A power unit detection system for high voltage inverter

Abstract

A power unit detection system for a high-voltage frequency converter relates to the technical field of high-voltage frequency converters. The system analyzes historical multi-source operation data and historical external scene data of a number of high-voltage frequency converters to obtain fault indicators in the multi-source operation data and three-dimensional vulnerability curves corresponding to various types of scene indicators; constructs critical feature sets of various types of operation indicators of the number of high-voltage frequency converters; constructs a risk warning data set, performs feature matching on the multi-source operation data and external scene data of the number of high-voltage frequency converters and the risk warning data set, and determines the detection methods of the number of high-voltage frequency converters according to the feature matching results; performs early warning detection on the number of high-voltage frequency converters according to the critical feature sets of various types of operation indicators in the risk warning data set; performs difference detection on the number of high-voltage frequency converters, and feeds back the high-voltage frequency converters that do not meet the requirements to the early warning detection module, so as to effectively improve the management and maintenance level of the high-voltage frequency converters.

Description

Technical Field

[0001] The invention relates to the technical field of high-voltage frequency converters, and in particular to a power unit detection system for high-voltage frequency converters. Background Art

[0002] The prior art CN115774159A "High-voltage inverter power unit fault detection system" monitors the environment where the high-voltage inverter is located through a real-scene monitoring unit to obtain environmental information, and simultaneously uses a data acquisition unit to obtain abnormal records of the high-voltage inverter. The abnormal records include abnormal time points and abnormal corresponding values, and the abnormal corresponding values ​​are the environmental impact assessment values ​​corresponding to the abnormal time points; then the data acquisition unit is used to transmit the abnormal records to the correlation analysis unit, and the correlation analysis unit is used to perform correlation analysis on the abnormal records, so as to determine key sections, idle sections, and ordinary sections; then different cycles are selected for different time periods, and the inspection cycle is determined in combination with real-time environmental impact assessment values.

[0003] The prior art CN112924797A "A fault location method and system for a high-voltage frequency conversion system" detects whether the high-voltage frequency converter is in a fault state. When it is detected that the frequency converter is in a fault state, the main control unit of the frequency converter sends data one by one to the three-phase power unit of the frequency converter through serial communication, and uses the fault location module set in the three-phase power unit to determine whether the corresponding three-phase power unit in the high-voltage frequency converter is faulty, and locate the source of the fault. In its corresponding positioning system, a power unit status real-time update function, a power unit waveform abnormality monitoring function and a communication fault location function are set.

[0004] Although the existing power unit detection system for high-voltage inverters has many advantages, there are still some problems that need to be solved in practical applications. These problems include that high-voltage inverters will generate a large amount of multi-source data. How to extract meaningful features from massive data is a technical problem. At the same time, the existing power unit detection system for high-voltage inverters often uses threshold monitoring technology for various indicators of the power unit. This is a static analysis method that fails to reflect the advantages of the online monitoring system, and the alarm threshold is difficult to set. There is a lack of sufficient basis. There are many factors that affect the failure of the power unit of the high-voltage inverter, such as the operating environment, service age, etc.; the individual differences caused by multiple influencing factors make it difficult to find universal standards or samples to identify the change in the operating state of the power unit of the high-voltage inverter, and the early faults of the power unit of the high-voltage inverter before the normal state changes to the fault state are often not easy to detect, because the signs of power unit failure of the high-voltage inverter may be weak or difficult to identify, which may lead to the inability to take preventive measures before the failure matures. Summary of the invention

[0005] In order to solve the above technical problems, the object of the present invention is to provide a power unit detection system for a high-voltage inverter, including a monitoring center, wherein the monitoring center is communicatively connected with a data acquisition module, a fault index analysis module, a scenario index analysis module, a critical feature analysis module, a risk warning module, a prediction module, an early warning detection module and a difference detection module;

[0006] The data acquisition module is connected to a plurality of high-voltage inverters in a distributed manner through an Internet of Things node, and is used to collect multi-source operation data and external scene data of a plurality of high-voltage inverters, mark the collection time, and set the collection cycle;

[0007] The fault index analysis module is used to analyze the historical multi-source operation data of a number of high-voltage inverters to obtain fault indicators in the multi-source operation data;

[0008] The scenario indicator analysis module is used to analyze the historical external scenario data of a number of high-voltage inverters to obtain three-dimensional vulnerability curves corresponding to various types of scenario indicators of the number of high-voltage inverters;

[0009] The critical feature analysis module is used to analyze the multi-source operation data before various types of operation indicators of the plurality of high-voltage inverters are marked as fault indicators, and to construct a critical feature set of various types of operation indicators of the plurality of high-voltage inverters;

[0010] The risk warning module is used to construct a risk warning data set, perform feature matching on the multi-source operation data and external scenario data of a plurality of high-voltage inverters with the risk warning data set, and determine the detection methods of the plurality of high-voltage inverters according to the feature matching results;

[0011] The prediction module is used to obtain the predicted numerical time series of various types of operating indicators of a number of high-voltage inverters;

[0012] The early warning detection module is used to perform early warning detection on a number of high-voltage inverters according to the critical feature set of each type of operating indicators in the risk warning data set;

[0013] The difference detection module is used to perform difference detection on a plurality of high-voltage frequency converters, and feed back high-voltage frequency converters that do not meet the requirements to the early warning detection module.

[0014] Furthermore, the scenario indicator analysis module analyzes historical external scenario data of a plurality of high-voltage inverters to obtain a three-dimensional vulnerability curve corresponding to each type of scenario indicator of the plurality of high-voltage inverters, including:

[0015] A statistical analysis is performed on the numerical time series corresponding to each type of scenario indicator in several historical collection cycles of the high-voltage inverter to obtain the average value and fluctuation coefficient corresponding to each type of scenario indicator in several historical collection cycles; a statistical analysis is performed on the average value, fluctuation coefficient and fault index corresponding to each type of scenario indicator in several historical collection cycles to obtain a three-dimensional vulnerability curve corresponding to each type of scenario indicator, wherein the three coordinate axes of the three-dimensional vulnerability curve corresponding to each type of scenario indicator respectively correspond to the average value, fluctuation coefficient and fault indicator generation probability.

[0016] Furthermore, the critical feature analysis module analyzes the multi-source operation data before various types of operation indicators of the plurality of high-voltage inverters are marked as fault indicators, and the process of constructing the critical feature set of various types of operation indicators of the plurality of high-voltage inverters includes:

[0017] Extract the timestamp when the type of operating indicator is marked as a fault indicator and the historical collection period to which the type of operating indicator belongs when the type of operating indicator is marked as a fault indicator, mark the numerical time series sequence corresponding to each type of operating indicator between the start timestamp of the historical collection period and the timestamp as a critical data set, and mark the numerical time series sequence corresponding to each type of operating indicator in the historical collection period where no fault indicator is generated as a normal data set;

[0018] Obtain the fluctuation coefficient range of each type of operating indicator in the critical data set and the normal data set, mark the type of operating indicator whose fluctuation coefficient range in the critical data set is consistent with that in the normal data set as a key type operating indicator, and mark the type of operating indicator whose fluctuation coefficient range in the critical data set is inconsistent with that in the normal data set as a key type operating indicator;

[0019] Obtain the correlation coefficients between each key type of operating indicators in the critical data set and the correlation coefficients between each key type of operating indicator and each non-key type of operating indicator, obtain the predicted numerical time series sequence of the key type of operating indicators, perform difference detection on the key type of operating indicators in the critical data set according to the predicted numerical time series sequence of the key type of operating indicators, obtain the difference coefficients of the key type of operating indicators, and construct a critical feature set of type operating indicators according to the correlation coefficients between each key type of operating indicators, the correlation coefficients between each key type of operating indicator and each non-key type of operating indicators, and the difference coefficients of the key type of operating indicators.

[0020] Furthermore, the risk warning module constructs a risk warning data set, performs feature matching on the multi-source operation data and external scenario data of a plurality of high-voltage inverters with the risk warning data set, and determines the detection methods of a plurality of high-voltage inverters according to the feature matching results, including:

[0021] A risk warning data set is constructed, wherein the risk warning data set includes three-dimensional vulnerability curves corresponding to various types of scenario indicators of a number of high-voltage inverters and critical feature sets of various types of operating indicators of a number of high-voltage inverters, and scenario feature matching is performed between the external scenario data of the number of high-voltage inverters and the risk warning data set to obtain the probability of generating fault indicators of the number of high-voltage inverters, and the probability of generating fault indicators of the number of high-voltage inverters is compared with a preset probability threshold, and early warning detection is performed on high-voltage inverters whose probability of generating fault indicators is greater than the probability threshold, and difference detection is performed on high-voltage inverters whose probability of generating fault indicators is less than or equal to the probability threshold.

[0022] Furthermore, the risk warning module matches the external scene data of a plurality of high-voltage inverters with the risk warning data set for scene features, and the process of obtaining the generation probability of the fault indicators of the plurality of high-voltage inverters includes:

[0023] According to the external scene data of the current acquisition period of the high-voltage inverter, the average value and fluctuation coefficient corresponding to each type of scenario indicator in the current acquisition period are obtained, and the average value and fluctuation coefficient corresponding to each type of scenario indicator are compared with the three-dimensional vulnerability curve corresponding to each type of scenario indicator to obtain the fault indicator generation probability of each type of scenario indicator, and the maximum fault indicator generation probability is screened out from the fault indicators of each type of scenario indicator as the high-voltage inverter fault indicator generation probability.

[0024] Furthermore, the process of the prediction module obtaining the predicted numerical time series of various types of operating indicators of a plurality of high-voltage inverters includes:

[0025] A prediction model is constructed based on deep learning, and the numerical time series sequences corresponding to various types of operating indicators in several historical acquisition cycles of several high-voltage inverters are used as training sets and test sets. The training set is input into the prediction model for training until the loss function training is stable, and the model parameters are saved. The prediction model is tested with the test set until it meets the preset requirements, and the prediction model is output. According to the prediction model, the predicted numerical time series sequences of various types of operating indicators of several high-voltage inverters are output.

[0026] Furthermore, the process of the early warning detection module performing early warning detection on a plurality of high-voltage inverters according to the critical feature set of each type of operating indicator in the risk warning data set includes:

[0027] Obtain the fluctuation coefficient range corresponding to each type of operating indicator in the current acquisition cycle according to the multi-source operating data of the current acquisition cycle of the high-voltage inverter, perform consistency comparison between the fluctuation coefficient range corresponding to each type of operating indicator in the current acquisition cycle and the fluctuation coefficient range corresponding to each type of operating indicator in the normal data set, and judge whether there is a key type of operating indicator in the current acquisition cycle according to the comparison result; if there is a key type of operating indicator in the current acquisition cycle, obtain the correlation coefficient between each key type of operating indicator in the current acquisition cycle, the correlation coefficient between each key type of operating indicator and each non-key type of operating indicator, and the difference coefficient of the key type of operating indicator; perform similarity comparison between the correlation coefficient between each key type of operating indicator, the correlation coefficient between each key type of operating indicator and each non-key type of operating indicator, and the difference coefficient of the key type of operating indicator and the critical feature set of each type of operating indicator, and obtain the similarity of the critical feature set of each type of operating indicator;

[0028] When the similarity of the critical feature set of the type operation indicator is greater than the preset similarity threshold, the type operation indicator is marked as a fault warning operation indicator, and a fault warning signal of the high-voltage inverter is generated according to the fault warning operation indicator;

[0029] If there is no key type of operating indicator in the current collection period or the similarity of the critical feature sets of each type of operating indicator is less than or equal to the preset similarity threshold, the high-voltage inverter is marked as operating normally.

[0030] Furthermore, the difference detection module performs difference detection on a plurality of high-voltage inverters, and feeds back the high-voltage inverters that do not meet the requirements to the early warning detection module, including:

[0031] Obtain predicted numerical time series of various types of operating indicators of the high-voltage inverter as judgment data, compare the numerical time series of various types of operating indicators in the current acquisition cycle of the high-voltage inverter with the judgment data, obtain the difference coefficient of each type of operating indicator, compare the difference coefficient of each type of operating indicator with a preset difference coefficient threshold, and if the difference coefficient of each type of operating indicator is less than or equal to the difference coefficient threshold, mark the high-voltage inverter as a normal operating state;

[0032] If the difference coefficient of the existence type operation indicator is greater than the difference coefficient threshold, an early warning detection tag of the high-voltage inverter is generated and fed back to the early warning detection module. The early warning detection module performs early warning detection on the high-voltage inverter according to the early warning detection tag of the high-voltage inverter.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. Comprehensiveness: The present invention not only focuses on the operating data of the high-voltage inverter itself, but also takes into account external environmental factors, providing more comprehensive fault prediction and early warning capabilities.

[0035] 2. Intelligence: Through machine learning and data analysis technology, the system can automatically identify fault modes and make predictions, improving the accuracy and timeliness of fault detection.

[0036] 3. Preventive maintenance: Through early warning detection, preventive measures can be taken in time to reduce unplanned downtime and improve equipment availability and reliability.

[0037] 4. Data-driven: The system is based on analysis of a large amount of historical data and real-time data, and can continuously optimize its own analytical capabilities and warning accuracy.

[0038] 6. Personalized maintenance: Through difference detection, the system can identify the differences between different high-voltage inverters and provide personalized maintenance suggestions.

[0039] 7. Scalability: The system supports distributed connections and can adapt to different numbers and types of high-voltage inverters, with good scalability and adaptability.

[0040] 8. Reduce costs: By discovering and handling potential faults in advance, reduce equipment maintenance costs and downtime losses, and improve production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a schematic diagram of a power unit detection system for a high-voltage inverter according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0043] like Figure 1 As shown, a power unit detection system for a high-voltage inverter includes a monitoring center, and the monitoring center is communicatively connected to a data acquisition module, a fault index analysis module, a scenario index analysis module, a critical feature analysis module, a risk warning module, a prediction module, an early warning detection module, and a difference detection module;

[0044] The data acquisition module is connected to a plurality of high-voltage inverters in a distributed manner through an Internet of Things node, and is used to collect multi-source operation data and external scene data of a plurality of high-voltage inverters, mark the collection time, and set the collection cycle;

[0045] The fault index analysis module is used to analyze the historical multi-source operation data of a number of high-voltage inverters to obtain fault indicators in the multi-source operation data;

[0046] The scenario indicator analysis module is used to analyze the historical external scenario data of a number of high-voltage inverters to obtain three-dimensional vulnerability curves corresponding to various types of scenario indicators of the number of high-voltage inverters;

[0047] The critical feature analysis module is used to analyze the multi-source operation data before various types of operation indicators of the plurality of high-voltage inverters are marked as fault indicators, and to construct a critical feature set of various types of operation indicators of the plurality of high-voltage inverters;

[0048] The risk warning module is used to construct a risk warning data set, perform feature matching on the multi-source operation data and external scenario data of a plurality of high-voltage inverters with the risk warning data set, and determine the detection methods of the plurality of high-voltage inverters according to the feature matching results;

[0049] The prediction module is used to obtain the predicted numerical time series of various types of operating indicators of a number of high-voltage inverters;

[0050] The early warning detection module is used to perform early warning detection on a number of high-voltage inverters according to the critical feature set of each type of operating indicators in the risk warning data set;

[0051] The difference detection module is used to perform difference detection on a plurality of high-voltage frequency converters, and feed back high-voltage frequency converters that do not meet the requirements to the early warning detection module.

[0052] It should be further explained that, in a specific implementation process, the fault indicator analysis module analyzes the historical multi-source operation data of a number of high-voltage inverters, and the process of obtaining the fault indicators in the multi-source operation data includes:

[0053] Obtain numerical time series sequences corresponding to various types of operating indicators in multi-source operating data of several high-voltage inverters within several historical acquisition cycles and numerical time series sequences corresponding to various types of scenario indicators in external scenario data, and simultaneously obtain threshold ranges corresponding to various types of operating indicators of several high-voltage inverters within several historical acquisition cycles, compare the numerical time series sequences corresponding to various types of operating indicators in the multi-source operating data with the corresponding threshold ranges to obtain deviation rate information of various types of operating indicators, compare the deviation rate information with a preset deviation rate threshold, obtain the accumulated time when the deviation rate information is greater than the deviation rate threshold, and when the accumulated time of a type of operating indicator is greater than the preset time threshold, mark the type of operating indicator as a fault indicator.

[0054] It should be further explained that, in the specific implementation process, the various types of operation indicators in the multi-source operation data include:

[0055] Input voltage (measure the voltage on the input side of the high-voltage inverter to ensure that it meets the design requirements), input current (monitor the input current to check whether there is overcurrent), output voltage (measure the voltage output from the inverter to the motor to ensure that it meets the needs of the motor), output current (monitor the output current to evaluate the motor load), DC bus voltage (measure the voltage of the DC link to ensure that it is stable and meets the requirements), harmonic content (analyze the harmonic components of the output current and voltage to ensure that the harmonic level is within an acceptable range), IGBT temperature (monitor the temperature of key power semiconductor devices such as IGBT to ensure that they are not overheated), capacitor temperature (measure the temperature of the DC link capacitor to prevent overheating and capacitor failure), IGBT turn-on and turn-off time, frequency and duty cycle of the PWM signal, mechanical parameters (including vibration and noise of the power unit);

[0056] The various types of scene indicators in external scene data include:

[0057] Ambient temperature (monitor the temperature of the environment where the inverter is located to ensure that it works within the appropriate temperature range), humidity (monitor the ambient humidity to prevent electrical failures caused by excessive humidity), dust concentration, etc.

[0058] It should be further explained that, in a specific implementation process, the scenario indicator analysis module analyzes the historical external scenario data of a number of high-voltage inverters, and the process of obtaining the three-dimensional vulnerability curves corresponding to various types of scenario indicators of the number of high-voltage inverters includes:

[0059] A statistical analysis is performed on the numerical time series corresponding to each type of scenario indicator in several historical collection cycles of the high-voltage inverter to obtain the average value and fluctuation coefficient corresponding to each type of scenario indicator in several historical collection cycles; a statistical analysis is performed on the average value, fluctuation coefficient and fault index corresponding to each type of scenario indicator in several historical collection cycles to obtain a three-dimensional vulnerability curve corresponding to each type of scenario indicator, wherein the three coordinate axes of the three-dimensional vulnerability curve corresponding to each type of scenario indicator respectively correspond to the average value, fluctuation coefficient and fault indicator generation probability.

[0060] It should be further explained that, in the specific implementation process, the calculation formula for the probability of generating fault indicators in the three-dimensional fragility curve is: ;

[0061] Where P represents the probability of generating fault indicators, Indicates the average value corresponding to the scene indicator, Indicates the volatility coefficient corresponding to the scenario indicator, Indicates the total number of fault indicators generated. Indicates the average value corresponding to the scene indicator and volatility The number of times the fault indicator is generated under the condition;

[0062] Among them, the volatility coefficient The calculation formula is: ;

[0063] in, It represents the value corresponding to the type scenario indicator at the tth moment, and n represents the total number of moments in the numerical time series sequence corresponding to the type scenario indicator.

[0064] It should be further explained that, in a specific implementation process, the critical feature analysis module analyzes multi-source operating data before various types of operating indicators of a number of high-voltage inverters are marked as fault indicators, and the process of constructing a critical feature set of various types of operating indicators of a number of high-voltage inverters includes:

[0065] Extract the timestamp when the type operation indicator is marked as a fault indicator and the historical collection period to which the type operation indicator belongs when it is marked as a fault indicator, obtain the numerical time series sequence corresponding to each type of operation indicator between the start timestamp of the historical collection period and the timestamp, and mark the numerical time series sequence corresponding to each type of operation indicator as a critical data set;

[0066] Obtain a historical collection period in which no fault indicator is generated, and mark the numerical time series corresponding to each type of operating indicator in the historical collection period as a normal data set;

[0067] Obtaining the fluctuation coefficient range of each type of operating indicator in the critical data set and the fluctuation coefficient range of each type of operating indicator in the normal data set, if the fluctuation coefficient range of the type operating indicator in the critical data set is consistent with the fluctuation coefficient range of the type operating indicator in the normal data set, marking the type operating indicator as a key type operating indicator;

[0068] If the fluctuation coefficient range of the type operation indicator in the critical data set is inconsistent with the fluctuation coefficient range of the type operation indicator in the normal data set, the type operation indicator is marked as a non-critical type operation indicator;

[0069] Obtain the correlation coefficients between each key type of operating indicators in the critical data set and the correlation coefficients between each key type of operating indicator and each non-key type of operating indicator, obtain the predicted numerical time series sequence of the key type of operating indicators, perform difference detection on the key type of operating indicators in the critical data set according to the predicted numerical time series sequence of the key type of operating indicators, obtain the difference coefficients of the key type of operating indicators, and construct a critical feature set of type operating indicators according to the correlation coefficients between each key type of operating indicators, the correlation coefficients between each key type of operating indicator and each non-key type of operating indicators, and the difference coefficients of the key type of operating indicators.

[0070] It should be further explained that, in the specific implementation process, the process of obtaining the correlation coefficient between the type operation indicators includes: ;

[0071] in, represents the correlation coefficient between type operation index i and type operation index j, It represents the value of type operation index i at time t. It represents the average value corresponding to the numerical time series of type operation index i. It represents the value of type operation index j at the tth moment. It represents the average value corresponding to the numerical time series of type operation index j, and n represents the total number of moments.

[0072] It should be further explained that, in the specific implementation process, the risk warning module constructs a risk warning data set, performs feature matching on the multi-source operation data and external scenario data of a number of high-voltage inverters with the risk warning data set, and determines the detection methods of a number of high-voltage inverters according to the feature matching results. The process includes:

[0073] A risk warning data set is constructed, wherein the risk warning data set includes three-dimensional vulnerability curves corresponding to various types of scenario indicators of a number of high-voltage inverters and critical feature sets of various types of operating indicators of a number of high-voltage inverters, and scenario feature matching is performed between the external scenario data of the number of high-voltage inverters and the risk warning data set to obtain the probability of generating fault indicators of the number of high-voltage inverters, and the probability of generating fault indicators of the number of high-voltage inverters is compared with a preset probability threshold, and early warning detection is performed on high-voltage inverters whose probability of generating fault indicators is greater than the probability threshold, and difference detection is performed on high-voltage inverters whose probability of generating fault indicators is less than or equal to the probability threshold.

[0074] It should be further explained that, in a specific implementation process, the process of the prediction module obtaining the predicted numerical time series of various types of operating indicators of a plurality of high-voltage inverters includes:

[0075] A prediction model is constructed based on deep learning, and the numerical time series sequences corresponding to various types of operating indicators in several historical acquisition cycles of several high-voltage inverters are used as training sets and test sets. The training set is input into the prediction model for training until the loss function training is stable, and the model parameters are saved. The prediction model is tested with the test set until it meets the preset requirements, and the prediction model is output. According to the prediction model, the predicted numerical time series sequences of various types of operating indicators of several high-voltage inverters are output.

[0076] It should be further explained that, in the specific implementation process, the risk warning module matches the external scene data of several high-voltage inverters with the risk warning data set for scene features, and the process of obtaining the generation probability of the fault indicators of several high-voltage inverters includes:

[0077] According to the external scene data of the current acquisition period of the high-voltage inverter, the average value and fluctuation coefficient corresponding to each type of scene indicator in the current acquisition period are obtained, and the average value and fluctuation coefficient corresponding to each type of scene indicator are compared with the three-dimensional vulnerability curve corresponding to each type of scene indicator, and the Euclidean distance between the average value corresponding to each type of scene indicator and each average value of the three-dimensional vulnerability curve corresponding to each type of scene indicator is obtained. Similarly, the Euclidean distance between the fluctuation coefficient corresponding to each type of scene indicator and each fluctuation coefficient of the three-dimensional vulnerability curve corresponding to each type of scene indicator is obtained, and the Euclidean distance between each average value in the three-dimensional vulnerability curve type corresponding to each type of scene indicator and each type of scene indicator is obtained. The sum of the Euclidean distances corresponding to the fluctuation coefficients is obtained, and the average value and fluctuation coefficient corresponding to the minimum sum of the Euclidean distances in the three-dimensional vulnerability curve type corresponding to each type of scenario indicator are screened out to obtain the fault indicator generation probability corresponding to the average value and fluctuation coefficient in the three-dimensional vulnerability curve type corresponding to each type of scenario indicator, and the fault indicator generation probability corresponding to the average value and fluctuation coefficient in the three-dimensional vulnerability curve type corresponding to each type of scenario indicator is used as the fault indicator generation probability of each type of scenario indicator, and the fault indicator generation probability of each type of scenario indicator is obtained, and the maximum fault indicator generation probability is screened out from the fault indicators of each type of scenario indicator as the high-voltage inverter fault indicator generation probability.

[0078] It should be further explained that, in the specific implementation process, the process of the early warning detection module performing early warning detection on a number of high-voltage inverters according to the critical feature set of each type of operating indicator in the risk warning data set includes:

[0079] According to the multi-source operation data of the current acquisition cycle of the high-voltage inverter, the fluctuation coefficient range corresponding to each type of operation indicator in the current acquisition cycle is obtained, and the fluctuation coefficient range corresponding to each type of operation indicator in the current acquisition cycle is compared with the fluctuation coefficient range corresponding to each type of operation indicator in the normal data set for consistency, and the type of operation indicator whose corresponding fluctuation coefficient range in the current acquisition cycle is inconsistent with the corresponding fluctuation coefficient range in the normal data set is marked as a key type of operation indicator, and the type of operation indicator whose corresponding fluctuation coefficient range in the current acquisition cycle is consistent with the corresponding fluctuation coefficient range in the normal data set is marked as a non-key type of operation indicator;

[0080] If there are key-type operating indicators in the current collection period, then obtain the correlation coefficients between the key-type operating indicators in the current collection period, the correlation coefficients between the key-type operating indicators and the non-key-type operating indicators, and the difference coefficients of the key-type operating indicators, and compare the similarity of the correlation coefficients between the key-type operating indicators, the correlation coefficients between the key-type operating indicators and the non-key-type operating indicators, and the difference coefficients of the key-type operating indicators with the critical feature sets of the various types of operating indicators to obtain the similarity of the critical feature sets of the various types of operating indicators;

[0081] It should be further explained that, in the specific implementation process, the calculation formula for obtaining the similarity of the critical feature set is: in, Indicates similarity, Indicates the correlation coefficient between the key type operating indicator z and the non-key type operating indicator v in the current collection cycle, represents the correlation coefficient between the key type operation index z and the non-key type operation index v in the critical feature set, represents a set of key type operating indicators, and R represents a set of non-key type operating indicators. represents the weight coefficient, Indicates the correlation coefficient between the key type operation indicator z and the key type operation indicator u in the current collection cycle, represents the correlation coefficient between the key type operation index z and the key type operation index u in the critical feature set, Indicates the difference coefficient of the key type operation indicator z in the current collection cycle, Represents the difference coefficient of the key type operation index z in the critical feature set.

[0082] When the similarity of the critical feature set of the type operation indicator is greater than the preset similarity threshold, the type operation indicator is marked as a fault warning operation indicator, and a fault warning signal of the high-voltage inverter is generated according to the fault warning operation indicator;

[0083] If there is no key type of operating indicator in the current collection period or the similarity of the critical feature sets of each type of operating indicator is less than or equal to the preset similarity threshold, the high-voltage inverter is marked as operating normally.

[0084] It should be further explained that, in a specific implementation process, the difference detection module performs difference detection on a plurality of high-voltage inverters, and feeds back the high-voltage inverters that do not meet the requirements to the early warning detection module, including:

[0085] Obtain predicted numerical time series of various types of operating indicators of the high-voltage inverter as judgment data, compare the numerical time series of various types of operating indicators in the current acquisition cycle of the high-voltage inverter with the judgment data, obtain the difference coefficient of each type of operating indicator, compare the difference coefficient of each type of operating indicator with a preset difference coefficient threshold, and if the difference coefficient of each type of operating indicator is less than or equal to the difference coefficient threshold, mark the high-voltage inverter as a normal operating state;

[0086] Among them, the calculation formula for obtaining the difference coefficient of each type of operation index is: ;

[0087] in, represents the coefficient of difference of type operation index i; It represents the value of type operation index i at time t. It represents the predicted value of type operation index i at time t;

[0088] If the difference coefficient of the existence type operation indicator is greater than the difference coefficient threshold, an early warning detection tag of the high-voltage inverter is generated and fed back to the early warning detection module. The early warning detection module performs early warning detection on the high-voltage inverter according to the early warning detection tag of the high-voltage inverter.

[0089] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A power unit detection system for a high voltage frequency converter, characterized in that: It includes a monitoring center, which is communicatively connected to a data acquisition module, a fault indicator analysis module, a scenario indicator analysis module, a critical feature analysis module, a risk warning module, a prediction module, an early warning detection module and a difference detection module; The data acquisition module is connected to a plurality of high-voltage inverters in a distributed manner through an Internet of Things node, and is used to collect multi-source operation data and external scene data of a plurality of high-voltage inverters, mark the collection time, and set the collection cycle; The fault indicator analysis module is used to analyze the historical multi-source operation data of a number of high-voltage inverters to obtain fault indicators in the multi-source operation data; The scenario indicator analysis module is used to analyze the historical external scenario data of a plurality of high-voltage inverters to obtain three-dimensional vulnerability curves corresponding to various types of scenario indicators of the plurality of high-voltage inverters, including: Performing statistical analysis on the numerical time series corresponding to each type of scenario indicator in several historical collection cycles of the high-voltage inverter, obtaining the average value and fluctuation coefficient corresponding to each type of scenario indicator in several historical collection cycles, performing statistical analysis on the average value, fluctuation coefficient and fault index corresponding to each type of scenario indicator in several historical collection cycles, and obtaining the three-dimensional vulnerability curve corresponding to each type of scenario indicator, wherein the three coordinate axes of the three-dimensional vulnerability curve corresponding to each type of scenario indicator respectively correspond to the average value, fluctuation coefficient and fault indicator generation probability; The critical feature analysis module is used to analyze multi-source operating data before various types of operating indicators of a plurality of high-voltage inverters are marked as fault indicators, and to construct a critical feature set of various types of operating indicators of a plurality of high-voltage inverters, including: Extract the timestamp when the type of operating indicator is marked as a fault indicator and the historical collection period to which the type of operating indicator belongs when the type of operating indicator is marked as a fault indicator, mark the numerical time series sequence corresponding to each type of operating indicator between the start timestamp of the historical collection period and the timestamp as a critical data set, and mark the numerical time series sequence corresponding to each type of operating indicator in the historical collection period where no fault indicator is generated as a normal data set; Obtain the fluctuation coefficient range of each type of operating indicator in the critical data set and the normal data set, mark the type of operating indicator whose fluctuation coefficient range in the critical data set is consistent with that in the normal data set as a key type operating indicator, and mark the type of operating indicator whose fluctuation coefficient range in the critical data set is inconsistent with that in the normal data set as a key type operating indicator; Obtain the correlation coefficients between each key type of operating indicators in the critical data set and the correlation coefficients between each key type of operating indicators and each non-key type of operating indicators, obtain the predicted numerical time series sequence of the key type of operating indicators, perform difference detection on the key type of operating indicators in the critical data set according to the predicted numerical time series sequence of the key type of operating indicators, obtain the difference coefficients of the key type of operating indicators, and construct a critical feature set of the type of operating indicators according to the correlation coefficients between each key type of operating indicators, the correlation coefficients between each key type of operating indicators and each non-key type of operating indicators, and the difference coefficients of the key type of operating indicators; The risk warning module is used to construct a risk warning data set, perform feature matching on the multi-source operation data and external scene data of a plurality of high-voltage inverters with the risk warning data set, and determine the detection methods of a plurality of high-voltage inverters according to the feature matching results, including: Constructing a risk warning data set, the risk warning data set includes three-dimensional vulnerability curves corresponding to various types of scenario indicators of a number of high-voltage inverters and critical feature sets of various types of operating indicators of a number of high-voltage inverters, matching the external scenario data of the number of high-voltage inverters with the risk warning data set for scenario features, obtaining the probability of generating fault indicators of the number of high-voltage inverters, comparing the probability of generating fault indicators of the number of high-voltage inverters with a preset probability threshold, performing early warning detection on high-voltage inverters whose probability of generating fault indicators is greater than the probability threshold, and performing difference detection on high-voltage inverters whose probability of generating fault indicators is less than or equal to the probability threshold; The prediction module is used to obtain the predicted numerical time series of various types of operating indicators of a plurality of high-voltage inverters, including: Construct a prediction model based on deep learning, use the numerical time series sequences corresponding to various types of operating indicators in several historical acquisition cycles of several high-voltage inverters as training sets and test sets, input the training sets into the prediction model for training until the loss function training is stable, and save the model parameters, test the prediction model with the test set until it meets the preset requirements, output the prediction model, and output the predicted numerical time series sequences of various types of operating indicators of several high-voltage inverters according to the prediction model; The early warning detection module is used to perform early warning detection on a number of high-voltage inverters according to the critical feature set of each type of operating indicator in the risk warning data set, including: Obtain the fluctuation coefficient range corresponding to each type of operating indicator in the current acquisition cycle according to the multi-source operating data of the current acquisition cycle of the high-voltage inverter, perform consistency comparison between the fluctuation coefficient range corresponding to each type of operating indicator in the current acquisition cycle and the fluctuation coefficient range corresponding to each type of operating indicator in the normal data set, and judge whether there is a key type of operating indicator in the current acquisition cycle according to the comparison result; if there is a key type of operating indicator in the current acquisition cycle, obtain the correlation coefficient between each key type of operating indicator in the current acquisition cycle, the correlation coefficient between each key type of operating indicator and each non-key type of operating indicator, and the difference coefficient of the key type of operating indicator; perform similarity comparison between the correlation coefficient between each key type of operating indicator, the correlation coefficient between each key type of operating indicator and each non-key type of operating indicator, and the difference coefficient of the key type of operating indicator and the critical feature set of each type of operating indicator, and obtain the similarity of the critical feature set of each type of operating indicator; When the similarity of the critical feature set of the type operation indicator is greater than the preset similarity threshold, the type operation indicator is marked as a fault warning operation indicator, and a fault warning signal of the high-voltage inverter is generated according to the fault warning operation indicator; If there is no key type of operating indicator in the current collection period or the similarity of the critical feature set of each type of operating indicator is less than or equal to the preset similarity threshold, the high-voltage inverter is marked as operating normally; The difference detection module is used to perform difference detection on a plurality of high-voltage inverters, and feed back the high-voltage inverters that do not meet the requirements to the early warning detection module, including: Obtain predicted numerical time series of various types of operating indicators of the high-voltage inverter as judgment data, compare the numerical time series of various types of operating indicators in the current acquisition cycle of the high-voltage inverter with the judgment data, obtain the difference coefficient of each type of operating indicator, compare the difference coefficient of each type of operating indicator with a preset difference coefficient threshold, and if the difference coefficient of each type of operating indicator is less than or equal to the difference coefficient threshold, mark the high-voltage inverter as a normal operating state; If the difference coefficient of the type operation indicator is greater than the difference coefficient threshold, an early warning detection tag of the high-voltage inverter is generated and fed back to the early warning detection module. The early warning detection module performs early warning detection on the high-voltage inverter according to the early warning detection tag of the high-voltage inverter.

2. A power unit detection system for a high voltage frequency converter according to claim 1, characterized in that: The risk warning module matches the external scene data of a plurality of high-voltage inverters with the risk warning data set for scene features, and the process of obtaining the generation probability of the fault indicators of the plurality of high-voltage inverters includes: According to the external scene data of the current acquisition period of the high-voltage inverter, the average value and fluctuation coefficient corresponding to each type of scenario indicator in the current acquisition period are obtained, and the average value and fluctuation coefficient corresponding to each type of scenario indicator are compared with the three-dimensional vulnerability curve corresponding to each type of scenario indicator to obtain the fault indicator generation probability of each type of scenario indicator, and the maximum fault indicator generation probability is screened out from the fault indicators of each type of scenario indicator as the high-voltage inverter fault indicator generation probability.

Images