Evaluation of partial discharge signals

Abstract

The present invention relates to a method for evaluating partial discharge signals, wherein the partial discharge signals are detected at only one location or at multiple locations near the insulation of the electrical outer conductor of a polyphase alternating current and are superimposed on each other, and the alternating currents flowing in the outer conductors of the polyphase alternating current have a fixed phase shift relative to each other. In this method, at least one characteristic variable (K) of the partial discharge signal is defined, and the phase of the alternating current at the time point when the partial discharge signal is detected is associated with each partial discharge signal. A characteristic variable value of each characteristic variable (K) is determined for each partial discharge signal, and a digital tuple is associated with each partial discharge signal, the digital tuple being formed by each characteristic variable value of the partial discharge signal and the phase associated with the partial discharge signal. The cluster (C) of the partial discharge signals is determined in a multi-dimensional space by using the points (V) formed by the digital tuples, and the partial discharge activity is determined for each outer conductor from the determined cluster (C).

Description

Technical Field

[0001] The present invention relates to the evaluation of partial discharge signals, where the partial discharge signals are detected at only one location or at multiple locations near the insulation of the electrical outer conductor in polyphase alternating current and are superimposed on each other. In other words, the present invention relates to the "single-channel" evaluation of partial discharge signals, without separately recording and evaluating the partial discharge signals of individual outer conductors for polyphase alternating current. Background Art

[0002] Partial discharge is, for example, a partial discharge caused in local inhomogeneities of an insulating material, such as small cavities generated inside the insulation of an electrical conductor due to a strong inhomogeneity of the electric field. Partial discharge can locally damage the insulation and lead to insulation failure in the long term. Therefore, in particular, the partial discharge of an electric machine is regularly inspected. Partial discharge causes short electromagnetic pulses, which have a typical pulse duration of less than 1 μs and frequency components up to the UNF frequency range. Generally, partial discharge signals are searched for in the frequency range between approximately 100 kHz and 10 MHz, where the partial discharge pulses are usually the strongest.

[0003] An outer conductor is an electrical conductor that is under voltage during normal operation and can contribute to the transmission or distribution of electrical energy, but is not a neutral conductor (central conductor).

[0004] For the detection of partial discharge, there are a large number of partially standardized measurement methods, such as according to DIN EN 60270. Almost only off-line measurements are performed on electric machines, where the normal operation of the electric machine is interrupted and high costs are incurred, such as due to the operation interruption and the costs for the measuring equipment. Therefore, such measurements can only be carried out occasionally.

[0005] Online measurements can enable continuous monitoring of a machine during the normal operation of the electric machine and in particular can identify trends regarding the partial discharge activity at the machine and the correlation between the partial discharge activity and different environmental influences such as temperature, air humidity, and air pressure. The information obtained therefrom can enable a reliable prediction of the period during which the machine affected by partial discharge can still operate reliably. The partial discharge signals detected by the measurement are usually plotted in a network phase-synchronized manner in the form of a partial discharge histogram and the intervals of their corresponding energy content are evaluated as the partial discharge rate.

[0006] In the case where a polyphase alternating current is formed by a plurality of alternating currents, partial discharges generated by different alternating currents overlap over time. Therefore, generally, partial discharge signals are detected and evaluated separately for each of these alternating currents. If, for cost-saving reasons, one only wants to use one partial discharge measuring unit or one evaluation unit, that is, to detect partial discharge signals only at one location or to evaluate partial discharge signals detected at multiple locations but overlapping with each other, then a problem arises, namely that the conventional evaluation of partial discharge signals cannot provide an account of individual alternating currents. Summary of the Invention

[0007] An object of the present invention is to provide a method for improving the evaluation of partial discharge signals, where the partial discharge signals are detected only at one location near the insulation of the electrical outer conductor of a polyphase alternating current or are detected at multiple locations and superimposed on each other.

[0008] According to the present invention, this object is achieved by a method having the features of the present invention.

[0009] Advantageous designs of the present invention are the subject matter of the dependent claims.

[0010] In the method according to the present invention, partial discharge signals are evaluated, where the partial discharge signals are detected only at one location near the insulation of the electrical outer conductor of a polyphase alternating current or are detected at multiple locations and superimposed on each other, and the alternating currents flowing in the outer conductors of the polyphase alternating current have a fixed phase shift relative to each other. For this, at least one characteristic variable is defined, based on which the partial discharge signals can be compared with each other. The phase of the alternating current at the time point when the partial discharge signal is detected is associated with each partial discharge signal. For each partial discharge signal, a characteristic variable value of each characteristic variable is determined. A digital tuple is associated with each partial discharge signal, and the digital tuple is formed by each characteristic variable value of the partial discharge signal and the phase associated with the partial discharge signal. Then, in a multi-dimensional space, clusters of partial discharge signals are determined using the points formed by the digital tuples, and the partial discharge activity is determined for each outer conductor from the determined clusters.

[0011] In one design, the partial discharge signal is decoupled through the connection of at least one winding temperature sensor. A partial discharge detection method is obtained by decoupling one or more partial discharge signals via one winding temperature sensor or through multiple winding temperature sensors, and this partial discharge detection method measures the superimposed activity in all phases, that is, partial discharge. In order to separate the activity of each phase from this superposition, clusters are used, that is, a clustering method. This is beneficial for being able to track the activity of each phase and thus identify changes in phases that can only be identified in the subsequent overall activity at an early stage.

[0012] Thus, the method according to the invention provides for the evaluation of partial discharge signals in a multi-dimensional space, the coordinates of the partial discharge signals being at least one characteristic variable of the partial discharge signals and the phase of the alternating current. In this space, each partial discharge signal is represented by a point which is determined by at least one characteristic variable value of the partial discharge signal and the phase at the time point at which the partial discharge signal is detected. Cluster analysis is used to combine the points representing the partial discharge signals into different clusters in order to associate the partial discharge signals with different outer conductors or to determine the partial discharge activity for each outer conductor. Thus, the method according to the invention enables a phase- or outer conductor-dependent evaluation of partial discharge signals using only one measuring unit instead of separate measuring units for each outer conductor, and thus the method can be implemented at low cost. In addition, the method according to the invention enables an online evaluation of partial discharge signals, for example during the normal operation of an electric machine carrying polyphase alternating current, since cluster analysis does not require special operating conditions.

[0013] In one design of the invention, the pulse duration, pulse height, charge, energy and / or repetition rate of the partial discharge signal are defined as characteristic variables. For example, the pulse duration constitutes the first characteristic variable, the pulse height constitutes the second characteristic variable, etc.

[0014] In another design of the invention, the clusters are determined using a partitioned clustering method, a hierarchical clustering method, a density-based clustering method and / or a method using a neural network. For example, the k-means algorithm, mean shift, Gaussian mixture and / or DBSCAN (density-based spatial clustering of applications with noise) can be used for clustering.

[0015] In another design of the invention, the cluster centroid is determined for each cluster and the cluster is associated with an outer conductor which is determined by the phase of the cluster centroid. For example, an outer conductor is associated with each phase and the cluster is associated with the outer conductor which is associated with the phase of the cluster centroid of the cluster. In particular, it can be provided that each outer conductor is associated with at least one phase angle interval of the phase angle characterizing the phase. The design of the invention utilizes that partial discharges at the outer conductor usually accumulate in a determined phase angle interval of the alternating current flowing in the outer conductor, for example in a phase angle interval in which the alternating current or the corresponding alternating voltage has a relatively strong slope. Since the alternating currents of the individual outer conductors have a fixed phase shift relative to each other, one of the outer conductors can thus be associated with a phase angle interval of the phase angle characterizing the alternating current, in which the probability of partial discharge at the outer conductor increases in the corresponding phase angle interval.

[0016] In another design of the present invention, in order to determine the partial discharge activity of the outer conductor, at least one activity variable for the partial discharge signal is defined, the partial discharge signal is detected in a predetermined time window and the partial discharge signal is associated with the outer conductor. For example, a plurality of partial discharge signals are defined as activity variables, the partial discharge signal is detected in the time window and the partial discharge signal is associated with the outer conductor. Alternatively or additionally, the activity variable is formed by the characteristic variable value of the partial discharge signal, the partial discharge signal is detected in a predetermined time window and the partial discharge signal is associated with the outer conductor. This realizes the quantitative evaluation and assessment of the partial discharge activity at each outer conductor.

[0017] In an improved version of the above design of the present invention, the time distribution of the activity values of at least one activity variable of each outer conductor is detected and evaluated. Here, it can be set that for each time distribution of the activity values, an anomaly recognition is performed, the anomalous activity values are obtained by using this anomaly recognition, and the anomalous activity values are removed from the corresponding distribution of the activity values. It can also be set that for at least one activity variable and each outer conductor, a regression curve of the time-related change of the activity variable is obtained from the time distribution of the activity values. It can also be set that the first derivative of each regression curve with respect to time is obtained. By evaluating the time distribution of the activity values of each outer conductor in this way, it is possible to advantageously identify the trend of the partial discharge activity at each outer conductor and derive a prediction of the future development of the partial discharge activity. When the regression curve exceeds a predetermined threshold or the first derivative of the regression curve exceeds a predetermined threshold, a warning and / or alarm signal can also be generated.

[0018] The evaluation unit according to the present invention is arranged to evaluate the partial discharge signal of the method according to the present invention, associate the phase of the alternating current at the time point of detecting the partial discharge signal with each partial discharge signal, obtain the characteristic variable value of each characteristic variable for each partial discharge signal, associate a digital tuple with each partial discharge signal, the digital tuple is formed by each characteristic variable value of the partial discharge signal and the phase associated with the partial discharge signal, find the cluster of the partial discharge signal by using the points formed by the digital tuples in a multi-dimensional space, and obtain the partial discharge activity for each outer conductor from the obtained cluster.

[0019] The evaluation unit can in particular be arranged to execute the computer program according to the present invention. The computer program according to the present invention includes instructions which, when the computer program is executed, cause the evaluation unit to execute the method according to the present invention.

[0020] Thus, the method according to the present invention can be realized by using the above advantages, especially in a computer-implemented manner.

[0021] Alternatively, the method can be executed or partially executed in at least one so-called edge device or in an application of a computer cloud. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above-described features, characteristics, and advantages of the present invention, as well as the types and manners of achieving them, become clearer and more understandable in connection with the description of the following embodiments, which are explained in more detail in conjunction with the accompanying drawings. Shown herein:

[0023] Figure 1 A flowchart showing an embodiment of a method for evaluating partial discharge signals according to the present invention,

[0024] Figure 2 Showing clusters of partial discharge signals and the cluster centroids,

[0025] Figure 3 Showing the time distribution of the activity values of the activity variables of partial discharge activity and the regression curve of the time distribution of the activity variables,

[0026] Figure 4 Showing the regression curve of the time distribution of the activity variables of the outer conductor of a three-phase alternating current.

[0027] Corresponding components are provided with the same reference numerals in the figures. DETAILED DESCRIPTION

[0028] Figure 1 A flowchart 100 showing an embodiment of the method according to the present invention, the method having method steps 101 to 109 for evaluating partial discharge signals, the partial discharge signals being detected at only one location near the insulation of the electrical outer conductor of a polyphase alternating current, or being detected at multiple locations and superimposed on each other, and the alternating currents flowing in the outer conductors of the polyphase alternating current having a fixed phase shift relative to each other. For example, partial discharge signals are detected using an antenna and filtered using a frequency filter, the pass range of the frequency filter being typical for the partial discharge signals. However, any other sensor can also be used to detect partial discharge signals, such as a capacitive sensor device having at least one sensor capacitor or an inductively coupled device having at least one sensor coil. This is not important for the present invention.

[0029] In a first method step 101, at least one characteristic variable K is defined, based on which the partial discharge signals can be compared with each other. Such a characteristic variable can be, for example, the pulse duration, pulse height, charge, energy, or repetition rate of the partial discharge signals.

[0030] In a second method step 102, the phase of the alternating current at the time point of detecting the partial discharge signal is associated with each partial discharge signal. By the phase angle of one of the alternating currents Limited phase.

[0031] In the third method step 103, the characteristic variable value of each characteristic variable K is determined for each partial discharge signal.

[0032] In the fourth method step 104, a digital tuple is associated with each partial discharge signal, the digital tuple being formed by the characteristic variable value of each characteristic variable of the partial discharge signal and the phase associated with the partial discharge signal. For example, if the characteristic variable K is the pulse duration and the pulse height, then the entries of the digital tuple associated with the partial discharge signal are the characteristic variable value of the pulse duration, the characteristic variable value of the pulse height, and the phase associated with the partial discharge signal.

[0033] In the fifth method step 105, for a predetermined time window, clusters C of the partial discharge signals detected in this time window are determined separately in the multi-dimensional space S using the points V formed by the digital tuples. For example, the clusters C are determined using a partitioning clustering method (such as based on the k-means algorithm), a hierarchical clustering method, a density-based clustering method (such as DBSCAN), and / or a method using a neural network.

[0034] In the sixth method step 106, the outer conductor is associated with each cluster C. For this purpose, each outer conductor is associated with at least one phase angle interval I1 to I6, the cluster centroid P is determined for each cluster C, and the cluster C is associated with the following outer conductor, which is associated with the phase of the cluster centroid P of the cluster C (i.e., the coordinate value of the cluster centroid P in the space S). value).

[0035] Figure 2 Exemplarily, clusters C determined in a time window for a three-phase alternating current with an alternating current having a phase shift of 120° to each other and their cluster centroids P are shown, wherein the points V of the respective clusters C are described by the same and different symbols than the other clusters C. The phase angle intervals I1 to I4 are associated with the first outer conductor. The phase angle intervals I2 to I5 are associated with the second outer conductor. The phase angle intervals I3 to I6 are associated with the third outer conductor. Each phase angle interval I1 to I6 has a width of 60°. The phase angle intervals I1 and I3 each have three cluster centroids P. The phase angle intervals I2 and I4 each have one cluster centroid P. The phase angle interval I5 has no cluster centroid P. The phase angle interval I6 has four cluster centroids P. Thus, a total of four cluster centroids P are associated with the first outer conductor in the time window, one cluster centroid P is associated with the second outer conductor, and seven cluster centroids P are associated with the third outer conductor.

[0036] In the seventh method step 107, the partial discharge activity is determined for each outer conductor in a time window by means of the determined cluster C. For this purpose, at least one activity variable A for the partial discharge signal is defined, the partial discharge signal is detected in the corresponding time window and the partial discharge signal is associated with the outer conductor. For example, a plurality of partial discharge signals are defined as the activity variable A, the partial discharge signal is detected in the time window and the partial discharge signal is associated with the outer conductor. Alternatively or additionally, the activity variable A is formed by characteristic variable values of the partial discharge signal, the partial discharge signal is detected in the time window and the partial discharge signal is associated with the outer conductor. For example, the value of the characteristic variable K of the cluster center of gravity P is used as the activity variable A, and / or the activity variable A is formed by the maximum value, the minimum value, the standard deviation, the ratio of the maximum value to the average value and / or the statistical moment of the characteristic variable values of the partial discharge signals of the assigned cluster C.

[0037] In the eighth method step 108, the time distribution of the activity values is detected for at least one activity variable A of each outer conductor, for example the distribution over a plurality of months. Anomaly detection (also known as outlier detection in English) is also carried out for each time distribution, and the anomalous activity values are determined by means of this anomaly detection and the anomalous activity values are removed from the corresponding distribution of the activity values. For anomaly detection, the time distribution of the activity values is analyzed, for example, by means of a density-based clustering method such as DBSCAN. Such anomaly detection is known, for example, from M.M. Breunig et al., LOF: identifying density-based local outliers, Proceedings of the 2000 ACM SIGMOD international conference on Management of data, pages 93-104, doi: 10.1145 / 342009.335388.

[0038] In the ninth method step 109, for at least one active variable A and each outer conductor, regression curves R, R1 to R3 of the change of the active variable A related to time t are obtained from the time distribution of the active values, for example, by using a known method of so-called support vector machine regression. It can also be set to obtain the first derivative of each regression curve R, R1 to R3 with respect to time. The time development of the partial discharge activity of each outer conductor is monitored based on the regression curves R, R1 to R3 and, if necessary, the first derivative of the regression curves. For example, when the regression curves R, R1 to R3 exceed a predetermined threshold or the first derivative of the regression curves R, R1 to R3 exceeds a predetermined threshold, a warning and / or alarm signal is automatically generated.

[0039] The method described according to method steps 101 to 109 can be extended, for example, such that before the cluster C is obtained, anomaly recognition in the multi-dimensional space S is performed in the fourth method step 104 to eliminate abnormal partial discharge signals.

[0040] Figure 3 Exemplarily, the time distribution of the active values of the active variable A of the outer conductor and the regression curve R of the change of the active variable A related to time t are shown. The abnormal active values are far from the regression curve R and are represented by a symbol different from other active values A.

[0041] Figure 4 Exemplarily, the regression curves R1 to R3 of the time change of the active variable A of the three outer conductors of a three-phase alternating current are shown. The regression curve R1 of the first outer conductor extends above the regression curve R2 of the second outer conductor and below the regression curve R3 of the third outer conductor, but has a high slope within the time interval Δt. From this, for example, it can be concluded that the highest partial discharge activity occurs at the third outer conductor, the lowest partial discharge activity occurs at the second outer conductor, and the partial discharge activity at the first outer conductor increases very rapidly within the time interval Δt.

[0042] Method steps 101 to 109 are performed, for example, by an evaluation unit for evaluating partial discharge signals. A computer program is executed on the evaluation unit, and the computer program includes instructions that, when the computer program is executed, cause the evaluation unit to perform method steps 101 to 109. The evaluation unit can particularly have at least one so-called neuromorphic integrated circuit. Alternatively, the method is executed or partially executed in at least one so-called edge device or in an application program in a computer cloud (Cloud).

[0043] Although the present invention has been described and illustrated in detail by means of sectional embodiments, the present invention is not limited to the disclosed examples, and those skilled in the art can derive other variant solutions therefrom without departing from the scope of protection of the present invention.

Claims

1. A method for evaluating partial discharge signals, wherein the partial discharge signals are detected at only one location or at multiple locations near the insulation of the electrical outer conductor of a polyphase alternating current and are superimposed on each other, and the alternating currents flowing in the outer conductors of the polyphase alternating current have a fixed phase shift relative to each other, where - Define at least one characteristic variable (K) based on which partial discharge signals can be compared with one another locally, where the pulse duration, pulse height, charge, energy, and / or repetition rate of the partial discharge signal are defined as characteristic variables (K), - associating the phase of the alternating current at the time point of detecting the partial discharge signal with each partial discharge signal, - determining the characteristic variable value of each characteristic variable (K) for each partial discharge signal, - associating a digital tuple with each partial discharge signal, the digital tuple being formed by each characteristic variable value of the partial discharge signal and the phase associated with the partial discharge signal, - determining clusters (C) of the partial discharge signals using points (V) formed by the digital tuples in a multi-dimensional space (S) and - determining the partial discharge activity for each outer conductor from the determined clusters (C), wherein a cluster centroid (P) is determined for each cluster (C) and the cluster (C) is associated with the outer conductor determined by the phase of the cluster centroid (P), and wherein the trend of the partial discharge activity at each outer conductor is identified by evaluating the time distribution of the activity values of the respective outer conductors.

2. The method according to claim 1, wherein, The clusters (C) are determined using a partitioning clustering method, a hierarchical clustering method, a density-based clustering method, and / or a method using a neural network.

3. The method according to claim 1 or 2, wherein The outer conductors are associated with each phase and the clusters (C) are associated with the outer conductor associated with the phase of the cluster centroid (P) of the cluster (C).

4. The method according to claim 3, wherein, Each outer conductor is associated with at least one phase angle interval (I1 to I6) characterizing the phase angle (φ) of the phase.

5. The method according to claim 1 or 2, wherein To determine the partial discharge activity of the outer conductor, at least one activity variable (A) for the partial discharge signal is defined, the partial discharge signal is detected in a predetermined time window, and the partial discharge signal is associated with the outer conductor.

6. The method according to claim 5, wherein, A plurality of the partial discharge signals are defined as activity variables (A), the partial discharge signals are detected in the time window, and the partial discharge signals are associated with the outer conductor.

7. The method according to claim 5, wherein The activity variable (A) is formed by the characteristic variable values of the partial discharge signal, the partial discharge signal is detected in the time window, and the partial discharge signal is associated with the outer conductor.

8. The method according to claim 5, wherein The time distribution of the activity values of at least one activity variable (A) of each outer conductor is detected and evaluated.

9. The method according to claim 8, wherein, Anomaly identification is performed on each time distribution of the activity values, the anomalous activity values are determined using the anomaly identification, and the anomalous activity values are removed from the corresponding distributions of the activity values.

10. The method according to claim 8, wherein, For at least one activity variable (A) and each outer conductor, a regression curve (R, R1, to R3) of the time (t)-related change of the activity variable (A) is determined from the time distribution of the activity values.

11. The method according to claim 10, wherein, The first derivative of each regression curve (R, R1, to R3) with respect to time (t) is determined.

12. An evaluation unit for evaluating partial discharge signals of the method according to any one of the preceding claims, wherein, The evaluation unit is configured to - associate the phase of the alternating current at the time point of detecting the partial discharge signal with each partial discharge signal, - determine a characteristic variable value of each characteristic variable (K) for each partial discharge signal, - associate a digital tuple with each partial discharge signal, the digital tuple being formed by each characteristic variable value of the partial discharge signal and the phase associated with the partial discharge signal, - determine a cluster (C) of the partial discharge signals by using points (V) formed by the digital tuples in a multi-dimensional space (S), and - determine the partial discharge activity for each outer conductor from the determined cluster (C).

13. A computer program product comprising a computer program, the computer program comprising instructions which, when the computer program is executed, cause an evaluation unit to perform the method according to any one of claims 1 to 11.

Images