Apparatus, system and method for identifying an error signal window in a measurement signal

Abstract

The present invention relates to an apparatus 2 for detecting error signal windows indicative of a signal error in a measurement signal of an electrical installation 8 for transmitting electric power. In this case, a measurement signal detected by a sensor during power transmission is analyzed by a model at least during an initialization phase 12 to identify a plurality of initial coefficient tuples 14. Each of the plurality of initial coefficient tuples 14 represents a coefficient of the model. From this, a noise tuple 16 is detected as a predicted value by the apparatus 2. In a testing phase 18 following the initialization phase 12, further signal windows of the measurement signal are analyzed by the model using the apparatus 2. The resulting coefficients generate one measurement tuple 20 for each signal window. A distance D from each measurement tuple to the noise tuple is used to check whether the analyzed signal error represents a signal window. If the analyzed signal error represents a signal window, the respective signal window is determined to be an error signal window. Furthermore, the present invention relates to a system 26 for transmitting electric power and a method for operating the apparatus 2. In this case, the system 26 comprises the apparatus 2.

Description

[Technical Field]

[0001] electricity equipment This can be configured to transmit power. Such electrical equipment could be, for example, an electrical cable having multiple wires used to transmit power. Another electrical piece of equipment could be, for example, a transformer. of Similarly, they may be used to transmit power, or to quantify current and voltage, for example, in electrical converters or elevated lines. In particular, electrical installations are cables for high-voltage DC or high-voltage AC power transmission. When high voltage is used to transmit power, partial discharge (TE) may occur. Partial discharge refers to a locally limited discharge that bypasses the insulation between multiple conductors of the cable mentioned above. Partial discharge can occur, for example, due to defects in the conductors and / or non-uniformity in the cable insulation. When a partial discharge occurs, the occurrence of the partial discharge generates a signal indicating the partial discharge, which has a corresponding signal pattern. This signal is transmitted by the electrical installation. In this case, however, This signal teeth, Department Is it the point of partial discharge? raden Up to the connection point of the air equipment transmission Therefore, the signal caused by the partial discharge is detected at the connection point of the electrical equipment before it is detected. To some extent This will proceed along the transmission path. electric The electromagnetic induction elements, capacitance elements and / or resistive elements of the gas equipment are Department Discharge can cause signal attenuation and / or distortion. Therefore, in many cases, Department Discharge ÷ Of the signals, Damping So Only the signal and / or distorted signal, contact It can be detected at the continuation point. Department The signal caused by minute discharge is also called the error signal component of the measurement signal, or simply the error signal. That is separate. Error signal but , electric Another air equipment defectThis can also be caused by (Fehler).

[0002] defect However, in order to determine whether or not it is caused by electrical equipment during power transmission to In reality, for many people, electric Detectable measurement signals at the connection points of the ventilation equipment The Recorded, defect During power transmission ni To determine whether the issue is occurring in the equipment, manual inspection, such as with an oscilloscope, is performed. However, this method is less robust than possible defect assessment and is also extremely time-consuming. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] German patent number 102018126743 [Non-patent literature]

[0004] [Non-Patent Document 1] N-Dimensional cumulative function, and other useful facts about gaussians and normal densities von Michaeol Bensimhoun(Jerusalem,06 / 2009) [Overview of the project] [Problems that the invention aims to solve]

[0005] The problem that the present invention will address is the power transmission by electrical equipment defect The objective is to provide an apparatus, system, and method that enables efficient and error-robust detection of [unspecified problem]. [Means for solving the problem]

[0006] According to a first aspect of the present invention, the above problem is solved by an apparatus having the features described in claim 1. That is, an apparatus is provided that has an input signal interface for receiving a digital measurement signal indicating an analog signal detected at a connection point of electrical equipment. This Digital measurement The signal is , check the emitted signal no sa a sequence consisting of sampled values. Further, the apparatus has a processor unit. This processor unit The system is configured to assign one signal window to each sampling value of the measurement signal. Each signal window consists of a portion of a sequence comprising each sampling value of the measurement signal and a predetermined number of sampling values ​​that precede them in time. In the initialization stage In initializes M signal windows as initialization windows setting and selects K noise windows from these M initialization windows setting and analyzes each noise window by a predetermined X - order model So and constructs an initial coefficient tuple assigned to each noise window 、 and determines the number The person in charge of and from the initial coefficient tuples of the noise windows rough forms a noise tuple Predicted values as It is configured to make a decision. where X is an even number from 1 to 5 .difference Furthermore, in the inspection stage In the processor unit sets a plurality of signal windows as measurement windows setting and analyzes each measurement window by a predetermined X - order model So and constructs a measurement tuple assigned to each measurement window 、 and determines the number The person in charge of and for each measurement window So detects the distance of the measurement tuple with respect to the noise tuple rough and from a number of measurement windows The measurement forms an error signal window, where each error signal window indicates a signal error of the measurement signal. In this case, the signal error is the error signal component of the measurement signal Multiple and The system is configured to identify measurement windows where the distance between the measurement tuples belonging to the measurement window and the noise tuple is greater than a default first limit value G as an error signal window. and be .

[0007] When power is transmitted by electrical equipment, signals indicating the voltage, power, current, and / or other values ​​indicating power transmission may be detected at the connection point. Inspect The output signals indicate, in particular, the power transmitted by the power system, the corresponding current, the corresponding voltage, and / or the time progression of the corresponding power. The digital measurement signals provided to the device's input signal interface indicate the signals detected at the connection points of the electrical installation. De A digital measurement signal consists of a sequence of multiple sampled values ​​of the detected signal. De Digital measurement signals are simply called measurement signals.

[0008] However, in many cases, measurement A constant signal indicates not only the power transmitted by the electrical equipment (or the corresponding current, voltage, and / or power), but also noise caused by the surrounding conditions and the power transmission by the electrical equipment. defect Error messages that may occur The number show.

[0009] For the configuration of the device's processor section, noise and electrical equipment during power transmission are important. Defects in equipment Caused by Measurement It is different from the error signal component of a constant signal. but , basic A basic premise Therefore, in the initialization phase, first, electric It is desirable that a characteristic initial coefficient tuple for noise generated during power transmission be obtained by a mathematical model. This model is used to model the transmission function or transmission path of electrical equipment during power transmission. The model can be adapted to the specific transmission characteristics of the electrical equipment depending on the fitting of the coefficients. In particular, the processor is configured to fit the coefficients of the model based on a signal window so that the model approximates the actual transmission function of the electrical equipment as accurately as possible.

[0010] Furthermore, the processor unit uses M signal windows as initialization windows during the initialization phase. setting It is configured to do so. In this case, these M signal windows but These are assigned to M corresponding sampling values. These M sampling values ​​are sequential to the measurement signal. M These are individual sample values. measurement A constant signal may have useful components, error signal components, and noise components. In this case, Yes The components used represent power transmission, corresponding voltage, corresponding current, and / or corresponding power during power transmission. workman The R signal component is, electric Possible to occur during force transmission defect Caused by It can be. Miscellaneous The sound components are, electric This indicates noise generated during force transmission and / or during the measurement-technical detection of the measurement signal. vinegar Therefore, preferably, P The losser section selects K noise windows from M initialization windows. setting It is configured to be this way. ni P The lossr unit processes each of the M initialization windows, for example, the corresponding spectrum Based on the rule , a portion of the measurement signal indicated by each initialization window, zero of a portion of the measurement signal indicated by each initialization window point passing The number of and / or other analytical parameters can be identified. therefore Which of the M initialization windows is the noise window? setting For example, what should be done? vinegar Pectol, Faith Energy of the number, Passing the zero point Number and / or minutes Based on the analysis parameters, the processor unit makes the determination. In this way, the processor unit can identify, for example, K noise windows. In each noise window, The noise component in each instance is very large or dominant. Assumed .

[0011] The processor unit is configured to analyze each noise window using a predetermined X-order model and determine the corresponding coefficients based on that analysis. Therefore, TheseThe coefficients are those of this default X-th order model. These coefficients may also be called model coefficients. Mo Dell is a mathematical model. Furthermore, this model can be configured to show a subsignal composed of a signal window by fitting the corresponding coefficients. minutes Analysis is, P Since the lossr unit performs this for each noise window, So In order to show the subsignals indicated by each noise window as accurately as possible, the model is designed to: Responsible person The numbers are adapted each time. In particular, X is integer 2 Therefore, the model The person in charge To determine the number, it is possible to analyze each noise window or the subsignal representing each noise window using a default model very efficiently and quickly. This assigns a coefficient to each noise window, which is commonly called an initial coefficient tuple. Thus, each noise window is assigned exactly one dedicated initial coefficient tuple. be .

[0012] The processor unit takes the corresponding K initial coefficient tuples and uses the predicted values ​​as a noise tuple. decision It is configured to do so. In this case, rough A sound tuple is, first time The arithmetic mean of the initial coefficient tuples is also acceptable. Each initial coefficient tuple is composed of multiple coefficients. Rube If it is composed of a cul, the noise tuple is, for example, the corresponding Rube Add the ctol, Addition The result can be calculated by subsequently dividing by the number of these initial coefficient tuples. The processor unit, calculation It can be configured for output. Therefore, rough The sound tuple can construct predicted values ​​for these initial coefficient tuples. roughThe tone tuple can also be understood as a noise center (Rauschzentrum) for these initial coefficient tuples, because these initial coefficient tuples are distributed around this noise center in geometric space.

[0013] During the initialization phase, M signal windows are initialized by the processor as initialization windows. setting The inspection phase follows this initialization phase. Therefore, P The Rossessa section is Multiple Using a numerical signal window as the measurement window setting It is configured to do so. In particular, these measurement windows are different from the measurement windows identified in advance during the preceding initialization phase. In particular, these measurement windows are first time Periodization window More than the number Significantly It is large. The basic idea of ​​this invention is the inspection stage. Measurement This involves identifying fixed windows as error signal windows, and these error signal windows represent a portion of the measurement signal indicated by each signal window. In that Rezo Reno Signal error 。 Therefore, the number of measurement windows is greater than the number of initialization windows. Significantly Bigger is more important.

[0014] The processor unit determines the corresponding coefficient. Therefore Furthermore, the system is configured to analyze each measurement window using a default X-order model during the testing phase. Therefore, each measurement window is analyzed using the same default X-order model that was already used in the initialization phase to analyze the noise window. Thus, measurement The analysis of fixed window is, P This can be performed in a similar manner to the noise window analysis performed by the lossr unit. Therefore, So Each measurement window ga pu The analysis is performed by the reduction unit using the Xth-order model. The coefficients of the X-th order model are determined. The coefficients determined for each measurement window are called measurement tuples. So Each is assigned to a measurement window. be . Already Fixed model two In the case of the next model, there are two coefficients each. Measure This is determined for each fixed window. However, if the model is a higher-order model, a number of coefficients corresponding to that order are determined for each measurement window. measurement A definite tuple measurement When detected within a fixed window, this measurement tuple is relative to the noise tuple in geometric space. first time Phase In this case It has a predetermined distance. The dimension of the geometric space may correspond to the degree of the model. If the default model is of a higher degree, particularly a cubic, quartic, or quintic model, the distance may be a distance defined in the corresponding space. P The lossr section processes each measurement tuple for each noise tuple. distance It is configured to calculate the distance. Therefore, depending on the number of measurement windows, there are a corresponding number of measurement tuples and a corresponding number of distances. P Detected by the losser. If the measurement tuple is at a short distance from the noise tuple, the corresponding measurement window is likely to show a portion of the measurement signal with a large noise component. However, if the distance between the measurement tuple and the noise tuple is greater, especially greater than the default first limit value, the corresponding measurement window is likely to show a portion of the measurement signal with a signal error. This is because, Signal error , This measurement window During analysis, This leads to the calculation of different coefficients for this default model. Therefore, the measurement tuple resulting from these other coefficients is For noise tuples, at a greater distance It will be positioned. Therefore, measurement definite tuple and rough Depending on the distance between each sound tuple, From multiple measurement windows Measurement signal The belief Measurement window showing error code of As an error signal window , non Always efficiently identify It is possible .

[0015] Therefore, according to the above explanation, for each measurement window, for the noise tuple Measurement tuple It has been proposed that the processor be configured to detect the distance between the noise tuple and the noise tuple, and from these measurement windows, identify measurement windows where the corresponding measurement tuple is at a distance greater than a predetermined first limit value relative to the noise tuple as error signal windows. This ensures that each error signal window represents a portion of the measurement signal that has a signal error.

[0016] The configuration of the processor section for identifying the error signal window mentioned last allows for the rapid and robust identification of the error signal window in which a signal error is expected. These error signal windows can be evaluated for further analysis. In further analysis, in particular, Electrical equipment During power transmission Departure The signal that is being produced error The types and characteristics and / or types and / or characteristics of defects may be identified. P The transducer section can be appropriately configured for this purpose.

[0017] Furthermore, the configuration of this processor unit has the advantage that manual analysis of the entire measurement signal can be omitted. Rather, further evaluation to measure It concentrates on a predetermined portion of the constant signal, i.e., the error signal window. do This is made possible by the configuration of the processor unit. This allows power transmission to defect The overall effort required to detect and / or analyze them can be reduced.

[0018] As already mentioned above, The processor unit is configured to assign one signal window to each sampling value of the measurement signal. In this way This results in multiple signal windows. sa Pumping value Move each time Move Signal windowIt was also mentioned that a sequence occurs. Each signal window may contain, for example, at least 16, at least 32, or at least 64 sampling values. Preferably, each signal window has, for example, 128 sampling values. The initialization phase may be very short. For this reason, P The losser unit may be configured to perform a check step after each initialization step. measurement The number of fixed windows is first time Periodization window It can be significantly larger than the number of initial values. Therefore, the number of measurement windows is initialized window The number may be at least 10 times the number of . Furthermore, the processor unit may be further configured to execute the initialization phase at least twice per second. The number of measurement windows may be determined according to the sampling frequency. The number of repetitions of the initialization phase may be increased. Thus, the processor unit may be configured to execute the initialization phase, for example, 2 to 10 times per second. For this reason to measure The number of fixed windows can be appropriately adjusted. As mentioned above , noise tuple teeth , can be placed in the center of the initial coefficient tuples ru. therefore, rough A sound tuple can reside at the noise center of a "cloud" consisting of multiple initial coefficient tuples of the noise window. In particular, The sampling frequency for the sampling value of the measurement signal is, Inspection stage In small Even without that, 10,000 signal windows are available in the measurement window. In contrast, settings It appears that it is being done. Selected It has been selected. In particular, Faith Window No. number It can be significantly more than that.

[0019] moreover As mentioned above The processor unit is configured to detect the distance of the corresponding measurement tuple to the noise tuple for each measurement window. ru.The distance between each measurement tuple and the noise tuple is the Mahalanobis distance between the two tuples. For information on calculating the Mahalanobis distance, see the following publication: N-Dimensional cumulative function, and other useful facts about gaussians and normal densities by Michael Bensimhoun (Jerusalem, 06 / 2009).

[0020] The default X-th order model is a linear model and / or appropriate It can be an LPC model of order. The coefficients corresponding to the X-order model may be, for example, the alpha coefficient, the PARCOR coefficient, or the prediction error coefficient and / or the normalization error.

[0021] moreover As described above, the processor unit is selected from multiple measurement windows, The corresponding measurement tuple is at a distance greater than the default first limit value relative to the noise tuple. measuring Identify the fixed window as the error signal window. ni kamo It has been done ru. This limit can be, for example, 1 to 5 times the standard deviation of the predicted value or noise tuple. Already Certain limits can be defined, for example, based on expertise and prior examinations. Other possibilities for defining such limits are equally possible.

[0022] A preferred configuration of the device is: P The losser unit zeros out a portion of the measurement signal indicated by each initialization window for each initialization window. point passing It is characterized by being configured to detect the number of [something]. Furthermore, P The Rossessa section is Multiple From the number initialization window, Most zeros The system is configured to identify K initialization windows, each with a point passing count, as noise windows. first time Shown by the periodization window measurement Part of a constant signal In zero point passing Having a large number means measurementThis part of the constant signal to It has been confirmed that this means there is a lot of noise. Of these initialization windows, the most Noise Many Miscellaneous To detect the sound window, these initialization windows Regarding zero point passing Detecting the number of zeros in the measurement signal has been demonstrated as an efficient method. point passing teeth, measurement The value of the constant signal ,value Zero or another default of In the signal value , especially in a short period of time To reach or Through this Existing when passing Miscellaneous. The number of sound windows K is smaller than the number of initialization windows M, especially Remarkably It is small. That is, a number K can be, for example, at most two-thirds, at most one-half, or at most one-third of a number M. ,degree The initialization window is a noise window, and the processor section setting Whether it will be done is very robust. Confirmed This is possible. This noise window is used in the initialization phase to detect the corresponding K initial coefficient tuples and to detect the noise tuple from these initial coefficient tuples. It is known that noise does not provide meaningful correlation. Therefore, zero point passing The use of and robust determination of the noise window makes it possible to efficiently and simultaneously robustly determine the noise tuple. As a result, for each measurement window during the inspection phase, rough For a sound tuple measuring distance of a constant tuple can check It is possible to output. Furthermore, this allows for the identification of the error signal window. Additionally, zero point passing Determining the noise window using detection means rough The sound window is detected in real time or simultaneously in time during the initialization phase and / or before the inspection phase. Making it possible to release .

[0023] A preferred configuration of the device is: P The Rosser part 、Each initialization window, as indicated by that initialization window ru A portion of the measurement signal Response It is characterized by being configured to detect the energy of the number. In this case, P The Rossessa section is Multiple The system is further configured to identify K initialization windows with the minimum signal energy from a set of number initialization windows as noise windows. This is because, first time Indicated by the periodization window ta measurement Part of a constant signal In Signal energy The fact that it is small This is part of the measurement signal. to This is because it has been confirmed that this means there is a lot of noise. Of these initialization windows, the Noisy and messy To detect the sound window, these initialization windows Regarding Detecting the minimum signal energy has been demonstrated as an efficient method. Therefore, rough The sound window is zero point passing Instead of using I believe It can also be identified using the energy of the number. ,degree The initialization window can be very robustly determined to be identified by the processor as a noise window. The noise window is K during the initialization phase. Detect K initial coefficient tuples, and then detect a noise tuple from these K initial coefficient tuples. Therefore It is used for this purpose. It is known that noise does not provide meaningful correlation. Therefore, the use of minimal signal energy and robust determination of the noise window make it possible to efficiently and simultaneously robustly determine the noise tuple. As a result, in the inspection phase, the distance of the corresponding measurement tuple to this noise tuple can be detected for each measurement window. This also makes it possible to identify the error signal window. Furthermore , most By detecting small signal energy Noise window Identification allows the noise window to be detected in real time or simultaneously in time during the initialization phase and / or before the inspection phase.

[0024] Faith Instead of the number energy, so-called "spectral flatness" can also be used. The processor unit can be appropriately configured for this purpose.

[0025] Another preferred configuration of the device is X characterized in that the next model is configured as the X-th LPC model. The LPC model is basically known from the prior art. The LPC model can be described, for example, by the following recurrence equation: y(k)=e(k)+Σ1 N a i ·y(k-i). In this case, k is a discrete time variable, i.e., a natural number greater than zero, y(k) is the value of the discrete measurement signal at the discrete time point k, and N is the degree of approximation. In this case, N coincides with the value X of the X-th order of the LPC model. a i is a so-called linear predictor of the N-th order in particular, and e(k) is the prediction error. The predictor a i is the coefficient of the LPC model in particular. Since the values of the discrete measurement signal are known, the linear identifier or the coefficients of the model can be determined by the processor unit such that the sum of squared errors of the approximation of the model with respect to the discrete measurement signal is minimized. In this case, a part of the measurement signal indicated by each window is considered respectively. The sum of squared errors can be expressed by Q according to the following equation. Q=Σ1 N e 2 (k)=Σ1 N (y(k)-Σ1 N a i ·y(k-i)) 2 . For this purpose, the sum of squared errors of the linear identifier a i or the coefficients of the model of differentiate death 、 So each result of compare with zero death and obtain a system of simultaneous equations consisting of N first-order equations as a result It can be solved. Therefore to , P The decliner section can be appropriately configured to determine the coefficients of the LPC model. and For details regarding the determination of the number, please refer further to German Patent No. 102018126743.

[0026] surely, X The following model is usefully constructed as an Xth-order LPC model. However, it can also be proposed that this Xth-order model is constructed as another Xth-order mathematical model. Therefore, this model can be constructed, for example, using the Fourier transform or wavelet transform.

[0027] outfit Another preferred configuration of the setting is characterized in that M is an integer of at least 100, and more particularly at least 10,000. Furthermore, it has been usefully proposed that K is an integer smaller than M, in particular. Specifically, K is less than 0.5 times M, and more particularly, K is at most 0.1 times M. This can ensure robust and simple identification of multiple noise windows.

[0028] outfit Another preferred configuration of the setup is characterized in that each initialization phase lasts for a maximum of 0.1 seconds, and in particular for a maximum of 0.05 seconds. The inspection phase may last for, for example, several seconds, or even several minutes, or even longer. The short duration of the initialization phase allows for a very rapid transition to the inspection phase. outfit When a placement is used by a user, generally, Yu -za is, first time They don't realize they're in the critical development stage.

[0029] outfit Another preferred configuration of the arrangement is, P The lossr unit analyzes each error signal window using a default N-th order model. So Assigned to each error signal window 、The analysis is configured to determine the coefficients that make up the error tuple, wherein N is an integer of at least 6, and especially 8. Inspect Inspection stage In It is possible. In particular, N The next model is X It is of the same type as the next model. Both models, namely the Nth-order model and the Xth-order model, can each be formed by the corresponding LPC model of the same order. X The next model is, in particular, a second-order or third-order model. The Nth-order model is, in particular, a sixth-order model or a higher-order model such as an eighth-order model. However, the order may be greater. In particular, N is at least 10, 12, or 14. Inspect During the review phase, the response of The measurement tuple is compared to the noise tuple. do Measurement window with a distance greater than the default first limit value from , P By Rossessa Teso Each has a single error signal window. It is identified As a result, So It is inferred that a portion of the measurement signal indicated by each error signal window contains a signal error. be According to the preferred configuration of the device mentioned last, P The Rossessa section is workman The Rah signal window is configured to be analyzed by a default N-th order model, i.e., a model of at least the sixth order. The corresponding coefficients are determined by this analysis. Therefore, the analysis can be performed to determine the coefficients of a model of at least the sixth order. This allows, workman A portion of the measurement signal indicated by the error signal window can be examined with great accuracy. However, even when analyzing the error signal window using the default Nth-order model, significant computational effort is not required. However, since only a limited number of error signal windows are identified, this identification is often possible in real time and during the inspection phase. In other words, only the error signal windows are analyzed intensively. In particular, by using coefficients, a portion of the measurement signal indicated by each error signal window can be examined. extreme values fruit number Department and empty number The unit can be detected by the processor unit. The processor unit can be configured for this purpose. number Department and empty number From this, it can be confirmed whether a portion of the measurement signal indicated by each error signal window is due to a defect, particularly partial discharge. By fitting the default first limit value, in many cases only or a large portion of the error signal window is due to a defect. especially It may be guaranteed that a portion of the measured signal, including signal components caused by partial discharge, is shown. However, it may be proposed that the order N of the model be limited in order to keep the computational effort required to analyze the error signal window within an acceptable level. Thus, N may be, for example, at most 20 or 30.

[0030] In another preferred configuration of the device, the processor unit is: Error tuples in the same error group are such that they are each less than the default second limit value G. , error tuple multiple error group It is characterized by being configured to distribute. Therefore, all error signal windows assigned to the same error group are, It shows the same signal error. A defect in the power transmission system within the electrical equipment will cause an error signal in the detected measurement signal. occurrence This can happen. Therefore, multiple error signal windows may indicate multiple error signals caused by the same defect during power transmission by the electrical equipment. For example, if an unwanted partial discharge occurs in the electrical equipment during power transmission, multiple error signal windows may indicate a portion of the measurement signal that is strongly affected by the partial discharge. This often means that the error tuples detected by the processor based on these error signal windows are affected by the same defect, namely the partial discharge. Therefore, these error tuples are the same error It is proposed that they be assigned to groups. It has been demonstrated that such error tuples have only narrow intervals from one another. Therefore, the maximum interval between error tuples can be determined by a default second limit value. Thus, the processor will determine that error tuples within the same error group are each smaller than the default second limit value. distanceTo have multiple error groups, the error tuple is divided into multiple error groups. Distribution It is beneficially configured to do so. Furthermore, it can be proposed that error tuples of the same error group are positioned at least approximately the same distance from noise tuples. In particular, the deviation in the distance of error tuples from noise tuples is smaller than the default second limit value. This can ensure that error tuples of the same error group exhibit the same signal error. These error tuples can be caused by the same defect during power transmission by electrical equipment. error The group can also be called a cluster. In particular, different error group number This indicates the number of different defects during power transmission by the electrical equipment. For example, if three partial discharges occur during power transmission by the electrical equipment, then three error The default first limit is set so that the group is generated by the processor unit. value And a default second limit value may be specified and / or selected in advance. In this case, each error The group has multiple error tuples, each having the same partial discharge window as a factor.

[0031] Error tuples into multiple error groups Distribution By doing so, the number of defects during power transmission, particularly partial discharges, can be identified very easily and quickly. Furthermore, the location and intensity of partial discharges can be easily and quickly analyzed based on error tuples and corresponding error signal windows.

[0032] Another preferred configuration of the device is that the processor unit is the same error The error tuples in the group's error signaling window are less than the default second limit value of each other. distance Multiple error signal windows are set to each have error It is characterized by being configured to distribute to groups. As a result, each is the same errorAll error signal windows in a group indicate the same signal error. This configuration is similar to the above configuration of the device. However, here, instead of grouping error tuples, error signal windows are grouped. Therefore, for this configuration of the device, please refer to the above preferred description, preferred features, technical effects and / or advantages.

[0033] In another preferred configuration of the device, the processor unit detects multiple different defects in the electrical equipment. error It is characterized by being configured to make decisions based on groups. For example, the processor unit determines that the error tuple is one of three error group Distribution Either the error signal window is opened, or the error signal window is divided into three error groups. Distribution If so, the processor unit will determine different error groups based on these multiple error groups. Number of defects We can infer that, The number of defects is Raa Group number It corresponds to.

[0034] outfit In another preferred configuration of the arrangement, P The Rosser part , measurement It is configured to generate an image signal that displays a constant signal as a signal graph. P The Rossessa section is measurement Based on the sampling value of the constant signal, error It is characterized by being configured to visually display identical portions of the signal graphs assigned to the error signal window of the group. In connection with this, the same error In the group's error signal window Regarding Mention Even in that case, the same applies to error tuples in the same error group. This is because Each error tuple is assigned to one of several error signal windows. Therefore. same defect Some of the signal graphs caused by this can be displayed in the same way, for example. , Department Discharge defect When this occurs in electrical equipment during power transmission, multiple error signal windows are generated by the corresponding signal components of this partial discharge. defect This may be displayed. As mentioned above, these error signal windows are the same error Assigned to a group be .however ,workman Assigned to the Rah signal window was Error tuple of ,same error Assign to group to This is also possible. In both cases, this error The group ,workman Does it include a signal window, or ha Assigned by Rataple Tae This includes error signal windows. These error signal windows are caused by partial discharge. defect or another defect This can be displayed very clearly and visually. workman LAR signal window (or corresponding Ru Ta Some of the corresponding signal graphs identified by the same group (Pull) are visually displayed identically by the processor. this The display can be done, for example, by the same color (e.g., red, green, or blue). However, the corresponding part of the signal graph of , displayed identically by the same dashed line, line thickness and / or other features. do It is also possible. outfit The device may be configured to provide an image signal. Therefore, outfit The device may, for example, have an output signal interface. In this case, P The filter unit processes the image signal transmission It is configured to control the output signal interface in particular to transmit. picture The image signal is sent to a remote unit. transmission It may be done and, if necessary, displayed on the unit.

[0035] outfit Another preferred configuration of the device is characterized by having a display unit. Furthermore, P The Rossessa section is table Control the display unit so that it displays an image based on the image signal. and, picture The statue Faith The graph can be configured to be displayed visually. In this case, picture The signal graphs visually displayed by the image are ,same character error group no E Assigned to the Rah signal window measurement signal A portion based on the sampled values ​​can be displayed visually identically. Therefore, Faith The display device can very quickly identify which parts of the graph are caused by the same defect, particularly defects resulting from the same partial discharge.

[0036] In another preferred configuration of the device, the processor unit performs the initialization phase multiple times. death, The noise tuple is configured to be newly detected at each initialization stage. This allows for updating of possible noise changes.

[0037] In another preferred configuration of the device, the processor unit is configured to perform at least one test step after each initialization step. Preferably, it is proposed that one test step is performed by the processor unit after each initialization step. However, it may also be proposed that multiple test steps be performed after each initialization step. In this case, these test steps may be performed sequentially by the processor unit.

[0038] According to a second aspect of the present invention, the above problems are solved by a system having the features described in claim 14. That is, a system for transmitting power is proposed. In this case, the system has electrical equipment for transmitting power signals from the power supply interface of an electrical installation to the power distribution interface of the electrical installation. Furthermore, the system has a sensor unit and a device. The device is configured according to a first aspect of the present invention and / or according to one of a plurality of corresponding preferred configurations. With respect to the device of the system, refer to the above preferred descriptions, preferred features, technical effects and / or advantages as described in relation to the device according to the first aspect and / or according to one of a plurality of corresponding preferred configurations. The sensor unit of the system is located at the connection point of the electrical installation between the power supply interface and the power distribution interface. The sensor unit is configured to detect power signals and generate a digital measurement signal indicating the power signal detected at the connection point. To transmit the measurement signal to the signal interface, the sensor unit is connected to the signal interface of the device. For the advantages and technical effects of the system, please refer to the technical effects and advantages already described in relation to the apparatus and / or one of the corresponding preferred configurations according to the first aspect of the present invention. Further explanation is omitted.

[0039] In a preferred configuration of the system, the electrical installation is characterized by comprising high-voltage lines, transformers, rotating electrical machinery, gas-insulated lines, or gas-insulated switchgear. In particular, the electrical installation has high-voltage lines extending from the power supply point of the electrical installation to the power distribution point of the electrical installation. The high-voltage lines may be configured to transmit power by high DC voltage or high AC current. Alternatively and / or further, the electrical installation may have transformers. If the electrical installation has transformers instead of high-voltage lines, the power supply point may be configured on the primary side of the transformer, and the power distribution point may be configured on the secondary side of the transformer. Thus, power can be transmitted from the power supply point, i.e., the primary side of the transformer, to the power distribution point, i.e., the secondary side of the transformer. The above technical effects and / or preferred configurations also apply to other possible configurations of the electrical installation.

[0040] According to a third aspect of the present invention, the above problem is solved by a method having the features described in claim 16. That is, a sequence consisting of multiple sampled values That is A method is proposed for operating a device having an input signal interface for receiving digital measurement signals. In this case, the measurement signals represent signals detected at connection points of electrical equipment. The method includes at least the step of: a) assigning each one signal window to each sampling value of a digital measurement signal by a processor unit, wherein each signal window consists of a part of a sequence consisting of each sampling value of the measurement signal and a predetermined number of sampling values ​​that precede it in time. The method also includes the step of: b) assigning M signal windows as initialization windows by the processor unit during the initialization stage. setting The steps are: c) Select K initialization windows from M initialization windows as noise windows setting The steps are: d) analyze each noise window using a default X-order model and assign to each noise window 、The method is configured to perform the steps of: e) determining the coefficients that constitute the initial coefficient tuple, where X is an even number between 1 and 5, and e) determining the predicted value as the noise tuple from the initial coefficient tuple of the noise window. Furthermore, in the inspection stage, the processor unit performs: f) multiple signal windows as measurement windows setting The steps are: g) analyze each measurement window using a default X-order model and assign to each measurement window 、 The steps are: h) determining the coefficients that make up the measurement tuple, and for each measurement window, detecting the distance of the corresponding measurement tuple to the noise tuple, and i) from multiple measurement windows, Belonging to this measurement window The steps include identifying measurement windows as error signal windows in which corresponding measurement tuples are each at a distance greater than a default first limit value relative to a noise tuple, wherein each error signal window is configured to perform the steps of indicating a signal error in the measurement signal.

[0041] The method corresponds at least substantially to an apparatus according to a first aspect of the present invention. Therefore, with respect to the method, refer at least similarly to the preferred descriptions, preferred features, advantages and / or technical effects described with respect to an apparatus according to a first aspect of the present invention and / or a corresponding set of preferred configurations.

[0042] A preferred configuration of this method is characterized in that the initialization stage is performed multiple times by the processor unit, particularly 2 to 10 times per second. Therefore, in order to distinguish between the noise component and the error component of the measurement signal as accurately as possible, the noise tuple can be regularly updated by performing the initialization stage multiple times. In particular, this method is characterized in that at least one inspection stage is performed by the processor unit after each initialization stage.

[0043] Another preferred configuration of the method is characterized in that the Xth-order model is composed of the Xth-order LPC model. Again, refer to the corresponding preferred description, preferred features, effects and / or advantages, as already described for the corresponding preferred configuration of the apparatus according to the first aspect of the present invention.

[0044] Another preferred configuration of the method is characterized in that step c) includes sub-steps c.1) and c.2). In c.1), for each initialization window, the number of zero passes in a portion of the measurement signal indicated by each initialization window is detected. In c.2), from the M initialization windows, K initialization windows having the most zero passes are identified as noise windows. Each sub-step of both sub-steps c.1) and c.2) is performed by the processor unit. In step c.1), the number of zero passes in a portion of the measurement signal indicated by each initialization window is detected. Then, in step c.2), K noise windows are identified by the processor unit from the M initialization windows based on the number of zero passes. With regard to the configuration of the method, refer to the preferred descriptions, preferred features, technical effects and advantages already described above with respect to the corresponding preferred configuration of the apparatus according to the first aspect of the present invention. Further explanation is omitted here.

[0045] In another preferred configuration of the method, step c) is characterized by comprising sub-steps c.1) and c.2). In c.1), for each initialization window, a portion of the signal energy of the measurement signal indicated by each initialization window is detected. In c.2), from the M initialization windows, K initialization windows having the minimum signal energy are identified as noise windows. Each sub-step of both sub-steps c.1) and c.2) is performed by the processor unit. In step c.1), a portion of the measurement signal indicated by each initialization window is detected. Signal energyIt is detected. Then, in step c.2), K noise windows are identified by the processor unit based on the minimum signal energy from M initialization windows. With regard to the configuration of the method, refer to the preferred description, preferred features, technical effects and advantages already described above with respect to the corresponding preferred configuration of the apparatus according to the first aspect of the present invention. Further explanation is omitted here.

[0046] Another preferred configuration of the method is that, in the inspection stage, the processor unit: j) analyzes each error signal window by a predetermined Nth-order model and assigns to each error signal window 、 The method is characterized by being configured to perform a further step of determining the corresponding coefficients that constitute the error tuple, where N is an integer of at least 6, particularly 8.

[0047] In another preferred configuration of the method, the processor unit, during the inspection phase, checks that: k) error tuples in the same error group are less than a default second limit value. distance Each of the error tuples has one of the following characteristics, and the error tuple is assigned to multiple error groups. Distribute It is configured in such a way that the error signal window corresponds of A key feature is that all error signal windows assigned to the same error group by an error tuple indicate the same signal error.

[0048] With regard to the preferred configuration of the method mentioned last, refer to the preferred description, preferred features, effects and / or advantages already described above for both corresponding preferred configurations of the apparatus according to the first aspect of the present invention.

[0049] Further features, advantages, and applicability of the present invention will be described with reference to the following embodiments and drawings. In this case, the features described and / or illustrated themselves, and any combination thereof, constitute the subject matter of the present invention regardless of the configuration described in the individual claims or the claims referenced by the individual claims. Furthermore, in the drawings, the same reference numerals indicate the same or similar subject matter. [Brief explanation of the drawing]

[0050] [Figure 1] A schematic representation of a preferred configuration of the apparatus and system is shown below. [Figure 2] The distribution of multiple coefficient tuples is illustrated as an example. [Figure 3] A portion of the measurement signal is shown as an example. [Figure 4] The structure of the method is illustrated with a schematic flowchart. [Figure 5] A schematic flowchart illustrates a preferred alternative configuration of the method. [Modes for carrying out the invention]

[0051] Figure 1 schematically shows a preferred configuration of device 2. Device 2 includes an input signal interface 4 and a processor unit 10. Furthermore, device 2 includes a display unit 22. Preferably, it is proposed that device 2 also has an output signal interface 24.

[0052] Furthermore, Figure 1 schematically shows an electrical installation 8. The electrical installation 8 is used to transmit power. Therefore, the electrical installation 8 may be configured, for example, for high-voltage direct current transmission. However, it is also possible that the electrical installation 8 is configured to transmit power in a different manner. Therefore, the electrical installation 8 may be configured, for example, for high-voltage alternating current transmission. The electrical installation 8 has, in particular, a power supply interface 28 and a power distribution interface 30. At least one wire 38 may be laid from the power supply interface 28 to the power distribution interface 30. Current can be passed through the wire 38. However, the preferred configuration of the electrical installation 8 shown in Figure 1 is merely illustrative. Therefore, in a different configuration, the electrical installation 8 may consist of a transformer. The primary terminals of the transformer constitute the power supply interface 28, and the secondary terminals of the transformer constitute the power distribution interface 30.

[0053] The electrical equipment 8 and device 2 are system Mu2 It may constitute part of 6. Furthermore, it may be proposed that system 26 have a sensor unit 32. This sensor unit 32 is connected to the input signal interface 4 of device 2, in particular via a signal line 36. To electrically detect power transmission, the sensor unit 32 may be located at the connection point 6 of the electrical equipment 8. Thus, the sensor unit 32 may be configured to detect, for example, the voltage and / or current flowing through the wire 38. Thus, the sensor unit 32 is configured to detect the power signal of the electrical equipment 8. Furthermore, the sensor unit 32 is configured to generate an analog measurement signal indicating the power signal detected at the connection point. This measurement signal may be transmitted from the sensor unit 32 to the input signal interface 4 of device 2 via the signal line 36. This measurement signal is digitized by the processor unit 10. The digital measurement is a sequence consisting of multiple sampled values ​​of the detected signal. The input signal interface 4 of device 2 is a sequence consisting of multiple sampled values. That is It is configured to receive digital measurement signals.

[0054] The above and below descriptions of device 2 may apply specifically to device 2. However, the same descriptions of device 2 also apply to system 26, because system 26 may include device 2.

[0055] Figure 3 shows a typical portion of the measured signal. On the time axis The data is recorded and outlined. Essentially, at least a portion of the measured signal represents the power signal detected at the connection point 6 of the electrical equipment 8. However, if a defect (Fehler), such as a partial discharge, occurs in the wire 38, this defect, particularly the partial discharge, introduces a corresponding signal component into the measured signal. In this case, it is difficult to determine which portion of the measured signal contains the signal component that particularly highlights each defect.

[0056] Therefore, device 2 is used at least to identify error signal windows of the measurement signal. In this case, each error signal window indicates a signal error in the measurement signal. In this case, these error signal windows indicate multiple different signal errors. However, at least multiple error signal windows indicate the same signal error in the measurement signal. case This is also the case when these signal errors are caused by the same defect, such as partial discharge. Therefore, each error signal may be part of the measurement signal that arises from a defect in the electrical equipment 8 during power transmission, particularly partial discharge.

[0057] To ensure the above-described desired objectives of device 2, a beneficial configuration of the processor unit 10, as described below, has been proposed.

[0058] The processor unit 10 is measurement Each sampling value of the constant signal Each has a window for one signal It is configured to be assigned. This signal window is measured Each sampling value of the constant signal and This measurement signal It has a predetermined number of sampling values ​​that precede it in time. measuringEach is composed of a part of a sequence consisting of multiple sampled values ​​of a constant signal. 。 Each signal window is, measurement It may contain the same number of sample values ​​as the constant signal. Therefore, for example, each signal window may contain: measurement It may be proposed to include 32, 64, or 128 sampling values ​​for a constant signal. Multiple sampling values ​​in one signal window are: measurement It is part of a sequence consisting of multiple sampled values ​​of a constant signal. Therefore, So The multiple sampling values ​​in each signal window are multiple sampling values ​​that are directly consecutive in time. P The lossr unit 10 processes, for example, one sampling value of the measurement signal. One signal window generated , assigned to each sampling value In this case, this signal window is measurement The sampling values ​​of each constant signal and the sampling values ​​of each measurement signal are compared. do It includes a predetermined number of sampling values ​​that precede it in time, for example, 127 sampling values ​​that precede it in time. In this example, this signal window contains a total of 128 sampling values. measurement Each sampling value of the constant signal One signal window Since it is allocated, it will be generated appropriately. Ta The signal window can also be described as a fluctuating sequence consisting of multiple signal windows. In describing the initialization stage 12 and the inspection stage 18 below, measurement It must be considered that the assignment of these signal windows to each sampling value of the constant signal continues, and in particular, is performed in parallel in time.

[0059] The processor unit 10 is in the initialization stage 12 In It is configured to perform several steps as described below. Therefore, the processor unit 10 performs initialization step 12 In Using M signal windows as initialization windows setting It is configured in this way. In this case, the M signal windows are provided by the processor unit 10. measurementMultiple sampling values ​​of a constant signal Continuously Assignment Ta It is part of multiple signal windows. The number M is an integer between 100 and 50,000. In particular, the number M is an integer between 1,000 and 10,000. Therefore, in initialization stage 12, for example, 10,000 initialization windows are generated by the processor unit 10. setting It is possible. From these initialization windows K noise windows are set The number K is smaller than the number M. Therefore, the number K can be an integer, and in particular it can be less than 0.5 or less than 0.1 times the number M. For example, the number K may be 1,000. For example, 1,000 noise windows are generated by the processor unit 10. numerous From the initialization window setting It is possible. degree The initialization window is a noise window setting Whether or not this will happen is indicated by these initialization windows. ta measurement Part of the constant signal Number of times the zero point is passed This can be determined by detecting or by detecting the signal energy. P The lossr section is defined by each initialization window, as shown by each initialization window. measurement Part of a constant signal Number of times the zeros are passed It may be configured to determine the or the corresponding signal energy. Thus, each initialization window This includes the number of zeros passed through or a specific signal energy. It is assigned .child These initialization windows The number of zero intersections or determined The signal energy is, Generally different That ru. Most times passing zeros Or the initialization window to which the minimum signal energy is assigned. but This shows a portion of the measurement signal that has the most noise components. Therefore, the processor unit 10 selects from M initialization windows, Most times passing zeros Alternatively, the system is configured to identify an initialization window with the minimum signal energy as a noise window. Therefore, these noise windows constitute a subset of all initialization windows. In this case, a portion of the measured signal indicated by these noise windows is High noise component It is characterized by the following.

[0060] An X-order model can be stored in the processor unit 10, particularly in the associated memory unit. This model is a mathematical model configured to model power transmission by electrical equipment. This model may have coefficients to fit the transmission function or transmission distance shown by the model. Therefore, by fitting the coefficients of the model, it is achieved to approximate the actual transmission function during power transmission by the electrical equipment 8 very well, depending on the selected coefficients. This model is also called the default model (vorbestimmte Modell). In the initialization stage 12, the processor unit analyzes each noise window using the default X-order model and responds accordingly. of It is configured to determine the coefficients. Therefore, a set of coefficients is determined for each noise window. The coefficients are different for each window. For quick determination, it has been shown to be beneficial for X to be an even number between 1 and 5. Therefore, X may be, for example, the number 2. In this case, the default model is a quadratic model. The number of coefficients corresponds in particular to the order of the model. Therefore, in the example above, two coefficients are provided for the quadratic model. Noise window About The coefficients determined by the processor unit 10 constitute one initial coefficient tuple 14 assigned to each noise window. Since the noise can differ for each noise window, the initial coefficient tuple 14 can also differ for each noise window. Therefore, in the initialization stage 12, the processor unit 10 selects from the initial coefficient tuples 14 of multiple noise windows. The predicted values ​​are set as noise tuple 16. It is configured to determine the noise tuple 16. Therefore, the noise tuple 16 may also be the arithmetic mean of the K initial coefficient tuples 14. Each initial coefficient tuple 14 consists of two coefficients, and for example, 1,000 noise windows, i.e., 1,000 initial coefficient tuples 14 decisionIf so, the noise tuple 16 may represent predicted values ​​for these initial coefficient tuples 14. In short, the noise tuple 16 may be placed at the "center of the noise". When another signal window is analyzed by the default model as an initialization window and the corresponding coefficients are determined, it is determined whether the corresponding signal window should be evaluated as noise based on the distance between the noise tuple 16 and the tuple consisting of these coefficients. Alternatively, if the distance is sufficiently large, the signal window represents a portion of the measured signal that has a signal error.

[0061] The processor unit 10 is configured to perform the following steps in the inspection stage 18. That is, in the inspection stage 18, the processor unit 10 uses multiple signal windows as measurement windows. setting It is configured to do so. In this case, preferably the measurement window is different from the initialization window. ru. Therefore, for example, the signal window following the last initialization window is the measurement window. setting This is possible. In particular, the number of measurement windows is significantly larger than the number of initialization windows. Therefore, in the inspection stage 18, for example, at least 10,000 signal windows are measured windows. Set This is possible. However, a larger number, for example 20,000, 30,000, or 100,000, is preferable. Furthermore, the processor unit 10 analyzes each measurement window using a predetermined X-order model and responds accordingly. of It is configured to determine the coefficients. In this case, the default X-th order model is the same X-th order model that was already used in initialization stage 12, that is, the one assigned to each measurement window. 、 Constitute measurement tuple 20 The person in charge of The number is determined for each measurement window. Therefore, a measurement tuple 20 consisting of each coefficient is provided for each measurement window. be .

[0062] Figure 2 shows that multiple coefficient tuples 34 correspond to ofThis is schematically shown by several points. Each coefficient tuple 34 consists of multiple coefficients determined by the X-order model during the analysis of the signal window. This analysis is performed on the noise window in the initialization stage 12. This results in multiple initial coefficient tuples 14. In Figure 2, since there are many of these initial coefficient tuples 14, Cloud of multiple points This is shown as follows: The noise tuple 16 is located at the center of the cloud, which consists of multiple initial coefficient tuples 14. Therefore, the noise tuple 16 is the predicted value for these initial coefficient tuples 14. forming .

[0063] In inspection stage 18, the measurement window is analyzed by the same X-order model. As a result, the corresponding multiple coefficients are obtained from the measurement tuple 20 above. forming . these One of the measurement tuples 20 is marked with an "x" in Figure 2. mark It is being done. Measurement tuple 20 corresponds of To determine whether or not the measurement signal indicated by the measurement window shows some noise. to , or to determine whether this measurement tuple 20 indicates a signal error in a portion of the measurement signal shown by each measurement window. to The default first limit value G is setting If the distance D between each measurement tuple 20 and the noise tuple 16 is greater than the first limit value G, then a portion of the measurement signal indicated by the corresponding measurement window is considered to be a signal error caused by a defect during power transmission by the electrical equipment 8. However, if the distance D is less than the limit value G, This measurement tuple 20 is, A portion of the measurement signal indicated by the corresponding measurement window is a signal error. What to show Rather, to noise Being more affected show.

[0064] Therefore, in the inspection step 18, the processor unit 10 calculates the distance D of the measurement tuple 20 to the noise tuple 16 for each measurement window. detection death, From multiple measurement windows, correspondence of Measurement tuple 20 The noise tuple 16 has a distance D greater than the default first limit value G. The measurement window is configured to be identified as an error signal window. This ensures that each of these identified error signal windows represents a portion of the measurement signal containing a signal error. because Each of these parts of the measurement signal indicated by the error signal window is affected by a defect during power transmission by the electrical equipment 8. Because they are receiving it. The signal error in the measurement signal is, for example, a portion of the measurement signal caused by each defect. Therefore, each error signal window also indicates the signal error in the measurement signal.

[0065] Testing stage 18 In particular Each of the error signal windows is an important part of the measurement signals that can be used to inspect for potential defects during power transmission by the electrical equipment 8.

[0066] Therefore, preferably, device 2 outputs the output signal transmit It has been proposed that the system has a configured output signal interface 24. The output signal may directly or indirectly indicate an error signal window. Therefore, the output signal may, for example, be in the measurement signal. to It may include data for identifying the error signal window. However, the output signal gaso It may also be proposed to indicate each signal window. To transmit, in particular transmit, the processor unit 10 of device 2 may be configured to control the output signal interface 24. Information regarding the error signal window detected by device 2 may be provided to other devices and / or units by the output signal.

[0067] In place of or in addition to the output signal interface 24, the device 2 may have a display unit 22. Furthermore, the processor unit 10 may be configured to generate an image signal that shows the measurement signal as a signal graph. The processor unit 10 may be configured to transmit the image signal to the display unit 22. The display unit 22 may be configured to generate an image on the display unit 22 based on the image signal.

[0068] Furthermore, the processor unit 10, Based on the sampled value of the measurement signal, it is assigned to an error signal window of the same error group. Signal graph shown by image signal one It has been beneficially proposed that the parts be configured to appear visually identical. In this case, the processor unit 10 has multiple error signal windows. error group Distribution It can be configured to do the same. error The error signal window of the group is the same when power is transmitted by the electrical equipment 8. defect This could indicate a single error signal caused by each of the following.

[0069] Figure 3 shows a portion of the measurement signal. error The signal window 40 of the group is shown as a solid line, and the second error The group signal window 42 is shown as a dotted line. In particular, the processor unit 10 is configured to control the display unit 22 so that the display unit 22 displays the corresponding image based on the image signal. As a result, the image is the same as the signal graph, and in particular parts of the signal graph. index Visually display using (Indizierung).

[0070] Initialization stage 12 and inspection stage 18 Regarding teeth, X Next model The It has already been explained that each is used for analysis. The Xth-order model in question is, in particular, the Xth-order LPC model. Alternatively, another mathematical model may be used. Another Mathematical models can be created, for example, using Fourier transforms or wavelet transforms.

[0071] moreover ,So Each signal window of , analysis using a higher-order model vinegar It has been proven that this is beneficial. In this case, this model is The aboveA higher-order model of the same type as the X-order model may also be used. Therefore, the processor unit 10 analyzes each error signal window using a default N-order model and assigns to each error signal window based on this analysis. 、 Construct an error tuple The person in charge of It is usefully proposed that the model be configured to determine a number. In this case, N is an integer of at least 6, and in particular an integer of at least 8. Similarly, the Nth-order model may be stored by the processor unit 10, and in particular by the associated memory unit. Furthermore, it is usefully proposed that the Nth-order model is an Nth-order LPC model. The Nth order may be at least 6, at least 8, and in particular at least 10, 12, or 14, as described above. The degree is higher. This has the advantage that the error signal window can be analyzed more accurately, allowing for more precise inspection of each defect. Furthermore, the error signal window can be grouped based on error tuples. For example, multiple error tuples in the same error group may be smaller than the default second limit value G. distance It has been proposed that it is beneficial to distribute multiple error tuples into multiple error groups, each having a corresponding error. Therefore, the correspondence of error signal windows of Each error tuple is the same error All error signal windows assigned to a group may indicate the same or at least similar signal errors. Reha The same defect can be caused during power transmission by the electrical equipment 8. Therefore, the same partial discharge, in particular, can be caused by the same defect. defect Error tuples and / or error signal windows that have a cause of ,same error Assign to group It is possible Therefore, the number of groups may correspond to the number of different defects during power transmission by the electrical equipment 8.

[0072] Figure 4 shows a preferred configuration of the method of the present invention. The method comprises steps a) to i). According to step a), one signal window is assigned to each sampling value of the measurement signal. This signal window is defined as each sampling value of the measurement signal and a predetermined number of sampling values ​​of the measurement signal that precede it in time. Consists of Each step consists of parts of the sequence. Step a) is performed sequentially, i.e., concurrently with initialization step 12 and / or inspection step 18.

[0073] Steps b) to e) of the method are assigned to the initialization stage 12. The method is configured so that the processor unit 10 performs the following steps in the initialization stage 12: b) set M signal windows as initialization windows, and c) select K initialization windows from the M initialization windows as noise windows. setting d) Analyze each noise window using the default X-order model and assign to each noise window 、 Construct an initial coefficient tuple. The person in charge of Determine the number, in this case X is an even number between 1 and 5, and e) determine the predicted value as a noise tuple from the initial coefficient tuple of the noise window.

[0074] Furthermore, the method is configured such that in the inspection step 18 the processor unit 10 performs the following steps: f) multiple signal windows as measurement windows setting g) Analyze each measurement window using the default X-order model and assign to each measurement window 、 Constitute a measurement tuple The person in charge of Determine the number, h) for each measurement window, detect the distance of the corresponding measurement tuple to the noise tuple, i) from multiple measurement windows, A measurement window in which the measurement tuples belonging to these measurement windows are located at a distance greater than the default first limit value relative to the noise tuple is defined as an error signal window. Identify .therefore Each error signal window indicates a signal error in the measured signal.

[0075] As can be seen from Figure 4, the inspection stage 18 follows the initialization stage 12. That is, In initialization stage 12 Steps b) to e) , This is executed by the lossr unit 10. After that, In inspection stage 18 Steps f) to i) are executed. After that, initialization stage 12 can be started again. However, it is also possible to run inspection stage 18 multiple times after initialization stage 12, before a new initialization stage 12 is started again. For this method, see the preferred description, beneficial features, technical effects and / or advantages described above.

[0076] Figure 5 shows another preferred configuration of the method of the present invention. In this case, refer to steps a) to i) above for explanation. The method according to Figure 5 includes substeps c.1) and c.2) performed by the processor unit 10 in step c): c.1) For each initialization window, the number of zero passes and / or signal energy of the portion of the measurement signal shown by each initialization window is detected; c.2) From the M initialization windows, K initialization windows having the most zero passes and / or the least signal energy are identified as noise windows.

[0077] Furthermore, the method shown in Figure 5 includes step j) of the inspection stage 18 performed by the processor unit 10. According to step j), each error signal window is analyzed by a predetermined N-th order model and assigned to each error signal window 、 Construct an error tuple The person in charge of The number is determined. In this case, N is an integer of at least 6, and especially 8.

[0078] Furthermore, it should be noted that “having” does not exclude other elements or steps, and “one” does not exclude multiple. It should also be noted that features described with reference to one embodiment of the above multiple embodiments may be used in combination with other features of another embodiment. Reference numerals in the claims should not be considered limiting. Although this application relates to the invention described in the claims, it may also encompass the following configurations as other embodiments. 1. A device (2) having an input signal interface (4) for receiving a digital measurement signal having a sequence of multiple sampled values ​​indicating a signal detected at a connection point (6) of electrical equipment (8), and a processor unit (10), The processor unit (10) is configured to assign each sampling value to a single signal window, each consisting of a part of a sequence of these sampling values ​​of the measurement signal, which has each sampling value and a predetermined number of sampling values ​​of the measurement signal that precedes it in time, The processor unit (10) determines M signal windows as initialization windows during the initialization stage (12), identifies K noise windows from these M initialization windows, analyzes each noise window using a predetermined X-order model, determines the corresponding coefficients that constitute the initial coefficient tuple (14) assigned to each noise window, and detects one predicted value as a noise tuple (16) from the initial coefficient tuple (14) of the noise window, where X is configured to be an even number between 1 and 5. The processor unit (10) identifies a plurality of signal windows as measurement windows during the inspection stage (18), analyzes each measurement window using a predetermined X-order model, determines the corresponding coefficients that constitute the measurement tuple (20) assigned to each measurement window, detects the distance of the corresponding measurement tuple (20) to the noise tuple (16) for each measurement window, identifies a plurality of measurement windows as error signal windows from the plurality of measurement windows, and the device (2) is configured such that each error signal window indicates a signal error in the measurement signal as a result of the corresponding measurement tuple (20) of these measurement windows having a distance greater than a predetermined first limit value G to the noise tuple (16). 2. The processor unit (10) is configured to detect, for each initialization window, the corresponding number of zero crossovers and / or the corresponding signal energy of a portion of the measurement signal indicated by each initialization window. The apparatus (2) according to claim 1, wherein the processor unit (10) is further configured to identify K initialization windows having the most zero crossings and / or the minimum signal energy from among the plurality of initialization windows as noise windows. 3. The Xth-order model is the apparatus (2) described in any one of 1 to 2 above, which is configured as the Xth-order LPC model. 4. Apparatus (2) according to any one of the above 1 to 3, wherein M is an integer of at least 100, in particular at least 10000, and K is an integer smaller than M. 5. Each initialization stage (12) is a device (2) described in any one of the above 1 to 4, which lasts for a maximum of 0.1 seconds, and especially for a maximum of 0.05 seconds. 6. The processor unit (10) is configured to analyze each error signal window using a predetermined N-th order model and to determine the corresponding coefficients that constitute the error tuple assigned to each error signal window, where N is an integer of at least 6, particularly 8, as described in any one of 1 to 5 above. 7. The device (2) described in 6 above is configured such that the processor unit (10) classifies a plurality of defect tuples into a plurality of error groups such that a plurality of defect tuples in the same error group have an interval of one less than a default second limit value G between them, and as a result, all error signal windows that are assigned to the same error group by the corresponding defect tuples in the error signal windows indicate the same signal error. 8. The device (2) according to any one of 1 to 6 above, wherein the processor unit (10) classifies a plurality of error signal windows into a plurality of error groups such that a plurality of defect tuples of a plurality of error signal windows in the same error group have an interval of one less than a default second limit value G between them, as a result, all error signal windows in each of the same error groups indicate the same signal error. 9. The device (2) according to any one of 7 to 8, wherein the processor unit (10) is configured to detect the number of different defects in the electrical equipment (8) based on a plurality of error groups. 10. The processor unit (10) is configured to generate an image signal that displays the measurement signal as a signal graph, The device (2) according to any one of 7 to 9 above, wherein the processor unit (10) is configured to visually display identical portions of the signal graph based on the sampling values ​​of the measurement signals assigned to the same error group error signal windows. 11. The device (2) has a display unit (22), The device (2) according to 10, wherein the processor unit (10) controls the display unit (22) so that the display unit (22) displays an image based on the image signal, and as a result, the image is configured to visually display the signal graph. 12. The device (2) according to any one of 1 to 11 above, wherein the processor unit (10) is configured such that the noise tuple (16) is newly detected for each initialization step (12) as a result of executing the initialization step (12) multiple times. 13. The apparatus (2) according to 12, wherein the processor unit (10) is configured to perform at least one test step (18) after each initialization step (12). 14. A system (26) for transmitting power, The system (26) comprises the electrical equipment (8) for transmitting power signals from the power supply interface (28) of the electrical equipment (8) to the power distribution interface (30) of the electrical equipment (8), a sensor unit (32), and the device (2) described in any one of 1 to 13 above. The sensor unit (32) is located at the connection point (6) of the electrical equipment (8) between the power supply interface (28) and the power distribution interface (30), The sensor unit (32) is configured to detect the power signal and generate a digital measurement signal indicating the power signal detected at the connection point (6). To transmit the measurement signal to the signal interface, the sensor unit (32) is connected to the signal interface of the device (2) (26). 15. The aforementioned electrical equipment (8) is the system (26) described in 14 above, which is configured as a high-voltage line, a transformer, a rotating electrical machine, a gas-insulated line, or a gas-insulated switchgear. 16. A method for operating a device (2) having an input signal interface (4) for receiving a digital measurement signal having a sequence of multiple sampled values ​​indicating a signal detected at a connection point (6) of an electrical installation (8), The method in question is as follows: a) The processor unit (10) includes at least the step of assigning to each sampling value of the measurement signal a signal, each of which is composed of a part of a sequence of these sampling values ​​of the measurement signal, having each sampling value of the measurement signal and a predetermined number of sampling values ​​that precede the measurement signal in time, This method involves the following being performed by the processor unit (10) during the initialization stage (12): b) A step of determining M signal windows as initialization windows, c) A step of identifying K initialization windows as noise windows from the M initialization windows, d) Analyze each noise window using a default X-th order model and determine the corresponding coefficients that make up the initial coefficient tuple (14) assigned to each noise window, where X is an even number from 1 to 5, e) The system is configured to perform the step of detecting one predicted value as a noise tuple (16) from the initial coefficient tuple (14) of the noise window, The method is performed by the processor unit (10) during the inspection stage (18) as follows: f) A step of identifying multiple signal windows as measurement windows, g) The step of analyzing each measurement window using a predetermined X-order model and determining the corresponding coefficients that constitute the measurement tuple (20) assigned to each measurement window, h) For each of the measurement windows, the step of detecting the distance D of the corresponding measurement tuple (20) to the noise tuple (16), i) The method is configured to perform the step of identifying a plurality of measurement windows from the plurality of measurement windows as error signal windows, and each error signal window indicates a signal error in the measurement signal as a result of the corresponding measurement tuple (20) of these measurement windows as error signal windows having a distance D greater than a default first limit value G for the noise tuple (16). 17. The initialization step (12) is performed multiple times, particularly 2 to 10 times per second, by the processor unit (10), according to the method described in 16 above. 18. The Xth-order model is the method described in any one of the above 16 to 17, which is the Xth-order LPC model. 19. Step c) is as follows: c.1) For each initialization window, detect the corresponding number of corresponding zero crossovers of the portion of the measurement signal shown by each initialization window, and / or for each initialization window, detect the corresponding signal energy of the portion of the measurement signal shown by each initialization window; c.2) The method according to any one of 16 to 18, further comprising step c.2) identifying K initialization windows having the most zero crossings from the M initialization windows as noise windows, and / or identifying K initialization windows having the minimum signal energy from the M initialization windows as noise windows. 20. The above method is performed by the processor unit (10) during the inspection step (18) as follows: j) The method according to any one of 16 to 19 above, further comprising the steps of analyzing each error signal window by a default N-th order model and determining the corresponding coefficients that constitute the defect tuple corresponding to each error signal window, where N is an integer of at least 6, particularly 8. 21. The above method is performed by the processor unit (10) during the inspection step (18) as follows: k) The method according to 20, wherein the method is further configured to classify a plurality of defect tuples into a plurality of error groups such that a plurality of defect tuples in the same error group have an interval of one less than a default second limit value G between them, and as a result, all error signal windows that are assigned to the same error group by the corresponding defect tuples in the error signal windows indicate the same signal error.

Explanation of Symbols

[0079] a1 First coefficient a2 Second coefficient[[ID=​​​​​​​​6 Connection Points 8. Electrical equipment 10. Processor section 12 Initialization Stage 14 Initial coefficient tuple 16 Noise Tuple 18. Testing Stage 20 measurement tuples 22 Display section 24 Output signal interfaces 26 Systems 28 Power supply interface 30 Distribution Interfaces 32 Sensor device 34 coefficient tuple 36 signal lines 38 Electric wire 40 1 error Group signal window 42. 2 error Group signal window

Claims

1. An apparatus (2) having an input signal interface (4) for receiving a digital measurement signal which is a sequence consisting of multiple sampling values ​​and a processor unit (10), The aforementioned digital measurement signal indicates the power signal detected by the sensor unit (32) at the connection point (6) between the electrical equipment (8) for transmitting power and the sensor unit (32). The sensor unit (32) is located between the power supply interface (28) and the power distribution interface (30) of the electrical equipment (8). The processor unit (10) is configured to assign one signal window to each sampling value of the digital measurement signal, Each signal window is composed of a part of a sequence consisting of each sampling value of the digital measurement signal and a predetermined number of sampling values ​​that precede that sampling value in time. The processor unit (10) in the initialization stage (12) - Set M signal windows as initialization windows, - From the M initialization windows mentioned above, K initialization windows are set as noise windows. - Analyze each noise window using a default X-order model, and determine the coefficients that make up the initial coefficient tuple (14) assigned to each noise window. - The system is configured to determine the predicted value as a noise tuple (16) from the initial coefficient tuple (14) of the noise window, where X is an even number between 1 and 5. The processor unit (10) in the inspection stage (18) - Set multiple signal windows as measurement windows, - Analyze each measurement window using the X-order model, and determine the coefficients that constitute the measurement tuple (20) assigned to each measurement window. - For each measurement window, the distance D of the measurement tuple (20) to the noise tuple (16) is detected. - The system is configured to identify, from among multiple measurement windows, an error signal window in which the measurement tuple (20) belonging to that measurement window has a distance D greater than a predetermined first limit value G with respect to the noise tuple (16). Each of the aforementioned error signal windows indicates a signal error in the digital measurement signal, the device (2).

2. The processor unit (10) is configured to determine, for each initialization window, the number of zero-point passes and / or signal energy of a portion of the measurement signal indicated by each initialization window, The apparatus (2) according to claim 1, characterized in that the processor unit (10) is further configured to identify K initialization windows having the most zeros passed and / or the least signal energy from among a plurality of initialization windows as noise windows.

3. The apparatus (2) according to any one of claims 1 to 2, characterized in that the Xth-order model is configured as an Xth-order LPC model.

4. The apparatus (2) according to any one of claims 1 to 3, characterized in that M is an integer of at least 100 or at least 10000, and K is an integer smaller than M.

5. The apparatus (2) according to any one of claims 1 to 4, characterized in that each initialization stage (12) lasts for a maximum of 0.1 seconds or a maximum of 0.05 seconds.

6. The processor unit (10) is configured to analyze each error signal window using a predetermined N-th order model and to determine the coefficients that constitute the error tuple assigned to each error signal window. The apparatus (2) according to any one of claims 1 to 5, wherein N is an integer of at least 6 or 8.

7. The processor unit (10) is configured to distribute error tuples into multiple error groups such that error tuples in the same error group are each at a distance less than a predetermined second limit value G from each other. The apparatus (2) according to claim 6, characterized in that all error signal windows assigned to the same error group by the error tuple indicate the same signal error.

8. The processor unit (10) is configured to distribute the error signal windows to multiple error groups such that the error tuples in the same error group's error signal window are each less than a predetermined second limit value G from each other. The apparatus (2) according to any one of claims 1 to 6, characterized in that all error signal windows of the same error group indicate the same signal error.

9. The apparatus (2) according to any one of claims 7 to 8, characterized in that the processor unit (10) is configured to determine the number of different types of defects in the electrical equipment (8) based on the number of error groups.

10. The processor unit (10) is configured to generate an image signal that displays the digital measurement signal as a signal graph. The apparatus (2) according to any one of claims 7 to 9, characterized in that the processor unit (10) is configured to visually display identical portions of the signal graphs assigned to the same error signal window based on the sampling value of the digital measurement signal.

11. The device (2) has a display unit (22), The apparatus (2) according to claim 10, wherein the processor unit (10) is configured to control the display unit (22) so that the display unit (22) displays an image based on an image signal, and the image visually displays a signal graph.

12. The apparatus (2) according to any one of claims 1 to 11, characterized in that the processor unit (10) is configured to perform the initialization step (12) multiple times, and a noise tuple (16) is newly detected for each initialization step (12).

13. The apparatus (2) according to claim 12, characterized in that the processor unit (10) is configured to perform at least one inspection step (18) after each initialization step (12).

14. A system (26) for transmitting power, The system (26) comprises an electrical facility (8) for transmitting a power signal from the power supply interface (28) of the electrical facility (8) to the power distribution interface (30) of the electrical facility (8), a sensor unit (32), and the device (2) described in any one of claims 1 to 13. The sensor unit (32) is located at the connection point (6) of the electrical equipment (8) between the power supply interface (28) and the power distribution interface (30). The sensor unit (32) is configured to detect a power signal and generate a digital measurement signal indicating the power signal detected at the connection point (6). The system (26) is connected to the signal interface of the device (2) in order to transmit the digital measurement signal to the signal interface.

15. The system (26) according to claim 14, characterized in that the electrical equipment (8) is configured as a high-voltage line, a transformer, a rotating electrical machine, a gas-insulated line, or a gas-insulated switchgear.

16. A method for operating a device (2) having an input signal interface (4) for receiving a digital measurement signal which is a sequence of multiple sample values ​​indicating a power signal detected by a sensor unit (32) at a connection point (6) with a sensor unit (32) of an electrical equipment (8) for transmitting power, wherein the sensor unit (32) is located between a power supply interface (28) and a power distribution interface (30) of the electrical equipment (8), This method is a) A step of assigning one signal window to each sampling value of a digital measurement signal by the processor unit (10), wherein each signal window is composed of a part of a sequence consisting of each sampling value of the digital measurement signal and a predetermined number of sampling values ​​that precede the sampling value in time, In this method, during the initialization stage (12), the processor unit (10) performs the following: b) A step of setting M signal windows as initialization windows, c) A step of setting K initialization windows as noise windows from the M initialization windows, d) A step of analyzing each noise window using a predetermined X-th order model and determining the coefficients that constitute the initial coefficient tuple (14) assigned to each noise window, wherein X is an even number from 1 to 5. e) The system is configured to perform the step of determining a predicted value as a noise tuple (16) from the initial coefficient tuple (14) of the noise window, In this method, during the inspection stage (18), the processor unit (10) performs the following: f) A step of setting multiple signal windows as measurement windows, g) A step of analyzing each measurement window using the X-order model and determining the coefficients that constitute the measurement tuple (20) assigned to each measurement window, h) For each measurement window, the step of detecting the distance D of the measurement tuple (20) to the noise tuple (16), i) A method configured to perform the steps of: determining from a plurality of measurement windows, an error signal window having a distance D greater than a predetermined first limit value G for the measurement tuple (20) belonging to that measurement window relative to the noise tuple (16), wherein each of the error signal windows indicates a signal error in the digital measurement signal.

17. The method according to 16, characterized in that the initialization step (12) is performed multiple times or 2 to 10 times per second by the processor unit (10).

18. The method according to any one of claims 16 to 17, characterized in that the Xth-order model is an Xth-order LPC model.

19. Step c) is c. 1) A sub-step c. 1) for each initialization window, detecting the number of zeros passed through a portion of the measurement signal shown by that initialization window, and / or for each initialization window, detecting the signal energy of a portion of the measurement signal shown by that initialization window, The method according to any one of claims 16 to 18, further comprising the sub-step c. 2) identifying K initialization windows having the most zeros passed from M initialization windows as noise windows, and / or identifying K initialization windows having the minimum signal energy from M initialization windows as noise windows.

20. The above method is performed by the processor unit (10) in the inspection step (18): j) The system is configured to further perform the step of analyzing each error signal window using a default N-th order model and determining the coefficients that constitute the error tuple assigned to each error signal window, The method according to any one of claims 16 to 19, characterized in that N is at least an integer of 6 or 8.

21. The above method is performed by the processor unit (10) in the inspection step (18): k) The system is further configured to perform the step of distributing multiple error tuples into multiple error groups such that multiple error tuples in the same error group are each less than a default second limit value G from each other, The method according to 20, characterized in that all error signal windows assigned to the same error group by the error tuple of the error signal window indicate the same signal error.

Images