A CLCC life prediction and failure risk analysis method, system, device and medium

Abstract

The present invention relates to a CLCC life prediction and fault risk analysis method, system, equipment and medium, wherein the method includes: constructing a CLCC life prediction and fault risk analysis model; constructing a CLCC simulation system, simulating the normal working condition and fault working condition of the CLCC to repeat multiple life test experiments, and obtaining a training data set; training the CLCC life prediction and fault risk analysis model; using the trained CLCC life prediction and fault risk analysis model to perform life prediction values ​​and fault risk analysis; the system is used to implement the above method, including a data acquisition and processing module, a calculation module and a life prediction and fault risk analysis module. Compared with the prior art, the present invention visualizes the decision-making process of the model, intuitively displays the process of life prediction and fault risk analysis, and helps identify the key fault types that affect the life of CLCC components, and the method adopts an improved neural network model so that it also has good processing performance in the face of complex working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of power system equipment monitoring, and in particular to a CLCC life prediction and failure risk analysis method, system, equipment and medium. Background Art

[0002] Controllable line commutated converter (CLCC) plays an important role in power conversion and transmission in modern power systems. With long-term operation, key components in CLCC, such as thyristors, insulators, and capacitors, will be affected by voltage fluctuations, temperature changes, and the external environment, and will gradually age or fail. Once these components fail, the entire system may be shut down, or even cause more serious economic losses and safety hazards. Therefore, life prediction and failure risk analysis of CLCC are key links to ensure its long-term stable operation, optimize maintenance strategies, and reduce operating costs.

[0003] Existing methods for component life prediction and fault diagnosis mainly rely on empirical formulas, statistical analysis or physical modeling. These methods cannot fully utilize the massive sensor data generated during the operation of modern equipment, and have limited effectiveness in processing multidimensional time series data under complex working conditions. At the same time, with the increasing complexity of equipment, traditional methods have significant deficiencies in dealing with multimodal data fusion and long-term dependency modeling, making it difficult to meet actual industrial needs.

[0004] Therefore, it is a problem that needs to be solved to provide an accurate method for CLCC life prediction and failure risk analysis that can fully utilize various types of data during equipment operation and is applicable to actual complex working conditions. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a CLCC life prediction and failure risk analysis method, system, equipment and medium.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] According to a first aspect of the present invention, a CLCC life prediction and failure risk analysis method is provided, the method comprising:

[0008] Constructing a CLCC life prediction and failure risk analysis model, the model includes a neural network and a visualization module; the visualization module visualizes the decision-making process of the CLCC life prediction and failure risk analysis model according to the feature extraction and output results of the neural network;

[0009] Construct a CLCC simulation system to simulate the normal working condition and faulty working condition of CLCC and repeat the life test for many times to obtain the sensor data of each component and the expected life value during the test; the sensor data includes voltage data, current data and temperature data;

[0010] Preprocessing the sensor data and the expected life data to obtain a preprocessed data set, and training a CLCC life prediction and failure risk analysis model based on the preprocessed data set;

[0011] Actual sensor data is collected, processed, and input into a trained CLCC life prediction and failure risk analysis model, the life prediction value and failure probability value of each component in the CLCC are output, and a visualization view is generated; the visualization view is used to visualize the decision importance of the actual sensor data.

[0012] As a preferred technical solution, the neural network is a U-type CNN-ResNet neural network, including an input layer, an encoder, a decoder and an output layer; the encoder includes multiple convolutional layers, multiple residual layers and multiple pooling layers; the decoder includes multiple anti-pooling layers.

[0013] As a preferred technical solution, the step of outputting the life prediction value and failure probability value of each component in the CLCC includes:

[0014] Downward collection: The input layer receives the processed actual sensor data and transmits it to the encoder, and is preliminarily processed by multiple convolutional layers; the result of the preliminary processing is processed by the pooling layer and input into the residual-pooling structure for downward collection. The residual-pooling structure is a multi-layer structure, and each layer includes a residual layer and a pooling layer; each downward collection will output the multi-scale features of the sensor, and the multi-head attention mechanism is used to reprocess the multi-scale features and output the first sensor features;

[0015] Skip connection: The first sensor feature output by the encoder is passed to the decoder through the skip connection;

[0016] Up-collection: The decoder receives the first sensor feature and the first sensor feature through the jump connection, processes them through the unpooling layer and performs up-sampling operation, and outputs the second sensor feature;

[0017] Output life prediction value and fault probability value: the output layer receives the second sensor feature output from the encoder, and inputs the second sensor feature into the life prediction branch and the fault risk analysis branch respectively; the life prediction branch receives the second sensor feature, and the fully connected layer of the life prediction branch outputs a real value, and the real value is processed using a linear activation function to output the life prediction value; the fault risk analysis branch receives the second sensor feature, and the fully connected layer of the fault risk analysis branch outputs a probability value, and the sigmoid activation function is used to process the probability value to output the fault probability value.

[0018] As a preferred technical solution, the generated visualization view is generated by the visualization module using the Grad-CAM method, specifically including:

[0019] Calculate the gradients of the last convolutional layer and the output layer;

[0020] Calculate the weight of each feature corresponding to the sensor data based on the gradient;

[0021] Calculate the Grad-CAM thermal value of each feature corresponding to the sensor data based on the weight;

[0022] The Grad-CAM thermal value is processed using an activation function, and a Grad-CAM thermal map, i.e., a visualization view, is drawn based on the processed Grad-CAM thermal value.

[0023] As a preferred technical solution, the pretreatment includes:

[0024] The sensor data is subjected to outlier removal, standardization, and sliding window segmentation to obtain first data;

[0025] Marking the first data, wherein the marking includes component number, time step, sensor data, remaining life, fault warning, and health status;

[0026] A preprocessed data set is constructed based on the labeled first data.

[0027] As a preferred technical solution, the method for obtaining the remaining life is to count the actual working hours of the corresponding component, and calculate the difference between the expected life value and the actual working hours, and the difference is the remaining life of the corresponding component;

[0028] The fault warning includes a warning mark and fault information, the warning mark is Boolean data, and the warning mark is set to 1 when the sensor data is abnormal;

[0029] The method for obtaining the health status is: evaluate the deviation between the abnormal sensor data and the normal sensor data and the duration of the abnormal sensor data, and form the health status of the corresponding component based on the evaluation results. The health status is a value of [0,1] and decreases as the component is used.

[0030] As a preferred technical solution, the training process of the CLCC life prediction and failure risk analysis model includes the following steps:

[0031] Dividing the preprocessed data set into a training set, a validation set and a test set according to preset rules;

[0032] The training set is used to train the CLCC life prediction and failure risk analysis model, and the mean square error is used as the loss function for the life prediction branch in the model, and the binary cross entropy is used as the loss function for the failure risk analysis branch in the model;

[0033] Use the validation set to evaluate the performance of the CLCC life prediction and failure risk analysis model, and optimize the model based on the performance evaluation results;

[0034] The performance of the optimized CLCC life prediction and failure risk analysis model was tested using the test set, and the model with the best performance was selected as the final CLCC life prediction and failure risk analysis model.

[0035] According to a second aspect of the present invention, a CLCC life prediction and failure risk analysis system is provided, which is used to implement the above-mentioned CLCC life prediction and failure risk analysis method, including:

[0036] Data acquisition and processing module: using various sensors in CLCC to collect actual sensor data and process the actual sensor data;

[0037] Calculation module: Use the CLCC life prediction and failure risk analysis model to output life prediction value, failure probability value and visualization view based on the processed actual sensor data;

[0038] Life prediction and failure risk analysis module: Based on the life prediction value and failure probability value, the life prediction and failure risk analysis of each component in the CLCC are realized, and the information of high failure risk components or abnormal behavior of the CLCC is obtained according to the visual view, and the information includes component number and failure type.

[0039] According to a third aspect of the present invention, there is provided an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and the method described above is implemented when the processor executes the program.

[0040] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method described when executed by a processor.

[0041] Compared with the prior art, the present invention combines the UNet structure, residual neural network (ResNet), attention mechanism and visualization technology, generates a visualization view through visualization technology, intuitively displays the key factors of the CLCC life prediction and fault risk analysis model in the process of life prediction and fault risk analysis, and helps to identify the key fault types that affect the life of CLCC components; through multi-scale feature extraction and adaptive feature weighting, the remaining life and failure risk of each CLCC component are accurately predicted, so that it has good processing performance even in the face of complex working conditions, enhancing its interpretability and practicality in actual industrial applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a diagram of the architecture of the neural network of the present invention;

[0043] Figure 2 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0045] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0046] Example 1

[0047] Deep learning technology has gradually shown great potential in the field of equipment health monitoring and predictive maintenance. Deep learning models such as convolutional neural networks (CNN) and recurrent neural networks (RNN) can automatically extract effective features from multi-dimensional sensor data and adapt to long-time series data patterns. However, a single network structure is often difficult to capture multi-scale features or accurately identify complex fault modes. However, the UNet structure can extract multi-level features at different time scales through the encoder-decoder network, and retain key detail information through jump connections to prevent feature loss; the residual network (ResNet) effectively alleviates the gradient vanishing problem in deep neural networks through residual connections and enhances the expressiveness of features; the attention mechanism helps the model to adaptively adjust the focus in time steps and feature dimensions, so that the model can focus on the key fault modes in the equipment, thereby improving the prediction accuracy and reliability of the model. Therefore, in this embodiment, a multi-structure neural network combining the UNet architecture, residual neural network and attention mechanism is proposed. In combination with the above content, this embodiment proposes a CLCC life prediction and failure risk analysis method.

[0048] The flowchart of this method is as follows Figure 2 As shown, specifically including:

[0049] S1. Build a CLCC life prediction and failure risk analysis model, including neural network and visualization modules.

[0050] S11. Constructing a neural network:

[0051] The neural network is a U-type CNN-ResNet neural network, and its detailed structure is as follows Figure 1 As shown, it includes an input layer, an encoder, a decoder and an output layer; the encoder includes multiple convolutional layers, multiple residual layers and multiple pooling layers, and the decoder includes multiple anti-pooling layers;

[0052] Each layer of the encoder is connected to the decoder in a residual manner to avoid the gradient vanishing problem and ensure the feature transfer effect of the deep network; all convolutional layers use the selected convolution kernel to extract the interaction characteristics between sensor data; the residual block is composed of jump connections between the input and the convolutional layer, and the input of each layer is directly added to the output of the convolutional layer to ensure the fluidity of the features in the network and prevent information loss in the deep network.

[0053] S12. Construct a visualization module: Construct a visualization module based on Grad-CAM technology. The visualization module visualizes the decision-making process of the CLCC life prediction and failure risk analysis model according to the feature extraction and output results of the neural network.

[0054] S2. Construct a CLCC simulation system to simulate the normal working conditions and fault conditions of CLCC and repeat multiple life test experiments. Obtain the sensor data of each component and the expected life value during the test. The sensor data includes voltage data, current data and temperature data. The sensor data of each component is recorded at fixed time intervals to form multi-dimensional time series data. For each time period of each component, establish a corresponding relationship between life and operating status. When a component fails, record its final working time and performance, and mark the end point of the component's life. For components that have not yet completely failed, estimate the time when they will fail based on the data change trend.

[0055] S3. Train the CLCC life prediction and failure risk analysis model.

[0056] S31. Preprocess the sensor data and the expected life data.

[0057] S311, removing outliers, standardizing and performing sliding window segmentation on the sensor data to obtain first data.

[0058] S312, marking the first data, the marking includes component number, time step, sensor data, remaining life, fault warning and health status. Taking component number 1 as an example, its marking information is shown in Table 1:

[0059] Table 1 Component 1 marking information table

[0060]

[0061] Among them, the method for obtaining "remaining life" is to count the actual working hours of the corresponding components, and calculate the difference between the expected life value and the actual working hours. The difference is the remaining life of the corresponding component; the "warning flag" is Boolean data, and the warning flag is set to 1 when the sensor data is abnormal; the method for obtaining "health status" is: evaluate the deviation between the abnormal sensor data and the normal sensor data and the duration of the abnormal sensor data, and form the health status of the corresponding component based on the evaluation results. The health status is a value of [0,1] and decreases as the component is used; "fault information" includes component circuit failure a, part function damage or inability to work b, and communication failure c.

[0062] S313: construct a preprocessed data set based on the labeled first data.

[0063] S32. Training model.

[0064] S321, dividing the preprocessed data set into a training set, a validation set and a test set according to preset rules.

[0065] S322, using the training set to train the CLCC life prediction and failure risk analysis model, and using the mean square error as the loss function for the life prediction branch in the model, and using the binary cross entropy as the loss function for the failure risk analysis branch in the model, the implementation code is:

[0066] model.compile(optimizer='adam',

[0067] loss={'lifetime_output': 'mean_squared_error', 'fault_prob_output': 'binary_crossentropy'},

[0068] metrics={'lifetime_output': 'mae', 'fault_prob_output': 'accuracy'}).

[0069] S323. Use the validation set to evaluate the performance of the CLCC life prediction and failure risk analysis model, and optimize the model based on the performance evaluation results.

[0070] S324. Use the test set to perform performance testing on the optimized CLCC life prediction and failure risk analysis model, and select the model with the best performance as the final CLCC life prediction and failure risk analysis model.

[0071] S4. Perform life prediction and failure risk analysis.

[0072] S41. Output the life prediction value and failure probability value of each component in CLCC.

[0073] S411, down collection: the input layer receives the processed actual sensor data and transmits it to the encoder, and the multi-layer convolution layer performs preliminary processing; the result of the preliminary processing is processed by the pooling layer and input into the residual-pooling structure for down collection, and the residual-pooling structure is a multi-layer structure and each layer includes a residual layer and a pooling layer; each down collection will output the multi-scale features of the sensor, and the multi-head attention mechanism is used to reprocess the multi-scale features and output the first sensor features; wherein, the implementation code of the data processed by the convolution layer, the pooling layer and the residual layer is as follows:

[0074] conv1=Conv1D(filters=64,kernel_size=3,activation='relu',padding='same')(input_data)

[0075] conv1= Conv1D(filters=64, kernel_size=3, activation='relu', padding='same')(conv1)

[0076] residual1 = Add()([input_data, conv1])

[0077] pool1 = MaxPooling1D(pool_size=2)(residual1);

[0078] The code for the processing of the first sensor feature is:

[0079] attn_output,attn_weights = MultiHeadAttention(num_heads=8, key_dim=64)(pool1, pool1)

[0080] attn_output = LayerNormalization()(attn_output).

[0081] S412, skip connection: passing the first sensor feature output by the encoder to the decoder through the skip connection.

[0082] S413, up-collection: The decoder receives the first sensor feature and the first sensor feature through the jump connection, processes through the depooling layer and performs up-sampling operation, and outputs the second sensor feature. The implementation code is:

[0083] up1 = UpSampling1D(size=2)(attn_output) reverse pooling

[0084] concat1 = Concatenate()([up1, residual1]) skip connection

[0085] conv3=Conv1D(filters=64,kernel_size=3,activation='relu',padding='same')(concat1).

[0086] S414, output life prediction value: the output layer receives the second sensor feature output from the encoder, and inputs the second sensor feature into the life prediction branch; the life prediction branch receives the second sensor feature, and the fully connected layer of the life prediction branch outputs a real value, and the real value is processed using a linear activation function to output the life prediction value.

[0087] S415, outputting a fault probability value: the output layer receives the second sensor feature outputted from the encoder, and also inputs the second sensor feature into the fault risk analysis branch; the fault risk analysis branch receives the second sensor feature, and the fully connected layer of the fault risk analysis branch outputs a probability value, and the sigmoid activation function is used to perform the probability value processing to output the fault probability value.

[0088] The implementation codes of step S414 and step S415 are:

[0089] lifetime_output = Dense(1, activation='linear', name='lifetime_output')(conv3)

[0090] fault_prob_output=Dense(1, activation='sigmoid', name='fault_prob_output')(conv3)

[0091] model = Model(inputs=input_data, outputs=[lifetime_output, fault_prob_output]);

[0092] S42, generate a visualization view. This step is performed using the Grad-CAM method, and the code for its implementation is:

[0093] with tf.GradientTape() as tape:

[0094] conv_outputs, predictions = grad_model(input_data)

[0095] loss = predictions[0][0] is assumed to be the failure probability output

[0096] gradients = tape.gradient(loss, conv_outputs)[0]

[0097] weights = tf.reduce_mean(gradients, axis=(0, 1))

[0098] grad_cam = np.zeros(conv_outputs.shape[0:2], dtype=np.float32)

[0099] for i, w in enumerate(weights):

[0100] grad_cam += w * conv_outputs[0, :, i]

[0101] grad_cam = np.maximum(grad_cam, 0)

[0102] grad_cam = grad_cam / np.max(grad_cam);

[0103] The following steps are involved:

[0104] S421. Calculate the gradients of the last convolutional layer and the output layer.

[0105] S422. Calculate the weight of each feature corresponding to the sensor data based on the gradient.

[0106] S423, calculating the Grad-CAM thermal value of each feature corresponding to the sensor data based on the weight.

[0107] S424. Use the activation function to process the Grad-CAM thermal value, and draw a Grad-CAM thermal map, i.e., a visualization view, according to the processed Grad-CAM thermal value.

[0108] Example 2

[0109] The above is an introduction to the method embodiment. The following is a further explanation of the solution of the present invention through a system embodiment.

[0110] In order to implement the method provided in the above embodiment, this embodiment proposes a CLCC life prediction and failure risk analysis system, including:

[0111] Data acquisition and processing module: uses various sensors in CLCC to collect actual sensor data and processes the actual sensor data;

[0112] Calculation module: Use the CLCC life prediction and failure risk analysis model to output life prediction value, failure probability value and visualization view based on the processed actual sensor data;

[0113] Life prediction and failure risk analysis module: Based on the life prediction value and failure probability value, the life prediction and failure risk analysis of each component in the CLCC are realized, and the information of high failure risk components or abnormal behavior of CLCC is obtained according to the visual view, which includes component number and failure type.

[0114] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0115] Example 3

[0116] The electronic device of the present invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0117] Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0118] The processing unit performs the various methods and processes described above, such as methods S1 to S4. For example, in some embodiments, methods S1 to S4 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S1 to S4 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S4 in any other appropriate manner (e.g., by means of firmware).

[0119] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0120] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0121] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0122] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A CLCC life prediction and failure risk analysis method, characterized in that: The method includes: A CLCC life prediction and failure risk analysis model is constructed, the model comprising a neural network and a visualization module; the visualization module visualizes the decision-making process of the CLCC life prediction and failure risk analysis model according to the feature extraction and output results of the neural network; the neural network is a U-type CNN-ResNet neural network, comprising an input layer, an encoder, a decoder and an output layer; the encoder comprises multiple convolutional layers, multiple residual layers and multiple pooling layers; the decoder comprises multiple anti-pooling layers; Construct a CLCC simulation system, simulate the normal working condition and fault working condition of CLCC and repeat the life test for many times, obtain the sensor data of each component and the expected life value during the test; the sensor data includes voltage data, current data and temperature data; The sensor data and expected life data are preprocessed to obtain a preprocessed data set, and the CLCC life prediction and fault risk analysis model is trained based on the preprocessed data set; the preprocessing includes: removing outliers, standardizing and sliding window segmenting the sensor data to obtain first data; marking the first data, and the marking includes component number, time step, sensor data, remaining life, fault warning and health status; constructing a preprocessed data set based on the marked first data; wherein, the method for obtaining the remaining life is to count the actual working hours of the corresponding component, and calculate the difference between the expected life value and the actual working hours, and the difference is the remaining life of the corresponding component; the fault warning includes a warning mark and fault information, and the warning mark is Boolean data, and when the sensor data is abnormal, the warning mark is set to 1; the method for obtaining the health status is: evaluating the deviation between the abnormal sensor data and the normal sensor data and the duration of the abnormal sensor data, and forming the health status of the corresponding component based on the evaluation result, and the health status is a value of [0,1] and decreases with the use time of the component; Collecting actual sensor data, processing the actual sensor data and inputting it into the trained CLCC life prediction and failure risk analysis model, outputting the life prediction value and failure probability value of each component in the CLCC and generating a visualization view; the visualization view is used to visualize the decision importance of the actual sensor data; The step of outputting the life prediction value and failure probability value of each component in the CLCC includes: Downward collection: The input layer receives the processed actual sensor data and transmits it to the encoder, and is preliminarily processed by multiple convolutional layers; the result of the preliminary processing is processed by the pooling layer and input into the residual-pooling structure for downward collection. The residual-pooling structure is a multi-layer structure, and each layer includes a residual layer and a pooling layer; each downward collection will output the multi-scale features of the sensor, and the multi-head attention mechanism is used to reprocess the multi-scale features and output the first sensor features; Skip connection: The first sensor feature output by the encoder is passed to the decoder through the skip connection; Up-collection: The decoder receives the first sensor feature and the first sensor feature through the jump connection, processes them through the unpooling layer and performs up-sampling operation, and outputs the second sensor feature; Output life prediction value and fault probability value: the output layer receives the second sensor feature output from the encoder, and inputs the second sensor feature into the life prediction branch and the fault risk analysis branch respectively; the life prediction branch receives the second sensor feature, and the fully connected layer of the life prediction branch outputs a real value, and the real value is processed using a linear activation function to output the life prediction value; the fault risk analysis branch receives the second sensor feature, and the fully connected layer of the fault risk analysis branch outputs a probability value, and the sigmoid activation function is used to process the probability value to output the fault probability value.

2. A CLCC life prediction and failure risk analysis method according to claim 1, characterized in that: The generated visualization view is generated by the visualization module using the Grad-CAM method, specifically including: Calculate the gradients of the last convolutional layer and the output layer; Calculate the weight of each feature corresponding to the sensor data based on the gradient; Calculate the Grad-CAM thermal value of each feature corresponding to the sensor data based on the weight; The Grad-CAM thermal value is processed using an activation function, and a Grad-CAM thermal map, i.e., a visualization view, is drawn based on the processed Grad-CAM thermal value.

3. A CLCC life prediction and failure risk analysis method according to claim 1, characterized in that: The training process of the CLCC life prediction and failure risk analysis model includes the following steps: Dividing the preprocessed data set into a training set, a validation set and a test set according to preset rules; The training set is used to train the CLCC life prediction and failure risk analysis model, and the mean square error is used as the loss function for the life prediction branch in the model, and the binary cross entropy is used as the loss function for the failure risk analysis branch in the model; Use the validation set to evaluate the performance of the CLCC life prediction and failure risk analysis model, and optimize the model based on the performance evaluation results; The performance of the optimized CLCC life prediction and failure risk analysis model was tested using the test set, and the model with the best performance was selected as the final CLCC life prediction and failure risk analysis model.

4. A CLCC life prediction and failure risk analysis system, characterized in that: The system is used to implement the CLCC life prediction and failure risk analysis method as described in any one of claims 1 to 3, comprising: Data acquisition and processing module: using various sensors in CLCC to collect actual sensor data and process the actual sensor data; Calculation module: Use the CLCC life prediction and failure risk analysis model to output life prediction value, failure probability value and visualization view based on the processed actual sensor data; Life prediction and failure risk analysis module: Based on the life prediction value and failure probability value, the life prediction and failure risk analysis of each component in the CLCC are realized, and the information of high failure risk components or abnormal behavior of the CLCC is obtained according to the visual view, and the information includes component number and failure type.

5. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images