Voiceprint library construction method and device for phase modifier rotating equipment

Abstract

The invention belongs to the technical field of phase modifier voiceprint construction, and particularly relates to a voiceprint library construction method and device for phase modifier rotating equipment. Comprising the following steps: acquiring operation sound data of equipment in an initial state, extracting initial voiceprint feature information, acquiring equipment sound data and extracting monitored voiceprint feature information according to a certain monitoring period in an equipment operation process, acquiring voiceprint feature change information through comparison, judging whether the voiceprint feature change information is abnormal or not, and if so, judging whether the voiceprint feature change information is abnormal or not. And if not, extracting different voiceprint feature information, constructing an analysis time period, combining with equipment operation state data, constructing a multi-modal feature tag set by using the different voiceprint feature information, combining the multi-modal feature tag set with the initial voiceprint feature information, generating a feature set tag set, performing dynamic feature value calculation, and finally constructing a voiceprint library. According to the method, the constructed voiceprint library is not only based on the sound features, but also associated with the multi-dimensional operation state of the equipment, so that the applicability and accuracy of the voiceprint library are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of phasor machine voiceprint construction, and particularly relates to a method and device for constructing a voiceprint library for a phasor machine rotating device. Background Art

[0002] Phasor machine rotating devices are widely used in power systems. As important devices for regulating the reactive power of the power grid and balancing the voltage, the reliability of their operation directly affects the stability and operation efficiency of the power grid. In recent years, with the development of phasor machine rotating device technology, the monitoring and maintenance of their operation status have become key links in ensuring the performance of the devices and extending their service life.

[0003] Traditional device status monitoring methods mainly rely on the measurement and analysis of single physical quantities such as vibration, temperature, and current. Although these methods can provide certain status information, it is difficult to comprehensively reflect the small changes or potential hazards existing in the device operation process. In addition, some monitoring methods rely too much on specific monitoring parameters. Once the device operation environment changes, the accuracy and stability of the monitoring results may be affected, and even misjudgment or missed judgment may occur.

[0004] As a natural by-product generated during device operation, the sound signal contains rich device status information. The research on device status monitoring based on sound signals has gradually become a hot topic. However, existing monitoring methods based on sound signals mainly focus on fixed voiceprint feature extraction and comparison, and fail to conduct in-depth analysis in combination with the multi-dimensional operation status of the device (such as temperature, load, etc.), resulting in limited applicability and accuracy of the monitoring methods. Especially when the device operation environment or operation status changes, the applicability of traditional voiceprint monitoring methods may decrease, and it is difficult to effectively cope with the changes in voiceprint features during long-term operation. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and device for constructing a voiceprint library for a phasor machine rotating device, which can improve the applicability and reliability of voiceprint monitoring and adapt to the changes in the device operation environment and status.

[0006] The technical solutions adopted by the present invention are specifically as follows:

[0007] A method for constructing a voiceprint library for a phasor machine rotating device includes:

[0008] Obtain the initial operation sound data of the phasor machine rotating device, and obtain the initial voiceprint feature information according to the initial operation sound data;

[0009] Construct a monitoring period, obtain the operating sound data of the synchronous condenser rotating equipment within the monitoring period, obtain the monitored voiceprint feature information based on the operating sound data within the monitoring period, compare the monitored voiceprint feature information with the initial voiceprint feature information, and obtain the voiceprint feature change information;

[0010] Judge whether the voiceprint feature change information meets the first preset condition. If not, obtain the differential voiceprint feature information of the synchronous condenser rotating equipment based on the monitored voiceprint feature information;

[0011] Construct an analysis period, obtain the operating status data of the synchronous condenser rotating equipment within the analysis period, and construct a multi-modal feature label set corresponding to the equipment operating status based on the differential voiceprint feature information;

[0012] Construct a feature set label set containing multiple states based on the multi-modal feature label set and the initial voiceprint feature information, obtain the dynamic feature values of each feature label in the feature set label set and assign label retrieval levels, and construct a voiceprint library.

[0013] In a preferred solution, the steps of obtaining the initial voiceprint feature information based on the initial operating sound data of the synchronous condenser rotating equipment include:

[0014] Obtain the initial operating sound data of the synchronous condenser rotating equipment;

[0015] Divide the initial operating sound data into multiple sound segment information;

[0016] Extract features from each sound segment to generate corresponding initial sound feature vectors;

[0017] Summarize multiple initial sound feature vectors, and mark the summary result as the initial voiceprint feature information.

[0018] In a preferred solution, the steps of constructing a monitoring period, obtaining the operating sound data of the synchronous condenser rotating equipment within the monitoring period, obtaining the monitored voiceprint feature information based on the operating sound data within the monitoring period, comparing the monitored voiceprint feature information with the initial voiceprint feature information, and obtaining the voiceprint feature change information include:

[0019] Construct a monitoring period;

[0020] Obtain the operating sound data of the synchronous condenser rotating equipment within the monitoring period, and mark it as the monitored operating sound data;

[0021] Extract the monitored voiceprint feature information corresponding to the monitored operating sound data;

[0022] Decompose the monitored voiceprint feature information into multiple monitored sound feature vectors;

[0023] Obtain the voiceprint feature change value according to multiple monitored voice feature vectors;

[0024] Obtain the voiceprint feature evaluation threshold;

[0025] Judge whether the voiceprint feature change value exceeds the voiceprint feature evaluation threshold;

[0026] If the voiceprint feature change value does not exceed the voiceprint feature evaluation threshold, it is determined that the voiceprint feature change information is stable;

[0027] If the voiceprint feature change value exceeds the voiceprint feature evaluation threshold, it is determined that the voiceprint feature change information is fluctuating;

[0028] The calculation of the voiceprint feature change value is as follows:

[0029] , where represents the voiceprint feature change value, i represents the number of multiple monitored voice feature vectors, i = 1, 2, 3…n, and n represents the number of monitored voice feature vectors, represents the i-th monitored voice feature vector.

[0030] In a preferred solution, to judge whether the voiceprint feature change information meets the first preset condition, if not, the steps of obtaining the differential voiceprint feature information of the synchronous condenser rotating device according to the monitored voiceprint feature information include:

[0031] Obtain the voiceprint difference evaluation threshold;

[0032] Obtain the corresponding voiceprint feature change value according to the voiceprint feature change information;

[0033] Judge whether the voiceprint feature change value exceeds the voiceprint difference evaluation threshold;

[0034] If the voiceprint feature change value exceeds the voiceprint difference evaluation threshold, it is determined that the operation sound of the synchronous condenser rotating device is abnormal, and the abnormal difference between the voiceprint feature change value and the voiceprint difference evaluation threshold is obtained;

[0035] Obtain the differential voiceprint feature information according to the abnormal difference, the monitored voiceprint feature information and the initial voiceprint feature information.

[0036] In a preferred solution, the steps of obtaining the differential voiceprint feature information according to the abnormal difference, the monitored voiceprint feature information and the initial voiceprint feature information include:

[0037] Obtain the abnormal difference between the voiceprint feature change value and the voiceprint difference evaluation threshold;

[0038] Obtain a difference table, where the difference table includes multiple abnormal difference intervals and the corresponding voiceprint similarity thresholds for each abnormal difference interval;

[0039] Obtain the corresponding voiceprint similarity threshold from the difference table according to the abnormal difference value;

[0040] Obtain a plurality of corresponding monitoring sound feature vectors according to the monitored voiceprint feature information;

[0041] Obtain a plurality of corresponding initial sound feature vectors according to the initial voiceprint feature information and the monitored voiceprint feature information;

[0042] Obtain the voiceprint feature similarity value based on each monitoring sound feature vector and the corresponding initial sound feature vector;

[0043] Determine whether a plurality of voiceprint feature similarity values exceed the obtained voiceprint similarity threshold;

[0044] If there is a voiceprint feature similarity value exceeding the voiceprint similarity threshold, summarize the voiceprint feature similarity values exceeding the voiceprint similarity threshold, and mark the summary result as differential voiceprint feature information;

[0045] The voiceprint feature similarity value is calculated as:

[0046] , where Y represents the voiceprint feature similarity value, A represents the initial sound feature vector, and B represents the monitoring sound feature vector.

[0047] In a preferred solution, the steps of constructing an analysis period, obtaining the operation state data of the synchronous condenser rotating device within the analysis period, and constructing a multi-modal feature tag set corresponding to the device operation state according to the differential voiceprint feature information include:

[0048] Obtain the acquisition frequency of the operation state of the synchronous condenser rotating device;

[0049] Obtain the acquisition duration according to the acquisition frequency;

[0050] Divide the monitoring period into multiple analysis periods according to the acquisition duration;

[0051] Obtain the operation state information of the synchronous condenser rotating device within each analysis period;

[0052] Summarize a plurality of operation state information to form a set of time series features, and mark it as analysis operation state data;

[0053] According to the time interval corresponding to the differential voiceprint feature information, obtain the corresponding operation state data of the synchronous condenser rotating device, and summarize it as differential operation state data;

[0054] Extract the voice features matching the differential operation state data to form differential voice feature vectors;

[0055] Obtain the same operating status information in the differential operating status data and the analysis operating status data, associate the same operating status information with the differential voice feature vectors, and summarize the association results into a label form to construct differential feature labels;

[0056] Summarize multiple differential feature labels and mark the summary result as a multi-modal feature label set.

[0057] In a preferred solution, constructing a feature set label set containing multiple states according to the multi-modal feature label set and the initial voiceprint feature information, obtaining the dynamic feature values of each feature label in the feature set label set and assigning label retrieval levels, the steps of constructing a voiceprint library include:

[0058] Extract and mark the basic feature labels of the device in the normal operating state according to the initial voiceprint feature information;

[0059] Integrate the multi-modal feature label set with the basic feature labels to obtain a feature set label set containing multiple states, where the feature set label set includes multiple feature labels;

[0060] Extract multiple voiceprint feature vectors of each feature label in the feature set label set;

[0061] Obtain the dynamic feature value corresponding to each feature label according to the multiple voiceprint feature vectors of each feature label;

[0062] Obtain a label level table, where the label level table includes multiple dynamic feature interval values and the label retrieval levels corresponding to each dynamic feature interval value;

[0063] Obtain the corresponding target dynamic feature interval value from the label level table according to the dynamic feature value;

[0064] Obtain the corresponding label retrieval level from the label level table according to the target dynamic feature interval value;

[0065] Construct a voiceprint library according to the label retrieval levels of each feature label in the feature set label set, and optimize the high-priority labels to the fast retrieval path;

[0066] The dynamic feature value is calculated as:

[0067] , where represents the dynamic feature value, h represents the number of multiple voiceprint feature vectors, h = 2, 3, 4... t, t represents the number of voiceprint feature vectors of the feature label, represents the hth voiceprint feature vector, represents the (h - 1)th voiceprint feature vector.

[0068] In a preferred embodiment, it further includes: constructing an update period, obtaining the new voiceprint feature change information generated by the synchronous condenser rotating device within the update period, and returning the new voiceprint feature change information as the voiceprint feature change information to the step of determining whether the voiceprint feature change information meets the first preset condition. This step specifically includes:

[0069] Obtain the end time node of the monitoring period and mark it as the start time of the update period;

[0070] Obtain the monitoring duration of the monitoring period;

[0071] Obtain multiple dynamic feature values within the feature set label set;

[0072] Obtain the update duration of the update period according to the multiple dynamic feature values and the monitoring duration;

[0073] Obtain the end time of the update period according to the update duration and the start time of the update period;

[0074] Obtain the update period according to the start time and the end time of the update period;

[0075] Obtain the new voiceprint feature change information generated by the synchronous condenser rotating device within the update period;

[0076] Return the new voiceprint feature change information as the voiceprint feature change information to the step of determining whether the voiceprint feature change information meets the first preset condition;

[0077] The update duration of the update period is calculated as:

[0078] , where E represents the update duration, q represents the number of multiple dynamic feature values, q = 1, 2, 3... r, represents the qth dynamic feature value, r represents the number of dynamic feature values, and T represents the monitoring duration.

[0079] The present invention also provides a voiceprint library construction device for a synchronous condenser rotating device, which is used to implement the above-mentioned voiceprint library construction method for a synchronous condenser rotating device, including:

[0080] An initial voiceprint module, which is used to obtain the initial operation sound data of the synchronous condenser rotating device and obtain the initial voiceprint feature information according to the initial operation sound data;

[0081] A feature change module, which is used to construct a monitoring period, obtain the operation sound data of the synchronous condenser rotating device within the monitoring period, obtain the monitored voiceprint feature information according to the operation sound data within the monitoring period, compare the monitored voiceprint feature information with the initial voiceprint feature information, and obtain the voiceprint feature change information;

[0082] A feature comparison module, configured to determine whether the voiceprint feature change information conforms to a first preset condition. If not, differential voiceprint feature information of the synchronous condenser rotating device is obtained according to the monitored voiceprint feature information;

[0083] A feature tagging module, configured to construct an analysis period, obtain operation status data of the synchronous condenser rotating device within the analysis period, and construct a multi-modal feature tag set corresponding to the device operation status according to the differential voiceprint feature information;

[0084] A voiceprint library module, configured to construct a feature set tag set including multiple states according to the multi-modal feature tag set and the initial voiceprint feature information, obtain dynamic feature values of each feature tag in the feature set tag set and assign tag retrieval levels, and construct a voiceprint library;

[0085] An update module, configured to construct an update period, obtain new voiceprint feature change information generated by the synchronous condenser rotating device within the update period, and return the new voiceprint feature change information as the voiceprint feature change information to the feature comparison module.

[0086] And, a voiceprint library construction terminal for a synchronous condenser rotating device, including:

[0087] One or more processors;

[0088] A storage device, on which one or more programs are stored;

[0089] When the one or more programs are executed by the one or more processors, the one or more processors implement a voiceprint library construction method for a synchronous condenser rotating device.

[0090] The technical effects achieved by the present invention are:

[0091] The present invention provides a voiceprint library construction method for a synchronous condenser rotating device, which can obtain the operation sound data of the device in real time, and accurately reflect the change of the device operation status through the voiceprint feature change information and the differential voiceprint feature information, generate a multi-modal feature tag set in combination with the device operation status data, make the voiceprint library not only based on the sound features, but also associated with the multi-dimensional operation status of the device, improve the applicability and accuracy of the voiceprint library, can adjust the voiceprint library in real time according to the latest operation data, avoid the voiceprint library from becoming invalid due to the change of the device operation environment, and ensure its long-term reliability. Description of the Drawings

[0092] Figure 1 is a flowchart of a voiceprint library construction method for a synchronous condenser rotating device provided by an embodiment of the present invention;

[0093] Figure 2 is a functional module diagram of a voiceprint library construction device for a synchronous condenser rotating device provided by an embodiment of the present invention. Detailed implementation manners

[0094] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the detailed implementation manners of the present invention with reference to the accompanying drawings of the specification.

[0095] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0096] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The appearances of "in a preferred embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other with other embodiments.

[0097] Thirdly, the present invention is described in detail with reference to the schematic diagrams. When describing the embodiments of the present invention in detail, for the sake of illustration, the schematic diagrams are only examples and should not limit the scope of protection of the present invention herein.

[0098] The first embodiment of the present invention provides a method for constructing a voiceprint library for a synchronous condenser rotating device. Refer to Figure 1 , including:

[0099] S1. Obtain the initial operation sound data of the synchronous condenser rotating device, and obtain the initial voiceprint feature information according to the initial operation sound data;

[0100] S2. Construct a monitoring period, obtain the operation sound data of the synchronous condenser rotating device within the monitoring period, obtain the monitoring voiceprint feature information according to the operation sound data within the monitoring period, compare the monitoring voiceprint feature information with the initial voiceprint feature information, and obtain the voiceprint feature change information;

[0101] S3. Determine whether the voiceprint feature change information meets the first preset condition. If not, obtain the differential voiceprint feature information of the synchronous condenser rotating device according to the monitoring voiceprint feature information;

[0102] S4. Construct an analysis period, obtain the operation state data of the synchronous condenser rotating device within the analysis period, and construct a multi-modal feature tag set corresponding to the device operation state according to the differential voiceprint feature information;

[0103] S5. Construct a feature set tag set including multiple states according to the multi-modal feature tag set and the initial voiceprint feature information, obtain the dynamic feature values of each feature tag in the feature set tag set and assign a tag retrieval level, and construct a voiceprint library.

[0104] As in the above steps S1 to S5, the operation sound data of the device in the initial state is obtained, and the initial voiceprint feature information is extracted by processing the sound data. During the operation of the device, the device sound data is obtained according to a certain monitoring period and the monitoring voiceprint feature information is extracted. At the same time, the monitoring voiceprint feature information is compared with the initial voiceprint feature information to obtain the voiceprint feature change information. By setting a first preset condition, it is determined whether the voiceprint feature change information is abnormal. If an abnormality is detected, the difference voiceprint feature information is extracted to reflect the change of the device operation state. By constructing an analysis period and combining the device operation state data (such as device temperature, load, etc.), the difference voiceprint feature information is used to analyze the change of the device operation state. Generate a multimodal feature label set related to the equipment operation status, combine the multimodal feature label set with the initial voiceprint feature information, generate a feature set label set, and finally build a voiceprint library by dynamically calculating the feature values of multiple feature labels in the feature set to achieve accurate modeling of the equipment operation voiceprint. It can obtain the equipment's operation sound data in real time, and accurately reflect the changes in the equipment's operation status through voiceprint feature change information and differential voiceprint feature information. Combined with the equipment operation status data, a multimodal feature label set is generated, so that the voiceprint library is not only based on sound features, but also associated with the multi-dimensional operation status of the equipment, thereby improving the applicability and accuracy of the voiceprint library.

[0105] In a preferred embodiment, the steps of obtaining initial operation sound data of the phase regulator rotating device and obtaining initial voiceprint feature information according to the initial operation sound data include:

[0106] S101, obtaining initial operation sound data of a phase condenser rotating device;

[0107] S102, acquiring corresponding multiple sound segment information according to the initial operation sound data;

[0108] S103, obtaining an initial sound feature vector in each sound clip information;

[0109] S104: Aggregate multiple initial sound feature vectors, and mark the aggregation result as initial voiceprint feature information.

[0110] In the above steps S101 to S104, sound data of the synchronous condenser rotating device in the initial operating state is acquired through sensors or audio acquisition devices, ensuring that these data are representative under normal operating conditions of the device. The acquired initial operating sound data is processed and divided into multiple sound segments. Each sound segment can be segmented according to a time interval (such as 1 second, 1 minute, 2 minutes, etc.) or specific features (such as energy change) to ensure good timeliness and resolution of the sound data. Feature extraction is performed on each sound segment to generate corresponding initial sound feature vectors. These feature vectors may include, but are not limited to, the following parameters: frequency domain features (such as spectrum, power spectral density); time domain features (such as signal peak value, mean value, variance); time-frequency domain features (such as MFCC, Chroma features). The initial sound feature vectors of all sound segments are summarized to generate a comprehensive feature set, which is marked as the initial voiceprint feature information. This embodiment can carefully capture the sound characteristics of the synchronous condenser rotating device in the initial operating state, ensure the representativeness and accuracy of the voiceprint features, and the generated initial voiceprint feature information provides a benchmark for subsequent analysis of voiceprint feature changes, enabling rapid comparison and accurate identification of abnormalities when the device operating state changes.

[0111] In a preferred embodiment, a monitoring period is constructed, the operating sound data of the synchronous condenser rotating device within the monitoring period is acquired, and the monitored voiceprint feature information is obtained based on the operating sound data within the monitoring period. The steps of comparing the monitored voiceprint feature information with the initial voiceprint feature information to obtain the voiceprint feature change information include:

[0112] S201. Construct a monitoring period;

[0113] S202. Acquire the operating sound data of the synchronous condenser rotating device within the monitoring period and mark it as monitored operating sound data;

[0114] S203. Extract the monitored operating sound data to obtain the corresponding monitored voiceprint feature information;

[0115] S204. Decompose the monitored voiceprint feature information into multiple monitored sound feature vectors;

[0116] S205. Obtain the voiceprint feature change value based on the multiple monitored sound feature vectors;

[0117] S206. Obtain the voiceprint feature evaluation threshold;

[0118] S207. Determine whether the voiceprint feature change value exceeds the voiceprint feature evaluation threshold;

[0119] If the voiceprint feature change value does not exceed the voiceprint feature evaluation threshold, it is determined that the voiceprint feature change information is stable;

[0120] If the voiceprint feature change value exceeds the voiceprint feature evaluation threshold, it is determined that the voiceprint feature change information is fluctuating.

[0121] In the above steps S201 to S207, a monitoring period is set for regularly collecting the running sound data of the device. The monitoring period can be dynamically adjusted according to the running frequency or scenario requirements of the device to ensure that the collected data is representative and continuous. During each monitoring period, the sound data generated during the device operation is collected and marked as the monitored running sound data. The monitored running sound data is processed to extract the corresponding monitored voiceprint feature information. The monitored voiceprint feature information is further decomposed into multiple monitored sound feature vectors. Each feature vector represents the sound characteristics of a certain segment within the monitoring period, such as frequency distribution, time-domain variation, time-frequency domain characteristics, etc. The multiple monitored sound feature vectors are compared with the initial voiceprint feature information or the voiceprint feature information of the previous period to calculate the voiceprint feature change value. The calculation formula of the voiceprint feature change value is , where, represents the voiceprint feature change value, i represents the number of multiple monitored sound feature vectors, i = 1, 2, 3…n, represents the i-th monitored sound feature vector. A voiceprint feature evaluation threshold is set to determine whether the amplitude of the voiceprint change is within an acceptable range. If the voiceprint feature change value does not exceed the evaluation threshold, it is determined that the voiceprint feature change information is stable, indicating that the device is operating normally. If the voiceprint feature change value exceeds the evaluation threshold, it is determined that the voiceprint feature change information is fluctuating, indicating that there may be an abnormality or a state change. This embodiment can continuously track the changes in the running sound of the device and realize the dynamic monitoring of the device running state. The monitoring period and evaluation threshold can be flexibly adjusted according to the characteristics of different devices to adapt to various types of synchronous condenser rotating devices and different running scenarios.

[0122] In a preferred embodiment, to determine whether the voiceprint feature change information meets the first preset condition, if not, the steps of obtaining the differential voiceprint feature information of the synchronous condenser rotating device according to the monitored voiceprint feature information include:

[0123] S301. Obtain the voiceprint difference evaluation threshold;

[0124] S302. Obtain the corresponding voiceprint feature change value according to the voiceprint feature change information;

[0125] S303. Determine whether the voiceprint feature change value exceeds the voiceprint difference evaluation threshold;

[0126] If the voiceprint feature change value exceeds the voiceprint difference evaluation threshold, it is determined that the running sound of the synchronous condenser rotating device is abnormal, and the abnormal difference between the voiceprint feature change value and the voiceprint difference evaluation threshold is obtained;

[0127] S304. Obtain differential voiceprint feature information based on the abnormal difference, the monitored voiceprint feature information, and the initial voiceprint feature information.

[0128] In the above steps S301 to S304, a voiceprint difference evaluation threshold is set to determine whether the change in voiceprint features exceeds the normal range. This threshold can be obtained through historical data training, expert experience, or dynamic adaptive adjustment to ensure the rationality and accuracy of the evaluation criteria. Extract the voiceprint feature change value from the voiceprint feature change information, and compare the voiceprint feature change value with the voiceprint difference evaluation threshold. If the voiceprint feature change value does not exceed the voiceprint difference evaluation threshold, it indicates that the running sound of the device is normal and no further processing is required. If the voiceprint feature change value exceeds the voiceprint difference evaluation threshold, it is determined that the running sound of the device is abnormal, and the abnormal difference (i.e., the difference between the change value and the threshold) is calculated. In this embodiment, when the device sound is abnormal, based on the abnormal difference, the monitored voiceprint feature information, and the initial voiceprint feature information, the differential voiceprint feature information is obtained, which can keenly identify the abnormality of the device sound state and avoid missed reports or false alarms.

[0129] In a preferred embodiment, the step of obtaining differential voiceprint feature information based on the abnormal difference, the monitored voiceprint feature information, and the initial voiceprint feature information includes:

[0130] S3041. Obtain the abnormal difference between the voiceprint feature change value and the voiceprint difference evaluation threshold;

[0131] S3042. Obtain a difference table, where the difference table includes multiple abnormal difference intervals and the corresponding voiceprint similarity thresholds for each abnormal difference interval;

[0132] S3043. Obtain the corresponding voiceprint similarity threshold from the difference table according to the abnormal difference;

[0133] S3044. Obtain the corresponding multiple monitored voice feature vectors according to the monitored voiceprint feature information;

[0134] S3045. Obtain the corresponding multiple initial voice feature vectors according to the initial voiceprint feature information;

[0135] S3046. Obtain the voiceprint feature similarity value based on each monitored voice feature vector and the corresponding initial voice feature vector;

[0136] S3047. Determine whether multiple voiceprint feature similarity values exceed the voiceprint similarity threshold;

[0137] If there is a voiceprint feature similarity value that exceeds the voiceprint similarity threshold, summarize the voiceprint feature similarity values that exceed the voiceprint similarity threshold, and mark the summary result as differential voiceprint feature information.

[0138] In the above steps S3041 to S3047, by calculating the difference between the voiceprint feature change value and the voiceprint difference evaluation threshold, the specific degree of abnormal operation sound of the device is determined. According to the abnormal difference, the corresponding voiceprint similarity threshold is obtained from the pre-set difference table. The corresponding voice feature vectors are extracted from the monitored voiceprint feature information and the initial voiceprint feature information respectively. The monitored voice feature vectors in each group are compared with the corresponding initial voice feature vectors, and their voiceprint feature similarity values are calculated. The calculation formula of the voiceprint feature similarity value is: , where Y represents the voiceprint feature similarity value, A represents the initial voice feature vector, and B represents the monitored voice feature vector. Compare each voiceprint feature similarity value with the voiceprint similarity threshold. If some voiceprint feature similarity values exceed the voiceprint similarity threshold, it indicates that the voice characteristics reflected by these feature vectors have changed significantly. Summarize the voiceprint feature similarity values that exceed the voiceprint similarity threshold, and mark the result as the differential voiceprint feature information. This embodiment accurately extracts the differential features of the device operation sound, avoids the ambiguity in simple change determination. The difference table dynamically maps the abnormal interval and the voiceprint similarity threshold, realizes the flexible adjustment of the analysis accuracy, adapts to different devices and working conditions, compares the monitored features with the initial features one by one, ensures that all possible differential features are extracted, and provides more comprehensive voiceprint information.

[0139] In a preferred embodiment, the steps of constructing an analysis period, obtaining the operation state data of the synchronous condenser rotating device within the analysis period, and constructing a multi-modal feature tag set corresponding to the device operation state according to the differential voiceprint feature information include:

[0140] S401. Obtain the acquisition frequency of the operation state of the synchronous condenser rotating device;

[0141] S402. Obtain the acquisition duration according to the acquisition frequency;

[0142] S403. Divide the monitoring period into multiple analysis periods according to the acquisition duration;

[0143] S404. Obtain the operation state information of the synchronous condenser rotating device within each analysis period;

[0144] S405. Summarize the multiple operation state information to form a set of time series features, and mark it as the analysis operation state data;

[0145] S406. Obtain the corresponding operation state data of the synchronous condenser rotating device according to the time interval corresponding to the differential voiceprint feature information, and summarize it as the differential operation state data;

[0146] S407. Extract the voice features matching the differential operation state data to form the differential voice feature vector;

[0147] S408. Obtain the same operating status information in the differential operating status data and the analysis operating status data, associate the same operating status information with the differential sound feature vector, and summarize the association result into a label form to construct a differential feature label;

[0148] S409. Aggregate multiple differential feature labels and mark the aggregation result as a multi-modal feature label set.

[0149] In the above steps S401 to S409, the frequency of device operating status data collection is determined by system parameters or monitoring requirements (for example, 1Hz means collecting data once per second). Calculate the total duration within a single collection period based on the collection frequency to ensure reasonable division of the analysis period. Subdivide the monitoring period into multiple time segments (analysis periods) to give the data a time dimension and support time series analysis. Collect device operating status data within each analysis period to reflect the operating condition characteristics (such as speed, load, temperature, etc.) of the device during that period. Aggregate the operating status information of multiple analysis periods into analysis operating status data to form a set of time series features. According to the time interval corresponding to the differential voiceprint feature information, obtain the relevant device operating status information and summarize it as differential operating status data, which reflects the characteristics of the device operating status during the abnormal period. Extract the sound features matching the differential operating status data to form a differential sound feature vector. Compare the differential operating status data with the analysis operating status data to find the same operating status information, associate it with the differential sound feature vector, and summarize the association result into a label form to describe the abnormal sound features under a specific operating status. Organize multiple differential feature labels according to the operating status and sound features to construct a multi-modal feature label set containing multi-dimensional information such as time, status, and sound. In this embodiment, the association analysis of the operating status information and the sound features forms a multi-modal feature label set, which helps to comprehensively reflect the characteristic changes during device abnormalities, can accurately describe the operating conditions corresponding to abnormal sounds, and provides an intuitive basis for abnormal analysis.

[0150] In a preferred embodiment, the steps of constructing a voiceprint library by constructing a feature set label set containing multiple states according to the multi-modal feature label set and the initial voiceprint feature information, obtaining the dynamic feature values of each feature label in the feature set label set and assigning label retrieval levels include:

[0151] S501. Extract and mark the basic feature labels of the device in the normal operating state according to the initial voiceprint feature information;

[0152] S502. Combine the multi-modal feature label set with the basic feature labels to obtain a feature set label set containing multiple states, where the feature set label set includes multiple feature labels;

[0153] S503. Extract multiple voiceprint feature vectors for each feature label in the feature set label set;

[0154] S504. Obtain the dynamic feature value corresponding to each feature label according to the multiple voiceprint feature vectors of each feature label;

[0155] S505. Obtain a label hierarchy table, where the label hierarchy table includes multiple dynamic feature interval values and the label retrieval level corresponding to each dynamic feature interval value;

[0156] S506. Obtain the corresponding target dynamic feature interval value according to the dynamic feature value;

[0157] S507. Obtain the corresponding label retrieval level from the label hierarchy table according to the target dynamic feature interval value;

[0158] S508. Construct a voiceprint library according to the label retrieval level of each feature label in the feature set label set, and optimize the high-priority labels to the fast retrieval path.

[0159] In the above steps S501 to S508, according to the initial voiceprint feature information, extract and mark the basic feature labels of the device in the normal operation state as the initial benchmark of the voiceprint library, integrate the multi-modal feature label set (reflecting the voiceprint and operation characteristics of the device in the abnormal or special state) with the initial operation feature labels, generate a feature set label set including multiple states, covering the entire life cycle of the device operation, extract multiple voiceprint feature vectors for each feature label in the feature set label set, these vectors are the acoustic feature descriptions of the device in a specific operation state, perform statistical analysis on the voiceprint feature vectors of each feature label, generate a dynamic feature value, and the calculation formula of the dynamic feature value is , in the formula, represents the dynamic feature value, h represents the number of multiple voiceprint feature vectors, h = 2, 3, 4... t, and t represents the number of voiceprint feature vectors of the feature label, represents the h-th voiceprint feature vector, It is represented as the (h - 1)-th voiceprint feature vector. The label hierarchy table defines intervals of multiple dynamic feature values and assigns corresponding label retrieval levels (such as high priority, medium priority, etc.) to each interval to support efficient label classification and retrieval. According to each dynamic feature value, the corresponding target dynamic feature interval value is matched to determine the classification level of the feature label. According to the target dynamic feature interval value, the corresponding label retrieval level is found from the label hierarchy table, and a retrieval priority is assigned to each feature label. A voiceprint library is constructed according to the label retrieval levels of each feature label in the feature set label set. The high-priority labels are optimized to the fast retrieval path to improve the access efficiency of the voiceprint library. In this embodiment, each label in the voiceprint library not only stores its voiceprint feature information but also includes the association with the multi-modal features of the device operation state, providing complete support for subsequent voiceprint comparison and device status diagnosis. The voiceprint library can not only cover the features of the device in the normal operation state but also record the differential features in the abnormal or specific states, improving the applicable range of the voiceprint library. The data in the voiceprint library is not only statically stored but also can describe the dynamic behavior of the device operation characteristics. By assigning retrieval levels to feature labels through the label hierarchy table and optimizing important or high-frequency labels to the fast retrieval path, the query efficiency of the voiceprint library is greatly improved.

[0160] In a preferred embodiment, it further includes: constructing an update period, obtaining the new voiceprint feature change information generated by the synchronous condenser rotating device within the update period, and returning the new voiceprint feature change information as the voiceprint feature change information to the step of determining whether the voiceprint feature change information meets the first preset condition;

[0161] This step includes:

[0162] S601. Obtain the end time node of the monitoring period and mark it as the start time of the update period;

[0163] S602. Obtain the monitoring duration of the monitoring period;

[0164] S603. Obtain multiple dynamic feature values within the feature set label set;

[0165] S604. Obtain the update duration of the update period according to the multiple dynamic feature values and the monitoring duration;

[0166] S605. Obtain the end time of the update period according to the update duration and the start time of the update period;

[0167] S606. Obtain the update period according to the start time and the end time of the update period;

[0168] S607. Obtain the new voiceprint feature change information generated by the synchronous condenser rotating device within the update period;

[0169] S608: Return the new voiceprint feature change information as the voiceprint feature change information to the step of determining whether the voiceprint feature change information meets the first preset condition.

[0170] As in the above steps S601 to S608, after the monitoring period ends, its end time node is automatically set as the start time of the update period. According to the monitoring duration of the monitoring period, a reference for the subsequent dynamic update duration is provided. By combining multiple dynamic characteristic values and the monitoring duration of the monitoring period, the update duration is calculated so that the length of the update period can adapt to the frequency and fluctuation of the device characteristic changes. The calculation formula for the update duration is: , where E represents the update duration, q represents the number of multiple dynamic feature values, q=1,2,3…r, The calculated update duration is added to the start time of the update cycle to determine the end time of the update cycle, and a complete update cycle range is dynamically generated to provide a clear time window for the extraction of new voiceprint feature change information. During the update cycle, new voiceprint feature change information is extracted according to the new sound data generated by the device operation, including the changed voiceprint feature vector, feature change value, and corresponding operation status label, etc. The extracted new voiceprint feature change information is returned to the step of determining whether the voiceprint feature change information meets the first preset condition, and participates in the evaluation and screening of voiceprint feature changes to form a closed-loop feedback mechanism to ensure that new changes in device operation can be adjusted and adapted in time, and new changes in device operation can be captured and analyzed in real time to maintain the real-time and reliability of the voiceprint library. The feature data in the voiceprint library can be corrected and supplemented in time to avoid diagnostic deviations caused by the invalidation of old data, improve data accuracy, and automatically adjust according to the speed of device feature changes to effectively balance the efficiency of system resource use and the timeliness of data updates.

[0171] Based on the same inventive concept, the present invention also provides a voiceprint library construction device for phase-shifting rotating equipment, which is used to implement the above-mentioned voiceprint library construction method for phase-shifting rotating equipment, see Figure 2 , the device comprises:

[0172] An initial voiceprint module is used to obtain the initial operation sound data of the phase-shifting rotating device, and obtain the initial voiceprint feature information according to the initial operation sound data;

[0173] The feature change module is used to construct a monitoring cycle, obtain the operating sound data of the phase-shifting rotating equipment within the monitoring cycle, obtain the monitoring voiceprint feature information according to the operating sound data within the monitoring cycle, compare the monitoring voiceprint feature information with the initial voiceprint feature information, and obtain the voiceprint feature change information;

[0174] A feature comparison module, configured to determine whether the voiceprint feature change information conforms to a first preset condition. If not, it obtains the differential voiceprint feature information of the synchronous condenser rotating device according to the monitored voiceprint feature information;

[0175] A feature labeling module, configured to construct an analysis period, obtain the operation state data of the synchronous condenser rotating device within the analysis period, and construct a multi-modal feature label set corresponding to the device operation state according to the differential voiceprint feature information;

[0176] A voiceprint library module, configured to construct a feature set label set including multiple states according to the multi-modal feature label set and the initial voiceprint feature information, obtain the dynamic feature values of each feature label in the feature set label set and assign label retrieval levels, and construct a voiceprint library;

[0177] An update module, configured to construct an update period, obtain the new voiceprint feature change information generated by the synchronous condenser rotating device within the update period, and return the new voiceprint feature change information to the feature comparison module as the voiceprint feature change information.

[0178] As described above, the initial voiceprint module collects the sound data of the device during initial operation (such as the sound signal during normal startup or no-load condition), and analyzes it into the initial voiceprint feature information. The feature change module continuously collects the sound data during device operation within the set monitoring period, extracts the corresponding voiceprint feature information, and compares it with the initial voiceprint feature data to generate the voiceprint feature change information. The feature comparison module determines whether the voiceprint feature change information conforms to the first preset condition. If it does, the voiceprint feature is considered normal and no further processing is required. If not, the differential part in the monitored voiceprint feature information is extracted to generate the differential voiceprint feature information. The feature labeling module associates the device operation state data with the differential voiceprint feature information to construct a multi-modal feature label set corresponding to a specific state. The voiceprint library module further generates a feature set label set according to the multi-modal feature label set and the initial voiceprint feature information, and finally constructs a complete voiceprint library of the device through dynamic feature value calculation and label retrieval level division. The update module sets an update period according to the dynamic changes during device operation, continuously collects new voiceprint feature change information, and returns it to the feature comparison module to form a dynamically updated closed-loop system, continuously optimizing the content of the voiceprint library, enhancing the ability to identify long-term abnormal trends of the device, ensuring that the voiceprint library always reflects the latest operation state of the device. Through the construction of the multi-modal feature label set, the device operation state is deeply associated with the voiceprint feature, facilitating a more comprehensive operation state analysis. The voiceprint library not only supports real-time operation state monitoring but also provides a data basis for the historical analysis of the device operation state.

[0179] And, a voiceprint library construction terminal for a synchronous condenser rotating device, including:

[0180] One or more processors;

[0181] A storage device on which one or more programs are stored;

[0182] When the one or more programs are executed by one or more processors, the one or more processors implement a method for constructing a voiceprint library for a synchronous condenser rotating device.

[0183] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an all-hardware embodiment, an all-software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0184] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0185] These computer program instructions can 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, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0186] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent replacements, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A method for constructing a voiceprint library for a synchronous condenser rotating device, characterized in that Including: Obtain the initial operating sound data of the synchronous condenser rotating device, and obtain the initial voiceprint feature information according to the initial operating sound data; Construct a monitoring period, obtain the operating sound data of the synchronous condenser rotating device within the monitoring period, obtain the monitoring voiceprint feature information according to the operating sound data within the monitoring period, compare the monitoring voiceprint feature information with the initial voiceprint feature information, and obtain the voiceprint feature change information; Judge whether the voiceprint feature change information meets the first preset condition. If not, obtain the differential voiceprint feature information of the synchronous condenser rotating device according to the monitoring voiceprint feature information; Construct an analysis period, obtain the operating state data of the synchronous condenser rotating device within the analysis period, and construct a multi-modal feature label set corresponding to the device operating state according to the differential voiceprint feature information; Construct a feature set label set including multiple states according to the multi-modal feature label set and the initial voiceprint feature information, obtain the dynamic feature values of each feature label in the feature set label set and assign label retrieval levels, and construct a voiceprint library.

2. The method for constructing a voiceprint library for a synchronous condenser rotating device according to claim 1, wherein The steps of obtaining the initial operating sound data of the synchronous condenser rotating device and obtaining the initial voiceprint feature information according to the initial operating sound data include: Obtain the initial operating sound data of the synchronous condenser rotating device; Divide the initial operating sound data into multiple sound segment information; Extract features from each sound segment to generate corresponding initial sound feature vectors; Summarize multiple initial sound feature vectors, and mark the summary result as the initial voiceprint feature information.

3. The method for constructing a voiceprint library for a synchronous condenser rotating device according to claim 1, wherein The steps of constructing a monitoring period, obtaining the operating sound data of the synchronous condenser rotating device within the monitoring period, obtaining the monitoring voiceprint feature information according to the operating sound data within the monitoring period, comparing the monitoring voiceprint feature information with the initial voiceprint feature information, and obtaining the voiceprint feature change information include: Construct a monitoring period; Obtain the operating sound data of the synchronous condenser rotating device within the monitoring period and mark it as the monitoring operating sound data; Extract the monitoring voiceprint feature information corresponding to the monitoring operating sound data; Decompose the monitoring voiceprint feature information into multiple monitoring sound feature vectors; Obtain the voiceprint feature change value according to multiple monitoring sound feature vectors; Obtain the voiceprint feature evaluation threshold; Judge whether the voiceprint feature change value exceeds the voiceprint feature evaluation threshold; If the voiceprint feature change value does not exceed the voiceprint feature evaluation threshold, it is determined that the voiceprint feature change information is stable; If the voiceprint feature change value exceeds the voiceprint feature evaluation threshold, it is determined that the voiceprint feature change information is fluctuating; The calculation of the voiceprint feature change value is: , where represents the voiceprint feature change value, i represents the number of multiple monitored voice feature vectors, i = 1, 2, 3... n, and n represents the number of monitored voice feature vectors, represents the i-th monitored voice feature vector.

4. The method for constructing a voiceprint library for a synchronous condenser rotating device according to claim 3, wherein, The steps of judging whether the voiceprint feature change information meets the first preset condition. If not, obtaining the differential voiceprint feature information of the synchronous condenser rotating device according to the monitoring voiceprint feature information include: Obtain the voiceprint difference evaluation threshold; Obtain the corresponding voiceprint feature change value according to the voiceprint feature change information; Judge whether the voiceprint feature change value exceeds the voiceprint difference evaluation threshold; If the voiceprint feature change value exceeds the voiceprint difference evaluation threshold, it is determined that the operating sound of the synchronous condenser rotating device is abnormal, and the abnormal difference between the voiceprint feature change value and the voiceprint difference evaluation threshold is obtained; Obtain differential voiceprint feature information based on the abnormal difference, monitored voiceprint feature information, and initial voiceprint feature information.

5. The method for constructing a voiceprint library for a synchronous condenser rotating device according to claim 4, wherein, The step of obtaining differential voiceprint feature information based on the abnormal difference, monitored voiceprint feature information, and initial voiceprint feature information includes: Obtain the abnormal difference between the voiceprint feature change value and the voiceprint difference evaluation threshold; Obtain a difference table, where the difference table includes multiple abnormal difference intervals and the corresponding voiceprint similarity thresholds for each abnormal difference interval; Obtain the corresponding voiceprint similarity threshold from the difference table according to the abnormal difference; Obtain the corresponding multiple monitored voice feature vectors according to the monitored voiceprint feature information; Obtain the corresponding multiple initial voice feature vectors according to the initial voiceprint feature information; Obtain the voiceprint feature similarity value based on each monitored voice feature vector and the corresponding initial voice feature vector; Judge whether multiple voiceprint feature similarity values exceed the obtained voiceprint similarity threshold; If there is a voiceprint feature similarity value that exceeds the voiceprint similarity threshold, summarize the voiceprint feature similarity values that exceed the voiceprint similarity threshold, and mark the summary result as differential voiceprint feature information; The calculation of the voiceprint feature similarity value is: , where Y represents the voiceprint feature similarity value, A represents the initial voice feature vector, and B represents the monitored voice feature vector.

6. The method for constructing a voiceprint library for a synchronous condenser rotating device according to claim 5, wherein The step of constructing the analysis period, obtaining the operation state data of the synchronous condenser rotating equipment within the analysis period, and constructing a multi-modal feature label set corresponding to the equipment operation state according to the differential voiceprint feature information includes: Obtain the acquisition frequency of the operation state of the synchronous condenser rotating equipment; Obtain the acquisition duration according to the acquisition frequency; Divide the monitoring period into multiple analysis periods according to the acquisition duration; Obtain the operation state information of the synchronous condenser rotating equipment within each analysis period; Summarize multiple operation state information to form a set of time series features, and mark it as analysis operation state data; According to the time interval corresponding to the differential voiceprint feature information, obtain the corresponding operation state data of the synchronous condenser rotating equipment, and summarize it as differential operation state data; Extract the voice features matching the differential operation state data to form differential voice feature vectors; Obtain the same operation state information in the differential operation state data and the analysis operation state data, associate the same operation state information with the differential voice feature vectors, and summarize the association result in the form of a label to construct differential feature labels; Summarize multiple differential feature labels, and mark the summary result as a multi-modal feature label set.

7. The method for constructing a voiceprint library for a synchronous condenser rotating device according to claim 6, characterized in that, The step of constructing a feature set label set containing multiple states according to the multi-modal feature label set and the initial voiceprint feature information, obtaining the dynamic feature value of each feature label in the feature set label set and assigning a label retrieval level, and constructing a voiceprint library includes: Extract and mark the basic feature labels of the equipment in the normal operation state according to the initial voiceprint feature information; Integrate the multi-modal feature label set with the basic feature labels to obtain a feature set label set containing multiple states, where the feature set label set includes multiple feature labels; Extract multiple voiceprint feature vectors of each feature label in the feature set label set; Obtain the dynamic feature value of the corresponding feature label according to the multiple voiceprint feature vectors of each feature label; Obtain a label level table, where the label level table includes multiple dynamic feature interval values and the label retrieval levels corresponding to each dynamic feature interval value; Obtain the corresponding target dynamic feature interval value from the label level table according to the dynamic feature value; Obtain the corresponding label retrieval level from the label level table according to the target dynamic feature interval value; Construct a voiceprint library based on the label retrieval levels of each feature label in the feature set label set, and optimize the high-priority labels to the fast retrieval path; The dynamic feature value is calculated as: , where represents the dynamic eigenvalue, h represents the number of multiple voiceprint feature vectors, h = 2, 3, 4... t, and t represents the number of voiceprint feature vectors of the feature label represents the h-th voiceprint feature vector represents the (h - 1)-th voiceprint feature vector 8. The method for constructing a voiceprint library for a synchronous condenser rotating device according to claim 1, wherein The method further includes: constructing an update period, obtaining new voiceprint feature change information generated by the synchronous condenser rotating device within the update period, and returning the new voiceprint feature change information as the voiceprint feature change information to the step of determining whether the voiceprint feature change information meets the first preset condition; This step includes: Obtain the end time node of the monitoring period and mark it as the start time of the update period; Obtain the monitoring duration of the monitoring period; Obtain multiple dynamic feature values within the feature set label set; Obtain the update duration of the update period according to the multiple dynamic feature values and the monitoring duration; Obtain the end time of the update period according to the update duration and the start time of the update period; Obtain the update period according to the start time and the end time of the update period; Obtain new voiceprint feature change information generated by the synchronous condenser rotating device within the update period; Return the new voiceprint feature change information as the voiceprint feature change information to the step of determining whether the voiceprint feature change information meets the first preset condition; The update duration of the update period is calculated as: , where E represents the update duration, q represents the number of multiple dynamic eigenvalues, q = 1, 2, 3... r, represents the q-th dynamic eigenvalue, r represents the number of dynamic eigenvalues, and T represents the monitoring duration.

9. An acoustic fingerprint library construction device for a synchronous condenser rotating device, which is applied to the acoustic fingerprint library construction method for a synchronous condenser rotating device according to any one of claims 1 to 8, and is characterized in that Includes: An initial voiceprint module, configured to obtain the initial operation sound data of the synchronous condenser rotating device, and obtain the initial voiceprint feature information according to the initial operation sound data; A feature change module, which constructs a monitoring period, obtains the operation sound data of the synchronous condenser rotating device within the monitoring period, obtains the monitoring voiceprint feature information according to the operation sound data within the monitoring period, compares the monitoring voiceprint feature information with the initial voiceprint feature information, and obtains the voiceprint feature change information; A feature comparison module, configured to determine whether the voiceprint feature change information meets the first preset condition. If not, obtain the differential voiceprint feature information of the synchronous condenser rotating device according to the monitoring voiceprint feature information; A feature label module, configured to construct an analysis period, obtain the operation state data of the synchronous condenser rotating device within the analysis period, and construct a multi-modal feature label set corresponding to the device operation state according to the differential voiceprint feature information; A voiceprint library module, configured to construct a feature set label set including multiple states according to the multi-modal feature label set and the initial voiceprint feature information, obtain the dynamic feature values of each feature label in the feature set label set and assign label retrieval levels, and construct a voiceprint library; An update module, configured to construct an update period, obtain new voiceprint feature change information generated by the synchronous condenser rotating device within the update period, and return the new voiceprint feature change information as the voiceprint feature change information to the feature comparison module.

10. A voiceprint library construction terminal for a synchronous condenser rotating device, characterized in that, Includes: One or more processors; A storage device, on which one or more programs are stored; When one or more programs are executed by one or more processors, such that the one or more processors implement the method for constructing a voiceprint library for a synchronous condenser rotating device according to any one of claims 1 to 8.

Images