A test method based on the frequent start and stop phenomenon of inverter

Abstract

The invention discloses a testing method based on the phenomenon of frequent start and stop of an inverter, and relates to the technical field of inverter detection. The specific steps of the testing method are as follows: S100, detection preparation: collecting inverter model, parameters, structure, fault history and installation environment information, selecting and calibrating an infrared thermal imager and an ultrasonic detection device according to inverter characteristics and environment. The invention introduces a data fusion algorithm, organically fuses temperature data collected by the infrared thermal imager with structural data acquired by an ultrasonic detection probe, accurately calculates temperature data fusion weights and ultrasonic data fusion weights, and performs correlation model training and optimization based on an inverter physical model and fault mechanism, thereby achieving comprehensive and accurate diagnosis of inverter faults. Compared with traditional single sensor detection, the method can more sensitively capture temperature changes on the inverter surface and abnormal conditions of the internal structure, and effectively improves the accuracy and reliability of fault detection.

Description

Technical Field

[0001] The present invention relates to the technical field of inverter detection, and in particular to a testing method based on the frequent start-stop phenomenon of an inverter. Background Art

[0002] In the current power system, inverters play a vital role. As a key device for power conversion, inverters are responsible for converting DC power into AC power and providing stable power for various power loads. With the continuous development of power systems and the improvement of their intelligence levels, the requirements for inverter performance and stability are becoming increasingly higher. However, in the actual operation of inverters, the problem of frequent start and stop has become increasingly prominent, which not only affects the service life and performance of the inverter, but also poses a threat to the stable operation of the entire power system.

[0003] To address the problem of frequent start and stop of inverters, traditional testing methods often use contact sensors for detection. This method requires the sensor to be installed directly on the inverter and its working status is judged by monitoring the operating parameters of the inverter. However, traditional testing methods have many shortcomings. On the one hand, the installation of the sensor will cause certain interference to the normal operation of the inverter and affect the performance and stability of the inverter. On the other hand, in some special environments, the installation and maintenance of the sensor face great difficulties, which not only increases the detection cost, but also poses a threat to the safety of the detection personnel. In addition, the traditional method is not accurate enough when diagnosing the reasons for the frequent start and stop of the inverter. There are many misjudgments and missed judgments, which makes it difficult to adapt to the detection needs of inverters under different models and working conditions.

[0004] Therefore, a testing method based on the frequent start and stop phenomenon of the inverter was developed. The non-contact detection technology was adopted to realize real-time monitoring and fault diagnosis of the inverter operating status through infrared thermal imaging and ultrasonic detection means, which provided strong guarantee for the stable operation of the inverter and the safety of the power system. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the prior art and provide a testing method based on the frequent start and stop phenomenon of the inverter. This method not only avoids the interference of sensor installation on the inverter performance, but also improves the accuracy and safety of detection. At the same time, by introducing data fusion algorithm and adaptive threshold learning model, the present invention can more sensitively capture the fault characteristics of the inverter, reduce the false alarm and missed alarm rates, and provide strong technical support for the stable operation of the inverter and the safety and reliability of the power system.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a test method based on the frequent start and stop phenomenon of the inverter, the specific steps of the test method are:

[0007] S100, test preparation: collect the inverter model, parameters, structure, fault history and installation environment information, select and calibrate the infrared thermal imager and ultrasonic detection equipment according to the inverter characteristics and environment, and plan the infrared thermal imager scanning path, frequency and ultrasonic detection probe position, signal parameters and data collection time interval according to the inverter structure and fault-prone parts;

[0008] S200, data acquisition: infrared thermal imagers and ultrasonic detection probes are installed around the inverter. The surface temperature data of the inverter is collected by scanning the infrared thermal imager according to the predetermined path and frequency, and transmitted to the data processing center. The ultrasonic detection probe transmits and receives ultrasonic signals, and the built-in signal processing circuit is used to amplify, filter, and digitize the signals. After processing, the data is transmitted to the data processing center;

[0009] S300, data processing: perform quality inspection on infrared thermal imaging temperature data, remove abnormal values, perform temperature compensation processing based on the impact of ambient temperature on inverter surface temperature measurement, extract amplitude, time delay, and frequency features from ultrasonic detection data and perform normalization;

[0010] S400, data fusion: According to the inverter physical model and fault mechanism, the intrinsic relationship between temperature change and internal structure change is analyzed, the correlation model is trained and optimized through historical fault data and normal operation data, the parameters in the model are determined, and the fusion weights are assigned to the temperature data and ultrasonic data according to the correlation model. The temperature data fusion weight w is set T (t) is calculated as: α avg (t) is the average temperature anomaly coefficient of the entire inverter surface at time t, α mid is an intermediate threshold, μ is the weight adjustment coefficient, t represents time, and the ultrasonic data fusion weight w S (t) The calculation formula is: S (t) = 1-w T (t), use appropriate data fusion algorithm to fuse the preprocessed temperature data and ultrasonic data, let F temp-struc (t) is the fused fault indication value, and the fusion formula is: Where α(x,y,t) represents the temperature anomaly coefficient, and t represents time;

[0011] S500, adaptive threshold learning: obtain the detection data of the inverter under different working conditions from the historical database, annotate these data in detail, divide the annotated historical data into a training set and a validation set, and build an adaptive threshold learning model. The input of the model is the fused data set, and the output is the temperature threshold and the ultrasonic signal abnormality threshold. The model is trained using the training set, and the performance of the model is evaluated using the validation set, so that the model can accurately predict the temperature threshold and the ultrasonic signal abnormality threshold.

[0012] S600, fault diagnosis: During the operation of the inverter, the data set collected, preprocessed and fused in real time is input into the trained adaptive threshold learning model. The model calculates the corresponding temperature threshold and ultrasonic signal abnormality threshold based on the current data, and compares the temperature part of the real-time fused data with the temperature threshold. At the same time, the ultrasonic part is compared with the ultrasonic signal abnormality threshold. If the temperature data exceeds the temperature threshold or the ultrasonic data exceeds the ultrasonic signal abnormality threshold, it is determined that the inverter has a fault. When it is determined that a fault exists, the characteristics of the fused data are further analyzed, and the fault type and location as well as the cause of the frequent start and stop are determined in combination with the inverter model and mechanism.

[0013] Furthermore, in the S300, in the data processing, the influence of the inverter surface temperature measurement is subjected to temperature compensation processing by a linear compensation model, and the formula is: T B (x,y,t)=T(x,y,t)+a×(T env (t)-T ref ), where a is the compensation coefficient, T ref is the reference ambient temperature, and T(x,y,t) is the temperature at coordinate (x,y) on the inverter surface detected by the infrared thermal imager at time t.

[0014] Furthermore, in the S400, the temperature anomaly data detected by infrared thermal imaging is quantified by the temperature anomaly coefficient in the data fusion to create a temperature anomaly association model, and the temperature anomaly coefficient α(x, y, t) of each point is calculated, and the formula is: Among them, β is the time adjustment coefficient related to the inverter start and stop frequency, t0 is the last start time of the inverter, i and j are index variables used to traverse the coordinates of all sampling points on the inverter surface, t represents time, T avg (t) is the average temperature of the entire inverter surface at time t, and the calculation formula is:

[0015] Furthermore, in the S400, in the data fusion, the structural abnormality index is calculated to quantify the structural abnormality detected by ultrasound to create a structural abnormality association model, assuming that S(f,t) is the frequency spectrum of the ultrasonic signal received at time t, and the frequency range is [f min ,fmax ], the calculation formula of the structural abnormality index γ(t) is: Among them, S ref (f) is the reference ultrasonic spectrum when the inverter is operating normally, and λ is the spectrum adjustment coefficient related to the start and stop of the inverter.

[0016] Furthermore, in the S400, the weight adjustment coefficient μ in the data fusion is adjusted according to the service life of the inverter, and the value of μ increases by 0.05-0.1 for every additional year of service life.

[0017] Furthermore, in S400, the intermediate threshold α in data fusion mid According to the statistical analysis of the temperature anomaly coefficient under normal operation and fault conditions of the inverter, a large amount of temperature anomaly coefficient data of the inverter under normal operation is collected and its average value is calculated as μ normal , collect the temperature anomaly coefficient data under known fault conditions, and also calculate the average value as μ fault , the calculation formula is:

[0018] Furthermore, in the S500, an adaptive temperature threshold learning model is constructed by using an adaptive temperature threshold formula in the adaptive threshold learning, and the formula is: Among them, F history (t) represents the time series of historical fusion fault indication values, F normal (t) represents the time series of the historical fusion fault indication value when the inverter is operating normally, η represents the temperature threshold learning rate, k represents the time window size, α th (t-1) is the adaptive temperature threshold at the previous moment. The initial value is determined according to the design specifications and experience of the inverter. avg (i) represents the average temperature anomaly coefficient of the entire inverter surface at time i, t represents time, and the calculation formula is: Where α(x,y,i) is the temperature anomaly coefficient at coordinate (x,y) at time i, N is the total number of temperature sampling points, and t represents time.

[0019] Furthermore, in the S500, a structural threshold learning model is constructed by using an adaptive structural threshold formula in the adaptive threshold learning, and the formula is: Among them, F history (t) represents the time series of historical fusion fault indication values, F normal (t) represents the time series of the historical fusion fault indication value when the inverter is operating normally, ρ represents the structural threshold learning rate, k represents the time window size, and γ th The initial value (t-1) is determined according to the characteristics and empirical values ​​of the ultrasonic signal under the normal state of the inverter structure, γ(i) is the structural abnormality index at time i, and t represents time.

[0020] Furthermore, in S500, the temperature threshold learning rate η and the structure threshold learning rate ρ in the adaptive threshold learning are dynamically adjusted according to the change rate of the fused data. The greater the change rate of the fused data, the greater the learning rate. The maximum value of the learning rate does not exceed 0.1, and the minimum value is not less than 0.001.

[0021] Furthermore, in the S500, fault type analysis and determination in adaptive threshold learning, the temperature data significantly exceeds the threshold and is concentrated in a specific power module area. Combined with the physical model of the inverter and the fault mechanism analysis, it is found that the power module has an electrical fault such as overload or short circuit. The ultrasonic data is abnormal and shows that the distance change between specific structural components exceeds the normal range. It is found that the component has a mechanical fault such as looseness or displacement. The abnormal conditions of the temperature and ultrasonic data and their association with different parts of the inverter are comprehensively considered to accurately determine the fault type and location.

[0022] Compared with the prior art, this test method based on the frequent start-stop phenomenon of the inverter has the following beneficial effects:

[0023] 1. The present invention introduces a data fusion algorithm to organically integrate the temperature data collected by the infrared thermal imager and the structural data obtained by the ultrasonic detection probe. By accurately calculating the temperature data fusion weight and the ultrasonic data fusion weight, and training and optimizing the correlation model based on the inverter physical model and fault mechanism, a comprehensive and accurate diagnosis of inverter faults is achieved. Compared with traditional single sensor detection, this method can more sensitively capture the temperature changes on the surface of the inverter and the abnormal conditions of the internal structure, effectively improving the accuracy and reliability of fault detection.

[0024] 2. The present invention avoids the installation difficulties, safety risks and data errors caused by the direct contact of sensors with the inverter in the traditional method by adopting non-contact infrared thermal imagers and ultrasonic detection equipment. At the same time, by constructing an adaptive threshold learning model, the temperature threshold and ultrasonic signal abnormality threshold are learned and predicted from the historical database, so that the model can dynamically adjust the threshold according to the real-time operating status of the inverter, thereby realizing accurate identification of inverter faults under different working conditions. It not only improves the practicability and flexibility of the test, but also reduces the false alarm and missed alarm rates caused by improper threshold setting, providing a strong guarantee for the stable operation of the inverter.

[0025] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 It is a flow chart of a testing method based on the frequent start and stop phenomenon of the inverter;

[0028] Figure 2 Flowchart of the data fusion steps. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] Embodiment 1:

[0031] Inverter testing in industrial plants

[0032] Test preparation (S100), there is an inverter for driving a large motor in an industrial plant, model XYZ-500, with a rated power of P rated The structure is modular. There have been several frequent start-stop failures for unknown reasons. The installation environment temperature fluctuates greatly, reaching as high as Winter minimum According to the inverter characteristics and environment, a resolution of R IR Pixel and temperature measurement range is to Accuracy: ±ΔT IR The infrared thermal imager was calibrated and the transmission frequency of the ultrasonic detection equipment was selected as f US The beam angle is θ US The infrared thermal imager scan path is planned to start from the upper left corner of the inverter and scan horizontally to the lower right corner line by line. The scanning frequency is set to Make sure the ultrasonic detection probe is located at the four corners of the inverter housing. The signal parameter is the transmission pulse width. Sampling frequency The data collection time interval is

[0033] In the data collection stage (S200), an infrared thermal imager and an ultrasonic detection probe are installed around the inverter. The infrared thermal imager scans the surface of the inverter according to a predetermined path and frequency. The collected temperature data is transmitted to the data processing center in real time. The ultrasonic detection probe transmits and receives ultrasonic signals, and the signals are amplified by the built-in amplifier circuit. US times, bandpass filter filtering (passband frequency is ), analog-to-digital converter digital processing (resolution b US After that, the data is transmitted to the data processing center.

[0034] In the data preprocessing stage (S300), the infrared thermal imaging temperature data is quality checked. If the temperature values ​​of individual pixels are found to be significantly deviated from the surrounding points, they are judged as abnormal values ​​and removed. Assuming that the current ambient temperature is T env , the reference ambient temperature is set as T ref The compensation coefficient is determined to be a through testing. The surface temperature measurement of the inverter is compensated by the linear compensation model. The formula is: T B (x,y,t)=T(x,y,t)+a×(T env (t)-T ref ), where T B (x, y, t) is the compensated temperature, T(x, y, t) is the temperature at the coordinate (x, y) on the inverter surface detected by the infrared thermal imager at time t, and the amplitude A is extracted from the ultrasonic detection data. amp , time delay t delay , frequency f freq The characteristics are normalized to 0-1, the time delay characteristic value is normalized to -1-1, and the frequency characteristic value is normalized to the frequency spectrum range. Normalize.

[0035] In the data fusion stage (S400), the physical model and previous fault mechanisms of the inverter are analyzed, and it is found that the temperature change is closely related to the connection structure change of the internal power module. By collecting a large amount of historical fault data (including N f Secondary power module failure and N n Normal operation data) and normal operation data are used to train and optimize the correlation model, determine the parameters in the model, and assume that at a certain time t, the average temperature anomaly coefficient α of the entire inverter surface avg (t), calculate the temperature anomaly coefficient α(x,y,t) of each point, the formula is: Where β is the time adjustment coefficient related to the inverter start and stop frequency, t0 is the last start time of the inverter, i and j are index variables used to traverse the coordinates of all sampling points on the inverter surface, T avg(t) is the average temperature of the entire inverter surface, and the calculation formula is: N is the total number of temperature sampling points, which is calculated through ultrasonic detection data. The calculation formula is: Where S red (f) is the reference ultrasonic spectrum when the inverter is operating normally, λ is the spectrum adjustment coefficient related to the start and stop of the inverter, and the temperature data fusion weight is calculated as follows: Then the ultrasonic data fusion weight w S (t) = 1-w T (t), assuming the fused fault indication value is F temp-struc (t), calculated according to the fusion formula, the formula is: Where γ(t) is the structural abnormality index at that moment, and the intermediate threshold α mid According to the statistical analysis of the temperature anomaly coefficient under normal operation and fault conditions of the inverter, a large amount of temperature anomaly coefficient data of the inverter under normal operation is collected and its average value is calculated as μ normal , collect the temperature anomaly coefficient data under known fault conditions, and also calculate the average value as μ fault , the calculation formula is: The weight adjustment coefficient μ is adjusted according to the service life Y of the inverter. When the service life increases by 1 year, the value of μ increases by Δμ.

[0036] In the adaptive threshold learning stage (S500), the inverter is obtained from the historical database under different working conditions such as different loads L and different ambient temperatures T env The total number of detection data is N total The data are annotated in detail, including normal operation, power module failure, and structural looseness failure. The annotated historical data are sorted according to r train :r val The ratio is divided into training set and validation set. The adaptive temperature threshold learning model is constructed by the adaptive temperature threshold formula, which is: where F history represents the time series of historical fusion fault indication values, F normal (t) represents the time series of the historical fusion fault indication value when the inverter is operating normally, η represents the temperature threshold learning rate, k represents the time window size, α th (t-1) is the adaptive temperature threshold at the previous moment. The initial value is determined according to the design specifications and experience of the inverter. avg (i) represents the average temperature anomaly coefficient of the entire inverter surface at time i. The structural threshold model is constructed by the adaptive structural threshold formula, which is: Where ρ represents the structural threshold learning rate, γ thThe initial value (t-1) is determined according to the characteristics and empirical values ​​of the ultrasonic signal under the normal state of the inverter structure. γ(i) is the structural abnormality index at time i. The temperature threshold learning rate η and the structure threshold learning rate ρ are determined according to the change rate r of the fusion data. change Dynamic adjustment: the greater the rate of change of fusion data, the greater the learning rate, but the maximum value of the learning rate does not exceed η max , the minimum value is not less than η min , the input of the model is the fused data set, and the output is the temperature threshold α th (t) and ultrasonic signal abnormal threshold γ th (t), use the training set to train the model, after n iter After iterations, the performance of the model was evaluated using the validation set, and it was found that the model accuracy reached A acc , can accurately predict the temperature threshold and ultrasonic signal abnormality threshold.

[0037] In the fault diagnosis stage (S600), during the operation of the inverter, the real-time collected, pre-processed and fused data set is input into the trained adaptive threshold learning model. The model calculates the corresponding temperature threshold and ultrasonic signal abnormality threshold according to the current data. The average temperature abnormality coefficient calculated by the temperature part of the real-time fusion data is α real (t), exceeds the currently calculated temperature threshold α th (t); the structural abnormality index calculated from the ultrasound data is γ real (t), exceeds the ultrasonic signal abnormal threshold γ th (t), it is determined that the inverter is faulty. It is found that the temperature anomaly is mainly concentrated in a power module area of ​​the inverter. Combined with the inverter model and mechanism, it is determined that the power module has an electrical fault such as overload or short circuit, which may be the cause of the frequent start and stop of the inverter. The maintenance personnel conduct detailed inspection and repair of the power module or related components based on the diagnosis results.

[0038] In summary, when inspecting the XYZ-500 inverter used to drive large motors in industrial plants, first select and calibrate the infrared thermal imager and ultrasonic detection equipment according to its characteristics and environment, plan the parameters of the detection path, collect data according to the plan and transmit and process them, perform quality inspection, remove outliers and perform temperature compensation on the infrared thermal imaging data, extract features from the ultrasonic data and normalize them, determine the weights and fuse the data through the analysis model in the fusion stage, use historical data to train the model to obtain an accurate threshold in the adaptive threshold learning stage, input real-time data into the model to compare the threshold to determine the fault in the fault diagnosis stage, and combine the data characteristics with the inverter mechanism to determine the fault type and location, such as electrical faults in the power module or mechanical faults in the structural components, so as to provide a basis for maintenance personnel to solve the problem of frequent start and stop of the inverter.

[0039] Embodiment 2:

[0040] Inverter testing in photovoltaic power plants

[0041] Test preparation (S100), there are a number of inverters of model ABC-300 in the photovoltaic power station, which are used to convert the DC power generated by the photovoltaic panels into AC power and feed it into the power grid. These inverters are compact in structure, and the heat dissipation method is natural air cooling. They are installed in outdoor open-air environments, where the ambient temperature varies widely and is affected by direct sunlight. We selected an infrared thermal imager suitable for outdoor use, which has dustproof and waterproof functions and a resolution of R′ IR Pixels (set to m'×n'), the temperature measurement range is to Accuracy: ±ΔT IR′ , and calibrate, the ultrasonic testing equipment selects a probe suitable for detecting small structures, and the transmission frequency is f′ US , the beam angle is θ′ US , the infrared thermal imager scanning path is planned to perform a circular scan around the inverter, and the scanning period is The ultrasonic detection probe is installed in the middle of the upper and lower planes of the inverter. The signal parameter is the transmission pulse width. Sampling frequency The data collection time interval is

[0042] Data collection (S200), after the detection equipment is installed, the infrared thermal imager scans the inverter surface according to the path and frequency, collects temperature data and transmits it, and the ultrasonic detection probe works to transmit and receive ultrasonic signals. The signals are processed (amplified A′ US After being multiplied, filtered and digitized, it is transmitted to the data processing center.

[0043] Data processing (S300), check the infrared thermal imaging temperature data, remove abnormal values, and calculate the current ambient temperature T' env and reference ambient temperature T′ ref And the compensation coefficient a′ (determined by experiment) is used for temperature compensation. The influence of the inverter surface temperature measurement is processed by the linear compensation model. The formula is: T′ B (x,y,t)=T′(x,y,t)+a′×(T′ env (t)-T′ ref ), extract the amplitude, time delay and frequency features of ultrasonic detection data and normalize them.

[0044] Data fusion (S400) analyzes the physical model and fault mechanism of the inverter, determines the relationship between temperature and the change of the heat dissipation structure of internal electronic components, and uses N′f Historical fault data (including heat sink fault, capacitor fault) and N′ n The normal operation data is used to train the optimized correlation model. Assume that at time t, the average temperature anomaly coefficient a′ avg (t), calculate the temperature anomaly coefficient α(x,y,t) of each point, the formula is: Where β′ is the time adjustment coefficient related to the inverter start and stop frequency, t ′ 0 is the last startup time of the inverter, i and j are index variables used to traverse the coordinates of all sampling points on the inverter surface, N′ is the total number of temperature sampling points, and the intermediate threshold α ′ mid The weight adjustment coefficient μ′ is determined based on the statistical analysis of the temperature anomaly coefficient under normal operation and fault conditions of the inverter. It is adjusted according to the service life Y′ of the inverter. For every additional year of service life, the value of μ′ increases by Δμ′. The temperature data fusion weight is calculated. The ultrasonic data fusion weight is:

[0053] w′ S (t) = 1 - w′ T (t), and calculate the fused fault indication value according to the fusion formula, which is: '

[0054] Where γ(t) is the structural abnormality index at that moment, which is calculated by ultrasonic detection data. The calculation formula is:

[0055] Adaptive threshold learning (S500) obtains the inverter's performance under different light intensity I and different ambient temperature T' from the historical database. env N' under working conditions total Group test data, press r' after marking train :r′ val Divide the training set and validation set, build an adaptive threshold learning model, and achieve an accuracy rate of A after training. acc ,The adaptive temperature threshold learning model is constructed by the adaptive temperature threshold, and the formula is: where F′ history (t) represents the time series of historical fusion fault indication values, F′ normal (t) represents the time series of the historical fusion fault indication value when the inverter is operating normally, η′ represents the temperature threshold learning rate, k′ represents the time window size, and α′ th (t-1) is the adaptive temperature threshold at the previous moment. The initial value is determined according to the design specifications and experience of the inverter. avg(i) represents the average temperature anomaly coefficient of the entire inverter surface at time i. The structural threshold model is constructed by the adaptive structural threshold formula, which is: Where ρ′ represents the structural threshold learning rate, γ′ th The initial value (t-1) is determined according to the characteristics and empirical values ​​of the ultrasonic signal under the normal state of the inverter structure. γ′(i) is the structural abnormality index at time i. The temperature threshold learning rate η′ and the structural threshold learning rate ρ′ are determined according to the change rate r′ of the fused data. change Dynamic adjustment: the greater the rate of change of fusion data, the greater the learning rate, but the maximum value of the learning rate does not exceed η max , the minimum value is not less than the minimum value is not less than η′ min .

[0074] During the operation of the inverter, the data set collected, preprocessed and fused in real time is input into the trained adaptive threshold learning model. The model calculates the corresponding temperature threshold α′ according to the current data. th (t real ) and the abnormal threshold value γ′ of ultrasonic signal fh (t real ), assuming that the average temperature anomaly coefficient calculated from the temperature part of the real-time fusion data is α′ real (t real ), the abnormal index of the result calculated from the ultrasonic data is γ′ real (t real ), if α′ real (t real )>α′ th (t real ) or γ′ real (t real )>γ′ th (t real ), the inverter is judged to have a fault, and the characteristics of the fused data are further analyzed. If it is found that the temperature anomaly is mainly concentrated near the heat sink, and the ultrasonic data shows that the heat sink fixing structure is abnormal, combined with the inverter model and mechanism, it is judged that the heat sink has poor heat dissipation and the fixing structure is loose, which may be the cause of the frequent start and stop of the inverter. The operation and maintenance personnel will conduct a detailed inspection of the heat sink and its fixing structure, clean the dust, and tighten the loose parts. If the temperature anomaly is concentrated in the capacitor electronic component area, and the ultrasonic data does not show obvious structural problems, the capacitor component has an electrical fault, and the electronic component needs to be further tested for electrical performance and repaired. By comprehensively analyzing the abnormal conditions of the temperature and ultrasonic data and their association with different parts of the inverter, the fault type and location can be accurately determined so that targeted measures can be taken to restore the normal operation of the inverter.

[0083] In summary, for the detection of ABC-300 inverters in photovoltaic power stations, the detection equipment is selected and the detection plan is planned according to its outdoor environment and its own structural characteristics. Data collection is carried out according to the set path and parameters. After collection, the infrared thermal imaging and ultrasonic data are processed separately. In the data fusion stage, the weights are determined by combining the physical model and historical data to complete the fusion. In the adaptive threshold learning stage, the data set is divided to train the model to obtain the threshold. During fault diagnosis, the real-time data is compared with the threshold calculated by the model to determine the fault. If the temperature or ultrasonic data is abnormal, the data characteristics are further analyzed. For example, if the temperature and structural data near the heat sink are abnormal, it is judged that there are problems with the heat dissipation and fixed structure, or the electrical fault is judged based on the abnormal temperature in the electronic component area. In this way, the fault type and location are determined to help operation and maintenance personnel implement targeted maintenance to ensure the normal operation of the inverter and solve the frequent start and stop conditions.

[0084] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A testing method based on the frequent start-stop phenomenon of the inverter, characterized in that: The specific steps of this test method are: S100, test preparation: collect the inverter model, parameters, structure, fault history and installation environment information, select and calibrate the infrared thermal imager and ultrasonic detection equipment according to the inverter characteristics and environment, and plan the infrared thermal imager scanning path, frequency and ultrasonic detection probe position, signal parameters and data collection time interval according to the inverter structure and fault-prone parts; S200, data acquisition: infrared thermal imagers and ultrasonic detection probes are installed around the inverter. The surface temperature data of the inverter is collected by scanning the infrared thermal imager according to the predetermined path and frequency, and transmitted to the data processing center. The ultrasonic detection probe transmits and receives ultrasonic signals, and the built-in signal processing circuit is used to amplify, filter, and digitize the signals. After processing, the data is transmitted to the data processing center; S300, data processing: perform quality inspection on infrared thermal imaging temperature data, remove abnormal values, perform temperature compensation processing based on the impact of ambient temperature on inverter surface temperature measurement, extract amplitude, time delay, and frequency features from ultrasonic detection data and perform normalization; S400, data fusion: According to the inverter physical model and fault mechanism, the intrinsic relationship between temperature change and internal structure change is analyzed, the correlation model is trained and optimized through historical fault data and normal operation data, the parameters in the model are determined, and the fusion weights are assigned to the temperature data and ultrasonic data according to the correlation model. The temperature data fusion weight w is set T (t) is calculated as: α avg (t) is the average temperature anomaly coefficient of the entire inverter surface at time t, α mid is an intermediate threshold, μ is the weight adjustment coefficient, t represents time, and the ultrasonic data fusion weight w S (t) The calculation formula is: S (t) = 1-w T (t), use appropriate data fusion algorithm to fuse the preprocessed temperature data and ultrasonic data, let F temp-struc (t) is the fused fault indication value, and the fusion formula is: Where α(x,y,t) represents the temperature anomaly coefficient, and t represents time; S500, adaptive threshold learning: obtain the detection data of the inverter under different working conditions from the historical database, annotate these data in detail, divide the annotated historical data into a training set and a validation set, and build an adaptive threshold learning model. The input of the model is the fused data set, and the output is the temperature threshold and the ultrasonic signal abnormality threshold. The model is trained using the training set, and the performance of the model is evaluated using the validation set, so that the model can accurately predict the temperature threshold and the ultrasonic signal abnormality threshold. S600, fault diagnosis: During the operation of the inverter, the data set collected, preprocessed and fused in real time is input into the trained adaptive threshold learning model. The model calculates the corresponding temperature threshold and ultrasonic signal abnormality threshold based on the current data, and compares the temperature part of the real-time fused data with the temperature threshold. At the same time, the ultrasonic part is compared with the ultrasonic signal abnormality threshold. If the temperature data exceeds the temperature threshold or the ultrasonic data exceeds the ultrasonic signal abnormality threshold, it is determined that the inverter has a fault. When it is determined that a fault exists, the characteristics of the fused data are further analyzed, and the fault type and location as well as the cause of the frequent start and stop are determined in combination with the inverter model and mechanism.

2. A test method based on the frequent start and stop phenomenon of the inverter according to claim 1, characterized in that: In the step S300, in the data processing, the influence of the inverter surface temperature measurement is subjected to temperature compensation processing by using a linear compensation model, and the formula is: T B (x,y,t)=T(x,y,t)+a×(T env (t)-T ref ), where a is the compensation coefficient, T ref is the reference ambient temperature, and T(x,y,t) is the temperature at coordinate (x,y) on the inverter surface detected by the infrared thermal imager at time t.

3. A test method based on the frequent start and stop phenomenon of the inverter according to claim 1, characterized in that: In the S400, the temperature anomaly data detected by infrared thermal imaging is quantified by the temperature anomaly coefficient in the data fusion to create a temperature anomaly association model, and the temperature anomaly coefficient α(x, y, t) of each point is calculated. The formula is: Among them, β is the time adjustment coefficient related to the inverter start and stop frequency, t0 is the last start time of the inverter, i and j are index variables used to traverse the coordinates of all sampling points on the inverter surface, t represents time, T avg (t) is the average temperature of the entire inverter surface at time t, and the calculation formula is:

4. A test method based on the frequent start-stop phenomenon of the inverter according to claim 1, characterized in that: In the S400, in the data fusion, the structural abnormality index is calculated to quantify the structural abnormality detected by ultrasound to create a structural abnormality association model. Suppose S(f,t) is the frequency spectrum of the ultrasonic signal received at time t, and the frequency range is [f min ,f max ], the calculation formula of the structural abnormality index γ(t) is: Among them, S ref (f) is the reference ultrasonic spectrum when the inverter is operating normally, and λ is the spectrum adjustment coefficient related to the start and stop of the inverter.

5. A test method based on the frequent start and stop phenomenon of the inverter according to claim 1, characterized in that: In the above S400, the weight adjustment coefficient μ in the data fusion is adjusted according to the service life of the inverter. For every additional year of service life, the value of μ increases by 0.05-0.

1.

6. A test method based on the frequent start and stop phenomenon of the inverter according to claim 1, characterized in that: S400, the intermediate threshold α in data fusion mid According to the statistical analysis of the temperature anomaly coefficient under normal operation and fault conditions of the inverter, a large amount of temperature anomaly coefficient data of the inverter under normal operation is collected and its average value is calculated as μ normal , collect the temperature anomaly coefficient data under known fault conditions, and also calculate the average value as μ fault , the calculation formula is:

7. A test method based on the frequent start and stop phenomenon of the inverter according to claim 1, characterized in that: In the above S500, an adaptive temperature threshold learning model is constructed by using an adaptive temperature threshold formula in the adaptive threshold learning, and the formula is: Among them, F history (t) represents the time series of historical fusion fault indication values, F normal (t) represents the time series of the historical fusion fault indication value when the inverter is operating normally, η represents the temperature threshold learning rate, k represents the time window size, α th (t-1) is the adaptive temperature threshold at the previous moment. The initial value is determined according to the design specifications and experience of the inverter. avg (i) represents the average temperature anomaly coefficient of the entire inverter surface at time i, t represents time, and the calculation formula is: Where α(x,y,i) is the temperature anomaly coefficient at coordinate (x,y) at time i, N is the total number of temperature sampling points, and t represents time.

8. A test method based on the frequent start and stop phenomenon of an inverter according to claim 1, characterized in that: In the above S500, a structural threshold learning model is constructed by using an adaptive structural threshold formula in the adaptive threshold learning, and the formula is: Among them, F history (t) represents the time series of historical fusion fault indication values, F normal (t) represents the time series of the historical fusion fault indication value when the inverter is operating normally, ρ represents the structural threshold learning rate, k represents the time window size, and γ th The initial value (t-1) is determined according to the characteristics and empirical values ​​of the ultrasonic signal under the normal state of the inverter structure, γ(i) is the structural abnormality index at time i, and t represents time.

9. A test method based on the frequent start and stop phenomenon of an inverter according to claim 1, characterized in that: In the S500, the temperature threshold learning rate η and the structure threshold learning rate ρ in the adaptive threshold learning are dynamically adjusted according to the change rate of the fused data. The greater the change rate of the fused data, the corresponding increase in the learning rate. The maximum value of the learning rate does not exceed 0.1, and the minimum value is not less than 0.

001.

10. A test method based on the frequent start and stop phenomenon of an inverter according to claim 1, characterized in that: In the S500, the fault type analysis and determination in the adaptive threshold learning, the temperature data significantly exceeds the threshold and is concentrated in the specific power module area. Combined with the physical model of the inverter and the fault mechanism analysis, it is found that the power module has an electrical fault such as overload or short circuit. The ultrasonic data is abnormal and shows that the distance change between specific structural components exceeds the normal range. It is a mechanical fault such as looseness or displacement of the component. The abnormal conditions of the temperature and ultrasonic data and their association with different parts of the inverter are comprehensively considered to accurately determine the fault type and location.

Images