A method and system for monitoring and early warning of power supply safety status of liquid-cooled cables

Abstract

The present invention discloses a method and system for monitoring and warning the power supply safety status of a liquid-cooled cable, including: obtaining the historical operating parameters of the liquid-cooled cable and the corresponding temperature field measurement values, using Gaussian regression to construct a temperature prediction model for the liquid-cooled cable, reading the temperature field data under different states, and obtaining the state change law of the liquid-cooled cable; constructing a diagnostic model for the power supply status of the liquid-cooled cable, mining the hidden layer features of the temperature field data, and training the mapping relationship between the hidden layer features and different abnormal states; using data perception to obtain the current operating parameters of the liquid-cooled cable and perform preprocessing, and using the temperature prediction model and the diagnostic model to predict the temperature field data and diagnose and classify the power supply status, and generate a safety warning according to the diagnosis and classification results. The present invention collects information and monitors and warns the safety status of the liquid-cooled cable in operation, provides real-time safety status information of the liquid-cooled cable, and ensures the safe operation of the cable of the charging equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of cable status monitoring, and more specifically, to a power supply safety status monitoring and early warning method and system for a liquid-cooled cable. Background Art

[0002] In the context of low carbon and environmental protection, new energy vehicles have won the favor of consumers due to their environmental protection and low travel costs. However, problems such as slow charging speed and low efficiency still restrict their development. At present, the main ways to meet user needs are to increase the voltage platform of the whole vehicle or increase the current transmitted by the charging pile. But no matter which method is used, it inevitably puts higher requirements on the conductivity of the cable.

[0003] The technical route of liquid-cooled high-power charging cable mainly utilizes liquid cooling technology. By cooling and lowering the temperature, the cable maintains a low constant temperature during the charging process, overcomes the thermal damage to the charging gun, charging cable, and charging pile caused by heat generation, and thus meets the high current and heat faced by fast charging. Its main principle is to add a liquid cooling circulation system inside the cable, and absorb the generated heat by the flow of cooling medium to keep the cable working temperature within a reasonable range, thereby ensuring that the cable is not easy to overheat and burn during long-term charging, making fast charging safer. And the difference between liquid-cooled cables and ordinary cables is that liquid-cooled charging cables can achieve cooling and heat dissipation while conducting electricity, transmit large currents while reducing the cable diameter, and are lighter than traditional charging cables. However, the real-time monitoring and early warning of the safety status of traditional charging cables usually only extracts low-frequency signals when the cable is operating abnormally, resulting in low accuracy in abnormal state identification and long fault location warning time. Therefore, there is an urgent need for a high-precision power supply safety status monitoring and efficient early warning method for liquid-cooled cables. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a power supply safety status monitoring and early warning method and system for liquid-cooled cables. The purpose is to realize the analysis and learning of monitoring information through information perception combined with deep learning methods, determine whether there is a coolant leakage and other problems, and ensure the safe operation of the cable of the charging equipment.

[0005] A first aspect of the present invention provides a method for monitoring and early warning of a power supply safety state of a liquid-cooled cable, comprising:

[0006] Obtain historical operating parameters of the liquid-cooled cable and corresponding temperature field measurement values, select characteristic variables according to the temperature field changes, and construct a data set based on the characteristic variables;

[0007] Gaussian regression is used to build a temperature prediction model for liquid-cooled cables. Gaussian regression training is performed through the data set to obtain historical operating parameters in different states. The temperature field data in different states are read through the trained temperature prediction model to analyze and obtain the state change law of the liquid-cooled cables.

[0008] Constructing a diagnostic model for the power supply status of a liquid-cooled cable, mining the hidden features of the temperature field data according to the state change law, and training the mapping relationship between the hidden features and different working conditions;

[0009] The current operating parameters of the liquid-cooled cable are acquired and preprocessed using data perception, and the temperature field data is predicted and the power supply status is diagnosed and classified using the temperature prediction model and the diagnosis model, and a safety warning is generated based on the diagnosis and classification results.

[0010] In this solution, characteristic variables are selected according to the change of temperature field, and a data set is constructed based on the characteristic variables, specifically:

[0011] The historical operating parameters of the liquid-cooled cable under different working conditions are obtained and preprocessed, the original operating parameter sequence is converted into an original sample set using a sliding window method, and the original sample set is decomposed and reconstructed using a wavelet packet to generate a number of reconstructed sample signals;

[0012] Calculating the spread entropy of the reconstructed sample signal to characterize the volatility of different reconstructed sample signals, and selecting the reconstructed sample signal not less than a preset spread entropy threshold to generate a preliminary sample set;

[0013] The temperature field change of the liquid cooling cable is taken as the dependent variable, and the Pearson correlation coefficient is used to calculate the correlation between each reconstructed sample signal and the dependent variable in the preliminary sample set, and the reconstructed sample signal with the largest correlation is selected;

[0014] The mRMR algorithm is improved based on the Pearson correlation coefficient. The remaining reconstructed sample signals in the preliminary sample set are introduced one by one according to the selected reconstructed sample signals. The feature correlation is sorted according to the principle of maximum correlation between the reconstructed sample signals and the dependent variable and minimum redundancy between the reconstructed sample signals.

[0015] Using SVM as the prediction model, verifying by adopting S-fold cross-validation strategy according to the feature correlation ranking, training in turn to obtain S prediction models, ranking the classification prediction accuracy as the importance, and selecting a preset number of reconstructed sample signals as feature variables;

[0016] The characteristic parameters of the historical operating parameters are selected according to the characteristic variables, and a data set is constructed by combining the selected characteristic parameters with the temperature change of the liquid-cooled cable.

[0017] In this solution, the operating parameters of different working conditions are obtained, the temperature field data under different conditions are read through the trained temperature prediction model, and the state change law of the liquid cooling cable is analyzed and obtained, specifically:

[0018] The data set is divided into a training set and a test set for Gaussian regression training, a covariance function is selected as a kernel function, Bayesian optimization is used as an optimizer for Gaussian regression, Gaussian process regression is performed through the training set and the test set, and a trained temperature prediction model is output when the error meets the requirements;

[0019] The historical operating parameters of different working conditions are clustered and analyzed, and the Euclidean distance is used as the similarity measurement function. The working condition labels are divided and set according to the cluster similarity calculation to generate the historical operating parameter subsets corresponding to each working condition label.

[0020] Different subsets of historical operating parameters are encoded using an autoencoder to obtain potential feature distributions, and features are reconstructed according to the potential feature distributions. An adversarial mechanism is introduced to generate potential representations consistent with each historical operating parameter by adding an identifier and mapping them to corresponding positions, and to penalize reconstructed data that deviates too much from the potential feature distributions.

[0021] The reconstructed data is used to enhance the original historical operating parameter subset, and the enhanced historical operating parameter subset is used as a model input of a temperature prediction model to predict corresponding temperature field data.

[0022] In this scheme, the state change law of the liquid cooling cable is analyzed and obtained, specifically:

[0023] Obtain a temperature field data set corresponding to a subset of historical operating parameters enhanced by different operating state tags, and read a liquid inlet temperature interval and a liquid outlet temperature interval in a liquid cooling cable from the temperature field data set;

[0024] The operating parameter deviations between different historical operating parameter subsets are read, and the time series change characteristics of the liquid inlet temperature range and the liquid outlet temperature range of different working condition status labels are analyzed. The state change law of the liquid cooling cable is generated according to the operating parameter deviations and time series change characteristics.

[0025] In this solution, a diagnostic model of the power supply status of the liquid cooling cable is constructed, and the hidden layer features of the temperature field data are mined according to the state change law, specifically:

[0026] A double hidden layer extreme learning machine was constructed, the state change law of the liquid cooling cable was imported, the input weights and biases of the first hidden layer and the second hidden layer were randomly generated, and the output weight matrix of the double hidden layer extreme learning machine was obtained through training.

[0027] By connecting several double hidden layer extreme learning machines in series, the temperature field data corresponding to different working conditions are encoded layer by layer. The input of each hidden layer node is the transpose of the output weight matrix of the previous hidden layer node, so as to mine the deep hidden layer features of the temperature field data.

[0028] In this solution, the mapping relationship between the hidden layer features and different working conditions is trained as follows:

[0029] Obtain the hidden layer features of the temperature field data set corresponding to each working condition label, perform a graph representation based on the association between the hidden layer features and the working condition label and the association information between different working condition labels, and use a graph attention network to generate a graph structure of a normal working condition and various types of abnormal working conditions;

[0030] Graph convolution is used to obtain the embedded representation of the working condition state node and the hidden layer feature node, and the attention mechanism is introduced to obtain the attention weights of the neighbor nodes in the graph structure. The attention weights are used to perform weighted aggregation of the neighbor nodes to update the embedded representation of the working condition state node and the hidden layer feature node.

[0031] The updated embedded representations of the working condition state node and the hidden layer feature node are average pooled respectively to obtain the final embedded representations of the working condition state node and the hidden layer feature node on the associated information;

[0032] The CTR prediction model is used to match the working state nodes and the hidden layer feature nodes, the mapping relationship between the hidden layer features and different working states is trained, and a diagnostic model for the power supply state of the liquid cooling cable is constructed.

[0033] In this solution, data perception is used to obtain the current operating parameters of the liquid cooling cable and preprocess them, and the temperature field data prediction and power supply status diagnosis classification are performed through the temperature prediction model and diagnosis model, specifically:

[0034] Real-time perception of the operating parameters of the liquid-cooled cable, importing the current operating parameters into the temperature prediction model to obtain the temperature field prediction data of the liquid-cooled cable after a preset time;

[0035] The current temperature field data and the temperature field prediction data are imported into the diagnosis model, the corresponding hidden layer features are read through the diagnosis model, the obtained hidden layer features are represented by graphs to obtain corresponding embedded representations, and the embedded representations of the hidden layer features are learned and updated and then matched using the CTR prediction model;

[0036] The matching degree between the current hidden layer feature and the working condition is obtained by inner product calculation, and the corresponding working condition label is generated. If it is an abnormal working condition label, warning information is generated according to the label category.

[0037] A second aspect of the present invention provides a liquid-cooled cable power supply safety status monitoring and early warning system, comprising a memory, a processor, a data perception module, a temperature field prediction module, a power supply status diagnosis module and an operation early warning module, wherein the memory and the processor store and execute a liquid-cooled cable power supply safety status monitoring and early warning method program;

[0038] The data sensing module uses data sensing to obtain the operating parameters of the liquid cooling cable and performs preprocessing, and extracts features according to preset feature variables to construct a model input data set;

[0039] The temperature field prediction module uses Gaussian regression to build a temperature prediction model for the liquid cooling cable, and uses the temperature prediction model to predict the temperature field data of the liquid cooling cable after a preset time based on the input data set;

[0040] The power supply status diagnosis module mines the hidden features of the temperature field data based on the state change law of the liquid cooling cable, trains the mapping relationship between the hidden features and different working conditions, and diagnoses the abnormal state through the predicted temperature field data;

[0041] The operation warning module generates a safety warning according to the diagnosis and classification results of the working condition of the liquid-cooled cable, and sends it to the operation and maintenance personnel according to a preset method.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention collects information and monitors and warns the safety status of the liquid-cooled cable in operation, provides real-time safety status information of the liquid-cooled cable, and realizes analysis and learning of operating parameter information through real-time information perception combined with a deep learning method, thereby determining whether the liquid-cooled cable has abnormal problems such as overtemperature or coolant leakage, thereby ensuring the safe operation of the cable of the charging equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments or exemplary embodiments of the present invention, the drawings required for use in the embodiments or exemplary descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to the drawings without paying creative work.

[0045] Figure 1 A flow chart showing a method for monitoring and early warning of power supply safety status of a liquid-cooled cable according to the present invention is shown;

[0046] Figure 2 A flow chart showing the reading of temperature field data under different states by a temperature prediction model is shown;

[0047] Figure 3A flow chart for constructing a diagnostic model for the power supply status of a liquid cooling cable is shown;

[0048] Figure 4 A block diagram of a liquid-cooled cable power supply safety status monitoring and early warning system of the present invention is shown. DETAILED DESCRIPTION

[0049] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0050] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0051] Figure 1 A flow chart of a method for monitoring and early warning the power supply safety status of a liquid-cooled cable is shown.

[0052] like Figure 1 As shown, this embodiment provides a method for monitoring and early warning of the power supply safety status of a liquid cooling cable, comprising:

[0053] S102, obtaining historical operating parameters of the liquid-cooled cable and corresponding temperature field measurement values, selecting characteristic variables according to temperature field changes, and constructing a data set based on the characteristic variables;

[0054] S104, using Gaussian regression to build a temperature prediction model for the liquid-cooled cable, performing Gaussian regression training on the data set to obtain historical operating parameters in different states, reading temperature field data in different states through the trained temperature prediction model, and analyzing and obtaining the state change law of the liquid-cooled cable;

[0055] S106, constructing a diagnostic model for the power supply state of the liquid cooling cable, mining hidden features of the temperature field data according to the state change law, and training a mapping relationship between the hidden features and different working conditions;

[0056] S108, using data perception to obtain the current operating parameters of the liquid cooling cable and perform preprocessing, using the temperature prediction model and diagnosis model to perform temperature field data prediction and power supply status diagnosis classification, and generating a safety warning based on the diagnosis classification results.

[0057] It should be noted that the historical operating parameters such as the coolant inlet temperature, coolant outlet temperature, excitation current, cable core conductivity, cable core relative magnetic permeability, insulation thermal conductivity, convective heat transfer coefficient and external environment temperature under different working conditions of the liquid-cooled cable are obtained and preprocessed. The preprocessing mainly includes data missing value processing, outlier processing, data standardization processing and data division, etc. The original operating parameter sequence is converted into an original sample set using the sliding window method, and the original sample set is decomposed and reconstructed using wavelet packets to generate a number of reconstructed sample signals. Wavelet packets are a multi-resolution time-frequency analysis method that is suitable for processing non-stationary signals and can better decompose the high-frequency details of the signal without repeated decomposition or missing signals in a certain frequency band. The spread entropy of the reconstructed sample signal is calculated to characterize the volatility of different reconstructed sample signals. The larger the spread entropy, the higher the volatility of the feature, and vice versa. The reconstructed sample signal that is not less than the preset spread entropy threshold is selected to generate a preliminary sample set.

[0058] Taking the temperature field change of the liquid-cooled cable as the dependent variable, the Pearson correlation coefficient is used to calculate the correlation between each reconstructed sample signal in the preliminary sample set and the dependent variable, and the reconstructed sample signal with the largest correlation is selected; the mRMR algorithm is used to select variables with small redundancy between features and large correlation between features and dependent variables. It has the advantages of fast calculation speed and good stability. The mRMR algorithm is improved based on the Pearson correlation coefficient, and the Pearson correlation coefficient is used to calculate the redundancy between features and the correlation between features and dependent variables. According to the selected reconstructed sample signals, the remaining reconstructed sample signals in the preliminary sample set are introduced one by one, and the feature correlation is sorted according to the principle of maximizing the correlation between the reconstructed sample signals and the dependent variable and minimizing the redundancy between the reconstructed sample signals; SVM is used as the prediction model, and in order to reduce the impact of randomness on the results, the S-fold cross-validation strategy is adopted for verification according to the feature correlation sorting, and the data sets corresponding to the temperature field measurement values ​​are divided into S groups, and each group of data is used as a verification set in turn, and the remaining S-1 groups of data are used as training sets. S prediction models are trained, and the classification prediction accuracy is used as the importance ranking, and a preset number of reconstructed sample signals are selected as feature variables; the feature parameters of the historical operating parameters are selected according to the feature variables, and the selected feature parameters are used to construct a data set in combination with the temperature change of the liquid cooling cable.

[0059] Figure 2 A flow chart showing the reading of temperature field data under different states through a temperature prediction model.

[0060] According to an embodiment of the present invention, the operating parameters of different working conditions are obtained, the temperature field data under different conditions are read through the trained temperature prediction model, and the state change law of the liquid cooling cable is analyzed and obtained, specifically:

[0061] S202, using the data set to be divided into a training set and a test set for Gaussian regression training, selecting a covariance function as a kernel function, using Bayesian optimization as an optimizer for Gaussian regression, performing Gaussian process regression through the training set and the test set, and outputting a trained temperature prediction model when the error meets the requirements;

[0062] S204, clustering analysis is performed on the historical operating parameters of different operating conditions, using Euclidean distance as a similarity measurement function, dividing and setting the operating condition labels according to cluster similarity calculation, and generating a subset of historical operating parameters corresponding to each operating condition label;

[0063] S206, using an autoencoder to encode different subsets of historical operating parameters to obtain a potential feature distribution, reconstructing features according to the potential feature distribution, introducing an adversarial mechanism, adding an identifier to generate a potential representation consistent with each historical operating parameter and mapping it to a corresponding position, and penalizing reconstructed data that deviates too much from the potential feature distribution;

[0064] S208, using the reconstructed data to enhance the original historical operating parameter subset, using the enhanced historical operating parameter subset as a model input of a temperature prediction model to predict corresponding temperature field data.

[0065] It should be noted that the optimizer in the Gaussian process regression model is set to Bayesian optimization, the basis function type is linear, and the kernel function type is square exponential covariance function. The training set and test set divided by the data set are input into the Gaussian process regression for training and testing. According to the Bayesian principle, the Gaussian process establishes a priori functions in the data set. If the error requirements are met, the temperature prediction model is directly output. If the requirements are not met, the hyperparameters in the covariance function need to be adjusted and retrained. The regression equation of the Gaussian process is: For input variables Output results The mean of For input variables Output results The variance of is a Gaussian distribution, Input variables for training, Output the training results. Introduce the adversarial mechanism, add a discriminator to the latent feature layer, preset the contraction constraint of the latent feature space, map the reconstructed data to the same position, and punish the reconstructed data with too large distribution deviation; perform alternating training according to consistency constraints and contraction constraints, and obtain reconstructed data to perform data expansion on the historical operation parameter subset.

[0066] Obtain a temperature field data set corresponding to a subset of historical operating parameters enhanced with different operating status labels, and read a liquid inlet temperature range and a liquid outlet temperature range in the liquid-cooled cable from the temperature field data set; read the operating parameter deviations between different historical operating parameter subsets, and analyze the time series change characteristics of the liquid inlet temperature range and the liquid outlet temperature range of different operating status labels, and generate a state change law of the liquid-cooled cable according to the operating parameter deviations and time series change characteristics. A monitoring instance of the liquid-cooled cable under the normal working condition label is selected, a liquid-cooling control database is constructed to store the monitoring instance, and clustering is performed by ambient temperature in the liquid-cooling control database. A standard ambient temperature interval is selected in the clustering result, and the distribution deviation of the coolant inlet temperature data subset corresponding to each other ambient temperature interval and the coolant inlet temperature data subset corresponding to the standard temperature interval is obtained, and the corresponding uncertainty coefficient is determined according to the distribution deviation; the current operating parameters of the liquid-cooled cable are obtained and retrieved in the liquid-cooling control database using similarity calculation, and a monitoring instance that meets the preset similarity standard is selected, and the coolant inlet temperature interval corresponding to the monitoring instance is read, and the corresponding uncertainty coefficient is read in the liquid-cooling control database according to the current ambient temperature to adjust the coolant inlet temperature interval, and the coolant inlet temperature mean is extracted according to the adjusted coolant inlet temperature interval.

[0067] Figure 3 A flow chart for constructing a diagnostic model for the power supply status of a liquid cooling cable is shown.

[0068] According to an embodiment of the present invention, a diagnostic model for the power supply status of a liquid cooling cable is constructed, specifically:

[0069] S302, constructing a double hidden layer extreme learning machine, importing the state change law of the liquid cooling cable, randomly generating input weights and biases of the first hidden layer and the second hidden layer, and training to obtain the output weight matrix of the double hidden layer extreme learning machine;

[0070] S304, using a plurality of double hidden layer extreme learning machines in series to encode the temperature field data corresponding to different working conditions layer by layer, wherein the input of each hidden layer node is the transpose of the output weight matrix of the previous hidden layer node, and the deep hidden layer features of the temperature field data are mined;

[0071] S306, obtaining the hidden layer features of the temperature field data set corresponding to each operating state label, performing a graph representation based on the association between the hidden layer features and the operating state label and the association information between different operating state labels, and using a graph attention network to generate a graph structure of a normal operating state and various types of abnormal operating states;

[0072] S308, using graph convolution to obtain the embedded representation of the working condition state node and the hidden layer feature node, and introducing the attention mechanism to obtain the attention weights of the neighbor nodes in the graph structure, and using the attention weights to perform weighted aggregation of the neighbor nodes to update the embedded representation of the working condition state node and the hidden layer feature node;

[0073] S310, average pooling the updated embedded representations of the operating state node and the hidden layer feature node respectively to obtain the final embedded representations of the operating state node and the hidden layer feature node on the associated information;

[0074] S312, using the CTR prediction model to match the working state node and the hidden layer feature node, training the mapping relationship between the hidden layer feature and different working states, and building a diagnostic model for the power supply state of the liquid cooling cable.

[0075] It should be noted that when constructing a double hidden layer extreme learning machine, the number of nodes in the two hidden layers is preset, and the number of nodes in the hidden layer is set to be greater than the number of input nodes to achieve sparse feature expression; the input of each hidden layer node is the transpose of the output weight matrix of the previous hidden layer node, and the temperature field data is mapped to a higher dimensional space through layer-by-layer abstract extraction of several double hidden layer extreme learning machines to improve recognition accuracy and generalization performance.

[0076] According to the association between the hidden layer features and the operating state labels and the association information between different operating state labels, a graphical representation is performed. Different operating state labels are divided into normal operating state, slightly abnormal operating state and abnormal operating state, etc. Abnormal operating state is divided into fault categories such as temperature abnormality, leakage abnormality and bonding degradation. The CTR prediction model is used to calculate the inner product of the hidden layer features and the corresponding embedded representation of different operating states, and the inner product is used to characterize the matching degree between the two.

[0077] It should be noted that the operating parameters of the liquid-cooled cable are sensed in real time, and the current operating parameters are imported into the temperature prediction model to obtain the temperature field prediction data of the liquid-cooled cable after a preset time; the current temperature field data and the temperature field prediction data are imported into the diagnosis model, and the corresponding hidden layer features are read through the diagnosis model, and the acquired hidden layer features are represented by a graph to obtain the corresponding embedded representation, and the embedded representation of the hidden layer features is learned and updated and then matched using the CTR estimation model; the matching degree between the current hidden layer features and the operating status is obtained according to the inner product calculation, and the corresponding operating status label is generated. If it is an abnormal operating status label, an early warning message is generated according to the label category and sent to the operation and maintenance personnel to ensure the timely operation and maintenance of the charging cable.

[0078] Figure 4 A block diagram of a liquid-cooled cable power supply safety status monitoring and early warning system is shown.

[0079] This embodiment provides a liquid-cooled cable power supply safety state monitoring and early warning system 4, including a memory 41, a processor 42, a data perception module 43, a temperature field prediction module 44, a power supply state diagnosis module 45 and an operation early warning module 46, the memory 41 and the processor 42 store and execute a liquid-cooled cable power supply safety state monitoring and early warning method program;

[0080] The data sensing module 43 uses data sensing to obtain the operating parameters of the liquid cooling cable and performs preprocessing, and extracts features according to preset feature variables to construct a model input data set;

[0081] The temperature field prediction module 44 uses Gaussian regression to construct a temperature prediction model for the liquid cooling cable, and predicts the temperature field data of the liquid cooling cable after a preset time based on the input data set through the temperature prediction model;

[0082] The power supply status diagnosis module 45 mines the hidden features of the temperature field data based on the state change law of the liquid cooling cable, trains the mapping relationship between the hidden features and different working conditions, and diagnoses the abnormal state through the predicted temperature field data;

[0083] The operation warning module 46 generates a safety warning according to the diagnosis and classification results of the working condition of the liquid-cooled cable, and sends it to the operation and maintenance personnel according to a preset method.

[0084] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0085] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0086] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0087] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0088] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0089] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for monitoring and early warning the power supply safety status of a liquid-cooled cable, characterized in that: The following steps are involved: Obtain historical operating parameters of the liquid-cooled cable and corresponding temperature field measurement values, select characteristic variables according to the temperature field changes, and construct a data set based on the characteristic variables; Gaussian regression is used to build a temperature prediction model for liquid-cooled cables. Gaussian regression training is performed through the data set to obtain historical operating parameters of different working conditions. The temperature field data under different working conditions are read through the trained temperature prediction model to analyze and obtain the state change law of the liquid-cooled cables. Constructing a diagnostic model for the power supply status of a liquid-cooled cable, mining the hidden features of the temperature field data according to the state change law, and training the mapping relationship between the hidden features and different working conditions; Using data perception to obtain the current operating parameters of the liquid cooling cable and perform preprocessing, using the temperature prediction model and diagnosis model to perform temperature field data prediction and power supply status diagnosis classification, and generating a safety warning based on the diagnosis classification results; A diagnostic model for the power supply status of the liquid-cooled cable is constructed, and the hidden features of the temperature field data are mined according to the state change law, and the mapping relationship between the hidden features and different working conditions is trained, specifically: A double hidden layer extreme learning machine was constructed, the state change law of the liquid cooling cable was imported, the input weights and biases of the first hidden layer and the second hidden layer were randomly generated, and the output weight matrix of the double hidden layer extreme learning machine was obtained through training. By connecting several double hidden layer extreme learning machines in series, the temperature field data corresponding to different working conditions are encoded layer by layer, and the input of each hidden layer node is the transpose of the output weight matrix of the previous hidden layer node, so as to mine the deep hidden layer features of the temperature field data; Obtain the hidden layer features of the temperature field data set corresponding to each working condition label, perform a graph representation based on the association between the hidden layer features and the working condition label and the association information between different working condition labels, and use a graph attention network to generate a graph structure of a normal working condition and various types of abnormal working conditions; Graph convolution is used to obtain the embedded representation of the working condition state node and the hidden layer feature node, and the attention mechanism is introduced to obtain the attention weights of the neighbor nodes in the graph structure. The attention weights are used to perform weighted aggregation of the neighbor nodes to update the embedded representation of the working condition state node and the hidden layer feature node. The updated embedded representations of the working condition state node and the hidden layer feature node are average pooled respectively to obtain the final embedded representations of the working condition state node and the hidden layer feature node on the associated information; The CTR prediction model is used to match the working state nodes and the hidden layer feature nodes, the mapping relationship between the hidden layer features and different working states is trained, and a diagnostic model for the power supply state of the liquid cooling cable is constructed.

2. The method for monitoring and early warning of power supply safety status of a liquid-cooled cable according to claim 1, characterized in that: Characteristic variables are selected according to the temperature field changes, and a data set is constructed based on the characteristic variables, specifically: The historical operating parameters of the liquid-cooled cable under different working conditions are obtained and preprocessed, the original operating parameter sequence is converted into an original sample set using a sliding window method, and the original sample set is decomposed and reconstructed using a wavelet packet to generate a number of reconstructed sample signals; Calculating the spread entropy of the reconstructed sample signal to characterize the volatility of different reconstructed sample signals, and selecting the reconstructed sample signal not less than a preset spread entropy threshold to generate a preliminary sample set; The temperature field change of the liquid cooling cable is taken as the dependent variable, and the Pearson correlation coefficient is used to calculate the correlation between each reconstructed sample signal and the dependent variable in the preliminary sample set, and the reconstructed sample signal with the largest correlation is selected; The mRMR algorithm is improved based on the Pearson correlation coefficient. The remaining reconstructed sample signals in the preliminary sample set are introduced one by one according to the selected reconstructed sample signals. The feature correlation is sorted according to the principle of maximum correlation between the reconstructed sample signals and the dependent variable and minimum redundancy between the reconstructed sample signals. Using SVM as the prediction model, verifying by adopting S-fold cross-validation strategy according to the feature correlation ranking, training in turn to obtain S prediction models, ranking the classification prediction accuracy as the importance, and selecting a preset number of reconstructed sample signals as feature variables; The characteristic parameters of the historical operating parameters are selected according to the characteristic variables, and a data set is constructed by combining the selected characteristic parameters with the temperature change of the liquid-cooled cable.

3. The method for monitoring and early warning of power supply safety status of a liquid-cooled cable according to claim 1, characterized in that: Obtain the historical operating parameters of different working conditions, read the temperature field data of different working conditions through the trained temperature prediction model, and analyze the state change law of the liquid cooling cable, specifically: The data set is divided into a training set and a test set for Gaussian regression training, a covariance function is selected as a kernel function, Bayesian optimization is used as an optimizer for Gaussian regression, Gaussian process regression is performed through the training set and the test set, and a trained temperature prediction model is output when the error meets the requirements; The historical operating parameters of different working conditions are clustered and analyzed, and the Euclidean distance is used as the similarity measurement function. The working condition labels are divided and set according to the cluster similarity calculation to generate the historical operating parameter subsets corresponding to each working condition label. Different subsets of historical operating parameters are encoded using an autoencoder to obtain potential feature distributions, and features are reconstructed according to the potential feature distributions. An adversarial mechanism is introduced to generate potential representations consistent with each historical operating parameter by adding an identifier and mapping them to corresponding positions, and to penalize reconstructed data that deviates too much from the potential feature distributions. The reconstructed data is used to enhance the original historical operating parameter subset, and the enhanced historical operating parameter subset is used as a model input of a temperature prediction model to predict corresponding temperature field data.

4. The method for monitoring and early warning of power supply safety status of a liquid-cooled cable according to claim 1, characterized in that: The state change rules of the liquid cooling cable are analyzed and obtained, specifically: Obtain a temperature field data set corresponding to a subset of historical operating parameters enhanced by different operating state tags, and read a liquid inlet temperature interval and a liquid outlet temperature interval in a liquid cooling cable from the temperature field data set; The operating parameter deviations between different historical operating parameter subsets are read, and the time series change characteristics of the liquid inlet temperature range and the liquid outlet temperature range of different working condition status labels are analyzed. The state change law of the liquid cooling cable is generated according to the operating parameter deviations and time series change characteristics.

5. The method for monitoring and early warning of power supply safety status of a liquid-cooled cable according to claim 1, characterized in that: The current operating parameters of the liquid cooling cable are obtained and preprocessed by using data perception, and the temperature field data is predicted and the power supply status diagnosis and classification are performed by using the temperature prediction model and the diagnosis model, specifically: Real-time perception of the operating parameters of the liquid-cooled cable, importing the current operating parameters into the temperature prediction model to obtain the temperature field prediction data of the liquid-cooled cable after a preset time; The current temperature field data and the temperature field prediction data are imported into the diagnosis model, the corresponding hidden layer features are read through the diagnosis model, the obtained hidden layer features are represented by graphs to obtain corresponding embedded representations, and the embedded representations of the hidden layer features are learned and updated and then matched using the CTR prediction model; The matching degree between the current hidden layer features and the working condition is obtained by inner product calculation, and the corresponding working condition label is generated. If it is an abnormal working condition label, warning information is generated according to the label category.

6. A liquid-cooled cable power supply safety status monitoring and early warning system, characterized in that: Implementing the power supply safety status monitoring and early warning method of the liquid-cooled cable as described in any one of claims 1 to 5, the system includes a memory, a processor, a data perception module, a temperature field prediction module, a power supply status diagnosis module and an operation early warning module, the memory and the processor store and execute a power supply safety status monitoring and early warning method program for the liquid-cooled cable; The data sensing module uses data sensing to obtain the operating parameters of the liquid cooling cable and performs preprocessing, and extracts features according to preset feature variables to construct a model input data set; The temperature field prediction module uses Gaussian regression to build a temperature prediction model for the liquid cooling cable, and uses the temperature prediction model to predict the temperature field data of the liquid cooling cable after a preset time based on the input data set; The power supply status diagnosis module mines the hidden features of the temperature field data based on the state change law of the liquid cooling cable, trains the mapping relationship between the hidden features and different working conditions, and diagnoses the abnormal state through the predicted temperature field data; The operation warning module generates a safety warning according to the diagnosis and classification results of the working condition of the liquid-cooled cable, and sends it to the operation and maintenance personnel according to a preset method.

Images