Intelligent diagnosis method and system, computer equipment and storage medium

Abstract

The invention relates to the technical field of equipment monitoring, and discloses an intelligent diagnosis method and system, computer equipment and a storage medium, and the method comprises the steps: data collection: collecting multiple types of data information of the operation of related equipment; data labeling and weighting: labeling the existing data and the acquired data information according to the actual operation condition of the equipment, and endowing different weights to different labeled data; feature extraction: extracting a multi-scale feature vector from the labeled data set by using a specific neural network and fusing a multi-scale feature extraction network; and prediction evaluation: inputting the extracted feature vectors into a prediction model, and evaluating the operation state of the system according to a state prediction result and a score thereof output by the prediction model. Through accurate data acquisition and processing, faults are effectively predicted and diagnosed, so that the emergency maintenance and downtime of the charging pile is reduced, the maintenance cost is reduced, accidental faults during use of a user are reduced, the customer experience is improved, additional hardware is not needed, and the cost performance is high.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment monitoring, and in particular to an intelligent diagnosis method, system, computer equipment and storage medium. Background Art

[0002] In recent years, with the continuous improvement of global environmental awareness and the rapid development of new energy technologies, electric vehicles (EVs) have become one of the mainstream means of transportation. This trend has greatly promoted the construction of electric vehicle charging infrastructure, and the number of charging piles has shown a trend of substantial growth.

[0003] However, as the number of charging piles increases rapidly, many problems are also exposed. In daily use, the internal components of charging piles will age over time, and they will also be disturbed by many external uncertain factors, which can easily cause various faults, leading to interruptions in charging services and a decline in user experience. When a fault occurs, the charging pile maintenance personnel cannot arrive at the site in time for maintenance, and the charging pile will remain in a faulty state and cannot provide charging services, seriously affecting the operation of the charging pile and the user experience.

[0004] The above information is presented as background information only to assist with understanding the present disclosure and no determination or admission is made as to whether any of the above may be used as prior art with respect to the present disclosure. Summary of the invention

[0005] The present invention provides an intelligent diagnosis method, system, computer equipment and storage medium to solve the problems existing in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An intelligent diagnosis method comprises the following steps:

[0008] Data collection steps: Collect various types of data information on the operation of related equipment;

[0009] Data labeling and weighting steps: label the existing data and collected data information according to the actual operation of the equipment, and assign different weights to different labeled data;

[0010] Feature extraction step: Use a specific neural network to fuse a multi-scale feature extraction network to extract multi-scale feature vectors from the labeled data set;

[0011] Prediction and evaluation step: the extracted feature vector is input into the prediction model, and the operating state of the system is evaluated according to the state prediction result and the score output by the prediction model.

[0012] Optionally, in the data collection step, the multiple types of data information include multiple physical quantity data collected by sensors installed in the equipment, equipment operation log data obtained from the cloud platform, and equipment maintenance record data obtained from operation and maintenance channels;

[0013] In the data labeling and weighting step, for the collected data information, part of the data is randomly sent to the backend database and labeled. The new data set is given a weight a, and the old data weight is b. The calculation of a and b is determined based on the number of data updates and batches;

[0014] In the feature extraction step, the multi-scale feature extraction network is a parallel multi-branch structure, including branches with different convolution kernel sizes or different convolution combinations;

[0015] In the prediction evaluation step, the prediction model predicts the data value of the next time step based on the input historical data sequence, and the Attention network in the prediction model follows a specific formula to perform state calculation and weighted summation operations in the actual model training prediction task.

[0016] Optionally, the intelligent diagnosis method further includes a data preprocessing step:

[0017] The K-means algorithm is used to remove noise from the data. The objective function calculation formula of the K-means algorithm is:

[0018]

[0019] J is the objective function value, is the ith cluster, is the cluster center of the ith cluster, and x is the data point in the cluster. When calculating, first select K initial cluster centers, assign data points to the cluster centers closest to them to form K clusters, recalculate the cluster centers, and repeat the assignment and calculation of cluster centers until the cluster centers no longer change or change very little.

[0020] The Z-score algorithm is used to normalize the data. The calculation formula is:

[0021]

[0022] Z is the standardized value, X is the original value, and u is the mean of the sample, which reflects the center position of the data. is the standard deviation of the sample, which measures the dispersion of the data points relative to the mean. After standardization, Z = 0 means that the data point is equal to the mean, Z > 0 means it is above the mean, and Z < 0 means it is below the mean. Data with large absolute values ​​of Z-score are considered outliers and are removed.

[0023] In the data preprocessing step, when the K-means algorithm removes noise, the i-th cluster center is the average value of all data points in the cluster.

[0024] Optionally, the intelligent diagnosis method also includes the step of constructing a neural network model: selecting a multilayer perceptron (MLP) as the basic neural network architecture, the number of input layer nodes is 5, corresponding to the operating temperature, ambient humidity, charging current, operating time, and charging time respectively, and the output node is the expected life of the component to construct a prediction model.

[0025] Optionally, the intelligent diagnosis method further includes a model training step:

[0026] The data set is divided into training set, validation set and test set according to the following proportions, where the training set accounts for 70% of the total data set, the validation set accounts for 15% of the total data set, and the test set accounts for 15% of the total data set;

[0027] Send historical data to the input layer of the neural network, obtain the output result through neuron calculation, and save the weighted sum, activation function output and weight value;

[0028] Calculate the output error, using mean square error, and calculate the error signal of each neuron from the output layer by backpropagating the error according to the chain rule;

[0029] Calculate the gradient, calculate the neuron weight gradient and bias gradient based on the error signal and the saved value;

[0030] Update weights, use the gradient descent algorithm to update the neuron connection weights and biases, and repeat the operation until the predetermined error value reaches the standard to complete the model training;

[0031] In the model training step, when calculating the gradient, the weight gradient and bias gradient of each neuron are calculated separately according to the error signal, weighted sum, activation function output and weight value according to specific mathematical calculation rules.

[0032] Optionally, the intelligent diagnosis method further includes a model application and feedback adjustment step: feedback adjustment based on the model prediction results and actual conditions to improve the model accuracy and apply it to actual component health status detection;

[0033] In the model application and feedback adjustment step, the basis for feedback adjustment is the difference between the component health status predicted by the model and the actual detected component health status.

[0034] The present invention also provides an intelligent diagnosis system for implementing any of the above intelligent diagnosis methods, comprising:

[0035] The data collection module is used to perform the data collection steps: collecting various types of data information of the operation of related equipment;

[0036] The data labeling and weighting module is used to perform data labeling and weighting steps: labeling the existing data and the collected data information according to the actual operation of the equipment, and assigning different weights to different labeled data;

[0037] A feature extraction module is used to perform the feature extraction step: using a specific neural network to fuse a multi-scale feature extraction network to extract a multi-scale feature vector from the labeled data set;

[0038] The prediction and evaluation module is used to perform the prediction and evaluation steps: input the extracted feature vector into the prediction model, and evaluate the operating state of the system according to the state prediction result and the score output by the prediction model.

[0039] Optionally, the intelligent diagnosis system further includes:

[0040] The data preprocessing module is used to perform data preprocessing steps, including:

[0041] The K-means algorithm is used to remove noise from the data. The objective function calculation formula of the K-means algorithm is as follows: first, K initial cluster centers are selected, data points are assigned to the cluster centers closest to them to form K clusters, cluster centers are recalculated, and the assignment and calculation of cluster centers are repeated until the cluster centers no longer change or change very little;

[0042] The data were normalized using the Z-score algorithm. The calculation formula is: after normalization, Z = 0 means the data point is equal to the mean, Z > 0 means it is higher than the mean, and Z < 0 means it is lower than the mean. Data with large absolute values ​​of Z-score are considered outliers and are removed.

[0043] A neural network model building module is used to execute the steps of building a neural network model, including: selecting a multi-layer perceptron (MLP) as the basic neural network architecture, the number of input layer nodes is 5, corresponding to the working temperature, ambient humidity, charging current, working time, charging time, and the output node is the expected life of the component;

[0044] The model training module is used to perform model training steps, including:

[0045] The data set is divided into training set, validation set and test set according to the following proportions, where the training set accounts for 70% of the total data set, the validation set accounts for 15% of the total data set, and the test set accounts for 15% of the total data set;

[0046] Send historical data to the input layer of the neural network, obtain the output result through neuron calculation, and save the weighted sum, activation function output and weight value;

[0047] Calculate the output error, using mean square error, and calculate the error signal of each neuron from the output layer by backpropagating the error according to the chain rule;

[0048] Calculate the gradient, calculate the neuron weight gradient and bias gradient based on the error signal and the saved value;

[0049] Update weights, use the gradient descent algorithm to update the neuron connection weights and biases, and repeat the operation until the predetermined error value reaches the standard to complete the model training;

[0050] A model application and feedback adjustment module is used to perform model application and feedback adjustment steps, including: feedback adjustment based on model prediction results and actual conditions to improve model accuracy and apply it to actual component health status detection;

[0051] In the model application and feedback adjustment step, the basis for feedback adjustment is the difference between the component health status predicted by the model and the actual detected component health status.

[0052] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements any of the above intelligent diagnosis methods when executing the computer program.

[0053] The present invention also provides a storage medium containing computer executable instructions, wherein the computer executable instructions are executed by a computer processor to implement any of the above intelligent diagnosis methods.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention provides an intelligent diagnosis method, system, computer device and storage medium, which can effectively predict and diagnose faults through accurate data collection and processing, thereby reducing emergency maintenance and downtime of charging piles, reducing maintenance costs, reducing unexpected failures during user use, and improving customer experience. It does not require additional hardware and is cost-effective.

[0056] The present invention has other features and advantages, which will be apparent from or will be described in detail in the accompanying drawings and the following detailed description incorporated herein, which together serve to explain the specific principles of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 is a flow chart of an intelligent diagnosis method provided by an embodiment of the present invention;

[0059] Figure 2 is another flow chart of an intelligent diagnosis method provided by an embodiment of the present invention;

[0060] Figure 3 is another flow chart of an intelligent diagnosis method provided by an embodiment of the present invention;

[0061] Figure 4 is another flow chart of an intelligent diagnosis method provided by an embodiment of the present invention;

[0062] Figure 5 is a diagnostic flow chart of an intelligent diagnostic method provided by an embodiment of the present invention;

[0063] Figure 6 It is a structural block diagram of an intelligent diagnosis system provided by an embodiment of the present invention.

[0064] Attached figure numbers: 10, data acquisition module; 20, data labeling and weighting module; 30, feature extraction module; 40, prediction and evaluation module; 50, data preprocessing module; 60, neural network model building module; 70, model training module; 80, model application and feedback adjustment module. DETAILED DESCRIPTION

[0065] In order to explain in detail the possible application scenarios, technical principles, specific schemes that can be implemented, and the purposes and effects that can be achieved, the following is a detailed description of the specific embodiments listed in conjunction with the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0066] Reference to "embodiment" herein means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The term "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it particularly limit its independence or association with other embodiments. In principle, in the present application, as long as there is no technical contradiction or conflict, the various technical features mentioned in the embodiments can be combined in any way to form a corresponding implementable technical solution.

[0067] Unless otherwise defined, the technical terms used in this document have the same meanings as those generally understood by those skilled in the art to which this application belongs; the use of relevant terms in this document is only for describing specific embodiments and is not intended to limit this application.

[0068] In the description of this application, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that three relationships may exist, for example, A and / or B, which means: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" logical relationship.

[0069] In the present application, terms such as “first” and “second” are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship of quantity, priority or sequence between these entities or operations.

[0070] Without further limitations, in this application, the words "include", "comprises", "has" or other similar expressions used in the sentences are intended to cover non-exclusive inclusion. These expressions do not exclude the presence of additional elements in the process, method or product including the elements, so that the process, method or product including a series of elements may include not only those limited elements, but also other elements not explicitly listed, or also include elements inherent to such process, method or product.

[0071] Similar to the understanding in the Examination Guidelines, in this application, expressions such as "greater than", "less than", "exceed" and the like are understood to exclude the number itself; expressions such as "above", "below", "within" and the like are understood to include the number itself. In addition, in the description of the embodiments of this application, "multiple" means more than two (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups", "multiple times", etc., unless otherwise clearly and specifically limited.

[0072] In the description of the embodiments of the present application, space-related expressions used, such as "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "vertical", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicate the orientation or position relationship based on the orientation or position relationship shown in the specific embodiments or drawings, and are only for the convenience of describing the specific embodiments of the present application or facilitating the reader's understanding, and do not indicate or imply that the referred device or component must have a specific position, a specific orientation, or be constructed or operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present application.

[0073] Unless otherwise expressly specified or limited, in the description of the embodiments of the present application, the terms such as "install", "connect", "connect", "fix", "set", etc. used should be understood in a broad sense. For example, the "connection" can be a fixed connection, a detachable connection, or an integrated setting; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For technicians in the technical field to which the present application belongs, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0074] Most of the existing technologies on the market lack effective means to predict possible failures of charging piles in advance and accurately diagnose the health status of their internal components.

[0075] The present invention aims to ensure that charging piles can operate stably and efficiently, improve their operating efficiency and safety, meet the charging needs of the majority of electric vehicle users, and provide an intelligent diagnostic solution that can collect and process relevant data in a scientific and reasonable manner, accurately extract feature information, and make an effective assessment of the system operating status based on the corresponding prediction model, so as to discover potential fault hazards in advance and realize intelligent management.

[0076] Please refer to Figure 1 The embodiment of the present invention provides an intelligent diagnosis method. In order to facilitate the explanation of the technical solution of the present invention, a charging pile is taken as an example for introduction below.

[0077] The intelligent diagnosis method provided by the present invention includes:

[0078] S1. Data collection: Collect various types of data information on the operation of related equipment.

[0079] Specifically, the data of the charging pile is collected first. The various data of the equipment during operation are collected through the sensors installed in the charging pile, including working temperature, humidity, charging voltage, current, charging time and number of contactor closures; the log of equipment operation is obtained through the cloud platform, and then the historical data in the log is extracted; the maintenance records of the charging pile are obtained by collecting operation and maintenance personnel to facilitate subsequent data processing.

[0080] By acquiring data through multiple channels, including physical quantity data collected by sensors, operation log data from the cloud platform, and maintenance record data provided by operation and maintenance personnel, it can cover all aspects of charging pile operation, provide a strong basis for subsequent analysis, ensure the comprehensiveness and diversity of data, and help to more accurately understand the operating status of charging piles and discover potential problems.

[0081] S2. Data labeling and weighting: Label the existing data and collected data information based on the actual operation of the equipment, and assign different weights to different labeled data.

[0082] The data labeling and weighting step is to label the existing data and collected data information according to the actual operation of the equipment, and assign different weights to different labeled data.

[0083] For the collected data information, part of the data is randomly sent to the background database and marked. The new data set is given weight a and the old data weight is b. The calculation of a and b is determined based on the number of data updates and batches.

[0084] In this step, the data is labeled and weighted according to the actual operation of the equipment, so that different data can reflect their different values ​​in subsequent analysis. The weights of new and old data are determined according to their update times and batches, which can better reflect the current operation trend of the equipment. Through reasonable data labeling and weighting, the role of important data can be highlighted, the accuracy of diagnosis can be improved, and the analysis results can be more in line with the actual situation.

[0085] S3. Feature extraction: Use a specific neural network to fuse a multi-scale feature extraction network to extract multi-scale feature vectors from the labeled data set.

[0086] The feature extraction step is to use a specific neural network to fuse a multi-scale feature extraction network to extract a multi-scale feature vector from the labeled data set, where the multi-scale feature extraction network is a parallel multi-branch structure, including branches with different convolution kernel sizes or different convolution combinations.

[0087] By using a specific neural network to fuse a multi-scale feature extraction network, this parallel multi-branch structure can mine features in the labeled data set from multiple perspectives, and feature vectors of different scales help capture complex relationships and potential patterns in the data. Multi-scale feature extraction can mine more valuable information, more accurately reflect the status of the internal components of the charging pile, and improve the ability to identify potential factors of failure.

[0088] S4. Prediction evaluation: The extracted feature vector is input into the prediction model, and the operating status of the system is evaluated according to the state prediction result and the score output by the prediction model.

[0089] The prediction evaluation step is to input the extracted feature vector into the prediction model, and evaluate the system's operating status based on the state prediction results and scores output by the prediction model. The prediction model predicts the data value of the next time step based on the input historical data sequence, and the Attention network in the model follows a specific formula to perform state calculation and weighted summation operations in the actual model training prediction task.

[0090] In this step, the extracted feature vector is input into the prediction model, which predicts the data value of the next time step based on the historical data sequence. The Attention network in the model performs state calculation and weighted summation operations through specific formulas, which can focus on key data features and then evaluate the system operation status. By learning and analyzing historical data, the operation status of the charging pile can be predicted in advance, possible faults can be discovered in time, and the normal operation of the charging pile can be guaranteed.

[0091] Please refer to Figure 2 In this embodiment, the intelligent diagnosis method further includes:

[0092] S5. Data preprocessing: Use the K-means algorithm to remove noise from the data and use the Z-score algorithm to normalize the data.

[0093] It is understandable that the data directly obtained generally contains noise and default values ​​and cannot be used directly to train the model. It needs to be preprocessed to remove the noise, default values ​​and outliers in the data, and then the data is normalized. After data preprocessing, the noise and outliers in the original data can be removed, the data quality can be improved, and the accuracy and reliability of subsequent model training can be ensured.

[0094] This embodiment uses the K-means algorithm to remove noise from the data and uses the Z-score algorithm to normalize the data to adapt it to the machine learning model and ensure the accuracy of the data in the model.

[0095] Specifically, the data preprocessing steps include:

[0096] The K-means algorithm is used to remove noise from the data. The objective function calculation formula of the K-means algorithm is:

[0097]

[0098] J is the objective function value, is the ith cluster, is the cluster center of the ith cluster, and x is the data point in the cluster; its goal is to make the data points in the same cluster as similar as possible, and the data points between different clusters as different as possible. When calculating, first select K initial cluster centers, then assign each data point to the cluster center closest to it to form K clusters, and then recalculate the cluster center of each cluster. Usually, the cluster center is the average value of all points in the cluster. Finally, reassign the data of each point to the cluster center closest to it, and then calculate the cluster center until the cluster center does not change or changes very little.

[0099] The Z-score algorithm is used to normalize the data. The calculation formula is:

[0100]

[0101] Z is the standardized value, X is the original value, u is the mean of the sample, which reflects the center position of the data, is the standard deviation of the sample, and measures the dispersion of the data points relative to the mean; after standardization, Z = 0 means that the data point is equal to the mean, Z > 0 is higher than the mean, and Z < 0 is lower than the mean. The data with a large absolute value of Z-score are considered outliers and are removed;

[0102] In the data preprocessing step, when the K-means algorithm removes noise, the i-th cluster center is the average value of all data points in the cluster.

[0103] Furthermore, this embodiment also includes:

[0104] S6. Build a neural network model: Select multilayer perceptron as the basic neural network architecture.

[0105] MLP is a feedforward neural network in which information propagates unidirectionally in the network, from the input layer through multiple hidden layers to the output layer. The number of nodes in the input layer is equal to the number of factors that affect the life of the component. In this embodiment, five factors are considered: operating temperature, ambient humidity, charging current, operating time, and charging time. The output node is the expected life of the component.

[0106] Therefore, the number of input layer nodes in this step is 5, corresponding to the working temperature, ambient humidity, charging current, working time, and charging time, and the output node is the expected life of the components to build a prediction model.

[0107] The step of building the neural network model is to select the multi-layer perceptron (MLP) as the basic neural network architecture. The number of input layer nodes is 5, corresponding to the working temperature, ambient humidity, charging current, working time, and charging time respectively. The output node is the expected life of the component.

[0108] Please refer to Figure 4 Furthermore, this embodiment also includes:

[0109] S7. Model training: Divide the dataset into training set, validation set and test set.

[0110] The training set accounts for seventy percent of the total data set, the validation set accounts for fifteen percent of the total data set, and the test set accounts for fifteen percent of the total data set.

[0111] S8. Send the historical data to the input layer of the neural network and obtain the output result through neuron calculation.

[0112] During the calculation process, the weighted sum, activation function output and weight value of each neuron are saved.

[0113] S9. Calculate the output error until the error signals of all neurons are calculated.

[0114] The output result is compared with the true value. The error calculation method uses the mean square error (MSE). Starting from the output layer, according to the chain rule, the contribution of each neuron to the output error is calculated, and the error signal is passed backward to backpropagate the error until the error signals of all neurons are calculated.

[0115] S10. Calculate the gradient. Calculate the weight gradient and bias gradient of each neuron based on the error signal and the saved weighted sum, activation function output and weight value.

[0116] S11. Update the weights. According to the calculated gradient values, use the gradient descent algorithm to update the connection weights and biases between neurons.

[0117] Then repeat steps S8-S11, and send all samples of the training data set to the neural network for training until the predetermined error value reaches a certain requirement, and the model training is completed.

[0118] S12. Feedback and adjustment are made based on the model’s prediction results and actual conditions to continuously improve the model’s accuracy, and finally the model is applied to actual component health status testing.

[0119] In this step, the basis for feedback adjustment is the difference between the component health status predicted by the model and the actual detected component health status. This feedback adjustment mechanism can enable the model to continuously adapt to changes in the actual operation of the charging pile, maintain the accuracy of component health status diagnosis, and ensure long-term effective fault prediction capabilities.

[0120] Please refer to Figure 5 , the diagnostic process of this embodiment is as follows: first, the health status of the components is detected, and then historical data is collected. These data will be preprocessed, including noise processing using the K-means algorithm and normalization using the Z-score algorithm. Then the neural network is constructed. In this process, the output error, neuron weights and gradients are calculated in turn, and the neural network is constructed. Then the weights are updated to determine whether the error meets the predetermined standard. If the standard is met, the health status of the components can be predicted; if the standard is not met, the weights need to be updated again, and the above operations are repeated until the health status of the components can be accurately predicted.

[0121] Please refer to Figure 6 Based on the above-mentioned embodiments, an intelligent diagnosis system is provided in this embodiment to implement the above-mentioned intelligent diagnosis method.

[0122] Specifically, the data collection module 10 is used to execute data collection steps to collect various types of data information of the operation of related equipment.

[0123] The data labeling and weighting module 20 is used to perform data labeling and weighting steps, label the existing data and the collected data information according to the actual operation of the equipment, and assign different weights to different labeled data.

[0124] The feature extraction module 30 is used to perform the feature extraction step, and extracts a multi-scale feature vector from the labeled data set by fusing a multi-scale feature extraction network with a specific neural network.

[0125] The prediction and evaluation module 40 is used to perform the prediction and evaluation steps, input the extracted feature vector into the prediction model, and evaluate the operating state of the system according to the state prediction result and the score output by the prediction model.

[0126] The data preprocessing module 50 is used to perform data preprocessing steps, including using the K-means algorithm to remove noise in the data and using the Z-score algorithm to normalize the data.

[0127] The neural network model building module 60 is used to execute the steps of building a neural network model, select a multi-layer perceptron (MLP) as a basic neural network architecture, and set input layer and output layer nodes.

[0128] The model training module 70 is used to perform model training steps, including data set division, data input and calculation, error calculation, gradient calculation and weight update operations.

[0129] The model application and feedback adjustment module 80 is used to execute the model application and feedback adjustment steps, and to compare and adjust the model prediction results and the actual situation.

[0130] Based on the foregoing embodiments, a computer device is provided in this embodiment, including a memory and a processor, wherein the memory stores a computer program, and the above-mentioned intelligent diagnosis method is implemented when the processor executes the computer program.

[0131] Based on the foregoing embodiments, a storage medium in this embodiment includes computer executable instructions, and these instructions are executed by a computer processor to implement the above intelligent diagnosis method.

[0132] Based on the above embodiments, the beneficial effects of the present invention are as follows:

[0133] There is no need to add additional hardware equipment. Only by collecting, processing and analyzing existing data, and using software algorithms to achieve intelligent diagnosis, costs are reduced and it has a high cost-effectiveness ratio. By predicting faults in advance and diagnosing the health status of components, emergency maintenance time and downtime can be effectively reduced. Preventive measures before a fault occurs or take measures at the early stage of a fault can avoid losses caused by long-term equipment downtime, thereby reducing maintenance costs. In addition, it can effectively reduce unexpected failures when users use charging piles, ensure the stability of charging services, provide users with a more reliable charging experience, and bring a better customer experience.

[0134] Finally, it should be noted that although the above embodiments have been described in the specification and drawings of this application, this does not limit the scope of patent protection of this application. All technical solutions generated by replacing or modifying equivalent structures or equivalent processes based on the essential concept of this application using the contents recorded in the specification and drawings of this application, as well as directly or indirectly implementing the technical solutions of the above embodiments in other related technical fields, are included in the scope of patent protection of this application.

Claims

1. An intelligent diagnosis method, characterized in that: The following steps are involved: Data collection: Collect various types of data information on the operation of related equipment; Data labeling and weighting: Label the existing data and collected data information according to the actual operation of the equipment, and assign different weights to different labeled data; Feature extraction: Use a specific neural network to fuse a multi-scale feature extraction network to extract multi-scale feature vectors from the labeled data set; Prediction evaluation: The extracted feature vector is input into the prediction model, and the operating status of the system is evaluated based on the state prediction results and scores output by the prediction model.

2. The intelligent diagnosis method according to claim 1, characterized in that: In the data collection step, the multiple types of data information include various physical quantity data collected by sensors installed in the equipment, equipment operation log data obtained from the cloud platform, and equipment maintenance record data obtained from operation and maintenance channels; In the data labeling and weighting step, for the collected data information, part of the data is randomly sent to the backend database and labeled. The new data set is given a weight a, and the old data weight is b. The calculation of a and b is determined based on the number of data updates and batches; In the feature extraction step, the multi-scale feature extraction network is a parallel multi-branch structure, including branches with different convolution kernel sizes or different convolution combinations; In the prediction evaluation step, the prediction model predicts the data value of the next time step based on the input historical data sequence, and the Attention network in the prediction model follows a specific formula to perform state calculation and weighted summation operations in the actual model training prediction task.

3. The intelligent diagnosis method according to claim 1, characterized in that: Also includes data preprocessing: The K-means algorithm is used to remove noise from the data. The objective function calculation formula of the K-means algorithm is: ； J is the objective function value, is the ith cluster, is the cluster center of the i-th cluster, and x is the data point in the cluster; When calculating, first select K initial cluster centers, assign data points to the cluster centers closest to them to form K clusters, recalculate cluster centers, and repeat the assignment and calculation of cluster centers until the cluster centers no longer change or change very little; The Z-score algorithm is used to normalize the data. The calculation formula is: ； Z is the standardized value, X is the original value, and u is the mean of the sample, which reflects the center position of the data. is the standard deviation of the sample, a measure of how spread out the data points are relative to the mean; After standardization, Z = 0 means that the data point is equal to the mean, Z > 0 means it is above the mean, and Z < 0 means it is below the mean. Data with large absolute values ​​of Z-score are considered outliers and are removed. In the data preprocessing step, when the K-means algorithm removes noise, the i-th cluster center is the average value of all data points in the cluster.

4. The intelligent diagnosis method according to claim 1, characterized in that: It also includes building a neural network model: a multilayer perceptron is selected as the basic neural network architecture, the number of input layer nodes is 5, corresponding to the operating temperature, ambient humidity, charging current, operating time, and charging time, and the output node is the expected life of the component to build a prediction model.

5. The intelligent diagnosis method according to claim 1, characterized in that: Also includes model training: The data set is divided into training set, validation set and test set according to the following proportions, where the training set accounts for 70% of the total data set, the validation set accounts for 15% of the total data set, and the test set accounts for 15% of the total data set; Send historical data to the input layer of the neural network, obtain the output result through neuron calculation, and save the weighted sum, activation function output and weight value; Calculate the output error, using mean square error, and calculate the error signal of each neuron from the output layer by backpropagating the error according to the chain rule; Calculate the gradient, calculate the neuron weight gradient and bias gradient based on the error signal and the saved value; Update weights, use the gradient descent algorithm to update the neuron connection weights and biases, and repeat the operation until the predetermined error value reaches the standard to complete the model training; In the model training step, when calculating the gradient, the weight gradient and bias gradient of each neuron are calculated separately according to the error signal, weighted sum, activation function output and weight value according to specific mathematical calculation rules.

6. The intelligent diagnosis method according to claim 1, characterized in that: It also includes model application and feedback adjustment: feedback adjustment based on model prediction results and actual conditions to improve model accuracy and apply it to actual component health status detection; In the model application and feedback adjustment step, the basis for feedback adjustment is the difference between the component health status predicted by the model and the actual detected component health status.

7. An intelligent diagnostic system, characterized in that: A method for implementing an intelligent diagnosis method according to any one of claims 1 to 6, comprising: The data collection module is used to perform the data collection steps: collecting various types of data information of the operation of related equipment; The data labeling and weighting module is used to perform data labeling and weighting steps: labeling the existing data and the collected data information according to the actual operation of the equipment, and assigning different weights to different labeled data; A feature extraction module is used to perform the feature extraction step: using a specific neural network to fuse a multi-scale feature extraction network to extract a multi-scale feature vector from the labeled data set; The prediction and evaluation module is used to perform the prediction and evaluation steps: input the extracted feature vector into the prediction model, and evaluate the operating state of the system according to the state prediction result and the score output by the prediction model.

8. The intelligent diagnosis system according to claim 7, characterized in that: Also includes: The data preprocessing module is used to perform data preprocessing steps, including: The K-means algorithm is used to remove noise from the data. The objective function calculation formula of the K-means algorithm is as follows: first, K initial cluster centers are selected, data points are assigned to the cluster centers closest to them to form K clusters, cluster centers are recalculated, and the assignment and calculation of cluster centers are repeated until the cluster centers no longer change or change very little; The data were normalized using the Z-score algorithm. The calculation formula is: after normalization, Z = 0 means the data point is equal to the mean, Z > 0 means it is higher than the mean, and Z < 0 means it is lower than the mean. Data with large absolute values ​​of Z-score are considered outliers and are removed. A neural network model building module is used to execute the steps of building a neural network model, including: selecting a multi-layer perceptron as a basic neural network architecture, the number of input layer nodes is 5, corresponding to operating temperature, ambient humidity, charging current, operating time, charging time, and the output node is the expected life of the component; The model training module is used to perform model training steps, including: The data set is divided into training set, validation set and test set, where the training set accounts for 70% of the total data set, the validation set accounts for 15% of the total data set, and the test set accounts for 15% of the total data set; Send historical data to the input layer of the neural network, obtain the output result through neuron calculation, and save the weighted sum, activation function output and weight value; Calculate the output error, using mean square error, and calculate the error signal of each neuron from the output layer by backpropagating the error according to the chain rule; Calculate the gradient, calculate the neuron weight gradient and bias gradient based on the error signal and the saved value; Update weights, use the gradient descent algorithm to update the neuron connection weights and biases, and repeat the operation until the predetermined error value reaches the standard to complete the model training; A model application and feedback adjustment module is used to perform model application and feedback adjustment steps, including: feedback adjustment based on model prediction results and actual conditions to improve model accuracy and apply it to actual component health status detection; In the model application and feedback adjustment step, the basis for feedback adjustment is the difference between the component health status predicted by the model and the actual detected component health status.

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 intelligent diagnosis method according to any one of claims 1 to 6 is implemented.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions are executed by a computer processor to implement the intelligent diagnosis method according to any one of claims 1 to 6.

Images