CN113433409A - Electric automobile IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning - Google Patents
Electric automobile IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning Download PDFInfo
- Publication number
- CN113433409A CN113433409A CN202110737833.6A CN202110737833A CN113433409A CN 113433409 A CN113433409 A CN 113433409A CN 202110737833 A CN202110737833 A CN 202110737833A CN 113433409 A CN113433409 A CN 113433409A
- Authority
- CN
- China
- Prior art keywords
- charging equipment
- fault diagnosis
- direct current
- data
- type common
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000003745 diagnosis Methods 0.000 title claims abstract description 57
- 238000013135 deep learning Methods 0.000 title claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000012360 testing method Methods 0.000 claims abstract description 14
- 238000012795 verification Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 239000002245 particle Substances 0.000 claims description 27
- 230000003044 adaptive effect Effects 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 238000012217 deletion Methods 0.000 claims description 6
- 230000037430 deletion Effects 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 6
- 238000009413 insulation Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 5
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000001816 cooling Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000004891 communication Methods 0.000 claims description 2
- 238000012423 maintenance Methods 0.000 abstract description 5
- 238000005457 optimization Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000013178 mathematical model Methods 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006353 environmental stress Effects 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000013024 troubleshooting Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Electric Propulsion And Braking For Vehicles (AREA)
Abstract
The invention designs a deep learning-based fault diagnosis method for an electric automobile IGBT type common direct current bus charging device, which specifically comprises the following steps: firstly, analyzing running state data of the IGBT type common direct current bus charging equipment of the electric automobile, and preprocessing the running state data to establish a data set; secondly, dividing a data set into a training set, a verification set and a test set; then, establishing a charging equipment fault diagnosis model based on the self-adaptive deep confidence network, and training the charging equipment fault diagnosis model by using a training set; then, evaluating the performance of the fault diagnosis model by using a verification set and a test set; and finally, inputting the running state data of the charging equipment into a fault diagnosis model meeting the requirements to obtain a diagnosis result. The method can effectively improve the fault diagnosis accuracy of the IGBT type common direct current bus charging equipment, and provides a basis for daily maintenance and repair of the charging equipment.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis, and particularly relates to a deep learning-based fault diagnosis method for an electric vehicle IGBT type common direct current bus charging device.
Background
In recent years, as the number of electric vehicles increases, there are increasing demands for charging quality and charging time, and it is desired to complete high-quality charging of vehicles in a short time. The IGBT type common direct current bus charging equipment (charging equipment for short) of the electric automobile has the advantages of high output power, short charging time, high electric energy quality and the like, and is installed and used in a large quantity. However, most of such charging devices are installed in outdoor areas and are subjected to large environmental stresses such as rain and dew for a long time, so that the functional maintenance of the charging devices faces more and more problems. For example, core components such as a charging module and an insulation detection module of the charging device often fail, which reduces the safety and reliability of the charging device. Meanwhile, a large amount of labor and time are required for the periodic maintenance of the charging equipment and the troubleshooting of the fault reason after the fault occurs, so that the maintenance and guarantee costs are higher and higher. Therefore, the effective charging equipment fault diagnosis method has important significance for safe operation of the charging equipment and reduction of maintenance cost.
At present, fault diagnosis methods are mainly divided into two main categories, namely qualitative analysis and quantitative analysis, wherein analytical model-based methods and data-driven methods in quantitative analysis are widely used in equipment fault diagnosis. The method based on the analytic model can often obtain higher fault diagnosis accuracy, but the internal structure of the charging equipment is complex, an accurate mathematical model is difficult to establish, and the established model is only suitable for equipment with a specific model, so that the universality is poor, and the method based on the analytic model is difficult to apply. During the operation process of the equipment, a large amount of operation data are accumulated, and the fault diagnosis of the equipment is realized by analyzing and processing the operation data, extracting the characteristics of fault data and mapping the relation between the fault data and the fault type based on a data driving method. The method is independent of an accurate mathematical model, and is relatively simple to implement and high in universality. The fault data of the charging equipment contains abundant fault information of the charging equipment under different working conditions, and the fault diagnosis method based on deep learning has the characteristics of strong coupling, multi-dimensionality and the like, can fully utilize the information, and can overcome the defects that the traditional fault diagnosis method is poor in generalization capability and easy to fall into local optimal solution.
Disclosure of Invention
The invention provides a method for diagnosing faults of an IGBT type common direct current bus device of an electric vehicle based on deep learning, aiming at the problem of fault diagnosis of the charging device of the electric vehicle. According to the method, a Nesterov momentum method and convergence of an adaptive learning rate acceleration model are introduced on the basis of a typical deep belief network model, the adaptive deep belief network model is constructed, the training speed of the model is improved, the improved particle swarm optimization is used for optimizing the hyper-parameters of the deep belief network, and the performance of model fault diagnosis is improved.
In order to achieve the purpose, the invention provides the following technical scheme:
a deep learning-based fault diagnosis method for an electric automobile IGBT type common direct current bus charging device specifically comprises the following steps:
step 1: analyzing the running state data of the IGBT type common direct current bus charging equipment of the electric automobile, and preprocessing the running state data to establish a data set;
step 2: dividing a data set into a training set, a verification set and a test set;
and step 3: establishing a charging equipment fault diagnosis model based on a self-adaptive deep confidence network, and training the charging equipment fault diagnosis model by using a training set;
and 4, step 4: evaluating the performance of the fault diagnosis model by using the verification set and the test set;
and 5: and inputting the running state data of the charging equipment into a charging equipment fault diagnosis model meeting the requirements to obtain a diagnosis result.
In step 1 of the present invention, the operating state of the charging device is divided into a normal state and a fault state, wherein the fault state includes, but is not limited to, a dc bus output overvoltage fault, a dc bus output overcurrent fault, a dc bus output contactor fault, an electric vehicle BMS communication fault, a charging device insulation detection fault, a charging module over-temperature fault, a charging module input open-phase fault, a charging device output overvoltage fault, a charging device connector fault, and the like.
In step 1 of the present invention, the operation state data of the charging device includes, but is not limited to, charging device input voltage, charging device input current, charging device output voltage, charging device output current, charging device output power, vehicle charging demand voltage, vehicle charging demand current, charging device dc bus voltage, dc circuit breaker state, battery temperature, battery set temperature, cooling fan state, auxiliary unit output voltage, charging duration, insulation detection module state, and other information.
In step 1 of the invention, the data set is preprocessed, which specifically comprises the following operations:
(1) processing missing values, and respectively executing interpolation or deletion operation according to the importance degree of the data;
(2) processing abnormal values, namely executing deletion operation on abnormal data in the data set;
(3) and (4) normalizing the data, and mapping the range of the data between [0,1] by a range normalization method.
In step 3 of the invention, the adaptive deep belief network optimizes the network by using a Nesterov momentum method, and performs equivalent transformation on an initial formula to accelerate the training speed, wherein the initial calculation formula of the Nesterov momentum method is as follows:
θt=θt-1+vt (2)
Let thetat=θ′tThen there is
Where θ ═ w, a, b represents parameters of the network, and a and b are offsets of the visible layer and the hidden layer, respectivelyW is the connection weight of the visible layer and the hidden layer, vtRepresenting the velocity after the t-th iteration, vt-1Represents the speed after the t-1 iteration, beta is the momentum value, and alpha is the learning rate.
In step 3 of the invention, the adaptive deep belief network updates network parameters by using an independent adaptive learning rate on the basis of introducing Nesterov momentum, and the calculation formula is as follows:
in the formula (I), the compound is shown in the specification,representing the gradient of the parameter theta after the t-th training, if the weight gradient direction of the current parameter is consistent with the weight gradient direction of the previous training, psitXi is increased, and psi is obtained otherwisetReduced by a factor of 1-xi, the initial value of the parameter ψ is set to 1, and the parameter ξ is set to 0.1.
In step 3 of the invention, the adaptive deep belief network selects the hyper-parameters of the network by adopting an improved particle swarm algorithm, the inertia factor of the improved particle swarm algorithm is updated by adopting a random weight mode, and the calculation formula is as follows:
in the formula, muminMinimum value of random inertial weight, mumaxMaximum value of random inertial weight, U (0,1) is [0,1]]N (0,1) is a random number that follows a standard normal distribution, and σ is used to measure the degree of deviation between the random inertial weight ω and its mathematical expectation.
In step 3 of the invention, the process of training the adaptive deep belief network by using the training set data to obtain the charging equipment fault diagnosis model specifically comprises two stages, wherein the first stage is pre-training, and each limited Boltzmann machine is trained according to the sequence from bottom to top, so that the extraction of the high-level characteristics of the input data and the initialization of the connection weight of the training network are realized; the second stage is fine tuning, which is to fine tune the parameters according to the sequence from top to bottom by using a back propagation algorithm and tag data to further optimize the network structure. After the above process is completed, the performance of the fault diagnosis model is evaluated using the test set data.
The beneficial effect of this application lies in:
(1) according to the method for diagnosing the faults of the IGBT type common direct current bus charging equipment of the electric automobile based on deep learning, after data of the charging equipment are preprocessed, implicit characteristics in original data of the charging equipment are extracted by using the deep learning method, and mapping between the fault data and the fault types is achieved;
(2) the electric vehicle IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning is based on data driving, an accurate charging equipment physical model does not need to be established, dependence on a traditional signal processing technology and priori knowledge is reduced, and the method has strong universality;
(3) according to the electric vehicle IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning, a Nesterov momentum method and an independent self-adaptive learning rate are added on the basis of a deep confidence network, a self-adaptive deep confidence network model is constructed, the convergence speed of the model is improved, the improved particle swarm optimization is used for optimizing the hyper-parameters of the network, and the classification performance of the charging equipment fault diagnosis model is further enhanced;
(4) the method for diagnosing the faults of the IGBT type common direct current bus charging equipment of the electric automobile based on deep learning provides a feasible implementation case for diagnosing the faults of the high-power charging equipment of the electric automobile.
Drawings
FIG. 1 is a schematic flow chart of a deep learning-based fault diagnosis method for an IGBT type common DC bus charging device of an electric vehicle according to the invention;
FIG. 2 is a schematic structural diagram of an IGBT type common DC bus charging device of an electric vehicle according to the present invention;
FIG. 3 is a flowchart of the Nesterov momentum algorithm of the present invention;
FIG. 4 is a flow chart of an improved particle swarm algorithm of the present invention.
Detailed Description
The following describes the embodiments of the present invention in further detail with reference to the drawings. The procedures, conditions, experimental methods and the like for carrying out the present invention are general knowledge and common general knowledge in the art except for the contents specifically mentioned below, and the present invention is not particularly limited.
Fig. 1 is a schematic flow chart of a deep learning-based fault diagnosis method for an electric vehicle IGBT-type common dc bus charging device according to the present invention. The structure of the IGBT-type common dc bus charging device selected in this embodiment is shown in fig. 2, and the fault diagnosis method for the IGBT-type common dc bus charging device includes the following steps:
step 1: and analyzing the running state data of the IGBT type common direct current bus charging equipment of the electric automobile, and preprocessing the running state data to establish a data set. Specifically, a data set is constructed by selecting data samples of data types including, but not limited to, charging device input voltage, charging device input current, charging device output voltage, charging device output current, charging device output power, vehicle charging demand voltage, vehicle charging demand current, charging device direct current bus voltage, direct current breaker state, battery temperature, battery set temperature, cooling fan state, auxiliary unit output voltage, charging duration, insulation detection module state, and the like, and the data samples are labeled according to fault types.
Further, the preprocessing of the data comprises the following operations:
(1) processing missing values, and respectively executing interpolation or deletion operation according to the importance degree of the data;
(2) processing abnormal values, namely executing deletion operation on abnormal data in the data set;
(3) and (3) normalizing the data, mapping the range of the data between [0,1] by a range normalization method, and calculating according to the formula:
in the formula, xminAnd xmaxRespectively representing the minimum value and the maximum value in the charging equipment running state data sample; x is the number ofoutIs input data xinAnd (5) normalizing the result.
Step 2: the data set is divided into a training set, a validation set, and a test set. Specifically, the data set is divided into a training set, a verification set and a test set according to the ratio of 6:2:2, the training set data is used for building a fault diagnosis model, and the verification set and the test set data are used for evaluating the classification performance of the fault diagnosis model.
And step 3: and establishing a charging equipment fault diagnosis model based on the self-adaptive deep confidence network, and training the charging equipment fault diagnosis model by using a training set. The training process of the self-adaptive deep confidence network is divided into two stages of pre-training and fine tuning.
The pre-training of the self-adaptive deep belief network adopts a layer-by-layer greedy learning algorithm, and each limited Boltzmann machine is trained according to the sequence from bottom to top, so that the extraction of high-level characteristics of input data and the initialization of the connection weight of the training network are realized. The restricted Boltzmann machine is used as a basic composition unit of the deep confidence network and is a probability distribution model based on energy. It consists of a visible layer v ═ (v)1,v2,v3,…,vn)TAnd a hidden layer h ═ (h)1,h2,h3,…,hm)TAnd (4) forming.
For a set of states (v, h) of visible and hidden layer elements, the energy function of the restricted boltzmann machine can be expressed as:
E(v,h|θ)=-aTv-bTh-vTwh (2)
where θ ═ { w, a, b } is the parameter set of the constrained boltzmann machine, where w ∈ Rn×mIs a visible layer and a hidden layerA ∈ R of the connection weight matrix ofnIs the visible layer bias, b ∈ RmIs the hidden layer bias.
From equation (2), the joint probability distribution of the visible layer and the hidden layer of the restricted boltzmann machine can be calculated:
wherein Z (theta) is Σv,hexp (-E (v, h | θ)) is called the partition function.
The edge distribution of the visible layer node unit can be calculated by equation (3):
according to the structural characteristics of the limited boltzmann machine, the probability that the ith unit of the visible layer and the jth unit of the hidden layer are activated can be expressed as:
wherein σ (x) is 1/(1+ e)-x)。
Further, in the training process of the limited Boltzmann machine, a Nesterov momentum method is used to accelerate the training process of the model, and the Nesterov momentum algorithm flow is shown in FIG. 3. When the algorithm is executed, temporary parameter updating is needed before each parameter updating, namelyFor calculating the parameter gradient, which results in slower operation speed than typical momentum algorithms. To solve the problem, the equivalent expression of the Nesterov momentum method is obtained by carrying out the following transformation:
Where θ ═ w, a, b represents parameters of the network, a and b represent offsets of the visible layer and the hidden layer, respectively, w represents a connection weight between the visible layer and the hidden layer, and v represents a connection weight between the visible layer and the hidden layertRepresenting the velocity after the t-th iteration, vt-1Represents the speed after the t-1 iteration, beta is the momentum value, and alpha is the learning rate.
Further, an independent adaptive learning rate is introduced in the training of the limited boltzmann machine to achieve a satisfactory classification effect and training speed, and the updating method of the parameters of the limited boltzmann machine can be expressed as follows:
in the formula (I), the compound is shown in the specification,representing the gradient of the parameter theta after the t-th training, if the weight gradient direction of the current parameter is consistent with the weight gradient direction of the previous training, psitXi is increased, and psi is obtained otherwisetReduced by a factor of 1-xi, the initial value of the parameter ψ is set to 1, and the parameter ξ is set to 0.1.
Further, the hyper-parameters of the self-adaptive deep confidence network are selected by means of an improved particle swarm optimization. At present, the hyper-parameters of the deep confidence network are obtained by selecting a plurality of different sets of parameters for debugging and comparison through experience, which usually needs to consume a great deal of time and energy. Aiming at the problem, the invention provides a method for searching the optimal hyper-parameter of the network by using a particle swarm algorithm and improving the particle swarm algorithm. In the operation process of the improved particle swarm optimization, the inertia factors are adjusted in a random weight mode to reduce training time. The flow chart of the improved particle swarm algorithm is shown in fig. 4, and specifically comprises the following steps:
step 3.1: and (5) initializing parameters. Determining the number N of particles of a particle swarm, wherein each particle of the particle swarm represents a group of depth confidence network parameters to be optimized; maximum iteration number M;
step 3.2: within a certain range, the position X of each particlei d=0And velocity Vi d=0Carrying out initialization;
step 3.3: adjusting an inertia factor of the particle swarm algorithm by adopting a random weight method, wherein the calculation formula is as follows:
In the formula, muminMinimum value of random inertial weight, mumaxMaximum value of random inertial weight, U (0,1) is [0,1]]N (0,1) is a random number following a standard normal distribution, σ is used to measure the degree of deviation between the random inertial weight ω and its mathematical expectation;
step 3.4: and taking the classification error of the deep belief network as a fitness function of the particle swarm, and evaluating each particle. Searching particles with maximum fitness in current iteration historical dataAnd globally optimal particles
Step 3.5: the position and velocity of each particle is updated using the following equation,
in the formula, Vi d+1Is the updated velocity of the ith particle, ω is the inertia factor, c1,c2The value range is [0,2 ] for learning factor],r1,r2Are randomly distributed in [0,1]]A random number in between, and a random number,is the position of the ith particle after the update;
step 3.6: recalculating the classification error by using the updated particles, and judging whether an iteration stop condition is met, wherein the iteration stop condition is that the current iteration time reaches the maximum iteration time, namely d is equal to M, or the classification error is lower than a set threshold;
step 3.7: if the iteration stop condition is met, determining the current global optimal particle as the hyper-parameter of the adaptive depth confidence network;
step 3.8: and if the iteration stop condition is not met, re-executing the step 3.3 and entering the next iteration.
In order to obtain network parameters of the deep confidence network, a contrast Divergence algorithm (CD-k) is used for quickly training a limited Boltzmann machine, and a gradient approximation calculation formula of each parameter of the limited Boltzmann machine is as follows:
in the formula, the parameter k represents the number of sampling times of gibbs sampling.
In the fine adjustment stage, a back propagation algorithm is adopted, the parameters are finely adjusted according to the sequence from top to bottom by using the label data, and the network structure is further optimized, so that the classification error is reduced, and the classification accuracy is improved. Unlike unsupervised pre-training, supervised fine tuning updates all parameters simultaneously until an iteration stop condition is reached.
And 4, step 4: the performance of the fault diagnosis model is evaluated using the validation set and the test set. After the fault diagnosis model is obtained, its performance needs to be evaluated. And if the test result accords with the expected effect, applying the test result to an actual project, otherwise, adjusting the structure and parameters of the network, and reconstructing a charging pile fault diagnosis model.
And 5: and inputting the running state data of the charging equipment into a charging equipment fault diagnosis model meeting the requirements to obtain a diagnosis result. The operator can refer to the information to check and maintain the charging equipment, so that the charging equipment is quickly restored to a normal operation state.
Although the present invention has been described with reference to the preferred embodiments, it is not limited thereto. Various changes and modifications may be effected therein by one skilled in the art without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (10)
1. A deep learning-based fault diagnosis method for an electric automobile IGBT type common direct current bus charging device is characterized by comprising the following steps:
step 1: analyzing the running state data of the IGBT type common direct current bus charging equipment of the electric automobile, and preprocessing the running state data to establish a data set;
step 2: dividing a data set into a training set, a verification set and a test set;
and step 3: establishing a charging equipment fault diagnosis model based on a self-adaptive deep confidence network, and training the charging equipment fault diagnosis model by using a training set;
and 4, step 4: evaluating the performance of the fault diagnosis model by using the verification set and the test set;
and 5: and inputting the running state data of the charging equipment into a charging equipment fault diagnosis model meeting the requirements to obtain a diagnosis result.
2. The deep learning-based fault diagnosis method for the IGBT-type common DC bus charging equipment of the electric vehicle according to claim 1, wherein the operation status of the charging equipment in step 1 is divided into a normal status and a fault status, wherein the fault status includes, but is not limited to, a DC bus output overvoltage fault, a DC bus output overcurrent fault, a DC bus output contactor fault, an electric vehicle BMS communication fault, a charging equipment insulation detection fault, a charging module over-temperature fault, a charging module input open-phase fault, a charging equipment output overvoltage fault, a charging equipment connector fault, and the like.
3. The method for diagnosing the faults of the IGBT type common direct current bus charging equipment of the electric vehicle based on deep learning according to claim 1, wherein the operation state data of the charging equipment in the step 1 comprises but is not limited to charging equipment input voltage, charging equipment input current, charging equipment output voltage, charging equipment output current, charging equipment output power, vehicle charging demand voltage, vehicle charging demand current, charging equipment direct current bus voltage, direct current breaker state, battery temperature, battery set temperature, cooling fan state, auxiliary unit output voltage, charging duration, insulation detection module state and other information.
4. The deep learning-based electric vehicle IGBT type common direct current bus charging equipment fault diagnosis method based on the claim 1 is characterized in that the data set is preprocessed in the step 1, and the method specifically comprises the following operations:
(1) processing missing values, and respectively executing interpolation or deletion operation according to the importance degree of the data;
(2) processing abnormal values, namely executing deletion operation on abnormal data in the data set;
(3) and (4) normalizing the data, and mapping the range of the data between [0,1] by a range normalization method.
5. The deep learning-based electric vehicle IGBT type common direct current bus charging equipment fault diagnosis method according to claim 1 is characterized in that the adaptive deep belief network in the step 3 optimizes network parameters by using a Nesterov momentum method, and performs equivalent transformation on a Nesterov initial calculation formula to accelerate training efficiency, wherein the Nesterov momentum method has the initial calculation formula:
θt=θt-1+vt (2)
Let thetat=θ′tThen there is
Where θ ═ w, a, b represents parameters of the network, a and b represent offsets of the visible layer and the hidden layer, respectively, w represents a connection weight between the visible layer and the hidden layer, and v represents a connection weight between the visible layer and the hidden layertRepresenting the velocity after the t-th iteration, vt-1Represents the speed after the t-1 iteration, beta is the momentum value, and alpha is the learning rate.
6. The deep learning-based electric vehicle IGBT type common direct current bus charging equipment fault diagnosis method according to claim 1 is characterized in that the adaptive deep belief network in the step 3 updates network parameters by using an independent adaptive learning rate on the basis of introducing Nesterov momentum, and the calculation formula is as follows:
in the formula (I), the compound is shown in the specification,representing the gradient of the parameter theta after the t-th training, if the weight gradient direction of the current parameter is consistent with the weight gradient direction of the previous training, psitXi is increased, and psi is obtained otherwisetReduced by a factor of 1-xi, the initial value of the parameter ψ is set to 1, and the parameter ξ is set to 0.1.
7. The deep learning-based electric vehicle IGBT type common direct current bus charging equipment fault diagnosis method according to claim 1 is characterized in that the adaptive deep belief network in the step 3 adopts an improved particle swarm algorithm to select the hyper-parameters of the network, the inertia factors of the improved particle swarm algorithm are updated in a random weight mode, and the calculation formula is as follows:
in the formula, muminMinimum value of random inertial weight, mumaxMaximum value of random inertial weight, U (0,1) is [0,1]]N (0,1) is a random number that follows a standard normal distribution, and σ is used to measure the degree of deviation between the random inertial weight ω and its mathematical expectation.
8. The deep learning-based electric vehicle IGBT type common direct current bus charging equipment fault diagnosis method according to claim 1 is characterized in that the process of training the adaptive deep confidence network by using the training set data in the step 3 to obtain the charging equipment fault diagnosis model specifically comprises two stages. The first stage is pre-training, each limited Boltzmann machine is trained according to the sequence from bottom to top, so as to realize the extraction of high-level characteristics of input data and the initialization of network parameters; the second stage is fine tuning, which is to fine tune the parameters according to the sequence from top to bottom by using a back propagation algorithm and tag data to further optimize the network structure. After the above process is completed, the performance of the fault diagnosis model is evaluated using the test set data.
9. The method for diagnosing the faults of the IGBT type common direct current bus charging equipment of the electric vehicle based on deep learning according to claim 1, wherein the IGBT type common direct current bus charging equipment of the electric vehicle in the step 1 refers to the charging equipment with the maximum output power of more than 120kW and less than 360 kW.
10. The method for diagnosing the faults of the IGBT type common direct current bus charging equipment of the electric vehicle based on deep learning according to claim 1, wherein the IGBT type common direct current bus charging equipment of the electric vehicle in the step 1 mainly comprises an AC/DC converter part, a common direct current bus part and a DC/DC converter part.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110737833.6A CN113433409A (en) | 2021-06-30 | 2021-06-30 | Electric automobile IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110737833.6A CN113433409A (en) | 2021-06-30 | 2021-06-30 | Electric automobile IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113433409A true CN113433409A (en) | 2021-09-24 |
Family
ID=77758155
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110737833.6A Pending CN113433409A (en) | 2021-06-30 | 2021-06-30 | Electric automobile IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113433409A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114184867A (en) * | 2021-12-10 | 2022-03-15 | 清湖光旭数据科技(北京)有限公司 | Charging facility fault detection method and system based on deep learning |
CN117520803A (en) * | 2023-10-23 | 2024-02-06 | 国网江苏省电力有限公司南京供电分公司 | Charging pile fault diagnosis method, computer readable medium and equipment |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012247308A (en) * | 2011-05-27 | 2012-12-13 | Denso Corp | Charging device and server |
CN106503790A (en) * | 2015-08-28 | 2017-03-15 | 余学飞 | A kind of Pressure wire temperature compensation of Modified particle swarm optimization neutral net |
CN106548230A (en) * | 2016-10-14 | 2017-03-29 | 云南电网有限责任公司昆明供电局 | Diagnosis Method of Transformer Faults based on Modified particle swarm optimization neutral net |
CN106769048A (en) * | 2017-01-17 | 2017-05-31 | 苏州大学 | Adaptive deep confidence network bearing fault diagnosis method based on Nesterov momentum method |
CN109063785A (en) * | 2018-08-23 | 2018-12-21 | 国网河北省电力有限公司沧州供电分公司 | charging pile fault detection method and terminal device |
CN109213122A (en) * | 2018-08-10 | 2019-01-15 | 合肥工业大学 | Method for diagnosing faults and computer storage medium for stamping equipment |
CN109816144A (en) * | 2018-12-18 | 2019-05-28 | 广东电网有限责任公司 | The short-term load forecasting method of distributed memory parallel computation optimization deepness belief network |
CN109886328A (en) * | 2019-02-14 | 2019-06-14 | 国网浙江省电力有限公司电力科学研究院 | A kind of electric car electrically-charging equipment failure prediction method and system |
CN110378286A (en) * | 2019-07-19 | 2019-10-25 | 东北大学 | A kind of Power Quality Disturbance classification method based on DBN-ELM |
CN111174370A (en) * | 2018-11-09 | 2020-05-19 | 珠海格力电器股份有限公司 | Fault detection method and device, storage medium and electronic device |
CN112036598A (en) * | 2020-06-24 | 2020-12-04 | 国网天津市电力公司电力科学研究院 | Charging pile use information prediction method based on multi-information coupling |
-
2021
- 2021-06-30 CN CN202110737833.6A patent/CN113433409A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012247308A (en) * | 2011-05-27 | 2012-12-13 | Denso Corp | Charging device and server |
CN106503790A (en) * | 2015-08-28 | 2017-03-15 | 余学飞 | A kind of Pressure wire temperature compensation of Modified particle swarm optimization neutral net |
CN106548230A (en) * | 2016-10-14 | 2017-03-29 | 云南电网有限责任公司昆明供电局 | Diagnosis Method of Transformer Faults based on Modified particle swarm optimization neutral net |
CN106769048A (en) * | 2017-01-17 | 2017-05-31 | 苏州大学 | Adaptive deep confidence network bearing fault diagnosis method based on Nesterov momentum method |
CN109213122A (en) * | 2018-08-10 | 2019-01-15 | 合肥工业大学 | Method for diagnosing faults and computer storage medium for stamping equipment |
CN109063785A (en) * | 2018-08-23 | 2018-12-21 | 国网河北省电力有限公司沧州供电分公司 | charging pile fault detection method and terminal device |
CN111174370A (en) * | 2018-11-09 | 2020-05-19 | 珠海格力电器股份有限公司 | Fault detection method and device, storage medium and electronic device |
CN109816144A (en) * | 2018-12-18 | 2019-05-28 | 广东电网有限责任公司 | The short-term load forecasting method of distributed memory parallel computation optimization deepness belief network |
CN109886328A (en) * | 2019-02-14 | 2019-06-14 | 国网浙江省电力有限公司电力科学研究院 | A kind of electric car electrically-charging equipment failure prediction method and system |
CN110378286A (en) * | 2019-07-19 | 2019-10-25 | 东北大学 | A kind of Power Quality Disturbance classification method based on DBN-ELM |
CN112036598A (en) * | 2020-06-24 | 2020-12-04 | 国网天津市电力公司电力科学研究院 | Charging pile use information prediction method based on multi-information coupling |
Non-Patent Citations (11)
Title |
---|
DEXIN GAO 等: "Fault Diagnosis Method of DC Charging Points for EVs Based on Deep Belief Network", 《WORLD ELECTRIC VEHICLE JOURNAL》 * |
JIANG H K 等: "Intelligent fault diagnosis of rolling bearings using an improved deep recurrent neural network", 《MEASUREMNT SCIENCE AND TECHNOLOGY》 * |
专祥涛: "《最优化方法基础》", 31 March 2018 * |
巨汉基等: "电动汽车充电桩数据采集终端设计", 《自动化与仪表》 * |
常雪松: "基于数据挖掘的电动汽车充电桩故障预测方法研究", 《万方数据库》 * |
林越等: "基于AP-HMM混合模型的充电桩故障诊断", 《广西师范大学学报(自然科学版)》 * |
沈长青等: "独立自适应学习率优化深度信念网络在轴承故障诊断中的应用研究", 《机械工程学报》 * |
赵翔 等: "一种基于深度神经网络的直流充电桩故障诊断方法", 《电测与仪表》 * |
郑鹏飞等: "基于深度信念网络在船用齿轮箱故障诊断中的应用研究", 《中国修船》 * |
闭应洲 等: "《数据挖掘与机器学习》", 31 January 2020 * |
顾兴健等: "基于LSTM神经网络的我国典型试航海域环境短期预报方法研究", 《中国造船》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114184867A (en) * | 2021-12-10 | 2022-03-15 | 清湖光旭数据科技(北京)有限公司 | Charging facility fault detection method and system based on deep learning |
CN117520803A (en) * | 2023-10-23 | 2024-02-06 | 国网江苏省电力有限公司南京供电分公司 | Charging pile fault diagnosis method, computer readable medium and equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110262463B (en) | Rail transit platform door fault diagnosis system based on deep learning | |
CN113433409A (en) | Electric automobile IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning | |
CN112415414A (en) | Method for predicting remaining service life of lithium ion battery | |
CN113283491A (en) | Fault diagnosis method of electric vehicle alternating current charging pile based on optimized deep confidence network | |
CN113159077A (en) | Three-phase inverter fault identification method based on hybrid convolutional neural network | |
CN112069727B (en) | Intelligent transient stability evaluation system and method with high reliability for power system | |
CN110298374B (en) | Driving track energy consumption analysis method and device based on deep learning | |
CN109858564B (en) | Improved Adaboost-SVM model generation method suitable for wind power converter fault diagnosis | |
CN112288147B (en) | Method for predicting insulation state of generator stator by BP-Adaboost strong predictor | |
CN116205140A (en) | IGBT residual service life prediction method based on LSTM-ARIMA | |
CN112784920A (en) | Cloud-side-end-coordinated dual-anti-domain self-adaptive fault diagnosis method for rotating part | |
CN113626586A (en) | Fault text analysis processing method for magnetic-levitation train | |
CN115099296A (en) | Sea wave height prediction method based on deep learning algorithm | |
CN117154690A (en) | Photovoltaic power generation power prediction method and system based on neural network | |
CN112766618A (en) | Anomaly prediction method and device | |
CN116484299A (en) | Charging pile fault diagnosis method based on integration of gradient lifting tree and multi-layer perceptron | |
CN116665483A (en) | Novel method for predicting residual parking space | |
CN110059938B (en) | Power distribution network planning method based on association rule driving | |
CN113791351B (en) | Lithium battery life prediction method based on transfer learning and difference probability distribution | |
CN111414943A (en) | Anomaly detection method based on mixed hidden naive Bayes model | |
CN114692956A (en) | Charging facility load prediction method and system based on multilayer optimization kernel limit learning machine | |
CN112731098B (en) | Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application | |
CN113640633A (en) | Fault positioning method for gas insulated switchgear | |
CN112748331A (en) | Circuit breaker mechanical fault identification method and device based on DS evidence fusion | |
CN112966429A (en) | Non-linear industrial process modeling method based on WGANs data enhancement |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210924 |