AC filter status diagnosis method based on recorded waveform image and artificial intelligence

Abstract

The present invention discloses an AC filter state diagnosis method based on waveform image recording and artificial intelligence, characterized in that the diagnosis method comprises the following steps: S1, obtaining a waveform data file of a certain fault recording from a fault recording device or a fault recording server of a converter station in an event-triggered or timed query manner, wherein the waveform data file is a file in COMTARDE format, which is a universal format for transient data exchange in a power system. The present invention can effectively avoid the tripping event of a circuit breaker caused by the legacy of the AC filter group of the converter station, and the present invention can analyze the health status of the AC filter group and its circuit breaker, and judge its operating conditions, so as to deal with the defects before the fault tripping occurs, and avoid the hidden danger of unsafe conditions to the power grid. At the same time, the use of an intelligent method can reduce the workload of manual analysis on the one hand, and improve the speed and accuracy of analysis on the other hand.

Description

Technical Field

[0001] The invention relates to the technical field of fault diagnosis, in particular to an AC filter state diagnosis method based on recorded waveform images and artificial intelligence. Background Art

[0002] In the DC transmission system, the converter will generate a large number of harmonics and consume a lot of reactive power when converting current. In order not to burden the power grid, each converter station will automatically put the corresponding AC filter group on the AC side according to the transmission power of the DC transmission system during operation, provide the reactive power required by the converter and eliminate harmonics to improve the power quality of the power grid. Due to the flexibility of DC transmission, the transmission power needs to be adjusted many times a day, so the AC filter is also frequently switched, making the switching frequency of its circuit breaker much higher than that of other line circuit breakers in the station. This undoubtedly puts pressure on the safe operation of the AC filter group and greatly tests the performance of the AC filter circuit breaker. At present, there have been many accidents in China where AC filters tripped and power was reduced due to circuit breaker failures, which greatly affected the level of DC power transmission.

[0003] In order to avoid the tripping of circuit breakers caused by abnormal AC filter groups in converter stations, a technical means to analyze the health status of AC filter groups and their circuit breakers is urgently needed to judge their operation status, so as to deal with defects before fault tripping occurs and avoid potential safety hazards to the power grid. At present, the analysis method of abnormal conditions of AC filter groups in converter stations mainly involves manual analysis of fault recording waveforms, and lacks automated intelligent analysis methods.

[0004] In view of the above situation, we need an AC filter status diagnosis method based on recorded waveform images and artificial intelligence. Summary of the invention

[0005] The purpose of the present invention is to provide an AC filter status diagnosis method based on recorded waveform images and artificial intelligence to solve the problems raised in the above background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solution: an AC filter status diagnosis method based on recorded waveform image and artificial intelligence, characterized in that the diagnosis method comprises the following steps:

[0007] S1. Obtaining a waveform data file of a certain fault recording from a fault recording device or a fault recording server of a converter station by event triggering or timing query. The waveform data file is a file in COMTARDE format, which is a common format for transient data exchange in power systems.

[0008] S2. Use known COMTARDE file analysis software to read the data of the corresponding COMTARDE format waveform data file, and convert the data into voltage analog signals, current analog signals and switch signals;

[0009] S3, automatically searching the moment when the switch signal converted in the fault recording file in step S2 changes, judging whether the circuit breaker is in the closed state or the open state, and further judging the recording type of the fault recording file according to the circuit breaker state, the recording type including closing recording, opening recording and other recordings,

[0010] When the circuit breaker is in the closed state, the fault recording file is recorded as the closed recording file, and the process goes to step S4.

[0011] When the circuit breaker is in the open state, the fault recording file is recorded as the open recording file, and the process goes to step S4.

[0012] When the circuit breaker is neither in the closed state nor in the open state, the fault recording file is recorded as other recordings and the analysis process is exited;

[0013] S4, recording the time when the switch signal changes, and extracting the data of the fault recording waveform in the corresponding time period according to the time;

[0014] S5. Retrieve the actual action moment of the circuit breaker, and generate multiple waveform images with different resolutions with the moment as the center. The resolution is determined by the number of data points displayed with the same size. The image is generated by displaying the data waveform in the form of a graphic through a drawing software or program and saving it as a file in a picture format;

[0015] S6, performing calculation and analysis on the extracted recorded wave data, and correspondingly generating a plurality of waveform images from the calculated data, wherein the calculation and analysis methods include: FFT analysis, wavelet analysis, and differential analysis;

[0016] The present invention uses FFT analysis, wavelet analysis, and differential analysis to analyze the recorded data in order to extract the characteristics of faults and anomalies. Different types of faults and anomalies have different effects when analyzed using different analysis methods. Therefore, multiple methods are used for analysis to obtain more comprehensive fault and anomaly feature points, making the final result more accurate.

[0017] S7, inputting the image files obtained in steps S5 and S6 into an artificial intelligence model for analysis, wherein the artificial intelligence model can learn automatically;

[0018] S8. Obtain output analysis results.

[0019] Furthermore, the converter station fault recording device in step S1 samples, records and stores the voltage waveform, current waveform and switch signal of the measurement points related to the AC filter when the AC filter circuit breaker of the converter station is actuated or other abnormal conditions occur, and then analyzes the equipment status and other abnormalities;

[0020] The fault recording server is a server used for storing and backing up fault recording data. The fault recording device can automatically store each recording data to the fault recording server.

[0021] Furthermore, the event triggering method in step S1 is to use each startup of the fault recording device as a condition for triggering analysis;

[0022] The method of the timed query is to access the fault recording device or the fault recording server at a fixed time period to determine whether there is new fault recording data. If there is new fault recording data, each fault recording data in the new fault recording data is analyzed in turn by executing the contents described in steps S2 to S8.

[0023] Furthermore, the method for determining whether the circuit breaker is in the closed state or the open state in step S3 is as follows:

[0024] The judgment is made by the time when the switch signal changes. The switch signal is a digital signal, that is, 0 or 1. For a certain switch signal,

[0025] When the switch signal changes from 0 to 1, it indicates the closing signal, that is, the circuit breaker is in the closed state.

[0026] When the switch signal changes from 1 to 0, it indicates a trip signal, that is, the circuit breaker is in the trip state;

[0027] The digital signal variation ranges corresponding to the closing signal and the opening signal may also be opposite, but both must be clearly defined.

[0028] In the present invention, the switch signal changing from 0 to 1 can represent both a closing signal and an opening signal. However, when the content represented is confirmed, the content represented by the corresponding switch signal changing from 1 to 0 is another signal, that is, when the switch signal changes from 0 to 1 to represent a closing signal, then the switch signal changes from 1 to 0 to represent an opening signal; similarly, when the switch signal changes from 0 to 1 to represent an opening signal, then the switch signal changes from 1 to 0 to represent a closing signal.

[0029] Furthermore, the switching quantity information change time in step S4 is the time when the switching quantity signal changes from 0 to 1 and from 1 to 0. For a recording file, assuming that the total number of recording points is N, this change time can be 0 to N-1, that is, the switching quantity signal changes at this point;

[0030] The data of the corresponding time period is data with a length of M data points starting from the time when the switch signal changes, and the length of M includes the transient data waveform of the filter short-circuit action.

[0031] Furthermore, in the step S5 of retrieving the actual action time of the circuit breaker, the time from the switch signal change to the actual action of different filter circuit breakers is different, and it is necessary to retrieve the actual action time of the circuit breaker from the switch signal change time backward.

[0032] When the fault recording file is a closing recording, the retrieval method is: starting from the moment when the switch signal changes, the first point where the change amplitude of the two points before and after the A, B, C three-phase current waveform corresponding to the recording exceeds the set first threshold is the actual action moment of the circuit breaker;

[0033] When the fault recording file is a trip recording, the retrieval method is: starting from the moment the switch signal changes, determine the first data point of the corresponding A, B, and C three-phase current waveforms that satisfies L consecutive values ​​that are all less than the set second threshold, which is the moment when the circuit breaker actually operates. The L is a preset value, and the second threshold value is close to 0.

[0034] In the present invention, after the actual action moment of the circuit breaker is retrieved, multiple waveform images are generated with different resolutions with this moment as the center because the length of the transient waveform that needs to be paid attention to is different for different types of abnormal situations, that is, the resolution of the waveform required is different. In addition, images with different resolutions can be displayed from different aspects such as macro and detail, which is convenient for analyzing waveform data.

[0035] Furthermore, the artificial intelligence model in step S7 is improved on the basis of an existing data-trained analysis model, so that the analysis model collects existing filter recording data files and known fault waveform cases, establishes a preliminary embedded artificial intelligence model through a machine learning algorithm, and continuously collects recording data.

[0036] After each waveform image is input into the artificial intelligence model, the analysis result will be obtained. At the same time, manual analysis and verification will be carried out. Then, the waveform image will be calibrated by manual analysis and then the artificial intelligence model will be trained. With the accumulation of data, the recognition standards and models will be continuously updated.

[0037] The collected waveform data also includes waveform images.

[0038] Furthermore, the method of outputting the analysis result in step S8 includes: file output and voice output.

[0039] If the output analysis result is determined to be abnormal, an alarm is issued for the output analysis result, and the alarm mode includes: voice alarm and switch signal alarm.

[0040] Compared with the prior art, the beneficial effects achieved by the present invention are: the present invention can effectively avoid the tripping of the circuit breaker caused by the legacy of the AC filter group in the converter station, and the present invention can analyze the health status of the AC filter group and its circuit breaker and judge its operation status, so as to deal with the defects before the fault tripping occurs, so as to avoid the hidden danger of unsafe conditions to the power grid. At the same time, the use of an intelligent approach can reduce the workload of manual analysis on the one hand, and improve the speed and accuracy of the analysis on the other hand. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0042] Figure 1 It is a structural schematic diagram of an AC filter state diagnosis method based on recorded waveform image and artificial intelligence of the present invention;

[0043] Figure 2 It is a schematic diagram of a waveform recording with a signal length of 200 data points corresponding to a resolution in the AC filter state diagnosis method based on waveform recording images and artificial intelligence of the present invention;

[0044] Figure 3 It is a schematic diagram of a waveform recording with a resolution corresponding to 50 data points in a waveform recording image and artificial intelligence based AC filter state diagnosis method of the present invention;

[0045] Figure 4 It is a schematic diagram of a recorded waveform with a resolution corresponding to 20 data points in a signal length in the AC filter state diagnosis method based on recorded waveform images and artificial intelligence of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] See also Figure 1-4 The present invention provides a technical solution: an AC filter status diagnosis method based on recorded waveform image and artificial intelligence, characterized in that the diagnosis method comprises the following steps:

[0048] S1. Obtaining a waveform data file of a certain fault recording from a fault recording device or a fault recording server of a converter station by event triggering or timing query. The waveform data file is a file in COMTARDE format, which is a common format for transient data exchange in power systems.

[0049] S2. Use known COMTARDE file analysis software to read the data of the corresponding COMTARDE format waveform data file, and convert the data into voltage analog signals, current analog signals and switch signals;

[0050] S3, automatically searching the moment when the switch signal converted in the fault recording file in step S2 changes, judging whether the circuit breaker is in the closed state or the open state, and further judging the recording type of the fault recording file according to the circuit breaker state, the recording type including closing recording, opening recording and other recordings,

[0051] When the circuit breaker is in the closed state, the fault recording file is recorded as the closed recording file, and the process goes to step S4.

[0052] When the circuit breaker is in the open state, the fault recording file is recorded as the open recording file, and the process goes to step S4.

[0053] When the circuit breaker is neither in the closed state nor in the open state, the fault recording file is recorded as other recordings and the analysis process is exited;

[0054] S4, recording the time when the switch signal changes, and extracting the data of the fault recording waveform in the corresponding time period according to the time;

[0055] S5. Retrieve the actual action moment of the circuit breaker, and generate multiple waveform images with different resolutions with the moment as the center. The resolution is determined by the number of data points displayed with the same size. The image is generated by displaying the data waveform in the form of a graphic through a drawing software or program and saving it as a file in a picture format;

[0056] S6, performing calculation and analysis on the extracted recorded wave data, and correspondingly generating a plurality of waveform images from the calculated data, wherein the calculation and analysis methods include: FFT analysis, wavelet analysis, and differential analysis;

[0057] S7, inputting the image files obtained in steps S5 and S6 into an artificial intelligence model for analysis, wherein the artificial intelligence model can learn automatically;

[0058] S8. Obtain output analysis results.

[0059] The converter station fault recording device in step S1 samples, records and stores the voltage waveform, current waveform and switch signal of the measurement points related to the AC filter when the AC filter circuit breaker of the converter station is actuated or other abnormal conditions occur, and then analyzes the equipment status and other abnormalities;

[0060] The fault recording server is a server used for storing and backing up fault recording data. The fault recording device can automatically store each recording data to the fault recording server.

[0061] The event triggering method in step S1 is to use each startup of the fault recording device as a condition for triggering analysis;

[0062] The method of the timed query is to access the fault recording device or the fault recording server at a fixed time period to determine whether there is new fault recording data. If there is new fault recording data, each fault recording data in the new fault recording data is analyzed in turn by executing the contents described in steps S2 to S8.

[0063] The method for determining whether the circuit breaker is in the closed state or the open state in step S3 is as follows:

[0064] The judgment is made by the time when the switch signal changes. The switch signal is a digital signal, that is, 0 or 1. For a certain switch signal,

[0065] When the switch signal changes from 0 to 1, it indicates the closing signal, that is, the circuit breaker is in the closed state.

[0066] When the switch signal changes from 1 to 0, it indicates a trip signal, that is, the circuit breaker is in the trip state;

[0067] The digital signal variation ranges corresponding to the closing signal and the opening signal may also be opposite, but both must be clearly defined.

[0068] In the present invention, the switch signal changing from 0 to 1 can represent both a closing signal and an opening signal. However, when the content represented is confirmed, the content represented by the corresponding switch signal changing from 1 to 0 is another signal, that is, when the switch signal changes from 0 to 1 to represent a closing signal, then the switch signal changes from 1 to 0 to represent an opening signal; similarly, when the switch signal changes from 0 to 1 to represent an opening signal, then the switch signal changes from 1 to 0 to represent a closing signal.

[0069] The switching quantity information change time in step S4 is the time when the switching quantity signal changes from 0 to 1 and from 1 to 0. For a recording file, assuming that the total number of recording points is N, this change time can be 0 to N-1, that is, the switching quantity signal changes at this point;

[0070] The data of the corresponding time period is data with a length of M data points starting from the time when the switch signal changes, and the length of M includes the transient data waveform of the filter short-circuit action.

[0071] In the step S5, when searching for the actual action time of the circuit breaker, the time from the switch signal change to the actual action time of different filter circuit breakers is different, and it is necessary to search for the actual action time of the circuit breaker from the switch signal change time backward.

[0072] When the fault recording file is a closing recording, the retrieval method is: starting from the moment when the switch signal changes, the first point where the change amplitude of the two points before and after the A, B, C three-phase current waveform corresponding to the recording exceeds the set first threshold is the actual action moment of the circuit breaker;

[0073] When the fault recording file is a trip recording, the retrieval method is: starting from the moment the switch signal changes, determine the first data point of the corresponding A, B, and C three-phase current waveforms that satisfies L consecutive values ​​that are all less than the set second threshold, which is the moment when the circuit breaker actually operates. The L is a preset value, and the second threshold value is close to 0.

[0074] In the present invention, after the actual action moment of the circuit breaker is retrieved, multiple waveform images are generated with different resolutions with this moment as the center because the length of the transient waveform that needs to be paid attention to is different for different types of abnormal situations, that is, the resolution of the waveform required is different. In addition, images with different resolutions can be displayed from different aspects such as macro and detail, which is convenient for analyzing waveform data.

[0075] The artificial intelligence model in step S7 is improved on the basis of an existing analysis model trained by data, so that the analysis model collects existing filter recording data files and known fault waveform cases, establishes a preliminary embedded artificial intelligence model through a machine learning algorithm, and continuously collects recording data.

[0076] After each waveform image is input into the artificial intelligence model, the analysis result will be obtained. At the same time, manual analysis and verification will be carried out. Then, the waveform image will be calibrated by manual analysis and then the artificial intelligence model will be trained. With the accumulation of data, the recognition standards and models will be continuously updated.

[0077] The collected waveform data also includes waveform images.

[0078] The method of outputting the analysis result in step S8 includes: file output and voice output.

[0079] If the output analysis result is determined to be abnormal, an alarm is issued for the output analysis result, and the alarm mode includes: voice alarm and switch signal alarm.

[0080] The specific implementation of the AC filter status diagnosis method based on recorded waveform images and artificial intelligence provided by the present invention can be a set of software programs running on a general-purpose computer or server, or it can be a device for protecting software and hardware combined with specific hardware. The specific hardware shown can be a high-speed DSP chip for data analysis, a single-core or multi-core ARM processor, FPGA, an artificial intelligence acceleration card, etc.

[0081] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0082] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An AC filter status diagnosis method based on recorded waveform image and artificial intelligence, characterized in that: The diagnostic method comprises the following steps: S1. Obtaining a waveform data file of a certain fault recording from a fault recording device or a fault recording server of a converter station by event triggering or timing query. The waveform data file is a file in COMTARDE format, which is a common format for transient data exchange in power systems. S2. Use known COMTARDE file analysis software to read the data of the corresponding COMTARDE format waveform data file, and convert the data into voltage analog signals, current analog signals and switch signals; S3, automatically searching the moment when the switch signal converted in the fault recording file in step S2 changes, judging whether the circuit breaker is in the closed state or the open state, and further judging the recording type of the fault recording file according to the circuit breaker state, the recording types including closing recording, opening recording and other recordings, When the circuit breaker is in the closed state, the fault recording file is recorded as the closed recording file, and the process goes to step S4. When the circuit breaker is in the open state, the fault recording file is recorded as the open recording file, and the process goes to step S4. When the circuit breaker is neither in the closed state nor in the open state, the fault recording file is recorded as other recordings and the analysis process is exited; S4, recording the time when the switch signal changes, and extracting the data of the fault recording waveform in the corresponding time period according to the time; S5. Retrieve the actual action moment of the circuit breaker, and generate multiple waveform images with different resolutions with the moment as the center. The resolution is determined by the number of data points displayed with the same size. The image is generated by displaying the data waveform in the form of a graphic through a drawing software or program and saving it as a file in a picture format; In the step S5, when searching for the actual action time of the circuit breaker, the time from the switch signal change to the actual action time of different filter circuit breakers is different, and it is necessary to search for the actual action time of the circuit breaker from the switch signal change time backward. When the fault recording file is a closing recording, the retrieval method is: starting from the moment when the switch signal changes, the first point where the change amplitude of the two points before and after the A, B, C three-phase current waveform corresponding to the recording exceeds the set first threshold is the actual action moment of the circuit breaker; When the fault recording file is a switch-off recording, the retrieval method is as follows: starting from the moment when the switch quantity signal changes, the first data point of the corresponding A, B, and C three-phase current waveforms that satisfies L consecutive values ​​that are all less than the set second threshold is determined as the actual action moment of the circuit breaker, where L is a preset value and the second threshold value is close to 0; S6, performing calculation and analysis on the extracted recorded wave data, and correspondingly generating a plurality of waveform images from the calculated data, wherein the calculation and analysis methods include: FFT analysis, wavelet analysis, and differential analysis; S7, inputting the image files obtained in steps S5 and S6 into an artificial intelligence model for analysis, wherein the artificial intelligence model can learn automatically; S8. Obtain output analysis results.

2. The AC filter status diagnosis method based on recorded waveform image and artificial intelligence according to claim 1 is characterized in that: The converter station fault recording device in step S1 samples, records and stores the voltage waveform, current waveform and switch signal of the measurement points related to the AC filter when the AC filter circuit breaker of the converter station is actuated or other abnormal conditions occur, and then analyzes the equipment status and other abnormalities; The fault recording server is a server used for storing and backing up fault recording data. The fault recording device can automatically store each recording data to the fault recording server.

3. The AC filter status diagnosis method based on recorded waveform image and artificial intelligence according to claim 1 is characterized in that: The event triggering method in step S1 is to use each startup of the fault recording device as a condition for triggering analysis; The method of the timed query is to access the fault recording device or the fault recording server at a fixed time period to determine whether there is new fault recording data. If there is new fault recording data, each fault recording data in the new fault recording data is analyzed in turn by executing the contents described in steps S2 to S8.

4. The AC filter status diagnosis method based on recorded waveform image and artificial intelligence according to claim 1 is characterized in that: The method for determining whether the circuit breaker is in the closed state or the open state in step S3 is as follows: The judgment is made by the time when the switch signal changes. The switch signal is a digital signal, that is, 0 or 1. For a certain switch signal, When the switch signal changes from 0 to 1, it indicates the closing signal, that is, the circuit breaker is in the closed state. When the switch signal changes from 1 to 0, it indicates a trip signal, that is, the circuit breaker is in the trip state; The digital signal variation ranges corresponding to the closing signal and the opening signal may also be opposite, but both must be clearly defined.

5. The AC filter status diagnosis method based on recorded waveform image and artificial intelligence according to claim 1 is characterized in that: The switching quantity information change time in step S4 is the time when the switching quantity signal changes from 0 to 1 and from 1 to 0. For a recording file, assuming that the total number of recording points is N, this change time can be 0 to N-1, that is, the switching quantity signal changes at this point; The data of the corresponding time period is data with a length of M data points starting from the time when the switch signal changes, and the length of M includes the transient data waveform of the filter short-circuit action.

6. The AC filter status diagnosis method based on recorded waveform image and artificial intelligence according to claim 1 is characterized in that: The artificial intelligence model in step S7 is improved on the basis of an existing analysis model trained by data, so that the analysis model collects existing filter recording data files and known fault waveform cases, establishes a preliminary embedded artificial intelligence model through a machine learning algorithm, and continuously collects recording data. After each waveform image is input into the artificial intelligence model, the analysis result will be obtained. At the same time, manual analysis and verification will be carried out. Then, the waveform image will be calibrated by manual analysis and then the artificial intelligence model will be trained. With the accumulation of data, the recognition standards and models will be continuously updated. The collected waveform data also includes waveform images.

7. The AC filter status diagnosis method based on recorded waveform image and artificial intelligence according to claim 1 is characterized in that: The method of outputting the analysis result in step S8 includes: file output and voice output. If the output analysis result is determined to be abnormal, an alarm is issued for the output analysis result, and the alarm mode includes: voice alarm and switch signal alarm.

Images