Online Intelligent Diagnosis Method and Device for Ultra-High Frequency Partial Discharge Faults in Hydro-generators

Abstract

This application provides an online intelligent diagnostic method and device for ultra-high frequency partial discharge faults in hydro-generators, belonging to the field of partial discharge monitoring technology. The method includes: acquiring monitoring signals collected by a coupling device within the current monitoring period; if the conditions for enabling the interference signal library are met, inputting the amplitude sequence obtained after discretizing the monitoring signals into a pre-built machine learning model to generate a diagnostic result indicating whether the hydro-generator is operating normally; if the abnormal operating state of the hydro-generator is diagnosed and the monitoring signal is determined not to be an interference signal using the pre-built interference signal library, proceeding to the next step; otherwise, accumulating the interference signal types in the interference signal library and proceeding to the next step; if the amplitude of the monitoring signal collected in the current monitoring period is greater than an amplitude threshold, and the number of monitoring periods with amplitudes greater than the amplitude threshold continuously accumulates to a period accumulation threshold, proceeding to the next step; and identifying partial discharge events.

Description

Technical Field

[0001] This application belongs to the field of partial discharge monitoring technology, specifically relating to an online intelligent diagnostic method for ultra-high frequency partial discharge faults in hydro-generators, an online intelligent diagnostic device for ultra-high frequency partial discharge faults in hydro-generators, a machine-readable storage medium, and a computer device. Background Technology

[0002] As the core equipment of a hydropower station, the health of the insulation system of a hydro-generator directly affects the safety and stability of the entire power system. In actual production, the detection of partial discharge (PD) signals is an important means of reflecting the insulation status of large power equipment. With the continuous expansion of hydropower installed capacity, higher requirements are placed on the monitoring and evaluation of the insulation status of hydro-generators. In recent years, compared with low-frequency (100kHz~500kHz) partial discharge monitoring technology, ultra-high frequency (300MHz~3GHz) partial discharge monitoring technology has been gradually adopted due to the following advantages: 1) It has higher sensitivity and can effectively capture early partial discharge signals, thus detecting insulation defects in advance; 2) It can filter out more low-frequency interference, thus having stronger anti-interference capabilities and being able to detect partial discharge signals more accurately in complex electromagnetic environments; 3) It can provide higher fault location accuracy; 4) It uses an ultra-high frequency partial discharge antenna as a coupling device to collect partial discharge signals, and this non-contact signal acquisition method ensures online partial discharge monitoring under normal operation of the hydro-generator. Therefore, ultra-high frequency partial discharge monitoring technology provides data support for online preventive maintenance of hydro-generators. However, the complex operating conditions of hydro-generators result in significant interference in the partial discharge data collected using ultra-high frequency partial discharge monitoring technology, making it difficult to perform quantitative analysis of the partial discharge data using thresholds. Currently, the main method for partial discharge identification is qualitative analysis based on single-unit trend comparison. However, single-unit trend comparison requires high professional skills from on-site technicians, has low identification efficiency, and is affected by subjective factors.

[0003] In summary, how to effectively utilize the partial discharge data collected using UHF partial discharge monitoring technology to achieve accurate online diagnosis of partial discharge faults remains a technical challenge that urgently needs to be overcome in this field. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide an online intelligent diagnostic method for ultra-high frequency partial discharge faults in hydro-generators, an online intelligent diagnostic device for ultra-high frequency partial discharge faults in hydro-generators, a machine-readable storage medium, and a computer device, so as to realize the effective use of partial discharge data collected by ultra-high frequency partial discharge monitoring technology, so as to accurately diagnose partial discharge faults online.

[0005] To achieve the above objectives, the first aspect of this application provides an online intelligent diagnosis method for ultra-high frequency partial discharge faults in hydro-generators, comprising:

[0006] Acquire the monitoring signals collected by the UHF partial discharge coupling device during the current monitoring period;

[0007] If the conditions for enabling the interference signal library are met, the amplitude sequence obtained after discretizing the monitoring signal is input into the pre-built machine learning model to generate a diagnostic result of whether the hydro-generator is working normally. If the abnormal working state of the hydro-generator is diagnosed and the monitoring signal is determined not to be an interference signal using the pre-built interference signal library, the process jumps to trigger the next step. Otherwise, the interference signal types in the interference signal library are accumulated, and the process jumps to trigger the next step.

[0008] Determine whether the amplitude of the monitoring signal acquired in the current monitoring cycle is greater than the amplitude threshold. If so, and the number of monitoring cycles in which the amplitude of the corresponding monitoring signal is greater than the amplitude threshold has been accumulated to reach the cycle accumulation threshold, jump to trigger the next step. Otherwise, enter the next monitoring cycle and jump to trigger the acquisition of the monitoring signal acquired by the UHF partial discharge coupling device in the current monitoring cycle.

[0009] Identify the partial discharge event to which the monitoring signals collected in the current monitoring period belong.

[0010] In a specific embodiment of this application, identifying the partial discharge event to which the monitoring signal acquired during the current monitoring period belongs includes:

[0011] The feature parameters of the monitoring signal collected in the current monitoring period are extracted, and the partial discharge events corresponding to the feature parameters of the extracted monitoring signal are matched from the constructed partial discharge case library. Each partial discharge event in the partial discharge case library has its own corresponding feature parameters.

[0012] In a specific embodiment of this application, the conditions for enabling the interference signal library include either the number of types of interference signals in the interference signal library reaching a first preset value or the cumulative number of monitoring cycles reaching a second preset value.

[0013] In a specific embodiment of this application, based on a first matching process, the monitored signal is determined not to be an interference signal using a pre-constructed interference signal database. The first matching process includes:

[0014] Extract a first feature from the monitoring signal, wherein the first feature includes at least an energy feature;

[0015] Determine whether the difference between the first statistical value and the first feature of the extracted monitoring signal is within a preset range. If not, determine that the monitoring signal is not an interference signal. The first statistical value is the statistical value of the first feature of each interference signal in the constructed interference signal library. The type of the first statistical value includes at least one of the average value of the first feature of all types of interference signals in the interference signal library, the maximum value of the first feature of all types of interference signals in the interference signal library, and the minimum value of the first feature of all types of interference signals in the interference signal library.

[0016] In a specific embodiment of this application, the energy characteristic is power spectral density.

[0017] In a specific embodiment of this application, determining whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold includes:

[0018] The interference status of the production site where the ultra-high frequency partial discharge coupling device is located is obtained;

[0019] Analyze the interference status to obtain the number of interference sources that have been collected at the production site;

[0020] If the number of interference sources is greater than a third preset value, an amplitude threshold is adaptively generated, and it is determined whether the amplitude of the monitoring signal collected in the current monitoring cycle is greater than the amplitude threshold. Otherwise, a preset amplitude threshold based on expert experience is obtained, and it is determined whether the amplitude of the monitoring signal collected in the current monitoring cycle is greater than the amplitude threshold.

[0021] In a specific embodiment of this application, the feature parameters of the monitoring signal collected in the current monitoring period are extracted, including:

[0022] The monitoring signal is discretized to obtain a pulse sequence;

[0023] Pulses with amplitudes lower than the amplitude threshold in the pulse sequence are removed to obtain the target pulse sequence, which is then used as the discretized monitoring signal.

[0024] Extract the characteristic parameters of the discretized monitoring signal.

[0025] In specific embodiments of this application, the feature parameters include at least one of phase, amplitude, waveform features, amplitude statistical features, and spectrum features.

[0026] In a specific embodiment of this application, the method further includes:

[0027] Determine whether the amplitude of the monitored signal meets the shutdown conditions; if so, issue a control signal that the hydro generator needs to be shut down, and then jump to trigger execution. If the interference signal library activation conditions are met, input the amplitude sequence obtained after discretizing the monitored signal into the pre-built machine learning model; otherwise, jump to trigger execution. If the interference signal library activation conditions are met, input the amplitude sequence obtained after discretizing the monitored signal into the pre-built machine learning model.

[0028] The shutdown conditions include at least one of the following: the amplitude of the monitoring signal is greater than the second statistical value, or the amplitude of the monitoring signal is greater than one of a plurality of maximum partial discharge amplitudes.

[0029] Wherein, the maximum partial discharge amplitude is the maximum value of the partial discharge amplitude of the partial discharge event in the corresponding constructed laboratory severe partial discharge case library. Each partial discharge event in the laboratory severe partial discharge case library has its own corresponding maximum partial discharge amplitude. The second statistical value is the statistical value of each maximum partial discharge amplitude. The type of the second statistical value includes at least one of the maximum value of each maximum partial discharge amplitude and the average value of each maximum partial discharge amplitude. The laboratory severe partial discharge case library includes at least partial discharge events that occur when a hydro-generator operates under extreme conditions.

[0030] The second aspect of this application provides an online intelligent diagnostic device for ultra-high frequency partial discharge faults in hydro-generators, comprising:

[0031] The acquisition module is used to acquire the monitoring signals collected by the UHF partial discharge coupling device during the current monitoring period;

[0032] The first judgment module is used to determine whether the conditions for enabling the interference signal library are met. If so, the amplitude sequence obtained after discretizing the monitoring signal is input into the pre-built machine learning model to generate a diagnostic result of whether the hydro-generator is working normally. When the abnormal working state of the hydro-generator is diagnosed and the monitoring signal is determined not to be an interference signal by using the pre-built interference signal library, the module jumps to trigger the execution of judging whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold. Otherwise, the module accumulates the interference signal types in the interference signal library and jumps to trigger the execution of judging whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold.

[0033] The second judgment module is used to determine whether the amplitude of the monitoring signal collected in the current monitoring cycle is greater than the amplitude threshold. If so, and the number of monitoring cycles in which the amplitude of the corresponding monitoring signal is greater than the amplitude threshold is continuously accumulated to reach the cycle accumulation threshold, the module jumps to trigger the execution of identifying the partial discharge event to which the monitoring signal collected in the current monitoring cycle belongs. Otherwise, the module enters the next monitoring cycle and jumps to trigger the execution of acquiring the monitoring signal collected by the UHF partial discharge coupling device in the current monitoring cycle.

[0034] The identification module is used to identify the partial discharge event to which the monitoring signal collected in the current monitoring cycle belongs.

[0035] A third aspect of this application provides a computer device, comprising:

[0036] The memory is configured to store instructions; and

[0037] The processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement the online intelligent diagnosis method for ultra-high frequency partial discharge faults of a hydro-generator according to the first aspect of this application.

[0038] A fourth aspect of this application provides a machine-readable storage medium storing instructions for causing a machine to perform the online intelligent diagnosis method for ultra-high frequency partial discharge faults of a hydro-generator according to the first aspect of this application.

[0039] The above technical solution combines machine learning models, interference signal libraries, amplitude thresholds, and utilizes the characteristic of periodic repetition of partial discharge signals to achieve rapid and accurate identification of partial discharge signals, thereby enabling fast and accurate online diagnosis of partial discharge faults.

[0040] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0041] The accompanying drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the following detailed description to explain the embodiments of this application, but do not constitute a limitation on the embodiments of this application. In the drawings:

[0042] Figure 1 A flowchart illustrating an online intelligent diagnosis method for ultra-high frequency partial discharge faults of hydro-generators according to an embodiment of this application is shown.

[0043] Figure 2 This illustration schematically shows another flowchart of the online intelligent diagnosis method for ultra-high frequency partial discharge faults of hydro-generators according to an embodiment of this application;

[0044] Figure 3 This schematic diagram illustrates the composition of an online intelligent diagnostic device for ultra-high frequency partial discharge faults of a hydro-generator according to an embodiment of this application.

[0045] Figure 4 A schematic block diagram of a computer device according to an embodiment of this application is shown. Detailed Implementation

[0046] The specific embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the embodiments of this application.

[0047] If the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0048] Ultra-high frequency (UHF) partial discharge (PD) monitoring technology can provide data support for online preventive maintenance of hydroelectric generators. However, the challenge in its application lies in how to accurately identify PD signals and determine PD events using UHF PD data.

[0049] To apply ultra-high frequency partial discharge monitoring technology for online diagnosis of partial discharge faults in hydro-generators, this application proposes an online intelligent diagnosis method for ultra-high frequency partial discharge faults in hydro-generators, the method comprising:

[0050] Acquire the monitoring signals collected by the UHF partial discharge coupling device during the current monitoring period;

[0051] If the conditions for enabling the interference signal library are met, the amplitude sequence obtained after discretizing the monitoring signal is input into the pre-built machine learning model to generate a diagnostic result on whether the hydro-generator is working properly. If the abnormal working state of the hydro-generator is diagnosed and the monitoring signal is determined not to be an interference signal using the pre-built interference signal library, the process jumps to trigger the next step. Otherwise, the interference signal types in the interference signal library are accumulated, and the process jumps to trigger the next step.

[0052] Determine whether the amplitude of the monitoring signal acquired in the current monitoring cycle is greater than the amplitude threshold. If so, and the number of monitoring cycles in which the amplitude of the corresponding monitoring signal is greater than the amplitude threshold has been accumulated to the cycle accumulation threshold, jump to trigger the next step. Otherwise, enter the next monitoring cycle and jump to trigger the acquisition of the monitoring signal acquired by the ultra-high frequency partial discharge coupling device in the current monitoring structure.

[0053] Identify the partial discharge event to which the monitoring signals collected in the current monitoring period belong.

[0054] In this application, obtaining an amplitude sequence by discretizing the monitoring signal specifically means: discretizing the monitoring signal to obtain a pulse sequence, and the amplitude sequence is composed of the amplitudes of each pulse in the pulse sequence.

[0055] As in the above embodiment, when the interference signal library activation condition is not met, while accumulating the types of interference signals in the interference signal library, based on the characteristic of the periodic recurrence of partial discharge signals, it is determined whether the monitoring signals collected in multiple consecutive monitoring cycles are all greater than the amplitude threshold to determine whether the monitoring signal is a partial discharge signal. When the interference signal library activation condition is met, a machine learning model is used to diagnose whether the hydro-generator is working normally in the current monitoring cycle. If the abnormal working state of the hydro-generator is diagnosed, the interference signal library is used to determine whether the monitoring signal is a partial discharge signal or an interference signal. If it is identified that the monitoring signal collected in the current monitoring cycle is not an interference signal, the periodic recurrence of partial discharge signals and the amplitude threshold are further used to determine whether the monitoring signal is a partial discharge signal, thus achieving partial discharge signal identification and subsequent partial discharge fault diagnosis. It can be seen that the embodiments of this application combine machine learning models, interference signal libraries, amplitude thresholds, and the characteristic of periodic recurrence of partial discharge signals to achieve rapid and accurate identification of partial discharge faults at different diagnostic stages.

[0056] It is important to understand that the diagnosis of UHF partial discharge faults at different diagnostic stages specifically refers to the following: Because the types of interference signals in the interference signal library need to be accumulated gradually, in the early stages of UHF partial discharge fault diagnosis, after the machine learning model determines that the turbine generator's operating state is abnormal, it is difficult to distinguish between partial discharge signals and interference signals using the interference signal library. Therefore, the amplitude threshold and the characteristic of the periodic repetition of partial discharge signals are combined to identify partial discharge signals. After the types of interference signals in the interference signal library accumulate to a certain extent, the interference signal library is activated. At this point, the machine learning model and the interference signal library are used to initially identify whether the monitored signal is a partial discharge signal. This is because the monitored signal still contains interference such as periodic pulse interference after preprocessing. That is, when the machine learning model diagnoses that the turbine generator's operating state is abnormal in the current monitoring cycle, the monitored signal collected in the current monitoring cycle can be a power frequency cycle carrying interference signals, a power frequency cycle that is interfered with and carries partial discharge pulses, or a power frequency cycle that is not interfered with and carries partial discharge pulses. Then, the amplitude threshold and the characteristic of the periodic repetition of partial discharge signals are combined to finally identify whether the monitored signal is a partial discharge signal.

[0057] Figure 1 A flowchart illustrating an online intelligent diagnosis method for ultra-high frequency partial discharge faults in hydro-generators according to an embodiment of this application is shown. Figure 1As shown, the online intelligent diagnosis method for ultra-high frequency partial discharge faults of hydro-generators may include steps 102 to 110. It is understood that the above-described online intelligent diagnosis method for ultra-high frequency partial discharge faults of hydro-generators may include all steps 102 to 110, or may include only some of them.

[0058] Step 102: Obtain the monitoring signal collected by the UHF partial discharge coupling device during the current monitoring period.

[0059] Specifically, in this application, the UHF partial discharge coupling device can be an UHF partial discharge sensor, an UHF partial discharge antenna, or a coupler composed of a transformer and a coupling capacitor, etc.

[0060] Step 104: Execute the first judgment process, which is as follows:

[0061] Determine whether the conditions for enabling the interference signal library are met; if so, preprocess and discretize the collected monitoring signals to obtain an amplitude sequence, input the amplitude sequence into a pre-built machine learning model to generate a diagnostic result of whether the hydro-generator is working properly, and jump to trigger the execution of step 106; otherwise, accumulate the interference signal types in the interference signal library and jump to trigger the execution of step 108.

[0062] Specifically, in the early stages of online diagnosis of partial discharge faults in this application, because interference signals have extremely strong randomness and randomness, the types of interference signals in the interference signal library are few. When there are few types of interference signals, the identification of partial discharge signals using the combination of machine learning models and interference signal libraries is ineffective. That is, the reliability of the distinction between partial discharge signals and interference signals is not high at this time. Therefore, it is necessary to first determine whether the conditions for enabling the interference signal library are met. The conditions for enabling the interference signal library mentioned in step 104 can be set according to specific needs.

[0063] As an example, the conditions for enabling the interference signal library include either the number of different types of interference signals in the interference signal library reaching a first preset value or the cumulative number of monitoring cycles reaching a second preset value.

[0064] As an example, the following process can be used to accumulate interference signal types in the interference signal library: Discretize the monitoring signals collected during the monitoring period, input the amplitude sequence obtained after discretization into a machine learning model, and when the diagnostic results generated by the machine learning model indicate that the hydro-generator is in an abnormal operating state, determine whether the monitoring signal is an interference signal based on the extracted energy characteristics of the monitoring signal. If no interference signal of this type is matched in the interference signal library for the current state, manually mark and classify the interference signal of this type, thereby realizing the accumulation of interference signal types.

[0065] Under the operating conditions of a hydro-generator, the types of interference signals in the interference signal library can include corona discharge, electromagnetic interference, pulse interference, interference signals caused by poor grounding during monitoring signal acquisition, electronic noise from the acquisition equipment itself during monitoring signal acquisition, environmental noise, erroneous signals generated by the acquisition equipment due to temperature changes during monitoring signal acquisition, false discharge signals caused by humidity changes, signal transmission interference, and human interference. The sources of these interference signals are shown in Table 1 below.

[0066] Table 1

[0067]

[0068] As an example, a machine learning model can be obtained by training and optimizing the hyperparameters of an initial classification network. The classification network can be an RBF network, an extreme learning machine, a random forest, etc.

[0069] As an example, preprocessing may include bandpass filtering, normalization, etc.

[0070] For example, normalization is performed using the following formula:

[0071] ;

[0072] in, x(t) This represents the monitoring signal after bandpass filtering, where max(|x(t)|) is the signal... x(t) Maximum absolute amplitude, normalized signal x b (t) The range is [-1, 1].

[0073] Step 106: Execute the second judgment process, which is as follows:

[0074] When the diagnostic results generated by the machine learning model indicate that the hydro generator is in an abnormal operating state, the existing interference signal library is used to determine whether the monitoring signal collected in the current monitoring cycle is an interference signal. If so, the partial discharge fault diagnosis of the current monitoring cycle ends and the next monitoring cycle begins; otherwise, the process jumps to trigger step 108.

[0075] As an optional embodiment of this application, based on the first matching process, the monitoring signal collected in the current monitoring period is determined not to be an interference signal using the constructed interference signal library. The first matching process specifically refers to:

[0076] Extract the first feature of the monitoring signal collected in the current monitoring period;

[0077] Determine whether the difference between the first statistical value and the first feature of the extracted monitoring signal is within a preset range. If not, determine that the monitoring signal collected in the current monitoring cycle is not an interference signal.

[0078] Specifically, in this application, the first feature includes at least an energy feature. For example, the energy feature is power spectral density. Distinguishing between partial discharge signals and interference signals based on the significant difference in energy density between them utilizes this principle. The first statistical value is the statistical value of the first feature corresponding to each interference signal in the constructed interference signal library. The type of the first statistical value includes at least one of the following: the average value of the first features of all types of interference signals in the interference signal library; the maximum value of the first features of all types of interference signals in the interference signal library; and the minimum value of the first features of all types of interference signals in the interference signal library. For example, as an example, the first statistical value is the maximum value of the first features corresponding to each interference signal in the constructed interference signal library.

[0079] Step 108: Execute the third judgment process, which is as follows:

[0080] Determine whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold. If the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold, and the number of monitoring periods in which the amplitude of the corresponding monitoring signal is greater than the amplitude threshold has been accumulated to the period accumulation threshold, jump to trigger the execution of step 110. Otherwise, enter the next monitoring period and jump to trigger the execution of step 102.

[0081] In a preferred embodiment of this application, the amplitude threshold can be a preset fixed amplitude threshold or an adaptive amplitude threshold. The selection of the preset fixed amplitude threshold and the adaptive amplitude threshold can be determined in combination with the number of interference sources under the specific operating conditions of the hydro-generator. When the number of interference sources is relatively simple, the preset fixed amplitude threshold is selected, and under the background of complex interference sources, the adaptive amplitude threshold is selected.

[0082] For example, determining whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold includes the following steps:

[0083] The interference status at the production site where the UHF partial discharge coupling device is located was obtained;

[0084] Analyze the interference status to determine the number of interference sources that have been collected at the production site;

[0085] If the number of interference sources collected at the production site is greater than the third preset value, an amplitude threshold will be generated adaptively; otherwise, a preset amplitude threshold based on expert experience will be obtained.

[0086] Determine whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold.

[0087] As described in the above embodiments, the amplitude threshold based on expert experience refers to the empirical amplitude threshold. By determining whether to use a preset fixed amplitude threshold or an adaptively generated amplitude threshold according to the state of the interference source at the production site, the accuracy of partial discharge signal identification is improved.

[0088] As an example, unsupervised clustering methods can be used to adaptively generate amplitude thresholds.

[0089] As another example, the Ostu thresholding method can be used to adaptively generate the amplitude threshold, as follows:

[0090] The monitoring signals collected in the current monitoring cycle are discretized to obtain a pulse sequence;

[0091] Find the amplitude threshold that maximizes the inter-class variance. The amplitude threshold divides the amplitude of each pulse in the pulse sequence into two categories, i.e., into two classes. The inter-class variance is the inter-class variance of the two classes.

[0092] Step 110: Identify the partial discharge event to which the monitoring signal collected in the current monitoring cycle belongs.

[0093] It is important to understand that the monitoring signals collected by the UHF partial discharge coupling device differ under different partial discharge events. The specific partial discharge event to which the monitoring signal belongs refers to the specific partial discharge event under which the monitoring signal was generated.

[0094] As an optional embodiment of this application, the partial discharge event to which the monitoring signal belongs can be identified using an existing partial discharge case library. Accordingly, step 110 includes the following sub-steps:

[0095] The feature parameters of the monitoring signal collected in the current monitoring period are extracted, and the partial discharge events corresponding to the feature parameters of the extracted monitoring signal are matched from the constructed partial discharge case library. Each partial discharge event in the partial discharge case library has its own corresponding feature parameters.

[0096] Specifically, the partial discharge signals corresponding to different types of partial discharge events differ in the time domain and / or frequency domain. Specifically, the amplitude, frequency, and waveform of the partial discharge signal change with the location and intensity of the partial discharge. Therefore, as mentioned above, the characteristic parameters refer to these features that can reflect the differences of the partial discharge signal in the time domain or frequency domain. As in the above embodiment, the specific time domain features and / or frequency domain features used are not limited. Feasible time domain features and / or frequency domain features need to be determined in combination with the specific type of hydropower turbine generator unit of the hydropower station.

[0097] For example, characteristic parameters include time-domain features such as phase, amplitude, waveform characteristics, and amplitude statistics. Waveform characteristics include one or more of rise time, fall time, duration, and pulse width. Amplitude statistics include one or more of the average amplitude and standard deviation of amplitude. Characteristic parameters can also include frequency-domain features such as spectral characteristics, which refer to the intensity and frequency distribution of different frequency components in the spectrum.

[0098] As an improved embodiment of this application, an amplitude threshold is used to remove interfering data points. Accordingly, the characteristic parameters of the monitoring signal collected in the current monitoring period are extracted in the following manner:

[0099] The monitoring signal is discretized to obtain a pulse sequence;

[0100] Pulses with amplitudes below the amplitude threshold in the pulse sequence are removed to obtain the target pulse sequence, which is then used as the discretized monitoring signal.

[0101] Extract the characteristic parameters of the discretized monitoring signal.

[0102] In the above improved embodiment, interference data points are removed by using an amplitude threshold, thereby improving the efficiency and accuracy of feature parameter extraction.

[0103] It is important to understand that the interference signal types in the partial discharge (PD) case library are continuously updated. When the current PD case library is insufficient to classify the monitored signal, manual analysis is required to classify the signal, and the newly classified PD events are added to the PD case library. Manual analysis to determine new PD event types can be performed using offline PD testing methods. For example, PD tests can be conducted offline at various UHF PD coupling device locations on the stator winding, collecting PD signals from each location and extracting their characteristic parameters. By matching these extracted characteristic parameters with the characteristic parameters of the monitoring signal to be classified, the PD event type of the current monitoring signal can be determined. Preferably, before extracting the characteristic parameters of these PD signals, the PD signals collected at each location are corrected using a maximum attenuation coefficient. The maximum attenuation coefficient can be determined through the process described in the standard embodiment.

[0104] For example, test partial discharge signals are injected into each preset injection point of the stator winding bar of the hydro generator. Then, the partial discharge signals are collected at the locations of the ultra-high frequency partial discharge coupling device. The transfer function and attenuation coefficient of the stator winding bar are determined by a network vector analyzer, thereby determining the maximum attenuation coefficient. The partial discharge signals collected at each location are reconstructed through the transfer function, and the reconstructed partial discharge signals are multiplied by the maximum attenuation coefficient to complete the correction of the partial discharge signals.

[0105] As an improved embodiment of this application, Figure 2 This schematically illustrates another flowchart of the online intelligent diagnosis method for ultra-high frequency partial discharge faults of hydro-generators according to embodiments of this application, such as... Figure 2 As shown, this improved embodiment is similar to Figure 1 The difference in the illustrated embodiment is that step 103 is included before step 104.

[0106] Step 103, perform a shutdown diagnostic. The shutdown diagnostic specifically includes:

[0107] Determine whether the amplitude of the monitoring signal collected in the current monitoring cycle meets the shutdown conditions; if so, issue a control signal that the hydro-generator needs to be shut down, and then jump to trigger the execution of step 104; otherwise, jump directly to trigger the execution of step 104.

[0108] In this application, the shutdown conditions include at least the monitoring signal amplitude being greater than a second statistical value and the monitoring signal amplitude being greater than one of a plurality of maximum partial discharge amplitudes. The maximum partial discharge amplitude is the maximum value of the partial discharge amplitude of the partial discharge event in the corresponding constructed laboratory severe partial discharge case library. Each partial discharge event in the laboratory severe partial discharge case library has its own corresponding maximum partial discharge amplitude. The second statistical value is the statistical value of each maximum partial discharge amplitude. The type of the second statistical value includes at least one of the maximum value among the maximum partial discharge amplitudes and the average value among the maximum partial discharge amplitudes. The laboratory severe partial discharge case library includes at least partial discharge events that occur when the hydro-generator operates under extreme conditions.

[0109] In some comparative embodiments, when a partial discharge event is determined to have occurred within the current monitoring period through single-unit trend comparison, it is impossible to provide a control signal indicating whether the hydro-generator needs to be shut down. However, in the embodiments described above, this application pre-constructs a laboratory severe partial discharge case library, matches the amplitude of the monitoring signal collected in the current monitoring period with the laboratory severe partial discharge case library to determine whether the shutdown conditions are met, and issues a control signal indicating that the hydro-generator needs to be shut down when the shutdown conditions are met. This avoids insulation deterioration, equipment damage, or even system collapse caused by the hydro-generator continuing to operate and the partial discharge problem not being addressed in a timely manner, thus avoiding production interruptions and economic losses caused by the aforementioned problems.

[0110] As an example, when using the established partial discharge case library to identify the partial discharge event to which the monitoring signal belongs, the partial discharge case library may include all partial discharge events in the aforementioned laboratory severe partial discharge case library and other ordinary partial discharge events, etc. Ordinary partial discharge events refer to partial discharge events of hydro-generators under normal operating conditions.

[0111] As an example, one type of partial discharge (PD) event in the laboratory severe PD case library is a PD case under extreme humidity conditions. Specifically, this involves simulating partial discharge in a high-humidity environment because moisture can significantly alter the properties of insulating materials, increasing the probability of PD occurrence. The maximum PD amplitude under this condition is obtained through PD simulation. Another typical PD event in the laboratory severe PD case library is a PD case under high-temperature conditions. Specifically, this involves simulating PD under conditions higher than normal operating temperatures because high temperatures may accelerate the aging of insulating materials, thereby triggering PD. The maximum PD amplitude under this condition is obtained through PD simulation. A third PD event in the laboratory severe PD case library is a PD case under strong electromagnetic interference environments. Specifically, this involves simulating PD in an environment with strong electromagnetic interference. In specific applications, the complex electromagnetic interference environment of a real power system is simulated to obtain the maximum PD amplitude under this condition. Another type of partial discharge (PD) event in the laboratory severe PD case library is the PD case resulting from insulation material aging. Specifically, this involves using aged insulation materials or accelerated aging experiments to simulate the long-term operation of a hydro-generator, examining the impact of insulation aging on PD, and obtaining the maximum PD amplitude under these conditions. Another type of PD event in the laboratory severe PD case library is the PD case caused by mechanical stress on the hydro-generator. Specifically, during PD testing of the hydro-generator, mechanical vibration or impact is applied to the hydro-generator. Because mechanical stress or impact may cause damage to the insulation structure, resulting in PD, the maximum PD amplitude under these conditions is obtained by examining the PD caused by this insulation damage. A third type of PD event in the laboratory severe PD case library is the PD case caused by different types of insulation defects. Specifically, this involves constructing insulation defect models with different types and sizes, such as bubbles, cracks, and impurities, examining the impact of these different insulation defects on PD, and obtaining the maximum PD amplitude under these conditions. Another type of partial discharge (PD) event in the laboratory severe PD case library is the PD case of a hydro-generator under extreme operating conditions. Specifically, this involves simulating extreme operating conditions of the hydro-generator, such as rapid voltage increase, rapid voltage decrease, and overload, as these extreme conditions can exacerbate PD phenomena. The maximum PD amplitude under these extreme conditions is obtained through PD simulation. Another type of PD event in the laboratory severe PD case library is the PD case under combined fault conditions. Specifically, this involves conducting PD simulation tests on a hydro-generator under multiple extreme conditions simultaneously, such as high temperature, high humidity, and mechanical stress. The combined impact of these combined fault conditions on PD is examined, and the maximum PD amplitude under these conditions is obtained. Table 2 below shows the PD event types in the laboratory severe PD case library constructed in a specific application example.

[0112] Table 2

[0113]

[0114] As is known, in actual production, the amount of signal data collected by the UHF partial discharge monitoring method is enormous, and most of it consists of signals from the normal operating state of hydroelectric generators. Therefore, there is a lack of abundant partial discharge signal samples that can be used to train the classification network. Consequently, based on the training methods in the common embodiments, the model is prone to overfitting and poor generalization ability. Therefore, as an improvement to the above embodiments of this application, this application proposes a method for constructing a machine learning model based on multi-stage training. Accordingly, after training and hyperparameter optimization of the classification network, a machine learning model for signal diagnosis is obtained, specifically including the following steps:

[0115] Step A1: The classification network is trained in the first stage using the first training set to obtain the first diagnostic model. The first training set consists of the amplitude sequence obtained by discretizing the normal operating state signal of the hydro-generator.

[0116] In this application, the samples used in the first stage of training can be referred to as positive samples.

[0117] Step A2: Use the second training set to train the first diagnostic model in the second stage to obtain the second diagnostic model. The second training set consists of a specific amplitude sequence, and the historical monitoring signal corresponding to the specific amplitude sequence is the abnormal working state signal of the hydro-generator.

[0118] In this application, the samples used in the second stage of training can be referred to as negative samples.

[0119] Step A3: Use the third training set to train the second diagnostic model in the third stage. The third training set consists of a target amplitude sequence, which is a historical amplitude sequence that meets preset conditions.

[0120] Specifically, the preset conditions mentioned in this application refer to: the diagnostic result obtained by the input second diagnostic model is that the hydro-generator is in an abnormal operating state, and then the monitoring signal corresponding to the input amplitude sequence is determined by manual inspection to be an abnormal operating state signal of the hydro-generator. The samples used in the third stage of training are the samples obtained after being identified by the second diagnostic model and manually labeled, which are also negative samples.

[0121] In the improved embodiment described above, multi-stage training fully utilizes both positive and negative samples. Furthermore, the third training set consists of expanded negative samples identified by the second diagnostic model and manually labeled. This sample expansion is more efficient than directly manually labeling negative samples from a large amount of historical monitoring signals in the ordinary embodiment. Therefore, multi-stage training improves the efficiency of building the machine learning model and the recognition accuracy of the resulting model.

[0122] In summary, the above-described improved embodiments enhance the efficiency and accuracy of partial discharge identification.

[0123] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0124] Corresponding to the above embodiment of the online intelligent diagnosis method for ultra-high frequency partial discharge faults in hydro-generators, Figure 3 The diagram illustrates the components of the online intelligent diagnostic device 400 for ultra-high frequency partial discharge faults of hydro-generators provided in this application embodiment. For ease of explanation, only the parts relevant to this application embodiment are shown.

[0125] like Figure 3 As shown, the online intelligent diagnostic device 400 for ultra-high frequency partial discharge faults of hydro-generators includes:

[0126] The acquisition module 410 is used to acquire the monitoring signals collected by the UHF partial discharge coupling device during the current monitoring period;

[0127] The first judgment module 420 is used to determine whether the conditions for enabling the interference signal library are met. If so, the amplitude sequence obtained after discretizing the monitoring signal is input into the pre-built machine learning model to generate a diagnostic result of whether the hydro-generator is working normally. When the abnormal working state of the hydro-generator is diagnosed and the monitoring signal is determined not to be an interference signal by using the pre-built interference signal library, the module jumps to trigger the execution of judging whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold. Otherwise, the module accumulates the interference signal types in the interference signal library and jumps to trigger the execution of judging whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold.

[0128] The second judgment module 430 is used to determine whether the amplitude of the monitoring signal collected in the current monitoring cycle is greater than the amplitude threshold. If so, and the number of monitoring cycles in which the amplitude of the corresponding monitoring signal is greater than the amplitude threshold is continuously accumulated to reach the cycle accumulation threshold, the module jumps to trigger the execution of identifying the partial discharge event to which the monitoring signal collected in the current monitoring cycle belongs. Otherwise, the module enters the next monitoring cycle and jumps to trigger the execution of acquiring the monitoring signal collected by the UHF partial discharge coupling device in the current monitoring cycle.

[0129] The identification module 440 is used to identify the partial discharge event to which the monitoring signal collected in the current monitoring cycle belongs.

[0130] The process by which each module in the online intelligent diagnostic device 400 for ultra-high frequency partial discharge faults of hydro-generators provided in this application implements its respective function can be specifically referred to the foregoing. Figure 1 The descriptions of the embodiments shown and other related method embodiments are not repeated here.

[0131] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. Their specific functions and technical effects can be found in the method embodiments section, and will not be repeated here.

[0132] Figure 4 A schematic block diagram of a computer device according to an embodiment of the present application is shown. In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown below. Figure 4 As shown in the figure, the computer device includes a processor A01, a network interface A02, a display screen A04, an input device A05, and a memory (not shown) connected via a system bus. The processor A01 provides computing and control capabilities. The memory includes internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 stored in the non-volatile storage medium A06. The network interface A02 is used for communication with external terminals via a network connection. When the computer program is executed by the processor A01, it implements an online intelligent diagnostic method for ultra-high frequency partial discharge faults in hydroelectric generators. The display screen A04 can be a liquid crystal display (LCD) or an e-ink display. The input device A05 can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0133] Those skilled in the art will understand that Figure 4The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0134] In one embodiment, the online intelligent diagnostic device 400 for ultra-high frequency partial discharge faults of hydro-generators provided in this application can be implemented as a computer program, which can be implemented in various ways, such as... Figure 4 The computer device shown operates on this device. The computer device's memory can store various program modules that constitute the online intelligent diagnostic device 400 for ultra-high frequency partial discharge faults in hydro-generators. The computer program, composed of these program modules, causes the processor to execute the steps in the online intelligent diagnostic method for ultra-high frequency partial discharge faults in hydro-generators described in the various embodiments of this application.

[0135] In one embodiment, this application also provides a machine-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the online intelligent diagnosis method for ultra-high frequency partial discharge faults of hydro-generators described in the above embodiments.

[0136] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0137] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0138] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for online intelligent diagnosis of ultra-high frequency partial discharge faults of a hydroelectric generator, characterized in that, The method comprises: acquiring monitoring signals collected by an ultra-high frequency partial discharge coupling device in a current monitoring period; if a condition for enabling the interference signal library is met, inputting an amplitude sequence obtained after discretizing the monitoring signals into a pre-constructed machine learning model to generate a diagnosis result of whether the hydroelectric generator is working normally, and if it is determined that the hydroelectric generator is in an abnormal working state and that the monitoring signals are not interference signals by using the interference signal library that has been constructed, jumping to trigger the execution of the next step, otherwise accumulating the types of interference signals in the interference signal library and jumping to trigger the execution of the next step; determining whether the amplitude of the monitoring signals collected in the current monitoring period is greater than an amplitude threshold value, if yes, and the number of monitoring periods in which the amplitudes of the corresponding monitoring signals are greater than the amplitude threshold value is continuously accumulated to reach a period accumulation threshold value, jumping to trigger the execution of the next step, otherwise entering the next monitoring period and jumping to trigger the execution of acquiring the monitoring signals collected by the ultra-high frequency partial discharge coupling device in the current monitoring period; identifying a partial discharge event to which the monitoring signals collected in the current monitoring period belong, specifically: extracting feature parameters of the monitoring signals collected in the current monitoring period, and matching a partial discharge event corresponding to the extracted feature parameters of the monitoring signals from a pre-constructed partial discharge case library, wherein each partial discharge event in the partial discharge case library has its own corresponding feature parameters.

2. The method according to claim 1, characterized in that, The condition for enabling the interference signal library includes one of that the number of types of interference signals in the interference signal library reaches a first preset value, and that the number of monitoring periods accumulated for execution reaches a second preset value.

3. The method according to claim 1, characterized in that, Based on the first matching process, it is determined that the monitoring signals are not interference signals by using the interference signal library that has been constructed, and the first matching process comprises: extracting a first feature of the monitoring signals, the first feature at least including an energy feature; determining whether the difference between a first statistical value and the first feature of the extracted monitoring signals is within a preset range, if not, determining that the monitoring signals are not interference signals, wherein the first statistical value is a statistical value of the first feature of each interference signal in the interference signal library, and the type of the first statistical value includes at least one of the average value of the first feature of all types of interference signals in the interference signal library, the maximum value of the first feature of all types of interference signals in the interference signal library, and the minimum value of the first feature of all types of interference signals in the interference signal library.

4. The method according to claim 3, characterized in that, The energy feature is a power spectral density.

5. The method according to claim 1, wherein, Determining whether the amplitude of the monitoring signals collected in the current monitoring period is greater than an amplitude threshold value comprises: learning the interference state of the production site where the ultra-high frequency partial discharge coupling device is located; analyzing the interference state to obtain the number of interference sources that have been collected in the production site; if the number of interference sources is greater than a third preset value, adaptively generating an amplitude threshold value, and determining whether the amplitude of the monitoring signals collected in the current monitoring period is greater than the amplitude threshold value, otherwise learning a preset amplitude threshold value based on expert experience, and determining whether the amplitude of the monitoring signals collected in the current monitoring period is greater than the amplitude threshold value.

6. The method according to claim 1, wherein, Extracting feature parameters of the monitoring signals collected in the current monitoring period comprises: discretizing the monitoring signals to obtain a pulse sequence; remove the pulses with the amplitudes lower than the amplitude threshold value in the pulse sequence to obtain a target pulse sequence, and take the target pulse sequence as the discretized monitoring signal; extract a characteristic parameter of the discretized monitoring signal.

7. The method according to claim 1 or 6, characterized in that, The characteristic parameter includes at least one of a phase, an amplitude, a waveform characteristic, an amplitude statistical characteristic, and a frequency spectrum characteristic.

8. The method according to claim 1, wherein, Further comprising: determining whether the amplitude of the monitoring signal meets a shutdown condition; if yes, issuing a control signal that the hydroelectric generator needs to be shut down, and then jumping to trigger the execution of: if the disturbance signal library enabling condition is met, inputting the amplitude sequence obtained after discretizing the monitoring signal into a pre-constructed machine learning model, otherwise jumping to trigger the execution of: if the disturbance signal library enabling condition is met, inputting the amplitude sequence obtained after discretizing the monitoring signal into a pre-constructed machine learning model; The shutdown condition at least includes that the amplitude of the monitoring signal is greater than a second statistical value, and the amplitude of the monitoring signal is greater than one of a plurality of maximum partial discharge amplitudes. The maximum partial discharge amplitude is the maximum value of the partial discharge amplitude of the partial discharge event in the self-corresponding constructed laboratory severe partial discharge case library, each partial discharge event in the laboratory severe partial discharge case library has its own corresponding maximum partial discharge amplitude, the second statistical value is a statistical value of each maximum partial discharge amplitude, and the type of the second statistical value at least includes one of the maximum value of each maximum partial discharge amplitude and the average value of each maximum partial discharge amplitude. The laboratory severe partial discharge case library at least includes a partial discharge event occurring under extreme conditions of the hydroelectric generator.

9. An online intelligent diagnosis device for ultra-high frequency partial discharge faults of a hydroelectric generator, characterized in that, Comprising: an acquisition module configured to acquire a monitoring signal collected by an ultra-high frequency partial discharge coupling device in a current monitoring period; a first determination module configured to determine whether a disturbance signal library enabling condition is met, if yes, input an amplitude sequence obtained after discretizing the monitoring signal into a pre-constructed machine learning model to generate a diagnosis result of whether the hydroelectric generator is working normally, and if the hydroelectric generator is determined to be in an abnormal working state and it is determined that the monitoring signal is not a disturbance signal by using the constructed disturbance signal library, jump to trigger the execution of determining whether the amplitude of the monitoring signal collected in the current monitoring period is greater than an amplitude threshold value, otherwise accumulate a type of disturbance signal in the disturbance signal library, and jump to trigger the execution of determining whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold value; a second determination module configured to determine whether the amplitude of the monitoring signal collected in the current monitoring period is greater than the amplitude threshold value, if yes, and the number of monitoring periods in which the amplitude of the corresponding monitoring signal is greater than the amplitude threshold value is continuously accumulated to reach a period accumulation threshold value, jump to trigger the execution of identifying a partial discharge event to which the monitoring signal collected in the current monitoring period belongs, otherwise enter a next monitoring period, and jump to trigger the execution of acquiring the monitoring signal collected by the ultra-high frequency partial discharge coupling device in the current monitoring period. The recognition module is used for recognizing a partial discharge event to which a monitoring signal collected in a current monitoring period belongs, specifically, a feature parameter of the monitoring signal collected in the current monitoring period is extracted, and a partial discharge event corresponding to the extracted feature parameter of the monitoring signal is matched from a constructed partial discharge case library, wherein each partial discharge event in the partial discharge case library has its own corresponding feature parameter.

10. A computer device, comprising: Comprising: a memory configured to store instructions; and a processor configured to call the instructions from the memory and capable of realizing the water-turbine-generator ultra-high-frequency partial-discharge fault online intelligent diagnosis method according to any one of claims 1 to 8 when executing the instructions.

11. A machine-readable storage medium, characterized in that, The machine-readable storage medium has instructions stored thereon, the instructions being used for causing a machine to execute the water-turbine-generator ultra-high-frequency partial-discharge fault online intelligent diagnosis method according to any one of claims 1 to 8.

Images