A fault monitoring data transmission system and method applied to an excitation power cabinet

Abstract

The application discloses a kind of fault monitoring data transmission systems and methods applied to excitation power cabinet, it is related to excitation power cabinet technical field, comprising: call the historical fault monitoring record of excitation power cabinet, deploy sensor and collect real-time monitoring data, determine the target sensor type corresponding to each fault type;Capture the repair behavior of each agent, obtain the repair capacity of each agent to different fault types, build the ability image of agent;Determine the fault degree of various fault types in excitation power cabinet;Based on the ability image of agent and the fault degree of fault type, judge each agent's target value, determine target agent and carry out repair work to excitation power cabinet.The application determines target agent by analyzing fault monitoring record, combining with agent ability image, reduces the problem of unreasonable task matching, realizes the accurate distribution transmission of fault monitoring task data, and guarantees that fault monitoring and repair work efficiently promote.

Description

Technical Field

[0001] This invention relates to the field of excitation power cabinet technology, specifically a fault monitoring data transmission system and method for excitation power cabinets. Background Technology

[0002] Large synchronous motors rely on the excitation system power cabinet to regulate the excitation current, thereby controlling the magnetic field strength. Currently, power cabinet fault monitoring largely depends on collaborative inspection operations by maintenance agents. In actual industrial scenarios, the aging degree, heat dissipation performance, and component wear of each cabinet vary individually. Traditional solutions use manual assignment or random distribution for fault monitoring tasks, without assessing the cabinet status in conjunction with the real-time operating conditions of the equipment. However, due to the different capability profiles of each agent, manual assignment or random distribution does not consider differentiated maintenance criteria, leading to unreasonable task matching and significantly reduced maintenance efficiency. This not only reduces the accuracy of fault identification but also delays the timing of fault assessment and handling. Therefore, it is necessary to combine the actual operating conditions of the excitation power cabinet to achieve accurate distribution and transmission of fault monitoring task data, ensuring the efficient progress of fault monitoring and maintenance work. Summary of the Invention

[0003] The purpose of this invention is to provide a fault monitoring data transmission system and method for excitation power cabinets, so as to solve the problems raised in the prior art.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A fault monitoring data transmission method for excitation power cabinets includes the following steps: Retrieve historical fault monitoring records of the excitation power cabinet, deploy several sensors on the excitation power cabinet, and determine the target sensor type corresponding to each fault type based on the real-time monitoring data of each sensor within the time window before the fault occurs, according to the monitoring records corresponding to different fault types. Based on the historical repair behavior of each agent, the repair capabilities of each agent for different types of faults are quantified, and a capability profile of each agent is constructed accordingly. The real-time monitoring data of each sensor is transmitted to the cloud platform for analysis. Based on the target sensor type corresponding to each fault type, the degree of fault of each fault type in the excitation power cabinet is determined. Based on the capability profile of the intelligent agent and the degree of failure of various fault types, the target value of each intelligent agent for the excitation power cabinet is determined, and then the target intelligent agent is identified. The fault monitoring task data is transmitted to the target intelligent agent so that the target intelligent agent can perform maintenance work on the excitation power cabinet.

[0005] Preferably, the target sensor type corresponding to each fault type is determined, including: Extract the fault time, fault power cabinet and fault type corresponding to the fault monitoring record. Deploy several sensors on the excitation power cabinet and extract the real-time monitoring data of each sensor within the preceding time window before the fault time. The real-time monitoring data is the time series data of the sensor value changing over time. From the normal operating period of the excitation power cabinet without faults, the stable operating period is selected and extracted, and based on the stable operating period, the normal fluctuation range corresponding to each sensor type is calculated. Within the preceding time window corresponding to fault type A, the number of data points for each sensor type that exceed the corresponding normal fluctuation range is counted, and the sensor type with the largest number is taken as the target sensor type for fault type A, thereby determining the target sensor type corresponding to each fault type.

[0006] Preferably, the process of selecting and extracting stable operating periods includes: obtaining a normal operating period during which the excitation power cabinet is fault-free, collecting audio signals at multiple times within the operating period, converting the audio data at each time time into corresponding spectrograms, calculating the cosine similarity between any two spectrograms, and if the cosine similarity is greater than a preset similarity threshold, then the operating period is taken as a stable operating period.

[0007] It should be noted that most faults will show abnormal changes in relevant monitoring parameters before they become apparent, allowing for fault prediction. For example, an abnormal pre-current indicates a potential problem in the power device or rectifier circuit, which can easily lead to faults such as abnormal thyristor conduction, loss sharing imbalance, or overcurrent. This corresponds to the target sensing type for the fault type determined in this step. The normal fluctuation range then serves as the basis for judging whether the relevant parameters deviate from the normal range. In this scheme, the normal fluctuation range can be determined based on the stable operating period, as follows: Preferably, the normal fluctuation range corresponding to each sensor type is obtained by: acquiring several sensor values ​​corresponding to a certain sensor type during a stable operating period, calculating the average value µ and standard deviation σ, obtaining the normal fluctuation range [µ-3σ, µ+3σ] of the sensor type based on the 3σ principle, and then calculating the normal fluctuation range of all sensor types during the stable operating period based on the 3σ principle.

[0008] Preferably, construct capability profiles for various intelligent agents, including: Retrieve the repair records of a certain intelligent agent for the excitation power cabinet fault, and extract the corresponding fault location, repair time, fault type and fault level from the repair records. The intelligent agent is a maintenance robot, and the repair time is the time interval from the start of the intelligent agent's operation to the successful repair of the fault. The intelligent agent follows a standardized repair process when performing fault repair operations. Based on the standardized repair process, reliable records are selected and extracted from the repair records. Obtain multiple reliable records of the agent's repair of fault type A. Based on the corresponding repair time and fault level, obtain the agent's repair capability value for fault type A: Where T is the number of reliable records, Dee t Let S be the fault level corresponding to the t-th reliable record. t Let t be the repair time corresponding to the t-th reliable record; obtain the repair capability value of the agent for each type of fault, and construct the capability profile of each agent accordingly.

[0009] It should be noted that the formula y=1-e -x It is a function where y takes values ​​from 0 to 1 when x takes the value x>0, and y increases as x increases. Therefore, in this scheme, Dee t / S t The larger the value, the greater the agent's ability to repair fault type A, while Dee t This represents the fault level corresponding to the t-th reliable record. A larger value indicates a greater ability of the agent to repair the fault type. S t The repair time is used to represent the repair time corresponding to the t-th reliable record. The shorter the repair time, the greater the agent's ability to repair the fault type. Therefore, Dee is used here. t / S t Substituting these variables into the formula can effectively quantify the repair capability of the intelligent agent.

[0010] Preferably, reliable records are selected and extracted from the repaired records, including: Obtain the standardized repair process corresponding to fault type A. The standardized repair process includes multiple standard execution steps with a sequential relationship. The actual repair process of a certain repair record is analyzed and broken down into each actual execution step. If an actual execution step belongs to a step in the standardized repair process, and the previous standard step in the standardized repair process also appears in the actual repair process and is located before the actual execution step, and the next standard step in the standardized repair process also appears in the actual repair process and is located after the actual execution step, then the actual execution step is considered a reliable step. The number of reliable steps N0 in the actual repair process is counted. Based on the number of standard execution steps N in the standardized repair process, the reliability of the repair record is obtained as N0 / N. If the reliability is greater than the preset threshold, the repair record is considered a reliable record.

[0011] Preferably, the severity of various fault types in the excitation power cabinet is determined, including: Based on the target sensing type L corresponding to the current fault type A AReal-time monitoring data is used to construct a front-end histogram H. A ; Extract the fault monitoring record R with fault type A, and collect the target sensor type L before the fault occurred. A Based on the real-time monitoring data, construct the baseline histogram corresponding to the fault monitoring record R, and then construct the baseline histogram corresponding to each fault monitoring record with fault type A. Extract the fault level of each fault monitoring record and calculate the pre-historical diagram H for each record. A The cosine similarity between each baseline histogram is used to calculate the fault severity of fault type A in the excitation power cabinet. The cosine similarity is then multiplied by the corresponding fault level in sequence, and the average of all the product results is calculated.

[0012] Preferably, a pre-construction histogram H is constructed. A The steps include: determining the target sensing type L A Divide the monitoring data into multiple intervals, count the number of data samples in each interval, calculate the probability distribution of each interval, and construct a preliminary histogram H based on the interval probabilities. A .

[0013] Preferably, the step of determining the target intelligent agent includes: retrieving the capability profile of any intelligent agent, combining the repair capability value of the intelligent agent for various fault types, and the fault severity of various fault types in the excitation power cabinet, multiplying the fault severity of each type of fault by the corresponding repair capability value and summing them to calculate the target value of the intelligent agent; obtaining the target value of each intelligent agent, and taking the intelligent agent with the largest target value as the target intelligent agent for overhauling the excitation power cabinet.

[0014] In determining the target value of the intelligent agent, this solution combines the agent's repair capability value for various fault types and the fault severity of each fault type. High-value evaluation items are selected by screening the fault severity, allowing the agent's capabilities to be fully demonstrated when handling high-level faults. This effectively reduces interference from inefficient evaluations and makes the target value indicators more realistic and better reflect the agent's maintenance capabilities.

[0015] A fault monitoring data transmission system for excitation power cabinets includes a memory and a processor, as well as a computer program stored in the memory and running on the processor. The processor is coupled to the memory, and the processor implements the aforementioned fault monitoring data transmission method for excitation power cabinets when executing the computer program.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides a fault monitoring data transmission system and method for excitation power cabinets, including: retrieving historical fault monitoring records of the excitation power cabinet; deploying sensors and collecting real-time monitoring data; determining the target sensor type corresponding to each fault type; capturing the repair behavior of each intelligent agent to obtain the repair capabilities of each intelligent agent for different fault types, and constructing a capability profile of the intelligent agent; determining the fault severity of various fault types in the excitation power cabinet; and based on the capability profile of the intelligent agent and the fault severity of the fault type, judging the target value of each intelligent agent, determining the target intelligent agent, and performing maintenance work on the excitation power cabinet. This invention, by analyzing fault monitoring records and combining intelligent agent capability profiles, determines the target intelligent agent, reduces the problem of unreasonable task matching, achieves accurate distribution and transmission of fault monitoring task data, and ensures efficient progress of fault monitoring and maintenance work. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a fault monitoring data transmission method for an excitation power cabinet according to the present invention. Detailed Implementation

[0019] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0020] Example: Figure 1 As shown, this invention provides a technical solution for a fault monitoring data transmission method applied to an excitation power cabinet, comprising the following steps: Retrieve historical fault monitoring records from the excitation power cabinet, deploy several sensors on the excitation power cabinet, and based on the real-time monitoring data of each sensor within the time window before the fault occurred, corresponding to the monitoring records for different fault types, determine the target sensor type corresponding to each fault type, including: Extract the fault time, fault power cabinet, and fault type corresponding to the fault monitoring record. In this embodiment, the fault type includes open circuit of thyristor, abnormal voltage equalization of rectifier, abnormal current equalization, etc. Deploy several sensors on the excitation power cabinet and extract the real-time monitoring data of each sensor within the preceding time window before the fault time. The real-time monitoring data is the time series data of the sensor value changing over time. The real-time monitoring data refers to the time series data continuously collected by the sensor at a fixed sampling frequency, such as once every 5 seconds.

[0021] From the normal operating period of the excitation power cabinet without faults, the stable operating period is selected and extracted, specifically: To obtain a fault-free normal operating period of a certain excitation power cabinet, audio signals at multiple times within this operating period are collected, and the audio data at each time time is converted into corresponding spectrograms. The cosine similarity between any two spectrograms is calculated. If the cosine similarity is greater than the preset similarity threshold, then this operating period is regarded as a stable operating period.

[0022] It should be noted that when the excitation power cabinet is fault-free and operating stably, the sound is characterized by stable frequency, uniform amplitude, and no abnormal noise. Since the spectrum diagram describes the relationship between amplitude and frequency with frequency as the horizontal axis and signal amplitude as the vertical axis, it can reflect the acoustic characteristics of the equipment operation. Therefore, if the spectrum diagrams corresponding to any two moments during normal operation are similar, it indicates that the period is in a stable operating condition. In this solution, the similarity of spectrum diagrams is characterized by cosine similarity. Since cosine similarity is an existing technology, it will not be elaborated on here. Since the value of cosine similarity ranges from 0 to 1, the closer it is to 1, the more similar the two spectrum diagrams are. Therefore, the similarity threshold here is 0.98, and the specific value can be determined according to the actual operating conditions.

[0023] It should be noted that most faults will show abnormal changes in relevant monitoring parameters before they become apparent, allowing for fault prediction. For example, an abnormal pre-current indicates a potential problem in the power device or rectifier circuit, which can easily lead to faults such as abnormal thyristor conduction, loss sharing imbalance, or overcurrent. This corresponds to the target sensing type for the fault type determined in this step. The normal fluctuation range then serves as the basis for judging whether the relevant parameters deviate from the normal range. In this scheme, the normal fluctuation range can be determined based on the stable operating period, as follows: Obtain several sensor values ​​corresponding to a certain sensor type during a stable operating period, and calculate the average value µ and standard deviation σ. Based on the 3σ principle, obtain the normal fluctuation range [µ-3σ, µ+3σ] of the sensor type. Then, based on the 3σ principle, calculate the normal fluctuation range of all sensor types during the stable operating period.

[0024] The 3σ principle is an existing technology. Here, we take the calculation steps to determine the normal fluctuation range of current as an example: First, collect several currents during the stable operating period, and then calculate the average value µ and the standard deviation σ. The normal fluctuation range can be obtained as [µ-3σ, µ+3σ]. This is because the probability of the data falling within the range of the average value ± 3 times the standard deviation is 99.73%, and the probability of falling outside this range is only 0.27%, which is a low probability event. Under stable operating conditions, the parameters rarely exceed this range. Once they exceed the limit, it can be basically determined that the parameters are abnormal. Therefore, the 3σ principle can be used to define the normal fluctuation range to determine whether the current data is abnormal.

[0025] Within the preceding time window corresponding to fault type A, the number of data points for each sensor type that exceed the corresponding normal fluctuation range is counted, and the sensor type with the largest number is taken as the target sensor type for fault type A, thereby determining the target sensor type corresponding to each fault type.

[0026] Based on the historical repair behavior of each agent, the repair capabilities of each agent for different fault types are quantified, and a capability profile of each agent is constructed accordingly. Retrieve the repair records of a certain intelligent agent for the excitation power cabinet fault, and extract the corresponding fault location, repair time, fault type and fault level from the repair records. The intelligent agent is a maintenance robot, and the repair time is the time interval from the start of the intelligent agent's operation to the successful repair of the fault. All fault repair operations performed by the intelligent agent follow a standardized repair process. Based on this process, reliable records are selected and extracted from the repair logs, including: Obtain the standardized repair process corresponding to fault type A. The standardized repair process includes multiple standard execution steps with a sequential relationship. The actual repair process of a certain repair record is analyzed and broken down into each actual execution step. If an actual execution step belongs to a step in the standardized repair process, and the previous standard step in the standardized repair process also appears in the actual repair process and is located before the actual execution step, and the next standard step in the standardized repair process also appears in the actual repair process and is located after the actual execution step, then the actual execution step is considered a reliable step. Here's an example: Suppose the standardized repair process is a→b→c→d→e, while the actual repair process is a→b→q→c→e, where a, b, c, d, e, and q are the respective execution stages. Taking the actual execution step b as an example, execution step b is a step in the standardized repair process. The standard step a preceding execution step b in the standardized repair process also appears in the actual repair process and is located before execution step b. The standard step c following execution step b in the standardized repair process also appears in the actual repair process and is located after execution step b. Therefore, the actual execution step b can be regarded as a reliable step. However, for the actual execution step c, although it belongs to the standardized repair process, the subsequent standard step d in the standardized repair process does not appear in the actual repair process. Therefore, although execution step c belongs to the standardized repair process, the absence of the subsequent standard step d in the standardized repair process indicates that there is a standard step omission in the actual repair process and the operation does not meet the specifications. Thus, execution step c cannot be considered a reliable step. Therefore, in the embodiment, only a, b, and e are reliable steps, a total of 3.

[0027] The number of reliable steps (N0) in the actual repair process is counted. Based on the number of standard execution steps (N) in the standardized repair process, the reliability of the repair record is calculated as N0 / N. If the reliability is greater than a preset threshold, the repair record is considered a reliable record. In this embodiment, N0 is 3 and N is 5, so the reliability is 0.6. In this embodiment, the threshold can be set to 0.8, therefore this repair record cannot be considered a reliable record. The specific threshold value can be determined according to the actual situation, and will not be elaborated here.

[0028] Obtain multiple reliable records of the agent's repair of fault type A. Based on the corresponding repair time and fault level, obtain the agent's repair capability value for fault type A: Where T is the number of reliable records, Dee t Let S be the fault level corresponding to the t-th reliable record. t Let t be the repair time corresponding to the t-th reliable record; obtain the repair capability value of the agent for each type of fault, and construct the capability profile of each agent accordingly.

[0029] In this embodiment, the fault level includes 5 levels, from level 1 to level 5. The higher the level, the more serious the fault.

[0030] It should be noted that the formula y=1-e -x It is a function where y takes values ​​from 0 to 1 when x takes the value x>0, and y increases as x increases. Therefore, in this scheme, Dee t / S t The larger the value, the greater the agent's ability to repair fault type A, while Deet This represents the fault level corresponding to the t-th reliable record. A larger value indicates a greater ability of the agent to repair the fault type. S t The repair time is used to represent the repair time corresponding to the t-th reliable record. The shorter the repair time, the greater the agent's ability to repair the fault type. Therefore, Dee is used here. t / S t Substituting these variables into the formula can effectively quantify the repair capability of the intelligent agent.

[0031] The real-time monitoring data of each sensor is transmitted to the cloud platform for analysis. Based on the target sensor type corresponding to each fault type, the severity of various fault types in the excitation power cabinet is determined, including: Based on the target sensing type L corresponding to the current fault type A A Real-time monitoring data is used to construct a front-end histogram H. A Specifically: For target sensing type L A Divide the monitoring data into multiple intervals, such as target sensing type L. A When the current is ampere, it is divided into several intervals according to 0.6-0.8, 0.8-1, 1-1.2, and 1.2-1.4, denoted by ampere. The number of data samples in each interval is counted, and the probability distribution corresponding to each interval is calculated. Based on the interval probabilities, a preliminary histogram H is constructed. A ; Histograms, using data intervals and corresponding probability distributions as feature vectors, fully preserve the overall distribution pattern and discrete characteristics of target sensing data. They can objectively depict the parameter patterns corresponding to the operating state. Before the occurrence of the same type of fault, the abnormal changes in target sensing parameters have fixed characteristics, and the corresponding histogram shapes are highly similar. Therefore, in this scheme, the histogram matching degree is quantified by cosine similarity, which can accurately reflect the similarity between the current operating condition and historical fault precursors, providing a reliable basis for fault severity assessment.

[0032] Extract the fault monitoring record R with fault type A, and collect the target sensor type L before the fault occurred. A Based on the real-time monitoring data, construct the baseline histogram corresponding to the fault monitoring record R, and then construct the baseline histogram corresponding to each fault monitoring record with fault type A. Extract the fault level of each fault monitoring record and calculate the pre-historical diagram H for each record. A The cosine similarity between each baseline histogram is used to calculate the fault severity of fault type A in the excitation power cabinet. This is then multiplied sequentially by the corresponding fault level, and the average of all products is calculated. The specific formula is as follows: Where M is the number of fault monitoring records, and since each fault monitoring record corresponds to a baseline histogram, the number of baseline histograms is also M, sim m For the preceding histogram H A The cosine similarity between Dee and the m-th baseline histogram m This represents the fault level corresponding to the m-th fault monitoring record.

[0033] Based on the capability profiles of the intelligent agents and the severity of various fault types, the target values ​​of each intelligent agent for the excitation power cabinet are determined, thereby identifying the target intelligent agent. Fault monitoring task data is then transmitted to the target intelligent agent, enabling it to perform maintenance operations on the excitation power cabinet, including: Retrieve the capability profile of any intelligent agent, combine it with the agent's repair capability values ​​for various fault types, and the fault severity of various fault types in the excitation power cabinet. Multiply the fault severity of each type of fault by the corresponding repair capability value and sum them to calculate the target value of the intelligent agent, which is represented by the formula: Where P is the target value of the agent, G is the number of fault types, and W... g Y represents the repair capability value for the g-th fault type. g The fault severity is the g-th fault type.

[0034] In determining the target value of the intelligent agent, this solution combines the agent's repair capability value for various fault types and the fault severity of each fault type. High-value evaluation items are selected by screening the fault severity, allowing the agent's capabilities to be fully demonstrated when handling high-level faults. This effectively reduces interference from inefficient evaluations and makes the target value indicators more realistic and better reflect the agent's maintenance capabilities.

[0035] Calculate the target value for each agent, and select the agent with the largest target value as the target agent for overhauling the excitation power cabinet.

[0036] This embodiment also provides a fault monitoring data transmission system for excitation power cabinets, including a memory, a processor, and a computer program stored in the memory and running on the processor. The processor is coupled to the memory, and when the processor executes the computer program, it implements the aforementioned fault monitoring data transmission method for excitation power cabinets. Since this fault monitoring data transmission method for excitation power cabinets has already been described in detail above, it will not be repeated here.

[0037] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0038] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0039] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fault monitoring data transmission method applied to an excitation power cabinet, characterized in that, Includes the following steps: Retrieve historical fault monitoring records of the excitation power cabinet, deploy several sensors on the excitation power cabinet, and determine the target sensor type corresponding to each fault type based on the real-time monitoring data of each sensor within the time window before the fault occurs, according to the monitoring records corresponding to different fault types. Based on the historical repair behavior of each agent, the repair capabilities of each agent for different types of faults are quantified, and a capability profile of each agent is constructed accordingly. The real-time monitoring data of each sensor is transmitted to the cloud platform for analysis. Based on the target sensor type corresponding to each fault type, the degree of fault of each fault type in the excitation power cabinet is determined. Based on the capability profile of the intelligent agent and the degree of failure of various fault types, the target value of each intelligent agent for the excitation power cabinet is determined, and then the target intelligent agent is identified. The fault monitoring task data is transmitted to the target intelligent agent so that the target intelligent agent can perform maintenance work on the excitation power cabinet.

2. The fault monitoring data transmission method for excitation power cabinet according to claim 1, characterized in that, Determine the target sensor type corresponding to each fault type, including: Extract the fault time, fault power cabinet and fault type corresponding to the fault monitoring record, deploy several sensors on the excitation power cabinet, and extract the real-time monitoring data of each sensor within the preceding time window before the fault time. The real-time monitoring data is the time series data of the sensor value changing over time. From the normal operating period of the excitation power cabinet without faults, the stable operating period is selected and extracted, and based on the stable operating period, the normal fluctuation range corresponding to each sensor type is calculated. Within the preceding time window corresponding to fault type A, the number of data points for each sensor type that exceed the corresponding normal fluctuation range is counted, and the sensor type with the largest number is taken as the target sensor type for fault type A, thereby determining the target sensor type corresponding to each fault type.

3. The fault monitoring data transmission method for excitation power cabinet according to claim 2, characterized in that, The process of filtering and extracting stable operating periods includes: obtaining a normal operating period during which the excitation power cabinet is fault-free; collecting audio signals at multiple times within the operating period; converting the audio data at each time time into corresponding spectrograms; calculating the cosine similarity between any two spectrograms; and if the cosine similarity is greater than a preset similarity threshold, then the operating period is considered a stable operating period.

4. The fault monitoring data transmission method for excitation power cabinet according to claim 2, characterized in that, To determine the normal fluctuation range for each sensor type, the process includes: obtaining several sensor values ​​corresponding to a certain sensor type during a stable operating period, calculating the average value µ and standard deviation σ, obtaining the normal fluctuation range [µ-3σ, µ+3σ] for the sensor type based on the 3σ principle, and then calculating the normal fluctuation range for all sensor types during the stable operating period based on the 3σ principle.

5. The fault monitoring data transmission method for an excitation power cabinet according to claim 1, characterized in that, Construct capability profiles for various intelligent agents, including: Retrieve the repair records of a certain intelligent agent for the excitation power cabinet fault, and extract the corresponding fault location, repair time, fault type and fault level from the repair records. The intelligent agent is a maintenance robot, and the repair time is the time interval from the start of the intelligent agent's operation to the successful repair of the fault. The intelligent agent follows a standardized repair process when performing fault repair operations. Based on the standardized repair process, reliable records are selected and extracted from the repair records. Obtain multiple reliable records of the agent's repair of fault type A. Based on the corresponding repair time and fault level, obtain the agent's repair capability value for fault type A: Where T is the number of reliable records, Dee t Let S be the fault level corresponding to the t-th reliable record. t Let t be the repair time corresponding to the t-th reliable record; obtain the repair capability value of the agent for each type of fault, and construct the capability profile of each agent accordingly.

6. The fault monitoring data transmission method for an excitation power cabinet according to claim 5, characterized in that, Reliable records are extracted from the repaired records, including: Obtain the standardized repair process corresponding to fault type A. The standardized repair process includes multiple standard execution steps with a sequential relationship. The actual repair process of a certain repair record is analyzed and broken down into each actual execution step. If an actual execution step belongs to a step in the standardized repair process, and the previous standard step in the standardized repair process also appears in the actual repair process and is located before the actual execution step, and the next standard step in the standardized repair process also appears in the actual repair process and is located after the actual execution step, then the actual execution step is considered a reliable step. The number of reliable steps N0 in the actual repair process is counted. Based on the number of standard execution steps N in the standardized repair process, the reliability of the repair record is obtained as N0 / N. If the reliability is greater than a preset threshold, the repair record is considered a reliable record.

7. The fault monitoring data transmission method for an excitation power cabinet according to claim 1, characterized in that, Determine the severity of various fault types in the excitation power cabinet, including: Based on the target sensing type L corresponding to the current fault type A A Real-time monitoring data is used to construct a front-end histogram H. A ; Extract the fault monitoring record R with fault type A, and collect the target sensor type L before the fault occurred. A Based on the real-time monitoring data, construct the baseline histogram corresponding to the fault monitoring record R, and then construct the baseline histogram corresponding to each fault monitoring record with fault type A. Extract the fault level of each fault monitoring record and calculate the pre-historical diagram H for each record. A The cosine similarity between each baseline histogram is used to calculate the fault severity of fault type A in the excitation power cabinet. The cosine similarity is then multiplied by the corresponding fault level in sequence, and the average of all the product results is calculated.

8. A fault monitoring data transmission method for an excitation power cabinet according to claim 7, characterized in that, Constructing the pre-histogram H A The steps include: determining the target sensing type L A Divide the monitoring data into multiple intervals, count the number of data samples in each interval, calculate the probability distribution of each interval, and construct a preliminary histogram H based on the interval probabilities. A .

9. A fault monitoring data transmission method for an excitation power cabinet according to claim 5, characterized in that, The steps for determining the target intelligent agent include: retrieving the capability profile of any intelligent agent, combining the repair capability value of the intelligent agent for various fault types, and the fault severity of various fault types in the excitation power cabinet, multiplying the fault severity of each type of fault by the corresponding repair capability value and summing the results to calculate the target value of the intelligent agent; obtaining the target value of each intelligent agent, and taking the intelligent agent with the largest target value as the target intelligent agent for overhauling the excitation power cabinet.

10. A fault monitoring data transmission system for excitation power cabinets, characterized in that, It includes a memory and a processor, as well as a computer program stored in the memory and running on the processor. The processor is coupled to the memory, and when the processor executes the computer program, it implements a fault monitoring data transmission method for an excitation power cabinet as described in any one of claims 1-9.

Images