Cable residual life evaluation method and device based on health factors, terminal equipment and storage medium

Abstract

The invention discloses a cable residual life assessment method and device based on a health factor, terminal equipment and a storage medium. The method comprises the following steps: acquiring the health factor of a to-be-assessed cable; inputting each health factor into a preset time convolutional neural network model, so that the time convolutional neural network model determines the residual life of the to-be-evaluated cable according to each health factor; the training of the time convolutional neural network model comprises the following steps: acquiring data items of a cable sample in a purchasing stage, a construction stage, an operation stage and a maintenance stage; performing normalization processing and principal component analysis on each data item, and extracting a health factor for representing the health state of the cable sample; and taking the health factor and the actual residual life of each cable sample as input, taking the predicted residual life of each cable sample as output, and carrying out iterative training on the time convolutional neural network model until the time convolutional neural network model converges. According to the invention, the residual life of the cable can be evaluated.

Description

Technical Field

[0001] The present invention relates to the field of power equipment status assessment, and in particular to a cable remaining life assessment method, device, terminal equipment and storage medium based on health factors. Background Art

[0002] As an indispensable key medium for power transmission, the reliability and safety of cable operation are directly related to the stable operation and efficiency of the entire power system. Cables bear the key task of transmitting electric energy in the power system. Once a cable fails, it will directly affect the normal supply of electric energy and have a serious impact on production and life. As the cable ages, its performance will gradually decline. This is mainly because the cable will be affected by various factors such as electricity, heat, mechanical force, and environment during long-term use, leading to problems such as aging of the insulation layer, increased conductor resistance, and reduced mechanical strength. These problems will further affect the transmission efficiency and safety of the cable, and even cause failures. Cable failures will cause power supply interruptions, causing huge losses to production and life. Therefore, it is very important to be able to predict the remaining life of the cable, so that maintenance can be planned in advance, aging can be delayed, reliability can be improved, and unexpected power outages can be avoided. How to predict the remaining life of the cable is an urgent problem to be solved. Summary of the invention

[0003] The embodiments of the present invention provide a method, an apparatus, a terminal device and a storage medium for evaluating the remaining life of a cable based on a health factor, which can evaluate the remaining life of the cable.

[0004] An embodiment of the present invention provides a method for evaluating the remaining life of a cable based on a health factor, comprising: obtaining a health factor of a cable to be evaluated;

[0005] Inputting each of the health factors into a preset time convolutional neural network model, so that the time convolutional neural network model determines the remaining life of the cable to be evaluated according to each of the health factors;

[0006] The training of the temporal convolutional neural network model includes:

[0007] Obtain various data items of cable samples during the procurement stage, construction stage, operation stage and maintenance stage;

[0008] Each data item is normalized and subjected to principal component analysis to extract health factors used to characterize the health status of cable samples;

[0009] Taking the health factor and the actual remaining life of each cable sample as input and the predicted remaining life of each cable sample as output, the temporal convolutional neural network model is iteratively trained until the temporal convolutional neural network model converges.

[0010] Furthermore, various data items of the cable samples during the procurement stage, construction stage, operation stage and maintenance stage are obtained, including:

[0011] Obtain the cable nameplate parameters, factory test report and arrival sampling test report of the cable sample, and obtain various data items of the cable sample in the procurement stage;

[0012] Obtain the transportation and installation records and handover acceptance report of the cable samples, and obtain various data items of the cable samples during the construction phase;

[0013] Obtain the cable monitoring data of the cable sample during operation, the environmental data of the operation site, and the inspection records during operation, and obtain various data items of the cable sample during the operation stage;

[0014] Obtain the maintenance items and parts replacement status of the cable sample, and obtain various data items of the cable sample during the maintenance stage.

[0015] Furthermore, the normalization of each data item includes:

[0016] Classify each data item according to whether it is numerical data to obtain numerical data and non-numerical data;

[0017] Each non-numeric data is normalized by one-hot encoding;

[0018] For each numerical data, secondary classification is performed according to the variable relationship between the numerical value and the cable health status, and positively correlated numerical data, negatively correlated numerical data and inverted U-shaped numerical data are obtained; wherein the positively correlated numerical data is numerical data that is positively correlated with the cable health status, the negatively correlated numerical data is numerical data that is negatively correlated with the cable health status, and the inverted U-shaped numerical data is numerical data that is in an inverted U-shaped relationship with the cable health status;

[0019] The positive correlation numerical data are normalized by the following formula:

[0020]

[0021] Negatively correlated numerical data are normalized using the following formula:

[0022]

[0023] The inverted U-shaped numerical data are normalized using the following formula:

[0024]

[0025] Where x is the actual value of the numeric data; x min It is the preset minimum value under normal conditions; xmax is the preset maximum value under normal conditions; m and n are the preset boundary parameters.

[0026] Furthermore, the temporal convolutional neural network model includes: several TCN layers, attention mechanism layers, pooling layers and fully connected layers; each TCN layer includes: an expanded causal convolution layer, a weight normalization layer, an activation function layer and a random inactivation layer;

[0027] The dilated causal convolution layer is used to perform convolution processing on the input data, and then transmit the convolution processed data to the weight normalization layer; wherein, when the dilated causal convolution layer is a dilated causal convolution layer in the first-level TCN layer, the input data is the health factor of each cable sample; when the dilated causal convolution layer is not a dilated causal convolution layer in the first-level TCN layer, the input data is the output of the random inactivation layer in the previous-level TCN layer;

[0028] The weight normalization layer is used to perform weight normalization on the data after convolution processing to obtain weight normalized data, and transmit the weight normalized data to the activation function layer;

[0029] The activation function layer is used to perform nonlinear transformation on the weight normalized data to obtain nonlinear transformed data, and input the nonlinear transformed data into the random dropout layer;

[0030] The random deactivation layer is used to perform a random node shielding operation on the linearly transformed data, and input the data of the node shielding operation to the attention mechanism layer, or the dilated causal convolution layer in the TCN layer of the next level; wherein, when the random deactivation layer is the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the attention mechanism layer, and when the random deactivation layer is not the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the dilated causal convolution layer in the TCN layer of the next level;

[0031] The attention mechanism layer is used to perform attention weighting processing according to the output of the random inactivation layer of the last level and the health factor of each cable sample, and output the sequence after attention processing to the pooling layer;

[0032] The pooling layer is used to perform pooling processing on the sequence after attention processing, and transmits the pooled data to the fully connected layer, so that the fully connected layer outputs the predicted remaining life of the cable sample.

[0033] Furthermore, after the temporal convolutional neural network model is trained, it also includes:

[0034] Calculate the relative error and the root mean square error based on the actual remaining life of the cable sample and the corresponding predicted remaining life;

[0035] When it is determined that the relative error or the root mean square error is not within the corresponding preset range, the structure or parameters of the temporal convolutional neural network model are adjusted.

[0036] Based on the above method embodiment, the present invention provides a corresponding device embodiment;

[0037] An embodiment of the present invention provides a cable remaining life assessment device based on health factors, a data acquisition module, a remaining life prediction module and a model training module;

[0038] The data acquisition module is used to obtain the health factor of the cable to be evaluated;

[0039] The remaining life prediction module is used to input each of the health factors into a preset time convolutional neural network model, so that the time convolutional neural network model determines the remaining life of the cable to be evaluated according to each of the health factors;

[0040] The model training module is used to obtain various data items of cable samples in the procurement stage, construction stage, operation stage and maintenance stage; normalize each data item and perform principal component analysis to extract health factors used to characterize the health status of the cable samples; use the health factors and actual remaining life of each cable sample as input and the predicted remaining life of each cable sample as output to iteratively train the time convolutional neural network model until the time convolutional neural network model converges.

[0041] Furthermore, the model training module normalizes each data item, including:

[0042] Classify each data item according to whether it is numerical data to obtain numerical data and non-numerical data;

[0043] Each non-numeric data is normalized by one-hot encoding;

[0044] For each numerical data, secondary classification is performed according to the variable relationship between the numerical value and the cable health status, and positively correlated numerical data, negatively correlated numerical data and inverted U-shaped numerical data are obtained; wherein the positively correlated numerical data is numerical data that is positively correlated with the cable health status, the negatively correlated numerical data is numerical data that is negatively correlated with the cable health status, and the inverted U-shaped numerical data is numerical data that is in an inverted U-shaped relationship with the cable health status;

[0045] The positive correlation numerical data are normalized by the following formula:

[0046]

[0047] Negatively correlated numerical data are normalized using the following formula:

[0048]

[0049] The inverted U-shaped numerical data are normalized using the following formula:

[0050]

[0051] Where x is the actual value of the numeric data; x min It is the preset minimum value under normal conditions; x max is the preset maximum value under normal conditions; m and n are the preset boundary parameters.

[0052] Furthermore, the temporal convolutional neural network model includes: several TCN layers, attention mechanism layers, pooling layers and fully connected layers; each TCN layer includes: an expanded causal convolution layer, a weight normalization layer, an activation function layer and a random dropout layer;

[0053] The dilated causal convolution layer is used to perform convolution processing on the input data, and then transmit the convolution processed data to the weight normalization layer; wherein, when the dilated causal convolution layer is a dilated causal convolution layer in the first-level TCN layer, the input data is the health factor of each cable sample; when the dilated causal convolution layer is not a dilated causal convolution layer in the first-level TCN layer, the input data is the output of the random inactivation layer in the previous-level TCN layer;

[0054] The weight normalization layer is used to perform weight normalization on the data after convolution processing to obtain the weight normalized data, and transmit the weight normalized data to the activation function layer;

[0055] The activation function layer is used to perform nonlinear transformation on the weight normalized data to obtain nonlinear transformed data. The nonlinear transformed data is input into the random dropout layer;

[0056] The random deactivation layer is used to perform a random node shielding operation on the linearly transformed data, and input the data of the node shielding operation to the attention mechanism layer, or the dilated causal convolution layer in the TCN layer of the next level; wherein, when the random deactivation layer is the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the attention mechanism layer, and when the random deactivation layer is not the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the dilated causal convolution layer in the TCN layer of the next level;

[0057] The attention mechanism layer is used to perform attention weighting processing according to the output of the random inactivation layer of the last level and the health factor of each cable sample, and output the sequence after attention processing to the pooling layer;

[0058] The pooling layer is used to perform pooling processing on the sequence after attention processing, and transmits the pooled data to the fully connected layer, so that the fully connected layer outputs the predicted remaining life of the cable sample.

[0059] Based on the above method embodiment, the present invention provides another embodiment;

[0060] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the cable remaining life assessment method as described in any method embodiment of the present invention.

[0061] Based on the above method embodiment, the present invention provides another embodiment:

[0062] Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the cable remaining life assessment method as described in any method embodiment of the present invention.

[0063] The following beneficial effects are achieved by implementing the embodiments of the present invention:

[0064] The embodiment of the present invention provides a method, device, terminal equipment and storage medium for evaluating the remaining life of a cable based on a health factor. The method extracts the health factor of the cable to be evaluated, inputs the health factor of the cable to be evaluated into a trained time convolutional neural network model, and the time convolutional neural network model predicts the remaining life of the cable to be evaluated based on the health factor. In the process of training the time convolutional neural network model, various data items of the cable sample in the procurement stage, construction stage, operation stage and maintenance stage are collected, and then the main data used to characterize the health status of the cable sample is extracted through normalization processing and principal component analysis, so as to obtain the health factor. Finally, the health factor and the actual remaining life of each cable sample are used as input, and the predicted remaining life of each cable sample is used as output. The time convolutional neural network model is iteratively trained. After the training is completed, the time convolutional neural network model can learn the mapping relationship between the health factor and the remaining life of the cable, and the remaining life of the cable to be evaluated is subsequently predicted by the time convolutional neural network model. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a flow chart of a method for assessing the remaining life of a cable based on health factors provided in one embodiment of the present invention.

[0066] Figure 2 It is a schematic diagram of the structure of a temporal convolutional neural network model provided by one embodiment of the present invention.

[0067] Figure 3 It is a structural schematic diagram of a cable remaining life assessment device based on health factors in one embodiment of the present invention. DETAILED DESCRIPTION

[0068] 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 only 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.

[0069] See also Figure 1 , is a flow chart of a method for assessing the remaining life of a cable based on health factors provided by an embodiment of the present invention;

[0070] An embodiment of the present invention provides a method for evaluating the remaining life of a cable based on a health factor, comprising the following steps:

[0071] Step S1: Obtain the health factor of the cable to be evaluated.

[0072] The above health factor refers to data that can be used to characterize the health status of a cable sample. When evaluating a cable to be evaluated, the health factor of the cable to be evaluated is first obtained. The specific steps for determining the health factor are described below.

[0073] Step S2: input each of the health factors into a preset time convolutional neural network model, so that the time convolutional neural network model determines the remaining life of the cable to be evaluated according to each of the health factors; wherein the training of the time convolutional neural network model includes: obtaining various data items of the cable samples in the procurement stage, construction stage, operation stage and maintenance stage; normalizing each data item and performing principal component analysis to extract the health factor used to characterize the health status of the cable sample; taking the health factor and the actual remaining life of each cable sample as input and the predicted remaining life of each cable sample as output, the time convolutional neural network model is iteratively trained until the time convolutional neural network model converges.

[0074] For step S2, in a preferred embodiment, obtaining various data items of the cable sample in the procurement stage, construction stage, operation stage and maintenance stage includes:

[0075] Obtain the cable nameplate parameters, factory test report and arrival random inspection report of the cable sample to obtain the data items of the cable sample in the procurement stage; obtain the transportation and installation records and the handover and acceptance report of the cable sample to obtain the data items of the cable sample in the construction stage; obtain the cable monitoring data of the cable sample during operation, the environmental data of the operation site and the inspection records during operation to obtain the data items of the cable sample in the operation stage; obtain the maintenance items and parts replacement status of the cable sample to obtain the data items of the cable sample in the maintenance stage.

[0076] Specifically, in the present invention, a cable health assessment system is first established, and the cable health index is composed of data from four stages: procurement, construction, operation, and maintenance:

[0077] ① Data at the procurement stage, including: cable nameplate parameters, factory test report and arrival random inspection test report; cable nameplate parameters include but are not limited to: rated voltage, model and specification, manufacturer name or trademark, manufacturing date, etc.; factory test report includes but is not limited to: mechanical properties test report, thermal properties test report and chemical properties test report, etc.; random inspection test report includes but is not limited to: conductor resistance, insulation thickness, sheath thickness and voltage test;

[0078] ② Construction phase data, including transportation and installation records and handover and acceptance reports; the transportation and installation records include but are not limited to the following data items: cable length, laying method and topography; the handover and acceptance report includes but is not limited to the following data items: installation status of the cable body, accessories, cable channels and auxiliary facilities.

[0079] ③ Operation phase data, including cable monitoring data during operation, environmental data at the operation site, and inspection records during operation; cable monitoring data, including but not limited to the following data items: infrared temperature measurement, load current, and operating voltage; environmental data at the operation site, including but not limited to the following data items: operating environment temperature and humidity; inspection records, including but not limited to the following data items: defects in components such as the main body, intermediate joints, terminal joints, and lightning arresters;

[0080] ④ Maintenance phase data, including: maintenance items and parts replacement status; maintenance items, including but not limited to the following data items: insulation resistance of main insulation, insulation resistance of outer sheath and inner lining, AC withstand voltage test of cable main insulation and dielectric loss factor, parts replacement status, including but not limited to the following data items: number of water ingress, maintenance history and fault history.

[0081] The data items of the above four stages of each cable sample were collected and then normalized to ensure the consistency and comparability of the data. Subsequently, the principal component analysis (PCA) method was applied to the preprocessed data set to identify the core features of the above data and obtain the health factors.

[0082] In an optional embodiment, the health factors finally selected may be: manufacturing date, manufacturer name, thermal performance test, insulation thickness, sheath thickness, cable length, topography, load current, operating voltage, and number of water ingress;

[0083] In a preferred embodiment, the normalization of each data item includes:

[0084] Classify each data item according to whether it is numerical data to obtain numerical data and non-numerical data;

[0085] Each non-numeric data is normalized by one-hot encoding;

[0086] For each numerical data, secondary classification is performed according to the variable relationship between the numerical value and the cable health status, and positively correlated numerical data, negatively correlated numerical data and inverted U-shaped numerical data are obtained; wherein the positively correlated numerical data is numerical data that is positively correlated with the cable health status, the negatively correlated numerical data is numerical data that is negatively correlated with the cable health status, and the inverted U-shaped numerical data is numerical data that is in an inverted U-shaped relationship with the cable health status;

[0087] The positive correlation numerical data are normalized by the following formula:

[0088]

[0089] Negatively correlated numerical data are normalized using the following formula:

[0090]

[0091] The inverted U-shaped numerical data are normalized using the following formula:

[0092]

[0093] Where x is the actual value of the numeric data; x min It is the preset minimum value under normal conditions; x max is the preset maximum value under normal conditions; m and n are preset boundary parameters, and the specific values ​​can be set according to actual conditions.

[0094] Through the above normalization method, all data items are normalized to intuitively characterize the health status of the cable. For numerical data, the normalized value range is in the range of [0,1], where 0 represents the cable sample in the best health state, and 1 represents the worst health state of the cable sample. For non-numerical data, one-hot encoding can be used for normalization. For a non-numerical feature with n different values, one-hot encoding converts it into an n-dimensional binary vector, in which only one element is 1 and the rest are 0.

[0095] After extracting the health factor of each cable sample as described above, each cable sample is divided into a training set and a test set; the health factor and actual remaining life of each cable sample in the training set are used as input, and the predicted remaining life of each cable sample is used as output to iteratively train the temporal convolutional neural network model. After the model converges, the temporal convolutional neural network model can learn the mapping relationship between the health factor and the remaining life of the cable. When the temporal convolutional neural network model is used subsequently, the health factor of the cable to be evaluated is collected and then input into the temporal convolutional neural network model, and the model can predict the life of the cable to be evaluated;

[0096] It should be noted that the factor type of the health factor of the cable to be evaluated should be consistent with the factor type of the health factor corresponding to the cable sample when training the time convolutional neural network model. During training, the actual remaining life of each cable sample is extracted from the retirement appraisal report of each cable sample, and the retirement appraisal report records the actual remaining service life of each cable sample.

[0097] like Figure 2 As shown, in a preferred embodiment, the temporal convolutional neural network model includes: several TCN layers, attention mechanism layers, pooling layers and fully connected layers; each TCN layer includes: an expanded causal convolution layer, a weight normalization layer, an activation function layer and a random inactivation layer;

[0098] The dilated causal convolution layer is used to perform convolution processing on the input data, and then transmit the convolution processed data to the weight normalization layer; wherein, when the dilated causal convolution layer is a dilated causal convolution layer in the first-level TCN layer, the input data is the health factor of each cable sample; when the dilated causal convolution layer is not a dilated causal convolution layer in the first-level TCN layer, the input data is the output of the random inactivation layer in the previous-level TCN layer;

[0099] The weight normalization layer is used to perform weight normalization on the data after convolution processing to obtain weight normalized data, and transmit the weight normalized data to the activation function layer;

[0100] The activation function layer is used to perform nonlinear transformation on the weight normalized data to obtain nonlinear transformed data, and input the nonlinear transformed data into the random dropout layer;

[0101] The random deactivation layer is used to perform a random node shielding operation on the linearly transformed data, and input the data of the node shielding operation to the attention mechanism layer, or the dilated causal convolution layer in the TCN layer of the next level; wherein, when the random deactivation layer is the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the attention mechanism layer, and when the random deactivation layer is not the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the dilated causal convolution layer in the TCN layer of the next level;

[0102] The attention mechanism layer is used to perform attention weighting processing according to the output of the random inactivation layer of the last level and the health factor of each cable sample, and output the sequence after attention processing to the pooling layer;

[0103] The pooling layer is used to perform pooling processing on the sequence after attention processing, and transmits the pooled data to the fully connected layer, so that the fully connected layer outputs the predicted remaining life of the cable sample.

[0104] Schematically, the temporal convolutional neural network model constructed in the present invention includes two levels of TCN layers, each of which integrates a dilated causal convolution layer, a weight normalization layer, an activation function layer, and a random dropout layer;

[0105] For the dilated causal convolution layer, assume there is an input sequence X = {x1, x2, ..., xn}, and a convolution kernel f with width k, whose index range is {0, 1, ..., k-1}. In each TCN layer, the dilated causal convolution layer increases the receptive field by introducing a dilation rate d while maintaining causality, that is, ensuring that each output depends only on its previous input. Specifically, for the input sequence X and the dilated causal convolution operation of a certain layer, its calculation formula can be expressed as:

[0106]

[0107] In the formula, F(s) is the data after convolution processing, k is the size of the convolution kernel, d is the size of the expansion factor, and X s-d·i It is the result of convolution on past data.

[0108] The weight normalization layer can speed up the training process of the model and effectively solve the gradient explosion problem caused by the deepening of the network layers. At the same time, an activation function that can speed up the training process is selected.

[0109] In addition, the introduction of the random dropout layer helps prevent the model from overfitting. After the second random dropout layer, the data stream will be added to the input after the 1×1 convolution operation, that is, the output of the last level of random dropout layer will be concatenated with the health factor of the cable sample after the 1×1 convolution operation to form an input sequence, which is then input to the attention mechanism layer. This design not only prevents information leakage, but also ensures that the dimensions of the input and output data remain consistent.

[0110] For the attention mechanism layer, the role of this layer is to quantify the degree of attention the model pays to different parts of the input health factor sequence by calculating the weight distribution. For a given input sequence and a query information, the attention mechanism calculates the attention coefficient of each input element and then uses it to perform weighted summation on the input sequence to generate a weighted result.

[0111] In a preferred embodiment, after the temporal convolutional neural network model is trained, it also includes:

[0112] Calculate the relative error and the root mean square error based on the actual remaining life of the cable sample and the corresponding predicted remaining life;

[0113] When it is determined that the relative error or the root mean square error is not within the corresponding preset range, the structure or parameters of the temporal convolutional neural network model are adjusted.

[0114] Specifically, after the temporal convolutional neural network model is trained, it is evaluated using test set data to ensure the accuracy and reliability of the model in practical applications.

[0115] In order to objectively evaluate the prediction performance of the temporal convolutional neural network model, relative error (RE) and root mean square error (RMSE) are selected as key evaluation indicators. Relative error measures the deviation ratio between the predicted value and the actual value, while the root mean square error reflects the square root of the average of the squares of the difference between the predicted value and the actual value, which is more sensitive to outliers. Based on the verification results of these two evaluation indicators, we can further adjust the structure of the temporal convolutional neural network model or optimize its parameters to improve the prediction accuracy of the model. The calculation formulas of these two evaluation indicators are:

[0116]

[0117] Where RE is the relative error, RMSE is the root mean square error, RULp is the predicted remaining life of the cable sample, RULt is the true life of the cable sample, and j is the number of cable samples used for testing.

[0118] The above-mentioned cable remaining life assessment method based on health factors of the present invention can more comprehensively reflect the current status and future trends of the cable by comprehensively considering multiple health factors. The temporal convolutional neural network model can learn the complex relationship between these health factors and predict the remaining life accordingly, which improves the accuracy and reliability of the prediction, and can realize the real-time prediction and online monitoring of the remaining life of the cable, which helps to timely discover potential problems of the cable and take corresponding maintenance measures, thereby extending the service life of the cable and improving the safety and stability of the power system.

[0119] Based on the above method embodiment of the present invention, a corresponding device embodiment is provided;

[0120] like Figure 3 As shown, another embodiment of the present invention provides a device for assessing the remaining life of a cable based on a health factor, comprising:

[0121] Includes: data acquisition module, remaining life prediction module and model training module;

[0122] The data acquisition module is used to obtain the health factor of the cable to be evaluated;

[0123] The remaining life prediction module is used to input each of the health factors into a preset time convolutional neural network model, so that the time convolutional neural network model determines the remaining life of the cable to be evaluated according to each of the health factors;

[0124] The model training module is used to obtain various data items of cable samples in the procurement stage, construction stage, operation stage and maintenance stage; normalize each data item and perform principal component analysis to extract health factors used to characterize the health status of the cable samples; use the health factors and actual remaining life of each cable sample as input and the predicted remaining life of each cable sample as output to iteratively train the time convolutional neural network model until the time convolutional neural network model converges.

[0125] In a preferred embodiment, the model training module normalizes each data item, including:

[0126] Classify each data item according to whether it is numerical data to obtain numerical data and non-numerical data;

[0127] Each non-numeric data is normalized by one-hot encoding;

[0128] For each numerical data, secondary classification is performed according to the variable relationship between the numerical value and the cable health status, and positively correlated numerical data, negatively correlated numerical data and inverted U-shaped numerical data are obtained; wherein the positively correlated numerical data is numerical data that is positively correlated with the cable health status, the negatively correlated numerical data is numerical data that is negatively correlated with the cable health status, and the inverted U-shaped numerical data is numerical data that is in an inverted U-shaped relationship with the cable health status;

[0129] The positive correlation numerical data are normalized by the following formula:

[0130]

[0131] Negatively correlated numerical data are normalized using the following formula:

[0132]

[0133] The inverted U-shaped numerical data are normalized using the following formula:

[0134]

[0135] Where x is the actual value of the numeric data; x min It is the preset minimum value under normal conditions; x max is the preset maximum value under normal conditions; m and n are the preset boundary parameters.

[0136] In a preferred embodiment, the temporal convolutional neural network model includes: several TCN layers, attention mechanism layers, pooling layers and fully connected layers; each TCN layer includes: an expanded causal convolution layer, a weight normalization layer, an activation function layer and a random inactivation layer;

[0137] The dilated causal convolution layer is used to perform convolution processing on the input data, and then transmit the convolution processed data to the weight normalization layer; wherein, when the dilated causal convolution layer is a dilated causal convolution layer in the first-level TCN layer, the input data is the health factor of each cable sample; when the dilated causal convolution layer is not a dilated causal convolution layer in the first-level TCN layer, the input data is the output of the random inactivation layer in the previous-level TCN layer;

[0138] The weight normalization layer is used to perform weight normalization on the data after convolution processing to obtain the weight normalized data, and transmit the weight normalized data to the activation function layer;

[0139] The activation function layer is used to perform nonlinear transformation on the weight normalized data to obtain nonlinear transformed data. The nonlinear transformed data is input into the random dropout layer;

[0140] The random deactivation layer is used to perform a random node shielding operation on the linearly transformed data, and input the data of the node shielding operation to the attention mechanism layer, or the dilated causal convolution layer in the TCN layer of the next level; wherein, when the random deactivation layer is the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the attention mechanism layer, and when the random deactivation layer is not the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the dilated causal convolution layer in the TCN layer of the next level;

[0141] The attention mechanism layer is used to perform attention weighting processing according to the output of the random inactivation layer of the last level and the health factor of each cable sample, and output the sequence after attention processing to the pooling layer;

[0142] The pooling layer is used to perform pooling processing on the sequence after attention processing, and transmits the pooled data to the fully connected layer, so that the fully connected layer outputs the predicted remaining life of the cable sample.

[0143] It can be understood that the above-mentioned device item embodiment corresponds to the method item embodiment of the present invention, which can implement the cable remaining life assessment method based on health factors provided by any one of the above-mentioned method item embodiments of the present invention.

[0144] It should be noted that the device embodiments described above are merely schematic, wherein the units / modules described as separate components may or may not be physically separated, and the components displayed as units / modules may or may not be physical units / modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without creative work. The schematic diagram is merely an example of the device and does not constitute a limitation on the device, and may include more or fewer components than shown in the diagram, or a combination of certain components, or different components.

[0145] Based on the above method embodiment of the present invention, a corresponding terminal device embodiment is provided;

[0146] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the health factor-based cable remaining life assessment method provided by any one of the above-mentioned method embodiments of the present invention is implemented.

[0147] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0148] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The device may include, but is not limited to, a processor and a memory. Those skilled in the art may understand that, for example, the device may also include an input / output device, a network access device, a bus, etc.

[0149] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the device, and various interfaces and lines are used to connect various parts of the entire device.

[0150] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0151] Another embodiment is provided based on the above method embodiment of the present invention. Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute any one of the health factor-based cable remaining life assessment methods described above in the present invention.

[0152] Wherein, the storage medium is a computer-readable storage medium, and the module / unit integrated in the device can be stored in a computer-readable storage medium if it is implemented in the form of a software functional unit and sold or used as an independent product. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0153] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A cable remaining life assessment method based on health factors, characterized in that: include: Obtain the health factor of the cable to be evaluated; Inputting each of the health factors into a preset time convolutional neural network model, so that the time convolutional neural network model determines the remaining life of the cable to be evaluated according to each of the health factors; The training of the temporal convolutional neural network model includes: Obtain various data items of cable samples during the procurement stage, construction stage, operation stage and maintenance stage; Each data item is normalized and principal component analysis is performed to extract health factors used to characterize the health status of cable samples; Taking the health factor and the actual remaining life of each cable sample as input and the predicted remaining life of each cable sample as output, the temporal convolutional neural network model is iteratively trained until the temporal convolutional neural network model converges.

2. The cable remaining life assessment method based on health factor according to claim 1, characterized in that: Obtain various data items of cable samples during the procurement, construction, operation and maintenance stages, including: Obtain the cable nameplate parameters, factory test report and arrival sampling test report of the cable sample, and obtain various data items of the cable sample in the procurement stage; Obtain the transportation and installation records and handover acceptance report of the cable samples, and obtain various data items of the cable samples during the construction phase; Obtain the cable monitoring data of the cable sample during operation, the environmental data of the operation site, and the inspection records during operation, and obtain various data items of the cable sample during the operation stage; Obtain the maintenance items and parts replacement status of the cable sample, and obtain various data items of the cable sample during the maintenance stage.

3. The cable remaining life assessment method based on health factor according to claim 1, characterized in that: The normalization of each data item includes: Classify each data item according to whether it is numerical data to obtain numerical data and non-numerical data; Each non-numeric data is normalized by one-hot encoding; For each numerical data, secondary classification is performed according to the variable relationship between the numerical value and the cable health status, and positively correlated numerical data, negatively correlated numerical data and inverted U-shaped numerical data are obtained; wherein the positively correlated numerical data is numerical data that is positively correlated with the cable health status, the negatively correlated numerical data is numerical data that is negatively correlated with the cable health status, and the inverted U-shaped numerical data is numerical data that is in an inverted U-shaped relationship with the cable health status; The positive correlation numerical data are normalized by the following formula: Negatively correlated numerical data are normalized using the following formula: The inverted U-shaped numerical data are normalized using the following formula: Among them, x is the actual value of the numerical data; x min It is the preset minimum value under normal conditions; x max is the preset maximum value under normal conditions; m and n are the preset boundary parameters.

4. The cable remaining life assessment method based on health factor according to claim 3 is characterized in that: The temporal convolutional neural network model includes: several TCN layers, attention mechanism layers, pooling layers and fully connected layers; each TCN layer includes: an extended causal convolution layer, a weight normalization layer, an activation function layer and a random inactivation layer; The dilated causal convolution layer is used to perform convolution processing on the input data, and then transmit the convolution processed data to the weight normalization layer; wherein, when the dilated causal convolution layer is a dilated causal convolution layer in the first-level TCN layer, the input data is the health factor of each cable sample; when the dilated causal convolution layer is not a dilated causal convolution layer in the first-level TCN layer, the input data is the output of the random inactivation layer in the previous-level TCN layer; The weight normalization layer is used to perform weight normalization on the data after convolution processing to obtain the weight normalized data, and transmit the weight normalized data to the activation function layer; The activation function layer is used to perform nonlinear transformation on the weight normalized data to obtain nonlinear transformed data, and input the nonlinear transformed data into the random dropout layer; The random deactivation layer is used to perform a random node shielding operation on the linearly transformed data, and input the data of the node shielding operation to the attention mechanism layer, or the dilated causal convolution layer in the TCN layer of the next level; wherein, when the random deactivation layer is the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the attention mechanism layer, and when the random deactivation layer is not the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the dilated causal convolution layer in the TCN layer of the next level; The attention mechanism layer is used to perform attention weighting processing according to the output of the random inactivation layer of the last level and the health factor of each cable sample, and output the sequence after attention processing to the pooling layer; The pooling layer is used to perform pooling processing on the sequence after attention processing, and transmits the pooled data to the fully connected layer, so that the fully connected layer outputs the predicted remaining life of the cable sample.

5. The cable remaining life assessment method based on health factor according to claim 4, characterized in that: After the temporal convolutional neural network model is trained, the method further includes: Calculate the relative error and the root mean square error based on the actual remaining life of the cable sample and the corresponding predicted remaining life; When it is determined that the relative error or the root mean square error is not within the corresponding preset range, the structure or parameters of the temporal convolutional neural network model are adjusted.

6. A device for assessing the remaining life of a cable based on health factors, characterized in that: include: Data acquisition module, remaining life prediction module and model training module; The data acquisition module is used to obtain the health factor of the cable to be evaluated; The remaining life prediction module is used to input each of the health factors into a preset time convolutional neural network model, so that the time convolutional neural network model determines the remaining life of the cable to be evaluated according to each of the health factors; The model training module is used to obtain various data items of cable samples in the procurement stage, construction stage, operation stage and maintenance stage; Each data item is normalized and subjected to principal component analysis to extract health factors used to characterize the health status of the cable samples. The health factors and actual remaining life of each cable sample are used as input and the predicted remaining life of each cable sample is used as output to iteratively train the temporal convolutional neural network model until the temporal convolutional neural network model converges.

7. The device for assessing the remaining life of a cable based on health factors according to claim 6, characterized in that: The model training module normalizes each data item, including: Classify each data item according to whether it is numerical data to obtain numerical data and non-numerical data; Each non-numeric data is normalized by one-hot encoding; For each numerical data, secondary classification is performed according to the variable relationship between the numerical value and the cable health status, and positively correlated numerical data, negatively correlated numerical data and inverted U-shaped numerical data are obtained; wherein the positively correlated numerical data is numerical data that is positively correlated with the cable health status, the negatively correlated numerical data is numerical data that is negatively correlated with the cable health status, and the inverted U-shaped numerical data is numerical data that is in an inverted U-shaped relationship with the cable health status; The positive correlation numerical data are normalized by the following formula: Negatively correlated numerical data are normalized using the following formula: The inverted U-shaped numerical data are normalized using the following formula: Among them, x is the actual value of the numerical data; x min It is the preset minimum value under normal conditions; x max is the preset maximum value under normal conditions; m and n are the preset boundary parameters.

8. The device for assessing the remaining life of a cable based on health factors according to claim 7, characterized in that: The temporal convolutional neural network model includes: several TCN layers, attention mechanism layers, pooling layers and fully connected layers; each TCN layer includes: an extended causal convolution layer, a weight normalization layer, an activation function layer and a random inactivation layer; The dilated causal convolution layer is used to perform convolution processing on the input data, and then transmit the convolution processed data to the weight normalization layer; wherein, when the dilated causal convolution layer is a dilated causal convolution layer in the first-level TCN layer, the input data is the health factor of each cable sample; when the dilated causal convolution layer is not a dilated causal convolution layer in the first-level TCN layer, the input data is the output of the random inactivation layer in the previous-level TCN layer; The weight normalization layer is used to perform weight normalization on the data after convolution processing to obtain the weight normalized data, and transmit the weight normalized data to the activation function layer; The activation function layer is used to perform nonlinear transformation on the weight normalized data to obtain nonlinear transformed data, and input the nonlinear transformed data into the random dropout layer; The random deactivation layer is used to perform a random node shielding operation on the linearly transformed data, and input the data of the node shielding operation to the attention mechanism layer, or the dilated causal convolution layer in the TCN layer of the next level; wherein, when the random deactivation layer is the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the attention mechanism layer, and when the random deactivation layer is not the random deactivation layer in the last level of the TCN layer, the random deactivation layer inputs the data of the node shielding operation to the dilated causal convolution layer in the TCN layer of the next level; The attention mechanism layer is used to perform attention weighting processing according to the output of the random inactivation layer of the last level and the health factor of each cable sample, and output the sequence after attention processing to the pooling layer; The pooling layer is used to perform pooling processing on the sequence after attention processing, and transmits the pooled data to the fully connected layer, so that the fully connected layer outputs the predicted remaining life of the cable sample.

9. A terminal device, characterized in that: include: one or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the cable remaining life assessment method based on health factors as described in any one of claims 1-5.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for assessing the remaining life of a cable based on health factors as described in any one of claims 1 to 5 is implemented.

Images