Self-detection method and device of power converter and computer program product

Abstract

The invention belongs to the technical field of fault detection of power converters, and discloses a self-detection method and device of a power converter and a computer program product in order to solve the problem that the overall performance of the converter cannot be evaluated in the prior art. According to the method, fault self-detection of the power converter is carried out by extracting operation data used for representing working states of different elements in the converter, and a hidden Markov model obtained by training through historical operation data of the power converter is constructed at the same time; thus, the operation data of the different elements are input into the trained hidden Markov model, the real-time operation state of the power converter can be obtained, and fault self-detection of the power converter is completed. Therefore, fault self-detection is carried out on the basis of the operation data used for representing the working states of different elements, and compared with the traditional technology, the overall performance of the power converter can be effectively evaluated, so that effective fault detection of the power converter can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault detection of power converters, and in particular relates to a self-detection method, device and computer program product of a power converter. Background Art

[0002] As a universal power conversion topology device, DC-DC converter has many types, and the commonly used ones include bidirectional full-bridge DC-DC converters. Due to its advantages such as high power density, strong control flexibility and modular design, it has been widely used in various important power conversion systems, such as electric vehicles, energy storage systems and power electronic transformers. At the same time, with the increasing reliability requirements for power electronic converters, the fault diagnosis problem in DC-DC converters has attracted more and more attention. Among them, if the DC-DC converter fails, it will cause the loss of electronic equipment function and even cause the system to paralyze, resulting in huge losses. Therefore, it is very critical to accurately detect the fault of the DC-DC converter before use, which helps users take defensive measures in time, thereby reducing maintenance cycles and losses.

[0003] At present, the fault detection of DC-DC converters is mostly aimed at a specific component in the converter, such as fault detection of electrolytic capacitors based on Kalman filtering, fault detection of filters based on Gaussian regression method, etc.; although the aforementioned technology can realize the fault detection of a specific component in the converter, due to the problems of load type and tolerance of each component in the converter, the aforementioned existing technology has limitations in application and cannot evaluate the overall performance of the DC-DC converter, thereby failing to effectively realize the fault detection of the DC-DC converter; therefore, based on the aforementioned deficiencies, how to provide a self-detection method that can evaluate the overall performance of the DC-DC converter has become a problem to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to provide a self-detection method, device and computer program product for a power converter, so as to solve the problem that the prior art cannot evaluate the overall performance of a DC-DC converter, thereby failing to effectively implement fault detection of the DC-DC converter.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, a self-detection method for a power converter is provided, comprising: Acquire operating parameters of the target power converter during the self-detection process, wherein the operating parameters include a first voltage, a second voltage, and a third voltage, and the first voltage is the output voltage of the target power converter, the second voltage is the output voltage of the inverter circuit in the target power converter, and the third voltage is the voltage across the freewheeling tube in the target power converter; Based on the historical operating parameter set of the target power converter, and using the power converter fault identification training algorithm, the hidden Markov model is trained to obtain a fault detection model after the training is completed, wherein the power converter fault identification training algorithm is used to optimize the optimal state number of the hidden Markov model and the optimal model parameters corresponding to the optimal state number by using an evolutionary computing method, and each state in the optimal state number of the hidden Markov model corresponds to an operating state of the target power converter; The operating parameters are input into the fault detection model for fault detection processing, so as to obtain a self-detection result of the target power converter after the fault detection processing.

[0006] Based on the above disclosed content, the self-detection method of the power converter provided by the present invention obtains the operating parameters of the target power converter during the self-detection process, and the operating parameters include the output voltage of the target power converter, the output voltage of the inverter circuit in the target power converter and the voltage across the freewheeling tube in the target power converter; wherein the voltage across the freewheeling tube contains the fault information of the rectifier element in the converter, the output voltage of the entire converter contains the fault information of the internal filter, and the output voltage of the inverter circuit contains the fault information of the switching tube; in this way, the present invention is equivalent to collecting operating data used to characterize the working status of different components in the converter to perform fault self-detection of the target power converter.

[0007] In specific applications, the present invention constructs a fault detection model, which is trained using its historical operating parameters as a training set, wherein the fault detection model is a hidden Markov model with multiple state numbers, and each state corresponds to an operating state of the target power converter; thus, when performing self-detection, the aforementioned operating parameters are directly input into the constructed fault detection model, and the real-time operating state of the target power converter (that is, the self-detection result) can be obtained.

[0008] Through the above design, the present invention performs fault self-detection of the power converter by extracting operating data used to characterize the working states of different components in the converter, and simultaneously constructs a hidden Markov model trained using its historical operating data; in this way, the operating data of different components of the power converter are input into the aforementioned trained hidden Markov model, and the real-time operating state of the power converter can be obtained, and after obtaining the real-time operating state, the fault self-detection of the power converter can be completed; therefore, the present invention performs fault self-detection based on the operating data used to characterize the working states of different components, and compared with traditional technologies, it can effectively evaluate the overall performance of the power converter, thereby realizing effective fault detection of the power converter.

[0009] In one possible design, a hidden Markov model is trained based on a historical operating parameter set of a target power converter and using a power converter fault identification training algorithm, including: Initializing the state number of the hidden Markov model to obtain a plurality of initial state numbers, and determining the initial model parameters of the hidden Markov model at each initial state number according to each initial state number, wherein each initial state corresponding to any initial state number is respectively used to characterize an operating state of the target power converter; Based on each initial state number and the initial model parameter corresponding to each initial state number, a plurality of individual particles are generated, wherein each individual particle is configured with an initial position vector and an initial velocity vector, and each initial position vector includes an initial state number and the initial model parameter corresponding to the initial state number; Initialize the number of iterations s to 1, and determine the position vector and velocity vector of each particle individual at the s-th iteration, wherein when s is 1, the position vector and velocity vector of any particle individual at the s-th iteration are the initial position vector and initial velocity vector of any particle individual; According to the model parameters in the position vector of each particle individual at the s-th iteration, an initial hidden Markov model of each particle individual at the s-th iteration is established, and according to the state number in the position vector of each particle individual at the s-th iteration, a training set of each particle individual at the s-th iteration is determined from the historical operation parameter set, wherein the training set of any particle individual at the s-th iteration contains the historical operation parameters of the target power converter under the target operation state, and the target operation state is various operation states corresponding to the state number in the position vector of any particle individual at the s-th iteration; Using the training set of each particle individual at the s-th iteration, the initial hidden Markov model of each particle individual at the s-th iteration is trained, and after the training, the fitness of each particle individual at the s-th iteration is obtained based on the output of the trained initial hidden Markov model corresponding to each particle individual; According to the fitness of each individual particle, update the global optimal solution and the individual optimal solution of each individual particle, wherein the global optimal solution is the position vector corresponding to the individual particle with the largest fitness during the 1st to sth iterations; Determine whether the iteration stop condition is met; If not, then according to the updated global optimal solution and each updated individual optimal solution, the velocity vector corresponding to each individual particle is updated to obtain the updated velocity vector corresponding to each individual particle; Using the state number in the position vector of each particle individual at the s-th iteration and the updated velocity vector corresponding to each particle individual, the state number in the position vector of each particle individual at the s-th iteration is updated to obtain the updated state number corresponding to each particle individual; Based on the updated state number corresponding to each individual particle, the model parameters of the initial hidden Markov model corresponding to each individual particle are re-determined, and the updated position vector of each individual particle is formed by using the updated state number corresponding to each individual particle and the re-determined model parameters; s is incremented by 1, and the updated position vector and updated velocity vector of each individual particle are used as the position vector and velocity vector of each individual particle at the s-th iteration, and the initial hidden Markov model of each individual particle at the s-th iteration is reestablished according to the model parameters in the position vector of each individual particle at the s-th iteration, until the iteration stop condition is met, and based on the global optimal solution when the iteration stop condition is met, the optimal state number and the optimal model parameters corresponding to the optimal state number are determined, so as to obtain the fault detection model based on the optimal state number and the optimal model parameters corresponding to the optimal state number.

[0010] In a possible design, the velocity vector corresponding to each individual particle is updated according to the updated global optimal solution and each updated individual optimal solution, and the updated velocity vector corresponding to each individual particle is obtained, including: For any individual particle, generate a first random number and a second random number, wherein the first random number and the second random number are different from each other and both are in [0,1]; Obtaining a maximum number of iterations, and generating a first update operator according to the maximum number of iterations and the first random number, and generating a second update operator according to the maximum number of iterations and the second random number; Based on the first update operator and the updated individual optimal solution corresponding to any individual particle, calculate the individual perturbation solution of any individual particle at the s-th iteration, and calculate the global perturbation solution at the s-th iteration according to the second update operator and the updated global optimal solution; According to the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any particle individual at the s-th iteration, an updated velocity vector corresponding to any particle individual is calculated.

[0011] In one possible design, generating a first update operator according to the maximum number of iterations and the first random number includes: According to the following formula (1), the first update operator is calculated; (1) In the above formula (1), represents the first update operator, represents the maximum number of iterations, represents the first random number; Correspondingly, based on the first update operator and the updated individual optimal solution corresponding to any individual particle, calculating the individual perturbation solution of any individual particle at the s-th iteration includes: The product of the first update operator and the updated individual optimal solution corresponding to any individual particle is used as the individual perturbation solution of any individual particle at the s-th iteration.

[0012] In a possible design, according to the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any particle individual at the s-th iteration, an updated velocity vector corresponding to any particle individual is calculated, including: According to the updated global optimal solution and the position vector of each particle at the sth iteration, the update weight is calculated; Based on the update weight, a first acceleration factor is determined, and according to the first acceleration factor, a second acceleration factor is calculated; An updated velocity vector corresponding to any individual particle is calculated using the update weight, the first acceleration factor, the second acceleration factor, the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any individual particle at the s-th iteration.

[0013] In a possible design, the model parameters in the position vector of any particle individual at the sth iteration include: the state distribution matrix, the state transition probability matrix and the observation value output probability matrix of the initial hidden Markov model corresponding to the any particle individual; Among them, according to the updated global optimal solution and the position vector of each particle individual at the sth iteration, the update weight is calculated, including: Based on the position vector of each particle individual at the sth iteration and the updated global optimal solution, and according to the following formula (2), the distance between each particle individual and the target individual is calculated, where the target individual is the particle individual corresponding to the updated global optimal solution; (2) In the above formula (2), Indicates the number of particles in each individual The distance between each particle and the target individual, Sequentially indicates The state transfer probability matrix, observation value output probability matrix and state distribution matrix in the position vector of each particle at the sth iteration are: represent the state transition probability matrix, observation value output probability matrix and state distribution matrix in the updated global optimal solution in turn, Indicates the The number of states in the position vector of the particle individual at the sth iteration, represents a constant, where ,and represents the total number of individual particles, Represents the modulo operator; Based on the distance between each particle individual and the target individual, the average distance at the sth iteration is obtained; Obtain all average distances from the 1st to the sth iteration, and select the largest average distance from all the obtained average distances; The ratio between the average distance at the sth iteration and the maximum average distance is used as the update weight.

[0014] In a possible design, determining a first acceleration factor based on the update weight, and calculating a second acceleration factor according to the first acceleration factor include: Obtain an update weight factor, and use the product of the update weight factor and the update weight as the first acceleration factor; Calculating a difference between the updated weight factor and the first acceleration factor, and using the difference as the second acceleration factor; Correspondingly, using the update weight, the first acceleration factor, the second acceleration factor, the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any individual particle at the sth iteration, an updated velocity vector corresponding to any individual particle is calculated, which includes: According to the following formula (3), the updated velocity vector corresponding to any particle individual is calculated; (3) In the above formula (3), represents the updated velocity vector corresponding to any particle individual, represents the velocity vector of any particle individual at the sth iteration, represents the update weight, represents the first acceleration factor, represents the second acceleration factor, represent the first update operator and the second update operator in turn, represents the individual perturbation solution, represents the global perturbation solution, represents the number of states in the position vector of any particle individual at the sth iteration, Represent the third random number and the fourth random number respectively.

[0015] In a second aspect, a self-detection device for a power converter is provided, comprising: An acquisition unit, used to acquire operating parameters of the target power converter during the self-detection process, wherein the operating parameters include a first voltage, a second voltage, and a third voltage, and the first voltage is the output voltage of the target power converter, the second voltage is the output voltage of the inverter circuit in the target power converter, and the third voltage is the voltage across the freewheeling tube in the target power converter; A model training unit, used for training a hidden Markov model based on a historical operating parameter set of a target power converter and using a power converter fault identification training algorithm, so as to obtain a fault detection model after the training is completed, wherein the power converter fault identification training algorithm is used for optimizing the optimal state number of the hidden Markov model and the optimal model parameters corresponding to the optimal state number by using an evolutionary computing method, and each state in the optimal state number of the hidden Markov model corresponds to an operating state of the target power converter; The self-detection unit is used to input the operating parameters into the fault detection model to perform fault detection processing, so as to obtain a self-detection result of the target power converter after the fault detection processing.

[0016] In a third aspect, another self-detection device for a power converter is provided. Taking the device as an electronic device as an example, the device includes a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the self-detection method of the power converter as in the first aspect or any possible design of the first aspect.

[0017] In a fourth aspect, a storage medium is provided, on which instructions are stored. When the instructions are executed on a computer, the self-detection method of the power converter as in the first aspect or any possible design of the first aspect is executed.

[0018] In a fifth aspect, a computer program product comprising instructions is provided, which, when executed on a computer, causes the computer to execute the self-detection method of the power converter as in the first aspect or any possible design of the first aspect.

[0019] Beneficial effects: (1) The present invention performs fault self-detection of the power converter by extracting operating data used to characterize the working status of different components in the converter, and at the same time constructs a hidden Markov model trained using its historical operating data; thus, the operating data of different components of the power converter are input into the trained hidden Markov model, and the real-time operating status of the power converter can be obtained. After the real-time operating status is obtained, the fault self-detection of the power converter can be completed; thus, the present invention performs fault self-detection based on the operating data used to characterize the working status of different components. Compared with the traditional technology, it can effectively evaluate the overall performance of the power converter, thereby realizing effective fault detection of the power converter. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic flow chart of the steps of a self-detection method for a power converter provided by an embodiment of the present invention; Figure 2 A schematic diagram of the specific structure of a voltage acquisition circuit provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a self-detection device for a power converter provided by an embodiment of the present invention; Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0022] It should be understood that although the terms first, second, etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another unit. For example, a first unit can be referred to as a second unit, and similarly, a second unit can be referred to as a first unit without departing from the scope of the exemplary embodiments of the present invention.

[0023] It should be understood that the term "and / or" that may appear in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" that may appear in this article describes another type of association object relationship, indicating that two relationships may exist. For example, A / and B can represent two situations: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0024] Example: See also Figure 1 As shown, the self-detection method of the power converter provided in this embodiment constructs a fault detection model based on the hidden Markov model, and at the same time uses the operating data used to characterize the working status of different components in the power converter as model input data; in this way, the operating data of the different components extracted above are input into the fault detection model, and the self-detection result of the power converter can be obtained; thus, compared with the traditional technology, the method can effectively evaluate the overall performance of the power converter, thereby realizing effective fault detection of the power converter; therefore, it is very suitable for large-scale application and promotion; among them, for example, the method can be but not limited to running on the self-detection end side, optionally, the self-detection end can be but not limited to a processor, it can be understood that the aforementioned execution subject does not constitute a limitation on the embodiments of the present application, and accordingly, the operation steps of the method can be but not limited to the following steps S1 to S3.

[0025] S1. Obtaining operating parameters of the target power converter during the self-detection process, wherein the operating parameters include a first voltage, a second voltage, and a third voltage, and the first voltage is the output voltage of the target power converter, the second voltage is the output voltage of the inverter circuit in the target power converter, and the third voltage is the voltage across the freewheeling tube in the target power converter; in this embodiment, the target power converter may include, but is not limited to, a bidirectional full-bridge DC-DC converter, and specifically, the main circuit structure of the bidirectional full-bridge DC-DC converter is: after the three-phase input power passes through the three-phase rectifier circuit and the LC filter circuit, it is converted into a DC power supply and transmitted to the full-bridge inverter circuit, and after being processed by the full-bridge inverter circuit, the required AC power supply is obtained, which is converted into a DC power supply through a full-wave rectifier circuit and finally transmitted to the load; further, the full-bridge inverter circuit usually includes 4 switching tubes, which convert the DC power supply after rectification and filtering into an AC square wave of a preset frequency, and then, a full-wave rectifier circuit is formed by two Schottky diodes and a freewheeling tube to convert the AC square wave into a DC, and finally, it can be transmitted to the load through the output filter circuit.

[0026] Thus, through the aforementioned circuit structure analysis, it can be known that the aforementioned collected voltage across the freewheeling tube contains the fault information of the rectifier element in the converter, the output voltage of the entire converter contains the fault information of the internal filter, and the output voltage of the inverter circuit (i.e., the full-bridge inverter circuit) contains the fault information of the switching tube; thus, this embodiment is equivalent to collecting operating data used to characterize the working status of different components in the converter to perform fault self-detection of the target power converter; based on this, the aforementioned operating data used to characterize the working status of different components in the converter can be used to comprehensively evaluate the overall performance and operating status of the target power converter.

[0027] Furthermore, the following discloses one of the aforementioned operating data collection circuits, namely, a voltage collection circuit; In this embodiment, the voltage acquisition circuit may include, but is not limited to: a first operational amplifier B1, a second operational amplifier B2, and a third operational amplifier B3; wherein the connection structure of the above three operational amplifiers is: See also Figure 2 As shown, the non-inverting input terminal of the first operational amplifier B1 is electrically connected to one end of the first resistor R3, and the other end of the first resistor R3 serves as a sampling end and is electrically connected to the device under test (i.e., the output end of the freewheeling tube, the full-bridge inverter circuit, and the output end of the entire target power converter); wherein the other end of the first resistor R3 is also grounded through the second resistor R2.

[0028] At the same time, the inverting input terminal of the first operational amplifier B1 is electrically connected to one end of the first capacitor C2 and one end of the third resistor R9, respectively, wherein the other end of the first capacitor C2 and the other end of the third resistor R9 are both electrically connected to the output terminal of the first operational amplifier B1, and the output terminal of the first operational amplifier B1 is also electrically connected to the inverting input terminal of the second operational amplifier B2 through the fourth resistor R8.

[0029] In this embodiment, the inverting input terminal of the third operational amplifier B3 is electrically connected to one end of the fifth resistor R7, and one end and the resistance adjustment end of the first adjustable resistor Rq1, wherein the other end of the fifth resistor R7 is electrically connected to the negative electrode of the voltage regulator DS1, one end of the second capacitor C1 and one end of the sixth resistor R1, respectively, the other end of the sixth resistor R1 is electrically connected to a 6V power supply, and the positive electrode of the voltage regulator DS1 and the other end of the second capacitor C1 are grounded respectively.

[0030] The non-inverting input terminal of the third operational amplifier B3 is grounded through the seventh resistor R5, wherein the output terminal of the third operational amplifier B3 is electrically connected to the other end of the first adjustable resistor Rq1 and one end of the eighth resistor R6, and the other end of the eighth resistor R6 is electrically connected to the inverting input terminal of the second operational amplifier B2, and one end and the adjustment end of the second adjustable resistor Rq2.

[0031] Among them, the non-inverting input terminal of the second operational amplifier B2 is respectively electrically connected to the anode of the first diode D1, the cathode of the second diode D2 and one end of the ninth resistor R4, wherein the cathode of the first diode D1 and the anode of the second diode D2 are respectively electrically connected to the non-inverting input terminal of the second operational amplifier, and the other end of the ninth resistor R4 is grounded.

[0032] The other end of the second adjustable resistor Rq2 is electrically connected to the output end of the second operational amplifier B2, wherein the output end of the second operational amplifier B2 serves as the output end of the entire voltage collection circuit to output the collected voltage signal.

[0033] Thus, through the aforementioned circuit explanation, this embodiment uses the first operational amplifier B1 as a voltage follower, and simultaneously has the functions of buffering, isolation, and increasing load capacity; then, a 6V power supply is used to charge the second capacitor C1, and a voltage regulator DS1 is used to generate a corresponding reference voltage, and reverse amplification processing is performed. At this time, the reference voltage after reverse amplification processing and the sampling voltage can be amplified by the second operational amplifier B2 and then output; based on this, the aforementioned circuit can be used to realize the detection of the voltages of different components in the target power converter, thereby obtaining the aforementioned operating parameters.

[0034] Optionally, for example, a self-detection circuit can be set on the target power converter, and the self-detection circuit includes the aforementioned self-detection terminal (ie, the processor) and a voltage acquisition circuit; in this way, the self-detection of the target power converter can be achieved with the help of the self-detection circuit.

[0035] After obtaining the operating parameters of the target power converter, this embodiment constructs a fault detection model trained based on its historical operating parameters as a training set, wherein the model is obtained by training a hidden Markov model using the historical operating parameter set; therefore, by determining the optimal state number and optimal model parameters of the hidden Markov model, self-detection of the operating state of the target power converter can be achieved based on the aforementioned operating parameters; wherein the model training process can be but is not limited to as shown in the following step S2.

[0036] S2. Based on the historical operating parameter set of the target power converter, and using the power converter fault identification training algorithm, the hidden Markov model is trained to obtain a fault detection model after the training is completed, wherein the power converter fault identification training algorithm is used to use evolutionary computing to optimize the optimal state number of the hidden Markov model and the optimal model parameters corresponding to the optimal state number, and each state in the optimal state number of the hidden Markov model corresponds to an operating state of the target power converter; in specific implementation, the state of the hidden Markov model is the state recognition result corresponding to its input observation sequence, wherein, if it is assumed that the number of states of the hidden Markov model is 4, the 4 states represent the normal state, the fault state, the component slightly degraded state, and the component severely degraded state respectively, then the operating state obtained by model identification can be the normal state, the fault state, the component slightly degraded state or the component severely degraded state; based on this, determining the optimal state number of its model is of great significance to the operating state identification of the target power converter.

[0037] Among them, this embodiment uses the improved particle swarm algorithm to optimize the optimal state number of the hidden Markov model, and in the optimization process, determines the optimal model parameters corresponding to each state number; in this way, after the optimization is completed, the optimal state number of the hidden Markov model and the optimal model parameters corresponding to the optimal state number can be obtained.

[0038] Optionally, the optimization process of the optimal number of states and optimal model parameters of the hidden Markov model may include but is not limited to the following steps S201 to S211.

[0039] S201. Initializing the state number of the hidden Markov model to obtain multiple initial state numbers, and determining the initial model parameters of the hidden Markov model at each initial state number according to each initial state number, wherein each initial state corresponding to any initial state number is used to characterize an operating state of the target power converter; in this embodiment, multiple initial state numbers are pre-set, such as setting a minimum state number and a maximum state number, and taking multiple values ​​between the two to obtain multiple initial state numbers. At the same time, when setting different state numbers, it is also necessary to determine the corresponding state of each state. For example, if the minimum state number is set to 2, then when there are two state numbers, the corresponding operating states are the normal state and the fault state. For another example, if the initial state number is 3, then the corresponding operating states are the normal state, the fault state and the component slightly degraded state; for another example, if the initial state number is 4, then the corresponding operating states are the normal state, the fault state, the component slightly degraded state and the component severely degraded state; of course, when the state number is greater than or equal to 4, the state can be subdivided, that is, the fault state can be divided into switch tube fault, rectifier device fault, filter fault, etc., which will not be repeated here.

[0040] In this way, after obtaining multiple initial state numbers, the initial model parameters of the hidden Markov model under each initial state number can be determined, where the initial model parameters can include but are not limited to a state distribution matrix, a state transition probability matrix and an observation value output probability matrix, and the observation value output probability matrix is ​​obtained based on the state distribution matrix and the state transition probability matrix; of course, determining the initial model parameters according to the state number is a common technique for training hidden Markov models, which will not be repeated here.

[0041] Based on this, after completing the initialization of the model state number, the state number and initial model parameters obtained by the above initialization can be used to generate a particle population, that is, to generate multiple individual particles; wherein the generation process of the individual particles can be but is not limited to as shown in the following step S202.

[0042] S202. Based on each initial state number and the initial model parameters corresponding to each initial state number, multiple particle individuals are generated, wherein each particle individual is configured with an initial position vector and an initial velocity vector, and each initial position vector contains an initial state number and the initial model parameters corresponding to the initial state number; in this embodiment, the number of particle individuals is the same as the total number of initial state numbers, that is, one particle individual corresponds to one initial state number, that is, the initial position vector of the particle individual contains one initial state number and the corresponding initial model parameters.

[0043] Thus, after generating a plurality of individual particles, the evolutionary computation method of the individual particles can be used to search for the optimal number of states of the model, and the process can be but is not limited to the following steps S203 to S211.

[0044] S203. Initialize the number of iterations s to 1, and determine the position vector and velocity vector of each individual particle at the s-th iteration, wherein when s is 1, the position vector and velocity vector of any individual particle at the s-th iteration are the initial position vector and initial velocity vector of the individual particle; in this embodiment, after obtaining the position vector and velocity vector of each individual particle at the s-th iteration, the initial hidden Markov model corresponding to each individual particle can be trained, and the process can be but not limited to the following step S204.

[0045] S204. An initial hidden Markov model of each particle individual at the s-th iteration is established according to the model parameters in the position vector of each particle individual at the s-th iteration, and a training set of each particle individual at the s-th iteration is determined from the historical operating parameter set according to the state number in the position vector of each particle individual at the s-th iteration, wherein the training set of any particle individual at the s-th iteration contains the historical operating parameters of the target power converter under the target operating state, and the target operating state is various operating states corresponding to the state number in the position vector of any particle individual at the s-th iteration.

[0046] In this embodiment, the aforementioned step S204 is explained by taking an example. Assuming that the initial state number in the initial position vector corresponding to the particle individual A1 is 3, then the initial hidden Markov model of the particle individual A1 can be established based on the initial state number 3 and its corresponding initial model parameters. At this time, the initial parameters of the model have been determined, therefore, it is necessary to determine its corresponding training data.

[0047] Optionally, as explained above, the state numbers of the hidden Markov model correspond to an operating state of the target power converter respectively. Therefore, the historical operating parameters corresponding to its various states can be determined from the historical operating parameter set based on the state number of the initial hidden Markov model corresponding to the particle individual A1. Specifically, it is explained based on the above examples, that is, obtaining the historical operating parameters of the power converter under the operating states corresponding to the three model states.

[0048] Further, it is assumed that the three states of the initial hidden Markov model correspond to the normal state, the fault state and the slightly degraded state of the component respectively; at this time, the voltages at both ends of the freewheeling tube, the voltage output by the inverter circuit and the voltage output by the converter when the power converter is operating normally are respectively screened out from the historical operating parameter set (of course, the data is multiple, and can be data in multiple observation time periods, that is, data when the operation is normal at different times); similarly, the operating data of the power converter when it is faulty is screened out from the historical operating parameter set; and for the extraction of historical operating parameters when the component is degraded, the following method is used: In specific applications, the power converter mainly includes components such as resistors, inductors, diodes, capacitors, etc. Therefore, this embodiment uses equivalent resistance to analyze the component degradation process, and the equivalent relationship is as follows: (4) In the above formula (4), represents the equivalent resistance of the target power converter at time t, represents the initial value, represents a constant, represents the temperature, where Less than or equal to 1.2 When , it is determined that the target power converter is in normal operation; when Greater than 1.2 , and less than or equal to 2 When , it is determined that the target power converter is in a state of slight component degradation; and when Greater than 2 , and less than or equal to 3 When , it is determined that the target power converter is in a state of severe component degradation; and when Greater than 3 , it is determined that the target power supply is in a component failure state (which is also a type of fault state).

[0049] Therefore, based on the above formula (4), the equivalent resistance value of each historical operating parameter at the corresponding operating time and temperature can be calculated, that is, when the power converter collects each historical operating parameter, it will record the corresponding operating time and temperature; in this way, the operating time (the operating time refers to the operating time) and temperature corresponding to each historical operating parameter are substituted into the above formula (4), and the state of the components of the power converter when collecting each historical operating parameter can be obtained.

[0050] Therefore, based on the above formula (4), the historical operating parameters used to characterize the power converter in a state of slight component degradation can be screened out from the historical operating parameter set; then, the historical operating parameters in normal operation and the historical operating parameters in fault state can be combined to form a training set of the initial hidden Markov model corresponding to the particle individual A1.

[0051] Of course, when the number of states is any other value, the process of obtaining the training set is the same as the above example; at the same time, when the number of states is greater than 4, that is, the specific fault state is subdivided, such as switch tube fault, rectifier device fault, filter fault, component failure fault, etc., then, the operating parameters of the switch tube fault, the operating parameters of the rectifier fault, and the operating parameters of the component failure are extracted from the historical operating parameter set (the component failure judgment is also obtained through the above formula (4)); therefore, through the above method, the corresponding historical operating parameters can be obtained for each state of the initial hidden Markov model of each particle individual, thereby forming a training set of the initial hidden Markov model corresponding to each particle individual.

[0052] Next, the aforementioned training set may be used to train the initial hidden Markov model corresponding to the particle individual A1, wherein the training process may be, but is not limited to, as shown in the following step S205.

[0053] S205. Using the training set of each particle individual at the s-th iteration, train the initial hidden Markov model of each particle individual at the s-th iteration, and after the training, based on the output of the trained initial hidden Markov model corresponding to each particle individual, obtain the fitness of each particle individual at the s-th iteration; in specific implementation, for example, but not limited to, the Viterbi algorithm and the Baum-Welch algorithm (forward-backward algorithm) can be used to train the model; of course, the aforementioned technology is a commonly used training algorithm for hidden Markov models, and its principle will not be repeated; at the same time, since the training set of each particle individual contains the historical operating parameters of the power converter under different operating states; therefore, according to the aforementioned hidden Markov model training method, the training set will be divided into sub-training sets with the same number as the number of model states during training; based on this, after each initial hidden Markov model is trained, a model with the same number as the number of states will be obtained.

[0054] In this way, after completing the training of the initial hidden Markov model corresponding to the particle individual A1 using the aforementioned algorithm, the optimal model parameters of the initial hidden Markov model corresponding to the particle individual A1 at the sth iteration can be obtained; of course, the training process of the initial hidden Markov model corresponding to the remaining particle individuals is also the same, which will not be repeated here.

[0055] At the same time, after completing the training of the model, the output of each initial hidden Markov model can be used to calculate the log-likelihood loss of the corresponding conditional probability, so as to determine the fitness of the individual particles corresponding to each initial hidden Markov model based on this; wherein, the log-likelihood loss quantifies the accuracy of the classifier by penalizing incorrect classification, and minimizing the log-likelihood loss is equivalent to maximizing the accuracy of the classifier; thus, this embodiment uses the log-likelihood loss of the conditional probability of each initial hidden Markov model as the fitness of each individual particle; and after obtaining the fitness of each individual particle at the s-th iteration, the global optimal solution and the individual optimal solution can be updated according to the aforementioned fitness, and the process can be but not limited to as shown in the following step S206.

[0056] S206. Update the global optimal solution and the individual optimal solution of each particle according to the fitness of each individual particle, wherein the global optimal solution is the position vector corresponding to the particle with the largest fitness during the 1st to sth iterations; in this embodiment, if the maximum fitness at the sth iteration is greater than the fitness of the particle corresponding to the global optimal solution at the s-1th iteration, the position vector of the particle corresponding to the maximum fitness at the sth iteration is used as the updated global optimal solution, otherwise, no update is performed; of course, the updating process of the individual optimal solution of each individual particle is the same, which will not be elaborated here.

[0057] After the global optimal solution and the individual optimal solutions of each particle are updated, it can be determined whether the optimization search can be stopped. The determination process can be, but is not limited to, as shown in the following step S207.

[0058] S207. Determine whether the iteration stop condition is met; in this embodiment, the iteration stop condition may be, but is not limited to, the fitness of the individual particle corresponding to the updated global optimal solution is less than or equal to the fitness threshold, or s reaches the maximum number of iterations; wherein, if the aforementioned iteration stop condition is met, then the number of states in the global optimal solution when the iteration stop condition is met is directly used as the optimal state number; then, the model parameters of the trained initial hidden Markov model of the individual particle corresponding to the global optimal solution when the iteration stop condition is met are used as the optimal model parameters corresponding to the optimal state number.

[0059] On the contrary, if the aforementioned iteration stop condition is not met, it is necessary to update the position vector and velocity vector of each individual particle so as to continue the optimization by utilizing the evolution of the individual particles; wherein, the updating process of the position vector and the velocity vector may be, but is not limited to, as shown in the following steps S208 and S209.

[0060] S208. If not, then update the velocity vector corresponding to each individual particle according to the updated global optimal solution and each updated individual optimal solution to obtain an updated velocity vector corresponding to each individual particle; in specific implementation, take any individual particle as an example to illustrate the updating process of the velocity vector, which can be but not limited to the following steps S208a to S208d.

[0061] S208a. For any individual particle, generate a first random number and a second random number, wherein the first random number and the second random number are different from each other and both are in [0,1]. In this embodiment, after the first random number and the second random number are generated, two update operators can be generated according to the first and second random numbers, and the global optimal solution and the individual optimal solution can be perturbed thereby, so as to solve the problems of slow evolution speed and low precision when the individual particle converges to the local extreme value. The calculation process of the two update operators can be, but is not limited to, as shown in the following step S208b.

[0062] S208b. Obtain a maximum number of iterations, and generate a first update operator according to the maximum number of iterations and the first random number, and generate a second update operator according to the maximum number of iterations and the second random number. In this embodiment, the first update operator and the second update operator are calculated in the same manner, and the first update operator is taken as an example for explanation. The example may be, but is not limited to, calculating the first update operator according to the following formula (1).

[0063] (1) In the above formula (1), represents the first update operator, represents the maximum number of iterations, Represents the first random number.

[0064] Thus, based on the aforementioned formula (1), the aforementioned first update operator can be calculated; similarly, by replacing the first random number in the aforementioned formula (1) with the second random number, the second update operator can be calculated.

[0065] Based on this, after the first update operator and the second update operator are calculated based on the above formula (1), the global optimal solution and the individual optimal solution can be disturbed based on this, and the process can be but not limited to as shown in the following step S208c.

[0066] S208c. Based on the first update operator and the updated individual optimal solution corresponding to any individual particle, the individual perturbation solution of any individual particle at the s-th iteration is calculated, and based on the second update operator and the updated global optimal solution, the global perturbation solution at the s-th iteration is calculated; in specific implementation, the product between the first update operator and the updated individual optimal solution corresponding to any individual particle (specifically, the product between the first update operator and the number of states in the updated individual optimal solution corresponding to any individual particle is calculated) is used as the individual perturbation solution of any individual particle at the s-th iteration; similarly, the product between the second update operator and the updated global optimal solution is used as the global perturbation solution.

[0067] Thus, after obtaining the individual perturbation solution and the global perturbation solution based on the aforementioned step S208c, the velocity vector corresponding to any individual particle can be updated, and the process can be but not limited to as shown in the following step S208d.

[0068] S208d. According to the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any particle individual at the sth iteration, an updated velocity vector corresponding to any particle individual is calculated; in a specific implementation, the updating process of the velocity vector is as follows: S208d1: Calculate the update weight according to the updated global optimal solution and the position vector of each particle individual at the s-th iteration; in the present embodiment, it has been explained above that the model parameters in the position vector of any particle individual at the s-th iteration include: the state distribution matrix, state transition probability matrix and observation value output probability matrix of the initial hidden Markov model corresponding to any particle individual; therefore, the update weight can be calculated based on the aforementioned model parameters and the following steps, for example but not limited to.

[0069] Step 1: Based on the position vector of each particle individual at the sth iteration and the updated global optimal solution, the distance between each particle individual and the target individual is calculated according to the following formula (2), where the target individual is the particle individual corresponding to the updated global optimal solution.

[0070] (2) In the above formula (2), Indicates the number of particles in each individual The distance between each particle and the target individual, Sequentially indicates The state transfer probability matrix, observation value output probability matrix and state distribution matrix in the position vector of each particle at the sth iteration are: represent the state transition probability matrix, observation value output probability matrix and state distribution matrix in the updated global optimal solution in turn, Indicates the The number of states in the position vector of the particle individual at the sth iteration, represents a constant, with a value of 2, where ,and represents the total number of individual particles, Represents the modulo operator.

[0071] Thus, according to the above formula (2), the distance between each particle individual and the target individual can be calculated, and then the average distance can be calculated. The process is shown in the second step below.

[0072] Step 2: Based on the distances between each particle individual and the target individual, the average distance at the sth iteration is obtained; in this embodiment, the distances between each particle individual and the target individual are summed up and then divided by the total number of particle individuals to obtain the average distance at the sth iteration.

[0073] After obtaining the average distance at the sth iteration, the average distance in each historical iteration process can be obtained, and the process is shown in the third step below.

[0074] Step 3: Obtain all average distances during the 1st to sth iterations, and select the largest average distance from all the obtained average distances; in this embodiment, assuming that s is 5 at this time, then the average distance during the first iteration, the average distance during the second iteration, the average distance during the third iteration, the average distance during the fourth iteration, and the average distance during the fifth iteration are obtained; then, the largest average distance among the 5 average distances is selected to calculate the updated weight; wherein, the calculation process of the updated weight can be but is not limited to as shown in the following step 4.

[0075] Step 4: The ratio between the average distance at the sth iteration and the maximum average distance is used as the update weight.

[0076] Thus, after the update weights are calculated based on the first to fourth steps, two acceleration factors can be calculated based on the update weights, and the process is shown in the following step S208d2.

[0077] S208d2. Based on the updated weight, a first acceleration factor is determined, and based on the first acceleration factor, a second acceleration factor is calculated; in specific implementation, for example, but not limited to, the updated weight factor is first obtained; then, the product of the updated weight factor and the updated weight is used as the first acceleration factor; then, the difference between the updated weight factor and the first acceleration factor is calculated; finally, the difference can be used as the second acceleration factor; wherein, in this embodiment, the updated weight factor can be, but not limited to, 2.

[0078] After calculating the two acceleration factors, the velocity vector of any individual particle can be updated by combining the aforementioned update weights, two update operators, the global perturbation solution and the individual perturbation solution. The process can be but is not limited to that shown in the following step S208d3.

[0079] S208d3. Calculate an updated velocity vector corresponding to any individual particle by using the update weight, the first acceleration factor, the second acceleration factor, the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any individual particle at the sth iteration.

[0080] In a specific implementation, the updated velocity vector corresponding to any particle individual may be calculated by, for example but not limited to, using the following formula (3).

[0081] (3) In the above formula (3), represents the updated velocity vector corresponding to any particle individual, represents the velocity vector of any particle individual at the sth iteration, represents the update weight, represents the first acceleration factor, represents the second acceleration factor, represent the first update operator and the second update operator in turn, represents the individual perturbation solution, represents the global perturbation solution, represents the number of states in the position vector of any particle individual at the sth iteration, represent the third random number and the fourth random number respectively; in this embodiment, the value interval of the third random number and the fourth random number is also [0,1], and the initial velocity vector of any individual particle is a constant.

[0082] Therefore, after completing the update of the velocity vector of any individual particle through the aforementioned steps S208d1 to S208d3, the position vector can be updated; wherein, this embodiment only updates the state number in the position vector, and the process can be but is not limited to the following step S209.

[0083] S209. Using the state number in the position vector of each particle individual at the s-th iteration and the updated velocity vector corresponding to each particle individual, the state number in the position vector of each particle individual at the s-th iteration is updated to obtain the updated state number corresponding to each particle individual. In the specific implementation, any particle individual is taken as an example for explanation, wherein the example may be but is not limited to using the following formula (5) to update the state number of any particle individual.

[0084] (5) In the above formula (5), represents the updated state number corresponding to any particle individual, Represents the rounding function.

[0085] In this way, based on the above formula (5), the state number in the position vector corresponding to each individual particle can be updated. After the state number is updated, the updated state number can be used to re-determine the model parameters corresponding to the updated state number, so as to form a new position vector based on this. The updating process of the position vector is shown in the following step S210.

[0086] S210. Based on the updated state number corresponding to each individual particle, the model parameters of the initial hidden Markov model corresponding to each individual particle are re-determined, and the updated state number corresponding to each individual particle and the re-determined model parameters are used to form the updated position vector of each individual particle; in this embodiment, assuming that the state number of particle individual A1 before updating is 3 and the state number after updating is 4, then the model parameters corresponding to particle individual A1 when the state number is 4 are obtained. At this time, the state number 4 and its corresponding model parameters are used to form the updated position vector corresponding to the particle individual A1; in this way, the position vector of each individual particle can be updated in the above manner; then, the model training can be performed again, and the stop iteration condition can be judged again until the stop iteration condition is met; wherein, the iterative optimization process can be but is not limited to the following step S211, S211. Increment s by 1, and use the updated position vector and updated velocity vector of each individual particle as the position vector and velocity vector of each individual particle at the s-th iteration, and re-establish the initial hidden Markov model of each individual particle at the s-th iteration based on the model parameters in the position vector of each individual particle at the s-th iteration, until the iteration stop condition is met, and determine the optimal state number and the optimal model parameters corresponding to the optimal state number based on the global optimal solution when the iteration stop condition is met, so as to obtain the fault detection model based on the optimal state number and the optimal model parameters corresponding to the optimal state number.

[0087] In this embodiment, it is assumed that the iteration stop condition is met at the 8th iteration. At this time, the number of states in the global optimal solution obtained in the 8th iteration can be used as the optimal state number; then, the model parameters of the initial hidden Markov model after training of the particle individual corresponding to the global optimal solution obtained in the 8th iteration are used as the optimal model parameters corresponding to the optimal state number; based on this, the optimization of the state number and model parameters of the hidden Markov model can be completed.

[0088] Optionally, for example, in this embodiment, the optimal number of states is set to 4, which respectively correspond to the normal state, the fault state, the component slightly degraded state and the component severely degraded state.

[0089] Therefore, through the aforementioned steps S201 to S211, the optimization of the number of hidden Markov states and model parameters can be completed; at this time, a fault detection model can be constructed based on this; optionally, the initial hidden Markov model after training of the individual particle corresponding to the global optimal solution when the iteration stop condition is met can be used as the fault detection model.

[0090] After the fault detection model is trained, a self-detection process of the target power converter can be performed based on the operating parameters collected in step S1 , and the process can be but is not limited to that shown in the following step S3 .

[0091] S3. Input the operating parameters into the fault detection model for fault detection processing, so as to obtain the self-detection result of the target power converter after the fault detection processing; in this embodiment, it has been explained above that after the hidden Markov model is trained, the same number of models as its state number will be obtained, therefore, the operating parameters are normalized and input into the fault detection model (that is, the number of states is 4, then it corresponds to 4 hidden Markov models); at this time, the similarity probability of each model in each fault detection model is calculated; finally, the state corresponding to the model with the largest similarity probability is taken as the operating state of the target power converter, that is, the self-detection result; in this way, the self-detection result is that the target power converter is in a normal operating state, in a faulty operating state, in a slightly degraded component state, or in a severely degraded component state.

[0092] Therefore, through the self-detection method of the power converter described in detail in the aforementioned steps S1 to S3, the present invention constructs a fault detection model based on the hidden Markov model, and at the same time uses the operating data used to characterize the working states of different components in the power converter as model input data; in this way, the operating data of the different components extracted above are input into the fault detection model, and the self-detection result of the power converter can be obtained; therefore, compared with traditional technologies, the present invention can effectively evaluate the overall performance of the power converter, thereby realizing effective fault detection of the power converter; therefore, it is very suitable for large-scale application and promotion.

[0093] like Figure 3 As shown, the second aspect of this embodiment provides a hardware device for implementing the self-detection method of the power converter described in the first aspect of the embodiment, including: An acquisition unit is used to acquire operating parameters of a target power converter during a self-detection process, wherein the operating parameters include a first voltage, a second voltage, and a third voltage, and the first voltage is the output voltage of the target power converter, the second voltage is the output voltage of an inverter circuit in the target power converter, and the third voltage is the voltage across a freewheeling tube in the target power converter.

[0094] A model training unit is used to train a hidden Markov model based on a historical operating parameter set of a target power converter and using a power converter fault identification training algorithm to obtain a fault detection model after the training is completed, wherein the power converter fault identification training algorithm is used to use an evolutionary computing method to optimize the optimal state number of the hidden Markov model and the optimal model parameters corresponding to the optimal state number, and each state in the optimal state number of the hidden Markov model corresponds to an operating state of the target power converter.

[0095] The self-detection unit is used to input the operating parameters into the fault detection model to perform fault detection processing, so as to obtain a self-detection result of the target power converter after the fault detection processing.

[0096] The working process, working details and technical effects of the device provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0097] like Figure 4 As shown, the third aspect of this embodiment provides another self-detection device for a power converter. Taking the device as an electronic device as an example, the device includes: a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the self-detection method for the power converter as described in the first aspect of the embodiment.

[0098] For example, the memory may include, but is not limited to, random access memory (RAM), read only memory (ROM), flash memory, first input first output (FIFO) and / or first in last out (FILO), etc. Specifically, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). At the same time, the processor may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state.

[0099] In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. For example, the processor may not be limited to a microprocessor of the STM32F105 series, a reduced instruction set computer (RISC) microprocessor, an X86 architecture processor, or a processor with an integrated embedded neural network processing unit (NPU); the transceiver may be, but is not limited to, a wireless fidelity (WIFI) wireless transceiver, a Bluetooth wireless transceiver, a General Packet Radio Service (GPRS) wireless transceiver, a ZigBee protocol (a low-power LAN protocol based on the IEEE802.15.4 standard, ZigBee) wireless transceiver, a 3G transceiver, a 4G transceiver, and / or a 5G transceiver. In addition, the device may also include, but is not limited to, a power module, a display screen, and other necessary components.

[0100] The working process, working details and technical effects of the electronic device provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0101] The fourth aspect of this embodiment provides a storage medium storing instructions including the self-detection method of the power converter described in the first aspect of the embodiment, that is, the storage medium stores instructions, and when the instructions are run on a computer, the self-detection method of the power converter described in the first aspect of the embodiment is executed.

[0102] The storage medium refers to a carrier for storing data, which may include but is not limited to a floppy disk, a CD, a hard disk, a flash memory, a USB flash drive and / or a memory stick, and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0103] The working process, working details and technical effects of the storage medium provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0104] A fifth aspect of this embodiment provides a computer program product comprising instructions, which, when executed on a computer, causes the computer to execute the self-detection method for the power converter as described in the first aspect of the embodiment, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0105] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A self-detection method for a power converter, characterized in that: include: Acquire operating parameters of the target power converter during the self-detection process, wherein the operating parameters include a first voltage, a second voltage, and a third voltage, and the first voltage is the output voltage of the target power converter, the second voltage is the output voltage of the inverter circuit in the target power converter, and the third voltage is the voltage across the freewheeling tube in the target power converter; Based on the historical operating parameter set of the target power converter, and using the power converter fault identification training algorithm, the hidden Markov model is trained to obtain a fault detection model after the training is completed, wherein the power converter fault identification training algorithm is used to optimize the optimal state number of the hidden Markov model and the optimal model parameters corresponding to the optimal state number by using an evolutionary computing method, and each state in the optimal state number of the hidden Markov model corresponds to an operating state of the target power converter; The operating parameters are input into the fault detection model for fault detection processing, so as to obtain a self-detection result of the target power converter after the fault detection processing.

2. The method according to claim 1, characterized in that Based on the historical operating parameter set of the target power converter, and using the power converter fault identification training algorithm, the hidden Markov model is trained, including: Initializing the state number of the hidden Markov model to obtain a plurality of initial state numbers, and determining the initial model parameters of the hidden Markov model at each initial state number according to each initial state number, wherein each initial state corresponding to any initial state number is respectively used to characterize an operating state of the target power converter; Based on each initial state number and the initial model parameter corresponding to each initial state number, a plurality of individual particles are generated, wherein each individual particle is configured with an initial position vector and an initial velocity vector, and each initial position vector includes an initial state number and the initial model parameter corresponding to the initial state number; Initialize the number of iterations s to 1, and determine the position vector and velocity vector of each particle individual at the s-th iteration, wherein when s is 1, the position vector and velocity vector of any particle individual at the s-th iteration are the initial position vector and initial velocity vector of any particle individual; According to the model parameters in the position vector of each particle individual at the s-th iteration, an initial hidden Markov model of each particle individual at the s-th iteration is established, and according to the state number in the position vector of each particle individual at the s-th iteration, a training set of each particle individual at the s-th iteration is determined from the historical operation parameter set, wherein the training set of any particle individual at the s-th iteration contains the historical operation parameters of the target power converter under the target operation state, and the target operation state is various operation states corresponding to the state number in the position vector of any particle individual at the s-th iteration; Using the training set of each particle individual at the s-th iteration, the initial hidden Markov model of each particle individual at the s-th iteration is trained, and after the training, the fitness of each particle individual at the s-th iteration is obtained based on the output of the trained initial hidden Markov model corresponding to each particle individual; According to the fitness of each individual particle, update the global optimal solution and the individual optimal solution of each individual particle, wherein the global optimal solution is the position vector corresponding to the individual particle with the largest fitness during the 1st to sth iterations; Determine whether the iteration stop condition is met; If not, then according to the updated global optimal solution and each updated individual optimal solution, the velocity vector corresponding to each individual particle is updated to obtain the updated velocity vector corresponding to each individual particle; Using the state number in the position vector of each particle individual at the s-th iteration and the updated velocity vector corresponding to each particle individual, the state number in the position vector of each particle individual at the s-th iteration is updated to obtain the updated state number corresponding to each particle individual; Based on the updated state number corresponding to each individual particle, the model parameters of the initial hidden Markov model corresponding to each individual particle are re-determined, and the updated position vector of each individual particle is formed by using the updated state number corresponding to each individual particle and the re-determined model parameters; s is incremented by 1, and the updated position vector and updated velocity vector of each individual particle are used as the position vector and velocity vector of each individual particle at the s-th iteration, and the initial hidden Markov model of each individual particle at the s-th iteration is reestablished according to the model parameters in the position vector of each individual particle at the s-th iteration, until the iteration stop condition is met, and based on the global optimal solution when the iteration stop condition is met, the optimal state number and the optimal model parameters corresponding to the optimal state number are determined, so as to obtain the fault detection model based on the optimal state number and the optimal model parameters corresponding to the optimal state number.

3. The method according to claim 2, characterized in that According to the updated global optimal solution and each updated individual optimal solution, the velocity vector corresponding to each individual particle is updated to obtain the updated velocity vector corresponding to each individual particle, including: For any individual particle, generate a first random number and a second random number, wherein the first random number and the second random number are different from each other and both are in [0,1]; Obtaining a maximum number of iterations, and generating a first update operator according to the maximum number of iterations and the first random number, and generating a second update operator according to the maximum number of iterations and the second random number; Based on the first update operator and the updated individual optimal solution corresponding to any individual particle, calculate the individual perturbation solution of any individual particle at the s-th iteration, and calculate the global perturbation solution at the s-th iteration according to the second update operator and the updated global optimal solution; According to the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any particle individual at the s-th iteration, an updated velocity vector corresponding to any particle individual is calculated.

4. The method according to claim 3, characterized in that Generating a first update operator according to the maximum number of iterations and the first random number, comprising: According to the following formula (1), the first update operator is calculated; （1） In the above formula (1), represents the first update operator, represents the maximum number of iterations, represents the first random number; Correspondingly, based on the first update operator and the updated individual optimal solution corresponding to any individual particle, calculating the individual perturbation solution of any individual particle at the s-th iteration includes: The product of the first update operator and the updated individual optimal solution corresponding to any individual particle is used as the individual perturbation solution of any individual particle at the s-th iteration.

5. The method according to claim 3, characterized in that: According to the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any particle individual at the s-th iteration, an updated velocity vector corresponding to any particle individual is calculated, including: According to the updated global optimal solution and the position vector of each particle at the sth iteration, the update weight is calculated; Based on the update weight, a first acceleration factor is determined, and according to the first acceleration factor, a second acceleration factor is calculated; An updated velocity vector corresponding to any individual particle is calculated using the update weight, the first acceleration factor, the second acceleration factor, the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any individual particle at the s-th iteration.

6. The method according to claim 5, characterized in that The model parameters in the position vector of any particle individual at the sth iteration include: the state distribution matrix, the state transition probability matrix and the observation value output probability matrix of the initial hidden Markov model corresponding to the any particle individual; Among them, according to the updated global optimal solution and the position vector of each particle individual at the sth iteration, the update weight is calculated, including: Based on the position vector of each particle individual at the sth iteration and the updated global optimal solution, and according to the following formula (2), the distance between each particle individual and the target individual is calculated, where the target individual is the particle individual corresponding to the updated global optimal solution; （2） In the above formula (2), Indicates the number of particles in each individual The distance between each particle and the target individual, Sequentially indicates The state transfer probability matrix, observation value output probability matrix and state distribution matrix in the position vector of each particle at the sth iteration are: represent the state transition probability matrix, observation value output probability matrix and state distribution matrix in the updated global optimal solution in turn, Indicates the The number of states in the position vector of the particle individual at the sth iteration, represents a constant, where ,and represents the total number of individual particles, Represents the modulo operator; Based on the distance between each particle individual and the target individual, the average distance at the sth iteration is obtained; Obtain all average distances from the 1st to the sth iteration, and select the largest average distance from all the obtained average distances; The ratio between the average distance at the sth iteration and the maximum average distance is used as the update weight.

7. The method according to claim 5, characterized in that Based on the update weight, determining a first acceleration factor, and calculating a second acceleration factor according to the first acceleration factor, including: Obtain an update weight factor, and use the product of the update weight factor and the update weight as the first acceleration factor; Calculating a difference between the updated weight factor and the first acceleration factor, and using the difference as the second acceleration factor; Correspondingly, using the update weight, the first acceleration factor, the second acceleration factor, the first update operator, the second update operator, the individual perturbation solution, the global perturbation solution, and the position vector and velocity vector of any individual particle at the sth iteration, an updated velocity vector corresponding to any individual particle is calculated, which includes: According to the following formula (3), the updated velocity vector corresponding to any particle individual is calculated; （3） In the above formula (3), represents the updated velocity vector corresponding to any particle individual, represents the velocity vector of any particle individual at the sth iteration, represents the update weight, represents the first acceleration factor, represents the second acceleration factor, represent the first update operator and the second update operator in turn, represents the individual perturbation solution, represents the global perturbation solution, represents the number of states in the position vector of any particle individual at the sth iteration, Represent the third random number and the fourth random number respectively.

8. A self-detection device for a power converter, characterized in that: include: An acquisition unit, used to acquire operating parameters of the target power converter during the self-detection process, wherein the operating parameters include a first voltage, a second voltage, and a third voltage, and the first voltage is the output voltage of the target power converter, the second voltage is the output voltage of the inverter circuit in the target power converter, and the third voltage is the voltage across the freewheeling tube in the target power converter; A model training unit, used for training a hidden Markov model based on a historical operating parameter set of a target power converter and using a power converter fault identification training algorithm, so as to obtain a fault detection model after the training is completed, wherein the power converter fault identification training algorithm is used for optimizing the optimal state number of the hidden Markov model and the optimal model parameters corresponding to the optimal state number by using an evolutionary computing method, and each state in the optimal state number of the hidden Markov model corresponds to an operating state of the target power converter; The self-detection unit is used to input the operating parameters into the fault detection model to perform fault detection processing, so as to obtain a self-detection result of the target power converter after the fault detection processing.

9. A self-detection device for a power converter, characterized in that: include: A memory, a processor and a transceiver which are sequentially communicatively connected, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the self-detection method of the power converter as described in any one of claims 1 to 7.

10. A computer program product comprising instructions, characterized in that When the instructions are executed on a computer, the computer is enabled to execute the self-detection method for a power converter as claimed in any one of claims 1 to 7.

Images