Frequency converter fault prediction method and system based on artificial intelligence

Abstract

The invention discloses a frequency converter fault prediction method based on artificial intelligence. The method comprises the following steps: firstly, acquiring sensing data sets corresponding to a plurality of parameters of the frequency converter respectively, selecting a target parameter from the plurality of parameters, and further acquiring a plurality of target data sets corresponding to the target parameter in historical data; secondly, constructing and training an integrated learning model, calculating the correlation degree between any two data sets in each target data set of the target parameters through the model, and combining the correlation degree into a correlation array of the target data sets; then, analyzing and processing the correlation array corresponding to each target data set, calculating to obtain a plurality of representative vectors of each target data set, and constructing a spatial feature model based on the representative vectors; and finally, collecting a real-time data set of the target parameter in the current time period, calculating to obtain a real-time representative vector, and inputting the real-time representative vector into the spatial feature model to obtain a fault prediction result. The method can effectively improve the accuracy and efficiency of frequency converter fault prediction.

Description

Technical Field

[0001] The present application relates to the technical field of fault prediction, and in particular to an inverter fault prediction method and system based on artificial intelligence. Background Art

[0002] Frequency converters (VFDs) are critical equipment in fields such as industrial automation and power systems. Failures can lead to production interruptions, equipment damage, and energy waste. Existing VFD fault prediction technologies have shortcomings, such as low detection efficiency and accuracy, and difficulty handling complex fault modes. Traditional fault prediction methods often rely on manual experience or single models, which are difficult to adapt to the diversity and complexity of VFD operating conditions. They are often only able to predict known fault types and have limited generalization capabilities for unknown or new faults. A similar prior art example is Chinese patent application number CN118551266A, which proposes a fault diagnosis method for VFDs. The method includes: receiving dust concentration data from a dust detector; monitoring the VFD's operating status using a trained digital twin model to obtain operating status data; determining the dust concentration and operating status data as monitoring data; and then obtaining potential fault data, performance trend data, and remaining life data for the VFD. Based on the potential fault data, performance trend data, and remaining life data, a fault tree for the VFD is constructed. Causal reasoning is used to filter the primary cause of the potential fault from multiple potential causes, and output a fault diagnosis report for the VFD. This method is used to improve the accuracy of fault diagnosis. Similar prior art also includes Chinese patent application number CN119475162A, which proposes a fault detection method for a high-voltage inverter. This method uses an acquisition module installed inside the high-voltage inverter to complete real-time acquisition of the operating parameters of the high-voltage inverter, relating to the field of data processing technology. This method extracts data parameters of similar high-voltage inverters from a database to establish a fault diagnosis model. Historical fault data related to the high-voltage inverter is then introduced into the fault diagnosis model for training and optimization. The operating parameters of the high-voltage inverter to be fault-detected are then introduced into the fault diagnosis model for processing, and a fault diagnosis result is output. This allows for real-time monitoring of the likelihood of equipment failure during real-time data acquisition, rather than relying on human experience to determine the fault problem during the detection process. This improves both the efficiency and accuracy of fault detection. Both of these patents address the problem of inverter fault prediction, but the data complexity involved in the prediction process is high and the accuracy is insufficient. Summary of the Invention

[0003] The present application provides an artificial intelligence-based inverter fault prediction method and system, the method comprising: Based on historical sensor data sets corresponding to multiple parameters of the frequency converter, a target parameter is selected from the multiple parameters, and multiple target data sets corresponding to the target parameters are obtained; Constructing and training an integrated learning model, calculating the correlation between any two data sets based on each of the target data sets of the target parameters according to the integrated learning model, and combining the correlations into a correlation array of the target data sets; By analyzing and processing the association array corresponding to each target data set, a plurality of representative vectors corresponding to each target data set are calculated, and a spatial feature model is constructed based on the plurality of representative vectors corresponding to each target data set; A real-time data set of the target parameter in the current time period is collected, and a real-time representative vector corresponding to the real-time data set is calculated and input into the spatial feature model to obtain a fault prediction result.

[0004] As a preferred technical solution of the present invention, the acquisition of multiple target data sets corresponding to the target parameters includes: Obtain historical sensor data sets and corresponding fault information corresponding to the multiple parameters of the inverter, select the target parameter according to the criticality of each parameter to inverter fault detection, and obtain multiple target data sets corresponding to the target parameter from the historical sensor data sets according to the working mode, wherein the target parameters include multiple parameters of the inverter.

[0005] As a preferred technical solution of the present invention, the acquisition of the correlation array includes: Construct and train the integrated learning model, which includes multiple sub-models. The numerical combination of the target parameters collected at the same time in the target data set is used as the data group. Any two of the data groups are input into the integrated learning model. Each of the sub-models in the integrated model uses reasoning and analysis from different angles to input the two data groups, outputs the corresponding reasoning correlation, and calculates the correlation between the two data groups based on the numerical distribution of each reasoning correlation. The correlation between any two data groups in the target data set is combined into the correlation array.

[0006] As a preferred technical solution of the present invention, the acquisition of the representative vector of the target data set and the construction of the spatial feature model include: Obtaining a representative array based on the correlation array and the de-averaged array corresponding to the target data set, performing diagonal processing on the representative array to obtain a plurality of representative values, and taking vectors corresponding to m representative values ​​whose representative contributions are greater than a set threshold as representative vectors of the target data set, and adjusting the value of m based on inverter fault information corresponding to a data group in the target data set; Each representative vector is mapped into a multi-dimensional feature coordinate system, and the inverter fault information corresponding to the target data group is added to the position of the corresponding representative vector to obtain the spatial feature model.

[0007] As a preferred technical solution of the present invention, the real-time representative vector is input into the spatial feature model to obtain a fault prediction result, including: Each real-time representative vector corresponding to the real-time data set is input into the spatial feature model, and a target representative vector whose correlation with the real-time representative vector is within a set range is obtained through the spatial feature model. When the fault information of the target representative vector is normal, the fault information is predicted to be normal; otherwise, the fault information is predicted to be abnormal. When the predicted fault information is normal, the distance between the real-time representative vector and the representative vector with abnormal fault information and the closest distance is calculated, and according to the change of the distance, the fault information change trend is obtained. When the distance is less than the warning distance, a warning message is sent to the user terminal, and the predicted fault information is used as the fault prediction result, wherein the working mode corresponding to the target vector is the same as the working mode of the real-time representative vector.

[0008] As a preferred technical solution of the present invention, before inputting the real-time representative vector corresponding to the real-time data set into the spatial feature model, the method further includes: The historical representative vector corresponding to the historical data set within the preset time period before the current time period is input into the spatial feature model, and a representative vector that is under the same working mode as the historical data set and has a correlation greater than a set value is searched as a similar representative vector. When the fault information corresponding to the historical representative vector and the similar representative vector is inconsistent or the similar representative vector cannot be found, the historical representative vector replaces the representative vector under the working mode.

[0009] As a preferred technical solution of the present invention, the training of the ensemble learning model includes: A plurality of sample data sets corresponding to the target parameters in different operating modes of the frequency converter are selected from a database, and the data groups in each of the sample data sets are normalized. Any two of the data groups in the sample data sets are used as sample data pairs, and each sub-model in the integrated learning model is trained based on the sample data pairs and label data, wherein the analysis angles in the sub-models include at least a time angle, a frequency domain angle, and a statistical angle.

[0010] As an optimal technical solution of the present invention, the sub-model in the integrated learning model is a gradient boosting tree model.

[0011] The present invention also provides an artificial intelligence-based inverter fault prediction system for implementing the above method, the system comprising: an acquisition unit, configured to select a target parameter from a plurality of parameters of the frequency converter based on historical sensing data sets corresponding to the plurality of parameters, and acquire a plurality of target data sets corresponding to the target parameter; a computing unit, configured to construct and train an ensemble learning model, calculate the correlation between any two data sets based on each of the target data sets of the target parameters according to the ensemble learning model, and combine the correlations into a correlation array of the target data sets; a construction unit, configured to calculate a plurality of representative vectors of each target data set by analyzing and processing the correlation array corresponding to each target data set, and construct a spatial feature model based on the plurality of representative vectors corresponding to each target data set; The prediction unit is used to collect a real-time data set of the target parameter within the current time period, calculate and input a real-time representative vector corresponding to the real-time data set into the spatial feature model to obtain a fault prediction result.

[0012] The present invention also provides a computer-readable storage medium having instructions stored thereon, and the above-mentioned method is implemented when the instructions are executed by a processor.

[0013] Effect This patent proposes an AI-based inverter fault prediction method and system. This invention offers significant technical benefits. First, through multi-dimensional data processing and the application of an ensemble learning model, the system effectively improves the accuracy of inverter fault prediction. By selecting multiple key inverter parameters and combining them with historical sensor data, the system extracts a target data set. Based on this data, it calculates the correlation between each data set and constructs a high-precision correlation array. This process uses multiple sub-models within the ensemble learning model to analyze data from multiple perspectives, including time, frequency, and statistics, enabling the system to more comprehensively understand the data characteristics. Second, by extracting representative vectors and constructing spatial feature models, the system reduces interference from redundant data while retaining the most representative feature information. By diagonalizing the representative vectors and extracting the most contributing representative values, the system further reduces the data dimensionality and improves the computational efficiency of subsequent predictive analysis. Furthermore, the method utilizes a dynamic update mechanism to process real-time data. When the system detects a discrepancy between the historical representative vectors and the real-time data regarding fault information, it automatically replaces or adds representative vectors to ensure the model consistently accurately reflects the current operating status. This mechanism not only enhances the adaptability of the system, but also can update the fault prediction model in real time, effectively improving the timeliness and accuracy of the prediction results; most importantly, based on the spatial feature model, the system can quickly identify the most relevant fault mode by comparing real-time data with historical data, and give early warning of potential fault risks. By setting the warning distance, the system can monitor the changing trend of the fault and issue a warning within the warning distance to ensure the safe operation of the inverter. In short, the present invention can greatly improve the efficiency and accuracy of inverter fault prediction by combining integrated learning, spatial feature model and dynamic update mechanism, and provide reliable technical support for preventive maintenance of industrial equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 Flowchart of a frequency converter fault detection method based on artificial intelligence in an embodiment; Figure 2 Flowchart of a method for constructing a spatial feature model in an embodiment; Figure 3 This is a flow chart of a method for obtaining fault prediction results in an embodiment; Figure 42 is a structural diagram of an inverter fault detection system based on artificial intelligence in an embodiment. DETAILED DESCRIPTION

[0016] The embodiments of the present application provide a method and system for predicting inverter faults based on artificial intelligence. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device 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 units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 As shown, an inverter fault prediction method based on artificial intelligence in an embodiment of the present application includes: Step S1: acquiring historical sensor data sets corresponding to a plurality of parameters of the frequency converter, and selecting a target parameter and a plurality of target data sets corresponding to the target parameter from the plurality of parameters; Step S2: constructing and training an ensemble learning model, calculating the correlation between any two data sets based on each target data set of the target parameter according to the ensemble learning model, and combining the correlations into a correlation array of the target data set; Step S3: Analyzing and processing the correlation array corresponding to each target data set, calculating a plurality of representative vectors corresponding to each target data set, and constructing a spatial feature model based on the plurality of representative vectors corresponding to each target data set; Step S4: collecting a real-time data set of the target parameter in the current time period, calculating and inputting a real-time representative vector corresponding to the real-time data set into the spatial feature model to obtain a fault prediction result.

[0018] Furthermore, the acquisition of multiple target data sets corresponding to the target parameters includes: Obtain historical sensor data sets and corresponding fault information corresponding to the multiple parameters of the inverter, select the target parameter according to the criticality of each parameter to inverter fault detection, and obtain multiple target data sets corresponding to the target parameter from the historical sensor data sets according to the working mode, wherein the target parameters include multiple parameters of the inverter.

[0019] Specifically, by obtaining the above-mentioned historical sensor data sets corresponding to the above-mentioned multiple parameters of the above-mentioned inverter, and based on the historical maintenance records and maintenance detection information, the historical fault causes can be obtained. By analyzing the above-mentioned historical fault causes, the criticality of each of the above-mentioned multiple parameters to the above-mentioned inverter fault detection can be obtained, that is, the parameter reference degree reflecting the fault state of the above-mentioned inverter. For example: the output current, voltage, motor speed, mechanical vibration and temperature of the inverter are key parameters reflecting the abnormality of the inverter. In addition, multiple above-mentioned target data sets corresponding to the above-mentioned target parameters are obtained from the above-mentioned historical sensor data according to the working mode. The above-mentioned data group is an array composed of the numerical values ​​of the above-mentioned target parameters at the same time point. Through the above-mentioned technical solution, not only the influence of redundant parameters on fault prediction is reduced, making subsequent analysis more efficient, but also the accuracy of fault prediction can be improved.

[0020] Furthermore, the acquisition of the related array includes: Construct and train the integrated learning model, which includes multiple sub-models. The numerical combination of the target parameters collected at the same time in the target data set is used as the data group. Any two of the data groups are input into the integrated learning model. Each of the sub-models in the integrated model uses reasoning and analysis from different angles to input the two data groups, outputs the corresponding reasoning correlation, and calculates the correlation between the two data groups based on the numerical distribution of each reasoning correlation. The correlation between any two data groups in the target data set is combined into the correlation array.

[0021] Specifically, the above-mentioned target parameters are the main parameter variables related to inverter fault detection, such as: working mode, input and output voltage, input and output current, temperature, motor speed and mechanical vibration data, etc. The above-mentioned target data set is a time series data set under the same working mode, and the above-mentioned target data set is big data. For example, the above-mentioned historical data is the time series data of the above-mentioned target parameters collected every 30 seconds, and the multiple target parameter data groups collected every 7 days are used as the above-mentioned target data sets, and the above-mentioned inverter fault information corresponding to the above-mentioned data groups is stored separately according to the numerical collection time of the above-mentioned target parameters. Since the above-mentioned integrated learning model includes multiple sub-models, each sub-model is used to analyze and infer the above-mentioned two data groups from different angles and feature spaces, and the reasoning correlation corresponding to the output of each sub-model and the above-mentioned two data groups is obtained, and the above-mentioned reasoning correlation output of each sub-model is weighted, wherein the sum of the differences between each above-mentioned reasoning correlation and other reasoning outputs of other models is calculated respectively, and according to each above The sum of the differences corresponding to the above-mentioned inference correlations is weighted, and the weighted result is used as the correlation between the above-mentioned two data groups, wherein the sum of the weights of the above-mentioned inference correlations output by different models corresponding to the above-mentioned two data groups is 1, and the larger the sum of the above-mentioned differences, the smaller the weight corresponding to the above-mentioned inference correlation, indicating that the deviation between the above-mentioned inference correlation and other inference correlations output by other models is large, and therefore the accuracy is relatively low. Therefore, the weight of the above-mentioned inference correlation is relatively low. Conversely, the smaller the sum of the above-mentioned differences, the smaller the deviation between the above-mentioned inference correlation and other inference correlations output by other models, the higher the accuracy, and it can better reflect the true correlation between the two data groups. Therefore, the weight of the above-mentioned inference correlation is relatively high, and the above-mentioned correlations of any two of the above-mentioned data groups in the above-mentioned target data set are also composed into the above-mentioned correlation array. Through the above-mentioned technical solution, by quantifying the global correlation of the data groups in the above-mentioned target data set, similarity is also captured, and data features are captured, laying the foundation for obtaining the representative vector of the above-mentioned target data set.

[0022] Furthermore, the acquisition of the representative vector of the target data set and the construction of the spatial feature model are as follows: Figure 2 Shown, including: Obtaining a representative array based on the correlation array and the de-averaged array corresponding to the target data set, and performing diagonal processing on the representative array to obtain multiple representative values, and using the vectors corresponding to the minimum number of representative values ​​that satisfy the requirement that the sum of the representative contribution values ​​is greater than a set threshold as the representative vectors of the target data set, wherein m representative values ​​are the minimum number that can satisfy the set threshold, and the value of m is adjusted based on the inverter fault information corresponding to the data group in the target data set; The representative vector corresponding to each target data set is mapped to a multi-dimensional feature coordinate system, and the inverter fault information corresponding to the target data set is added to the positions of the corresponding multiple representative vectors to obtain the spatial feature model.

[0023] Specifically, by constructing the above-mentioned de-averaged array A, , where B is an n-dimensional n-column unit array, and e is an n-dimensional n-column all-1 array. The following example uses a target dataset consisting of four data groups. In actual applications, there are more than four data groups in the target dataset. For example, when the related array is When , the elements in the relevant array Indicates the correlation between the i-th data group and the j-th data group, and the above representative array is calculated by the calculation formula: , The calculation result is , where the values ​​above the diagonal represent the variance of the above data set in the feature space, and the values ​​on the off-diagonal represent the corresponding data The correlation value between groups is positive, indicating positive correlation, and negative, indicating negative correlation. By diagonalizing the above-mentioned de-meaned array and performing characteristic analysis, the representative value of the above-mentioned de-meaned array is obtained, which is also the eigenvalue. The above-mentioned representative value reflects the importance of the corresponding representative vector in the above-mentioned representative array. The representative contribution of the above-mentioned representative value is: the ratio of the absolute value of the representative value to the sum of the absolute values ​​of all representative values. When the representative values ​​are a1= 0.35, a2=0.2, a3= -0.12, and a4= -0.43, the above-mentioned representative contribution values ​​are 31.8%, 18.2%, 10.9%, and 39.1%, respectively. When the above-mentioned threshold value is set to 70%, m is taken as 2, and the eigenvector b1 corresponding to the above-mentioned a1= 0.35 and the above-mentioned a4= The characteristic vector b2 corresponding to -0.43 is used as the representative vector of the above-mentioned target data set, and also represents the characteristic data of the above-mentioned target data set. Each of the above-mentioned representative vectors is mapped to a multidimensional feature coordinate system, and the inverter fault information corresponding to the above-mentioned target data set is added to the position of the corresponding representative vector to obtain the above-mentioned spatial feature model. The above-mentioned technical solution, on the basis of selecting the target parameter according to the degree of influence of each parameter on the inverter fault through step S1 to reduce the influence of redundant parameters, further cooperates with this step to perform dimensionality reduction processing on the target data set to obtain a representative vector that can reflect the data characteristics of the target parameter in the above-mentioned target data set, and constructs a corresponding spatial feature model, which not only simplifies the subsequent calculation amount, but also improves the subsequent search efficiency.

[0024] Furthermore, the real-time representative vector is input into the spatial feature model to obtain the fault prediction result, such as Figure 3 Shown, including: Each real-time representative vector corresponding to the real-time data set is input into the spatial feature model, and a target representative vector whose correlation with the real-time representative vector is within a set range is obtained through the spatial feature model. When the fault information of the target representative vector is normal, the fault information is predicted to be normal; otherwise, the fault information is predicted to be abnormal. When the predicted fault information is normal, the distance between the real-time representative vector and the representative vector with abnormal fault information and the closest distance is calculated, and according to the change of the distance, the fault information change trend is obtained. When the distance is less than the warning distance, a warning message is sent to the user terminal, and the predicted fault information is used as the fault prediction result, wherein the working mode corresponding to the target vector is the same as the working mode of the real-time representative vector.

[0025] Specifically, by inputting each of the above-mentioned real-time representative vectors into the above-mentioned spatial feature model, the above-mentioned representative vector whose distance from the above-mentioned real-time representative vector is within the set range is calculated by the above-mentioned spatial feature model, that is, the above-mentioned target representative vector whose correlation is within the set range, and the working mode corresponding to the above-mentioned target vector is the same as the working mode of the above-mentioned real-time representative vector. Since the target parameters of the inverter may change in the future time period, and since the fault state will not appear until it accumulates to a certain extent during the working process, the future time period corresponds to the setting of the above-mentioned set range. The longer the future time period is set, the larger the setting range, and vice versa. The above-mentioned fault information corresponding to the above-mentioned target representative vector reflects the fault information corresponding to the above-mentioned real-time representative vector. Therefore, when the fault information of the above-mentioned target representative vector is normal, the fault information corresponding to the above-mentioned real-time representative vector is also normal. Therefore, the predicted fault information of the corresponding inverter is also normal. Conversely, when there is an abnormality in the fault information of the above-mentioned target representative vector, the above-mentioned fault information is normal. The predicted fault information is abnormal; since, when the predicted information is normal, it may be at the critical point between normal and abnormal states, in order to obtain abnormal information in time and in advance, the distance of the representative vector closest to the real-time representative distance and with abnormal fault information is calculated. The larger the distance, the closer the fault state corresponding to the real-time data set corresponding to the real-time representative vector is to the normal state. Conversely, the closer the fault state corresponding to the real-time data set corresponding to the real-time representative vector is to the abnormal state. As time goes by, the smaller the distance, the smaller the working state of the inverter is developing towards an abnormal state. Conversely, the working state of the inverter is moving towards or maintaining a normal working state, and when the distance is less than the warning distance, a warning message is sent to the user terminal. Through the above technical solution, not only the fault information of the inverter can be accurately obtained, but also the changing trend of the fault information can be obtained according to the change of the distance, and warning information can be issued in time, and measures can be taken in time to prevent the occurrence of inverter failure.

[0026] Furthermore, before inputting the real-time representative vector corresponding to the real-time data set into the spatial feature model, the method further includes: The historical representative vector corresponding to the historical data set within the preset time period before the current time period is input into the spatial feature model, and a representative vector that is under the same working mode as the historical data set and has a correlation greater than a set value is searched as a similar representative vector. When the fault information corresponding to the historical representative vector and the similar representative vector is inconsistent or the similar representative vector cannot be found, the historical representative vector replaces the representative vector under the working mode.

[0027] Specifically, since the correspondence between the numerical value of the target parameter and the fault information will change with the passage of time, in order to improve the prediction accuracy of the above-mentioned spatial prediction model, before the above-mentioned real-time representative vector corresponding to the above-mentioned real-time data set is input into the above-mentioned spatial feature model, the historical representative vector corresponding to the above-mentioned historical data set within the preset time period before the above-mentioned current time period is input into the above-mentioned spatial feature model, wherein the above-mentioned historical data set has the same working mode as the above-mentioned real-time data set, and the accuracy of the above-mentioned spatial feature vector is verified by searching for the above-mentioned similar representative vector, and when the fault information corresponding to the above-mentioned historical representative vector is inconsistent with the corresponding similar representative vector or the above-mentioned similar representative vector cannot be found, it indicates that the accuracy of the above-mentioned spatial feature model under the current working mode is insufficient, and therefore the above-mentioned historical representative vector is replaced by the representative vector under the working mode. Through the above-mentioned technical solution, the real-time accuracy of the above-mentioned spatial feature model can be guaranteed, thereby improving the prediction accuracy of the inverter fault information.

[0028] Furthermore, the training of the integrated learning model includes: selecting multiple sample data sets corresponding to the target parameters in different working modes of the frequency converter from the database, and normalizing the data groups in each of the sample data sets, taking any two of the data groups in the sample data sets as sample data pairs, and training each sub-model in the integrated learning model based on the sample data pairs and label data, wherein the analysis angles in the sub-models include at least time angle, frequency domain angle and statistical angle.

[0029] Specifically, by selecting multiple sample data sets under different working modes of the inverter corresponding to the above-mentioned target parameters from the database, and normalizing the data groups in the sample data sets, ensuring that data from different sensors or different types can be compared and analyzed on the same scale, a multi-submodel structure is introduced into the integrated learning model, and each sub-model analyzes the correlation of data groups from different angles, such as time angle, frequency domain angle and statistical angle, etc., to enhance the diversity of the model. Each sub-model is trained for specific data features, such as current, voltage, mechanical vibration and corresponding label data, etc. The above-mentioned label data is correlation data. The strategy of random sub-sample and random feature selection is used to improve the generalization ability of the model, thereby improving the calculation accuracy of the above-mentioned integrated learning model for the correlation between any two data groups in the sample data set.

[0030] Furthermore, the sub-model in the ensemble learning model is a gradient boosting tree model.

[0031] The present invention also provides an artificial intelligence-based inverter fault prediction system for implementing the above method, such as Figure 4 As shown, the system includes: an acquisition unit, configured to select a target parameter from a plurality of parameters of the frequency converter based on historical sensing data sets corresponding to the plurality of parameters, and acquire a plurality of target data sets corresponding to the target parameter; a computing unit, configured to construct and train an ensemble learning model, calculate the correlation between any two data sets based on the target data set of the target parameter according to the ensemble learning model, and combine the correlations into a correlation array of the target data set; a construction unit, configured to calculate a plurality of representative vectors of each target data set by analyzing and processing the correlation array corresponding to each target data set, and construct a spatial feature model based on the plurality of representative vectors corresponding to each target data set; The prediction unit is used to collect a real-time data set of the target parameter within the current time period, calculate and input a real-time representative vector corresponding to the real-time data set into the spatial feature model to obtain a fault prediction result.

[0032] The present invention also provides a computer-readable storage medium having instructions stored thereon, and the above-mentioned method is implemented when the instructions are executed by a processor.

[0033] In summary, this invention proposes an AI-based inverter fault prediction method and system with significant technical benefits. First, through multi-dimensional data processing and the application of an ensemble learning model, the system can effectively improve the accuracy of inverter fault prediction. By selecting multiple key inverter parameters and combining them with historical sensor data, the system extracts a target data set and calculates the correlation between each data set based on this data, constructing a high-precision correlation array. This process analyzes data from multiple perspectives, including time, frequency, and statistics, through multiple sub-models within the ensemble learning model, enabling the system to more comprehensively understand the data characteristics. Second, by extracting representative vectors and constructing spatial feature models, the system reduces interference from redundant data while retaining the most representative feature information. By diagonalizing the representative vectors and extracting the representative values ​​with the greatest contribution, the data dimension is further reduced, improving the computational efficiency of subsequent predictive analysis. Furthermore, the method processes real-time data through a dynamic update mechanism. When the system detects a discrepancy between the historical representative vectors and the real-time data regarding fault information, it automatically replaces or adds the representative vectors, ensuring that the model always accurately reflects the current operating status. This mechanism not only enhances the adaptability of the system, but also can update the fault prediction model in real time, effectively improving the timeliness and accuracy of the prediction results; most importantly, based on the spatial feature model, the system can quickly identify the most relevant fault mode by comparing real-time data with historical data, and give early warning of potential fault risks. By setting the warning distance, the system can monitor the changing trend of the fault and issue a warning within the warning distance to ensure the safe operation of the inverter. In short, the present invention can greatly improve the efficiency and accuracy of inverter fault prediction by combining integrated learning, spatial feature model and dynamic update mechanism, and provide reliable technical support for preventive maintenance of industrial equipment.

[0034] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0035] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0036] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A frequency converter fault prediction method based on artificial intelligence, characterized in that: The method comprises: Based on obtaining historical sensor data sets corresponding to multiple parameters of the frequency converter, a target parameter and multiple target data sets corresponding to the target parameter are selected from the multiple parameters; Constructing and training an integrated learning model, calculating the correlation between any two data sets based on each of the target data sets of the target parameters according to the integrated learning model, and combining the correlations into a correlation array of the target data sets; By analyzing and processing the association array corresponding to each target data set, a plurality of representative vectors corresponding to each target data set are calculated, and a spatial feature model is constructed based on the plurality of representative vectors corresponding to each target data set; A real-time data set of the target parameter in the current time period is collected, and a real-time representative vector corresponding to the real-time data set is calculated and input into the spatial feature model to obtain a fault prediction result.

2. The method according to claim 1, characterized in that Acquiring multiple target data sets corresponding to the target parameters includes: Obtain historical sensor data sets and corresponding fault information corresponding to the multiple parameters of the inverter, select the target parameter according to the criticality of each parameter to inverter fault detection, and obtain multiple target data sets corresponding to the target parameter from the historical sensor data sets according to the working mode, wherein the target parameters include multiple parameters of the inverter.

3. The method according to claim 1, characterized in that The acquisition of the related array includes: Construct and train the integrated learning model, which includes multiple sub-models. The numerical combination of the target parameters collected at the same time in the target data set is used as the data group. Any two of the data groups are input into the integrated learning model. Each of the sub-models in the integrated model uses reasoning and analysis from different angles to input the two data groups, outputs the corresponding reasoning correlation, and calculates the correlation between the two data groups based on the numerical distribution of each reasoning correlation. The correlation between any two data groups in the target data set is combined into the correlation array.

4. The method according to claim 1, wherein The acquisition of the representative vector of the target data set and the construction of the spatial feature model include: Obtaining a representative array based on the correlation array and the de-averaged array corresponding to the target data set, performing diagonal processing on the representative array to obtain a plurality of representative values, and taking vectors corresponding to m representative values ​​whose representative contributions are greater than a set threshold as representative vectors of the target data set, and adjusting the value of m based on inverter fault information corresponding to a data group in the target data set; Each representative vector is mapped into a multi-dimensional feature coordinate system, and the inverter fault information corresponding to the target data group is added to the position of the corresponding representative vector to obtain the spatial feature model.

5. The method according to claim 1, wherein Inputting the real-time representative vector into the spatial feature model to obtain a fault prediction result includes: Each real-time representative vector corresponding to the real-time data set is input into the spatial feature model, and a target representative vector whose correlation with the real-time representative vector is within a set range is obtained through the spatial feature model. When the fault information of the target representative vector is normal, the fault information is predicted to be normal; otherwise, the fault information is predicted to be abnormal. When the predicted fault information is normal, the distance between the real-time representative vector and the representative vector with abnormal fault information and the closest distance is calculated, and according to the change of the distance, the fault information change trend is obtained. When the distance is less than the warning distance, a warning message is sent to the user terminal, and the predicted fault information is used as the fault prediction result, wherein the working mode corresponding to the target vector is the same as the working mode of the real-time representative vector.

6. The method according to claim 1, characterized in that Before inputting the real-time representative vector corresponding to the real-time data set into the spatial feature model, the method further includes: The historical representative vector corresponding to the historical data set within the preset time period before the current time period is input into the spatial feature model, and a representative vector that is under the same working mode as the historical data set and has a correlation greater than a set value is searched as a similar representative vector. When the fault information corresponding to the historical representative vector and the similar representative vector is inconsistent or the similar representative vector cannot be found, the historical representative vector replaces the representative vector under the working mode.

7. The method according to claim 1, characterized in that The training of the integrated learning model includes: A plurality of sample data sets corresponding to the target parameters in different operating modes of the frequency converter are selected from a database, and the data groups in each of the sample data sets are normalized. Any two of the data groups in the sample data sets are used as sample data pairs, and each sub-model in the integrated learning model is trained based on the sample data pairs and label data, wherein the analysis angles in the sub-models include at least a time angle, a frequency domain angle, and a statistical angle.

8. The method according to claim 1, characterized in that The sub-model in the ensemble learning model is a gradient boosting tree model.

9. An artificial intelligence-based inverter fault prediction system, used to implement the method according to any one of claims 1 to 8, characterized in that: The system comprises: an acquisition unit, configured to acquire sensing data sets corresponding to a plurality of parameters of the frequency converter, select a target parameter from the plurality of parameters, and acquire a plurality of target data sets corresponding to the target parameters in historical data; a computing unit, configured to construct and train an ensemble learning model, calculate the correlation between any two data sets based on the target data set of the target parameter according to the ensemble learning model, and combine the correlations into a correlation array of the target data set; a construction unit, configured to calculate a plurality of representative vectors of each target data set by analyzing and processing the correlation array corresponding to each target data set, and construct a spatial feature model based on the plurality of representative vectors corresponding to each target data set; The prediction unit is used to collect a real-time data set of the target parameter within the current time period, calculate and input a real-time representative vector corresponding to the real-time data set into the spatial feature model to obtain a fault prediction result.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images