Motor insulation breakdown voltage prediction method based on continuous hidden Markov model

Abstract

The invention discloses a motor insulation breakdown voltage prediction method based on a continuous hidden Markov model, and the method comprises the steps: taking a plurality of groups of obtained experimental data of an insulating material of a motor in different motor operation time periods as a training set, and enabling the experimental data to comprise capacitance, insulation resistance, dielectric loss, maximum discharge capacity and breakdown voltage; constructing a continuous hidden Markov model by taking the test parameters as an observation sequence and taking the discretized breakdown voltage as a hidden state; performing iterative calculation on parameters of the continuous hidden Markov model through a Baum-Welch algorithm based on the training set, and determining optimal parameters of the continuous hidden Markov model; and based on the continuous hidden Markov model of the optimal parameters, determining a hidden state sequence with the maximum probability under the observation sequence at the current moment through a Viterbi algorithm, and determining the breakdown voltage at the current moment based on the hidden state sequence with the maximum probability. According to the invention, the breakdown voltage of the motor insulating material in the motor operation process can be effectively predicted.

Description

Technical Field

[0001] The present application relates to the technical field of motor life, and in particular to a method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model. Background Art

[0002] Motors are important equipment for energy conversion. They are also an important part of industrial automation, renewable energy, transportation and other fields. The development of motors not only has a far-reaching impact on the technical and economic levels, but also plays an increasingly important role in promoting environmental protection and sustainable development. Studying motor failures and analyzing the causes of motor failures can improve the operating reliability of motors, reduce the occurrence of failures, reduce maintenance costs, ensure production safety, extend equipment life, and improve investment returns. Motor failure research is not only related to the normal operation and economic benefits of equipment, but also plays a positive role in improving safety and promoting technological innovation.

[0003] The insulation damage of the motor is one of the important reasons for its failure. During the long-term operation, the motor will be subjected to the combined effects of multiple factors such as electricity, mechanical force and hot and cold conversion. The mechanical and dielectric properties of the stator main insulation will gradually deteriorate and age, eventually leading to insulation failure. If the motor insulation fails, it will not only endanger the generator itself, but also cause the entire system to lose power due to its suddenness, resulting in huge economic losses.

[0004] Breakdown voltage is an important indicator of insulation properties, but once the insulation material is broken down, it will lose its insulation effect, causing significant economic losses. For running motors, the breakdown voltage of the insulation material cannot be measured, so the breakdown voltage of the insulation material cannot be correctly predicted. Summary of the invention

[0005] In order to address the deficiencies of the prior art, the purpose of the present application is to provide a method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model, which can effectively predict the breakdown voltage of the motor insulation material during motor operation.

[0006] To achieve the above objectives, the present application provides a method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model, the method comprising: Taking the insulation material of the motor as the sample to be tested, obtaining multiple groups of experimental data of the sample to be tested in different motor operation time periods, and constructing a training set with the multiple groups of experimental data, wherein the experimental data include the test parameters and breakdown voltage of the sample to be tested, and the test parameters include capacitance, insulation resistance, dielectric loss, and maximum discharge capacity; A continuous hidden Markov model is constructed with the test parameters as the observation sequence and the discretized breakdown voltage as the hidden state. Based on the training set, the parameters of the continuous hidden Markov model are iteratively calculated by the Baum-Welch algorithm to determine the optimal parameters of the continuous hidden Markov model; Based on the continuous hidden Markov model with optimal parameters, the hidden state sequence with maximum probability under the observation sequence at the current moment is determined by Viterbi algorithm, and the breakdown voltage at the current moment is determined based on the hidden state sequence with maximum probability, wherein the observation sequence at the current moment includes the test parameters at the current moment.

[0007] Furthermore, a continuous hidden Markov model is constructed with the test parameters as the observation sequence and the discretized breakdown voltage as the hidden state, including: Each set of test parameters is regarded as an observation sequence O, expressed as ,in, , the observation vector It represents the observed value obtained at time t, including capacitance, insulation resistance, dielectric loss and maximum discharge capacity, and T represents the time length; The obtained multiple sets of breakdown voltages are discretized into N values, and the N values ​​are used to represent N hidden states to form a hidden state set S, which is expressed as ,in, represents the i-th hidden state, and N represents N different hidden states; Construct the actual hidden state sequence Q, expressed as ,in, represents the hidden state value at time t, The value range is ; The parameters of the continuous hidden Markov model include the initial state probability matrix , state transition probability matrix And the observation probability matrix , where the initial state probability Indicates that the hidden state value at the initial moment is The probability of state transition Indicates that the hidden state value at time t is , the hidden state value at time t+1 is The probability of observation Indicates that the hidden state value is When The probability of The Gaussian mixture model is used to fit the observation data and obtain the observation probability density of the output observation value under each hidden state. , where the observed probability density is The parameters of the Gaussian distribution based on the observed value are determined, and the parameters of the Gaussian distribution include the weight of the Gaussian distribution, the mean vector of the Gaussian distribution, and the covariance matrix of the Gaussian distribution.

[0008] Furthermore, the Gaussian mixture model is used to fit the observations to obtain the observation probability density of the output observations under each hidden state. ,include: Four Gaussian distributions are used to describe the data distribution of capacitance, insulation resistance, dielectric loss and maximum discharge capacity respectively. The observation probability density is determined by the weight, mean vector and covariance matrix of each Gaussian distribution. , the observed probability density It is expressed as: ; ; ; Where M represents the number of Gaussian distributions, M is 4, Represents hidden state The weight of the m-th Gaussian distribution, The hidden state is The mean vector of the m-th Gaussian distribution, The hidden state is The covariance matrix of the m-th Gaussian distribution is, Represents the mixed Gaussian probability density function, m=1,2,3,4, corresponding to capacitance, insulation resistance, dielectric loss, and maximum discharge capacity, respectively.

[0009] Furthermore, based on the training set, the parameters of the continuous hidden Markov model are iteratively calculated by the Baum-Welch algorithm to determine the optimal parameters of the continuous hidden Markov model, including: Initialize the initial state probability matrix, state transition probability matrix and observation probability matrix to obtain the initial parameters of the continuous hidden Markov model, and iteratively update the initial parameters of the continuous hidden Markov model. In each iteration, Calculate the forward probability of each hidden state value at each moment within the time length; Calculate the backward probability of each hidden state value at each moment within the time length; Determine the current maximum likelihood probability under the observation sequence using the current parameters of the continuous hidden Markov model; Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, calculate the conditional probability of each hidden state value at any time; Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, calculate the joint conditional probability of the hidden state value at any time and the hidden state value at the next time after the any time; Based on conditional probability and mixed Gaussian probability density function, calculate the joint probability of each hidden state value at any time; Based on the conditional probability, the joint conditional probability and the joint probability, the parameters of the Gaussian distribution, the initial state probability matrix and the state transition probability matrix are updated, and based on the updated parameters of the Gaussian distribution, an updated observation probability matrix is ​​determined; Determining updated parameters of the continuous hidden Markov model based on the updated initial state probability matrix, state transition probability matrix and observation probability matrix; Using the updated parameters of the continuous hidden Markov model, the updated maximum likelihood probability under the observation sequence is determined; If the number of iterations is greater than or equal to the number threshold, and / or the difference between the current maximum likelihood probability and the updated maximum likelihood probability meets the threshold, the iteration is stopped, and the updated parameters of the continuous hidden Markov model of the current iteration are used as the optimal parameters of the continuous hidden Markov model.

[0010] Furthermore, the forward probability of each hidden state value at each moment within the time length is calculated, including: Calculate the forward probability of each hidden state value at the initial moment, and calculate the forward probability of each hidden state at time t , the forward probability It is used to indicate that the hidden state value at time t is , the observation vector from the initial time to time t The probability of forward The calculation formula is: ; ; in, Indicates that the initial hidden state value is The forward probability of represents the initial state probability, Indicates that the initial hidden state value is , the probability that the observation vector is O1, N represents the number of hidden states, Indicates that the hidden state value at time t is The forward probability of Indicates that the hidden state value at time t is , the observation vector is O t The probability of Indicates that the hidden state value at time t-1 is The forward probability of represents the state transition probability; Calculate the backward probability of each hidden state value at each moment in the time length, including: Calculate the backward probability of each hidden state value at the termination time, and calculate the backward probability of each hidden state value at time t , backward probability It is used to indicate that the hidden state value at time t is , the observation vector from time t+1 to time T The probability of backward The calculation formula is: ; ; in, Indicates that the hidden state value at the termination time T is The backward probability of Indicates that the hidden state value at time t is The backward probability of It means that the hidden state value at time t is , the observation vector is O t+1 The probability of N represents the number of hidden states. Indicates that the hidden state value at time t+1 is The backward probability of represents the state transition probability.

[0011] Furthermore, the current maximum likelihood probability under the observation sequence is determined using the current parameters of the continuous hidden Markov model, including: The current maximum likelihood probability is calculated as: ; in, represents the current parameters of the continuous hidden Markov model, Representation parameters The maximum likelihood probability of the observation sequence under Indicates that the hidden state value at the end time T is The forward probability of Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, the conditional probability of each hidden state value at any time is calculated, including: Calculate the conditional probability of each hidden state value at any time It is expressed as: ; in, Indicates that the hidden state value at time t is The conditional probability of Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, the joint conditional probability of the hidden state value at any time and the hidden state value at the next time of the any time is calculated, including: Joint conditional probability It is expressed as: ; in, Indicates that the hidden state value at time t is and transfer to The joint conditional probability of Indicates that the hidden state value is , the observation vector is O t+1 The probability of Indicates that the hidden state value at time t+1 is The backward probability of Indicates that the hidden state value at time t is The forward probability of Based on the conditional probability and the mixed Gaussian probability density function, the joint probability of each hidden state value at any time is calculated, including: Joint Probability The calculation formula is: ; in, Indicates that the hidden state value at time t is The joint probability of the mth Gaussian mixture component, represents the mixed Gaussian probability density function, Represents hidden state The weight of the next m-th Gaussian distribution, M represents the number of Gaussian distributions, M is 4.

[0012] Further, based on the conditional probability, the joint conditional probability and the joint probability, the parameters of the Gaussian distribution, the initial state probability matrix and the state transition probability matrix are updated, and based on the updated parameters of the Gaussian distribution, an updated observation probability matrix is ​​determined, including: Updated initial state probability for: ; in, It means that given the observation sequence O, the hidden state value at the initial moment is The probability of Updated state transition probability for: ; in, Indicates that the hidden state value at time t is The conditional probability of Updated Gaussian distribution weights for: ; in, Indicates that the hidden state value at time t is The joint probability of the sth Gaussian mixture component; Updated mean vector of the Gaussian distribution for: ; Among them, O t represents the observation vector at time t; Covariance matrix of the updated Gaussian distribution for: .

[0013] Furthermore, based on the continuous hidden Markov model with optimal parameters, the hidden state sequence with the maximum probability under the observation sequence at the current moment is determined by the Viterbi algorithm, and the breakdown voltage value at the current moment is determined based on the hidden state sequence with the maximum probability, including: Set the two-dimensional matrix V to store the maximum probability of each hidden state value in each hidden state at each time step, and set Indicates that the hidden state value at time t is The maximum probability, path matrix Represents the hidden state at time t-1 that the hidden state sequence with the maximum probability at time t depends on ; Initialization: At the initial time t=1, the hidden state value is calculated as The maximum probability of , initialize the path matrix to 0: ; ; in, Table hidden state value The initial probability of Indicates that the hidden state value is , the probability of observing vector O1; Recursive calculation: For any time t and any hidden state , calculate the hidden state value at time t The maximum probability for: ; in, Indicates that the hidden state value at time t-1 is The maximum probability of is the transition probability, Indicates that the hidden state value is , the observation vector O t The probability of It means that the hidden state value at time t is the maximum value among all possible maximum probability values ​​of the hidden state; Record path information, from the hidden state at time t−1 The hidden state at time t is The optimal path of ; The calculation is terminated at time T, and the hidden state with the maximum probability at time T is: ; in, represents the hidden state value at time T, Indicates that the hidden state value at time T is The maximum probability of Path backtracking: Get the optimal hidden state sequence by backtracking the path matrix: ; in, represents the hidden state value at time t, is the path matrix that records the hidden state transfer path, indicating the hidden state at time t-1 that the hidden state sequence that reaches the maximum probability at time t depends on; Until we trace back to t=1, we get the hidden state sequence with the maximum probability: .

[0014] To achieve the above objectives, the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method for predicting the breakdown voltage of motor insulation based on the continuous hidden Markov model as described above is implemented.

[0015] To achieve the above objectives, the present application provides a computer storage medium on which a computer program is stored. When the computer program is run on a computer, the computer executes the motor insulation breakdown voltage prediction method based on the continuous hidden Markov model as described above.

[0016] The present application uses a continuous hidden Markov model to predict the breakdown voltage of the motor insulation material during the operation of the motor, which can play a significant auxiliary role in the life assessment and prediction of the motor and help improve the safety and reliability of electrical equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1A first flow chart of a method for predicting breakdown voltage of motor insulation based on a continuous hidden Markov model provided in an embodiment of the present application; Figure 2 A second flow chart of a method for predicting breakdown voltage of motor insulation based on a continuous hidden Markov model provided in an embodiment of the present application; Figure 3 A third flow chart of a method for predicting breakdown voltage of motor insulation based on a continuous hidden Markov model provided in an embodiment of the present application; Figure 4 A system block diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the specific implementation manner of the present application will be clearly and completely described below in conjunction with the drawings in the implementation manner of the present application.

[0019] During the operation of the motor, the four variables of the motor's insulation material, capacitance, insulation resistance, dielectric loss and maximum discharge, change over time and are observable. The breakdown voltage of the motor's insulation material will also change over time, but it is not observable. Therefore, the present application maps the breakdown voltage prediction problem of the motor's insulation material into a continuous hidden Markov model (CHMM), where the hidden state sequence is the continuous time-varying process of the breakdown voltage, and the observation sequence is the capacitance, insulation resistance, dielectric loss and maximum discharge obtained in the time-varying process, and the breakdown voltage of the motor's insulation material during the operation of the motor is predicted based on the continuous hidden Markov model.

[0020] Please refer to Figure 1 The present application provides a method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model, the method comprising the steps of: Step S101, taking the insulating material of the motor as the sample to be tested, obtaining multiple groups of experimental data of the sample to be tested in different motor operation time periods, and constructing a training set with the multiple groups of experimental data, wherein the experimental data include the test parameters and breakdown voltage of the sample to be tested, and the test parameters include capacitance, insulation resistance, dielectric loss, and maximum discharge capacity.

[0021] Since the breakdown voltage can only be measured once, it is necessary to perform multiple sets of repeated experiments on the same sample to be tested. Taking the insulation material of the motor as the sample to be tested, obtain multiple sets of experimental data of the sample to be tested in different motor operation time periods. The experimental data includes the test parameters and breakdown voltage of the sample to be tested. The test parameters include capacitance, insulation resistance, dielectric loss, and maximum discharge. For example, data collection is performed every 20 hours. At the initial moment of the motor operation, the capacitance, insulation resistance, dielectric loss, maximum discharge, and breakdown voltage data of the motor insulation material are collected as the first set of experimental data; after the motor has been running for 20 hours, the capacitance, insulation resistance, dielectric loss, maximum discharge, and breakdown voltage data of the motor insulation material are collected as the second set of experimental data, and so on until the motor insulation material fails, and multiple sets of experimental data are obtained to ensure that comprehensive data on the changes of various variables over time are obtained.

[0022] Step S102, constructing a continuous hidden Markov model with the test parameters as the observation sequence and the discretized breakdown voltage as the hidden state.

[0023] The obtained capacitance, insulation resistance, dielectric loss and maximum discharge capacity of each group are used as observation sequences, capacitance, insulation resistance, dielectric loss and maximum discharge capacity are used as observation values, the obtained breakdown voltage is discretized, and the discretized breakdown voltage is used as the hidden state to construct a continuous hidden Markov model. The hidden state sequence is the continuous time-varying process of the breakdown voltage, and the observation sequence is the capacitance, insulation resistance, dielectric loss and maximum discharge amount obtained in the time-varying process. The breakdown voltage of the motor insulation material during the operation of the motor is predicted by the continuous hidden Markov model.

[0024] Step S103 : Based on the training set, the parameters of the continuous hidden Markov model are iteratively calculated by using the Baum-Welch algorithm to determine the optimal parameters of the continuous hidden Markov model.

[0025] The essence of training a hidden Markov model is the process of reestimating the model parameters. During the process of reestimating the model parameters, for a known observation sequence, the model parameters are iteratively calculated using the Baum-Welch algorithm until the parameters converge to meet the convergence conditions, and the optimal parameters of the hidden Markov model are obtained, so that the probability of the hidden Markov model generating the observation sequence O under the optimal parameters is maximum.

[0026] Step S104, based on the continuous hidden Markov model with optimal parameters, determine the hidden state sequence with the maximum probability under the observation sequence at the current moment through the Viterbi algorithm, and determine the breakdown voltage at the current moment based on the hidden state sequence with the maximum probability, wherein the observation sequence at the current moment includes the test parameters at the current moment.

[0027] The Viterbi algorithm is a dynamic programming algorithm that can solve the shortest path problem in any graph. The Viterbi algorithm is used to recursively calculate and store the optimal path probability of each state at each moment through dynamic programming, thereby avoiding repeated calculations and ultimately determining the most likely hidden state sequence under the observation sequence at the current moment, and then the breakdown voltage at the current moment.

[0028] In this implementation, the breakdown voltage of the motor insulation material during motor operation is predicted by a continuous hidden Markov model, which can greatly assist in the evaluation and prediction of the motor life and help improve the safety and reliability of electrical equipment.

[0029] An embodiment of the present application, such as Figure 2 As shown, a continuous hidden Markov model is constructed with the test parameters as the observation sequence and the discretized breakdown voltage as the hidden state, including the steps: Step S201, taking each set of test parameters as an observation sequence O, represented as ,in, , the observation vector It represents the observed value obtained at time t, including capacitance, insulation resistance, dielectric loss and maximum discharge capacity, and T represents the time length; Step S202, discretize the acquired multiple groups of breakdown voltages into N values, and the N values ​​are used to represent N hidden states to form a hidden state set S, which is expressed as ,in, represents the i-th hidden state, and N represents N different hidden states; Step S203, construct the actual hidden state sequence Q, expressed as ,in, represents the hidden state value at time t, The value range is ; Step S204: the parameters of the continuous hidden Markov model include the initial state probability matrix , state transition probability matrix And the observation probability matrix , where the initial state probability Indicates that the hidden state value at the initial moment is The probability of state transition Indicates that the hidden state value at time t is , the hidden state value at time t+1 is The probability of observation Indicates that the hidden state value is When The probability of Step S205: Use a Gaussian mixture model to fit the observations to obtain the observation probability density of the output observations in each hidden state. , where the observed probability density is The parameters of the Gaussian distribution based on the observed value are determined, and the parameters of the Gaussian distribution include the weight of the Gaussian distribution, the mean vector of the Gaussian distribution, and the covariance matrix of the Gaussian distribution.

[0030] Continuous Hidden Markov Model uses a quintuple N represents the number of hidden states, M represents the number of observations, π represents the initial state probability matrix, A represents the state transition probability matrix, and B represents the observation probability matrix.

[0031] In this embodiment, each set of test parameters is taken as an observation sequence O, which is expressed as , the number of observation values ​​M is 4, that is, capacitance, insulation resistance, dielectric loss and maximum discharge capacity are taken as observation values, among which, , the observation vector represents the observation value obtained at time t, and T represents the length of time.

[0032] In this embodiment, the multiple groups of breakdown voltages obtained are divided into N intervals in ascending order, each interval corresponds to a value, and the value represents a hidden state. N values ​​represent N hidden states to form a hidden state set S, which is expressed as ,in, represents the i-th hidden state, that is, the breakdown voltage is within the i-th interval, and N represents N different hidden states.

[0033] In this embodiment, the actual hidden state sequence Q is constructed, and the actual hidden state sequence Q is expressed as ,in, represents the hidden state value at time t, The value range is , T represents the length of time.

[0034] In this embodiment, the parameters of the continuous hidden Markov model include the initial state probability matrix , state transition probability matrix And the observation probability matrix . Initial state probability Indicates the hidden state value at the initial moment for The probability of the breakdown voltage when the motor runs for t=1 is The probability of state transition , , which means the hidden state value at time t is , the hidden state value at time t+1 is The probability of breakdown voltage at time t is , the breakdown voltage at time t+1 is The probability of observation. Indicates that the hidden state value is When probability.

[0035] In this embodiment, the Gaussian mixture model is a probability model composed of multiple single Gaussian models, each of which is a single Gaussian distribution, used to describe the distribution of data. The Gaussian mixture model can fit data with continuous probability distribution. Four Gaussian distributions are used to describe the data distribution of capacitance, insulation resistance, dielectric loss and maximum discharge capacity respectively. The Gaussian mixture model is used to fit the observed values ​​to obtain the observed probability density of the output observed values ​​under each hidden state. , where the observed probability density is The parameters of the Gaussian distribution based on the observed values ​​are determined, and the parameters of the Gaussian distribution include the weight of the Gaussian distribution, the mean vector of the Gaussian distribution, and the covariance matrix of the Gaussian distribution.

[0036] In one embodiment, four Gaussian distributions are used to describe the data distribution of capacitance, insulation resistance, dielectric loss and maximum discharge capacity respectively, and the observation probability density is determined by the weight, mean vector and covariance matrix of each Gaussian distribution. , the observed probability density It is expressed as: ; ; ; Where M represents the number of Gaussian distributions, M is 4, Represents hidden state The weight of the m-th Gaussian distribution, The hidden state is The mean vector of the m-th Gaussian distribution, The hidden state is The covariance matrix of the m-th Gaussian distribution is, Represents the mixed Gaussian probability density function, m=1,2,3,4, corresponding to capacitance, insulation resistance, dielectric loss, and maximum discharge capacity, respectively.

[0037] In one embodiment, Figure 3 As shown, based on the training set, the parameters of the continuous hidden Markov model are iteratively calculated by the Baum-Welch algorithm to determine the optimal parameters of the continuous hidden Markov model, including the steps of: Step S301, initializing the initial state probability matrix, the state transition probability matrix and the observation probability matrix, obtaining the initial parameters of the continuous hidden Markov model, and iteratively updating the initial parameters of the continuous hidden Markov model, wherein in each iteration process; Step S302, calculating the forward probability of each hidden state value at each moment within the time length, and calculating the backward probability of each hidden state value at each moment within the time length; Step S303, determining the current maximum likelihood probability under the observation sequence using the current parameters of the continuous hidden Markov model; Step S304, based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, calculate the conditional probability of each hidden state value at any time; Step S305, calculating the joint conditional probability of the hidden state value at any time and the hidden state value at the next time after the any time based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value; Step S306, calculating the joint probability of each hidden state value at any time based on the conditional probability and the mixed Gaussian probability density function; Step S307, updating the parameters of the Gaussian distribution, the initial state probability matrix, and the state transition probability matrix based on the conditional probability, the joint conditional probability, and the joint probability, and determining an updated observation probability matrix based on the updated parameters of the Gaussian distribution; Step S308, based on the updated initial state probability matrix, the state transition probability matrix and the observation probability matrix, determine the updated parameters of the continuous hidden Markov model, and determine the updated maximum likelihood probability under the observation sequence with the updated parameters of the continuous hidden Markov model; Step S309: If the number of iterations is greater than or equal to the number threshold, and / or the difference between the current maximum likelihood probability and the updated maximum likelihood probability meets the threshold, the iteration is stopped, and the updated parameters of the continuous hidden Markov model of the current iteration are used as the optimal parameters of the continuous hidden Markov model.

[0038] In the process of training the hidden Markov model, multiple sets of experimental data are used as training sets to perform model sequence optimization learning on the model parameters through the Baum-Welch algorithm. The problem of parameter optimization learning is described as the probability of the observation sequence when solving the parameters under the given observation sequence. maximum.

[0039] In one embodiment, calculating the forward probability of each hidden state value at each moment within the time length includes: Calculate the forward probability of each hidden state value at the initial moment, and calculate the forward probability of each hidden state at time t , forward probability It is used to indicate that the hidden state value at time t is , the observation vector from the initial time to time t The probability of forward The calculation formula is: ; ; in, Indicates that the initial hidden state value is The forward probability of represents the initial state probability, Indicates that the initial hidden state value is , the probability that the observation vector is O1, N represents the number of hidden states, Indicates that the hidden state value at time t is The forward probability of Indicates that the hidden state value at time t is , the observation vector is O t The probability of Indicates that the hidden state value at time t-1 is The forward probability of represents the state transition probability; Calculate the backward probability of each hidden state value at each moment in the time length, including: Calculate the backward probability of each hidden state value at the termination time, and calculate the backward probability of each hidden state value at time t , backward probability It is used to indicate that the hidden state value at time t is , the observation vector from time t+1 to time T The probability of backward The calculation formula is: ; ; in, Indicates that the hidden state value at the termination time T is The backward probability of Indicates that the hidden state value at time t is The backward probability of It means that the hidden state value at time t is , the observation vector is O t+1 The probability of N represents the number of hidden states. Indicates that the hidden state value at time t+1 is The backward probability of represents the state transition probability.

[0040] In one embodiment, the current maximum likelihood probability under the observation sequence is determined using the current parameters of the continuous hidden Markov model, including: The current maximum likelihood probability is calculated as: ; in, represents the current parameters of the continuous hidden Markov model, that is, the parameters of the current iteration, Representation parameters The maximum likelihood probability of the observation sequence under Indicates that the hidden state value at the end time T is The forward probability of Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, the conditional probability of each hidden state value at any time is calculated, including: Calculate the conditional probability of each hidden state value at any time It is expressed as; ; in, Indicates that the hidden state value at time t is The conditional probability of Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, the joint conditional probability of the hidden state value at any time and the hidden state value at the next time at any time is calculated, including: Joint conditional probability It is expressed as: ; in, Indicates that the hidden state value at time t is and transfer to The joint conditional probability of Indicates that the hidden state value is , the observation vector is O t+1 The probability of Indicates that the hidden state value at time t+1 is The backward probability of Indicates that the hidden state value at time t is The forward probability of Based on the conditional probability and the mixed Gaussian probability density function, the joint probability of each hidden state value at any time is calculated, including: Joint Probability The calculation formula is: ; in, Indicates that the hidden state value at time t is The joint probability of the mth Gaussian mixture component, represents the mixed Gaussian probability density function, Represents hidden state The weight of the next m-th Gaussian distribution, M represents the number of Gaussian distributions, M is 4.

[0041] In one embodiment, the parameters of the Gaussian distribution, the initial state probability matrix, and the state transition probability matrix are updated based on the conditional probability, the joint conditional probability, and the joint probability, and the updated observation probability matrix is ​​determined based on the updated parameters of the Gaussian distribution, including: Updated initial state probability for: ; in, It means that given the observation sequence O, the hidden state value at the initial moment is probability; Updated state transition probability for: ; in, Indicates that the hidden state value at time t is The conditional probability of Updated Gaussian distribution weights for; ; in, Indicates that the hidden state value at time t is The joint probability of the sth Gaussian mixture component; The mean vector of the updated Gaussian distribution for: ; Among them, O t represents the observation vector at time t; Covariance matrix of the updated Gaussian distribution for: .

[0042] Based on the updated Gaussian distribution parameters and the calculation formula of the observation probability matrix, the updated observation probability matrix is ​​obtained. With the updated initial state probability matrix, state transition probability matrix and observation probability matrix, the updated parameters of the continuous hidden Markov model can be determined: Through the calculation formula of the maximum likelihood probability, the updated maximum likelihood probability of the observation sequence under the updated parameters of the continuous hidden Markov model can be calculated. ,like ; in, is the preset threshold, the iteration is stopped, and the parameter is the optimal parameter of the continuous hidden Markov model. If the above formula does not hold, continue to iterate until the convergence condition is met.

[0043] In one embodiment, based on a continuous hidden Markov model with optimal parameters, a hidden state sequence with the maximum probability under an observation sequence at the current moment is determined by a Viterbi algorithm, and a breakdown voltage value at the current moment is determined based on the hidden state sequence with the maximum probability, including: Set the two-dimensional matrix V to store the maximum probability of each hidden state value in each hidden state at each time step, and set Indicates that the hidden state value at time t is The maximum probability, path matrix Represents the hidden state at time t-1 that the hidden state sequence with the maximum probability at time t depends on ; Initialization: At the initial time t=1, the hidden state value is calculated as The maximum probability of , initialize the path matrix to 0: ; ; in, Table hidden state value The initial probability of Indicates that the hidden state value is , the probability of observing vector O1; Recursive calculation: For any time t and any hidden state , calculate the hidden state value at time t The maximum probability for; ; in, Indicates that the hidden state value at time t-1 is The maximum probability of is the transition probability, Indicates that the hidden state value is , the observation vector O t The probability of It means that the hidden state value at time t is the maximum value among all possible maximum probability values ​​of the hidden state; Record path information, from the hidden state at time t−1 The hidden state at time t is The optimal path is: ; The calculation is terminated at time T. The hidden state with the maximum probability at time T is: ; in, represents the hidden state value at time T, Indicates that the hidden state value at time T is The maximum probability of Path backtracking: The optimal hidden state sequence obtained by backtracking the path matrix is: ; in, represents the hidden state value at time t, is the path matrix that records the hidden state transfer path, indicating the hidden state at time t-1 that the hidden state sequence that reaches the maximum probability at time t depends on; Until we trace back to t=1, we get the hidden state sequence with the maximum probability: .

[0044] Based on the continuous hidden Markov model with optimal parameters, the Viterbi algorithm and the known observation sequence at the current moment, namely the capacitance, insulation resistance, dielectric loss and maximum discharge capacity at the current moment, are used to recursively calculate the most likely hidden state sequence under the observation sequence at the current moment, and then the breakdown voltage at the current moment, thereby realizing the prediction of the breakdown voltage at the current moment.

[0045] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, any one of the above-mentioned methods for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model is implemented.

[0046] Based on the same inventive concept, the embodiment of the present application also provides a computer device for implementing the above-mentioned method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model. The implementation scheme for solving the problem provided by the computer device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more computer device embodiments provided below can refer to the above limitations on the method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model, which will not be repeated here.

[0047] Figure 4 It is a schematic diagram of the hardware structure of a computer device provided in an embodiment of the present application. Figure 4The computer device shown includes: a processor 401, a communication interface 402, a memory 403 and a communication bus 404. The processor 401, the communication interface 402 and the memory 403 communicate with each other via the communication bus 404. Figure 4 The connection method between the processor 401 , the communication interface 402 , and the memory 403 shown is merely exemplary. During implementation, the processor 401 , the communication interface 402 , and the memory 403 may also be communicatively connected to each other using other connection methods besides the communication bus 404 .

[0048] The memory 403 can be used to store a computer program, which may include instructions and data to implement the steps of any of the above methods for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model. In an embodiment of the present application, the memory 403 may be various types of storage media, such as random access memory (RAM), read only memory (ROM), non-volatile RAM (NVRAM), programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), flash memory, optical storage, and registers. The memory 403 may include a hard disk and / or a memory.

[0049] The processor 401 may be a general-purpose processor, which may be a processor that performs specific steps and / or operations by reading and executing a computer program (e.g., a computer program) stored in a memory (e.g., memory 403). The general-purpose processor may use data stored in the memory (e.g., memory 403) during the execution of the steps and / or operations. The general-purpose processor may be, for example, but not limited to, a central processing unit (CPU). In addition, the processor 401 may also be a dedicated processor, which may be a processor specially designed to perform specific steps and / or operations. The dedicated processor may be, for example, but not limited to, an ASIC and an FPGA. In addition, the processor 401 may also be a combination of multiple processors, such as a multi-core processor.

[0050] The communication interface 402 may include an input / output (I / O) interface, a physical interface, and a logical interface for interconnecting devices within the network device, as well as an interface for interconnecting the network device with other devices (e.g., network devices). The communication network may be Ethernet, a radio access network (RAN), a wireless local area network (WLAN), etc. The communication interface 402 may be a module, a circuit, a transceiver, or any device capable of implementing communication.

[0051] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 401 or an instruction in the form of software. The method disclosed in conjunction with the embodiment of the present application can be directly embodied as a hardware processor for execution, or a combination of hardware and software modules in the processor for execution. The software module can be located in a mature storage medium in the art such as a random access memory flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 403, and the processor 401 reads the information in the memory 403 and completes the steps of the above method in conjunction with its hardware. To avoid repetition, it is not described in detail here.

[0052] The above description of the embodiments is to facilitate the understanding and application of the present application by those skilled in the art. It is obvious that those familiar with the art can easily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present application is not limited to the above embodiments, and improvements and modifications made to the present application by those skilled in the art based on the disclosure of the present application should be within the scope of protection of the present application.

Claims

1. A method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model, characterized in that: The method comprises: Taking the insulation material of the motor as the sample to be tested, obtaining multiple groups of experimental data of the sample to be tested in different motor operation time periods, and constructing a training set with the multiple groups of experimental data, wherein the experimental data include the test parameters and breakdown voltage of the sample to be tested, and the test parameters include capacitance, insulation resistance, dielectric loss, and maximum discharge capacity; A continuous hidden Markov model is constructed using the test parameters as an observation sequence and the discretized breakdown voltage as a hidden state; Based on the training set, the parameters of the continuous hidden Markov model are iteratively calculated by the Baum-Welch algorithm to determine the optimal parameters of the continuous hidden Markov model; Based on the continuous hidden Markov model with optimal parameters, the hidden state sequence with maximum probability under the observation sequence at the current moment is determined by Viterbi algorithm, and the breakdown voltage at the current moment is determined based on the hidden state sequence with maximum probability, wherein the observation sequence at the current moment includes the test parameters at the current moment.

2. The method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model according to claim 1, characterized in that: The method of constructing a continuous hidden Markov model using the test parameters as an observation sequence and the discretized breakdown voltage as a hidden state includes: Each set of test parameters is regarded as an observation sequence O, expressed as ,in, , the observation vector represents the observed values ​​obtained at time t, including capacitance, insulation resistance, dielectric loss and maximum discharge capacity, and T represents the time length; The obtained multiple sets of breakdown voltages are discretized into N values, and the N values ​​are used to represent N hidden states to form a hidden state set S, which is expressed as ,in, represents the i-th hidden state, and N represents N different hidden states; Construct the actual hidden state sequence Q, expressed as ,in, represents the hidden state value at time t, The value range is ; The parameters of the continuous hidden Markov model include the initial state probability matrix , state transition probability matrix And the observation probability matrix , where the initial state probability Indicates that the hidden state value at the initial moment is The probability of state transition Indicates that the hidden state value at time t is , the hidden state value at time t+1 is The probability of observation Indicates that the hidden state value is When , the observation vector is probability; The Gaussian mixture model is used to fit the observation data and obtain the observation probability density of the output observation value under each hidden state. , where the observed probability density is The parameters of the Gaussian distribution based on the observed values ​​are determined, and the parameters of the Gaussian distribution include the weight of the Gaussian distribution, the mean vector of the Gaussian distribution, and the covariance matrix of the Gaussian distribution.

3. The method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model according to claim 2, characterized in that: The Gaussian mixture model is used to fit the observation data and obtain the observation probability density of the output observation value under each hidden state. ,include: Four Gaussian distributions are used to describe the data distribution of capacitance, insulation resistance, dielectric loss and maximum discharge capacity respectively. The observation probability density is determined by the weight, mean vector and covariance matrix of each Gaussian distribution. , the observed probability density It is expressed as: ； ； ； Where M represents the number of Gaussian distributions, M is 4, Represents hidden state The weight of the m-th Gaussian distribution, The hidden state is The mean vector of the m-th Gaussian distribution, The hidden state is The covariance matrix of the m-th Gaussian distribution is, Represents the mixed Gaussian probability density function, m=1,2,3,4, corresponding to capacitance, insulation resistance, dielectric loss, and maximum discharge capacity, respectively.

4. The method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model according to claim 2, characterized in that: Based on the training set, the parameters of the continuous hidden Markov model are iteratively calculated by the Baum-Welch algorithm to determine the optimal parameters of the continuous hidden Markov model, including: Initialize the initial state probability matrix, the state transition probability matrix and the observation probability matrix to obtain the initial parameters of the continuous hidden Markov model, and iteratively update the initial parameters of the continuous hidden Markov model, wherein in each iteration process, Calculate the forward probability of each hidden state value at each moment within the time length; Calculate the backward probability of each hidden state value at each moment within the time length; Determining the current maximum likelihood probability under the observation sequence with the current parameters of the continuous hidden Markov model; Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, calculate the conditional probability of each hidden state value at any time; Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, calculate the joint conditional probability of the hidden state value at any time and the hidden state value at the next time after the any time; Based on conditional probability and mixed Gaussian probability density function, calculate the joint probability of each hidden state value at any time; Based on the conditional probability, the joint conditional probability and the joint probability, the parameters of the Gaussian distribution, the initial state probability matrix and the state transition probability matrix are updated, and based on the updated parameters of the Gaussian distribution, an updated observation probability matrix is ​​determined; Determining updated parameters of the continuous hidden Markov model based on the updated initial state probability matrix, state transition probability matrix and observation probability matrix; Determining the updated maximum likelihood probability under the observation sequence using the updated parameters of the continuous hidden Markov model; If the number of iterations is greater than or equal to the number threshold, and / or the difference between the current maximum likelihood probability and the updated maximum likelihood probability meets the threshold, the iteration is stopped, and the updated parameters of the continuous hidden Markov model of the current iteration are used as the optimal parameters of the continuous hidden Markov model.

5. The method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model according to claim 4, characterized in that: Calculate the forward probability of each hidden state value at each moment in the time length, including: Calculate the forward probability of each hidden state value at the initial moment, and calculate the forward probability of each hidden state at time t , the forward probability It is used to indicate that the hidden state value at time t is , the observation vector from the initial time to time t The probability of forward The calculation formula is: ； ； in, Indicates that the initial hidden state value is The forward probability of represents the initial state probability, Indicates that the initial hidden state value is , the probability that the observation vector is O1, N represents the number of hidden states, Indicates that the hidden state value at time t is The forward probability of Indicates that the hidden state value at time t is , the observation vector is O t The probability of Indicates that the hidden state value at time t-1 is The forward probability of represents the state transition probability; Calculate the backward probability of each hidden state value at each moment in the time length, including: Calculate the backward probability of each hidden state value at the termination time, and calculate the backward probability of each hidden state value at time t , backward probability It is used to indicate that the hidden state value at time t is , the observation vector from time t+1 to time T The probability of backward The calculation formula is: ； ； in, Indicates that the hidden state value at the termination time T is The backward probability of Indicates that the hidden state value at time t is The backward probability of It means that the hidden state value at time t is , the observation vector is O t+1 The probability of N represents the number of hidden states. Indicates that the hidden state value at time t+1 is The backward probability of represents the state transition probability.

6. The method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model according to claim 5, characterized in that: Determining the current maximum likelihood probability under the observation sequence with the current parameters of the continuous hidden Markov model includes: The current maximum likelihood probability is calculated as: ； in, represents the current parameters of the continuous hidden Markov model, Representation parameters The maximum likelihood probability of the observation sequence under Indicates that the hidden state value at the end time T is The forward probability of Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, the conditional probability of each hidden state value at any time is calculated, including: Calculate the conditional probability of each hidden state value at any time It is expressed as: ； in, Indicates that the hidden state value at time t is The conditional probability of Based on the current maximum likelihood probability, the forward probability of each hidden state value, and the backward probability of each hidden state value, the joint conditional probability of the hidden state value at any time and the hidden state value at the next time of the any time is calculated, including: Joint conditional probability It is expressed as: ； in, Indicates that the hidden state value at time t is and transfer to The joint conditional probability of Indicates that the hidden state value is , the observation vector is O t+1 The probability of Indicates that the hidden state value at time t+1 is The backward probability of Indicates that the hidden state value at time t is The forward probability of Based on the conditional probability and the mixed Gaussian probability density function, the joint probability of each hidden state value at any time is calculated, including: Joint Probability The calculation formula is: ； in, Indicates that the hidden state value at time t is The joint probability of the mth Gaussian mixture component, represents the mixed Gaussian probability density function, Represents hidden state The weight of the next m-th Gaussian distribution, M represents the number of Gaussian distributions, M is 4.

7. The method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model according to claim 6, characterized in that: The parameters of the Gaussian distribution, the initial state probability matrix, and the state transition probability matrix are updated based on the conditional probability, the joint conditional probability, and the joint probability, and the updated observation probability matrix is ​​determined based on the updated parameters of the Gaussian distribution, including: Updated initial state probability for: ； in, Indicates that the hidden state value of a given observation sequence at the initial moment is probability; Updated state transition probability for: ； in, Indicates that the hidden state value at time t is The conditional probability of Updated Gaussian distribution weights for: ； in, Indicates that the hidden state value at time t is The joint probability of the sth Gaussian mixture component; The mean vector of the updated Gaussian distribution for: ； Among them, O t represents the observation vector at time t; Covariance matrix of the updated Gaussian distribution for: 。 8. The method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model according to claim 1, characterized in that: Based on the continuous hidden Markov model with optimal parameters, the Viterbi algorithm is used to determine the hidden state sequence with the maximum probability under the observation sequence at the current moment, and the breakdown voltage value at the current moment is determined based on the hidden state sequence with the maximum probability: Set the two-dimensional matrix V to store the maximum probability of each hidden state value in each hidden state at each time step, and set Indicates that the hidden state value at time t is The maximum probability, path matrix Represents the hidden state at time t-1 that the hidden state sequence with the maximum probability at time t depends on ; Initialization: At the initial time t=1, the hidden state value is calculated as The maximum probability of , initialize the path matrix to 0: ； ； in, Indicates that the hidden state value is The initial probability of Indicates that the hidden state value is , the probability that the observation vector is O1; Recursive calculation: For any time t and any hidden state , calculate the hidden state value at time t The maximum probability for: ； in, Indicates that the hidden state value at time t-1 is The maximum probability of is the transition probability, Indicates that the hidden state value is , the observation vector O t The probability of It means that the hidden state value at time t is the maximum value among all possible maximum probability values ​​of the hidden state; Record path information, from the hidden state at time t−1 The hidden state at time t is The optimal path of ； The calculation is terminated at time T, and the hidden state with the maximum probability at time T is: ； in, represents the hidden state value at time T, Indicates that the hidden state value at time T is The maximum probability of Path backtracking: Get the optimal hidden state sequence by backtracking the path matrix: ； in, represents the hidden state value at time t, is the path matrix that records the hidden state transfer path, indicating the hidden state at time t-1 that the hidden state sequence that reaches the maximum probability at time t depends on; Until we trace back to t=1, we get the hidden state sequence with the maximum probability: .

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model as described in any one of claims 1 to 8 is implemented.

10. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed on a computer, the computer is enabled to execute the method for predicting the breakdown voltage of motor insulation based on a continuous hidden Markov model according to any one of claims 1 to 8.

Images