Reliability monitoring method and system for power electronic equipment based on fiber grating sensing

Abstract

The present application relates to a reliability monitoring method and system for power electronic equipment based on fiber grating sensing, characterized in that it includes: obtaining spectral sensing characteristic data information of fiber grating sensors pre-arranged in power electronic equipment; obtaining power electronic equipment operating parameter information; inputting spectral sensing characteristic data information and power electronic equipment operating parameter information into a pre-trained learning model to obtain time series temperature information inside the power electronic equipment; converting the time series temperature information into minimum temperature information and temperature difference information according to an online rain flow counting algorithm; calculating the achievable number of power cycles according to the obtained minimum temperature information and temperature difference information to obtain the remaining service life information of the power electronic equipment. The present application provides a solution for monitoring the surface of power electronic equipment or components to achieve health status detection.

Description

Technical Field

[0001] The present application relates to the field of power electronic equipment, and in particular to a reliability monitoring method and system for power electronic equipment based on fiber grating sensing. Background Art

[0002] Improving the reliability of power electronic equipment is one of the most important research trends today, as power systems have gradually become the backbone of energy production. In many applications that require high-quality services, especially when the equipment is located in remote areas (high-voltage transmission systems, industrial systems, offshore wind farms), the cost of unplanned maintenance of power electronic equipment can be very high. In the existing framework, the ability to assess the health of power electronic equipment has become critical, both from a service quality perspective and from an economic perspective.

[0003] Most of today's detection methods require sensors to be placed in power electronic equipment. Generally speaking, although the design of built-in thermocouples can measure the temperature of each point to estimate the health of power electronic equipment, thermocouples are easily affected by strong electromagnetic fields and cause reading errors. In order to solve this error, existing technologies also use fiber grating sensors instead of thermocouples to measure the temperature of each point, but because they are embedded in the interior of power electronic equipment, the solution is complex and difficult to maintain and replace, which can easily cause damage to the electronic power system. Summary of the invention

[0004] In order to provide a solution for monitoring the surface of power electronic equipment to detect health conditions, the present application provides a method and system for monitoring the reliability of power electronic equipment based on fiber grating sensing.

[0005] In the first aspect, the present application provides a method for monitoring the reliability of power electronic equipment based on fiber grating sensing, which adopts the following technical solution:

[0006] A reliability monitoring method for power electronic equipment based on fiber grating sensing, comprising:

[0007] Acquiring spectral sensing characteristic data information of a fiber grating sensor pre-arranged in a power electronic device;

[0008] Obtain operating parameter information of power electronic equipment;

[0009] Inputting the spectral sensing characteristic data information and the power electronic equipment operation parameter information into the pre-trained learning model to obtain the time series temperature information inside the power electronic equipment;

[0010] The time series temperature information is converted into minimum temperature information and temperature difference information according to the online rain flow counting algorithm;

[0011] The achievable number of power cycles is calculated based on the acquired minimum temperature information and temperature difference information to obtain the remaining service life information of the power electronic equipment.

[0012] Among them, this method uses the characteristic of fiber grating sensors that have a low impact on the normal operation of power electronic equipment, and obtains the time-series temperature information inside the power electronic equipment through a learning model. On the one hand, it allows the power electronic equipment to operate normally, and on the other hand, there is no need to use complex circuit arrangements to collect the temperature inside the power electronic equipment. In addition, after obtaining the time-series temperature information through the learning model, the remaining service life information of the power electronic equipment can also be intuitively calculated, thereby providing a solution for monitoring the surface of the power electronic equipment to achieve health status detection.

[0013] Preferably, the power electronic equipment operation parameter information includes electrical characteristic information and control parameter characteristic information, and the training method of the learning model includes:

[0014] Acquiring spectral sensing characteristic data information of a fiber grating sensor pre-arranged in a power electronic device;

[0015] Taking electrical characteristic information, control parameter characteristic information and spectral sensing characteristic data information as input variable data sets;

[0016] The prediction model is trained according to the acquired input variable data set and output variable data set, wherein the output variable data set includes at least one or more of node temperature data of a preset node or historical node temperature data of a preset node.

[0017] Preferably, the method for acquiring the node temperature data of the preset node includes:

[0018] Obtain operating parameter information of power electronic equipment;

[0019] A thermal circuit model is constructed according to the operating parameter information of the power electronic equipment to obtain the node temperature data of the power electronic equipment at the preset nodes.

[0020] Preferably, the input variable data set includes a training data set and a test data set, and the method for training the prediction model includes:

[0021] Building a learning model based on an input variable data set in a training data set, wherein the learning model has a number of preset and yet-to-be-determined hyperparameters;

[0022] Input the input variable data set in the test data set into the learning model, and verify the result with the corresponding output variable data set to determine whether the preset termination training condition is met;

[0023] If the preset termination training condition is not met, the information of each hyperparameter is changed and updated for iterative optimization until the preset termination training condition is met.

[0024] By adopting the above technical solution, the prediction model is evolved towards the optimal prediction model by iterating and evolving the hyperparameter combination required by the prediction model, thereby improving the accuracy of the prediction model in the process of continuous training and iteration.

[0025] Preferably, the termination training condition includes calculating at least one or more of the mean absolute error, mean absolute percentage error, root mean square error and regression determination coefficient in sequence according to the result of the output target prediction model function and the corresponding output variable data set, and comparing the obtained result with the corresponding termination condition;

[0026] The mean absolute error is expressed as follows:

[0027]

[0028] The mean absolute percentage error is calculated using the formula:

[0029]

[0030] The root mean square error is expressed as:

[0031]

[0032] The regression coefficient of determination is expressed as:

[0033]

[0034] Where N is the number of input variable data sets in the test data set, e MAE is the mean absolute error, e MAPE is the mean absolute percentage error, e RMSE is the root mean square error, R is the regression determination coefficient, represents the output data obtained after the i-th input variable data set is input into the learning model, Represents the data of the i-th output variable dataset.

[0035] Preferably, the method for changing and updating each hyperparameter adopts a grid search method to combine each hyperparameter within a specified optimization range to update the information of each hyperparameter.

[0036] Preferably, before being divided into a training data set and a test data set, the obtained input variable data set and output variable data set are standardized;

[0037] The standardization process satisfies the formula:

[0038]

[0039] In the formula is the standardized input variable data set, x i is the input variable data set before standardization, and are the maximum and minimum values ​​of the input variable data set before standardization.

[0040] Preferably, the rain flow counting algorithm recursively calculates the time series temperature information in the order of time series, and the algorithm includes:

[0041] When the time series temperature information at the current moment is identified as a maximum value, it is determined whether there is a stored maximum value in the maximum value stack; when there is no stored maximum value in the maximum value stack, the current maximum value is stored in the maximum value stack, otherwise, the value in the maximum value stack is compared with the current maximum value;

[0042] If the current maximum value is greater than any value in the maximum value stack, then check the number of values ​​in the minimum value stack;

[0043] If there is only one value in the minimum value stack, the difference between the value in the minimum value stack and the first value in the maximum value stack is used as the temperature difference information, and then the first value in the maximum value stack is replaced by the current maximum value;

[0044] If there are multiple values ​​in the minimum value stack, the difference between the first value in the minimum value stack and the first value in the maximum value stack is used as the temperature difference information, and then the first value in the minimum value stack is removed, and the first value in the maximum value stack is replaced by the current maximum value;

[0045] If the current maximum value is smaller than any value in the maximum value stack, the current maximum value is stored in the maximum value stack, wherein the minimum temperature information is the value in the minimum value stack.

[0046] Preferably, after obtaining the minimum temperature information and the temperature difference information, the method for calculating the remaining service life information includes: obtaining specification parameter information of the power electronic device;

[0047] A corresponding preset coefficient relationship is assigned to each specification parameter information, and the remaining service life information is calculated with the acquired minimum temperature information and temperature difference information.

[0048] In the second aspect, the present application provides a power electronic equipment reliability monitoring system based on fiber grating sensing, which adopts the following technical solution:

[0049] A reliability monitoring system for power electronic equipment based on fiber grating sensing, comprising:

[0050] The FBG sensor module includes a fiber grating sensor, which is used to obtain spectral sensing characteristic data information of the fiber grating sensor pre-arranged on the surface of the power electronic equipment;

[0051] An operating parameter reading module is used to obtain operating parameter information of power electronic equipment;

[0052] A temperature measurement module is used to input the spectral sensing characteristic data information and the power electronic equipment operation parameter information into a pre-trained learning model to obtain the time series temperature information inside the power electronic equipment;

[0053] The health status monitoring module is used to convert the time series temperature information into minimum temperature information and temperature difference information according to the rain flow counting algorithm; and calculate the achievable number of power cycles based on the acquired minimum temperature information and temperature difference information to obtain the remaining service life information of the power electronic equipment.

[0054] In summary, the present application includes at least one of the following beneficial technical effects:

[0055] 1. This method uses the characteristics of fiber Bragg grating sensors that have little impact on the normal operation of power electronic equipment, and obtains the time-series temperature information inside the power electronic equipment by learning the model. On the one hand, the power electronic equipment can operate normally, and on the other hand, there is no need to use complex circuit arrangements to collect the temperature inside the power electronic equipment. In addition, after obtaining the time-series temperature information through the learning model, the remaining service life information of the power electronic equipment can also be intuitively calculated, thereby providing a solution for monitoring the surface of the power electronic equipment to achieve health status detection;

[0056] 2. Covers the modeling process of scenario-based learning models. The combination of thermal circuit models and fiber Bragg grating sensors on target power electronic devices provides a comprehensive dataset for the construction of learning models for temperature monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flow chart of a reliability monitoring method according to one embodiment of the present invention.

[0058] Figure 2 It is a flowchart of a method for constructing a learning model according to one embodiment of the present invention.

[0059] Figure 3 It is a flowchart of a training learning model according to one embodiment of the present invention.

[0060] Figure 4It is a flowchart of an online rainflow counting algorithm according to one embodiment of the present invention.

[0061] Figure 5 It is a schematic diagram of the arrangement of the sensing optical fiber in one embodiment of the present invention.

[0062] Description of the accompanying drawings: 1. Sensing optical fiber; 2. Electrolyte capacitor; 3. FBG demodulator. DETAILED DESCRIPTION

[0063] The following is combined with Figure 1-5 This application is described in further detail.

[0064] The embodiment of the present application discloses a reliability monitoring method and system for power electronic equipment based on fiber grating sensing, which collects corresponding data through a fiber grating sensor set on the surface of the power electronic equipment and obtains the temperature information of each internal point more accurately through machine learning. The obtained temperature information is used to establish a health status monitoring model, and the constructed health status detection model is used to simulate the operating status of the monitored power electronic equipment, especially the remaining life estimation and other information.

[0065] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. 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.

[0066] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0067] The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings.

[0068] The embodiment of the present invention provides a method for monitoring the reliability of power electronic equipment based on fiber grating sensing, referring to Figure 1 , the main process of the method is described as follows.

[0069] Step S100: Acquire spectral sensing characteristic data information of a fiber grating sensor pre-arranged in a power electronic device.

[0070] Among them, fiber grating sensors are generally used to be placed on the surface of components in power electronic equipment that need to be measured, or placed along the periphery of power electronic equipment. Fiber grating sensors are generally made of non-conductive thermal conductive materials, so they do not affect the thermal performance of power electronic equipment. Generally speaking, fiber grating sensors have the advantages of being customizable, immune to electromagnetic interference, and small space (generally the diameter of a single-mode optical fiber is 125 microns, and after high-temperature resistant coating, it will basically not exceed 500 microns). The spectral sensing characteristic data information includes the instantaneous central wavelength collected by the fiber grating sensor and the converted temperature information.

[0071] Step S200: Acquire operating parameter information of power electronic equipment.

[0072] The operating parameter information of the power electronic equipment includes the electrical characteristics, load characteristics, control characteristics and physical characteristic parameters related to the electronic components of the power electronic equipment, including at least the current, voltage, load power and switching frequency during the operation of the power electronic equipment.

[0073] Step S300: inputting the spectral sensing characteristic data information and the power electronic device operation parameter information into a pre-trained learning model to obtain the time series temperature information inside the power electronic device.

[0074] Reference Figure 2 , as specific training methods, specifically include:

[0075] Step S10: Acquire spectral sensing characteristic data information of a fiber grating sensor pre-arranged in a power electronic device.

[0076] Among them, the arrangement of the fiber grating sensor in the process of training the learning model is the same as that in the subsequent monitoring process. Similarly, the spectral sensing characteristic data information here also includes the instantaneous central wavelength collected by the fiber grating sensor and the converted temperature information.

[0077] Step S20: taking the electrical characteristic information, the control parameter characteristic information and the spectral sensing characteristic data information as input variable data sets.

[0078] Here, the electrical characteristic information includes the current I and voltage V of the power electronic equipment during operation, and the control parameter characteristic information includes the load power P and the switching frequency f s The spectral sensing characteristic data information includes the instantaneous central wavelength (CW 1 ,…CW N ) and the measured PE element surface temperature (T s1 ,…T sN), where N represents the number of fiber Bragg grating sensors. In order to reduce the error caused by ambient temperature, the ambient temperature T can also be added to the input variable data set. ambient Specifically, the input variable data set needs to contain a number of data in a time series, represented as x i =[CW 1 ,…CW N ,I,V,P,f s ,T ambient ,T s1 ,…T sN ] i n , i represents the i-th sample of the input feature.

[0079] Step S30: training a prediction model according to the acquired input variable data set and output variable data set, wherein the output variable data set includes at least one or more of node temperature data of a preset node or historical node temperature data of a preset node.

[0080] The data in the output variable data set represents the wide range of internal temperatures of the power electronic device, which may be only the node temperature data of the preset node, or only the historical node temperature data of the preset node (the historical temperature measured by the power electronic device with the same parameter value), or a combination of the two and an average thereof for representation. Here, a node refers to a point in each power electronic device where the temperature needs to be measured.

[0081] The method for obtaining node temperature data of a preset node includes:

[0082] Step S31: Acquire operating parameter information of power electronic equipment.

[0083] The operating parameter information of power electronic equipment includes the electrical characteristics, load characteristics, control characteristics of the power electronic equipment and physical characteristic parameters related to electronic components.

[0084] Step S32: constructing a thermal circuit model according to the operating parameter information of the power electronic device to obtain node temperature data of the power electronic device at a preset node.

[0085] Among them, the thermal circuit model of the power electronic equipment and the power loss model of the power electronic equipment can be constructed through the obtained operating parameter information of the power electronic equipment, and the power loss model will also be introduced into the constructed thermal circuit model, so that the constructed thermal circuit model can generate node temperature data of the power electronic equipment at the preset nodes. As an implementable implementation, the thermal circuit model in this embodiment uses the LPTN model.

[0086] Reference Figure 3, in the process of training the learning model, including:

[0087] Step S33: constructing a learning model according to the input variable data set in the training data set, wherein the learning model has a number of preset and pending hyperparameters.

[0088] Among them, as a training for the learning model, the input variable data set needs to be standardized first. After the standardized preprocessing, it will be converted into a standardized value, unify the comparison standard, ensure the reliability of the results, and eliminate the influence of different attributes between different indicators. Then the standardized input variable data set is randomly divided into a training data set and a test data set, which are used to perform different processing in subsequent steps. As for the division of the training data set and the test data set, in the embodiment of the present application, the standardized data set is divided into a training data set and a test data set using the method of time series data division, for example, the first 50% is used for training and the latter 50% is used for testing.

[0089] As a specific standardized preprocessing method, it meets the following requirements:

[0090]

[0091] In the formula is the standardized input variable data set, x i is the input variable data set before standardization, and are the maximum and minimum values ​​of the input variable data set before standardization.

[0092] Here, the learning model can deploy a variety of machine learning models, such as artificial neural networks, support vector regression, Bayesian models, random forests or other learning models. In this embodiment, the ANN model is preferred. In general, the perceptron is the basis of the neural network model. A single perceptron (neuron) is a linear model that aims to establish a linear hyperplane to classify linearly separable data sets. However, the single-layer perceptron model cannot handle nonlinear problems, that is, it cannot handle the "or problem". The solution to this problem is to construct a multi-layer perceptron structure model. The main structural difference between it and the single-layer perceptron is that the multi-layer perceptron structure model has several more hidden layers, which enables the neural network to handle nonlinear problems. The basic principle of multi-layer perceptron regression model training is the back propagation algorithm based on the gradient descent strategy. The weight of each neuron and the calculation and solution of the constant term use the chain derivation rule.

[0093] As the key hyperparameters of the ANN model, they include at least the number of hidden layers, the number of neurons in each layer, the type of activation function, the type of optimizer function, the number of training rounds and the learning rate.

[0094] Step S34: input the input variable data set in the test data set into the learning model, and verify the result with the corresponding output variable data set to determine whether the preset termination training condition is met.

[0095] Among them, the output variable data set also needs to be standardized and preprocessed before training, just like the input variable data set, and the input variable data set and the output variable data set after standardized preprocessing are still in a one-to-one mapping relationship at the same time.

[0096] In step S33, the data in the training data set is mainly used to build a learning model, and after the learning model is built, the data in the test data set is input into the built learning model in step S34, and then the output result of the learning model is compared with the corresponding data in the output variable data set to determine whether the hyperparameters corresponding to the learning model meet the required requirements. Here, the termination training conditions specifically include one or more of the mean absolute error, the mean absolute percentage error, the root mean square error, and the regression determination coefficient, and the above-mentioned termination training conditions are obtained by calculating the results of the output learning model with the corresponding output variable data set. Under the preferred embodiment, the learning model corresponding to the hyperparameters that need to meet all the termination training conditions at the same time is the final required learning model.

[0097] The mean absolute error is calculated using the formula:

[0098]

[0099] The mean absolute percentage error is calculated using the formula:

[0100]

[0101] The root mean square error uses the formula:

[0102]

[0103] The regression coefficient of determination adopts the formula:

[0104]

[0105] Where N is the number of input variable data sets in the test data set, e MAE is the mean absolute error, e MAPE is the mean absolute percentage error, e RMSE is the root mean square error, R is the regression determination coefficient, represents the output data obtained after the i-th input variable data set is input into the learning model, Represents the data of the i-th output variable data set. As an achievable method, set the termination threshold e MAE <5℃,e MAPE <10%,e RMSE <5℃,R>0.9.

[0106] Step S35: If the preset termination training condition is not met, change and update the information of each hyperparameter to perform iterative optimization until the preset termination training condition is met.

[0107] Wherein, in step S34, after a set of data in a training test set is input into the constructed learning model, the data corresponding to the output and output variable data sets are calculated to obtain the specific values ​​of each median training condition. If one item does not meet the corresponding termination threshold, it indicates that the hyperparameter selected by the current learning model does not meet the required requirements, and the hyperparameter needs to be iteratively optimized. As a specific optimization method, a grid search method can be used to combine each hyperparameter within a specified optimization range to update the information of each hyperparameter. The learning model corresponding to the updated hyperparameter information will repeat steps S34 and S35 until each termination training condition meets the required termination requirements. Wherein, the grid search method refers to the exhaustive method, that is, exhaustively and combined within the reasonable optimization range of each hyperparameter, and each combination of hyperparameters is brought into the constructed machine learning model to continuously determine whether the preset termination training conditions are met.

[0108] At this point, the learning model required in step S300 has been trained. In the subsequent actual detection of the power electronic device and execution of step S300, the time series temperature information inside the power electronic device can be obtained by collecting the spectral sensing characteristic data information and operating parameter information of the fiber grating sensor.

[0109] Step S400: converting the time series temperature information into minimum temperature information and temperature difference information according to the online rain flow counting algorithm.

[0110] Among them, the time series temperature information refers to a series of temperature values ​​obtained after inputting the learning model in time series. Since the electrical characteristic information, control parameter characteristic information and spectral sensor characteristic data information will be obtained in real time, a temperature value will be obtained after inputting into the learning model. Since the data of the input variable data set will be continuously obtained, the temperature value will also be continuously updated with the passage of time, thus forming a time series temperature information in time series. Here, the maximum value and the minimum value refer to the value of the temperature change in time series. In simple terms, for the definition of the maximum value, when the temperature values ​​at adjacent moments are all less than the current moment, it means that the temperature at the current moment is the maximum value of the time series temperature information. Conversely, when the temperature values ​​at adjacent moments are all greater than the current moment, it means that the temperature at the current moment is the minimum value of the time series temperature information. Here, the maximum value and the minimum value in the time series temperature information can also be judged by the definition of the derivative. When executing the online rain flow counting algorithm, the maximum value and the minimum value in the time series temperature information will be recursively calculated by the rain flow counting algorithm in the order of the time series.

[0111] Reference Figure 4 , the specific process of the online rainflow counting algorithm includes:

[0112] Step S410: when the time series temperature information at the current moment is identified as a maximum value, it is determined whether there is a stored maximum value in the maximum value stack.

[0113] The maximum value stack refers to a stack used to store maximum values.

[0114] Step S420: When there is no stored maximum value in the maximum value stack, the current maximum value is stored in the maximum value stack; otherwise, the value in the maximum value stack is compared with the current maximum value.

[0115] Step S430: If the current maximum value is greater than any value in the maximum value stack, the number of values ​​in the minimum value stack is detected.

[0116] Among them, the minimum value stack refers to a stack used to store minimum values, and both the minimum value stack and the maximum value stack will be assigned an initial value, which can be temperature information such as room temperature.

[0117] Step S431: If there is only one value in the minimum value stack, the difference between the value in the minimum value stack and the first value in the maximum value stack is used as the temperature difference information, and then the first value in the maximum value stack is replaced by the current maximum value.

[0118] Step S432: If there are multiple values ​​in the minimum value stack, the difference between the first value in the minimum value stack and the first value in the maximum value stack is used as the temperature difference information, and then the first value in the minimum value stack is removed, and the first value in the maximum value stack is replaced by the current maximum value.

[0119] Step S440: If the current maximum value is smaller than any value in the maximum value stack, the current maximum value is stored in the maximum value stack.

[0120] Step S450: If there are multiple values ​​in the maximum value stack, re-execute steps S410 to S450.

[0121] Among them, after step S410-step S450, the online rain flow counting algorithm will convert the data in the time series temperature information into the maximum value stack and the minimum value stack according to the corresponding storage logic according to each step, and the corresponding temperature difference information can be obtained through the limitation of each step. In other words, it is only necessary to read the value in the minimum value stack and the obtained temperature difference information to obtain the minimum temperature information and temperature difference information required for the subsequent steps.

[0122] Step S500: Calculate the achievable number of power cycles according to the acquired minimum temperature information and temperature difference information to obtain the remaining service life information of the power electronic device.

[0123] Among them, the remaining service life information of the power electronic equipment is not only related to the minimum temperature information and the temperature difference information, but also to the specific specifications of the power electronic equipment. In one embodiment, the calculation of the remaining service life information first needs to obtain the specification parameter information of the power electronic equipment, and then assign the corresponding preset coefficient relationship to each specification parameter information and calculate the remaining service life information with the obtained minimum temperature information and temperature difference information. Among them, the specification parameter information of the power electronic equipment is associated with the monitored power electronic equipment, so different power electronic equipment has different specific required specification parameter information, but because the monitored power electronic equipment will not be replaced, the required specification parameter information and the corresponding coefficient relationship can be pre-set. In a feasible embodiment, the remaining service life information satisfies the formula:

[0124]

[0125] Where N f is the remaining service life, ΔT j is the temperature difference information, T j(min) is the minimum temperature information, T on is the pulse duration, I B is the pin current, V Cis the voltage calibration value, D is the pin wire diameter, β 1 -β 6 are the coefficient values ​​corresponding to each parameter, and A is the corresponding calibration coefficient.

[0126] Among them, the remaining service life information is a calculation formula for the remaining service life information corresponding to semiconductor components in electronic power equipment. The corresponding pulse duration, pin current, voltage calibration value and pin welding wire diameter will change depending on the monitored object, and the coefficient values ​​corresponding to the specific parameters will also change depending on the usage scenarios of the monitored object.

[0127] In addition, the various data obtained through the above steps can be used to construct a pre-diagnostic fault indicator to indicate the possible existence of hardware faults by comparing the collected real-time values ​​with historical values. The input of the indicator can include the operating parameter information of the electronic power equipment, the minimum temperature information and temperature difference information output by the rain flow counting algorithm, the remaining service life, the time series temperature information of the output learning model, the electrical characteristic data, and the spectral sensing characteristic data information obtained by the fiber grating sensor. For example, under the same load conditions, the time series temperature of the output learning model continues to increase over time without tending to a steady state, which means that the increase in power loss can be used as a pre-indication of the occurrence of a fault. In addition, the real-time changes in the current / voltage of the power electronic equipment and the real-time impedance changes of the PE element can be used as additional indicators of the health status. In addition, the vibration sensing data provided by the fiber grating sensor can indicate the physical state of the power electronic equipment, and any vibration harmonics with an amplitude exceeding the safety threshold will be regarded as a foreshadowing of the occurrence of a hardware fault.

[0128] Based on the same inventive concept, the embodiment of the present application also discloses a reliability monitoring system for power electronic equipment based on fiber grating sensing, comprising:

[0129] The FBG sensor module includes a fiber grating sensor and is used to obtain spectral sensing characteristic data information of the fiber grating sensor pre-arranged on the surface of the power electronic equipment.

[0130] The operating parameter reading module is used to obtain the operating parameter information of the power electronic equipment.

[0131] The temperature measurement module is used to input the spectral sensing characteristic data information and the power electronic equipment operation parameter information into the pre-trained learning model to obtain the time series temperature information inside the power electronic equipment.

[0132] The health status monitoring module is used to convert the time series temperature information into minimum temperature information and temperature difference information according to the rain flow counting algorithm; and calculate the achievable number of power cycles based on the acquired minimum temperature information and temperature difference information to obtain the remaining service life information of the power electronic equipment.

[0133] The method for constructing the learning model is as described in the aforementioned steps S10 to S30.

[0134] In one practicable manner, the power electronic device may be an array-type electrolyte capacitor 2, which may be applicable to the formula in step S500 to detect the life information. The FBG sensor module includes a plurality of sensing optical fibers 1 having multiple grid regions, and the FBG sensor module may be composed of optical fibers of the same type (e.g., temperature), or a plurality of optical fibers of different types of sensors (e.g., temperature and vibration). The sensing optical fiber 1 is arranged on the surface of the array of electrolyte capacitors 2 in a preset manner, and one end is connected to the FBG demodulator 3 through an optical fiber connector. The data demodulated by the FBG demodulator 3 is the central wavelength data, and this part of the data is converted accordingly to obtain temperature data and vibration information to be measured as the input of the machine learning model.

[0135] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0136] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0137] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0138] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0139] As described above, the above embodiments are only used to introduce the technical solution of the present application in detail, but the description of the above embodiments is only used to help understand the method and core idea of ​​the present invention and should not be understood as limiting the present invention. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A reliability monitoring method for power electronic equipment based on fiber grating sensing, It is characterized in that include: Acquiring spectral sensing characteristic data information of a fiber grating sensor pre-arranged in a power electronic device; Obtain operating parameter information of power electronic equipment; Inputting the spectral sensing characteristic data information and the power electronic equipment operation parameter information into the pre-trained learning model to obtain the time series temperature information inside the power electronic equipment; The time series temperature information is converted into minimum temperature information and temperature difference information according to the online rain flow counting algorithm; Calculate the achievable number of power cycles based on the acquired minimum temperature information and temperature difference information to obtain the remaining service life information of the power electronic equipment; The power electronic equipment operation parameter information includes electrical characteristic information and control parameter characteristic information, and the training method of the learning model includes: Acquiring spectral sensing characteristic data information of a fiber grating sensor pre-arranged in a power electronic device; Taking electrical characteristic information, control parameter characteristic information and spectral sensing characteristic data information as input variable data sets; Training a learning model according to the acquired input variable data set and output variable data set, wherein the output variable data set includes at least one or more of node temperature data of a preset node or historical node temperature data of a preset node; The input variable data set includes a training data set and a test data set, and the method for training the learning model includes: Building a learning model based on an input variable data set in a training data set, wherein the learning model has a number of preset and yet-to-be-determined hyperparameters; Input the input variable data set in the test data set into the learning model, and verify the result with the corresponding output variable data set to determine whether the preset termination training condition is met; If the preset termination training condition is not met, the information of each hyperparameter is changed and updated for iterative optimization until the preset termination training condition is met; The termination training condition includes calculating at least one or more of the mean absolute error, the mean absolute percentage error, the root mean square error and the regression determination coefficient in sequence according to the result of the output learning model and the corresponding output variable data set, and comparing the obtained result with the corresponding termination condition; The mean absolute error is expressed as follows: The mean absolute percentage error is calculated using the formula: The root mean square error is expressed as: The regression coefficient of determination is expressed as: Where N is the number of input variable data sets in the test data set, e MAE is the mean absolute error, e MAPE is the mean absolute percentage error, e RMSE is the root mean square error, R is the regression determination coefficient, represents the output data obtained after the i-th input variable data set is input into the learning model, represents the data of the i-th output variable data set; The online rain flow counting algorithm recursively operates the time series temperature information according to the order of the time series, and the algorithm includes: when the time series temperature information at the current moment is identified as a maximum value, determining whether there is a maximum value stored in the maximum value stack; When there is no stored maximum value in the maximum value stack, the current maximum value is stored in the maximum value stack, otherwise, the value in the maximum value stack is compared with the current maximum value; If the current maximum value is greater than any value in the maximum value stack, then check the number of values ​​in the minimum value stack; If there is only one value in the minimum value stack, the difference between the value in the minimum value stack and the first value in the maximum value stack is used as the temperature difference information, and then the first value in the maximum value stack is replaced by the current maximum value; If there are multiple values ​​in the minimum value stack, the difference between the first value in the minimum value stack and the first value in the maximum value stack is used as the temperature difference information, and then the first value in the minimum value stack is removed, and the first value in the maximum value stack is replaced by the current maximum value; If the current maximum value is smaller than any value in the maximum value stack, the current maximum value is stored in the maximum value stack, wherein the minimum temperature information is the value in the minimum value stack.

2. The method according to claim 1, It is characterized in that The method for acquiring the node temperature data of the preset node includes: Obtain operating parameter information of power electronic equipment; A thermal circuit model is constructed according to the operating parameter information of the power electronic equipment to obtain the node temperature data of the power electronic equipment at the preset nodes.

3. The method according to claim 1, It is characterized in that The method for changing and updating each hyperparameter adopts a grid search method to combine each hyperparameter within a specified optimization range to update the information of each hyperparameter.

4. The method according to claim 1, It is characterized in that Before dividing into a training data set and a test data set, the obtained input variable data set and output variable data set are standardized; The standardization process satisfies the formula: In the formula is the standardized input variable data set, x i is the input variable data set before standardization, and are the maximum and minimum values ​​of the input variable data set before standardization.

5. The method according to claim 1, It is characterized in that After obtaining the minimum temperature information and the temperature difference information, the method for calculating the remaining service life information includes: Obtain specification parameter information of power electronic equipment; A corresponding preset coefficient relationship is assigned to each specification parameter information, and the remaining service life information is calculated with the acquired minimum temperature information and temperature difference information.

6. A reliability monitoring system for power electronic equipment based on fiber grating sensing, It is characterized in that The reliability monitoring method of power electronic equipment based on fiber grating sensing as claimed in claim 1 comprises: The FBG sensor module includes a fiber grating sensor, which is used to obtain spectral sensing characteristic data information of the fiber grating sensor pre-arranged on the surface of the power electronic equipment; An operating parameter reading module is used to obtain operating parameter information of power electronic equipment; A temperature measurement module is used to input the spectral sensing characteristic data information and the power electronic equipment operation parameter information into a pre-trained learning model to obtain the time series temperature information inside the power electronic equipment; The health status monitoring module is used to convert the time series temperature information into minimum temperature information and temperature difference information according to the rain flow counting algorithm; and calculate the achievable number of power cycles based on the acquired minimum temperature information and temperature difference information to obtain the remaining service life information of the power electronic equipment.

Images