformer. The measuring system can include multiple sensors, each designed to detect a physical and / or chemical property of the power transformer. It is not always possible to directly detect the property of a specific transformer component. Often, a sensor detects a property of the power transformer that is only indirectly dependent on the state of a transformer component. The measuring system can be located in close proximity to the power transformer, with its sensors attached directly to the transformer to detect the relevant physical and / or chemical properties.Each sensor can be coupled to an evaluation unit of the measuring system, such that the measuring system is configured via the evaluation unit to generate the status signal. The evaluation unit of the measuring system can also be configured to transmit the status signal to the signal input interface of the device. Generally, it is preferred that the measuring system and the device are separate. However, it is also possible for the measuring system and the device to be at least partially or fully integrated. For example, the device can form part of the measuring system, or vice versa.

[0035] Another advantageous embodiment of the system is characterized in that the measuring system is designed to sensorially detect at least two of the following properties of the power transformer: gas from hydrocarbon chains in the liquid insulating medium, water in the liquid insulating medium, fill level of the liquid insulating medium, electrical voltage, electrical current, winding tension force, rotational speed of a fan, electrical power of the fan motor, noise level of the fan, motor torque of a tap changer, switching sequence of the tap changer, temperature (e.g. hot spot temperature or top oil temperature), vibroacoustic intensity, and functionality of the dehumidifier, loss factor of one or more transformer bushings, capacitance of one or more transformer bushings, vibroacoustic intensity of the tap changer.The power transformer preferably includes a tank filled with coolant. The coolant is also referred to as an insulating fluid or insulating medium, as it is preferably electrically insulating. The coolant is preferably an oil, particularly a mineral oil. Often, the coolant is an ester derived from rapeseed oil. Thus, the coolant, which can also be called an insulating fluid, is, for example, an oil, preferably ester-based, and most preferably a liquid insulating medium such as rapeseed oil. The coolant preferably has two main functions: to cool the primary and secondary coils, and to act as an electrical insulator. The transformer core and the primary and secondary coils are also preferably located in the aforementioned tank. Therefore, the greatest amount of heat from the transformer is usually dissipated in this tank.The heat generated at the coils is dissipated by the coolant. The transformer can therefore have a piping system and a cooler, designed and arranged to circulate coolant from the tank through the piping system to the cooler and back to the tank, allowing the heat to be dissipated via the cooler. The power transformer is preferably designed to connect different electrical networks with different voltages. For example, the power transformer can electrically couple a first network with a lower voltage to a second network with a higher voltage, or vice versa, to transfer electrical energy from one network to the other. The electrical networks are preferably three-phase. Thus, each network can have its corresponding current and voltage on each of the three phases.The corresponding currents and / or voltages can be detected by sensors. For this purpose, the measuring system can include sensors that are appropriately designed and / or arranged. Theoretically, all currents and voltages of both networks can be detected by sensors. In practice, however, usually only a portion of the currents and voltages in the two networks are detected. Therefore, the measuring system can be designed to detect a portion of the electrical voltages and currents in the two networks to which the electrical power transformer can be coupled.

[0036] Another advantageous embodiment of the system is characterized by the display unit forming part of the measuring system. The measuring system is often located in close proximity to the power transformer. The device can be physically separate from the measuring system and the power transformer. In particular, the device can be formed by a computer network, especially a cloud network, which includes a data storage unit. This allows the device to incorporate a processor unit with particularly high computing power. Displaying the second image on the display device is especially advantageous when the display device is part of the measuring system and thus in close proximity to the power transformer. This is because this information is frequently required during the maintenance and / or repair of a power transformer.

[0037] Another advantageous embodiment of the system is characterized by the inclusion of a remote monitoring system, wherein the signal output interface of the device is coupled to the remote monitoring system via a direct or indirect, secondary signal connection in order to transmit the output signal to the remote monitoring system. The remote monitoring system is located separately from the device, the measuring system, and the power transformer. The remote monitoring system can, for example, be a control center where a large number of power transformers are monitored.

[0038] Another advantageous feature of the system is that the remote monitoring system is cloud-based. The remote monitoring system can therefore be comprised of a cloud network with a large number of computer units.

[0039] Another advantageous embodiment of the system is characterized by the fact that the system's display device forms part of the remote monitoring system. The display device can therefore be arranged separately from the device, the measuring system, and the power transformer.

[0040] Another advantageous embodiment of the system is characterized by the fact that the system includes the power transformer. The system can thus comprise the power transformer, the measuring system, and the device. Furthermore, it can be provided that the system also includes the remote monitoring system.

[0041] Further features, advantages, and applications of the present invention will become apparent from the following description of the exemplary embodiments and the figures. All features described and / or illustrated, individually and in any combination, constitute the subject matter of the invention, irrespective of their composition in the individual claims or their cross-references. In the figures, the same reference numerals denote identical or similar objects. Figure 1 shows an advantageous embodiment of the device and the system, each in a schematic representation. Figure 2 shows an advantageous embodiment of the Bayesian network in a schematic representation. Figure 3 shows an exemplary tabular representation of patterns. Figure 4 shows an exemplary tabular representation of patterns of a reference group. Figure 5 shows an exemplary representation of an image that can be displayed on a display unit.

[0042] In the Figure 1An advantageous embodiment of the device 2 is shown schematically. The device 2 comprises a processor unit 6, a signal input interface 4, and a memory unit 8. The processor unit 6 is connected to the memory unit 8 in such a way that the processor unit 6 can access data stored in the memory unit 8. Thus, the processor unit 6 can, for example, access a data record stored in the memory unit 8. Furthermore, the processor unit 6 is connected to the signal input interface 4 so that a signal received by the signal input interface 4 is made available to the processor unit 6 directly or indirectly. The signal input interface 4 can, for example, perform data preprocessing to make the data represented by the received signal available to the processor unit 6.It has proven advantageous for the device 2 to have a signal output interface 36. A signal generated by the device 2, and in particular by the processor unit 6, such as an output signal Q, can be transmitted via the signal output interface 36. The processor unit 6 can be directly or indirectly coupled to the signal output interface 36 to control it directly or indirectly, so that the output signal Q can be transmitted via the signal output interface 36.

[0043] In the Figure 1An advantageous embodiment of the system 28 is also shown, comprising the device 2 and a display unit 30. The system 28 is shown schematically by a dashed line to indicate that the display unit 30 does not necessarily have to be part of the device 2. The device 2 can have a dedicated display unit 34. The device 2 and the display unit 30 can be coupled via a signal connection running between the device 2 and the display unit 30, which is also referred to as the first signal connection. The first signal connection can extend from the signal output interface 36 of the device 2 to the display unit 30. The display unit 30 has a screen.

[0044] Where advantageous embodiments, preferred features, technical effects and / or advantages of device 2 are explained below, corresponding advantageous embodiments, preferred features, technical effects and / or advantages may apply analogously to system 28. Therefore, repetition is omitted.

[0045] In the Figure 1 In addition to the device 2, a measuring system 10 and a power transformer 12 are also shown schematically. The power transformer 12 has terminals 44 on the primary side and further terminals 46 on the secondary side. The power transformer 12 serves to transform a voltage from the primary side to the secondary side. Different electrical networks can be connected to the primary and secondary sides, so that electrical energy can be transferred from the primary side to the secondary side, or vice versa.

[0046] The power transformer 12 has a variety of transformer components 16. For example, the power transformer 12 can have the following transformer components 16: terminals 44, 46, a primary coil, a secondary coil, a core, a coolant tank, the coolant itself, a cooling system, a radiator, a fan for cooling the radiator, a tap changer, a motor for driving the tap changer, bushings, valves, solid insulation for winding insulation, and / or other technical components.

[0047] The measuring system 10 comprises several sensors 14. Each sensor 14 is arranged on the power transformer 12, so that each sensor 14 can detect a physical and / or chemical property of the power transformer 12. For example, one sensor 14 can be arranged to detect the temperature of the coolant in the power transformer 12. Another sensor 14 can, for example, be designed and / or arranged to detect a mechanical tension force between the windings of the primary and secondary coils. This tension force is also referred to as the winding tension force. Thus, the respective sensor 14 can be designed to detect the winding tension force. A further sensor 14 can be arranged, for example, to detect the electrical voltage between the terminals 44 on the primary side.Another sensor 14 of the measuring system 10 can, for example, be configured to detect the condition of the transformer bushings, wherein the sensor is configured, for example, to determine the capacitance and / or loss factor of the transformer bushings. Another sensor 14 of the measuring system 10 can, for example, be configured to detect vibroacoustic signals. This sensor 14 can, for example, be configured as a sensor for detecting acoustic signals that arise when a tap changer of the power transformer 12 is operated. The in . Figure 1 The sensors shown are only examples 14. In fact, many more sensors 14 and / or other sensors 14 may be provided for the measuring system 10.

[0048] The sensors 14 can be designed to detect forces, mechanical stresses, electrical currents, electrical voltages, temperatures, pressures, mass flows and / or acoustic sound.

[0049] Each of the transformer components 16 can exhibit different properties, in particular different physical and / or chemical properties. In practice, it has been found that not all properties of a transformer component 16 can necessarily be detected by a sensor 14. For example, the temperature in a primary or secondary coil can often only be detected indirectly via the temperature of the coolant. There is therefore a fundamental functional relationship between the temperature of the primary or secondary coil and the temperature of the coolant detected by a sensor 14. However, the detected temperature of the coolant does not directly indicate whether, and if so, what temperature, one of the two coils has.In particular, if the coolant temperature is too high, it cannot be clearly determined which of the two coils is, for example, in a faulty condition, so that the faulty coil releases too much heat to the coolant and causes the aforementioned excessive temperature of the coolant.

[0050] The example above illustrates that the physical and / or chemical properties of the power transformer 12 detected by the sensors 14 can indeed indicate that one of the transformer components 16 is in a non-normal, i.e., an abnormal, state. This non-normal or abnormal state is also referred to as the abnormal state of the respective transformer component 16. Therefore, a device 2 is required that is designed to determine, based on the multitude of physical and / or chemical properties detected by the sensors 14, a specific assessment of whether and which transformer component(s) 16 are defective when conspicuous properties of the power transformer 12 are detected by the sensors.

[0051] The problem explained above is solved with device 2, as shown in an exemplary and schematic way in the Figure 1The device 2 has the aforementioned signal input interface 4, which is configured for direct or indirect coupling with a measuring system 10. The measuring system 10 has several sensors 14, each configured to detect a physical and / or chemical property of the power transformer 12. The power transformer 12 has several transformer components 16. Each of the sensors 14 of the measuring system 10 is coupled to at least one of the transformer components 16 of the power transformer 12 via a respective associated direct or indirect connection. This is preferably a functional connection.Through the direct or indirect connection between the sensors 14 and the transformer components 16, the technical effect can be achieved that the property of the power transformer 12 detected by the respective sensor 14 is influenced by at least one of the transformer components 16.

[0052] The measuring system 10 can have an input interface 38, an output interface 40, and an evaluation unit 42. The sensors 14 are preferably coupled to the input interface 38 so that the properties detected by the sensors 14 are transmitted to the evaluation unit 42, preferably via a sensor signal. The evaluation unit 42 can be configured to evaluate the properties detected by the sensors 14 and generate a status signal S that represents several different status variables. Each status variable can, for example, indicate whether one of the sensors 14 has detected a chemical and / or physical property of the power transformer 12 that is greater than or less than a corresponding limit value.The status signal S represents different status variables, with each status variable being assigned to at least one of the sensors 14, such that each status variable is coupled to at least one of the transformer components 16 via a functional relationship. This is not a mechanical coupling, but a functional coupling in the sense of a functional relationship. Each status variable represents a status state of the property of the power transformer 12 detected by the at least one assigned sensor 14, either as a normal state N or as an abnormal state A. The status state can therefore be either the normal state N or the abnormal state A, with the respective status state being represented by the status variable.If, for example, a property of the transformer is the temperature of the coolant, then the status state of the coolant temperature can be in the normal state when the temperature is below a predetermined maximum temperature, and in an abnormal state when the temperature reaches the maximum temperature or is even higher.

[0053] The status signal S represents several of the different status variables. Furthermore, the signal input interface 4 of the device 2 is configured to receive the status signal S from the measuring system 10. After the measuring system 10 has detected the properties of the power transformer 12 using the sensors 14, the measuring system 10 can transmit the status signal S from a corresponding output interface 40 of the measuring system 10 to the signal input interface 4 of the device 2.

[0054] The processor unit 6 of the device 2 is preferably coupled to the signal input interface 4, so that the status variables represented by the status signal S are made available to the processor unit 6. The processor unit 6 of the device 2 is configured to determine an associated fault probability value for each transformer component 16 based on a Bayesian network 18 and based on the status states represented by the status variables of the status signal, wherein the respective fault probability value indicates a probability that the respective transformer component 16 is in a fault state. The fault state can be an abnormal state of the respective transformer component 16.

[0055] In the Figure 2Figure 18 shows an example of such a Bayesian network, which is designed as a directed graph 20 from the transformer components 16, which in this case are called the first transformer component T1 and the second transformer component T2, to the state variables S1, S2. The directed graph 20 is based on the first probability values, the second probability values, and the third probability values. The graph 20 is a directed, acyclic, and bipartite graph, such that there are only relationships from the transformer components T1, T2 to the state variables S1, S2. The state variables S1, S2 thus depend on the states of the transformer components T1, T2. However, the states of the transformer components T1, T2 do not depend on the states of the state variables S1, S2. Furthermore, there are no relationships between the state variables S1, S2.It is also preferred that graph 20 does not depict any relationships between the transformer components T1 and T2. The transformer components T1 and T2 form a first group of nodes in graph 20. The state states S1 and S2 form a second group of nodes in graph 20. The previously mentioned probability values ​​will now be discussed.

[0056] For each status variable S1, S2, a corresponding first probability value (1st W-value) is provided indicating that the respective status variable S1, S2 represents an incorrect state. In the Figure 2Table 48 shows that the first status variable S1 has a corresponding first probability value of 0.2. The second status variable S2 has a corresponding first probability value of 0.3. The first probability value of 0.2 for the first status variable S1 indicates that the state represented by S1 is incorrect. For example, if the first status variable S1 represents the normal state N, there is a 20% probability that the normal state N is incorrect and that the actual state should be an abnormal state A. The same applies to the second status variable S2 and its corresponding first probability value of 0.3.

[0057] For each transformer component T1, T2, a corresponding second probability value (2nd W-value) is provided indicating that a fault has occurred in the respective transformer component T1, T2. For example, if the first transformer component T1 is formed by the primary coil and the second transformer component T2 by the secondary coil, statistical analysis can determine, for instance, that there is a 20% probability that the primary coil has a fault and a 30% probability that the secondary coil has a fault. For each functional relationship, represented by an arrow in graph 20, from one of the transformer components T1, T2 to one of the status variables S1, S2, a corresponding third probability value (3rd W-value) is provided indicating that a fault in the respective transformer component T1, T2 causes an abnormal state A of the respective status variable S1, S2.For example, if the transformer component Ti is in an abnormal state A, then the third probability value with respect to the relationship between the first transformer component T1 and the first status variable S1 is 0.5. This means there is a 50% probability that if the transformer component T1 assumes the abnormal state A, then the status state represented by the status variable S1 is the abnormal state A. Therefore, the corresponding third probability value of 0.5 is assigned to the relationship between the transformer component T1 and the status variable S1. However, as can be seen from graph 20, there is also a case involving the transformer component T1 and the second status variable S2.Table 52 shows that if the first transformer component T1 is in abnormal state A, then, based on the corresponding third probability value, there is a 90% probability that the state represented by the second status variable S2 is abnormal state A. The device 2 has a memory unit 8, which stores a data set comprising the first probability values, the second probability values, and the third probability values. The processor unit 6 of the device 2 is configured to read the data set and / or the probability values ​​from the data set from the memory unit 8.

[0058] Based on the previously explained relationships between the first, second, and third probability values ​​and graph 20, the processor unit 6 can use the corresponding data to generate a Bayesian network 18. However, it is also possible that the Bayesian network 18 already exists, based on the directed graph 20 and the aforementioned probability values. The Bayesian network 18 can thus represent a specific mapping between the transformer components 16 or T1, T2 and the status variables S1, S2. The status signal S of the measuring system 10, received via the signal input interface 4 of the device 2, represents several different status variables—in this case, actually existing status variables. Based on the Bayesian network 18 and the status states represented by the status variables of the measuring signal, the processor unit 6 can therefore generate a Bayesian network 18 for each transformer component 16 or T2.The processor unit 6 determines a corresponding fault probability value for each transformer component 16 (T1, T2) indicating that the respective component is in a fault state. Each fault probability value thus represents the probability that the respective transformer component 16 (T1, T2) is actually in this fault state. The processor unit 6 can be configured to determine the fault probability values ​​by evaluating the Bayesian network 18 and taking into account the actual status states, for example, using one of the following algorithms: Sampling Algorithm or Variable Elimination Algorithm. The processor unit 6 can be trained and / or configured to execute the respective algorithm.

[0059] In particular, the processor unit 6 is configured to generate a multitude of patterns 22 based on the Bayesian network 18 and the first, second, and third probability values, where each pattern 22 represents a corresponding pattern state 24 as normal state N or abnormal state A for each status variable and for each transformer component T1, T2, and 16, respectively. Figure 3A table is presented showing a total of twelve patterns 22, numbered from No. 1 to No. 12, where each pattern 22 represents a pattern state 24 as a normal state (N) or abnormal state (A) for each transformer component T1, T2 and each status variable S1, S2. Pattern No. 1 thus represents the pattern states N, N, N, N in the following order for: first transformer component T1, second transformer component T2, first status variable S1, and second status variable S2. Similarly, the second pattern No. 2 represents the pattern states A, N, N, A. The remaining pattern states are derived from the Figure 3The table shown illustrates this. Processor unit 6 generates the pattern states 24 such that the probability values ​​shown in Tables 48, 50, and 52 are satisfied. In practice, this often requires generating a large number of patterns with their corresponding pattern states. For example, processor unit 6 is configured to generate a large number of at least 10,000 patterns 22. The in Figure 3The numerous patterns 22 shown are therefore only exemplary and do not claim to fulfill all the conditions of Tables 48, 50, and 52. However, considering the pattern states 24 in the patterns 22 for the first transformer component T1, it follows from the pattern states 24 for the first ten patterns that two pattern states 24 represent the abnormal state A and the remaining pattern states 24 represent the normal state N. This results in a probability value of 20% for the abnormal state A for the first transformer component T1. In the first Table 48 of the Figure 2 is a corresponding probability value for the abnormal state A with respect to the first transformer component T1, namely a probability value of 0.2.

[0060] A status signal S is transmitted from the measuring system 10 to the signal input interface 4 of the device 2, the status variables transmitted thereby representing the actual status states either as normal state N or as abnormal state A. If the first status variable S1 represents a normal state N and the second status variable S2 represents an abnormal state A, then the values ​​can be determined from the table of Figure 3 Exactly two patterns 22 are identified, which specify the first status variable S1 as the normal state N and the second status variable S2 as the abnormal state A. These are patterns numbered 2 and 8. In the Figure 4 Only these two patterns, numbered 2 and 8, are shown. The processor unit 6 is preferably configured to execute a corresponding process step. It is therefore provided that the processor unit 6 is configured to select from the multitude of patterns 22 ( Fig. 3) to determine a reference group 26 of patterns 22 (in the example, patterns no. 2, 8) such that for each pattern (no. 2 and no. 8) of reference group 26, the associated pattern states 24 for the status variables S1, S2 correspond to the status states of the status variables of the status signal S. Furthermore, the processor unit 6 is configured to determine, for each transformer component T1, T2, an associated fault probability value based on the pattern states 24 for the transformer components T1, T2 from the patterns 22 (no. 2, 8) of reference group 26, that the respective transformer component T1, T2 is in an abnormal state A. The table from Figure 3The patterns (patterns 2, 8) of reference group 26 are specified, exhibiting status variables S1, S2 that correspond to the actual status variables S1, S2 determined by measuring system 10. For each of the patterns 22 of the reference group, a pattern state 24 is also specified for each of the two transformer components T1, T2. The fault probability value for each transformer component T1, T2 can be determined from the mean of the specified pattern states. For the first transformer component T1, pattern no. 2 specifies an abnormal state A and pattern no. 8 specifies a normal state N. The mean is therefore 50%, so the fault probability value for the first transformer component T1 is 0.5. For the second transformer component T2, pattern state 24 is specified as normal state N in both patterns no. 2, 8, so the mean is 100% normal state N.For the second transformer component T2, the error probability value is therefore 0. In summary, it can be stated that if the measuring system 10 detects a normal state N for the first state variable and an abnormal state A for the second state variable, then a fault probability value of 0.5 is determined for the first transformer component T1 and a corresponding fault probability value of 0 for the second transformer component T2 using the device 2. This means that the second transformer component T2 is in a normal state N and that the first transformer component T1 is 50% likely to be in a fault state or abnormal state A. The fault state and abnormal state A can be used synonymously.

[0061] As from the Figure 1As can be seen schematically, the device 2 can be coupled to a separate display device 30 via a signal line. Alternatively or additionally, the device 2 can have an integrated display unit 34. The display unit 34 can be coupled to the processor unit 6. The processor unit 6 can be configured and / or designed to control the display unit 34 such that the display unit 34 reproduces an image that graphically represents the determined error probability values ​​and indicates and / or identifies the associated transformer components 16. This allows a person viewing the display unit 34 of the device 2 to immediately recognize which of the transformer components 16 has an abnormal state A or a fault condition and with what probability.Experience shows that such a finding cannot be derived directly and immediately from the status states generated by the measuring system 10.

[0062] It is possible that the device 2 has an output signal interface 36, wherein the device 2 is configured to generate an output signal Q that represents at least the largest of the determined error probability values ​​and the associated transformer component 16. Furthermore, the device 2 is preferably configured and / or configured to control the signal output interface 36 such that the output signal Q is transmitted, in particular sent, via the signal output interface 36, preferably to the display device 30.

[0063] It has already been explained that the device 2 can be part of a system 28, which, in addition to the device 2, also includes the display unit 30 with a screen. The device 2 can be designed and / or configured to transmit the output signal Q to the display unit 30 via the signal output interface 36, so that the display unit 30 shows an image that graphically represents the error probability value and indicates and / or identifies the associated transformer component 16. The display unit 30 can be designed accordingly for this purpose. A signal connection, in particular a wired and / or wireless signal connection, can exist between the signal output interface 36 and the display unit 30.

[0064] In practice, there is often interest in the transformer component 16 for which the highest error probability value has been determined, since in this case it can be assumed that the corresponding transformer component 16 is very likely in an abnormal state A. However, the error probability value represents a probability that the respective transformer component 16 is in an abnormal state A. Therefore, a residual probability remains that the respective transformer component 16 is not in the abnormal state A, but in a normal state N. In practice, there is therefore often also interest in the second-highest or third-highest error probability values. It is therefore preferably provided that the device 2 is configured to generate the output signal Q such that the output signal Q represents at least the two or three highest error probability values ​​and the corresponding transformer components 16.The display unit 30 of the system 28 is preferably configured to display a second image on the screen of the evaluation unit 30, indicating the transformer components 16 and associated fault probability values, based on the output signal Q. The display unit 30 can be configured such that the image displayed on the screen shows the transformer components 16 represented by the output signal Q in list form and the associated fault probability values ​​qualitatively by means of different colors and / or bars. Figure 5The screen image displayed is shown as an example, comprising the aforementioned image, in which the transformer components 16 are listed and the associated error probability values ​​are qualitatively represented by (colored) bars of varying lengths. The longer the bar, the greater the error probability determined for the respective transformer component 16. The transformer components 16 are listed according to the magnitude of their error probability values, with the transformer component 16 with the highest error probability value at the top.From this list-like representation of the transformer components 16 and their associated fault probability values, a person can see at a glance which of the transformer components 16 is most likely to be in an abnormal state A, and simultaneously, by comparing the color and / or length of the bars, can see which of the other transformer components 16 are also likely to be in an abnormal state A. This facilitates maintenance and / or troubleshooting in a power transformer 12.

[0065] The in Figure 1System 28, shown schematically, comprises the device 2 and the display unit 30. System 28 may additionally include the measuring system 10 with its associated sensors 14. The display unit 30 may be connected to or integrated into the measuring system 10. Furthermore, system 28 may also include the power transformer 12.

[0066] The measuring system 10 can include a first sensor 14 designed to detect the temperature of the coolant in the tank of the power transformer 12. A second sensor 14 can be designed to detect a mechanical stress or force acting on the primary coil. A third sensor 14 of the measuring system 10 can be designed to detect a voltage between the terminals 44 on the primary side of the power transformer 12. A fourth sensor 14 of the measuring system can be designed to detect the loss factor and / or capacitance of one or more bushings of the power transformer 12. A fifth sensor 14 of the measuring system 10 can be designed to detect acoustic signals generated when a tap changer of the power transformer 12 is operated. Furthermore, other sensors or additional sensors 14 are also possible for the measuring system 10.

[0067] Furthermore, it is preferably provided that the device 2 is configured to periodically re-acquire the fault probability values ​​for the transformer components 16, preferably always after a status signal S with new, multiple, different status variables is received from the measuring system 10. The measuring system 10 can also be configured to periodically evaluate the properties of the power transformer 12 acquired by the associated sensors 14. After each periodic acquisition of the properties by means of the associated sensors 14, the measuring system 10 can transmit a status signal S to the signal input interface 4 of the device 2. Subsequently, the determination of the fault probabilities for the transformer components 16 takes place by means of the processor unit 6.Furthermore, the output signal Q can be updated based on the most recently determined error probabilities by means of the processor unit 6, so that the display device always shows an up-to-date image of which of the transformer components 16 have the highest probability of a fault.

[0068] It should be further noted that "having" does not exclude any other elements or steps, and "a" or "an" does not exclude a plurality. It should also be noted that features described with reference to one of the above embodiments may also be used in combination with other features of other embodiments described above. Reference numerals in the claims are not to be considered as a limitation. REFERENCE MARK LIST

[0069] AA Normal state Q Output signal S Status signal N Normal state S1 Status variable S2 Status variable T1 Transformer component T2 Transformer component 2 Device 4 Device signal input interface 6 Processor unit 8 Storage unit 10 Measuring system 12 Power transformer 14 Sensor 16 Transformer component 18 Bayesian network 20 Graph 22 Pattern 24 Pattern state 26 Reference group 28 System 30 Display unit 34 Device display unit 36 ​​Device signal output interface 38 Input interface 40 Output interface 42 Evaluation unit 44 Primary side connections 46 Secondary side connections 48 First table (1 W values) 50 Second table (2 W values) 52 Third table (3 W values)

Claims

1. Device< / b> (2) for determining a fault probability value for a transformer component (16), comprising a signal input interface (4), a processor unit (6), and a storage unit (8), the signal input interface (4) being designed for direct or indirect coupling to a measurement system (10) for a power transformer (12), the measurement system (10) having a plurality of sensors (14) which are each designed to detect a physical and / or chemical property of the power transformer (12), the power transformer (12) having a plurality of transformer components (16), and wherein each sensor (14) is coupled to at least one of the transformer components (16) of the power transformer (12) via a respectively associated direct or indirect connection, so that the property of the power transformer (12) detected by the respective sensor (14) is influenced by at least one of the transformer components (16); the signal input interface (14) being designed to receive a status signal S of the measuring system (10), the status signal S representing a plurality of different status variables S1, S2, each status variable S1, S2 being assigned to at least one of the sensors (14), so that each status variable S1, S2 is in a functional relationship with at least one of the transformer components (16), wherein each status variable S1, S2 represents a status state of the property of the power transformer (12) detected by the at least one associated sensor (14) as a normal state N or as an abnormal state A; whereby a data set is stored by the memory unit (8), the data set having an associated first probability value for each status variable S1, S2, that the respective status variable S1, S2 represents an incorrect state, the data set having an associated second probability value for each transformer component (16), that an error has occurred in the respective transformer component (16), wherein the data set for each of the functional relationships of one of the transformer components (16) to one of the status variables S1, S2 has an associated third probability value that an error in the respective transformer component (16) causes an abnormal state A of the respective status variable S1, S2; and wherein the processor unit (6) is configured to determine for each transformer component (16) an associated fault probability value based on a Bayesian network (18) representing a directed graph (20) from the transformer components (16) to the status variables S1, S2 based on the first, second and third probability values, and based on the status states represented by the status variables S1, S2 that the respective transformer component (16) is in a fault state; characterized in that the Bayesian network (18) is formed as a bipartite graph (20) representing a first set of nodes, a second set of nodes, and functional relationships from the first set of nodes to the second set of nodes, the first set of nodes being determined by the first probability values, the second set of nodes being determined by the second probability values, and the functional relationships being determined by the third probability values.

2. The device (2) according to the preceding claim, characterized in that the processor unit (6) is configured to generate a plurality of patterns (22) based on the Bayesian network (18) representing a directed graph (20) from the transformer components (16) to the status variables S1, S2 based on the first, second and third probability values, each pattern (22) for each status variable S1, S2 and for each transformer component (16) having a respective associated pattern value, second and third probability values, to generate a plurality of patterns (22), each pattern (22) representing for each status variable S1, S2 and for each transformer component (16) a respective associated pattern state (24) as normal state N or abnormal state A; wherein the processor unit (6) is configured to determine a reference group (26) of patterns (22) from the plurality of patterns (22), so that for each pattern (22) of the reference group (26) the associated pattern states (24) for the status variables S1, S2 correspond to the status states of the status variables S1, S2 of the status signal S; and wherein the processor unit (6) is configured to determine for each transformer component (16) an associated fault probability value based on the pattern states (24) for the transformer components (16) from the patterns (22) of the reference group (26) that the respective transformer component (16) is in an abnormal state A.

3. Device (2) according to one of the preceding claims, characterized in that the device (2) has an output interface (40), wherein the device (2) is designed to generate an output signal Q which represents at least the largest of the error probability values and the associated transformer component (16), and wherein the device (2) is designed to transmit the output signal Q by means of the signal output interface (36).

4. Device (2) according to the preceding claim, characterized in that the device (2) is designed to generate the output signal Q in such a way that the output signal Q represents at least the two or three largest error probability values and the associated transformer components (16).

5. Device (2) according to one of the preceding claims, characterized in that the status variables S1, S2 represented by the status signal S are ternary status variables which each represent exactly one status state from the following group of possible status states: Normal state N, Abnormal state A or Non-installation state.

6. Device (2) according to the preceding claim, characterized in that the processor unit (6) is configured to generate the Bayesian network (18) based on the data set.

7. The device (2) according to any one of the preceding claims, characterized in that the processing unit (6) is configured to generate the patterns (22) by means of a likelihood weighting algorithm.

8. Device (2) according to one of the preceding claims 4 to 7, characterized in that the processor unit (6) is designed to generate the output signal Q in such a way that the output signal Q represents the associated error probability values and transformer components (16) as a first image in which the error probability values are graphically represented and the transformer components (16) are indexed and / or designated.

9. System (28), comprising: A device (2) according to any one of the preceding claims 4 to 7, and a display unit (30) with a screen, wherein the signal output interface (36) of the device (2) is coupled to the display unit (30) via a direct or indirect first signal connection to transmit the output signal Q to the display unit (30), and wherein the display unit (30) is designed to display a second image on the screen indicating the transformer components (16) and associated fault probability values based on the output signal Q.

10. System (28) according to the preceding claim, characterized in that the display unit (30) is designed so that the second image displays the transformer components (16) represented by the output signal Q in list form and the associated error probability values qualitatively by means of different colors and / or bars.

11. System (28) according to one of the preceding claims 9 to 10, characterized in that the system (28) comprises a measuring system (10) for the power transformer (12), wherein the measuring system (10) is coupled to the signal input interface (4) of the device (2) in order to transmit the status signal S to the signal input interface (4), wherein the measuring system (10) comprises a plurality of sensors (14), which are each designed to detect the physical and / or chemical property of the power transformer (12), wherein the power transformer (12) has a plurality of transformer components (16), and wherein each sensor (14) is coupled to at least one of the transformer components (16) of the power transformer (12) via a respectively associated direct or indirect connection, so that the property of the power transformer (12) detected by the respective sensor (14) is influenced by at least one of the transformer components (16).

12. System (28) according to the preceding claim, characterized in that the measuring system (10) is designed to detect at least two of the following properties of the power transformer (12) by sensors: Gas from hydrocarbon chains in the liquid insulating medium, water in the liquid insulating medium, filling level of the liquid insulating medium, electrical voltage, electrical current, winding tension force, rotational speed of a fan, electrical power of the fan motor, volume of the fan, motor torque of a tap changer, switching sequence of the tap changer, hot spot temperature, top oil temperature, dissipation factor of one or more bushings of the power transformer (12), capacitance of one or more bushings of the power transformer (12), acoustic signals of a tap changer, vibroacoustic intensity, and functionality of the dehumidifier.

13. System (28) according to one of the preceding claims 9 to 12, characterized in that the display unit (30) forms part of the measuring system (10).

14. The system (28) according to any one of the preceding claims 9 to 13, characterized in that the system (28) comprises a remote monitoring system, wherein the signal output interface (36) of the device (2) is coupled to the remote monitoring system via a direct or indirect second signal connection to transmit the output signal Q to the remote monitoring system.

15. System (28) according to the preceding claim, characterized in that the remote monitoring system is cloud-based.

16. System (28) according to any one of the preceding claims 14 to 15, characterized in that the display unit (30) forms part of the remote monitoring system.

17. The system (28) according to any one of the preceding claims 9 to 16, characterized in that the system (28) comprises the power transformer (12).