Data processing method and device based on neural network
The experimental data of the fuel cell system is preprocessed through convolutional neural networks and screened out effective data, solving the problems of low processing efficiency and poor reliability caused by large data volume and complexity, and achieving rapid and effective data analysis.
Patent Information
- Application Number
- CN202310546750.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-15
- Publication Date
- 2024-11-15
AI Technical Summary
The experimental data generated by the fuel cell system during the development and verification process is large and complex, resulting in a long processing time, affecting work efficiency, and it is difficult to quickly detect abnormalities in the event of system failure, affecting the reliability of the test.
The data processing method based on neural network is adopted to train pre-label samples through convolutional neural networks, generate a trained network model, and input experimental data into the model, output target categories to filter effective data, and improve data analysis and processing efficiency.
By quickly screening out target data that can participate in subsequent processing, the efficiency of data analysis and processing is improved, the consumption of manpower and time is reduced, and the reliability of data processing is enhanced.
Smart Images

Figure CN118962449A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fuel cell data analysis, and in particular to a data processing method and device based on a neural network. Background Art
[0002] A large amount of experimental data will be generated during the development and verification process of the fuel cell system. Each piece of data contains various indicators about the performance of the fuel cell system. Analyzing the experimental data is a key means to understand and develop new systems.
[0003] However, currently, a large amount of testing and verification is required during the system development process. The huge amount of data generated takes too long to process and analyze, and the complexity of the data also affects work efficiency during comparative analysis and call analysis. In addition, when a system failure occurs, it is difficult to find the abnormality of the system in the complex data, which in turn affects the reliability of the test. Summary of the invention
[0004] The purpose of the present invention is to provide a data processing method and device based on a neural network to alleviate the technical problem that experimental data is too complex and affects processing efficiency and reliability.
[0005] In a first aspect, an embodiment of the present invention provides a data processing method based on a neural network, comprising:
[0006] Train the convolutional neural network based on the pre-labeled samples to obtain a trained network model;
[0007] Inputting experimental data of the fuel cell system into the trained network model, and outputting the experimental data and the target category corresponding to the experimental data;
[0008] Based on the target category, the experimental data is processed accordingly.
[0009] In combination with the first aspect, an embodiment of the present invention provides a first possible implementation of the first aspect, wherein the step of training a convolutional neural network according to pre-labeled samples to obtain a trained network model includes:
[0010] Inputting the pre-labeled sample into a convolutional neural network, and comparing the sample result output by the convolutional neural network with the pre-labeled sample;
[0011] If the comparison result is less than the preset threshold, the training of the convolutional neural network meets expectations;
[0012] If the comparison result is not less than the preset threshold, the step of inputting the pre-labeled sample into the convolutional neural network is continued until the training of the convolutional neural network reaches the expected level.
[0013] In combination with the first aspect, an embodiment of the present invention provides a second possible implementation of the first aspect, wherein the step of inputting a pre-labeled sample into a convolutional neural network and comparing the sample result output by the convolutional neural network with the pre-labeled sample includes:
[0014] Marking invalid data in the fuel cell system experimental data to obtain a first pre-marked sample;
[0015] Inputting the first pre-labeled sample into a convolutional neural network, and outputting a first sample result corresponding to the first pre-labeled sample;
[0016] Re-marking the target feature points in the first pre-marked sample according to the first sample result, and removing invalid data in the first pre-marked sample to obtain a second pre-marked sample;
[0017] Inputting the second pre-labeled sample into the convolutional neural network, and outputting a second sample result corresponding to the second pre-labeled sample;
[0018] The second sample result is compared with the second pre-labeled sample.
[0019] In combination with the first aspect, an embodiment of the present invention provides a third possible implementation of the first aspect, wherein the step of inputting experimental data of the fuel cell system into the trained network model, and outputting the experimental data and the target category corresponding to the experimental data includes:
[0020] Clean the experimental data of the fuel cell system;
[0021] Inputting the cleaned experimental data into the trained network model to obtain a first target category corresponding to the experimental data, wherein the first target category is used to characterize the validity of the experimental data;
[0022] The experimental data after invalid data is eliminated is input into the trained network model to obtain a second target category corresponding to the experimental data after invalid data is eliminated. The second target category is used to evaluate the probability value of whether each test quantity in the experimental data is valid.
[0023] In combination with the first aspect, the embodiment of the present invention provides a fourth possible implementation of the first aspect, wherein the step of processing the experimental data accordingly based on the target category includes:
[0024] According to the comparison result of the second target category corresponding to the experimental data and the preset category threshold, the target experimental data participating in the data processing operation is determined.
[0025] In combination with the first aspect, an embodiment of the present invention provides a fifth possible implementation of the first aspect, wherein the step of determining the target experimental data participating in the data processing operation according to the comparison result of the second target category corresponding to the experimental data and the preset category threshold includes:
[0026] Determine a comprehensive score corresponding to each test quantity in the experimental data according to a weight coefficient corresponding to each test quantity in the experimental data after invalid data is eliminated and the second target category;
[0027] The comprehensive score corresponding to each test quantity is compared with the corresponding preset category threshold, and the target experimental data participating in the data processing operation is determined according to the comparison result.
[0028] In combination with the first aspect, an embodiment of the present invention provides a sixth possible implementation of the first aspect, wherein, before the step of training the convolutional neural network according to the pre-labeled samples, the method further includes:
[0029] A convolutional neural network is constructed, wherein the convolutional neural network includes five convolutional pooling layers and three fully connected layers; each of the convolutional pooling layers includes connected convolutional layers and pooling layers, and the convolutional layer includes 6 5×5 convolutional kernels.
[0030] In a second aspect, an embodiment of the present invention further provides a data processing device based on a neural network, comprising:
[0031] The training module trains the convolutional neural network based on the pre-labeled samples to obtain a trained network model;
[0032] An application module, inputting the experimental data of the fuel cell system into the trained network model, and outputting the experimental data and the target category corresponding to the experimental data;
[0033] The processing module processes the experimental data accordingly based on the target category.
[0034] In a third aspect, an embodiment provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method described in any of the aforementioned implementation methods are implemented.
[0035] In a fourth aspect, an embodiment provides a machine-readable storage medium, wherein the machine-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the steps of the method described in any one of the aforementioned implementations.
[0036] The embodiment of the present invention provides a data processing method and device based on a neural network. A convolutional neural network model is trained by pre-marking samples of experimental data validity, and a target category corresponding to the experimental data of an actual fuel cell system is determined based on the network model. If the target category is valid, the experimental data corresponding to the target category is used to participate in subsequent processing operations. The target data that can participate in subsequent processing operations can be quickly screened out from the complex experimental data, thereby improving the efficiency of data analysis and processing.
[0037] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and the drawings.
[0038] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0040] Figure 1 A flow chart of a data processing method based on a neural network provided by an embodiment of the present invention;
[0041] Figure 2 A schematic diagram of functional modules of a data processing device based on a neural network provided by an embodiment of the present invention;
[0042] Figure 3 A schematic diagram of the hardware architecture of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0043] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0044] At present, tools such as Excel are usually used to create a data processing template, and then specialized personnel conduct specific analysis on the data and finally output the required analysis results.
[0045] However, this traditional method cannot completely solve the work efficiency problems caused by huge amounts of data. It still takes up a lot of manpower and time, and cannot output the required analysis results in a timely manner.
[0046] Based on this, an embodiment of the present invention provides a neural network-based data processing method and device, which pre-processes complex experimental data through a convolutional neural network, and performs subsequent data processing operations according to the output experimental category, thereby improving the processing efficiency of the experimental data.
[0047] To facilitate understanding of this embodiment, a neural network-based data processing method disclosed in an embodiment of the present invention is first introduced in detail. This method can be applied to intelligent control devices such as host computers, servers, and vehicle computers.
[0048] Figure 1 A flow chart of a data processing method based on a neural network provided in an embodiment of the present invention.
[0049] Reference Figure 1 , the method can be implemented by the following steps, specifically including:
[0050] Step S102: training the convolutional neural network based on the pre-labeled samples to obtain a trained network model.
[0051] The pre-labeled samples may be understood as samples whose experimental sample data conditions have been labeled in advance; for example, the validity of the sample data has been labeled in advance.
[0052] Step S104, inputting the experimental data of the fuel cell system into the trained network model, and outputting the experimental data and the target category corresponding to the experimental data.
[0053] It can be understood that the trained network model can automatically, quickly and conveniently determine the availability of the experimental data of the fuel cell system. The target category can be understood as whether the experimental data is valid and can be used for subsequent processing operations.
[0054] Step S106: Process the experimental data accordingly based on the target category.
[0055] Here, according to the category corresponding to the experimental data, appropriate experimental data is selected to participate in the subsequent processing operation.
[0056] In a preferred embodiment of practical application, a convolutional neural network model is trained by pre-labeling samples of experimental data validity, and the target category corresponding to the experimental data of the actual fuel cell system is determined based on the network model. If this target category is valid, the experimental data corresponding to the target category is used to participate in subsequent processing operations, which can quickly screen out target data that can participate in subsequent processing operations from the complex experimental data, thereby improving the efficiency of data analysis and processing.
[0057] It should be noted that, before step S102, the method further includes:
[0058] Step 1.1), build a convolutional neural network.
[0059] Among them, the convolutional neural network can be understood as an 8-layer neural network architecture, including five convolutional pooling layers and three fully connected layers; the five convolutional pooling layers are connected in sequence and then connected to the three fully connected layers; each convolutional pooling layer includes connected convolutional layers and pooling layers, the convolutional layer includes 6 5×5 convolution kernels, its stride strid=1, the pooling layer stride strid=2, the pooling method is global pooling, and the activation function is Sigmoid.
[0060] Here, in order for the convolutional neural network structure to accurately determine the operability of the experimental data, the convolutional neural network model needs to be trained in combination with samples; illustratively, step S102 can be implemented by the following steps, including:
[0061] In step 2.1), the pre-labeled samples are input into the convolutional neural network, and the sample results output by the convolutional neural network are compared with the pre-labeled samples.
[0062] It can be known that the convolutional neural network recognizes the input sample data and can output sample results including corresponding category recognition results. The remaining pre-labeled categories are compared with the edited samples, which can determine the recognition ability of the convolutional neural network.
[0063] Step 2.2), if the comparison result is less than the preset threshold, the training of the convolutional neural network reaches the expected result.
[0064] At this point, it can be determined that the category recognition result of the convolutional neural network is not much different from the pre-marked category result, and the comparison difference meets the preset threshold range, then the convolutional neural network meets the expected requirements.
[0065] Step 2.3), if the comparison result is not less than the preset threshold, continue to execute step 2.1) of inputting the pre-labeled sample into the convolutional neural network. After multiple trainings, convergence is achieved until the training of the convolutional neural network reaches the expected level, and the most accurate model output is finally achieved.
[0066] It should be noted that if the category recognition result of the convolutional neural network is significantly different from the pre-labeled category result, the above step 2.1) is repeated until the difference between the category recognition result of the convolutional neural network and the pre-labeled category result meets the preset threshold.
[0067] As an optional embodiment, in order to make the recognition of the convolutional neural network more accurate, the operation of the convolutional neural network structure can be combined with the Adam optimizer, wherein step 2.1) includes:
[0068] Step 2.1.1), marking invalid data in the fuel cell system experimental data to obtain a first pre-marked sample.
[0069] Here, the invalid data can be marked manually or by machine, and the first pre-marked sample marked with the invalid data is then identified by the network model.
[0070] Step 2.1.2), input the first pre-labeled sample into the convolutional neural network, and output the first sample result corresponding to the first pre-labeled sample.
[0071] The convolutional neural network has undergone a large amount of recognition training, and the first pre-labeled samples have been labeled in more detail, so that the model can eventually recognize invalid data in the first pre-labeled samples (first sample structure).
[0072] Step 2.1.3), mark the target feature points in the first pre-marked sample again according to the first sample result, and remove invalid data in the first pre-marked sample to obtain a second pre-marked sample.
[0073] Here, after invalid data is eliminated from the first pre-marked sample based on the first sample result, the feature points in the sample are continuously marked in the second round to obtain the second pre-marked sample.
[0074] Step 2.1.4), input the second pre-labeled sample into the convolutional neural network, and output the second sample result corresponding to the second pre-labeled sample.
[0075] Preferably, the convolutional neural network may also include an evaluation model structure, such as a normalized function model softmax, so that the convolutional neural model can identify and score the second pre-labeled samples and output recognition results (second sample results) of each category.
[0076] Step 2.1.5), compare the second sample result with the second pre-labeled sample.
[0077] Here, the category recognition result in the experimental data recognized for the second time can be compared with the pre-marked feature points in the experimental data to determine the reliability of the convolutional neural model.
[0078] In practical applications, the model training process for each time can be as follows: the experimental sample data is input into the convolutional neural network architecture, which is divided into five layers. Each layer performs a convolution and pooling operation. After the first layer of convolution (using 6 convolution kernels of size 5*5), 6 298*298 feature matrices are obtained. Subsequently, 5*5 kernels are used for pooling to obtain 6 149*149 feature matrices. After pooling, the second layer is entered, and the convolution output is 30 145*145 feature matrices, and the convolution kernel size is 5*5. After convolution, secondary pooling is performed to obtain 30 72*72 feature matrices. Entering the third layer again, convolution obtains 60 68*68 feature matrices, and further pooling obtains 60 34*34 feature matrices. Entering the fourth layer again, convolution obtains 120 30*30 feature matrices, and further pooling obtains 120 15*15 feature matrices. Entering the last layer, the convolution obtains 240 11*11 feature matrices, and after pooling, 240 5*5 feature matrices are obtained. Then, after three fully connected layers, 10 categories are finally obtained, and the category probability is determined by softmax to judge the final sample result; the sample result can be compared with the pre-labeled sample.
[0079] On the basis of the above-mentioned embodiment, the trained network model can be used to pre-process the experimental data of the fuel cell system so as to improve the efficiency of data processing and analysis; here, step S104 may include:
[0080] Step 3.1), cleaning the experimental data of the fuel cell system.
[0081] Among them, dirty data in the experimental data or garbled data that does not conform to the preset format can be cleaned and cleared first.
[0082] Step 3.2), input the cleaned experimental data into the trained network model to obtain the first target category corresponding to the experimental data.
[0083] The first target category is used to characterize the validity of the experimental data, that is, based on the first target category, it can be known whether each experimental data is valid.
[0084] Step 3.3), the experimental data after invalid data is removed is input into the trained network model to obtain the second target category corresponding to the experimental data after invalid data is removed.
[0085] Here, after invalid data is eliminated based on the first target category, the second target category is obtained by performing network model recognition processing again. The second target category is used to evaluate the probability value of whether each test quantity in the experimental data is valid.
[0086] On this basis, step S106 can process the test volume whose effective probability value reaches the expected value in the experimental data according to the second target category, and ignore other data whose probability value does not reach the expected value in the complex experimental data, thereby achieving the effect of greatly improving the data processing efficiency; specifically, it includes:
[0087] Step 4.1), according to the comparison result of the second target category corresponding to the experimental data and the preset category threshold, the target experimental data participating in the data processing operation is determined.
[0088] Exemplarily, the weight coefficient corresponding to each test quantity in the experimental data after invalid data is eliminated is multiplied by the second target category (the probability value of each test quantity in the experimental data being valid) to determine the comprehensive score corresponding to each test quantity in the experimental data. The comprehensive score corresponding to each test quantity is then compared with the corresponding preset category threshold. If the comprehensive score of the test quantity exceeds the corresponding preset category threshold, the test quantity can be identified as the target experimental data participating in the subsequent data processing operation.
[0089] The trained convolutional neural network model of the embodiment of the present invention can process a large amount of data in batches. After obtaining the daily operation data of the vehicle and / or the experimental data of the fuel cell system, the data is given to the trained model. First, the data is cleaned, invalid data is marked according to the training results, and the marking results are output. The invalid data is manually removed and the data is reintegrated for identification and analysis again. At this time, the test quantities that need to be marked are scored. Finally, the model comprehensively scores the test quantity conditions of the system and the vehicle under the data and outputs the effective probability results, which saves manpower and time, can achieve real-time data processing and analysis, and timely feedback on system performance.
[0090] In some embodiments, Figure 2 As shown, an embodiment of the present invention provides a data processing device based on a neural network, comprising:
[0091] The training module trains the convolutional neural network based on the pre-labeled samples to obtain a trained network model;
[0092] An application module, inputting the experimental data of the fuel cell system into the trained network model, and outputting the experimental data and the target category corresponding to the experimental data;
[0093] The processing module processes the experimental data accordingly based on the target category.
[0094] In some embodiments, the training module is further specifically used to input pre-labeled samples into a convolutional neural network, and compare the sample results output by the convolutional neural network with the pre-labeled samples; if the comparison result is less than a preset threshold, the training of the convolutional neural network meets expectations; if the comparison result is not less than the preset threshold, continue to input the pre-labeled samples into the convolutional neural network until the training of the convolutional neural network meets expectations.
[0095] In some embodiments, the training module is further specifically used to mark invalid data in the fuel cell system experimental data to obtain a first pre-marked sample; input the first pre-marked sample into a convolutional neural network, and output a first sample result corresponding to the first pre-marked sample; mark the target feature points in the first pre-marked sample again according to the first sample result, and eliminate the invalid data in the first pre-marked sample to obtain a second pre-marked sample; input the second pre-marked sample into the convolutional neural network, and output a second sample result corresponding to the second pre-marked sample; and compare the second sample result with the pre-marked sample.
[0096] In some embodiments, the application module is further specifically used to clean the experimental data of the fuel cell system; input the cleaned experimental data into the trained network model to obtain a first target category corresponding to the experimental data, and the first target category is used to characterize the validity of the experimental data.
[0097] In some embodiments, the application module is further specifically used to input the experimental data after invalid data is eliminated into the trained network model to obtain a second target category corresponding to the experimental data after invalid data is eliminated, and the second target category is used to evaluate the probability value of whether each test quantity in the experimental data is valid.
[0098] In some embodiments, the processing module is further specifically used to determine the target experimental data involved in the data processing operation according to the comparison result between the second target category corresponding to the experimental data and the preset category threshold.
[0099] In some embodiments, the processing module is further specifically used to determine a comprehensive score corresponding to each test quantity in the experimental data based on the weight coefficient corresponding to each test quantity in the experimental data after eliminating invalid data and the second target category; compare the comprehensive score corresponding to each test quantity with the corresponding preset category threshold, and determine the target experimental data participating in the data processing operation based on the comparison results.
[0100] In some embodiments, before the step of training the convolutional neural network based on the pre-labeled samples, the device also includes: a construction module for constructing a convolutional neural network, wherein the convolutional neural network includes five convolutional pooling layers and three fully connected layers; each of the convolutional pooling layers includes connected convolutional layers and pooling layers, and the convolutional layer includes 6 5×5 convolution kernels.
[0101] An embodiment of the present invention provides an electronic device for implementing an electronic device. In this embodiment, the electronic device may be, but is not limited to, a personal computer (PC), a laptop computer, a monitoring device, a server, or other computer device with analysis and processing capabilities.
[0102] As an exemplary embodiment, see Figure 3 The electronic device 110 includes a communication interface 111, a processor 112, a memory 113 and a bus 114. The processor 112, the communication interface 111 and the memory 113 are connected via the bus 114. The memory 113 is used to store a computer program that supports the processor 112 to execute the method. The processor 112 is configured to execute the program stored in the memory 113.
[0103] The machine-readable storage medium mentioned in this article can be any electronic, magnetic, optical or other physical storage device that can contain or store information, such as executable instructions, data, etc. For example, the machine-readable storage medium can be: RAM (Radom Access Memory), volatile memory, non-volatile memory, flash memory, storage drive (such as hard disk drive), any type of storage disk (such as CD, DVD, etc.), or similar storage medium, or a combination thereof.
[0104] The non-volatile medium may be a non-volatile memory, a flash memory, a storage drive (such as a hard drive), any type of storage disk (such as a CD, DVD, etc.), or a similar non-volatile storage medium, or a combination thereof.
[0105] It can be understood that the specific operation methods of each functional module in this embodiment can refer to the detailed description of the corresponding steps in the above method embodiment, and will not be repeated here.
[0106] The computer-readable storage medium provided in the embodiment of the present invention stores a computer program. When the computer program code is executed, the method described in any of the above embodiments can be implemented. For specific implementation, please refer to the method embodiment, which will not be described in detail here.
[0107] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.
[0108] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0109] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.
[0110] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the aforementioned embodiments, those of ordinary skill in the art should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed by the present invention, or can easily conceive of changes, or make equivalent replacements for some of the technical features therein. Such modifications, changes or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention.
Claims
1. A data processing method based on a neural network, characterized in that: include: Train the convolutional neural network based on the pre-labeled samples to obtain a trained network model; Inputting experimental data of the fuel cell system into the trained network model, and outputting the experimental data and the target category corresponding to the experimental data; Based on the target category, the experimental data is processed accordingly.
2. The method according to claim 1, characterized in that The steps of training the convolutional neural network based on the pre-labeled samples to obtain the trained network model include: Inputting the pre-labeled sample into a convolutional neural network, and comparing the sample result output by the convolutional neural network with the pre-labeled sample; If the comparison result is less than the preset threshold, the training of the convolutional neural network meets expectations; If the comparison result is not less than the preset threshold, the step of inputting the pre-labeled sample into the convolutional neural network is continued until the training of the convolutional neural network reaches the expected level.
3. The method according to claim 2, characterized in that The step of inputting a pre-labeled sample into a convolutional neural network and comparing a sample result output by the convolutional neural network with the pre-labeled sample comprises: Marking invalid data in the fuel cell system experimental data to obtain a first pre-marked sample; Inputting the first pre-labeled sample into a convolutional neural network, and outputting a first sample result corresponding to the first pre-labeled sample; Re-marking the target feature points in the first pre-marked sample according to the first sample result, and removing invalid data in the first pre-marked sample to obtain a second pre-marked sample; Inputting the second pre-labeled sample into the convolutional neural network, and outputting a second sample result corresponding to the second pre-labeled sample; The second sample result is compared with the second pre-labeled sample.
4. The method according to claim 1, characterized in that: The step of inputting the experimental data of the fuel cell system into the trained network model and outputting the experimental data and the target category corresponding to the experimental data comprises: Clean the experimental data of the fuel cell system; Inputting the cleaned experimental data into the trained network model to obtain a first target category corresponding to the experimental data, wherein the first target category is used to characterize the validity of the experimental data; The experimental data after invalid data is eliminated is input into the trained network model to obtain a second target category corresponding to the experimental data after invalid data is eliminated. The second target category is used to evaluate the probability value of whether each test quantity in the experimental data is valid.
5. The method according to claim 4, characterized in that The step of processing the experimental data accordingly based on the target category includes: According to the comparison result of the second target category corresponding to the experimental data and the preset category threshold, the target experimental data participating in the data processing operation is determined.
6. The method according to claim 5, characterized in that The step of determining the target experimental data involved in the data processing operation according to the comparison result of the second target category corresponding to the experimental data and the preset category threshold comprises: Determine a comprehensive score corresponding to each test quantity in the experimental data according to a weight coefficient corresponding to each test quantity in the experimental data after invalid data is eliminated and the second target category; The comprehensive score corresponding to each test quantity is compared with the corresponding preset category threshold, and the target experimental data participating in the data processing operation is determined according to the comparison result.
7. The method according to claim 1, characterized in that Before the step of training the convolutional neural network according to the pre-labeled samples, the method further includes: A convolutional neural network is constructed, wherein the convolutional neural network includes five convolutional pooling layers and three fully connected layers; each of the convolutional pooling layers includes connected convolutional layers and pooling layers, and the convolutional layer includes 6 5×5 convolutional kernels.
8. A data processing device based on a neural network, characterized in that: include: The training module trains the convolutional neural network based on the pre-labeled samples to obtain a trained network model; An application module, inputting the experimental data of the fuel cell system into the trained network model, and outputting the experimental data and the target category corresponding to the experimental data; The processing module processes the experimental data accordingly based on the target category.
9. An electronic device, characterized in that: The method comprises a memory, a processor, and a program stored in the memory and capable of being run on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.
10. A computer-readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 7 is implemented.
