A PCS fault detection method and system for energy storage converter

Abstract

The present invention relates to a PCS fault detection method and system for an energy storage converter. The method comprises: using an audio acquisition device deployed in the PCS to acquire real-time operating audio data and environmental audio data within a preset time window to generate audio data to be diagnosed; utilizing a preset discrete event extraction mechanism to generate a discrete event segment sequence of the audio data to be diagnosed; extracting the temporal pattern characteristics of each discrete event segment based on the discrete event segment sequence to generate fault matching feature information for the discrete event segment sequence; and using the fault matching feature information to perform a comparison in a pre-established fault pattern library. If a known fault pattern is matched, the fault type and severity are determined. This improves the accuracy and timeliness of fault detection, providing important support for equipment maintenance and management.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a PCS fault detection method and system for an energy storage converter. Background Art

[0002] The Power Conversion System (PCS) is a key component in the entire energy storage system, and its operating status has a significant impact on the system's performance and safety. In actual applications, PCS failures are often not discovered until a certain period of time has passed. During this period, undetected failures can easily lead to accidents. In practice, the most common method for detecting PCS failures is for maintenance personnel to regularly check device status and alarm information. If an abnormality or alarm is detected, they contact professionals for troubleshooting and repair. However, this method requires users to frequently check and manually confirm whether a failure has occurred, making the overall process less intelligent and efficient. Summary of the Invention

[0003] The present invention addresses the technical problems existing in the prior art, improves the accuracy and timeliness of fault detection, and provides important support for equipment maintenance and management.

[0004] The present invention solves the above technical problems with the following technical solutions: A method for detecting PCS faults in an energy storage converter, comprising:

[0005] S101, based on the audio acquisition device deployed by the PCS, real-time acquisition of operating audio data and environmental audio data within a preset time window to generate audio data to be diagnosed;

[0006] S102, using a preset discrete event extraction mechanism to generate a discrete event segment sequence of the audio data to be diagnosed;

[0007] S103, based on the discrete event segment sequence, extracting the temporal pattern features of each discrete event segment, and generating fault matching feature information of the discrete event segment sequence;

[0008] S104 , using the fault matching feature information, a comparison is performed in a pre-established fault mode library. If a known fault mode is matched, the fault type and severity are determined.

[0009] Preferably, the audio collection device deployed by the PCS includes an internal audio collection device and an external audio collection device. The internal audio collection device is arranged inside the PCS and is used to collect operating audio data. The external audio collection device is arranged outside the PCS and is used to collect environmental audio data.

[0010] Preferably, the preset discrete event extraction mechanism specifically includes:

[0011] S201, performing short-time Fourier transform on the audio data to be diagnosed to obtain frequency distribution information;

[0012] S202, calculating, based on each frame of the audio signal, a rate of change of the frequency component thereof compared to the previous frame, i.e., a frequency change rate, obtained by comparing changes in amplitude values of the same frequency component of adjacent frames;

[0013] S203: Obtain the amplitude and frequency change rate of each frame of the audio data to be diagnosed, set preset thresholds, and determine the starting and ending points of the discrete event segment; the preset thresholds include amplitude thresholds and frequency mutation thresholds, which are set based on actual conditions and expert experience;

[0014] S204 , performing segmentation based on all the marked start points and end points of the audio data to be diagnosed, to obtain at least one discrete event segment, which constitutes a discrete time segment sequence.

[0015] Preferably, the timing pattern features include event duration, discrete interval duration, and event intensity value; event duration is the time length of a discrete event segment; discrete interval duration is the time interval between the discrete event segment and the previous discrete event segment; and event intensity value is the amplitude value of the discrete event segment.

[0016] Preferably, in said S103, generating the fault matching feature information of the discrete event segment sequence specifically includes:

[0017] S301, performing cluster analysis based on the temporal pattern characteristics of all discrete event segments to obtain several event types;

[0018] S302: For each cluster, count its occurrence frequency, occurrence interval sequence, and the central temporal pattern characteristics of the cluster to form the first characteristic information of the event type; the occurrence frequency is set as the ratio of the number of discrete event segments in the type to the total number of segments, and the occurrence interval sequence is set as the sequence consisting of the occurrence time intervals between adjacent discrete event segments in the type;

[0019] S303 , obtaining the event intensity value of each discrete event segment in the discrete event segment sequence, where the horizontal axis represents time and the vertical axis represents the time intensity value, and performing curve fitting to generate an intensity variation curve.

[0020] S304: Combine the first characteristic information of the event type and the intensity change curve into fault matching characteristic information.

[0021] Preferably, the pre-established fault mode library is constructed by:

[0022] S401, collecting various fault data that occurred during the past operation of the energy storage converter PCS, including operating audio data, fault type, fault severity, and operating conditions within the PCS within different historical preset time windows at the time of the fault and in the early period before the fault;

[0023] S402, generating operating audio data under different fault types through simulation experiments or emulation methods to enrich the data volume of the fault mode library;

[0024] S403: Generate corresponding fault matching feature information for the collected operating audio data within each historical preset time window, bind a label, and set the label content to the corresponding fault type, fault severity, and operating status parameters in the PCS;

[0025] S404 , classifying all fault matching feature information according to the fault type in the tag, obtaining a number of fault matching feature information corresponding to each fault type, and the fault severity and PCS internal operating status parameters bound to each fault matching feature information.

[0026] Preferably, in S304, when the number of event types in the discrete event segment sequence is greater than 1, the method further includes:

[0027] A1. Based on the pre-built discrete event stream analysis model, match the discrete event stream of the discrete event segment sequence. If the match is successful, execute step A2; otherwise, treat each event type as an independent individual and execute step A4.

[0028] A2. Based on all clusters of the discrete event segment sequence, first feature information corresponding to the clusters hit by the discrete event stream is combined into a first feature information group corresponding to the discrete event stream;

[0029] A3. Combining the first characteristic information group and the intensity change curve into fault matching characteristic information;

[0030] A4. The first characteristic information of each event type is combined with the intensity variation curve to form independent fault matching characteristic information, and the information is matched in the fault mode library respectively.

[0031] Preferably, the pre-built discrete event stream analysis model is specifically:

[0032] B1. Based on the pre-established fault mode library, obtain the discrete event segment sequences corresponding to all the collected operating audio data corresponding to all fault types, and generate the event type corresponding to each discrete event segment sequence;

[0033] B2. Filter target discrete event fragment sequences with more than one event type, count all event type combinations, calculate the frequency of occurrence of each event type combination, sort all occurrence frequencies in descending order, and eliminate event type combinations in the lower third to obtain target event type combinations and their corresponding target discrete event fragment sequences;

[0034] B3. Based on each target event type combination, obtain the discrete event streams corresponding to all corresponding target discrete event fragment sequences, that is, the occurrence chain of all discrete event fragments in chronological order;

[0035] B4. Generate a corresponding intensity change curve based on each discrete event stream corresponding to each target event type combination;

[0036] In A1, the discrete event stream matching the discrete event segment sequence includes:

[0037] C1. Obtain the intensity change curve of the discrete event segment sequence and match it with the intensity change curve of each discrete event flow in the discrete event flow analysis model;

[0038] C2. If there is an overlapping segment between the two curves, the matching is successful; otherwise, the matching fails.

[0039] C3. Obtain overlapping segments of the discrete event fragment sequence, and use the overlapping segments as the discrete event stream of the discrete event fragment sequence.

[0040] Preferably, the internal audio collection device is configured as a plurality of microphones with uniform coverage;

[0041] The S101 further includes:

[0042] Obtain the operating audio data collected by each microphone and its corresponding audio data to be diagnosed, as well as the key object corresponding to each audio data to be diagnosed;

[0043] The key object is set as the component object corresponding to the key sound source direction of the audio data to be diagnosed;

[0044] The S102 further includes:

[0045] All discrete event segment sequences of the audio data to be diagnosed are spliced and combined according to the sampling time.

[0046] A PCS fault detection system for an energy storage converter includes: an acquisition module, an extraction module, and a matching module;

[0047] The acquisition module is used to acquire the operating audio data and environmental audio data within a preset time window in real time based on the audio acquisition device deployed by the PCS, and generate audio data to be diagnosed;

[0048] The extraction module is used to generate a discrete event segment sequence of the audio data to be diagnosed using a preset discrete event extraction mechanism; based on the discrete event segment sequence, the temporal pattern features of each discrete event segment are extracted to generate fault matching feature information of the discrete event segment sequence;

[0049] The matching module is used to use the fault matching feature information to compare with the pre-established fault mode library. If a known fault mode is matched, the fault type and severity are determined.

[0050] The beneficial effects of the present invention are:

[0051] Traditional methods have difficulty capturing discontinuous changes in sound signals, resulting in insufficient early fault warning capabilities. By capturing discontinuous changes in sound signals, early fault warnings can be provided, and abnormal event patterns can be discovered to improve the accuracy and predictability of fault prediction. Based on the frequency distribution information of the audio data to be diagnosed, thresholds are set to identify and segment discrete acoustic events. This can capture discontinuous changes in sound signals, such as sudden knocking sounds, intermittent friction sounds, and other early fault indicators, thereby improving the sensitivity of fault detection. The temporal pattern features of discrete acoustic events are extracted to generate fault matching feature information for the audio data to be diagnosed. By matching with a pre-established fault pattern library, it can determine whether abnormal patterns exist, improving the accuracy and timeliness of fault detection and providing important support for equipment maintenance and management.

[0052] Taking into account the coherence between different discrete acoustic events and possible early fault development chains, by constructing a discrete event stream analysis model, it is possible to more accurately identify discrete event combinations related to faults, thereby improving the accuracy of fault detection; by identifying discrete event streams, it is possible to discover potential fault development chains in advance, thereby providing early warning before the fault occurs, enhancing the predictability of fault warnings; matching of discrete event streams is performed in the fault matching feature information generation stage, reducing the number of matches in the fault pattern library, thereby improving the efficiency of fault matching; in actual operation, the failure of the energy storage converter PCS may be caused by a combination of multiple discrete acoustic events. By considering discrete event streams, it is possible to better adapt to such complex fault scenarios and improve the comprehensiveness and accuracy of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 Schematic diagram of a flow chart of a method for detecting PCS faults in an energy storage converter according to an embodiment of the present invention;

[0054] Figure 2 The figure is a structural diagram of a PCS fault detection system for an energy storage converter according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0056] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the described features. In the description of this application, "plurality" means two or more, unless otherwise specifically specified.

[0057] In the description of this application, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art will recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in this application.

[0058] Example 1

[0059] Figure 1 The present invention is a flowchart of a method for detecting a PCS fault in an energy storage converter according to an embodiment of the present invention.

[0060] like Figure 1 As shown, a method for detecting PCS faults of an energy storage converter includes:

[0061] S101, an audio acquisition device deployed based on the PCS acquires operating audio data and environmental audio data within a preset time window in real time to generate audio data to be diagnosed.

[0062] Among them, the audio collection device deployed by the PCS includes an internal audio collection device and an external audio collection device. The internal audio collection device is set inside the PCS and is used to collect running audio data. The external audio collection device is set outside the PCS and is used to collect environmental audio data.

[0063] The preset time window is set according to actual needs and scenarios, for example, audio data of minutes is sampled each time.

[0064] Specifically, generating the audio data to be diagnosed includes: performing noise reduction processing on the operating audio data using characteristic information of the environmental audio data to obtain the audio data to be diagnosed.

[0065] As an example, the noise reduction processing can be set as follows: feature extraction of environmental audio data, and spectrum analysis and other methods can be used to obtain environmental feature information, such as the frequency distribution of environmental audio, energy concentration frequency band, etc.; spectrum analysis is performed on at least one audio segment in the running audio data to determine the noise information (such as the frequency range and intensity of the noise) and frequency feature information (such as the amplitude of different frequency components, etc.) in the running audio data; based on the environmental feature information, noise information and frequency feature information, noise reduction processing is performed on each audio segment in the running audio data, that is, the environmental noise frequency band in the running audio data is removed to obtain the audio data to be diagnosed.

[0066] It should be noted that the above-mentioned noise reduction process is only an example. Other noise reduction methods can refer to relevant existing technologies, and the present invention does not elaborate on or limit them.

[0067] S102 , using a preset discrete event extraction mechanism to generate a discrete event segment sequence of audio data to be diagnosed.

[0068] The discrete-time extraction mechanism is used to capture discontinuous changes in sound signals, such as sudden knocking sounds and intermittent friction sounds, which are often important indicators of early faults. It can classify and count discrete acoustic events and discover abnormal event patterns. Discrete event segments are segmented from the denoised audio data to be diagnosed. Each segment represents a specific discrete acoustic event, which may be related to the operating status or fault of internal PCS components. These events are captured and segmented into discrete event segments. All discrete event segments are organized into a discrete time segment sequence based on the sampling time sequence.

[0069] In some embodiments, the preset discrete event extraction mechanism specifically includes:

[0070] S201 , performing short-time Fourier transform on the audio data to be diagnosed to obtain frequency distribution information.

[0071] Exemplarily, a suitable window function (such as a Hanning window) and window length are selected, the audio data is framed, and the Fourier transform of each frame of data is calculated to obtain frequency distribution information, including frequency components and their corresponding amplitude values.

[0072] S202 , calculating the rate of change of the frequency component of each frame of the audio signal compared with the previous frame, that is, the frequency change rate, by comparing the amplitude value changes of the same frequency component of adjacent frames.

[0073] It should be noted that in order to obtain a comprehensive frequency change rate index, the change rates of all frequency components of interest can be weighted and summed or averaged. The weights can be set according to actual needs, such as giving a higher weight to changes in a specific frequency range.

[0074] S203: Obtain the amplitude and frequency change rate of each frame of the audio data to be diagnosed, set preset thresholds, and determine the starting and ending points of the discrete event segment. The preset thresholds include amplitude thresholds and frequency mutation thresholds, which are set based on actual conditions and expert experience and are not limited in the present invention.

[0075] Specifically include:

[0076] Amplitude threshold setting: Set an amplitude threshold based on the amplitude distribution of the audio signal. When the amplitude exceeds the threshold, it is considered that an acoustic event may have occurred. The amplitude threshold can be determined by analyzing the amplitude statistical characteristics of the audio signal. For example, calculate the mean and standard deviation of the amplitude and use the mean plus twice the standard deviation as the amplitude threshold.

[0077] Frequency mutation threshold setting: Calculate the frequency change rate between adjacent frames and set a frequency mutation threshold. When the frequency change rate exceeds the threshold, a frequency mutation is considered to have occurred, which may correspond to the start or end of an acoustic event. The frequency mutation threshold can be determined by analyzing the frequency change rate of the audio signal under normal operating conditions. For example, calculate the frequency change rate sequence of the audio signal over a period of time, then calculate the distribution characteristics of the sequence, including the mean and standard deviation, and use the mean value of the frequency change rate plus twice the standard deviation as the frequency mutation threshold.

[0078] Start and end point judgment: Combine the amplitude threshold and frequency mutation threshold to determine the start and end points of discrete acoustic events. The specific rules are as follows:

[0079] When the amplitude exceeds the amplitude threshold and the frequency change rate exceeds the frequency mutation threshold, the previous frame of the frame is determined to be the starting point of the acoustic event;

[0080] When the amplitude is lower than the amplitude threshold or the frequency change rate is lower than the frequency mutation threshold (and lasts for a period of time, which can be set as a small window much smaller than the preset time window), the previous frame of the frame is determined as the end point of the acoustic event.

[0081] S204 , performing segmentation based on all the marked start points and end points of the audio data to be diagnosed, to obtain at least one discrete event segment, which constitutes a discrete time segment sequence.

[0082] S103 , based on the discrete event segment sequence, extracting the temporal pattern features of each discrete event segment, and generating fault matching feature information of the discrete event segment sequence.

[0083] Specifically, the temporal pattern features include event duration, discrete interval duration, and event intensity value; event duration is the time length of the discrete event segment, that is, the time interval from the start to the end of the event segment; discrete interval duration is the time interval between the discrete event segment and the previous discrete event segment (the difference between the start time of the current discrete event segment and the start time of the previous discrete event segment. If the current discrete event segment is the first one, the discrete event segment at the end of the previous time window is extracted as the previous discrete event segment); the event intensity value is the amplitude value of the discrete event segment (the difference between the maximum and minimum values).

[0084] In some embodiments, generating fault matching feature information of the discrete event segment sequence specifically includes:

[0085] S301 , performing cluster analysis based on the temporal pattern features of all discrete event segments to obtain several event types.

[0086] Exemplarily, a K-Means clustering algorithm may be used to generate several clusters, each cluster corresponding to an event type and including temporal pattern features of at least one discrete event segment.

[0087] S302: For each cluster, count its occurrence frequency (the ratio of the number of discrete event segments in this type to the total number of segments), occurrence interval sequence (a sequence consisting of the time intervals between adjacent discrete event segments in this type), and the central temporal pattern feature of the cluster (the average value of all central temporal pattern features in the cluster) to form the first characteristic information of the event type.

[0088] S303 , obtaining the event intensity value of each discrete event segment in the discrete event segment sequence, where the horizontal axis represents time and the vertical axis represents the time intensity value, and performing curve fitting to generate an intensity variation curve.

[0089] S304: Combine the first characteristic information of the event type and the intensity change curve into fault matching characteristic information.

[0090] S104 , using the fault matching feature information, a comparison is performed in a pre-established fault pattern library. If a known fault pattern is matched, the fault type and severity are determined, and a fault warning report is generated.

[0091] In some embodiments, the pre-established failure mode library is specifically:

[0092] S401, collect various fault data that occurred during the past operation of the energy storage converter PCS, including operating audio data, fault type, fault severity (which can be set based on expert experience), and operating conditions within the PCS within different historical preset time windows at the time of the fault and in the early period before the fault.

[0093] S402 , generating operating audio data under different fault types through simulation experiments or emulation methods to enrich the data volume of the fault mode library.

[0094] S403: Generate corresponding fault matching feature information for the collected operating audio data within each historical preset time window (refer to the relevant contents of steps S101 to S105 for the specific generation method, which is also based on processing the operating audio data within the preset time window to generate the corresponding fault matching feature information, which is not described in detail in the present invention), bind a label, and set the label content to the corresponding fault type, fault severity, and operating status parameters within the PCS.

[0095] The operating state parameters in the PCS are set to parameters related to the operation of each component in the PCS, such as temperature, vibration frequency, and operating current.

[0096] S404 , classifying all fault matching feature information according to the fault type in the tag, obtaining a number of fault matching feature information corresponding to each fault type, and the fault severity and PCS internal operating status parameters bound to each fault matching feature information.

[0097] Specifically, the cosine similarity between the fault matching feature information and the fault matching feature information in the fault pattern library is calculated. If the similarity is greater than the preset similarity threshold (for example, the similarity threshold is set to 0.85), the known fault mode and its severity corresponding to the successfully matched fault matching feature information are bound to generate a fault warning report and send it to the management center.

[0098] In summary, traditional methods have difficulty capturing discontinuous changes in sound signals, resulting in insufficient early fault warning capabilities. By capturing discontinuous changes in sound signals, early fault warnings can be provided, and abnormal event patterns can be discovered to improve the accuracy and predictability of fault prediction. Based on the frequency distribution information of the audio data to be diagnosed, thresholds are set to identify and segment discrete acoustic events. This can capture discontinuous changes in sound signals, such as sudden knocking sounds, intermittent friction sounds, and other early fault indicators, thereby improving the sensitivity of fault detection. The temporal pattern features of discrete acoustic events are extracted to generate fault matching feature information for the audio data to be diagnosed. By matching with a pre-established fault pattern library, it can be determined whether there is an abnormal pattern, thereby improving the accuracy and timeliness of fault detection and providing important support for equipment maintenance and management.

[0099] Example 2

[0100] In Example 1, the fault matching feature information mainly consists of the first feature information and intensity change curve of at least one event type (there is no limit on whether the number of time types is single or combined), and does not take into account the coherence between different discrete acoustic events and the possible early development chain of faults. This may lead to matching failures, inaccurate matching, or misjudgment of new faults when matching the fault mode library due to ignoring the correlation or collaborative features between events, thereby reducing the efficiency and accuracy of fault detection and increasing safety hazards.

[0101] Since there may be coherence between different discrete acoustic events, a series of seemingly unrelated sound changes may also cause the evolution of failures in certain components or equipment states of the PCS due to their combination. That is, some faults are not caused by a single discrete event fragment, and there may be a fault development chain. Therefore, in Example 1, the composition of the fault matching feature information is extremely critical, because there may be more than one event type and its corresponding first feature information in the discrete event fragment sequence, and there may be a certain relationship between different event types (that is, different clusters). Event streams. If this is ignored, the fault matching feature information composed of a single first feature information may fail when matching the fault library, and the fault matching feature information of some fault types cannot fully reflect the root cause characteristics of the fault, that is, the correct fault type cannot be matched.

[0102] In some embodiments, in step S304, the event type combined with the intensity change curve can be understood as two cases: 1. Independent event type (single event type), the first feature information of each event type is respectively combined with the intensity change curve to form independent fault matching feature information, and matched in the fault mode library respectively, so as to match different fault types; 2. Associated event type (at least two event types), the first feature information of at least two associated event types is combined with the intensity change curve to form independent fault matching feature information;

[0103] When the number of event types in a discrete event segment sequence is greater than 1, it also includes:

[0104] A1. Based on the pre-built discrete event stream analysis model, match the discrete event stream of the discrete event segment sequence. If the match is successful, execute step A2; otherwise, treat each event type as an independent individual and execute step A4.

[0105] Specifically, the pre-built discrete event stream analysis model is:

[0106] B1. Based on the pre-established fault mode library, obtain the discrete event segment sequences corresponding to all the collected operating audio data corresponding to all fault types, and generate the event type corresponding to each discrete event segment sequence (obtained by clustering, the same as step S301).

[0107] Among them, based on the fault mode library, the operating audio data corresponding to all fault types are collected, discrete events are extracted for each operating audio data, a discrete event segment sequence is generated, and the event type of each discrete event segment sequence is obtained.

[0108] B2. Filter target discrete event fragment sequences with more than one event type, count all event type combinations, calculate the frequency of occurrence of each event type combination (the ratio of the number of occurrences to the number of target discrete event fragment sequences), sort all occurrence frequencies in descending order, and eliminate event type combinations in the lower third to obtain the target event type combinations and their corresponding target discrete event fragment sequences (not unique).

[0109] B3. Based on each target event type combination, obtain the discrete event flows corresponding to all corresponding target discrete event fragment sequences, that is, the occurrence chain of all discrete event fragments in chronological order. For example, a target discrete event fragment sequence corresponding to a certain target event type combination includes event type A (discrete event fragments A1 and A2) and event type B (discrete event fragments B1 and B2), and the corresponding discrete event flow is A1→B1→A2→B2.

[0110] B4. Generate a corresponding intensity change curve based on each discrete event stream corresponding to each target event type combination.

[0111] Specifically, the discrete event stream that matches the discrete event fragment sequence includes:

[0112] C1. Obtain the intensity change curve of the discrete event segment sequence and match it with the intensity change curve of each discrete event flow in the discrete event flow analysis model.

[0113] C2. If there is an overlapping segment between the two curves (when the difference between the curves is less than the preset error, it is considered overlapping. The preset error is set based on the actual situation and expert experience), the match is considered successful. Otherwise, the match fails.

[0114] C3. Obtain overlapping segments of the discrete event fragment sequence, and use the overlapping segments as the discrete event stream of the discrete event fragment sequence.

[0115] A2. Based on all clusters of the discrete event segment sequence, first feature information corresponding to (at least one) cluster hit by the discrete event stream is used to generate a first feature information group corresponding to the discrete event stream.

[0116] It should be noted that if the discrete event stream of the discrete event segment sequence does not completely cover the entire discrete event segment sequence, it means that there are still remaining independent discrete event segments, that is, there are independent event types (indicating that there is no association relationship). In this case, step A4 is executed.

[0117] A3. Combining the first characteristic information group and the intensity change curve into fault matching characteristic information.

[0118] A4. The first characteristic information of each event type is combined with the intensity variation curve to form independent fault matching characteristic information, and the information is matched in the fault mode library respectively.

[0119] Therefore, the coherence between different discrete acoustic events and the possible early development chain of faults are taken into consideration. By constructing a discrete event stream analysis model, the discrete event combination related to the fault can be identified more accurately, thereby improving the accuracy of fault detection; by identifying the discrete event stream, the potential fault development chain can be discovered in advance, so that an early warning can be issued before the fault occurs, enhancing the predictability of the fault warning; the discrete event stream is matched in the fault matching feature information generation stage, reducing the number of matches in the fault pattern library, thereby improving the efficiency of fault matching; in actual operation, the fault of the energy storage converter PCS may be caused by a combination of multiple discrete acoustic events. By considering the discrete event stream, it is possible to better adapt to this complex fault scenario and improve the comprehensiveness and accuracy of fault detection.

[0120] Example 3

[0121] With the development of power electronics technology, the structure of PCS is becoming increasingly complex, with a large number of interconnected components. Abnormalities in different components may produce different audio characteristics. It is necessary to comprehensively collect audio data and accurately determine the location of the sound source to better perform differentiated sound source audio analysis.

[0122] In some embodiments, the internal audio collection device may also be configured as a plurality of microphones with uniform coverage, and the deployment of the microphones may cover all components inside the PCS, such as capacitors, reactors, switch modules, and the like.

[0123] Step S101 also includes:

[0124] Obtain the operating audio data collected by each microphone, use the characteristic information of the ambient audio data to perform noise reduction on the operating audio data, generate the corresponding audio data to be diagnosed and the key object corresponding to each audio data to be diagnosed (set to the component object corresponding to the key sound source direction of the audio data to be diagnosed).

[0125] In some embodiments, a method for determining a key object corresponding to each audio data to be diagnosed specifically includes:

[0126] Utilize the preset sound source localization mechanism to determine the key object corresponding to each audio data to be diagnosed.

[0127] The preset sound source localization mechanisms include:

[0128] Based on the beamforming algorithm, the collected audio data to be diagnosed and the deployment position of the corresponding microphone are calculated to determine the direction of the sound source. Combined with the internal structure topology of the PCS, the sound source is mapped to the key object, the signal of the audio data to be diagnosed in the direction of the sound source is enhanced, and high-precision sound source positioning results are obtained. All audio data to be diagnosed and their corresponding key objects are generated.

[0129] It should be noted that the specific implementation method for determining the sound source direction of audio data collected by multiple microphones can refer to the relevant existing technology, and the present invention does not elaborate on or limit this.

[0130] Step S102 further includes:

[0131] All discrete event segment sequences of the audio data to be diagnosed are spliced and combined according to the sampling time (if two discrete event segment information appear at the same time point, they can be processed by taking the average) to generate the final discrete event segment sequence.

[0132] In summary, by setting up several evenly covered microphones inside the PCS, the collected audio data is more targeted, covering the operating information of each key component, providing a richer data basis for subsequent fault diagnosis, and helping to discover potential abnormal audio characteristics of different components; utilizing the preset sound source localization mechanism, the sound source direction is calculated based on the beamforming algorithm, and the sound source is mapped to the key object in combination with the PCS internal structure topology diagram, which can accurately determine the key object corresponding to each audio data to be diagnosed, that is, the component object corresponding to the key sound source direction, enhance the signal of the audio data to be diagnosed in the direction of the sound source, and obtain high-precision sound source localization results. When performing subsequent fault mode library matching, it can quickly locate the components that may have problems, thereby improving diagnostic efficiency and accuracy.

[0133] Example 4

[0134] In some embodiments, step S104 further includes:

[0135] S501, if a known fault mode is matched and the determined fault type is not unique, obtaining a key object of the audio data to be diagnosed corresponding to each discrete event segment in the discrete event segment sequence;

[0136] Compare the key objects with the determined fault types to determine the final fault type.

[0137] For example, when the determined fault types include "gear breakage" and "motor overheating", and the key object is the gear, the final fault type is determined to be "gear breakage".

[0138] S502, when no known fault mode is matched, obtaining a key object of the audio data to be diagnosed corresponding to each discrete event segment in the discrete event segment sequence;

[0139] Obtain standard operating audio data when the key object is operating normally, and compare it with the corresponding audio data to be diagnosed. If the difference is too large, the corresponding key component will be generated as a suspected fault object for feedback; otherwise, there will be no fault warning.

[0140] It should be noted that when comparing two audio data, it is possible to determine whether the difference is too large based on the difference between the time-amplitude fitting curves corresponding to the audio data (the calculus difference between the two fitting curves). A difference threshold can be set, which will not be elaborated in the present invention.

[0141] In summary, when multiple known fault modes are matched, the final fault type can be accurately determined by comparing the key objects with the fault types, avoiding diagnostic ambiguity caused by non-unique fault types, improving the accuracy of fault diagnosis, and enabling maintenance personnel to take more targeted maintenance measures; when no known fault mode is matched, by comparing with the standard operating audio data during normal operation of the key objects, the abnormalities of key components can be discovered in time, and suspected fault objects can be generated for feedback, which helps to discover potential fault hazards even when the fault mode library does not cover new fault types, thereby enhancing the predictability of fault warnings; using key objects for fault type comparison and anomaly detection narrows the scope of fault investigation, avoids comprehensive inspection of all components, improves the efficiency of fault diagnosis, and reduces maintenance time and costs.

[0142] Example 5

[0143] Figure 2 The figure is a structural diagram of a PCS fault detection system for an energy storage converter according to an embodiment of the present invention.

[0144] like Figure 2 As shown, a PCS fault detection system for an energy storage converter includes: an acquisition module, an extraction module, and a matching module;

[0145] The acquisition module is used to acquire the operating audio data and environmental audio data within a preset time window in real time based on the audio acquisition device deployed by the PCS, and generate audio data to be diagnosed;

[0146] The extraction module is used to generate a discrete event segment sequence of the audio data to be diagnosed using a preset discrete event extraction mechanism; based on the discrete event segment sequence, the temporal pattern features of each discrete event segment are extracted to generate fault matching feature information of the discrete event segment sequence;

[0147] The matching module is used to use the fault matching feature information to compare with the pre-established fault mode library. If a known fault mode is matched, the fault type and severity are determined.

[0148] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0149] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0151] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0152] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0153] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0154] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for detecting PCS faults in an energy storage converter, characterized in that: include: S101, based on the audio acquisition device deployed by the PCS, real-time acquisition of operating audio data and environmental audio data within a preset time window to generate audio data to be diagnosed; S102, using a preset discrete event extraction mechanism to generate a discrete event segment sequence of the audio data to be diagnosed; S103: Based on the discrete event segment sequence, extract the temporal pattern features of each discrete event segment and perform cluster analysis to obtain several event types. Generate first feature information and an intensity change curve of the event type as fault matching feature information for the discrete event segment sequence. The first feature information of each event type consists of an occurrence frequency, an occurrence interval sequence, and a central temporal pattern feature of the cluster. The occurrence frequency is set as the ratio of the number of discrete event segments in that type to the total number of segments. The occurrence interval sequence is set as the sequence consisting of the occurrence time intervals between adjacent discrete event segments in that type. S104, using the fault matching feature information, performing a comparison in a pre-established fault mode library, and if a known fault mode is matched, determining the fault type and severity; When the number of event types in a discrete event fragment sequence is greater than 1, the method includes: A1, matching the discrete event stream of the discrete event fragment sequence based on a pre-built discrete event stream analysis model; if the match is successful, executing step A2; otherwise, treating each event type as an independent individual and executing step A4; A2, based on all clusters of the discrete event fragment sequence, forming the first feature information corresponding to the cluster hit by its discrete event stream into a first feature information group corresponding to the discrete event stream; A3, forming the first feature information group and the intensity change curve into fault matching feature information; A4, forming the first feature information of each event type into independent fault matching feature information with the intensity change curve, and matching them in the fault mode library respectively.

2. The energy storage converter PCS fault detection method according to claim 1, characterized in that: The audio collection device deployed by the PCS includes an internal audio collection device and an external audio collection device. The internal audio collection device is set inside the PCS and is used to collect operating audio data. The external audio collection device is set outside the PCS and is used to collect environmental audio data.

3. The energy storage converter PCS fault detection method according to claim 1, characterized in that: The preset discrete event extraction mechanism specifically includes: S201, performing short-time Fourier transform on the audio data to be diagnosed to obtain frequency distribution information; S202, calculating, based on each frame of the audio signal, a rate of change of the frequency component thereof compared to the previous frame, i.e., a frequency change rate, obtained by comparing changes in amplitude values of the same frequency component of adjacent frames; S203: Obtain the amplitude and frequency change rate of each frame of the audio data to be diagnosed, set preset thresholds, and determine the starting and ending points of the discrete event segment; the preset thresholds include amplitude thresholds and frequency mutation thresholds, which are set based on actual conditions and expert experience; S204 , performing segmentation based on all the marked start points and end points of the audio data to be diagnosed, to obtain at least one discrete event segment, which constitutes a discrete time segment sequence.

4. The energy storage converter PCS fault detection method according to claim 3, characterized in that: The temporal pattern features include event duration, discrete interval duration, and event intensity value; event duration is the time length of a discrete event segment; discrete interval duration is the time interval between a discrete event segment and the previous discrete event segment; and event intensity value is the amplitude value of the discrete event segment.

5. The energy storage converter PCS fault detection method according to claim 3, characterized in that: The S103 specifically includes: S302: For each cluster, count its occurrence frequency, occurrence interval sequence, and central temporal pattern characteristics of the cluster to form first characteristic information of the event type; S303, obtaining the event intensity value of each discrete event segment in the discrete event segment sequence, where the horizontal axis is time and the vertical axis is the time intensity value, and performing curve fitting to generate an intensity variation curve; S304: Combine the first characteristic information of the event type and the intensity change curve into fault matching characteristic information.

6. The energy storage converter PCS fault detection method according to claim 1, characterized in that: The pre-established fault mode library is constructed in the following manner: S401, collecting various fault data that occurred during the past operation of the energy storage converter PCS, including operating audio data, fault type, fault severity, and operating conditions within the PCS within different historical preset time windows at the time of the fault and in the early period before the fault; S402, generating operating audio data under different fault types through simulation experiments or emulation methods to enrich the data volume of the fault mode library; S403: Generate corresponding fault matching feature information for the collected operating audio data within each historical preset time window, bind a label, and set the label content to the corresponding fault type, fault severity, and operating status parameters in the PCS; S404 , classify all fault matching feature information according to the fault type in the tag, and obtain a number of fault matching feature information corresponding to each fault type and the fault severity and PCS internal operating status parameters bound to each fault matching feature information.

7. The energy storage converter PCS fault detection method according to claim 5, characterized in that: The pre-built discrete event stream analysis model is specifically: B1. Based on a pre-established fault mode library, obtain discrete event segment sequences corresponding to all collected operating audio data corresponding to all fault types, and generate event types corresponding to each discrete event segment sequence; B2. Filter target discrete event fragment sequences with more than one event type, count all event type combinations, calculate the frequency of occurrence of each event type combination, sort all occurrence frequencies in descending order, and eliminate event type combinations in the lower third to obtain target event type combinations and their corresponding target discrete event fragment sequences; B3. Based on each target event type combination, obtain the discrete event streams corresponding to all corresponding target discrete event fragment sequences, that is, the occurrence chain of all discrete event fragments in chronological order; B4. Generate a corresponding intensity change curve based on each discrete event stream corresponding to each target event type combination; In A1, the discrete event stream matching the discrete event segment sequence includes: C1. Obtain the intensity change curve of the discrete event segment sequence and match it with the intensity change curve of each discrete event flow in the discrete event flow analysis model; C2. If there is an overlapping segment between the two curves, the matching is successful; otherwise, the matching fails. C3. Obtain overlapping segments of the discrete event fragment sequence, and use the overlapping segments as the discrete event stream of the discrete event fragment sequence.

8. The energy storage converter PCS fault detection method according to claim 2, characterized in that: The internal audio collection device is configured as a plurality of microphones with uniform coverage; The S101 further includes: Obtain the operating audio data collected by each microphone and its corresponding audio data to be diagnosed, as well as the key object corresponding to each audio data to be diagnosed; The key object is set as the component object corresponding to the key sound source direction of the audio data to be diagnosed; The S102 further includes: All discrete event segment sequences of the audio data to be diagnosed are spliced and combined according to the sampling time.

9. A PCS fault detection system for an energy storage converter, applied to a PCS fault detection method for an energy storage converter according to any one of claims 1 to 8, characterized in that: include: Acquisition module, extraction module, matching module; The acquisition module is used to acquire the operating audio data and environmental audio data within a preset time window in real time based on the audio acquisition device deployed by the PCS, and generate audio data to be diagnosed; The extraction module is used to generate a discrete event segment sequence of the audio data to be diagnosed by using a preset discrete event extraction mechanism; Based on the discrete event segment sequence, the temporal pattern features of each discrete event segment are extracted to generate the fault matching feature information of the discrete event segment sequence; The matching module is used to use the fault matching feature information to compare with the pre-established fault mode library. If a known fault mode is matched, the fault type and severity are determined.

Images