A fault detection method and system for fan inverter based on big data

Abstract

The present application discloses a fault detection method and system for a wind turbine inverter based on big data, which relates to the field of fault detection technology. The method comprises: deploying a sensor group to obtain real-time operating data of the wind turbine inverter; performing entity recognition on historical fault description data to obtain associated parameters of the fault type; assigning an initial weight to the associated parameters of each fault type; calculating the real-time deviation ratio, fault prediction accuracy error, and statistical deviation; calculating the dynamic adjustment coefficient obtained from the real-time feedback of each sensor parameter; calculating the error index of each parameter and optimizing the dynamic adjustment coefficient through a gradient descent algorithm; using the optimized dynamic adjustment coefficient, calculating a comprehensive fault index, and establishing a graded early warning mechanism. The method solves the technical problem that the existing technology is difficult to accurately and timely detect potential faults of the inverter, which leads to reduced wind turbine operating efficiency and even serious consequences, affecting normal production and operation.

Description

Technical Field

[0001] The present application relates to the field of fault detection technology, and in particular to a fault detection method and system for a fan inverter based on big data. Background Art

[0002] Wind turbines are essential equipment used in numerous fields, including industry, energy, and construction. Their stable operation plays an irreplaceable role in ensuring production processes, energy supply, and environmental control. As the core component of wind power generation systems, the importance of wind turbine inverters is self-evident. With the continuous development of wind power technology and the expansion of its application, the operational stability and reliability of wind turbine inverters in wind power plants are directly related to the power generation efficiency and safety of the entire system. Therefore, fault detection and maintenance of wind turbine inverters are particularly important. Traditional methods for detecting wind turbine inverter faults often rely on manual experience, which lacks scientificity and accuracy.

[0003] As the operating environment of wind turbines becomes more diverse and complex, existing technologies have difficulties in accurately and timely detecting potential faults of frequency converters, which can easily lead to reduced operating efficiency of wind turbines or even serious consequences, affecting normal production and operations. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for detecting faults in fan inverters based on big data, in order to solve the technical problem that existing technologies have difficulty in accurately and timely detecting potential inverter faults, which can lead to reduced fan operating efficiency and even serious consequences, affecting normal production and operations.

[0005] In view of the above technical problems, the present application provides a fault detection method and system for a fan inverter based on big data.

[0006] A first aspect of an embodiment of the present application provides a fault detection method for a wind turbine inverter based on big data, the method comprising:

[0007] Deploy a sensor group to obtain real-time operating data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The real-time operating data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switch status, current impact, communication delay, and data loss rate;

[0008] Based on historical fault description data, natural language processing technology is applied to perform entity recognition on the historical fault description data to obtain standardized fault type labels. Each fault type and statistical data are analyzed to determine the sensor parameters closely related to the fault type and obtain the associated parameters of the fault type.

[0009] Assigning an initial weight to the associated parameters of each fault type, wherein the initial weight is used to reflect the importance of the associated parameters in fault judgment;

[0010] Based on historical data and real-time operation data, obtain the real-time monitoring values ​​of sensor parameters and data values ​​under normal operating conditions, historical sampling times and collected data values, and calculate the real-time deviation ratio, fault prediction accuracy error and statistical deviation;

[0011] Calculate the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation, wherein the sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current impact, communication delay, and data loss rate;

[0012] Calculate the error index of each parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter, and optimize the value of the dynamic adjustment coefficient through the gradient descent algorithm;

[0013] The optimized dynamic adjustment coefficient is used to calculate the comprehensive fault index in real time and establish a graded early warning mechanism.

[0014] A second aspect of an embodiment of the present application provides a fault detection system for a wind turbine inverter based on big data, the system comprising:

[0015] A sensor group deployment module is used to deploy a sensor group to obtain real-time operating data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The real-time operating data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switch status, current impact, communication delay, and data loss rate.

[0016] A fault type associated parameter acquisition module is used to perform entity recognition on historical fault description data using natural language processing technology to obtain standardized fault type labels, analyze each fault type and statistical data, determine sensor parameters closely related to the fault type, and obtain associated parameters for the fault type;

[0017] An importance judgment module, wherein the importance judgment module is used to assign an initial weight to the associated parameters of each fault type, wherein the initial weight is used to reflect the importance of the associated parameters in fault judgment;

[0018] A deviation calculation module is used to obtain the real-time monitoring value of the sensor parameter and the data value under normal operating conditions, the historical sampling times and the collected data values ​​based on historical data and real-time operating data, and calculate the real-time deviation ratio, fault prediction accuracy error and statistical deviation;

[0019] A dynamic adjustment coefficient calculation module is used to calculate the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter based on the real-time deviation ratio, fault prediction accuracy error and statistical deviation, wherein the sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current impact, communication delay and data loss rate;

[0020] An error index calculation module is used to calculate the error index of each parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter, and optimize the value of the dynamic adjustment coefficient through a gradient descent algorithm;

[0021] A hierarchical early warning mechanism establishment module is used to use the optimized dynamic adjustment coefficient to calculate the comprehensive fault index in real time and establish a hierarchical early warning mechanism.

[0022] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0023] A sensor group is deployed to obtain real-time operation data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The real-time operation data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switch status, current impact, communication delay, and data loss rate. Based on historical fault description data, natural language processing technology is applied to perform entity recognition on the historical fault description data to obtain standardized fault type labels. Each fault type and statistical data are analyzed to determine the sensor parameters closely related to the fault type and obtain the associated parameters of the fault type. An initial weight is assigned to the associated parameters of each fault type. The initial weight is used to reflect the importance of the associated parameters in fault judgment. Based on historical data and real-time operating data, the real-time monitoring values ​​of sensor parameters and data values ​​under normal operating conditions, historical sampling times, and collected data values ​​are obtained to calculate the real-time deviation ratio, fault prediction accuracy error, and statistical deviation. Based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation, the dynamic adjustment coefficient obtained from the real-time feedback of each sensor parameter is calculated. The sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current surge, communication delay, and data loss rate. Based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter, the error index of each parameter is calculated, and the value of the dynamic adjustment coefficient is optimized using a gradient descent algorithm. The optimized dynamic adjustment coefficient is used to calculate the comprehensive fault index in real time and establish a graded early warning mechanism. This solves the technical problem that existing technologies have difficulty in accurately and timely detecting potential inverter faults, which leads to reduced wind turbine operating efficiency and even serious consequences, affecting normal production and operation.

[0024] The above description is only an overview of the technical solution of the present application. In order to more clearly illustrate the technical means of the present application and to implement it in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the accompanying drawings of the embodiments of the present disclosure are briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in precise order. Instead, various steps may be processed in reverse order or simultaneously as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0026] Figure 1A flowchart of a method for fault detection of a fan inverter based on big data provided in an embodiment of the present application;

[0027] Figure 2 A structural schematic diagram of a fault detection system for a fan inverter based on big data provided in an embodiment of the present application.

[0028] Explanation of the accompanying symbols: sensor group layout module 10, fault type associated parameter acquisition module 20, importance judgment module 30, deviation calculation module 40, dynamic adjustment coefficient calculation module 50, error index calculation module 60, hierarchical warning mechanism establishment module 70. DETAILED DESCRIPTION

[0029] This application solves the technical problem that the existing technology is difficult to accurately and timely detect potential faults of the inverter, resulting in reduced fan operating efficiency or even serious consequences, affecting normal production and operation, by providing a fault detection method and system for the fan inverter based on big data.

[0030] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0032] Example 1, as Figure 1 As shown, the present application provides a fault detection method for a fan inverter based on big data, wherein the method includes:

[0033] Deploy a sensor group to obtain real-time operating data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The real-time operating data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switch status, current impact, communication delay, and data loss rate;

[0034] Specifically, a sensor group is deployed at key locations on the wind turbine inverter to capture real-time operating data. The sensor group includes current sensors, voltage sensors, temperature sensors, vibration sensors, an IGBT status monitoring unit, and a communication status detection unit. Real-time operating data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching status, current surge, communication delay, and data loss rate.

[0035] Furthermore, the sensor group is arranged to obtain real-time operating data of the wind turbine inverter, and the sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit, including:

[0036] The current sensor is used to monitor input current parameters and output current parameters;

[0037] The voltage sensor is used to monitor input voltage parameters and output voltage parameters;

[0038] The temperature sensor is used to monitor the temperature parameters of the frequency converter;

[0039] The vibration sensor is used to monitor mechanical vibration parameters;

[0040] The IGBT status monitoring unit is used to monitor IGBT switch status parameters and current impact parameters;

[0041] The communication status detection unit is used to monitor communication delay parameters and data loss rate.

[0042] Specifically, a sensor group is deployed to obtain real-time operating data from the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The current sensor monitors input and output current parameters; the voltage sensor monitors input and output voltage parameters; the temperature sensor monitors inverter temperature parameters to prevent overheating; the vibration sensor monitors mechanical vibration parameters to prevent mechanical failure; the IGBT status monitoring unit monitors IGBT switch state parameters and current surge parameters to prevent switch failure; and the communication status detection unit monitors communication delay parameters and data loss rate to ensure the reliability and timeliness of data transmission.

[0043] Based on historical fault description data, natural language processing technology is applied to perform entity recognition on the historical fault description data to obtain standardized fault type labels. Each fault type and statistical data are analyzed to determine the sensor parameters closely related to the fault type and obtain the associated parameters of the fault type.

[0044] Furthermore, based on the historical fault description data, natural language processing technology is applied to perform entity recognition on the historical fault description data to obtain standardized fault type labels, and each fault type and statistical data are analyzed to determine the sensor parameters closely related to the fault type, and obtain associated parameters of the fault type, including:

[0045] Collect a large amount of historical fault description data from multiple data sources such as fault reports, maintenance logs, and work order records. The fault description data includes the name of the faulty device, fault phenomenon, fault cause, and repair process;

[0046] Use natural language processing technology to preprocess historical fault description data and perform entity recognition;

[0047] Perform semantic clustering on the extracted entities;

[0048] Based on the clustering results, faults with similar descriptions are grouped into the same category to generate standardized fault type labels;

[0049] Analyze each fault type and statistical data to evaluate the correlation between each sensor parameter and the occurrence of the fault;

[0050] Determine a set of closely related sensor parameters for each fault type and form a mapping relationship between the fault type and the associated parameters;

[0051] The mapping relationship obtained by analysis is stored in a data table, wherein the data table includes the fault type and the sensor parameters corresponding to the fault type, and the associated parameters of the fault type are obtained.

[0052] Specifically, a large amount of historical fault description data is collected from multiple data sources (such as fault reports, maintenance logs, work order records, etc.). This data should include key information such as the name of the faulty device, fault symptoms, fault causes, and repair process;

[0053] Preprocess the collected historical fault description data, including noise removal, word segmentation, and part-of-speech tagging. Use named entity recognition (NER) in natural language processing to extract key entities (such as device names and fault symptoms) from the historical fault description data.

[0054] Use text similarity calculation methods (such as cosine similarity, Jaccard similarity, etc.) to calculate the similarity between entities, and apply clustering algorithms (such as K-means, hierarchical clustering, etc.) to cluster entities;

[0055] Based on the clustering results, faults with similar descriptions are grouped into the same category to form standardized fault type labels;

[0056] Collect sensor parameter data related to the faulty equipment and use statistical methods (such as correlation analysis and chi-square test) to evaluate the correlation between sensor parameters and fault types;

[0057] Based on the analysis results, the sensor parameters closely related to each fault type are screened out to form a mapping relationship between the fault type and the associated parameters;

[0058] Design a data table structure, including the fault type and the sensor parameter field corresponding to the fault type, store the analyzed mapping relationship data in the form of a data table, and obtain the associated parameters of the fault type.

[0059] Assigning an initial weight to the associated parameters of each fault type, wherein the initial weight is used to reflect the importance of the associated parameters in fault judgment;

[0060] Furthermore, the associated parameters of each fault type are assigned an initial weight, and the initial weight is used to reflect the importance of the associated parameters in fault judgment, including:

[0061] Statistical methods are used to identify abnormal values ​​of each parameter, and the abnormal values ​​of each sensor parameter are matched with standardized fault type labels to form a data pair set;

[0062] Using the Pearson correlation coefficient, the correlation coefficient between each sensor parameter outlier sequence and the standardized fault type label sequence is calculated;

[0063] Combining expert experience and historical data, an initial weight is assigned to the associated parameters of each fault type;

[0064] The correlation coefficients of all sensor parameters are compared and sorted, and initial weights are determined according to the magnitudes of the correlation coefficients. The initial weights are used to reflect the importance of the associated parameters in fault judgment.

[0065] Specifically, statistical methods (such as the 3σ principle and boxplots) are used to identify abnormal values ​​of each sensor parameter. These abnormal values ​​may represent abnormal conditions of the equipment or precursors to failure. The identified abnormal values ​​of the sensor parameters are matched with standardized fault type labels to form a data pair set. Each data pair contains the abnormal value and its corresponding fault type label.

[0066] Calculate the Pearson correlation coefficient between each sensor parameter outlier sequence and the standardized fault type label sequence. The Pearson correlation coefficient measures the degree of linear correlation between two variables, and its value ranges from -1 to 1. A positive correlation coefficient indicates a positive correlation between the two variables, a negative correlation coefficient indicates a negative correlation, and a correlation coefficient close to 0 indicates no significant linear relationship between the two variables.

[0067] Domain experts assess the importance of the parameters associated with each fault type based on their experience and knowledge. They use historical data on fault cases and parameter changes to analyze the actual role of each parameter in fault diagnosis. Combining expert assessments with historical data analysis results, they assign an initial weight to the parameters associated with each fault type.

[0068] Compare and rank the correlation coefficients of all sensor parameters. Determine the initial weight of each parameter based on the correlation coefficient. The weight accurately reflects the importance of each parameter in fault diagnosis. Parameters with larger correlation coefficients are assigned higher weights because they have a stronger linear relationship with the fault type label and are therefore more important in fault diagnosis.

[0069] Based on historical data and real-time operation data, obtain the real-time monitoring values ​​of sensor parameters and data values ​​under normal operating conditions, historical sampling times and collected data values, and calculate the real-time deviation ratio, fault prediction accuracy error and statistical deviation;

[0070] Specifically, the monitoring values ​​of the current sensor parameters and the data values ​​under normal operating conditions are extracted from the real-time operation data, and the real-time deviation ratio is calculated using the real-time deviation ratio calculation formula; the occurrence of faults is predicted based on historical data, and the mean square error formula is used to calculate the fault prediction accuracy error to evaluate the accuracy of the prediction; the historical sampling times and historically collected data values ​​are obtained from the historical data, and the mean of each parameter in the historical data is calculated using the mean calculation formula; then the statistical deviation is calculated based on the calculated mean using the statistical deviation calculation formula.

[0071] Furthermore, based on the historical data and real-time operation data, obtaining the real-time monitoring value of the sensor parameter and the data value under normal operating conditions, the historical sampling times and the collected data values, and calculating the real-time deviation ratio, the fault prediction accuracy error and the statistical deviation include:

[0072] Based on historical data and real-time operation data, obtain the real-time monitoring values ​​of sensor parameters and the data values ​​under normal operating conditions, and calculate the real-time deviation ratio;

[0073] The real-time deviation ratio calculation formula is:

[0074]

[0075] Among them, d i is the real-time deviation ratio, i is the i-th sensor parameter, P i is the real-time monitoring value of the i-th parameter, P i,norm is the data value of parameter i under normal operating conditions;

[0076] Use the mean square error formula to calculate the fault prediction accuracy error e i ;

[0077] Based on historical data, the mean of each parameter is calculated to obtain the statistical deviation s i ;

[0078] The mean calculation formula is:

[0079]

[0080] Among them, μ i is the parameter P i The mean value in the historical data, N is the number of sampling times of the historical data, P i (j) is the parameter P i The data value collected at the jth time;

[0081] The statistical deviation calculation formula is:

[0082]

[0083] Among them, s i is the parameter P i Statistical deviation, N is the number of sampling times of historical data, μ i is the parameter P i The mean value in historical data, (P i (j) -μ i ) is the degree of deviation of a single measurement value from the mean.

[0084] Calculate the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation, wherein the sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current impact, communication delay, and data loss rate;

[0085] Specifically, based on the three previously calculated metrics of real-time deviation ratio, fault prediction accuracy error, and statistical deviation, the dynamic adjustment coefficient is calculated based on real-time feedback of each sensor parameter. Sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current surge, communication delay, and data loss rate.

[0086] Furthermore, the calculation of the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter according to the real-time deviation ratio, fault prediction accuracy error, and statistical deviation includes:

[0087] Adaptive adjustment is performed based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation, and the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter is calculated. The calculation formula is:

[0088] β i (t+1)=β i (t)+λ1×d i +λ2×e i +λ3×s i ;

[0089] Among them, β i (t+1) is the dynamic adjustment coefficient obtained by real-time feedback at time t+1, λ1, λ2, and λ3 are pre-set adjustment factors, and d i is the real-time deviation ratio, e i is the fault prediction accuracy error, s i is the statistical deviation, β i The range is 0.5≤β i ≤1.5.

[0090] Specifically, when calculating the dynamic adjustment coefficient derived from real-time feedback on each sensor parameter, we comprehensively consider three key factors: real-time deviation ratio, fault prediction accuracy error, and statistical bias. This adaptive adjustment process ensures that the system can dynamically respond to changes in sensor parameters, thereby optimizing monitoring efficiency, improving fault warning accuracy, and enhancing overall system reliability.

[0091] The dynamic adjustment coefficient calculation formula obtained by real-time feedback of sensor parameters is β i (t+1)=β i (t)+λ1×d i +λ2×e i +λ3×s i ;

[0092] Among them, β i (t+1) is the dynamic adjustment coefficient obtained by real-time feedback at time t+1, λ1, λ2, and λ3 are pre-set adjustment factors, and d i is the real-time deviation ratio, e i is the fault prediction accuracy error, s i is the statistical deviation, β i The range is 0.5≤β i ≤1.5.

[0093] Calculate the error index of each parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter, and optimize the value of the dynamic adjustment coefficient through the gradient descent algorithm;

[0094] Furthermore, the error index of each parameter is calculated based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter, and the value of the dynamic adjustment coefficient is optimized by the gradient descent algorithm, including:

[0095] Obtain the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter;

[0096] Calculating an initial dynamic adjustment coefficient for each sensor parameter based on the real-time deviation ratio, the fault prediction accuracy error, and the statistical deviation;

[0097] The error index of each sensor parameter is calculated, and the value of the dynamic adjustment coefficient is optimized by the gradient descent algorithm.

[0098] Specifically, based on the calculation formulas for the real-time deviation ratio, fault prediction accuracy error, and statistical deviation, the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter (including input current parameters, output current parameters, input voltage parameters, output voltage parameters, inverter temperature parameters, mechanical vibration parameters, IGBT switch state parameters, current impact parameters, communication delay parameters, and data loss rate) are obtained;

[0099] Combining the real-time deviation ratio, fault prediction accuracy error, and statistical deviation, an initial dynamic adjustment coefficient is calculated for each sensor parameter. This coefficient reflects the system's emphasis or adjustment intensity on each sensor parameter in the current state.

[0100] Then, the error index for each sensor parameter is calculated and the value of the dynamic adjustment coefficient is optimized using the gradient descent algorithm. The gradient descent algorithm is an iterative optimization algorithm that updates the value of the coefficient based on the gradient of the objective function (in this case, the error index) with respect to the dynamic adjustment coefficient.

[0101] Furthermore, the calculation of the error index of each sensor parameter and the optimization of the value of the dynamic adjustment coefficient by a gradient descent algorithm include:

[0102] Calculate the error index of each sensor parameter using the following formula:

[0103] E i =|P i real -P i norml |;

[0104] Among them, E i is the error index of sensor parameter i, P i real is the current measured value of sensor parameter i, P i normlis the data value of sensor parameter i under normal operating conditions;

[0105] The gradient descent algorithm is used to optimize the dynamic adjustment coefficient. The loss function is calculated as follows:

[0106] W i (t+1) =W i (t) -η×W i ×E i 2 ;

[0107] Among them, W i (t+1) is the dynamic adjustment coefficient after the t+1th round of iteration, W i (t) is the dynamic adjustment coefficient after the tth iteration, and η is the learning rate.

[0108] Specifically, first, the error index of each sensor parameter is calculated according to the formula: E i =|P i real -P i norml |. Among them, E i is the error index of sensor parameter i, P i real is the current measured value of sensor parameter i, P i norml is the data value of sensor parameter i under normal operating conditions;

[0109] Then, the gradient descent algorithm is used to optimize the dynamic adjustment coefficient, and the loss function is calculated as W i (t+1) =W i (t) -η×W i ×E i 2 Among them, W i (t+1) is the dynamic adjustment coefficient after the t+1th round of iteration, W i (t) is the dynamic adjustment coefficient after the tth iteration, and η is the learning rate. For each sensor parameter, the gradient of the error index with respect to its dynamic adjustment coefficient is calculated according to this step. The gradient descent algorithm is then used to update the value of the dynamic adjustment coefficient to achieve the optimization goal.

[0110] The optimized dynamic adjustment coefficient is used to calculate the comprehensive fault index in real time and establish a graded early warning mechanism.

[0111] Specifically, based on the dynamic adjustment coefficients previously optimized using the gradient descent algorithm, these coefficients reflect the importance and adjustment strength of different sensor parameters in the current system state. A comprehensive fault index is calculated in real time using a formula, and a graded early warning mechanism is established. We categorize fault warnings into four levels, each corresponding to a different level of urgency and response measures. This improves fault handling efficiency and minimizes the impact of faults on production operations. This graded early warning mechanism enables precise monitoring of equipment operating status, ensuring stable system operation and continuous optimization.

[0112] Furthermore, the optimized dynamic adjustment coefficient is used to calculate the comprehensive fault index in real time and establish a hierarchical early warning mechanism, including:

[0113] According to the fault type, associated parameters, initial weight and dynamic adjustment coefficient, a real-time fault index calculation formula is constructed to calculate the comprehensive fault index;

[0114] Calculate the real-time fault index, the real-time fault index calculation formula is:

[0115] α i * =α i ×β i ;

[0116] Among them, α i * is the real-time fault index, α i is the initial weight of the i-th parameter, β i Dynamic adjustment coefficients obtained for real-time feedback;

[0117] Calculate the comprehensive fault index, the comprehensive fault index calculation formula is:

[0118]

[0119] Among them, F is the comprehensive fault index, α i * is the real-time fault index, P i is the real-time monitoring value of the i-th parameter, P i,norm is the data value of parameter i under normal operating conditions;

[0120] Adopt the optimized dynamic adjustment coefficient W i Calculate the optimized comprehensive fault index using the following formula:

[0121]

[0122] Among them, F o is the optimized comprehensive fault index, n is the total number of sensors, Wi is the optimized dynamic adjustment coefficient, P i is the real-time monitoring value of the i-th parameter, P i,norm is the data value of parameter i under normal operating conditions;

[0123] Based on the optimized comprehensive fault index F o , establish a hierarchical early warning mechanism and divide fault warnings into four levels;

[0124] When F o <F low When , it is judged as normal level;

[0125] When F low ≤F o <F medium When it is detected, it is judged as a minor abnormality and enters the observation state;

[0126] When F medium ≤F o <F high When it is judged as a serious abnormality level, the intelligent management system recommends inspection and maintenance;

[0127] When F o ≥F high When it is judged as an emergency fault level, the intelligent management system starts to shut down and send an alarm.

[0128] Specifically, based on the previously obtained fault type, associated parameters, initial weights, and dynamic adjustment coefficients, a real-time fault index calculation formula is constructed to calculate the real-time comprehensive fault index. The real-time fault index calculation formula is:

[0129] α i * =α i ×β i ;

[0130] Among them, α i * is the real-time fault index, α i is the initial weight of the i-th parameter, β i Dynamic adjustment coefficients obtained for real-time feedback;

[0131] Then, based on the real-time fault index calculated previously, the comprehensive fault index is calculated. The calculation formula for the comprehensive fault index is:

[0132]

[0133] Among them, F is the comprehensive fault index, α i * is the real-time fault index, P i is the real-time monitoring value of the i-th parameter, Pi,norm is the data value of parameter i under normal operating conditions;

[0134] The optimized dynamic adjustment coefficient is used to calculate the optimized comprehensive fault index. The calculation formula is:

[0135]

[0136] Among them, F o is the optimized comprehensive fault index, n is the total number of sensors, W i is the optimized dynamic adjustment coefficient, P i is the real-time monitoring value of the i-th parameter, P i,norm is the data value of parameter i under normal operating conditions;

[0137] Based on the optimized comprehensive fault index, a graded early warning mechanism is established. This mechanism aims to achieve timely identification and response to potential problems by carefully dividing the severity of the fault warning. Specifically, the levels are divided into normal level, slight abnormality level, severe abnormality level and emergency fault level. Among them, the normal level means that the optimized comprehensive fault index is less than the slight abnormality threshold, the system runs smoothly, and no measures need to be taken; the slight abnormality level means that the optimized comprehensive fault index has increased slightly, but is still within an acceptable range, and it enters the observation state at this time; the severe abnormality level means that the optimized comprehensive fault index reaches or exceeds the severe abnormality threshold. At this time, the intelligent management system should issue a prompt and recommend that the operator immediately conduct inspection and maintenance, determine the potential source of the fault, and take necessary corrective measures; the emergency fault level means that the optimized comprehensive fault index rises sharply and reaches or exceeds the emergency fault threshold. At this time, the system should issue an emergency warning signal, the intelligent management system starts and stops running, and sends an alarm to prevent the fault from further expanding and causing serious consequences. The specific judgments are as follows:

[0138] When F o <F low When , it is judged as normal level;

[0139] When F low ≤F o <F medium When it is detected, it is judged as a minor abnormality and enters the observation state;

[0140] When F medium ≤F o <F high When it is judged as a serious abnormality level, the intelligent management system recommends inspection and maintenance;

[0141] When F o ≥F high When it is judged as an emergency fault level, the intelligent management system starts to shut down and send an alarm.

[0142] In summary, the embodiments of the present application have at least the following technical effects:

[0143] A sensor group is deployed to obtain real-time operation data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The real-time operation data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switch status, current impact, communication delay, and data loss rate. Based on historical fault description data, natural language processing technology is applied to perform entity recognition on the historical fault description data to obtain standardized fault type labels. Each fault type and statistical data are analyzed to determine the sensor parameters closely related to the fault type and obtain the associated parameters of the fault type. An initial weight is assigned to the associated parameters of each fault type. The initial weight is used to reflect the importance of the associated parameters in fault judgment. Based on historical data and real-time operation data, the real-time monitoring values ​​of sensor parameters and the data values ​​under normal operating conditions, the historical sampling times and the collected data values ​​are obtained, and the real-time deviation ratio, fault prediction accuracy error and statistical deviation are calculated; based on the real-time deviation ratio, fault prediction accuracy error and statistical deviation, the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter is calculated, and the sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current impact, communication delay, and data loss rate parameters; based on the real-time deviation ratio, fault prediction accuracy error and statistical deviation of each parameter, the error index of each parameter is calculated, and the value of the dynamic adjustment coefficient is optimized by the gradient descent algorithm; using the optimized dynamic adjustment coefficient, the comprehensive fault index is calculated in real time, and a hierarchical early warning mechanism is established.

[0144] Embodiment 2 is based on the same inventive concept as the method for fault detection of a fan inverter based on big data in the above embodiment. Figure 2 As shown, the present application provides a fault detection system for a wind turbine inverter based on big data. The system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0145] A sensor group layout module 10 is used to layout a sensor group to obtain real-time operating data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The real-time operating data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switch status, current impact, communication delay, and data loss rate.

[0146] A fault type associated parameter acquisition module 20 is configured to perform entity recognition on historical fault description data using natural language processing technology to obtain standardized fault type labels, analyze each fault type and statistical data, determine sensor parameters closely related to the fault type, and obtain associated parameters for the fault type;

[0147] An importance judgment module 30 is used to assign an initial weight to the associated parameters of each fault type, wherein the initial weight is used to reflect the importance of the associated parameters in fault judgment;

[0148] Deviation calculation module 40, which is used to obtain the real-time monitoring value of the sensor parameter and the data value under normal operating conditions, the historical sampling times and the collected data values ​​based on historical data and real-time operating data, and calculate the real-time deviation ratio, fault prediction accuracy error and statistical deviation;

[0149] A dynamic adjustment coefficient calculation module 50 is used to calculate a dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation. The sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current impact, communication delay, and data loss rate.

[0150] An error index calculation module 60 is used to calculate the error index of each parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter, and optimize the value of the dynamic adjustment coefficient through a gradient descent algorithm;

[0151] The hierarchical warning mechanism establishment module 70 is used to use the optimized dynamic adjustment coefficient to calculate the comprehensive fault index in real time and establish a hierarchical warning mechanism.

[0152] It should be noted that the above-mentioned order of the embodiments of the present application is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. The above description is of specific embodiments of this specification. 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 an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0153] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0154] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A fault detection method for a fan inverter based on big data, characterized in that: The method comprises: Deploy a sensor group to obtain real-time operating data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The real-time operating data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switch status, current impact, communication delay, and data loss rate; Based on historical fault description data, natural language processing technology is applied to perform entity recognition on the historical fault description data to obtain standardized fault type labels. Each fault type and statistical data are analyzed to determine the sensor parameters closely related to the fault type and obtain the associated parameters of the fault type. Assigning an initial weight to the associated parameters of each fault type, wherein the initial weight is used to reflect the importance of the associated parameters in fault judgment; Based on historical data and real-time operation data, obtain the real-time monitoring values ​​of sensor parameters and data values ​​under normal operating conditions, historical sampling times and collected data values, and calculate the real-time deviation ratio, fault prediction accuracy error and statistical deviation; Calculate the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation, wherein the sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current impact, communication delay, and data loss rate; Calculate the error index of each parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter, and optimize the value of the dynamic adjustment coefficient through the gradient descent algorithm; The optimized dynamic adjustment coefficient is used to calculate the comprehensive fault index in real time and establish a graded early warning mechanism.

2. A fault detection method for a fan inverter based on big data according to claim 1, characterized in that: The sensor group is arranged to obtain real-time operating data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit, including: The current sensor is used to monitor input current parameters and output current parameters; The voltage sensor is used to monitor input voltage parameters and output voltage parameters; The temperature sensor is used to monitor the temperature parameters of the frequency converter; The vibration sensor is used to monitor mechanical vibration parameters; The IGBT status monitoring unit is used to monitor IGBT switch status parameters and current impact parameters; The communication status detection unit is used to monitor communication delay parameters and data loss rate.

3. The method for fault detection of a fan inverter based on big data according to claim 1, characterized in that: Based on the historical fault description data, natural language processing technology is applied to perform entity recognition on the historical fault description data to obtain standardized fault type labels, and each fault type and statistical data are analyzed to determine the sensor parameters closely related to the fault type, and obtain the associated parameters of the fault type, including: Collect a large amount of historical fault description data from multiple data sources such as fault reports, maintenance logs, and work order records. The fault description data includes the name of the faulty device, fault phenomenon, fault cause, and repair process; Use natural language processing technology to preprocess historical fault description data and perform entity recognition; Perform semantic clustering on the extracted entities; Based on the clustering results, faults with similar descriptions are grouped into the same category to generate standardized fault type labels; Analyze each fault type and statistical data to evaluate the correlation between each sensor parameter and the occurrence of the fault; Determine a set of closely related sensor parameters for each fault type and form a mapping relationship between the fault type and the associated parameters; The mapping relationship obtained by analysis is stored in a data table, wherein the data table includes the fault type and the sensor parameters corresponding to the fault type, and the associated parameters of the fault type are obtained.

4. The method for fault detection of a fan inverter based on big data according to claim 1, characterized in that: The initial weight is assigned to the associated parameters of each fault type, and the initial weight is used to reflect the importance of the associated parameters in fault judgment, including: Statistical methods are used to identify abnormal values ​​of each parameter, and the abnormal values ​​of each sensor parameter are matched with standardized fault type labels to form a data pair set; Using the Pearson correlation coefficient, the correlation coefficient between each sensor parameter outlier sequence and the standardized fault type label sequence is calculated; Combining expert experience and historical data, an initial weight is assigned to the associated parameters of each fault type; The correlation coefficients of all sensor parameters are compared and sorted, and initial weights are determined according to the magnitudes of the correlation coefficients. The initial weights are used to reflect the importance of the associated parameters in fault judgment.

5. The method for fault detection of a fan inverter based on big data according to claim 1, characterized in that: The method of obtaining the real-time monitoring value of the sensor parameter and the data value under normal operating conditions, the historical sampling times and the collected data values ​​based on the historical data and the real-time operating data, and calculating the real-time deviation ratio, the fault prediction accuracy error and the statistical deviation includes: Based on historical data and real-time operation data, obtain the real-time monitoring values ​​of sensor parameters and the data values ​​under normal operating conditions, and calculate the real-time deviation ratio; The real-time deviation ratio calculation formula is: Among them, d i is the real-time deviation ratio, i is the i-th sensor parameter, P i is the real-time monitoring value of the i-th parameter, P i,norm is the data value of parameter i under normal operating conditions; Use the mean square error formula to calculate the fault prediction accuracy error e i ; Based on historical data, the mean of each parameter is calculated to obtain the statistical deviation s i ; The mean calculation formula is: Among them, μ i is the parameter P i The mean value in the historical data, N is the number of sampling times of the historical data, P i (j) is the parameter P i The data value collected at the jth time; The statistical deviation calculation formula is: Among them, s i is the parameter P i Statistical deviation, N is the number of sampling times of historical data, μ i is the parameter P i The mean value in historical data, (P i (j) -μ i ) is the degree of deviation of a single measurement value from the mean.

6. The method for fault detection of a fan inverter based on big data according to claim 1, characterized in that: The step of calculating the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter based on the real-time deviation ratio, the fault prediction accuracy error, and the statistical deviation includes: Adaptive adjustment is performed based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation, and the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter is calculated. The calculation formula is: b i (t+1)=β i (t)+λ1×d i +λ2×e i +λ3×s i ; Among them, β i (t+1) is the dynamic adjustment coefficient obtained by real-time feedback at time t+1, λ1, λ2, and λ3 are pre-set adjustment factors, and d i is the real-time deviation ratio, e i is the fault prediction accuracy error, s i is the statistical deviation, β i The range is 0.5≤β i ≤1.

5.

7. The method for fault detection of a fan inverter based on big data according to claim 1, characterized in that: The method of calculating the error index of each parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter and optimizing the value of the dynamic adjustment coefficient by the gradient descent algorithm includes: Obtain the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter; Calculating an initial dynamic adjustment coefficient for each sensor parameter based on the real-time deviation ratio, the fault prediction accuracy error, and the statistical deviation; The error index of each sensor parameter is calculated, and the value of the dynamic adjustment coefficient is optimized by the gradient descent algorithm.

8. The method for fault detection of a fan inverter based on big data according to claim 7, characterized in that: The calculation of the error index of each sensor parameter and the optimization of the value of the dynamic adjustment coefficient by the gradient descent algorithm include: Calculate the error index of each sensor parameter using the following formula: Among them, E i is the error index of sensor parameter i, P i real is the current measured value of sensor parameter i, P i norml is the data value of sensor parameter i under normal operating conditions; The gradient descent algorithm is used to optimize the dynamic adjustment coefficient. The loss function is calculated as follows: IN i (t+1) =In i (t) -η×W i ×E i 2 ; Among them, W i (t+1) is the dynamic adjustment coefficient after the t+1th round of iteration, W i (t) is the dynamic adjustment coefficient after the tth iteration, and η is the learning rate.

9. The method for fault detection of a fan inverter based on big data according to claim 1, characterized in that: The optimized dynamic adjustment coefficient is used to calculate the comprehensive fault index in real time and establish a hierarchical early warning mechanism, including: According to the fault type, associated parameters, initial weight and dynamic adjustment coefficient, a real-time fault index calculation formula is constructed to calculate the comprehensive fault index; Calculate the real-time fault index, the real-time fault index calculation formula is: α i * ＝α i ×β i ； Among them, α i * is the real-time fault index, α i is the initial weight of the i-th parameter, β i Dynamic adjustment coefficients obtained for real-time feedback; Calculate the comprehensive fault index, the comprehensive fault index calculation formula is: Among them, F is the comprehensive fault index, α i * is the real-time fault index, P i is the real-time monitoring value of the i-th parameter, P i,norm is the data value of parameter i under normal operating conditions; Adopt the optimized dynamic adjustment coefficient W i Calculate the optimized comprehensive fault index using the following formula: Among them, F o is the optimized comprehensive fault index, n is the total number of sensors, W i is the optimized dynamic adjustment coefficient, P i is the real-time monitoring value of the i-th parameter, P i,norm is the data value of parameter i under normal operating conditions; Based on the optimized comprehensive fault index F o , establish a hierarchical early warning mechanism and divide fault warnings into four levels; When F o <F low When , it is judged as normal level; When F low ≤F o <F medium When it is detected, it is judged as a minor abnormality and enters the observation state; When F medium ≤F o <F high When it is judged as a serious abnormality level, the intelligent management system recommends inspection and maintenance; When F o ≥F high When it is judged as an emergency fault level, the intelligent management system starts to shut down and send an alarm.

10. A fault detection system for a fan inverter based on big data, characterized in that: The system for implementing the method for fault detection of a wind turbine inverter based on big data according to any one of claims 1 to 9 comprises: A sensor group deployment module is used to deploy a sensor group to obtain real-time operating data of the wind turbine inverter. The sensor group includes a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, an IGBT status monitoring unit, and a communication status detection unit. The real-time operating data includes input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switch status, current impact, communication delay, and data loss rate. A fault type associated parameter acquisition module is used to perform entity recognition on historical fault description data using natural language processing technology to obtain standardized fault type labels, analyze each fault type and statistical data, determine sensor parameters closely related to the fault type, and obtain associated parameters for the fault type; An importance judgment module, wherein the importance judgment module is used to assign an initial weight to the associated parameters of each fault type, wherein the initial weight is used to reflect the importance of the associated parameters in fault judgment; A deviation calculation module is used to obtain the real-time monitoring value of the sensor parameter and the data value under normal operating conditions, the historical sampling times and the collected data values ​​based on historical data and real-time operating data, and calculate the real-time deviation ratio, fault prediction accuracy error and statistical deviation; A dynamic adjustment coefficient calculation module is used to calculate the dynamic adjustment coefficient obtained by real-time feedback of each sensor parameter based on the real-time deviation ratio, fault prediction accuracy error and statistical deviation, wherein the sensor parameters include input current, output current, input voltage, output voltage, inverter temperature, mechanical vibration, IGBT switching state, current impact, communication delay and data loss rate; An error index calculation module is used to calculate the error index of each parameter based on the real-time deviation ratio, fault prediction accuracy error, and statistical deviation of each parameter, and optimize the value of the dynamic adjustment coefficient through a gradient descent algorithm; A hierarchical early warning mechanism establishment module is used to use the optimized dynamic adjustment coefficient to calculate the comprehensive fault index in real time and establish a hierarchical early warning mechanism.

Images