Apparatus and system for indirectly determining temperature of power transformer

Abstract

The invention relates to a device (2) having a signal interface (4), a processor unit (6) and a memory unit (8), the signal interface (4) being configured to receive measurement signals representing first sensor data and second sensor data of a power transformer (10). Based on the sensor data and a simulation model for the power transformer (10), the processor unit is configured to determine a value for a temperature in the power transformer (10) which is not detected with a sensor. The invention also relates to a system (1) comprising a plurality of such devices (2).

Description

Technical Field

[0001] The present invention relates to a device for indirectly determining the temperature of a power transformer and a system for indirectly determining the temperature of a power transformer. Background Technology

[0002] Power transformers are known in principle in the prior art. A power transformer is used to convert electrical power from its primary side to its secondary side. Power transformers are typically constructed to convert very high power. Therefore, a power transformer can, for example, convert at least 30 kW, at least 50 kW, or at least 100 kW. Heat is generated due to the conversion of electrical power, and this heat is usually dissipated to the surrounding environment through the power transformer's cooling system. A power transformer has a primary side with a primary winding and a secondary side with a secondary winding. The power transformer has a housing forming a tank. The two windings are arranged in the tank of the power transformer, in which a coolant, particularly oil, is present, and the tank is connected to a cooler in the cooling system via a fluid loop. The cooling performance of the cooler typically depends on the ambient temperature. If the ambient temperature is low, a large temperature difference between the ambient temperature and the temperature on the cooler can contribute to good cooling performance on both the primary and secondary sides. If the ambient temperature rises, the cooling performance of the cooler may decrease. To prevent the coolant from overheating, power transformers are typically equipped with sensors to detect the temperature at at least one location on the power transformer.

[0003] Mathematical models suitable for modeling power transformers are known from the prior art. However, these models have the following drawbacks: several parameters of the model must be pre-defined so that the model ensures a robust mapping to actual power transformers. The parameters of the model must usually be determined based on the actual characteristics of the power transformer. In practice, it has been found that determining such parameters for the model is not insignificantly costly. Therefore, these models are rarely used in practice to model the characteristics of power transformers. Summary of the Invention

[0004] Therefore, the object of the present invention is to provide an apparatus and / or a system having such an apparatus, wherein the relevant temperature of a power transformer should be reliably and as simply as possible deduced by means of said apparatus.

[0005] According to a first aspect of the invention, the objective is achieved by a device according to the invention. Therefore, a device is provided having a signal interface, a processor unit, and a memory unit. The signal interface is configured to receive measurement signals representing first sensor data and second sensor data, wherein the first sensor data and the second sensor data respectively represent at least one parameter detected by a sensor in or on a power transformer. The processor unit is configured to generate, in successive time windows, sets of measurement data formed by the first and second sensor data received in the respective time windows. A simulation model of the power transformer is stored in the memory unit, the simulation model reflecting the transfer characteristics from input data to output data according to model parameters of the simulation model. The processor unit is configured to repeatedly execute a group of steps a) through c) for each time window: a) determining output data by executing a simulation model with input data and model parameters, the input data being determined by first sensor data belonging to a measurement data group for the corresponding time window, the model parameters being determined by pre-given reference parameters when the simulation model is first executed for the corresponding time window, and the model parameters being determined by respectively finally determined optimized parameters in other cases; b) determining a target value, the target value including at least one squared error between second sensor data belonging to a measurement data group for the corresponding time window and previously determined output data, weighted by a first weighting factor; and c) determining optimized parameters as new model parameters for the simulation model based on at least one target value determined in the corresponding time window. The simulation model also reflects model values ​​for temperature within the power transformer based on the input data and model parameters. The processor unit is configured to determine model values ​​for each time window by executing a simulation model with input data and model parameters, the input data being determined by first sensor data belonging to a measurement data group for the corresponding time window, the model parameters being determined by respectively finally determined optimized parameters for the corresponding time window.

[0006] Before delving further into the characteristics of the device according to the first aspect, it is important to emphasize that the device, and particularly its processor unit, can refine the model parameters used for the simulation model from one time window to another, enabling the determination of model values ​​for temperatures within the power transformer that have not previously been detected by sensors on or within the power transformer. This allows for particularly reliable monitoring of the power transformer's thermal state. Furthermore, the iterative optimization of the model parameters allows for a very simple start-up of the device. While the model parameters used during the initial simulation should be good estimates, they are not necessarily optimal. Therefore, these model parameters can be determined at startup using pre-defined reference parameters. After the model parameters used for the simulation model have been refined over multiple time windows, the processor unit can determine at least one temperature within the power transformer using the simulation model, and this determined temperature can be used to monitor the power transformer. This allows for better utilization of the transformer's power capacity, and more precisely, better utilization of the transformer's power capacity at its thermal power limit.

[0007] The features of the device will be described in more detail below.

[0008] The device includes a signal interface, a processor unit, and a memory unit. The signal interface can be a wired or wireless signal interface. The signal interface is configured to receive measurement signals representing data from a first sensor and data from a second sensor. The device may have another signal interface configured to transmit an output signal representing a model value for temperature determined by the processor unit. In principle, it is feasible for the signal interface for receiving the measurement signal and the other signal interface for transmitting the output signal to be formed by a common signal interface. However, the signal interface and the other signal interface can also be configured as separate signal interfaces.

[0009] The first sensor data and the second sensor data each represent at least one parameter detected by a sensor in or on the power transformer. Preferably, this relates to a parameter detected by a sensor. The corresponding sensor-detected parameter preferably relates to a mechanical, electrical, or other physical parameter. The first sensor data represents a preferred sensor-detectable and externally sourced parameter of the power transformer. The second sensor data represents a preferred sensor-detectable, however non-externally sourced parameter of the power transformer. The first sensor data can therefore represent sensor-detectable parameters of the power transformer that act on the power transformer externally and / or unaffected. These parameters may, for example, be the voltage applied to the terminals on the primary side of the power transformer, the current flowing through the primary windings, or the ambient temperature of the power transformer. External parameters of the power transformer can influence non-externally sourced parameters, such as the temperature of the coolant within the power transformer's tank. The parameter represented by the second sensor data is therefore preferably a parameter of the power transformer caused by the parameter represented by the first sensor data.

[0010] The parameters of the power transformer, represented by the first and second sensor data, are preferably detected at a predetermined frequency, such that the measurement signal preferably represents a continuous data stream of sampled values ​​in the form of the first and second sensor data. Therefore, the processor unit is configured to generate measurement data sets formed by the first and second sensor data received in the respective time windows within successive time windows. Thus, if a continuous data stream of the first and second sensor data is delivered to the processor unit via a signal interface by means of a measurement signal, the processor unit copies the first and second sensor data from the data stream at the respective time window to generate a measurement data set belonging to the respective time window. The time windows may partially overlap in time. For example, if a time window has a duration of 24 hours, and if the processor unit generates a corresponding measurement data set for each new time window of the past 24 hours every four hours, then the successive time windows at least partially overlap, and this also applies to the measurement data sets accordingly. In the previous example, for example, the last determined time window and the last determined time window may overlap over a sixteen-hour time period. Each time window may overlap with multiple preceding time windows. If the time windows overlap, it is preferable to copy data from the data stream to generate measurement data sets belonging to the respective time windows. Processor units can be configured accordingly for this purpose. However, it is also feasible in principle for the time windows to proceed directly and non-overlappingly. That is, the time windows can be determined to be non-overlapping. It is also feasible to set defined or arbitrary time pauses between each time window, during which data received via the measurement signal is not used to generate measurement data sets. Each time window preferably has a predetermined constant time parameter. If multiple measurement data sets are generated for a corresponding time window, that is, the time windows are moved forward in time at discrete, predetermined intervals, and a corresponding measurement data set is generated for each time window using the processor unit. As previously explained, it is preferable that the duration of each time window is, for example, 24 hours. Furthermore, the time windows can be moved forward by four hours each time. However, in practice, this does not involve the same time window, but rather a new time window with the same parameters.

[0011] A simulation model is stored in the memory unit of the device. The simulation model is preferably a mathematical model of the power transformer executable by a processor unit, wherein the simulation model reflects the transfer characteristics from input data to output data based on its model parameters. Input data may be formed from data representing external parameters of the power transformer that can be detected by sensors. Output data may be formed from data representing parameters of the power transformer that cannot be detected by sensors and / or representing non-external parameters of the power transformer. The simulation model preferably also reflects the transfer characteristics from input data to model values ​​based on its model parameters.

[0012] The simulation model offers the advantage that it can be used by a processor unit to estimate at least one or more states of a power transformer. The state of the power transformer can be, for example, the temperature existing within the power transformer, or other physical parameters of the power transformer that depend on external parameters of the power transformer. Therefore, the output data that may be generated during simulation using the simulation model generally relates to transformer characteristics similar to or the same as those of the second sensor data, which is represented by a measurement signal and reaches the processor unit via the measurement signal. However, the output data generated during simulation using the simulation model is preferably understood here as an estimate of the transformer characteristics, while the second sensor data detected by the sensor is preferably understood as the actual value for the corresponding transformer characteristic. In other words, the output data can represent an estimate of the transformer characteristics, and the second sensor data can represent the actual value detected by the sensor for the same sensor characteristic. Therefore, the output data and the second sensor data can be compared between the estimate and the actual value, where the result of the comparison represents the error of the estimate relative to the actual value, or the error of the actual value relative to the estimate. If the error is small, the simulation model can be considered to reflect the power transformer particularly well. To reduce errors, it is preferable to repeat the above groups (steps a), b), and c) for each time window in a group to optimize the model parameters of the simulation model. The aim here is to minimize the weighted square error between the second sensor data in the measurement data group belonging to the corresponding time window and the output data generated by simulation when performing step a) in the group of corresponding steps a) to c). Thus, in step c), the processor unit can be used to determine the optimized parameters as new model parameters for the simulation model, thereby improving the model parameters after repeatedly executing the group of steps a) to c), so that the weighted square error is generally reduced to a minimum. Preferably, the group of steps a) to c) is repeated for at least 100 time windows. Here, the group of steps a) to c) is repeated for each time window. That is, the aforementioned grouped repetition is performed.

[0013] Furthermore, the simulation model is constructed to generate model values ​​for the temperature within the power transformer based on input data and model parameters, particularly the model parameters updated with appropriate optimization parameters, during simulation. These model values ​​preferably represent the temperature at a predetermined location within the power transformer. This location can be any predetermined location within the power transformer. The temperature can be, for example, the hotspot temperature on one winding of the primary or secondary winding. However, the model value can also be the so-called top-oil temperature, especially when the coolant is formed from oil in the power transformer's tank. Preferably, the top-oil temperature is in the upper region of the tank.

[0014] To achieve the most accurate possible updates to the model values, it is stipulated that the processor unit determines the model values ​​by executing a simulation model for each time window. Here, the simulation model is executed with input data and model parameters, the input data being determined by first sensor data belonging to a measurement data set corresponding to the respective time window, and the model parameters being determined by the last determined optimized parameters for the respective time window. The processor unit can be configured accordingly for this purpose. Therefore, for each time window, a model value for one of the previously mentioned temperatures within the power transformer can be determined by the processor unit. This temperature can then be represented by an output signal, which can be transmitted, in particular, sent, via the device's signal interface or another signal interface. Thus, information regarding the previously mentioned temperature within the power transformer is provided to the receiver at a sufficient update rate.

[0015] The target value is determined in step b), which can be executed by the processor unit. The target value may be formed by the sum of multiple components. However, it is specified that the target value includes at least the following weighted squared error as a component, or the target value may even be formed solely by the following weighted squared errors. For example, it is specified that the target value includes, or is formed solely by, the squared error between the second sensor data weighted by a first weighting factor and the previously determined output data in the measurement data set assigned to the corresponding time window. A simple (non-quadratic) error can be formed, for example, by the difference between the second sensor data and the output data. Preferably, the processor unit is configured to determine this error. The second sensor data and the output data can be understood as data vectors, where a simple error can be formed by the difference between these two vectors. The squared error can in turn be formed by the sum of squared terms of the simple error. The processor unit can be configured for this purpose. The squared error is therefore preferably a one-dimensional parameter.

[0016] The processor unit is preferably configured to execute a pre-given optimization method to determine optimization parameters in step c). For example, the processor unit may be configured to repeatedly perform the group including steps a) through c) multiple times until a target value less than a pre-given threshold is reached in step b) and / or a pre-given interrupt criterion is met in steps b) or c). For example, the processor unit may be configured to perform a so-called SQP (Sequential Quadratic Programming) optimization method or an Active-Set-Verfahren method for steps a) through c) to solve the SQP problem. The processor unit may then be configured to determine optimization parameters in step c) based on the target value finally determined in step b) or based on multiple target values ​​from multiple previous steps b). If multiple target values ​​from previous steps b) are considered, it is preferable that these steps have been performed for the same time window. Therefore, in step c), the processor unit can determine optimization parameters based on one or more target values ​​such that a smaller target value is determined when step b) is executed again. The optimization process can be repeated multiple times until a particularly small target value is determined in step b) and optimization parameters or new model parameters for a new simulation model are determined in step c), so that the simulation model can reflect the power transformer particularly well and accurately.

[0017] An advantageous configuration of the device is characterized in that the temperatures of the power transformer reflected by the model values ​​include, or are not limited to, the hot spot temperature, the upper oil temperature, or the bottom oil temperature of the power transformer. The hot spot temperature, upper oil temperature, or bottom oil temperature are generally temperatures of particular interest and relevance to the operation of the power transformer. These temperatures are not always detected by sensors. Therefore, the simulation model provides a feasible solution whereby the processor unit uses the simulation model to determine model values, wherein the model values ​​reflect and thus represent at least one or more of the temperatures of the power transformer. Thus, multiple temperatures and / or particularly relevant temperatures can be determined using the simulation model, temperatures that are not necessarily detected by sensors and / or are not required to be detected by sensors.

[0018] Another advantageous configuration of the device is characterized in that the temperature reflected by the model value in the power transformer relates to the temperature in the upper region of the oil in the power transformer or to the temperature at a location in one of the windings of the power transformer. In particular, the temperature directly at a winding of the power transformer can only be detected by sensors at a high cost. This cost can be reduced by using a simulation model. This is because second sensor data and output data can be used to optimize the model parameters of the simulation model, where the second sensor data represents parameters of the power transformer detected by sensors. Such parameters represented by the second sensor data can, for example, be power transformer parameters that can be easily or cost-effectively detected by sensors. Therefore, it is feasible to determine the temperature at a predetermined location in the power transformer at a reduced cost.

[0019] Furthermore, it can be advantageously specified that at least one additional model value is provided, which reflects a corresponding physical parameter of the power transformer based on the input data and model parameters. The processor unit can also be configured to determine each of the model values ​​for each time window by executing a simulation model with the input data and model parameters. The additional model value may, for example, reflect the current, voltage, mechanical force, or other physical parameters of the power transformer. Therefore, when referring below to the model value for temperature within the power transformer, it is preferably specified that the associated preferred description, advantageous features, technical effects, and advantages can be similarly applied to each additional model value.

[0020] Another advantageous configuration of the device is characterized by the number of group repetitions of steps a) to c) being between 10 and 1000. Here, group repetition means repeating the entire group having steps a) to c). That is, steps a) to c) are performed separately so that repetition can then begin, or more precisely, repetition can begin again with steps a) to c). Thus, the steps are repeated in groups. An advantageous range for the number of repetitions is given by a minimum of ten repetitions and limiting the number of repetitions to a maximum of 1000 repetitions, which has been shown in practice to determine the model parameters used for the simulation model, thereby allowing the power transformer to be reflected particularly well by the simulation model and its associated model parameters.

[0021] Another advantageous configuration of the device is characterized in that the device is configured to determine a target value in step b) from the execution for the second time window such that the target value also includes the squared difference between the last determined model parameters, weighted by a second weighting factor, and the last determined optimized parameters for the previous time window. As already explained at the beginning, the processor unit is configured to generate sets of measurement data formed by the first and second sensor data received in the respective time windows, respectively. Thus, the second time window is the second time window in the sequence of time windows. The sets having steps a), b), and c) are executed multiple times sequentially by the processor unit of the device for each time window, wherein within each set, step b) follows step a) and step c) follows step b). In each step c), the processor unit determines the optimized parameters as new model parameters for the simulation model.

[0022] For each time window, step c) is executed multiple times due to the repetition of the group. Therefore, there is also a final execution of step c), so that for the corresponding time window, the finally determined optimization parameters are used as new model parameters for the simulation model and determined with reference to the corresponding signal window during the last execution of step c).

[0023] In step b) of the group of steps a) to c), a target value is determined. According to a first aspect of the invention, the target value includes at least one squared error, weighted by a first weighting factor, between second sensor data belonging to a group of measurement data for a corresponding time window and previously determined output data. According to a recently mentioned advantageous configuration of the device, it is preferably specified that the target value also includes, weighted by a second weighting factor, the squared difference between the last determined model parameters and the last determined optimized parameters for the previous time window. Therefore, when repeating the group of steps a) to c) for a second target window and from the second target window, target values ​​are determined, each formed by at least two components: a weighted squared error and a weighted squared difference. The first and second weighting factors determine which of these two components defines the target value with a larger share. It has proven advantageous that the second weighting factor is greater than the first weighting factor. In this case, the target value is more strongly influenced by the squared difference. A larger second weighting factor prevents rapid changes in the model parameters. Because the squared difference is primarily based on the simple difference between the last determined model parameters and the last determined optimization parameters for the previous time window. To determine the target value in step b) when executing for the second time window, the simple difference between the last determined model parameters and the last determined optimization parameters for the first time window can be determined first, for example, using a processor unit. The model parameters and optimization parameters can be understood as vectors, and therefore the simple difference between them can also be understood as a vector. The squared difference is here understood as the sum of squared terms of the simple differences of the vectors. Therefore, the squared difference is a one-dimensional value. The processor unit can be configured to determine the simple difference and / or the squared difference. The target value can be determined by the sum of the error weighted by a first weighting factor and the squared difference weighted by a second weighting factor. The processor unit is preferably configured to determine the target value accordingly.

[0024] Another advantageous configuration of the device is characterized in that the device is configured to determine a target value in step b), such that the target value also includes the squared difference between the finally determined model parameters and the reference parameters, weighted by a third weighting factor. Therefore, the target value can, for example, be formed by the sum of three components. The first component can be formed by the squared error weighted by a first weighting factor, the second component by the squared difference weighted by a second weighting factor, and the third component by the squared difference weighted by a third weighting factor. The squared differences weighted by the second and third weighting factors are distinct. The squared difference belonging to the second weighting factor refers to the squared difference as illustrated in the advantageous embodiments described above. The squared difference weighted by the direct weighting factor refers to the squared difference between the finally determined model parameters and the reference parameters. To determine this squared difference, reference is made in a similar manner to the advantageous descriptions, preferred features, and effects as for the aforementioned difference.

[0025] Reference parameters can be pre-defined. These reference parameters can be pre-defined, for example, as initial values ​​for the model parameters. These initial values ​​can be pre-defined by the user of the device and / or stored in the device's memory unit. The dimension of the reference parameters preferably corresponds to the dimension of the model parameters. If the model parameters and reference parameters are understood as vectors respectively, then the simple difference between the model parameters and the reference parameters also gives a vector. Therefore, the simple difference between the finally determined model parameters and the reference parameters can be a vector composed of the differences. The squared difference of the simple difference is preferably the sum of the squared terms of the simple difference or the vector to which it belongs. The processor unit is preferably configured to determine the simple difference and / or the squared difference. The third weighting factor is generally chosen to be smaller than the first and second weighting factors, thus resulting in a weaker weight. Therefore, when optimizing the model parameters, this is actively caused: the model parameters are adjusted only to the extent that they are favorable to the target value. The third weighting factor helps avoid slow biases caused by irrelevant or redundant model parameters, which can also be called drifts.

[0026] Another advantageous configuration of the device is characterized in that the first sensor data represents parameters that act on the power transformer from the outside and are detected by the sensor. One of these parameters is, for example, the ambient temperature acting on the power transformer from the outside. Therefore, the ambient temperature can be represented by the first sensor data. The ambient temperature affects the characteristics within the transformer. Therefore, the temperature of the coolant (especially oil) within the transformer is affected by the external temperature. The temperature of the coolant in the power transformer may also depend on the power of solar radiation acting on the power transformer from the outside. Therefore, the corresponding solar radiation power can be represented by the first sensor data as a parameter acting on the power transformer from the outside and detectable by the sensor. The ambient temperature and solar radiation power are parameters that are not affected by the state of the power transformer. Other parameters acting on the power transformer from the outside are, for example, the current, voltage, or electrical power on the primary or secondary side of the power transformer. All the parameters mentioned above act on the power transformer from the outside and cause changes in the physical parameters within the power transformer. The first sensor data preferably represents only the parameters acting on the power transformer from the outside and detected by the sensor. If the model parameters are selected such that the simulation model correctly reflects the characteristics of the power transformer, new output data can be generated using the first sensor data as input data. The new output data represents physical parameters within the power transformer, which may not have been detected by a sensor or at least cannot be detected by a sensor at a reasonable cost.

[0027] Another advantageous configuration of the device is characterized in that the first sensor data represents at least two parameters from the following group: current on the primary side of the power transformer, current on the secondary side of the power transformer, electrical power on the primary side of the power transformer, electrical power on the secondary side of the power transformer, ambient temperature outside the power transformer, solar radiation power outside the power transformer, and wind speed outside the power transformer. The aforementioned group of possible physical parameters acting externally on the power transformer and detectable by the sensor can be a closed or open group. Therefore, it is possible that the sensor data represents only at least two parameters from the aforementioned group. Preferably, the parameters in the aforementioned group all act externally on the power transformer and thereby cause changes in physical parameters within the power transformer.

[0028] Another advantageous configuration of the device is characterized in that the second sensor data represents parameters detected within the power transformer. These detected parameters preferably relate to physical parameters of the components of the power transformer arranged within the power transformer, particularly within the power transformer housing. For example, these could be physical parameters of the coolant (especially oil), physical parameters of the windings on the primary and / or secondary sides, and / or mechanical parameters generated by mechanical interactions between components within the power transformer. Therefore, the parameters represented by the second sensor data are preferably a result of the parameters represented by the first sensor data acting on the power transformer and the system behavior of the power transformer itself. If, for example, the first sensor data represents the external temperature, then the external temperature can act on the power transformer, and the system behavior of the power transformer causes a change in the temperature of the oil (which, for example, forms the coolant). Therefore, the oil temperature can be represented by the second sensor data because the oil temperature is a resulting parameter occurring within the power transformer. The parameters represented by the second sensor data are detectable by sensors. This also applies to the example described above where the second sensor data represents the oil temperature.

[0029] Another advantageous configuration of the device is characterized in that the second sensor data represents at least one parameter from the following group: oil temperature, oil temperature on the upper side of the power transformer, oil temperature on the lower side of the power transformer, winding temperature on the primary winding, winding temperature on the secondary winding, heat flow from the primary winding of the power transformer to the oil in the power transformer, heat flow from the secondary winding of the power transformer to the oil in the power transformer, and oil viscosity of the power transformer oil. The aforementioned possible groups of physical parameters can be closed or open. Therefore, it is feasible for the second sensor data to represent only at least two parameters from the aforementioned groups. However, it is preferable for the second sensor data to represent only at least one parameter from the aforementioned groups. Preferably, the parameters in the aforementioned groups are influenced by the parameters represented by the first sensor data and by the system behavior of the power transformer.

[0030] For parameters represented by data from the first and second sensors, the device or system having the device may have sensors tailored to each represented parameter detected by the sensors. For example, to measure temperature, the device or system may include one or more sensors. For voltage, the device or system may include corresponding sensors for detecting voltage. A similar approach applies to other parameters actually represented by data from the first and second sensors. Two or more sensors may be constructed integrated with each other. The corresponding integrated sensors may then be constructed, for example, for detecting current, voltage, and electrical power.

[0031] Another advantageous configuration of the device is characterized in that the first and second sensor data of the measurement data set represent the respective parameter as a sequence of periodic, sensor-detected values ​​of the parameter. The detected values ​​preferably form the values ​​of the parameter that occur during a time period corresponding to a time window. The first and second sensor data of the measurement data set thus represent parameters from the same time period corresponding to their respective time windows.

[0032] Another advantageous configuration of the device is characterized in that it is configured such that in each time window, a set of measurement data is generated from first and second sensor data, the first and second sensor data representing a parameter belonging to a pre-given time period. The duration of each pre-given time period is preferably the same for each set of measurement data and therefore the same for each time window. Preferably, each time window is of equal length in time. For example, each time window may have a duration of 24 hours.

[0033] Another advantageous configuration of the device is characterized in that the predetermined time period is greater than the maximum time constant of the power transformer. Preferably, the predetermined time period is at least twice, and particularly at least ten times, the maximum time constant of the power transformer. This provides the advantage that, compared to the dynamics of the power transformer determined by its time constant, adjustments to the model parameters, particularly those made by repeatedly performing steps a) to c) in groups, are performed only slowly. This, in turn, allows for robust optimization of the model parameters against disturbance variables.

[0034] Another advantageous configuration of the device is characterized in that the pre-given time period is at least two hours, at least four hours, at least eight hours, at least sixteen hours, at least twenty-four hours, or at least two days. In practice, it has been determined that the pre-given time period for each time window should not be too short to ensure that model parameters are adjusted only slowly. However, excessively long time periods may lead to overly slow adjustments to model parameters. Therefore, the pre-given time period is preferably at least two hours, four hours, eight hours, or twenty-four hours. However, it is also preferred that the pre-given time period be less than ten days, preferably less than five days, and most preferably less than four days.

[0035] Another advantageous configuration of the device is characterized by the fact that the simulation model is constructed based on the IEEE Annex G-Model 1 for power transformers. This model is available from the publication of the IEEE Guide for Loading Mineral-Oil-Immersed Transformers and Step-Voltage-Regulators on the website www.ieee.org.

[0036] Another advantageous configuration of the device is characterized in that the simulation model is constructed based on the IEC model for power transformers. The IEC model is published by the International Electrotechnical Commission (www.iec.ch) and is preferably based on the IEC 60076-7 standard. Version 2.0 2018-01 of the document entitled "Loading guide for mineral-oil-immersed power transformers" or "Power transformers - Part 7: Loading guide for mineral-oil-immersed power transformers" has proven to be an advantageous version.

[0037] Another advantageous configuration of the device is characterized in that the device is configured to transmit an output signal representing model values ​​via the signal interface or another interface. The processor unit may preferably be configured to generate the output signal. This output signal can provide the model values ​​to a higher-level unit and / or a higher-level device. The output signal can be sent to the appropriate unit or device via the signal interface or other signal interface. The higher-level device may, for example, be a control console for multiple power transformers and devices according to the first aspect of the invention.

[0038] Another advantageous configuration of the device is characterized in that the processor unit is configured to: determine the last determined optimized parameters for each time window as end parameters for the corresponding time window; determine trend parameters by approximation with multiple end parameters; compare the trend parameters with their respective pre-given limit values; and determine a first warning state for the power transformer when one of the limit values ​​is exceeded. Steps a) to c) are repeated multiple times for each time window. The optimized parameters are determined as end parameters for the corresponding time window during the last repetition of step c). The processor unit is configured accordingly for this purpose. Because the processor unit is also configured to generate multiple successive time windows with their respective measurement data, multiple end parameters are also determined based on these multiple time windows. The processor unit is configured to perform approximations based on the multiple end parameters and thereby determine the trend parameters. These trend parameters may, for example, represent end parameters for future time windows. However, the end parameters may also represent deviations relative to the last end parameter or deviations of the average of the end parameters relative to future approximated end parameters. Furthermore, the processor unit is configured to compare the approximated trend parameters with their respective pre-given limit values. The trend parameters preferably have the same dimension as the optimization parameters. Therefore, the trend parameters can be formed as a vector, where each term of the vector includes a corresponding limit value. Thus, each trend parameter can be compared with its pre-given limit value. The processor unit is preferably configured accordingly for this purpose. The comparison of the trend parameters with the limit values ​​provides a feasible approach: when one or more trend parameters exceed one or more of their respective limit values, a warning state is determined. Therefore, the processor unit is configured to determine a first warning state for the power transformer when one or more limit values ​​are exceeded. This first warning state can indicate that the power transformer will be in a problematic or critical state in the future. Therefore, with the help of the trend parameters, it can be read whether the simulation model, and thus the actual system behavior of the power transformer, is likely to become a problematic or critical state in the future.

[0039] According to a second aspect of the invention, the aforementioned objective is achieved by a system according to the invention, the system having a plurality of devices. Each device is constructed according to the first aspect of the invention and / or one of its advantageous configurations. Therefore, for each device of the system, reference is made similarly to the advantageous descriptions, preferred features, technical effects and / or advantages that have been explained with respect to the first aspect of the invention or one of its advantageous configurations.

[0040] A system configuration according to a second aspect of the invention is used to compare model parameters, each determined by the device, with each other, and thereby determine model parameter differences. The system is further configured to compare the model difference parameters with pre-defined threshold values, and, if one of the threshold values ​​is exceeded, identify the model parameter that caused the exceedance and the associated power transformer, and determine a second warning state for the identified power transformer.

[0041] The system preferably also includes a plurality of power transformers. Each power transformer is associated with one of the aforementioned devices in the system, such that the device associated with a corresponding power transformer is configured to receive first and second sensor data via measurement signals, the first and second sensor data representing parameters detected by sensors in or on the corresponding associated power transformer. Furthermore, it is preferably specified that the power transformers have the same type and / or are constructed identically. This allows for comparison of the system behavior of the power transformers with each other. The simulation models of the devices are adjusted by their respective model parameters so that the simulation models can simulate the system behavior of the respective associated power transformers. Therefore, comparing the model parameters of the plurality of devices can yield results similar to comparing the system behavior of the plurality of power transformers in the system. Thus, differences in model parameters can be used to determine whether the system behavior of the power transformers is strongly different. If one of the power transformers has a strongly deviated system behavior, particularly a strongly deviated system behavior from the average of the other power transformers, then the model parameters of the simulation model used for that deviating power transformer also deviate from the other model parameters of the other devices by a greater difference in model parameter differences. If the difference in model parameters exceeds a predefined threshold, this indicates that the model parameters causing the exceedance correspond to a simulation model of a device that receives measurement signals from a power transformer in a problematic or critical state. Therefore, the power transformer for which a second warning state needs to be determined can be identified via the model parameters. Thus, the processor unit is preferably configured to determine the model parameters causing the exceedance of the threshold for the simulation model, and based on the model parameters and the corresponding simulation model, identify the corresponding device and power transformer, the identified device receiving measurement signals representing first and second sensor data from the corresponding power transformer, the first and second sensor data representing parameters detected by sensors of the identified power transformer, respectively. Preferably, the system is also configured to determine a second warning state for the identified power transformer. If a second warning state is determined, subsequent actions can be taken to inspect the identified power transformer and, if necessary, introduce maintenance measures.

[0042] Other features, advantages, and possible applications of the invention can be derived from the following description and drawings of the embodiments. Here, all described and / or illustrated features, in themselves and in any combination thereof, also constitute the technical solutions of the invention, independent of their composition in the various claims or their references. Furthermore, in the drawings, the same reference numerals denote the same or similar objects. Attached Figure Description

[0043] Figure 1 A schematic diagram illustrates an advantageous implementation of a system with a power transformer and an advantageous implementation of a device.

[0044] Figure 2 An exemplary set of steps a) to c) is illustrated in the diagram. Detailed Implementation

[0045] Figure 1 An advantageous embodiment of system 1 is shown. System 1 includes a power transformer 10 and a device 2.

[0046] When the advantageous features, technical effects, and / or benefits of device 2 are described below, these descriptions may also apply to embodiments in which device 2 is not part of system 1, and / or to device 2 that is part of system 1. The power transformer 10 may constitute another part of system 1.

[0047] The power transformer 10 of System 1 has a housing 38 in which a primary-side coil group 20 and a secondary-side coil group 28 are arranged. The primary-side coil group 20 is electromagnetically coupled to the secondary-side coil group 28 via a magnetic core, particularly a toroidal core. Furthermore, the power transformer 10 has terminals 18 on the primary side, forming an electrical connection with the primary-side coil group 20. Similarly, terminals 30 are provided on the secondary side, electrically connected to the secondary-side coil group 28. Each of these two coil groups 20, 28 has multiple windings. During operation of the power transformer 10, a voltage drop occurs between the terminals 18 on the primary side, which in turn causes current to flow through the primary-side terminals 18 and the windings of the coil group 20. This consumes electrical power on the primary side. A voltage drop also occurs between the terminals 30 on the secondary side, and current flows through the secondary-side terminals 30 and through the windings of the coil group 28. Therefore, electrical power is provided on the secondary side. The power transformer 10 is therefore used to transfer and convert electrical power from the primary side to the secondary side. Heat is generated here, which can be received by a coolant 40, preferably formed of oil. The coolant 40 is also contained within the interior space of the housing 38 and can receive and conduct heat generated on the windings of the coil groups 20, 28 to the cooler 26. The cooler 26 is preferably an oil cooler for cooling the oil (which preferably forms the coolant 40).

[0048] The coolant 40 is heated by heat flow between the windings of the coil groups 20, 28 on the primary or secondary side. Therefore, particularly hot spots may appear in the coolant 40, especially on the primary side coils, where the coolant is exceptionally hot. Therefore, it is preferable to attach a winding temperature sensor 22 to or within the windings to determine the temperature on or within the windings. This temperature is also referred to as the hot spot temperature. The winding temperature sensor 22 is preferably constructed to detect the temperature on or within the windings. Figure 1 As can be schematically and exemplaryly seen, the winding temperature sensor 22 can, for example, be fixed to the winding of the coil assembly 20 on the primary side. The winding temperature sensor 22 can be connected to the detection unit 32 of the power transformer 10 via a signal connection device. Preferably, the power transformer 10 also has an additional sensor coupled to the detection unit 32. (As shown by...) Figure 1 As can be seen by way of example, the power transformer 10 also preferably has an oil temperature sensor 24. The oil temperature sensor 24 is arranged within the housing 38 of the power transformer 10, and more specifically, the oil temperature sensor is preferably arranged such that it detects the temperature of the coolant 40 in the upper region of the internal space of the housing 38. Thus, the oil temperature sensor 24 can be arranged, for example, to detect the so-called upper oil temperature. The oil temperature sensor 24 is coupled to the detection unit 32 via an associated signal connection device. Furthermore, it is preferably specified that the power transformer 10 has an ambient temperature sensor 14. The ambient temperature sensor 14 is not necessarily required to be fixed to the housing 38 of the power transformer 10, however, this is feasible. Rather, it is preferably specified that the ambient temperature sensor 14 forms part of the power transformer 10 without thermal contact with the housing 38. The ambient temperature sensor 14 can therefore be at least associated with the power transformer 10 and arranged near the housing 38. The ambient temperature sensor 14 is coupled to the detection unit 32 via an associated signal connection device. Furthermore, the power transformer 10 preferably includes a current sensor 16 for detecting the current flowing through the terminals 18 of the power transformer 10. Figure 1In this transformer, current sensor 16 is arranged, for example, on one of the terminals 18 on the primary side, to detect the current flowing through terminal 18. Current sensor 16 is coupled to detection unit 32 via an associated signal connection device. Ambient temperature sensor 14 and current sensor 16 respectively detect physical parameters or characteristics acting on the power transformer 10 from the outside. Winding temperature sensor 22 and oil temperature sensor 24 respectively detect physical parameters or characteristics existing within the power transformer 10, particularly within the housing 38 of the power transformer 10. Therefore, it is stipulated that, for example, ambient temperature sensor 14 and current sensor 16 generate first sensor data, and winding temperature sensor 22 and oil temperature sensor 24 generate second sensor data. Sensor data from sensors 14, 16, 22, and 24 can be transmitted to the detection unit via corresponding signal connection devices. Detection unit 32 can be configured to generate measurement signals representing the first and second sensor data. Detection unit 32 can send the measurement signals to signal interface 4 of device 2 via signal connection device 36. In this manner, device 2 is provided with first sensor data and second sensor data, the first sensor data and the second sensor data respectively representing at least one parameter detected by a sensor on or in power transformer 10.

[0049] While a power transformer 10 with multiple sensors is desirable, in practice, not all theoretically available and feasible sensors are often installed. In particular, it is possible that the power transformer 10 does not have a winding temperature sensor 22. For example, Figure 1 This is the case with the winding of coil group 28 on the secondary side 30. For further illustration, this is merely an example of... Figure 1 The power transformer 10 shown does not have a winding temperature sensor 22, although the temperature of the winding is of particular interest.

[0050] Starting from the previously stated assumption that the power transformer 10 does not have a winding temperature sensor 22, it is to be determined that the second sensor data cannot represent the winding temperature. The second sensor data may be fed solely by, for example, the oil temperature sensor 24. In other words, the second sensor data may, for example, represent the oil temperature of the coolant 40, which is oil, in the upper internal region of the housing 38. In order to still obtain conclusions about the temperature of, for example, the winding of the coil group 20 on the primary side, the device 2 is simplified to appear to be constructed for determining a so-called model value, which, for example, represents the temperature of the winding of the coil group 20 on the primary side, although this temperature is not detected by a sensor.

[0051] Device 2 has a signal interface 4, a processor unit 6, and a memory unit 8, wherein the memory unit 8 is preferably a memory unit for storing data. The processor unit 6 is configured to perform mathematical operations and / or signal processing. For this purpose, the processor unit 6 may include a microprocessor. The signal interface 4 is configured to receive measurement signals representing first sensor data and second sensor data. The signal interface 4 can receive the measurement signals from the detection unit 32 and via a signal connection device 36. The signal interface 4 is preferably coupled to the processor unit 6, thereby providing the processor unit 6 with the first sensor data and the second sensor data. The signal interface 4 and the processor unit 6 can be integrally constructed.

[0052] Processor unit 6 is configured to generate, in temporally successive time windows, sets of measurement data formed by first and second sensor data received in the respective time windows. The duration of a time window is preferably 24 hours. Furthermore, it is preferably specified that a new time window with an associated set of measurement data is generated every four hours. Therefore, a set of measurement data associated with a time window can partially include the same first and second sensor data. Furthermore, in the previously mentioned example, if two successive time windows are considered, each time window can overlap by 16 hours.

[0053] A mathematical simulation model of the power transformer 10 is stored in memory unit 8. Processor unit 6 is configured to load and execute the simulation model. The simulation model is constructed such that it represents the transmission characteristics of the power transformer 10. Because the simulation model is a mathematical model, input data can be processed by the simulation model during execution, resulting in output data. The output data therefore depends on the input data. However, the transmission characteristics of the simulation model are variable, more specifically, they are variable due to the associated model parameters. Therefore, the output data depends not only on the model parameters of the simulation model but also on the input data used to execute the simulation model. Processor unit 6 is configured to execute the simulation model with the input data and the associated model parameters for the simulation model, thereby generating output data. At least the model parameters used as starting parameters for executing the simulation model can be stored in memory unit 8 and / or loaded by processor unit 6. Furthermore, the simulation model is configured to generate not only output data but also model values ​​during execution, more specifically, model values ​​for the temperature within the power transformer 10 based on the input data and model parameters used during execution. Therefore, the model value can also be the result of executing the simulation model. Processor unit 6 is configured to execute the simulation model accordingly to determine the model value. The simulation model can, for example, be configured such that the model value represents the temperature on the winding of one of the coil groups 20, 28. Preferably, the simulation model is configured such that the model value represents the temperature on the winding of coil group 20 on the primary side 20. Depending on the configuration of the simulation model and preferably also based on the requirement that the simulation model should be used to determine which temperature in the power transformer 10, the model value can also reflect other temperatures within the power transformer 10. The model value can, for example, represent the upper oil temperature or the temperature of the coolant 40 in the cooler 26.

[0054] However, since the model values ​​determined by the simulation model depend not only on the input data but also on the model parameters used in the simulation model, it is desirable to advantageously select the model parameters so that the simulation model accurately reflects the transmission characteristics of the power transformer 10. Therefore, it is stipulated that repeated operations in groups, for example in... Figure 2The group 12 shown consists of steps a), b), and c), which determines the model parameters used in the simulation model, allowing the power transformer 10 to be simulated with particular accuracy. When repeated in groups, each group 12 consisting of steps a) to c) is executed completely before the same group 12 consisting of steps a) to c) is updated and thus repeated. When repeated in groups, steps a) to c) of the corresponding group 12 are not executed in parallel, but sequentially. Nor does the execution of individual steps of multiple parallel groups 12 occur. That is, each group 12 is executed sequentially, and within each group 12, steps a) to c) are executed sequentially. This results in repeated groups.

[0055] Processor unit 6 is configured to determine output data in step a) by executing a simulation model with input data and model parameters. The input data is determined by first sensor data belonging to a measurement data group corresponding to a specific time window. The model parameters are determined by pre-given reference parameters during the first execution of the simulation model for the corresponding time window, and by individually optimized parameters in other cases. The reference parameters can be initial parameters used for the model parameters. This enables the execution of the simulation model. The reference data can be stored in memory unit 8 and / or loaded from memory unit 8 by the processor unit.

[0056] exist Figure 1 In the example shown, the first sensor data is formed from the sensor data of ambient temperature sensor 14 and current sensor 16. Therefore, the first sensor data in this case can represent the ambient temperature of the power transformer 10 and the current on the primary side of the power transformer 10. Output data is generated by executing a simulation model with the first sensor data as input data and reference parameters as model parameters. The simulation model can be constructed such that the output data represents parameters of the simulated power transformer 10, which correspond to the parameters detected by the sensors in the second sensor data. Therefore, in Figure 1 In the example shown, output data can be generated by executing a simulation model, which represents the simulated upper oil temperature of the coolant 40 in the internal space of the housing 38 of the power transformer 10.

[0057] As specified in step b), a target value is determined using processor unit 6. This target value includes at least one squared error between second sensor data weighted by a first weighting factor and the output data previously determined in step a). Regarding Figure 1In the example shown, the second sensor data represents the upper oil temperature detected by oil temperature sensor 24. The output data determined in step a) represents the simulated upper oil temperature for the coolant 40 within the internal space of the housing 38 of the power transformer 10. Therefore, the error between the second sensor data and the output data can be either positive or negative, depending on the corresponding detected value. A squared error is used to avoid the presence of a mathematical sign for the error. The first weighting factor can be selected to calculate the target value accordingly.

[0058] An example mathematical function for determining the target value is shown below:

[0059]

[0060] Here, the objective value E(p) depends on three components. The first component, with the first integral sign, is represented by the first weighting factor c. a The weighted squared error between the second sensor data m and the output data s determined in step a). Each of the other two components of the target value begins with a summation sign. The second weighting factor c is considered in the second component. d And in the third component, the third weighting factor c is considered. r .

[0061] Preferably, device 2 is configured to determine the target value E in step b) from the start of execution for the second target window, such that the target value E also includes a second weighting factor c. d The weighted model parameters p in the final determination i And the final optimization parameter p used for the previous time window i,pre The difference of squares between (p) i -p i,pre ) 2 The second component of the objective value E can be determined by selecting a second weighting factor c. d This helps prevent rapid changes in the parameters of each model. As a result, the simulation model is robust to external disturbance parameters.

[0062] Preferably, device 2 is configured to determine the target value E in step b) such that the target value E also includes a third weighting factor c. r The weighted model parameters p in the final determination i With reference parameter p i,o The difference of squares between (p) i -p i,o ) 2 By relative to the first weighting factor c a Second weighting factor c d Appropriate selection of the third weighting factor c rThe third component of the objective value E can be chosen to be smaller in order to give other components greater weight. However, by appropriately selecting the third weighting factor c... r This can prevent the initial reference parameter p from being in the wrong place. i,o With the corresponding current model parameter p i A strong deviation occurs between them. This is based on the following: the initial reference parameter p. i,o This at least approximately allows for a sufficiently good simulation of the transmission characteristics of the power transformer 10 using a simulation model.

[0063] Furthermore, processor unit 6 is configured to determine, in step c), optimization parameters as new model parameters for the simulation model based on at least one target value determined in a corresponding time window from step b). Processor unit 6 can, for example, be configured to execute an optimization method to determine the optimization parameters. Here, the optimization method employs not only the target value determined in the previous step b), but also the optimization parameters determined in the previous step c), so as to determine new optimization parameters as model parameters for the simulation model based on the optimization method for the corresponding current determination of optimization parameters, which result in a smaller target value in the next step b). Therefore, the optimization method can be designed to keep the target value as small as possible and / or minimize the target value in step c). This is meaningful because the target value contains a weighted sum of squared errors or differences of squares, which should be kept as small as possible. Optimization methods are known in principle in the prior art and can be used when performing group 12 of steps a) to c). Preferably, processor unit 6 is configured to repeat group 12 of steps a) to c) between 10 and 1000 times. In practice, it has been determined that, during the final execution of step c), the optimized parameters can be set as new model parameters, which allow for a particularly advantageous simulation of the transmission characteristics of the power transformer 10. Furthermore, model values ​​for the temperature within the power transformer 10 can be determined using these model parameters via a simulation model, preferably not detected by sensors. Therefore, the processor unit 6 is configured to determine model values ​​for each time window by executing a simulation model with input data and model parameters, the input data being determined by first sensor data belonging to a measurement data set for the corresponding time window, and the model parameters for the corresponding time window being determined by the final optimized parameters. This allows for the determination of particularly accurate model values ​​for each time window, which thus also reflect or represent the temperature within the power transformer 10 with particular precision. The temperature can, for example, be the temperature of one winding in one of the windings of one of the coil groups 20, 28. Preferably, the device 2 has another interface 34 coupled to the processor unit 6. Furthermore, the device 2 can be configured to transmit an output signal representing the model value, particularly the final determined model value, via the other interface 34.

[0064] Additionally, it should be noted that the expression "having" does not exclude any other elements or steps, and "a" or "an" does not exclude a plurality. It should also be noted that the features described with reference to one embodiment of the above embodiments can also be used in combination with other features of other above embodiments. Reference numerals in the claims should not be considered limiting.

[0065] List of reference numerals

[0066] 1 system

[0067] 2 equipment

[0068] 4 signal interfaces

[0069] 6 processor units

[0070] 8 memory units

[0071] 10 power transformers

[0072] 12 groups

[0073] 14 Ambient Temperature Sensor

[0074] 16 Current Sensors

[0075] 18 Primary side terminals

[0076] 20 primary-side coil groups

[0077] 22 Winding Temperature Sensor

[0078] 24 oil temperature sensor

[0079] 26 Cooler

[0080] 28 secondary-side coil groups

[0081] 30 Secondary side terminals

[0082] 32 detection units

[0083] 34. Additional signal interfaces

[0084] 36 signal connection devices

[0085] 38 housing

[0086] 40 coolant

Claims

1. A device (2), said device having: - Signal interface (4). - Processor unit (6), and - Memory unit (8) in, The signal interface (4) is configured to receive measurement signals representing data from the first sensor and data from the second sensor. The first sensor data and the second sensor data each represent at least one parameter detected by a sensor in or on the power transformer (10). The processor unit (6) is configured to generate measurement data sets formed by first sensor data and second sensor data received in the corresponding time windows in successive time windows. The memory unit (8) stores a simulation model of the power transformer (10), which reflects the transmission characteristics from input data to output data based on the model parameters of the simulation model. The processor unit (6) is configured to repeatedly execute groups (12) of the following steps a) to c) for each time window: a) The output data is determined by executing the simulation model with input data and model parameters, wherein the input data is determined by first sensor data belonging to a measurement data group of the corresponding time window, and the model parameters are determined by pre-given reference parameters when the simulation model is executed for the first time for the corresponding time window, and the model parameters are determined by the respective last determined optimization parameters in other cases; b) Determine a target value, said target value comprising at least one squared error between second sensor data weighted by a first weighting factor and previously determined output data within a measurement data set belonging to a corresponding time window; and c) Based on at least one target value determined within the corresponding time window, determine optimization parameters as new model parameters for the simulation model, wherein the processor unit determines the optimization parameters according to one or more target values ​​such that a smaller target value is determined when step b) is executed again. The simulation model also reflects model values ​​for the temperature within the power transformer (10) based on input data and model parameters, and The processor unit (6) is configured to determine model values ​​for each time window by executing a simulation model with input data and model parameters, wherein the input data is determined by first sensor data belonging to the measurement data group of the corresponding time window, and the model parameters are determined for the corresponding time window by the respective final determined optimization parameters.

2. The device (2) according to claim 1, characterized in that, The temperature of the power transformer (10) reflected by the model value includes either the hot spot temperature, the upper oil temperature, or the bottom oil temperature of the power transformer (10).

3. The device (2) according to claim 1 or 2, characterized in that, The number of times to repeat steps a) to c) of group (12) is between 10 and 1000.

4. The device (2) according to claim 1 or 2, characterized in that, The device (2) is configured to, starting from the execution for the second time window, determine the target value in step b) such that the target value also includes the squared difference between the last determined model parameter weighted by the second weighting factor and the last determined optimization parameter for the previous time window.

5. The device (2) according to claim 1 or 2, characterized in that, The device (2) is configured to determine the target value in step b) such that the target value also includes the squared difference between the final determined model parameters and the reference parameters, weighted by a third weighting factor.

6. The device (2) according to claim 1 or 2, characterized in that, The first sensor data represents the parameters that are applied to the power transformer (10) from the outside and detected by the sensor.

7. The device (2) according to claim 1 or 2, characterized in that, The first sensor data represents at least two of the following parameter groups (12): current on the primary side of the power transformer (10), current on the secondary side of the power transformer (10), electrical power on the primary side of the power transformer (10), electrical power on the secondary side of the power transformer (10), ambient temperature (14) outside the power transformer (10), solar radiation power outside the power transformer (10), and wind speed outside the power transformer (10).

8. The device (2) according to claim 1 or 2, characterized in that, The second sensor data represents the parameters detected by the sensor inside the power transformer (10).

9. The device (2) according to claim 1 or 2, characterized in that, The second sensor data represents at least one of the following parameter groups (12): oil temperature, oil temperature on the upper side of the power transformer (10), oil temperature on the lower side of the power transformer (10), winding temperature on the primary winding, winding temperature on the secondary winding, heat flow from the primary winding of the power transformer (10) into the oil of the power transformer (10), heat flow from the secondary winding of the power transformer (10) into the oil of the power transformer (10), and oil viscosity of the power transformer (10).

10. The device (2) according to claim 1 or 2, characterized in that, The first and second sensor data of the measurement data set respectively represent the parameters to which they belong as a sequence of periodic and sensor-detected values ​​of the parameters.

11. The device (2) according to claim 1, characterized in that, The device (2) is configured such that in each time window a group of measurement data is generated, which consists of first sensor data and second sensor data, representing parameters belonging to a given time period.

12. The device (2) according to claim 11, characterized in that, The pre-given time period is greater than the maximum time constant of the power transformer (10).

13. The device (2) according to claim 11 or 12, characterized in that, The pre-given time period is at least two hours, at least four hours, at least eight hours, at least sixteen hours, at least twenty-four hours, or at least two days.

14. The device (2) according to claim 1 or 2, characterized in that, The simulation model is constructed based on the G model in the IEEE Appendix for power transformers (10).

15. The device (2) according to claim 1 or 2, characterized in that, The simulation model is constructed based on the IEC model used for the power transformer (10).

16. The device (2) according to claim 1 or 2, characterized in that, The device (2) is configured to transmit an output signal representing a model value via the signal interface (4) or another interface (34).

17. The device (2) according to claim 1 or 2, characterized in that, The processor unit (6) is configured to: determine the last determined optimization parameter for a time window as the end parameter for the corresponding time window; determine a trend parameter by approximating with multiple end parameters; compare the trend parameter with a pre-given limit value; and determine a first warning state for the power transformer (10) when one of the limit values ​​is exceeded.

18. A system (1), said system having: Multiple devices (2), each of which is constructed according to any one of claims 1 to 17, in, The system (1) is configured to compare the model parameters determined by each device (2) with each other, and thereby determine the model parameter differences, and The system (1) is configured to: compare the model parameter difference with the corresponding pre-given limit value; and when one of the limit values ​​is exceeded, identify the model parameter that caused the exceedance and the corresponding power transformer (10); and determine a second warning state for the identified power transformer (10).

Images