Condition monitoring methods, devices, systems, equipment, and media for cold shrink joints.

Abstract

This application discloses a method, apparatus, system, equipment, and medium for condition monitoring of cold-shrinkable cable joints. The method includes: acquiring multimodal time-series data of the target cold-shrinkable cable joint; preprocessing the multimodal time-series data to obtain a multimodal dataset; inputting the multimodal dataset into a pre-acquired hybrid model to obtain multimodal prediction data; the hybrid model consists of an LSTM model and a Transformer model, with the output of the LSTM model serving as the input to the Transformer model; and generating early warning information when the predicted temperature data, vibration data, or humidity data in the multimodal prediction data exceeds a preset acceptable temperature threshold, a preset acceptable vibration threshold, or a preset acceptable humidity threshold. This application enables timely detection and early warning of potential hazards before cable joint failures occur.

Description

Technical Field

[0001] This application relates to the field of cable joint technology, and in particular to a method, apparatus, system, equipment and medium for monitoring the condition of a cold shrink joint. Background Technology

[0002] With the continuous expansion of the power grid, the increasing proportion of renewable energy generation, and the rising demands of users for power quality, the drawbacks of traditional cold-shrink cable joints are becoming increasingly apparent. Traditional cold-shrink cable joints lack real-time operational status monitoring and alarm functions. This lack of functionality makes it difficult for maintenance personnel to detect potential hazards before a fault occurs, and makes it impossible to accurately locate the fault after it occurs. This results in longer locating and repair times after a cold-shrink cable joint failure, causing more serious impacts on power grid operation. Summary of the Invention

[0003] This application aims to provide a method, device, system, equipment, and medium for monitoring the condition of cold-shrink intermediate joints, which can detect potential hazards and provide early warnings in a timely manner before cable joint failures occur.

[0004] According to the first aspect of this application, the cold shrink joint includes a cold shrink joint body, a bolt-type connecting tube, a temperature measuring module, a vibration sensor, and a water immersion sensor. The cold shrink joint body is tightly fitted onto the surface of the cable's insulation layer. The bolt-type connecting tube is located inside the cold shrink joint body and is used to fix and connect the cable core. The bolt-type connecting tube has a slot. The temperature measuring module is embedded in the slot and is used to collect temperature data. The vibration sensor is installed at the end of the cold shrink joint body and is used to collect vibration data. The water immersion sensor is installed on one side of the cold shrink joint body and is used to collect humidity data. The condition monitoring method for the cold shrink joint includes: Acquire multimodal time-series data of the target cold-shrink joint, wherein the multimodal time-series data includes at least the temperature data, vibration data, and humidity data of the target cold-shrink joint; The multimodal time-series data is preprocessed to obtain a multimodal dataset; The multimodal dataset is input into a pre-acquired hybrid model to obtain multimodal prediction data; the hybrid model consists of an LSTM model and a Transformer model, and the output of the LSTM model is used as the input of the Transformer model; the multimodal prediction data includes at least temperature prediction data, vibration prediction data, and humidity prediction data; If the temperature prediction data in the multimodal prediction data is greater than a preset temperature qualification threshold, or the vibration prediction data is greater than a preset vibration qualification threshold, or the humidity prediction data is greater than a preset humidity qualification threshold, an early warning message is generated. The hybrid model is obtained through the following steps: Acquire historical multimodal time-series data of the target cold-shrink joint; The historical multimodal time series data is preprocessed to obtain a historical multimodal dataset; A basic hybrid model is built based on LSTM and Transformer models; The base hybrid model is trained using the historical multimodal dataset to obtain the hybrid model.

[0005] According to some embodiments of this application, the step of inputting the multimodal dataset into a pre-acquired hybrid model to obtain multimodal prediction data includes: The multimodal dataset is input into the LSTM model for training, and local feature sequences are output. The local feature sequence is input into the Transformer model, and a global feature vector is output based on the self-attention mechanism. The multimodal prediction data is obtained based on the local feature sequence and the global feature vector.

[0006] According to some embodiments of this application, obtaining the multimodal prediction data based on the local feature sequence and the global feature vector includes: The local feature sequence and the global feature vector output at the last time step are concatenated to obtain the fused feature vector; The fused feature vector is reduced in dimensionality and mapped to obtain the multimodal prediction data.

[0007] According to some embodiments of this application, a physical constraint term based on a current-temperature relationship model is added to the loss function of the hybrid model; wherein, the current-temperature relationship model is used to characterize the relationship between the current data flowing through the cold shrink joint and the temperature data of the cold shrink joint.

[0008] According to some embodiments of this application, the preset temperature qualification threshold, the preset vibration qualification threshold, and the preset humidity qualification threshold are obtained by the following steps: The simulated multimodal time-series data of the target cold shrink joint is obtained, and the simulated multimodal time-series data is the data collected during the process of simulating abnormal failures through manual intervention; By analyzing the historical multimodal time-series data, the simulated multimodal time-series data, and the national standard data for cable accessory operation, the preset temperature qualification threshold, the preset vibration qualification threshold, and the preset humidity qualification threshold are determined.

[0009] According to some embodiments of this application, the preprocessing of the multimodal time-series data to obtain a multimodal dataset includes: Missing values ​​in the multimodal time series data are filled in using linear interpolation; Outliers in the multimodal time series data are identified and replaced using the 3σ criterion; The multimodal time series data with different dimensions are normalized to obtain the multimodal dataset.

[0010] A condition monitoring device for a cold-shrink intermediate joint according to a second aspect embodiment of this application includes: The acquisition module is used to acquire multimodal time-series data of the target cold-shrink joint, wherein the multimodal time-series data includes at least the temperature data, vibration data and humidity data of the target cold-shrink joint; The preprocessing module is used to preprocess the multimodal time series data to obtain a multimodal dataset; The prediction module is used to input the multimodal dataset into a pre-acquired hybrid model to obtain multimodal prediction data; the hybrid model consists of an LSTM model and a Transformer model, and the output of the LSTM model will be used as the input of the Transformer model; the multimodal prediction data includes at least temperature prediction data, vibration prediction data, and humidity prediction data; The early warning module is used to generate early warning information when the temperature prediction data in the multimodal prediction data is greater than the temperature qualification threshold, or the vibration prediction data is greater than the vibration qualification threshold, or the humidity prediction data is greater than the humidity qualification threshold.

[0011] A cold shrink joint status monitoring system according to a third aspect of this application includes a joint status monitoring module, a communication module, and a host computer. The joint status monitoring module is used to execute the cold shrink joint status monitoring method as described in any of the first aspect embodiments above.

[0012] An electronic device according to a fourth aspect of this application includes a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the cold shrink joint status monitoring method as described in any of the first aspect embodiments above.

[0013] A computer-readable storage medium according to a fifth aspect embodiment of the present application stores computer-executable instructions for performing the cold shrink joint status monitoring method as described in the first aspect embodiment above.

[0014] In this embodiment, the cable cold shrink joint is equipped with a temperature measurement module, a vibration sensor, and a water immersion sensor to achieve real-time acquisition of temperature, vibration, and humidity data at the cable joint. The joint status monitoring module then executes the cold shrink joint status monitoring method. Through a hybrid model (LSTM+Transformer), multimodal data analysis and monitoring functions are realized to analyze the operating status information of the cable joint and monitor the conductor temperature, vibration, and water immersion of the cable joint. This enables timely detection of potential hazards and early warning before a fault occurs. In addition, the abnormal data acquisition source can help to accurately locate the fault.

[0015] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing this application. Attached Figure Description

[0016] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a schematic diagram of the structure of the cold shrink joint and the condition monitoring system of the cold shrink joint of this application; Figure 2 This is a flowchart illustrating an embodiment of the condition monitoring method for cold shrink joints of this application; Figure 3 This is a schematic diagram of an embodiment of the condition monitoring device for the cold shrink joint of this application; Figure 4 This is a schematic diagram of the hardware structure of an embodiment of the electronic device of this application. Detailed Implementation

[0017] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0018] In the description of this application, the use of terms such as "first," "second," etc., is for the purpose of distinguishing technical features only and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0019] In the description of this application, it should be understood that the orientation descriptions, such as up, down, etc., are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0020] In the description of this application, it should be noted that, unless otherwise explicitly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0021] The technical solution of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are some embodiments of this application, not all embodiments.

[0022] Figure 1 This is a schematic diagram of the structure of the cold shrink joint and the condition monitoring system of the cold shrink joint of this application; Figure 2 This is a flowchart illustrating an embodiment of the condition monitoring method for cold shrink joints of this application; Figure 3 This is a schematic diagram of an embodiment of the condition monitoring device for the cold shrink joint of this application; Figure 4 This is a schematic diagram of the hardware structure of an embodiment of the electronic device of this application.

[0023] See below. Figure 1 and Figure 2 The embodiments of this application are further described below. This application proposes a method for monitoring the condition of a cold-shrink joint. This method is applied to the condition monitoring of a cold-shrink joint, which includes a cold-shrink joint body 1, a bolt-type connecting tube 2, a temperature measuring module 3, a vibration sensor 4, and a water immersion sensor 5. The cold-shrink joint body 1 is tightly fitted to the surface of the cable insulation layer. The bolt-type connecting tube 2 is located inside the cold-shrink joint body 1 and is used to fix and connect the cable core; the bolt-type connecting tube 2 has a slot. The temperature measuring module 3 is embedded in the slot and is used to collect temperature data. The vibration sensor 4 is installed at the end of the cold-shrink joint body 1 and is used to collect vibration data. The water immersion sensor 5 is installed on one side of the cold-shrink joint body 1 and is used to collect humidity data. The condition monitoring method for this cold shrink joint includes the following steps: Step 101: Obtain multimodal time-series data of the target cold-shrink joint. The multimodal time-series data shall include at least the temperature data, vibration data, and humidity data of the target cold-shrink joint. Step 102: Preprocess the multimodal time series data to obtain a multimodal dataset; Step 103: Input the multimodal dataset into the pre-acquired hybrid model to obtain multimodal prediction data; the hybrid model consists of an LSTM model and a Transformer model, and the output of the LSTM model will be used as the input of the Transformer model; the multimodal prediction data includes at least temperature prediction data, vibration prediction data, and humidity prediction data; Step 104: If the temperature prediction data in the multimodal prediction data is greater than the preset temperature qualification threshold, or the vibration prediction data is greater than the preset vibration qualification threshold, or the humidity prediction data is greater than the preset humidity qualification threshold, generate an early warning message. The hybrid model is obtained through the following steps: Acquire historical multimodal time-series data of the target cold-shrink joint; Historical multimodal time series data are preprocessed to obtain a historical multimodal dataset; A basic hybrid model is built based on LSTM and Transformer models; The base mixture model is trained using a historical multimodal dataset to obtain the mixture model.

[0024] In this embodiment, the cable cold shrink joint is equipped with a temperature measurement module 3, a vibration sensor 4, and a water immersion sensor 5, which can realize real-time acquisition of temperature, vibration, and humidity data at the cable joint. During the execution of the cold shrink joint status monitoring method, a hybrid model (LSTM+Transformer) is used to realize multimodal data analysis and monitoring functions, analyze the operating status information of the cable joint, and monitor the conductor temperature, vibration, and water immersion of the cable joint. This enables timely detection of potential hazards and early warning before a fault occurs. In addition, the abnormal data acquisition source can also help to accurately locate the fault.

[0025] To facilitate understanding of the condition monitoring method for the cold shrink joint of this application, the cold shrink joint proposed in this application will be described first.

[0026] like Figure 1 As shown, the cold shrink joint includes a cold shrink joint body 1, a bolt-type connecting tube 2, a temperature measuring module 3, a vibration sensor 4, and a water immersion sensor 5. It also includes a first sheath 6 wrapped around the cold shrink joint body 1, a second sheath 7 wrapped around the first sheath 6, and a CT induction current-taking ring 8 fitted onto the outside of the cable. The first sheath 6 can be composed of a waterproof tape, a semi-conductive tape, and a copper mesh layer, while the second sheath 7 can be composed of PVC adhesive tape and a waterproof tape.

[0027] Specifically, the cable core is connected and fixed by a bolt-type connecting tube 2. The bolt-type connecting tube 2 can be made of aluminum alloy. The cable conductor is connected by unscrewing the bolt. The wall thickness of the bolt-type connecting tube 2 is larger than that of the conventional connecting tube, providing sufficient slotting space. At the same time, the bolt-type connecting tube 2 will not undergo significant deformation during installation, which can greatly reduce the probability of damage to the embedded sensor.

[0028] The temperature measurement module 3 is pre-embedded in a groove outside the bolt-type connecting pipe 2 at the factory. This groove is the temperature measurement chip groove and is cured with epoxy resin. It can realize real-time monitoring of the temperature of the connecting pipe and can detect abnormal heating at the connection point caused by overload or excessive contact resistance of the connecting pipe in advance, so as to provide timely warning and avoid the escalation of line faults. It can be understood that the pre-embedded temperature measurement module 3 is a wireless temperature measurement module.

[0029] The cold shrink joint body 1 is tightly wrapped around the surface of the cable insulation layer, which can effectively optimize the insulation shielding break, control electrical stress, and ensure insulation performance. The waterproof strip, semi-conductive strip, and copper mesh layer of the first protective layer 6 can serve to waterproof and restore the copper shielding layer. The vibration sensor 4 is installed at the end of the connector body using PVC tape and waterproof tape. The connecting wire is wrapped inside the tape and then covered with a layer of sealant to reduce the impact of the connecting wire on the wrapping performance and ensure that the connecting wire is always parallel to the connector body. The vibration sensor 4 is used to monitor the status of the cable and connector. When a connector malfunctions or vibration is caused by human error, it can promptly locate the connector to prevent damage to the power system and assist maintenance personnel in quickly troubleshooting. It is understood that the vibration sensor 4 can be a wired vibration sensor.

[0030] The water immersion sensor 5 is installed on one side of the connector using PVC tape and waterproof tape. The connecting wire is wrapped inside the tape and then sealed with a layer of sealant to reduce the impact of the connecting wire on the wrapping performance and ensure that the connecting wire is always parallel to the connector body. The water immersion sensor 5 is used to monitor whether water has entered the cable connector. Based on the water immersion status information, it promptly notifies maintenance personnel to drain the water and prevent the cable connector from being submerged in water for a long time, which could reduce the insulation performance of the cable connector and lead to cable connector failure. It is understood that the water immersion sensor 5 can be a wired water immersion sensor.

[0031] The PVC adhesive tape and waterproof tape of the second protective layer 7 are used to fix the cable joints and also enhance the waterproof function. The connecting wire of the CT induction power-taking ring 8 is wrapped inside the tape and covered with a layer of sealant to reduce the impact of the connecting wire on the wrapping performance, while ensuring that the connecting wire is always parallel to the connector body. A small CT induction power-taking ring 8 is added to the metal sheath of the cable to draw power using the principle of electromagnetic induction to power various sensors and other electrical equipment. The above-mentioned acquisition of multimodal time-series data of the target cold shrink joint includes at least temperature data, vibration data and humidity data of the target cold shrink joint. The temperature data is acquired through temperature measurement module 3, the vibration data is acquired through vibration sensor 4, and the humidity data is acquired through water immersion sensor 5.

[0032] Temperature module 3 collects the conductor temperature of the cold-shrink joint in real time, with a collection frequency that can be set to 1-5 minutes / time, which can be adjusted according to load fluctuations; vibration sensor 4 monitors the vibration acceleration of the joint caused by electrodynamic and mechanical loosening, including the X, Y, and Z axes, capturing instantaneous vibration peaks and continuous vibration characteristics; water immersion sensor 5 detects the humidity of the joint installation environment, with a range of 0-100%RH. The data collected by each sensor is time-series data including time steps and numerical values.

[0033] For the multimodal dataset input to the hybrid model, the pre-processed data from the temperature measurement module 3, vibration sensor 4, and water immersion sensor 5 are organized into 3D tensor numerical sequence data according to [Batch_size, Seq_len, Input_size], i.e., [number of samples, number of time steps, feature dimension], and then transmitted to the model for data processing. Since the temperature measurement module 3 measures the actual conductor temperature as a continuous floating-point number, the vibration sensor 4 measures the X, Y, and Z axis vibration acceleration as continuous floating-point numbers, and the water immersion sensor 5 measures the ambient relative humidity as a continuous floating-point number (0%RH~100%RH), there are a total of five indicators, hence Input_size=5. The number of time steps (Seq_len) refers to the "length of the historical data time window" that makes up a single input sample; that is, the model extracts and predicts features using "sensor data from the past N consecutive time points," and the value of N is determined by the acquisition frequency and business requirements.

[0034] Specifically, assuming a sampling frequency of 5 minutes per sample, a time window of the past 2 hours (Seq_len=24), a batch size of 32 samples (Batch_size=32), and 5 feature dimensions (Input_size=5), and normalization processing, the batch processing involves simultaneously processing 32 samples from different endpoints or time windows, ultimately forming a 3D tensor with dimensions [32, 24, 5]. Here, Batch_size=32 represents 32 independent input samples; Seq_len=24 represents that each sample contains data from 24 consecutive time points; and Input_size=5 represents that each time point contains data from 5 feature dimensions. This [32, 24, 8] 3D tensor is directly input into the input layer of the LSTM model in the hybrid model. The LSTM processes these 24 sequence data step-by-step through a gating mechanism, capturing the local temporal dynamic dependencies of indicators such as temperature and vibration.

[0035] Understandably, in the process of training the hybrid model, a historical multimodal dataset obtained from historical multimodal time series data is used. The historical multimodal time series data can be the operational data of the monitored object over the past 1 to 3 years as historical data. During the training process, the historical multimodal dataset can be divided into training set, validation set and test set according to a certain ratio, such as a 7:2:1 ratio. The training set is fed into the hybrid network for training, and the validation set guides the training process, while the test set is used to test the performance of the model.

[0036] The hybrid model can output temperature prediction data, vibration prediction data, and humidity prediction data based on the input temperature data, vibration data, and humidity data of the target cold shrink joint. Since sensor data may be affected by signal interference during transmission or by factors affecting the sensor data acquisition process, resulting in inaccurate data, by inputting time-series data collected by the sensor over a period of time into the hybrid model, more accurate real-time prediction data can be output. In some cases, the hybrid model can also be trained to output prediction data after a certain period of time.

[0037] If the temperature prediction data in the multimodal prediction data exceeds the preset temperature qualification threshold, or the vibration prediction data exceeds the preset vibration qualification threshold, or the humidity prediction data exceeds the preset humidity qualification threshold, an early warning message will be generated. If any one of the temperature, vibration, or water immersion prediction data is unqualified, then the entire multimodal prediction data sample is considered unqualified. An early warning message to be sent can be generated based on the unqualified data, while normal data will not be sent.

[0038] In some implementations, the multimodal dataset is input into a pre-acquired mixture model to obtain multimodal prediction data, including: The multimodal dataset is input into the LSTM model for training, and the output is a local feature sequence; The local feature sequence is input into the Transformer model, and a global feature vector is output based on the self-attention mechanism. Multimodal prediction data is obtained based on local feature sequences and global feature vectors.

[0039] In this embodiment, the prediction accuracy of the output multimodal prediction data is improved by using the LSTM model to process local dynamics and the Transformer model to extract global features.

[0040] LSTM local dynamic feature extraction can receive preprocessed time-series data sequences, i.e., with dimensions of [number of samples, time step, feature dimension], where the number of features includes temperature, vibration, and humidity. It captures local temporal dependencies through 3 layers of LSTM units. Each LSTM unit contains a forget gate, an input gate, and an output gate. The information flow is controlled by sigmoid and tanh activation functions. Finally, it outputs a local feature sequence, i.e., with dimensions of [number of samples, time step, hidden layer dimension], to handle dynamic changes in data over a short period of time.

[0041] Transformer global feature extraction encodes the local feature sequences output by LSTM, resulting in a feature matrix with location information. A multi-head self-attention layer calculates the association weights between each time step's features and all other time steps, extracting long-distance dependency features. A three-layer encoder module (self-attention + residual connection + Layer Normalization + fully connected layer) outputs a global feature vector, with dimensions [number of samples, hidden layer dimension], integrating key information from the entire time series.

[0042] In some implementations, multimodal prediction data is obtained based on local feature sequences and global feature vectors, including: The local feature sequence and global feature vector output at the last time step are concatenated to obtain the fused feature vector; The fusion feature vectors are reduced in dimensionality and mapped to obtain multimodal prediction data.

[0043] In this embodiment, the local feature sequence of the LSTM (note that the output of the last time step should be taken) is concatenated with the global feature vector of the Transformer to obtain the fused features. Then, the fused features are reduced in dimensionality and mapped through two fully connected layers to output the predicted values ​​of temperature, vibration and humidity.

[0044] In some implementations, the data obtained from the temperature measurement module 3, vibration sensor 4, and water immersion sensor 5 are arranged according to [Batch_size, Seq_len, Input_size], i.e. [number of samples, number of time steps, feature dimension], to form 3D tensor numerical sequence data, which is then transmitted to the LSTM model for data processing, including the following steps.

[0045] The input data undergoes a basic transformation to form three fundamental variables: cell state C. t-1 (Shape is [Batch_size, Hidden_units], where Hidden_unit is the number of hidden units, a hyperparameter, set to 48), hidden state h t-1 (Shape is [Batch_size, Hidden_units]), Input vector X at time t t(The shape is [Batch_size, Input_size]); There are two outputs: cell state C t Hidden state h t Where h t It is also the output at time t, with the same shape as described above.

[0046] The LSTM model consists of cell states and three gating units (forget gate, input gate, and output gate), with the cell state being the core. The first step in feeding data into the LSTM network model is to use the forget gate to retain information about the cell state from the previous time step.

[0047] f t =Sigmoid(W f ·[ h t-1 X t ] + b f ); Among them, W f Let b be the weight matrix of the forget gate. f Let be the bias vector of the forget gate. The second step in the LSTM network model is to determine what new information to add to the cell state, which is divided into two steps. First, h t-1 and X t The data is processed by the input gate to determine which information to update, and then h is used. t-1 and X t New candidate cell information is obtained through the Tanh activation function layer. In transferring old cell information C t-1 Updated to new cell information C t .

[0048] i t =Sigmoid(W i ·[ h t-1 , X t ] + b i ); =Tanh(W C ·[ h t-1 , X t ] + b C ); C t =f t C t-1 +i t ; Among them, W i W is the weight matrix of the input gate. Cb is the weight matrix for cell states; i Let b be the bias vector of the input gate. C The weight matrix represents the cell state; output i t , and C t .

[0049] The third step is to control the proportion of cell states transferred to hidden state via the output gate.

[0050] O t =Sigmoid(W O ·[X t , h t-1 ] + b O ) Among them, W O Let b be the weight matrix of the output gate. O This is the bias matrix for the output gate.

[0051] Simultaneously update the hidden state to obtain a vector of values ​​between -1 and 1: h t =O t ·Tanh (C t ); After processing through 3 LSTM model units, the hidden state h of all time steps of the LSTM is extracted. t The sequence X required to form the Transformer Transformer As input, it is fed into the Transformer network module. type. First, the output of the Transformer is preprocessed, and X is... Transformer Add location encoding PE: ; ; Where pos is the sequence position, ranging from (0, seq_len-1), seq_len is the sequence length, and i is the dimension index, ranging from (0, d). model / 2-1), d model The feature dimension of Transformer is defined as 256.

[0052] To obtain the final input X of the Transformer: X=X Transformer +PE; Where: the shape of X is [N, seq_len, d] model ([Number of samples, sequence length, Transformer feature dimension]).

[0053] Scaling dot product attention calculates the attention weights for query (Q), key (K), and value (V), assigning different weights to each word, using the following formula: ; in: These are the query, key, and value matrix, respectively. is the dimension of the key vector.

[0054] Then, the linear projection is computed in parallel across multiple heads, and the multiple heads are split: ; ; ; Among them, W Q W K W V These are the learnable weight matrices corresponding to the query, key, and value, respectively. Perform scaled dot product attention on each head and output the output. split And splice multiple heads to output. concat : ; Therefore, the final projection is: output=output concat ·W o ; Among them, W o This is the final projection weight matrix.

[0055] In the feedforward neural network, an independent nonlinear transformation is performed on the features at each location: FFN(x)=max(0,XW1+b1)W2+b2; Where W1 and W2 are weight matrices, and b1 and b2 are bias vectors.

[0056] Finally, layer normalization is performed to prevent gradient vanishing. ; Where μ is the mean of the feature dimension. Let be the variance, and γ and β be learnable parameters. =1e-5 =0.00001 The intelligent cold shrink joint requires monitoring three data points: temperature, vibration, and water immersion. The final prediction head performs predictions on a time-step basis and maps them to the three targets through a linear layer.

[0057] y pred_t =X enc_t ·W t +b t ; Among them, y pred_t Let X be the predicted output of the Transformer at time step t. enc_t For input, W t To predict the weight matrix, b t For predicting bias.

[0058] y of all time steps pred_t By concatenating the sequences according to time steps, the final predicted sequence is obtained: ; In some implementations, a physical constraint term based on a current-temperature relationship model is added to the loss function of the hybrid model; wherein, the current-temperature relationship model is used to characterize the relationship between the current data flowing through the cold shrink joint and the temperature data of the cold shrink joint.

[0059] In this embodiment, optimizing the loss function based on the current-temperature relationship improves the physical rationality of data monitoring. The current-temperature relationship model used here, i.e., the physical relationship between current and temperature, was obtained in previous laboratory experiments. It is used in the loss function of the Transformer network in the hybrid model. The Transformer itself has its own mean square error (MSE) loss function. Then, this physical constraint term of the current-temperature relationship is added to the basic loss function to force the model output to conform to the physical law of current-temperature. Specifically, the current-temperature relationship model is obtained by collecting joint temperature data under different load currents (0%~150% of rated current) in previous experiments and fitting the physical relationship model using the least squares method.

[0060] In some implementations, the rationality of the predicted temperature can be verified based on real-time current data and the temperature prediction value output by the hybrid model, combined with the offline fitted current-temperature relationship. According to standards such as GB / T12706, with the constraint that the predicted temperature does not exceed the maximum allowable temperature of the connector (e.g., 90℃), the maximum allowable current carrying capacity of the connector can be calculated, thus achieving dynamic capacity assessment. Specifically, since the relationship between current and temperature has already been obtained, the temperature data transmitted by the sensor is compared with the temperature data derived from the actual current. An allowable difference of ±5℃ can be set; if the difference is large, a certain warning will be issued.

[0061] In some implementations, the preset temperature acceptable threshold, preset vibration acceptable threshold, and preset humidity acceptable threshold are obtained by the following steps: The simulated multimodal time-series data of the target cold shrink joint were obtained. The simulated multimodal time-series data were collected during the process of simulating abnormal failures through manual intervention. By analyzing historical multimodal time-series data, simulated multimodal time-series data, and national standard data for cable accessory operation, preset temperature qualification thresholds, preset vibration qualification thresholds, and preset humidity qualification thresholds are determined.

[0062] In this embodiment, sufficient sample data covering the entire scenario is collected to support the threshold calculation between qualified and unqualified data.

[0063] Historical multimodal time series data, also known as historical operational data, can be used to collect normal operation data of the monitored object over the past 1 to 3 years as qualified samples, and data under fault / abnormal conditions as unqualified samples. It must include samples of different operating conditions, such as samples of diurnal temperature difference.

[0064] Simulated multimodal time-series data, also known as experimental simulation data, can simulate abnormal scenarios through artificial intervention, such as gradually increasing the core temperature to the standard requirement (90℃), increasing the vibration amplitude of intermediate joints in different directions to produce abnormal noises or affect operation, simulating different degrees of water immersion, and simultaneously recording data changes and critical state data from "qualified" to "unqualified".

[0065] The national standard data for cable accessory operation refers to the threshold calculation and confirmation based on the national standard requirements for cable accessory operation standards or specifications. This can be based on GB / T 12706.1-2022 "Extruded Insulated Power Cables and Accessories with Rated Voltages of 1kV (Um=1.2kV) to 35kV (Um=40.5kV) Part 1: Cables with Rated Voltages of 1kV (Um=1.2kV) and 3kV (Um=3.6kV)", where the maximum allowable core temperature for cables of different insulation grades is: PVC insulation: 70℃, cross-linked polyethylene (XLPE) insulation: 90℃. GB / T 13824-2023 "Methods for Measuring Vibration Intensity" specifies that the vibration intensity of electrical equipment is usually classified according to "ISO 2372", with slight vibration (≤0.28mm / s) considered safe and strong vibration (≥1.8mm / s) considered unqualified. GB / T 2900.10-2013 "Electrical Engineering Terminology - Cables" stipulates that the relative humidity of the operating environment for cable accessories should be ≤75%; the IP protection level should be determined according to requirements.

[0066] Simultaneously, by referencing historical operational data, experimental simulation data, and national standards, the thresholds for qualified and unqualified data are ultimately determined. The qualified threshold can be determined by selecting the minimum value among the critical state data from historical operational data and experimental simulation data, and the threshold under national standards. Alternatively, it can be input into the model for training, allowing the model to determine qualified and unqualified samples.

[0067] In some implementations, the multimodal time series data is preprocessed to obtain a multimodal dataset, including: Missing values ​​in multimodal time series data are filled using linear interpolation; Outliers in multimodal time series data are identified and replaced using the 3σ criterion. Multimodal time series data with different dimensions are normalized to obtain a multimodal dataset.

[0068] In this embodiment, missing values ​​are filled using linear interpolation, and outliers are identified using the 3σ criterion. These outliers can be replaced with the mean of adjacent normal data. Then, data of different dimensions are normalized and scaled to the range of [0, 1], thus completing the preprocessing of the original data.

[0069] The cold shrink joint status monitoring method provided in this application embodiment can be executed by a cold shrink joint status monitoring device 200. This application embodiment uses the cold shrink joint status monitoring device 200 executing the cold shrink joint status monitoring method as an example to illustrate the cold shrink joint status monitoring device 200 provided in this application embodiment.

[0070] Please see Figure 3 This is a structural schematic diagram of a condition monitoring device 200 for a cold-shrink intermediate joint provided in an embodiment of this application. Figure 3 As shown, the condition monitoring device 200 for the cold shrink joint includes: The acquisition module 201 is used to acquire multimodal time-series data of the target cold shrink joint. The multimodal time-series data includes at least the temperature data, vibration data and humidity data of the target cold shrink joint. Preprocessing module 202 is used to preprocess multimodal time series data to obtain multimodal dataset; The prediction module 203 is used to input the multimodal dataset into a pre-acquired hybrid model to obtain multimodal prediction data; the hybrid model consists of an LSTM model and a Transformer model, and the output of the LSTM model will be used as the input of the Transformer model; the multimodal prediction data includes at least temperature prediction data, vibration prediction data and humidity prediction data; The early warning module 204 is used to generate early warning information when the temperature prediction data in the multimodal prediction data is greater than the temperature qualified threshold, or the vibration prediction data is greater than the vibration qualified threshold, or the humidity prediction data is greater than the humidity qualified threshold.

[0071] In some implementations, the prediction module 203 can be used to: The multimodal dataset is input into the LSTM model for training, and the output is a local feature sequence; The local feature sequence is input into the Transformer model, and a global feature vector is output based on the self-attention mechanism. Multimodal prediction data is obtained based on local feature sequences and global feature vectors.

[0072] In some implementations, the prediction module 203 can be used to: The local feature sequence and global feature vector output at the last time step are concatenated to obtain the fused feature vector; The fusion feature vectors are reduced in dimensionality and mapped to obtain multimodal prediction data.

[0073] In some implementations, a physical constraint term based on a current-temperature relationship model is added to the loss function of the hybrid model; wherein, the current-temperature relationship model is used to characterize the relationship between the current data flowing through the cold shrink joint and the temperature data of the cold shrink joint.

[0074] In some implementations, the early warning module 204 can be used for: The simulated multimodal time-series data of the target cold shrink joint were obtained. The simulated multimodal time-series data were collected during the process of simulating abnormal failures through manual intervention. By analyzing historical multimodal time-series data, simulated multimodal time-series data, and national standard data for cable accessory operation, preset temperature qualification thresholds, preset vibration qualification thresholds, and preset humidity qualification thresholds are determined.

[0075] In some implementations, the preprocessing module 202 may be used for: Missing values ​​in multimodal time series data are filled using linear interpolation; Outliers in multimodal time series data are identified and replaced using the 3σ criterion. Multimodal time series data with different dimensions are normalized to obtain a multimodal dataset.

[0076] Since the cold shrink joint condition monitoring device 200 adopts all the technical solutions of the cold shrink joint condition monitoring method of the above embodiments, it has at least all the beneficial effects brought about by the technical solutions of the above embodiments, and will not be described again here.

[0077] like Figure 1 As shown, the cold shrink intermediate joint status monitoring system provided in this application embodiment includes a joint status monitoring module 9, a communication module 10, and a host computer 11. The joint status monitoring module 9 is used to execute the cold shrink intermediate joint status monitoring method as described above. The joint status monitoring module 9 includes the cold shrink intermediate joint status monitoring device 200 described above.

[0078] In this embodiment, the cable cold shrink joint is equipped with a temperature measurement module 3, a vibration sensor 4, and a water immersion sensor 5, which can realize real-time acquisition of temperature, vibration, and humidity data at the cable joint. Then, the joint status monitoring module 9 executes the cold shrink joint status monitoring method. Through a hybrid model (LSTM+Transformer), multimodal data analysis and monitoring functions are realized to analyze the operating status information of the cable joint and monitor the conductor temperature, vibration, and water immersion of the cable joint. This enables timely detection of potential hazards and early warning before a fault occurs. In addition, the abnormal data source can be used to help to accurately locate the fault.

[0079] The connector status monitoring module 9 receives signals from the temperature measurement module 3, vibration sensor 4, and water immersion sensor 5, performs data analysis and processing, and monitors the data. It draws power through the CT induction power ring 8. When the obtained multimodal prediction data is unqualified, it immediately generates an early warning message and transmits the abnormal unqualified data and early warning message to the host computer 11 through the communication module 10. The host computer 11 triggers an alarm, and platform personnel can quickly locate the fault point based on the signal transmission location. The temperature measurement module 3, vibration sensor 4, and water immersion sensor 5 can transmit the collected data to the joint status monitoring module 9 via LoRa transmission; the communication module 10 can include a signal generator, which can enhance the signal transmitted by the joint status monitoring module 9 to the host computer 11, and the host computer 11 can receive the transmitted signal of the joint status monitoring module 9 enhanced by the signal generator more stably.

[0080] The connector status monitoring module 9 features multimodal data analysis and monitoring capabilities. By processing multimodal signals using a deep learning network model, it enhances monitoring accuracy and reduces signal transmission pressure. It can monitor conductor temperature, vibration, and water immersion conditions of cable connectors, enabling data analysis and real-time monitoring of the connector's operational status. Through a hybrid model (LSTM + Transformer), in scenarios such as power grid load forecasting, LSTM handles local dynamics while Transformer extracts global features, improving prediction accuracy. Furthermore, by optimizing the loss function based on pre-trained current-temperature relationships, it achieves dynamic capacity assessment, enhancing the physical rationality of data monitoring. An edge computing endpoint is added, deploying lightweight AI algorithms in an edge computing box to clean data in real time, identify abnormal fluctuations, and upload only key early warning information to the cloud, reducing signal transmission pressure. In addition, it features an alarm function. By setting alarm thresholds, it can proactively send alarm information when the temperature exceeds the threshold, the connector is submerged in water, or strong vibrations occur, alerting maintenance personnel to the abnormal conditions of the connector and enabling further investigation and resolution of potential safety hazards.

[0081] Figure 4 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application.

[0082] This electronic device may include a processor 301 and a memory 302 storing computer program instructions.

[0083] Specifically, the processor 301 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0084] Memory 302 may include mass storage for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 302 may include removable or non-removable (or fixed) media. Where appropriate, memory 302 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 302 is non-volatile solid-state memory.

[0085] In some embodiments, memory 302 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Thus, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this disclosure.

[0086] The processor 301 reads and executes computer program instructions stored in the memory 302 to implement any of the cold shrink joint status monitoring methods in the above embodiments.

[0087] In one example, the electronic device may also include a communication interface 303 and a bus 310. For example, Figure 4 As shown, the processor 301, memory 302, and communication interface 303 are connected through bus 310 and complete communication with each other.

[0088] The communication interface 303 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0089] Bus 310 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 310 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0090] The electronic device can execute the cold shrink joint status monitoring method in the embodiments of this application, thereby achieving the combination Figure 2 and Figure 3 The described method and apparatus for monitoring the condition of cold shrink joints.

[0091] Furthermore, in conjunction with the cold shrink joint status monitoring method in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the cold shrink joint status monitoring methods in the above embodiments.

[0092] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0093] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0094] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0095] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0096] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for monitoring the condition of a cold-shrink joint, characterized in that, The cold-shrink joint includes a cold-shrink joint body, a bolt-type connecting tube, a temperature measuring module, a vibration sensor, and a water immersion sensor. The cold-shrink joint body is tightly fitted onto the surface of the cable's insulation layer. The bolt-type connecting tube is located inside the cold-shrink joint body and is used to fix and connect the cable core; the bolt-type connecting tube has a slot. The temperature measuring module is embedded in the slot and is used to collect temperature data. The vibration sensor is installed at the end of the cold-shrink joint body and is used to collect vibration data. The water immersion sensor is installed on one side of the cold-shrink joint body and is used to collect humidity data. The condition monitoring method for the cold shrink joint includes: Acquire multimodal time-series data of the target cold-shrink joint, wherein the multimodal time-series data includes at least the temperature data, vibration data, and humidity data of the target cold-shrink joint; The multimodal time-series data is preprocessed to obtain a multimodal dataset; The multimodal dataset is input into a pre-acquired hybrid model to obtain multimodal prediction data; the hybrid model consists of an LSTM model and a Transformer model, and the output of the LSTM model is used as the input of the Transformer model; the multimodal prediction data includes at least temperature prediction data, vibration prediction data, and humidity prediction data; If the temperature prediction data in the multimodal prediction data is greater than a preset temperature qualification threshold, or the vibration prediction data is greater than a preset vibration qualification threshold, or the humidity prediction data is greater than a preset humidity qualification threshold, an early warning message is generated. The hybrid model is obtained through the following steps: Acquire historical multimodal time-series data of the target cold-shrink joint; The historical multimodal time series data is preprocessed to obtain a historical multimodal dataset; A basic hybrid model is built based on LSTM and Transformer models; The base mixture model is trained using the historical multimodal dataset to obtain the mixture model; The step of inputting the multimodal dataset into a pre-acquired hybrid model to obtain multimodal prediction data includes: The multimodal dataset is input into the LSTM model for training, and local feature sequences are output. The local feature sequence is input into the Transformer model, and a global feature vector is output based on the self-attention mechanism. The multimodal prediction data is obtained based on the local feature sequence and the global feature vector; The step of obtaining the multimodal prediction data based on the local feature sequence and the global feature vector includes: The local feature sequence and the global feature vector output at the last time step are concatenated to obtain the fused feature vector; The fused feature vector is reduced in dimensionality and mapped to obtain the multimodal prediction data; The loss function of the hybrid model incorporates a physical constraint term based on a current-temperature relationship model; wherein the current-temperature relationship model is used to characterize the relationship between the current data flowing through the cold shrink joint and the temperature data of the cold shrink joint.

2. The condition monitoring method for cold-shrink intermediate joints according to claim 1, characterized in that, The preset temperature qualification threshold, the preset vibration qualification threshold, and the preset humidity qualification threshold are obtained by the following steps: The simulated multimodal time-series data of the target cold shrink joint is obtained, and the simulated multimodal time-series data is the data collected during the process of simulating abnormal failures through manual intervention; By analyzing the historical multimodal time-series data, the simulated multimodal time-series data, and the national standard data for cable accessory operation, the preset temperature qualification threshold, the preset vibration qualification threshold, and the preset humidity qualification threshold are determined.

3. The condition monitoring method for cold-shrink intermediate joints according to claim 1, characterized in that, The preprocessing of the multimodal time-series data to obtain a multimodal dataset includes: Missing values ​​in the multimodal time series data are filled in using linear interpolation; Outliers in the multimodal time series data are identified and replaced using the 3σ criterion; The multimodal time series data with different dimensions are normalized to obtain the multimodal dataset.

4. A condition monitoring device for a cold-shrink joint, characterized in that, The condition monitoring method for cold shrink joints as described in any one of claims 1 to 3 includes: The acquisition module is used to acquire multimodal time-series data of the target cold-shrink joint, wherein the multimodal time-series data includes at least the temperature data, vibration data and humidity data of the target cold-shrink joint; The preprocessing module is used to preprocess the multimodal time series data to obtain a multimodal dataset; The prediction module is used to input the multimodal dataset into a pre-acquired hybrid model to obtain multimodal prediction data; the hybrid model consists of an LSTM model and a Transformer model, and the output of the LSTM model will be used as the input of the Transformer model; the multimodal prediction data includes at least temperature prediction data, vibration prediction data, and humidity prediction data; The early warning module is used to generate early warning information when the temperature prediction data in the multimodal prediction data is greater than the temperature qualification threshold, or the vibration prediction data is greater than the vibration qualification threshold, or the humidity prediction data is greater than the humidity qualification threshold.

5. A condition monitoring system for a cold-shrink joint, characterized in that, It includes a joint status monitoring module, a communication module, and a host computer. The joint status monitoring module is used to execute the status monitoring method for cold shrink intermediate joints as described in any one of claims 1 to 3.

6. An electronic device, characterized in that, It includes a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions being executed by the processor to implement the steps of the cold shrink joint status monitoring method as described in any one of claims 1 to 3.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for causing a computer to perform the condition monitoring method for a cold-shrink joint as described in any one of claims 1 to 3.

Images