Inverter aging detection method and device based on multi-source data analysis

Abstract

The invention relates to the technical field of aging detection, and discloses an inverter aging detection method and device based on multi-source data analysis, and the method comprises the steps: carrying out the multi-dimensional data set and synchronizing a timestamp, and obtaining a fusion data set; segmenting the fusion data set to calculate fluctuation amplitude, and when the fluctuation amplitude exceeds a threshold value, marking as an abnormal time period to generate a multi-dimensional feature vector; calculating a feature slope distribution weight, obtaining a weighted feature vector, calculating the similarity between the weighted feature vector and the feature pattern library, matching patterns of which the similarity is greater than a preset threshold value, extracting historical multi-dimensional feature vectors, and calculating parameters to obtain a mathematical description result; parameters in the mathematical description result are updated according to the feature vectors, and an early warning threshold value and a fault threshold value are calculated; when the feature slope in the weighted feature vector is greater than an early warning threshold but less than a fault threshold, generating an early warning signal; calculating a comprehensive health degree score distinguishing grade; and generating maintenance guidance information according to the health state level, and completing inverter detection. The method can solve the problem of resource waste.

Description

Technical Field

[0001] This invention relates to the field of aging detection technology, and in particular to an inverter aging detection method and apparatus based on multi-source data analysis. Background Technology

[0002] Currently, inverters, as core equipment in power electronic systems, play an irreplaceable role in key applications such as new energy power generation, industrial drives, and grid regulation. With the increasing reliability requirements of power systems, inverter health monitoring has become a crucial link in ensuring the safe and stable operation of the system. In the semiconductor testing field, a series of professional testing methods, including wafer testing, chip coating / thickness measurement, wafer defect / particle monitoring, wafer parameter testing, functional testing, aging testing, and probe testing, are providing precise data support for the quality control of inverter core components, comprehensively ensuring the consistent and stable operation of inverters in complex and ever-changing power system environments.

[0003] In one existing technology, a single data source is primarily used for detection. The device status is determined solely by monitoring the current and voltage signals output by the inverter. This monitoring relies on simple threshold judgments, with a pre-set fixed threshold for current, voltage, or temperature. An alarm is triggered if the monitored data exceeds this threshold. However, when hidden, gradual aging problems occur within the inverter, it is difficult to detect them accurately using only current and voltage signals. Furthermore, fixed thresholds cannot adapt to complex and changing operating scenarios. When processing the massive amounts of multi-dimensional data generated by the inverter's long-term operation, it is difficult to accurately extract key characteristics of device performance degradation, leading to either excessive maintenance wasting resources or insufficient maintenance causing malfunctions.

[0004] In summary, existing technologies suffer from resource waste. Summary of the Invention

[0005] This invention provides an inverter aging detection method and device based on multi-source data analysis to solve the problem of resource waste in the prior art.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an inverter aging detection method based on multi-source data analysis, comprising: Acquire multi-dimensional datasets and historical fusion datasets for three consecutive cycles during inverter operation; Synchronize the timestamps of voltage, current, temperature, humidity, and power data in the multi-dimensional dataset of the first cycle to obtain a fused dataset; The fused dataset is segmented using a preset time window length. The fluctuation amplitude of the data within each time window is calculated. When the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, the time window is marked as abnormal, and a multi-dimensional feature vector is generated from the data corresponding to the time window. Calculate the characteristic slopes of voltage, current, and power in the multidimensional feature vector and assign weights to obtain a weighted feature vector; Calculate the similarity between the weighted feature vector and the feature pattern vector in the pre-established feature pattern library, match feature patterns with similarity greater than a preset matching threshold, generate aging type identifiers and aging degree quantification indicators, and obtain aging classification results; Based on the aging classification results, multi-dimensional feature vectors corresponding to the aging type are extracted from the historical fusion dataset, and the decay curve parameters are calculated to obtain the mathematical description results of the aging process. The decay curve parameters in the mathematical description result of the aging process are updated based on the weighted feature vector, and the warning threshold and fault threshold are calculated. When the slope of a certain feature in the weighted feature vector is greater than the warning threshold but less than the fault threshold, a warning signal is generated immediately; when the slope of a certain feature in the weighted feature vector is greater than the fault threshold, a fault signal is generated immediately. A comprehensive health score is calculated based on the warning signal, and the inverter's health status level is obtained by classifying the levels according to the score. Calculate the health status level of the multi-dimensional dataset for the second and third cycles, generate maintenance guidance information based on the health status level, and complete the inverter aging detection.

[0007] In one optional implementation, the fused dataset is segmented using a preset time window length, the fluctuation amplitude of the data within each time window is calculated, and when the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, the time window is marked as abnormal. A multi-dimensional feature vector is then generated from the data corresponding to that time window, including: Based on a preset time window length, a continuous sequence of time segments is segmented from the fused dataset to obtain multiple time window datasets; The time-domain mean and time-domain variance of the time window dataset are calculated, and the frequency-domain spectral energy features are extracted by fast Fourier transform to obtain the data feature set. Calculate the fluctuation range of the data feature set. When the fluctuation range exceeds a preset fluctuation range threshold, mark the time window as an abnormal time period to obtain an abnormal feature set. The time-domain mean, time-domain variance, and frequency-domain spectral energy features extracted from the abnormal feature set are integrated to obtain a multi-dimensional feature vector.

[0008] In one optional implementation, the step of calculating the characteristic slopes of voltage, current, and power in the multidimensional feature vector and assigning weights to obtain a weighted feature vector includes: Calculate the feature slopes of voltage, current, and power in the multidimensional feature vector. If the feature slope exceeds a preset slope threshold, it is marked as an abnormal feature point, and an abnormal mark set is obtained. Based on the weight values ​​assigned to voltage, current, and power in the set of anomaly markers, a preliminary weighted vector is obtained; The initial weighted vector is integrated with the time-domain mean, time-domain variance, and frequency-domain spectral energy characteristics, and the vector value is calculated to obtain the integrated vector value. If the integrated weight value deviates from the preset weight range, the weight ratio is adjusted to obtain a weighted feature vector.

[0009] In one optional implementation, the step of extracting multidimensional feature vectors corresponding to the aging type from the historical fusion dataset based on the aging classification results, and calculating decay curve parameters to obtain a mathematical description of the aging process, includes: Based on the aging classification results, multi-dimensional feature vectors corresponding to the aging types are extracted from the historical fusion dataset to obtain a classification feature set; The decay curve parameters are calculated for the classification feature set to obtain the slope and intercept of the decay curve; When the slope exceeds a preset slope threshold, the classification feature set is corrected to obtain a set of correction parameters; The set of correction parameters is matched with the aging type to generate a mathematical description of the decay curve, thus obtaining a mathematical description of the aging process.

[0010] In one optional implementation, updating the decay curve parameters in the mathematical description result of the aging process based on the weighted feature vector, and calculating the warning threshold and fault threshold, includes: The weighted feature vectors are grouped to obtain grouped feature sets; The decay curve parameters in the mathematical description result of the aging process are updated based on the grouped feature set to obtain the updated parameter set; The warning threshold and fault threshold are calculated based on the updated parameter set.

[0011] In one optional implementation, the step of calculating a comprehensive health score based on the warning signal, and classifying levels according to the score to obtain the health status level of the inverter includes: Retrieve historical status levels; The decay curve parameters and feature slope of the multidimensional feature vector are calculated based on the warning signal, and classification is performed to obtain a set of classification features; A comprehensive health score is obtained by weighting the feature slope and decay curve parameters in the classification feature set. If the overall health score is higher than the preset health status threshold, the inverter is marked as healthy, and the inverter health detection result is obtained. If the overall health score is lower than the preset health status threshold, the health status history of the status level is queried to obtain the corresponding inverter's health status level.

[0012] In one optional implementation, the calculation of the health status level of the multi-dimensional dataset for the second and third cycles, the generation of maintenance guidance information based on the health status level, and the completion of inverter aging detection include: Retrieve historical maintenance records; Calculate the health status level of the multi-dimensional dataset for the second and third periods. If the health status level shows a downward trend for three consecutive periods, the prediction error of the decay curve parameters is generated. When the prediction error is greater than the preset error threshold, the matching threshold for similarity with the feature pattern library is increased, and the prediction error of the decay curve parameters is recalculated. When the prediction error is less than the preset error threshold, the historical maintenance records are queried to generate maintenance guidance information and complete the inverter aging detection.

[0013] Secondly, the present invention provides an inverter aging detection device based on multi-source data analysis, comprising: The data acquisition module is used to acquire multi-dimensional datasets and historical fusion datasets for three consecutive cycles during the operation of the inverter. The data synchronization module is used to synchronize the timestamps of voltage, current, temperature, humidity and power data in the multi-dimensional dataset of the first cycle to obtain the fused dataset; The data segmentation module is used to segment the fused dataset using a preset time window length, calculate the fluctuation amplitude of the data within each time window, mark the time window as abnormal when the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, and generate a multi-dimensional feature vector from the data corresponding to the time window. The data calculation module is used to calculate the characteristic slopes of voltage, current, and power in the multidimensional feature vector and assign weights to obtain a weighted feature vector; The data matching module is used to calculate the similarity between the weighted feature vector and the feature pattern vector in the pre-established feature pattern library, match feature patterns with similarity greater than a preset matching threshold, generate aging type identifiers and aging degree quantification indicators, and obtain aging classification results. The data extraction module is used to extract multi-dimensional feature vectors corresponding to the aging type from the historical fusion dataset based on the aging classification results, and to calculate the decay curve parameters to obtain the mathematical description results of the aging process. The data update module is used to update the decay curve parameters in the mathematical description result of the aging process according to the weighted feature vector, and to calculate the warning threshold and the fault threshold. The data generation module is used to immediately generate a warning signal when the slope of a certain feature in the weighted feature vector is greater than the warning threshold but less than the fault threshold, and to immediately generate a fault signal when the slope of a certain feature in the weighted feature vector is greater than the fault threshold. The level calculation module is used to calculate a comprehensive health score based on the warning signal, and to distinguish levels according to the score to obtain the health status level of the inverter. The data detection module is used to calculate the health status level of the multi-dimensional dataset in the second and third cycles, generate maintenance guidance information based on the health status level, and complete the inverter aging detection.

[0014] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the inverter aging detection method based on multi-source data analysis as described above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the inverter aging detection method based on multi-source data analysis described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention acquires a multi-dimensional dataset of three consecutive cycles during inverter operation, synchronizes the timestamps, uses a sliding window technique to segment and merge the dataset, calculates multi-dimensional feature vectors, and simultaneously calculates feature slopes and assigns weights to obtain a weighted feature vector. This approach breaks through the limitations of traditional single data source monitoring, fully reflects the inverter's operating status, and compared to traditional simple detection methods, the data processing and feature extraction methods can more deeply mine the information behind the data, keenly perceive subtle changes in the equipment, and identify potential problems in the progressive aging process.

[0017] (2) This invention utilizes a preset time window length to segment the dataset and calculates time-domain and frequency-domain features, enabling more accurate extraction of key feature information related to device performance degradation. By calculating the similarity between the weighted feature vector and a pre-established feature pattern library, and matching feature patterns with similarity greater than a preset matching threshold, different aging types can be accurately identified, overcoming the difficulties in accurately extracting aging features from complex multi-source data and the challenges posed by the interweaving and strong concealment of feature patterns of different aging types.

[0018] (3) Based on the aging classification results, this invention extracts multi-dimensional feature vectors corresponding to the aging type from the historical fusion dataset and calculates the decay curve parameters to obtain the mathematical description results of the aging process. Simultaneously, the decay curve parameters in the mathematical description results of the aging process are dynamically updated based on the weighted feature vectors, and the warning threshold and fault threshold are calculated. This achieves a quantitative assessment of the aging process. Compared with the traditional fixed threshold judgment method, it can reasonably set aging thresholds, accurately and timely determine the health status of equipment, and avoid the problems of over-maintenance or under-maintenance. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the inverter aging detection method based on multi-source data analysis provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the inverter data aging detection structure provided by the present invention; Figure 3 This is a schematic diagram of the inverter aging detection device based on multi-source data analysis provided in the second embodiment of the present invention. Detailed Implementation

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0021] Reference Figure 1 The first embodiment of the present invention provides an inverter aging detection method based on multi-source data analysis, including the following steps: S11, obtain the multi-dimensional dataset and historical fusion dataset for three consecutive cycles during inverter operation; S12, synchronize the timestamps of voltage, current, temperature, humidity and power data in the multi-dimensional dataset of the first cycle to obtain the fused dataset; S13, the fused dataset is segmented using a preset time window length, the fluctuation amplitude of the data within each time window is calculated, and when the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, the time window is marked as abnormal, and a multi-dimensional feature vector is generated from the data corresponding to the time window. S14, calculate the characteristic slopes of voltage, current and power in the multidimensional feature vector and assign weights to obtain a weighted feature vector; S15, calculate the similarity between the weighted feature vector and the feature pattern vector in the pre-established feature pattern library, match feature patterns with similarity greater than the preset matching threshold, generate aging type identifier and aging degree quantification index, and obtain aging classification result; S16. Based on the aging classification results, extract the multi-dimensional feature vector corresponding to the aging type from the historical fusion dataset, and calculate the decay curve parameters to obtain the mathematical description result of the aging process; S17. Update the decay curve parameters in the mathematical description result of the aging process based on the weighted feature vector, and calculate the warning threshold and fault threshold. S18, when the slope of a certain feature in the weighted feature vector is greater than the warning threshold but less than the fault threshold, a warning signal is generated immediately; when the slope of a certain feature in the weighted feature vector is greater than the fault threshold, a fault signal is generated immediately. S19, Calculate a comprehensive health score based on the warning signal, classify the levels according to the score, and obtain the health status level of the inverter; S110, calculate the health status level of the multi-dimensional dataset for the second and third cycles, generate maintenance guidance information based on the health status level, and complete the inverter aging detection.

[0022] In step S11, the multi-dimensional dataset and historical fusion dataset for three consecutive cycles during the operation of the inverter are obtained.

[0023] It should be noted that multiple sensors are installed at different locations on the inverter to acquire voltage waveform data, current waveform data, temperature environmental data, humidity environmental data, and power output data. This data is then synthesized into a multi-dimensional dataset. The cycle length is determined based on the inverter's output frequency or operating cycle, and data is collected continuously for three such cycles. The collected multi-dimensional data is temporarily stored in a data buffer or temporary database, recorded using timestamps. The historical fusion dataset is stored in a long-term storage system and can be retrieved from the long-term storage system using database query languages ​​(such as SQL) or distributed file system APIs.

[0024] In step S12, the timestamps of voltage, current, temperature, humidity, and power data in the multi-dimensional dataset of the first cycle are synchronized to obtain the fused dataset.

[0025] It should be noted that the initial time points of voltage, current, temperature, humidity, and power data are synchronized using time series comparison tools to obtain the initially aligned dataset. These tools can be implemented using the Pandas library, which provides powerful data structures and functions for efficient processing of time series data, including alignment, cleaning, and transformation. Pandas stores time series data from different sources in a DataFrame structure, and then uses its built-in functions such as merging and joining to compare the time series and identify the correspondences between different dimensions of data at the timestamps. This data is then synchronized using a unified timestamp, and interpolation is used to complete missing or out-of-synchronization data points. Linear interpolation can be used for calculation; the process of linear interpolation involves calculating approximate values ​​for missing points based on the values ​​of known points and the time intervals, where the data changes linearly between two known points. For example, when voltage data is missing at a certain timestamp, an approximate value is obtained through linear interpolation. If the voltage is 218V at t=1s and 222V at t=2s, then the interpolated voltage at t=1.5s is 220V. This ensures all data are aligned on the same time axis, thus determining the synchronized data set. Based on this synchronized data set, the voltage, current, temperature, humidity, and power data are standardized using z-scores. The z-score value can be calculated using a formula. Taking voltage data as an example, if the calculated mean is 220V and the standard deviation is 2V, then the z-score value for a data point of 218V is (218-220) / 2 = -1, resulting in the standardized data set. Using these standardized data, a fusion dataset is constructed. The standardized data is integrated into a unified vector format to generate the fusion dataset. The standardized data is arranged in the order of timestamps, and a vector is created for each time point. The standardized data values ​​of different dimensions corresponding to that time point are placed into the vector in sequence. For example, the fusion data vector at a certain timestamp is [voltage z-score: -1, current z-score: 0.5, temperature z-score: 0.2, humidity z-score: -0.3, power z-score: 0.8].

[0026] In step S13, the fused dataset is segmented using a preset time window length, and the fluctuation amplitude of the data within each time window is calculated. When the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, the time window is marked as abnormal, and a multidimensional feature vector is generated from the data corresponding to the time window.

[0027] In one implementation, the fused dataset is segmented using a preset time window length, the fluctuation amplitude of the data within each time window is calculated, and when the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, the time window is marked as abnormal. A multi-dimensional feature vector is then generated from the data corresponding to that time window, including: Based on a preset time window length, a continuous sequence of time segments is segmented from the fused dataset to obtain multiple time window datasets; The time-domain mean and time-domain variance of the time window dataset are calculated, and the frequency-domain spectral energy features are extracted by fast Fourier transform to obtain the data feature set. Calculate the fluctuation range of the data feature set. When the fluctuation range exceeds a preset fluctuation range threshold, mark the time window as an abnormal time period to obtain an abnormal feature set. The time-domain mean, time-domain variance, and frequency-domain spectral energy features extracted from the abnormal feature set are integrated to obtain a multi-dimensional feature vector.

[0028] It should be noted that the data is divided into time windows of a preset length of 10 seconds, with a sliding step of 5 seconds, allowing for overlap between adjacent windows. Starting from the beginning of the fused dataset, a 10-second time segment is selected as the first window dataset. Then, the window slides for 5 seconds, selecting the next 10-second time segment as the second window dataset, and so on, until the entire fused dataset is covered. This process segments a continuous sequence of time segments from the fused dataset, resulting in multiple time window datasets. For each time window dataset, its time-domain mean and time-domain variance are calculated. For each data point in the time window dataset, its average value is calculated to obtain the time-domain mean. The square of the difference between each data point's value and the mean is calculated, and the average of these squared differences is then used to obtain the time-domain variance. Frequency domain spectral energy features are extracted using a Fast Fourier Transform (FFT). The data within the window is converted into a frequency domain signal using an FFT algorithm to obtain a spectrum. Based on the spectrum, the main energy concentration frequency bands are identified, resulting in a data feature set containing the time-domain mean, time-domain variance, and frequency domain spectral energy.

[0029] The fluctuation amplitude of the data feature set is calculated, where the fluctuation amplitude is the difference between the maximum and minimum values ​​of the signal within the window. If the fluctuation amplitude in the feature set exceeds a preset fluctuation amplitude threshold, the time window is marked as an abnormal time period, resulting in an abnormal feature set. If the threshold is not exceeded, the window is marked as a normal time period. The time-domain mean, variance, and frequency-domain spectral energy features of each window are integrated into a multi-dimensional feature vector. Principal component analysis (PCA) can be used for this integration. Statistical indicators of fluctuation amplitude, such as mean and standard deviation, can be calculated from historical data under normal operating conditions. The set threshold fluctuation amplitude is the mean plus a certain multiple of the standard deviation. The PCA algorithm implementation process involves first standardizing the data. Different features have different dimensions and numerical ranges, requiring data standardization. This process is the same as z-score standardization, eliminating dimensional differences between different features. The covariance matrix of the standardized data is then calculated, and eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors. Based on the magnitude of the eigenvalues, the eigenvectors corresponding to the top 50% largest eigenvalues ​​are selected as principal components. The data is then projected onto the selected principal component directions to obtain multidimensional eigenvectors.

[0030] In step S14, the characteristic slopes of voltage, current, and power in the multidimensional feature vector are calculated and weights are assigned to obtain a weighted feature vector.

[0031] In one implementation, the step of calculating the characteristic slopes of voltage, current, and power in the multidimensional feature vector and assigning weights to obtain a weighted feature vector includes: Calculate the feature slopes of voltage, current, and power in the multidimensional feature vector. If the feature slope exceeds a preset slope threshold, it is marked as an abnormal feature point, and an abnormal mark set is obtained. Based on the weight values ​​assigned to voltage, current, and power in the set of anomaly markers, a preliminary weighted vector is obtained; The initial weighted vector is integrated with the time-domain mean, time-domain variance, and frequency-domain spectral energy characteristics, and the vector value is calculated to obtain the integrated vector value. If the integrated weight value deviates from the preset weight range, the weight ratio is adjusted to obtain a weighted feature vector.

[0032] It should be noted that a preset slope calculation tool is used to calculate the characteristic slopes of voltage, current, and power in the multidimensional feature vector. Taking the voltage signal as an example, a 10-second sliding window (10,000 data points) is selected with a step size of 5 seconds. The voltage data within the window is fitted using the least squares method, resulting in a slope of 0.02V / s, representing the trend of voltage change over time. Similarly, the current slope is calculated to be 0.01A / s, and the power slope is 0.15W / s, reflecting the dynamic changes of each parameter. If the characteristic slope exceeds the preset slope threshold range, it is marked as an abnormal feature point, resulting in an abnormal marker set. The mean and standard deviation of the characteristic slopes of voltage, current, and power are calculated from a large amount of historical data from normal operation. Taking voltage as an example, assuming the calculated slope mean is 0.02V / s and the standard deviation is 0.01V / s, a preset slope threshold range can be set to the mean plus twice the standard deviation. Weight values ​​are assigned to the voltage, current, and power. Based on historical data analysis, voltage accounts for 40% of the inverter's state influence, current 30%, and power 30%. Therefore, the weight allocation is [0.4, 0.3, 0.3], resulting in a preliminary weighted vector. The time-domain mean, time-domain variance, and frequency-domain energy characteristics are obtained from the preliminary weighted vector. A data fusion tool is used to integrate the data with the preliminary weighted vector, calculating the vector value. Based on the weight values ​​of each feature, each element in the preliminary weighted vector is multiplied by its corresponding weight. The data fusion tool can use Pandas provided in Python, which is suitable for processing time series data and feature fusion. The results are then summed to obtain a comprehensive vector value, the integrated vector value. If the integrated vector value deviates from the preset weight range, the weight ratio is reduced to obtain a weighted feature vector. The entropy value of each feature (voltage, current, power) is calculated based on historical data. The smaller the entropy value, the higher the importance of the feature. The weight of each feature is calculated based on its entropy value, and the resulting weight range is the preset weight range.

[0033] In step S15, the similarity between the weighted feature vector and the feature pattern vector in the pre-established feature pattern library is calculated. Feature patterns with similarity greater than a preset matching threshold are matched to generate aging type identifiers and aging degree quantification indicators, and aging classification results are obtained.

[0034] It should be noted that a feature vector set is obtained from a pre-established feature pattern library. The cosine similarity value between the input weighted feature vector and each standard vector in the feature vector set is calculated to obtain a similarity set. The feature pattern library contains standard vectors for three aging types, forming the feature vector set. For example, capacitor aging pattern vector = [0.3, 0.08, 1600], switching device aging pattern vector = [0.2, 0.15, 1550], and heat dissipation system aging pattern vector = [0.22, 0.09, 1400]. For the features... The cosine similarity is calculated for both the vector set and the weighted feature vectors using the formula: Similarity = (Weighted Feature Vector × Standard Vector) / (Magnitude of Weighted Feature Vector × Magnitude of Standard Vector). Based on the similarity set, each similarity value is compared to a preset matching threshold. If the similarity value is less than the threshold, it is marked as a non-matching feature point; if it is greater, it is marked as a matching feature point, resulting in a matching feature set. The matching threshold is set to 0.95. A for loop iterates through each matching feature point in the set, searching the feature pattern library for the corresponding feature pattern vector. An aging type identifier is extracted from the feature pattern vector. For example, each feature pattern vector in the feature vector set has a corresponding aging type identifier, such as "capacitor aging," "switching device aging," or "heat dissipation system aging," generating an aging type set. For the set of aging types, calculate the quantitative index of each aging type in the set of aging types. First, calculate the Euclidean distance between the weighted feature vector and the standard vector, and then use the formula to calculate the quantitative index of each aging type. The formula is Euclidean distance / maximum Euclidean distance. The maximum Euclidean distance is set to 2000 to determine the aging classification result.

[0035] In step S16, based on the aging classification results, multidimensional feature vectors corresponding to the aging type are extracted from the historical fusion dataset, and decay curve parameters are calculated to obtain the mathematical description results of the aging process.

[0036] In one implementation, the step of extracting multi-dimensional feature vectors corresponding to the aging type from the historical fusion dataset based on the aging classification results, and calculating decay curve parameters to obtain a mathematical description of the aging process includes: Based on the aging classification results, multi-dimensional feature vectors corresponding to the aging types are extracted from the historical fusion dataset to obtain a classification feature set; The decay curve parameters are calculated for the classification feature set to obtain the slope and intercept of the decay curve; When the slope exceeds a preset slope threshold, the classification feature set is corrected to obtain a set of correction parameters; The set of correction parameters is matched with the aging type to generate a mathematical description of the decay curve, thus obtaining a mathematical description of the aging process.

[0037] It should be noted that, based on the aging classification results, a set of multi-dimensional feature vectors corresponding to capacitor aging, switching device aging, and heat dissipation system aging is obtained from the historical fusion dataset to obtain a classification feature set. A linear regression tool is used to calculate the slope and intercept of the decay curve for each aging type. The linear regression tool can use the scikit-learn library in Python, which provides linear regression models for convenient linear regression analysis. By defining the feature matrix and target vector, the `fit` method of the linear regression model is called for training to obtain the slope and intercept. If the slope of the decay curve parameters exceeds a preset parameter threshold, the classification feature set is processed using a moving average method to determine a corrected parameter set. The moving average method smooths the data by calculating the average of adjacent data points; for example, a simple moving average or a weighted moving average method can be used. A parameter mapping tool is used to match the corrected parameter set with the aging type to generate a mathematical description of the decay curve, thus obtaining a mathematical description of the aging process. The mathematical description uses a linear regression model Y=S*t+I, where Y is the aging quantification index, S is the slope representing the aging rate, I is the intercept representing the initial state, and t is the time change. Parameter mapping can be achieved using Python dictionaries, which match the corrected parameter set with the aging type. For example, a dictionary can be created where the keys are the aging types and the values ​​are the corresponding corrected parameter sets.

[0038] In step S17, the decay curve parameters in the mathematical description result of the aging process are updated according to the weighted feature vector, and the warning threshold and fault threshold are calculated.

[0039] In one implementation, updating the decay curve parameters in the mathematical description result of the aging process based on the weighted feature vector, and calculating the warning threshold and the fault threshold, includes: The weighted feature vectors are grouped to obtain grouped feature sets; The decay curve parameters in the mathematical description result of the aging process are updated based on the grouped feature set to obtain the updated parameter set; The warning threshold and fault threshold are calculated based on the updated parameter set.

[0040] It should be noted that the weighted feature vectors are grouped according to the degree of aging and the quantitative index. Based on the range of the aging degree and the quantitative index, the weighted feature vectors are divided into different groups. For example, the aging degree can be divided into three intervals: mild aging (0-33%), moderate aging (34%-66%), and severe aging (67%-100%). All weighted feature vectors are iterated over, and each vector is assigned to the corresponding group based on its quantitative index value. For example, if a weighted feature vector corresponds to an aging degree of 40%, it is assigned to the moderate aging group. A Kalman filter algorithm is used to fuse the grouped feature sets and calculate the weighted average. The weighted average is the smoothed feature value. The decay curve parameters S (slope) and I (intercept) are updated. S is calculated using the standard deviation of historical data; assuming a standard deviation of 0.5, then S = 0.5. I is calculated based on a linear regression model of running time and feature values; assuming 1000 hours of operation, I = 0.002 × number of hours = 2. Calculate the warning threshold and the fault threshold, setting the warning threshold as the 75th quantile of the slope and the fault threshold as the 95th quantile.

[0041] In step S18, when the slope of a certain feature in the weighted feature vector is greater than the warning threshold but less than the fault threshold, a warning signal is generated immediately; when the slope of a certain feature in the weighted feature vector is greater than the fault threshold, a fault signal is generated immediately.

[0042] It should be noted that when the slope of a feature in the weighted feature vector is less than the warning threshold, no signal is generated, indicating normal operation; when the slope of a feature in the weighted feature vector is greater than the warning threshold but less than the fault threshold, a warning signal is generated immediately; when the slope of a feature in the weighted feature vector is greater than the fault threshold, a fault signal is generated immediately.

[0043] In step S19, a comprehensive health score is calculated based on the warning signal, and the levels are distinguished according to the score to obtain the health status level of the inverter.

[0044] In one implementation, the step of calculating a comprehensive health score based on the warning signal, and classifying the levels according to the score to obtain the health status level of the inverter includes: Retrieve historical status levels; The decay curve parameters and feature slope of the multidimensional feature vector are calculated based on the warning signal, and classification is performed to obtain a set of classification features; A comprehensive health score is obtained by weighting the feature slope and decay curve parameters in the classification feature set. If the overall health score is higher than the preset health status threshold, the inverter is marked as healthy, and the inverter health detection result is obtained. If the overall health score is lower than the preset health status threshold, the health status history of the status level is queried to obtain the corresponding inverter's health status level.

[0045] It should be noted that the inverter's past health status records are retrieved from the historical database. These records contain health status information of the inverter at different points in time. When the emitted signal is a warning signal, the attenuation curve parameters and feature slope of the multi-dimensional feature vector are calculated. Based on the feature slope and attenuation curve parameters, the multi-dimensional feature vector is classified to form a classification feature set. The weights of the feature slope and attenuation curve parameters are determined according to their influence on the inverter's health status; the weight of the feature slope is set to 0.6, and the weight of the attenuation curve parameter is set to 0.4. The feature slope and attenuation curve parameters in the classification feature set are weighted and calculated using the formula: Comprehensive Health Score = Feature Slope × 0.6 + Attenuation Curve Parameter × 0.4. The calculated comprehensive health score is compared with a preset health status threshold. If the comprehensive health score is higher than the preset health status threshold, the inverter is marked as healthy; if it is lower than the threshold, the inverter is marked as having a health problem. The historical status record is then retrieved to find historical records with similar feature slope and attenuation curve parameters to the current one. Based on the retrieved historical records and the current feature slope and attenuation curve parameters, the inverter's health status level is determined. For example, the health status can be divided into three levels: healthy, sub-healthy, and faulty. Based on the matching results of the comprehensive health score and historical records, the inverter can be determined to be in the corresponding health status level.

[0046] In step S110, the health status level of the multi-dimensional dataset for the second and third cycles is calculated, maintenance guidance information is generated based on the health status level, and the inverter aging detection is completed.

[0047] In one implementation, the calculation of the health status levels of the multi-dimensional dataset for the second and third cycles, the generation of maintenance guidance information based on the health status levels, and the completion of inverter aging detection include: Retrieve historical maintenance records; Calculate the health status level of the multi-dimensional dataset for the second and third periods. If the health status level shows a downward trend for three consecutive periods, the prediction error of the decay curve parameters is generated. When the prediction error is greater than the preset error threshold, the matching threshold for similarity with the feature pattern library is increased, and the prediction error of the decay curve parameters is recalculated. When the prediction error is less than the preset error threshold, the historical maintenance records are queried to generate maintenance guidance information and complete the inverter aging detection.

[0048] It should be noted that the historical maintenance records of the inverter are retrieved from the maintenance database, including information such as past maintenance time, maintenance content, replaced components, and repaired faults. Similar to the calculation process of the multi-dimensional dataset in the first cycle, the health status level of the multi-dimensional dataset in the second and third cycles is calculated. If the health status level shows a downward trend for three consecutive cycles, the prediction error of the attenuation curve parameters is generated. The difference between the predicted value and the actual value is calculated to obtain the prediction error. The predicted value can be obtained through the linear regression model Y=S*t+I, and the actual value refers to the true value of the inverter's current health status obtained through actual measurement or calculation. The prediction error is compared with a preset error threshold. If the prediction error is greater than the preset error threshold, the matching threshold for similarity matching with the feature pattern library is increased, and the prediction error of the attenuation curve parameters is recalculated. After adjusting the matching threshold, similarity matching and prediction error calculation are performed again until the prediction error is less than the preset error threshold. When the prediction error is less than the preset error threshold, the retrieved historical maintenance records are queried to find historical records similar to the current health status level and feature pattern. Based on the retrieved historical maintenance records, maintenance guidance information is generated, including suggested maintenance time, maintenance content, and components requiring attention, to complete the inverter aging test.

[0049] In summary, this invention acquires a multi-dimensional dataset from three consecutive cycles of inverter operation, synchronizes the timestamps, segments and merges the dataset using a sliding window technique, calculates multi-dimensional feature vectors, and simultaneously calculates feature slopes and assigns weights to obtain a weighted feature vector. This approach breaks through the limitations of traditional single-data source monitoring, fully reflecting the inverter's operating status. Compared to traditional simple detection methods, the data processing and feature extraction methods can more deeply mine the information behind the data, keenly perceive subtle changes in the equipment, and identify potential problems in the progressive aging process. By segmenting the dataset using a preset time window length and calculating time-domain and frequency-domain features, key feature information on equipment performance degradation can be extracted more accurately. By calculating the similarity between the weighted feature vector and a pre-established feature pattern library, and matching feature patterns with similarity greater than a preset matching threshold, different aging types can be accurately identified, overcoming the difficulties of accurately extracting aging features from complex multi-source data and the challenges posed by the intertwining and strong concealment of feature patterns of different aging types. Based on the aging classification results, this invention extracts multi-dimensional feature vectors corresponding to the aging type from the historical fused dataset and calculates the decay curve parameters to obtain a mathematical description of the aging process. Simultaneously, the decay curve parameters in the mathematical description of the aging process are dynamically updated based on the weighted feature vector, and the warning threshold and fault threshold are calculated. This enables a quantitative assessment of the aging process. Compared with traditional fixed threshold judgment methods, it can reasonably set aging thresholds and accurately and timely determine the health status of equipment, avoiding the problems of over-maintenance or under-maintenance.

[0050] Reference Figure 3 The second embodiment of the present invention provides an inverter aging detection device based on multi-source data analysis, comprising: The data acquisition module is used to acquire multi-dimensional datasets and historical fusion datasets for three consecutive cycles during inverter operation. The data synchronization module is used to synchronize the timestamps of voltage, current, temperature, humidity and power data in the multi-dimensional dataset of the first cycle to obtain the fused dataset; The data segmentation module is used to segment the fused dataset using a preset time window length, calculate the fluctuation amplitude of the data within each time window, mark the time window as abnormal when the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, and generate a multi-dimensional feature vector from the data corresponding to the time window. The data calculation module is used to calculate the characteristic slopes of voltage, current, and power in the multidimensional feature vector and assign weights to obtain a weighted feature vector; The data matching module is used to calculate the similarity between the weighted feature vector and the feature pattern vector in the pre-established feature pattern library, match feature patterns with similarity greater than a preset matching threshold, generate aging type identifiers and aging degree quantification indicators, and obtain aging classification results. The data extraction module is used to extract multi-dimensional feature vectors corresponding to the aging type from the historical fusion dataset based on the aging classification results, and to calculate the decay curve parameters. The data update module is used to update the attenuation curve parameters according to the weighted feature vector, and to calculate the warning threshold and the fault threshold. The data generation module is used to generate a warning signal when the feature slope in the weighted feature vector is greater than the warning threshold but less than the fault threshold, and to generate a fault signal when the feature slope in the weighted feature vector is greater than the fault threshold. The level calculation module is used to calculate a comprehensive health score based on the warning signal, and to distinguish levels according to the score to obtain the health status level of the inverter. The data detection module is used to calculate the health status level of the multi-dimensional dataset in the second and third cycles, generate maintenance guidance information based on the health status level, and complete the inverter aging detection.

[0051] It should be noted that the inverter aging detection device based on multi-source data analysis provided in this embodiment of the invention is used to execute all the process steps of the inverter aging detection method based on multi-source data analysis in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0052] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program, such as an algorithm program, stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps described in the various embodiments of the inverter aging detection method based on multi-source data analysis, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the data detection module.

[0053] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0054] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0055] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0056] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0057] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0058] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0059] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An inverter aging detection method based on multi-source data analysis, characterized in that, include: Acquire multi-dimensional datasets and historical fusion datasets for three consecutive cycles during inverter operation; Synchronize the timestamps of voltage, current, temperature, humidity, and power data in the multi-dimensional dataset of the first cycle to obtain a fused dataset; The fused dataset is segmented using a preset time window length. The fluctuation amplitude of the data within each time window is calculated. When the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, the time window is marked as abnormal, and a multidimensional feature vector is generated from the data corresponding to the time window. Calculate the characteristic slopes of voltage, current, and power in the multidimensional feature vector and assign weights to obtain a weighted feature vector; Calculate the similarity between the weighted feature vector and the feature pattern vector in the pre-established feature pattern library, match feature patterns with similarity greater than a preset matching threshold, generate aging type identifiers and aging degree quantification indicators, and obtain aging classification results; Based on the aging classification results, multi-dimensional feature vectors corresponding to the aging type are extracted from the historical fusion dataset, and the decay curve parameters are calculated to obtain the mathematical description results of the aging process. The decay curve parameters in the mathematical description result of the aging process are updated based on the weighted feature vector, and the warning threshold and fault threshold are calculated. When the slope of a certain feature in the weighted feature vector is greater than the warning threshold but less than the fault threshold, a warning signal is generated immediately; when the slope of a certain feature in the weighted feature vector is greater than the fault threshold, a fault signal is generated immediately. A comprehensive health score is calculated based on the warning signal, and the inverter's health status level is obtained by classifying the levels according to the score. Calculate the health status level of the multi-dimensional dataset for the second and third cycles, generate maintenance guidance information based on the health status level, and complete the inverter aging detection.

2. The inverter aging detection method based on multi-source data analysis according to claim 1, characterized in that, The fused dataset is segmented using a preset time window length. The fluctuation amplitude of the data within each time window is calculated. When the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, the time window is marked as abnormal, and a multi-dimensional feature vector is generated from the data corresponding to that time window, including: Based on a preset time window length, a continuous sequence of time segments is segmented from the fused dataset to obtain multiple time window datasets; The time-domain mean and time-domain variance of the time window dataset are calculated, and the frequency-domain spectral energy features are extracted by fast Fourier transform to obtain the data feature set. Calculate the fluctuation range of the data feature set. When the fluctuation range exceeds a preset fluctuation range threshold, mark the time window as an abnormal time period to obtain an abnormal feature set. The time-domain mean, time-domain variance, and frequency-domain spectral energy features extracted from the abnormal feature set are integrated to obtain a multi-dimensional feature vector.

3. The inverter aging detection method based on multi-source data analysis according to claim 1, characterized in that, The calculation of the characteristic slopes of voltage, current, and power in the multidimensional feature vector and the assignment of weights to obtain a weighted feature vector include: Calculate the feature slopes of voltage, current, and power in the multidimensional feature vector. If the feature slope exceeds a preset slope threshold, it is marked as an abnormal feature point, and an abnormal mark set is obtained. Based on the weight values ​​assigned to voltage, current, and power in the set of anomaly markers, a preliminary weighted vector is obtained; The initial weighted vector is integrated with the time-domain mean, time-domain variance, and frequency-domain spectral energy characteristics, and the vector value is calculated to obtain the integrated vector value. If the integrated weight value deviates from the preset weight range, the weight ratio is adjusted to obtain a weighted feature vector.

4. The inverter aging detection method based on multi-source data analysis according to claim 1, characterized in that, Based on the aging classification results, the multi-dimensional feature vectors corresponding to the aging type are extracted from the historical fusion dataset, and the decay curve parameters are calculated to obtain the mathematical description results of the aging process, including: Based on the aging classification results, multi-dimensional feature vectors corresponding to the aging types are extracted from the historical fusion dataset to obtain a classification feature set; The decay curve parameters are calculated for the classification feature set to obtain the slope and intercept of the decay curve; When the slope exceeds a preset slope threshold, the classification feature set is corrected to obtain a set of correction parameters; The set of correction parameters is matched with the aging type to generate a mathematical description of the decay curve, thus obtaining a mathematical description of the aging process.

5. The inverter aging detection method based on multi-source data analysis according to claim 1, characterized in that, The step of updating the decay curve parameters in the mathematical description result of the aging process based on the weighted feature vector, and calculating the warning threshold and fault threshold, includes: The weighted feature vectors are grouped to obtain grouped feature sets; The decay curve parameters in the mathematical description result of the aging process are updated based on the grouped feature set to obtain the updated parameter set; The warning threshold and fault threshold are calculated based on the updated parameter set.

6. The inverter aging detection method based on multi-source data analysis according to claim 1, characterized in that, The process of calculating a comprehensive health score based on the warning signal, classifying levels according to the score, and obtaining the inverter's health status level includes: Retrieve historical status levels; The decay curve parameters and feature slope of the multidimensional feature vector are calculated based on the warning signal, and classification is performed to obtain a set of classification features; A comprehensive health score is obtained by weighting the feature slope and decay curve parameters in the classification feature set. If the overall health score is higher than the preset health status threshold, the inverter is marked as healthy, and the inverter health detection result is obtained. If the overall health score is lower than the preset health status threshold, the health status history of the status level is queried to obtain the corresponding inverter's health status level.

7. The inverter aging detection method based on multi-source data analysis according to claim 1, characterized in that, The calculation of the health status level of the multi-dimensional dataset for the second and third cycles, the generation of maintenance guidance information based on the health status level, and the completion of inverter aging detection include: Retrieve historical maintenance records; Calculate the health status level of the multi-dimensional dataset for the second and third periods. If the health status level shows a downward trend for three consecutive periods, the prediction error of the decay curve parameters is generated. When the prediction error is greater than the preset error threshold, the matching threshold for similarity with the feature pattern library is increased, and the prediction error of the decay curve parameters is recalculated. When the prediction error is less than the preset error threshold, the historical maintenance records are queried to generate maintenance guidance information and complete the inverter aging detection.

8. An inverter aging detection device based on multi-source data analysis, characterized in that, include: The data acquisition module is used to acquire multi-dimensional datasets and historical fusion datasets for three consecutive cycles during inverter operation. The data synchronization module is used to synchronize the timestamps of voltage, current, temperature, humidity, and power data in the multi-dimensional dataset of the first cycle to obtain a fused dataset; The data segmentation module is used to segment the fused dataset using a preset time window length, calculate the fluctuation amplitude of the data within each time window, mark the time window as abnormal when the fluctuation amplitude exceeds a preset fluctuation amplitude threshold, and generate a multi-dimensional feature vector from the data corresponding to the time window. The data calculation module is used to calculate the characteristic slopes of voltage, current, and power in the multidimensional feature vector and assign weights to obtain a weighted feature vector; The data matching module is used to calculate the similarity between the weighted feature vector and the feature pattern vector in the pre-established feature pattern library, match feature patterns with similarity greater than a preset matching threshold, generate aging type identifiers and aging degree quantification indicators, and obtain aging classification results. The data extraction module is used to extract multi-dimensional feature vectors corresponding to the aging type from the historical fusion dataset based on the aging classification results, and to calculate the decay curve parameters to obtain the mathematical description results of the aging process. The data update module is used to update the decay curve parameters in the mathematical description result of the aging process according to the weighted feature vector, and to calculate the warning threshold and the fault threshold. The data generation module is used to immediately generate a warning signal when the slope of a certain feature in the weighted feature vector is greater than the warning threshold but less than the fault threshold, and to immediately generate a fault signal when the slope of a certain feature in the weighted feature vector is greater than the fault threshold. The level calculation module is used to calculate a comprehensive health score based on the warning signal, and to distinguish levels according to the score to obtain the health status level of the inverter. The data detection module is used to calculate the health status level of the multi-dimensional dataset in the second and third cycles, generate maintenance guidance information based on the health status level, and complete the inverter aging detection.

9. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the inverter aging detection method based on multi-source data analysis as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the inverter aging detection method based on multi-source data analysis as described in any one of claims 1 to 7.

Images