A multi-parameter fusion-based non-invasive evaluation method and system for the repair effect of a high-voltage cable buffer layer

Abstract

The present application belongs to the technical field of high-voltage cable buffer layer repair, and relates to a non-invasive evaluation method and system for the repair effect of the high-voltage cable buffer layer based on multi-parameter fusion. The method can solve the problems of the existing evaluation methods for the repair effect of the high-voltage cable buffer layer, such as the inability to evaluate in real time online, the single evaluation index, and the lack of dynamic coupling modeling. By constructing a full-process mathematical model of the curing dynamics of the repair material, the three-dimensional pore structure of the buffer layer, and the injection seepage of the repair liquid, and combining the PINN algorithm with physical constraints, the repair effect evaluation of the live cable is realized without external excitation, and the evaluation accuracy and field applicability are improved.

Description

Technical Field

[0001] This invention belongs to the field of high-voltage cable buffer layer repair technology, specifically relating to a non-invasive evaluation method and system for the repair effect of high-voltage cable buffer layers based on multi-parameter fusion. Background Technology

[0002] Frequent ablation failures occur in the buffer layer of high-voltage cross-linked polyethylene cables. The root cause is an abnormally increased volume resistivity of the buffer layer, leading to electric field distortion and breakdown. Although conductive repair fluid injection technology can repair these faults, existing methods for evaluating repair effectiveness still have the following shortcomings: (1) Applying pulse voltage or measuring capacitive current requires power outage and connection to a dedicated detection device, which causes power outage and operational safety risks, and is not applicable to online real-time evaluation of live cables; (2) The evaluation index is singular, relying only on single electrical parameters such as the rate of change of capacitive current, surface resistivity or partial discharge amplitude. It does not consider the coupling effect of the curing characteristics of the repair material, the pore structure of the buffer layer and the parameters of the injection process, and ignores the influence of the whole process from liquid penetration to curing and molding, which leads to limited evaluation accuracy and easy misjudgment. (3) Lacking dynamic coupling modeling, it can only perform static detection after the repair is completed. It cannot predict the repair effect through process parameters, and it does not combine the curing kinetic law of the repair material, making it difficult to quantify the impact of the curing degree on the final performance of the buffer layer.

[0003] In view of this, it is very necessary to provide a non-invasive evaluation method and system for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion to solve the above-mentioned defects in the prior art. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing methods for evaluating the repair effect of high-voltage cable buffer layers, such as the inability to conduct online real-time evaluation due to reliance on external stimuli, the limited evaluation accuracy due to the single evaluation index, and the difficulty in quantifying the impact of curing degree due to the lack of dynamic coupling modeling. This invention provides a non-invasive evaluation method and system for the repair effect of high-voltage cable buffer layers based on multi-parameter fusion to solve the aforementioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A non-invasive evaluation method for the repair effect of high-voltage cable buffer layers based on multi-parameter fusion includes the following steps: Step S1: Collect the characteristics of the repair material, the pore structure of the buffer layer, and the parameters of the repair fluid injection process; substitute the collected parameters into the prior knowledge base constructed based on the curing kinetics and seepage mechanism to obtain the model training input parameters; Step S2: Construct the PINN model and a hybrid loss function based on the physical constraints of the injection process, and train the PINN model using measured data of cable repair until the preset accuracy standard is met, thus obtaining a well-trained repair effect evaluation model. Step S3: Input the real-time collected data of the cable to be evaluated into the trained repair effect evaluation model, output the predicted value of the repair effect evaluation index, and perform grade determination of the repair effect of the high-voltage cable buffer layer and optimize the injection process and curing parameters of the PINN model based on the predicted value.

[0006] Preferably, step S1 specifically includes: Step S11: Collect data on the characteristics of the repair material, the pore structure of the buffer layer, and the parameters of the repair fluid injection process; Step S12: Construct a prior knowledge base that can describe the entire repair process based on solidification kinetics and seepage mechanism; Step S13: Substitute the collected parameters into the prior knowledge base to obtain the inference parameters, and combine the inference parameters with some of the collected parameters to obtain the model training input parameters.

[0007] Preferably, step S11, which involves collecting the characteristics of the repair material, the pore structure of the buffer layer, and the parameters of the repair fluid injection process, specifically includes: Step S111: Select a conductive repair material. The repair material adopts a silicone-based, epoxy-based, or polyurethane-based conductive repair system, and collect the characteristic parameters of the repair material. Step S112: Use X-ray computed tomography (X-CT) technology to obtain a three-dimensional pore structure image of the buffer layer of the high-voltage cable to be repaired; process the three-dimensional pore structure image through image segmentation and morphology to extract the pore structure parameters of the buffer layer. Step S113: Real-time acquisition of repair fluid injection process parameters using high-precision sensors; The repair material characteristic parameters in step S111 include: the initial viscosity of the material. Activation energy of curing reaction Frequency factor Reaction order ambient temperature , curing degree Material density ; The pore structure parameters of the buffer layer in step S112 include: porosity. Average aperture Maximum aperture Pore ​​tortuosity Pore ​​connectivity ; The parameters for the repair fluid injection process in step S113 include: injection pressure difference. Instantaneous injection of traffic Injection time seepage cross-sectional area seepage length .

[0008] Preferably, the prior knowledge base describing the entire repair process in step S12 includes: A coupled model of the viscosity of repair materials as a function of temperature and degree of curing, a fully coupled curing kinetic equation describing the curing process of repair materials, an equivalent permeability model applicable to porous media with buffer layers, and a dynamic seepage equation that couples the curing process with seepage behavior.

[0009] Preferably, step S12 specifically includes: Step S121: Based on the collected characteristic parameters of the repair material, the Arrhenius equation and the autocatalytic curing kinetics theory are used to construct a coupled model of the viscosity of the repair material as a function of temperature and degree of curing, and a fully coupled curing kinetic equation describing the curing process of the repair material. Step S122: Based on the obtained pore structure parameters of the buffer layer, the Kozeny-Carman equation is used to construct an equivalent permeability model suitable for porous media in the buffer layer, which is used to quantify the seepage capacity of the remediation fluid in the pores. Step S123: Based on Darcy's law, the fully coupled curing kinetic equation and the equivalent permeability model are coupled to construct a dynamic seepage equation describing the seepage behavior of the repair fluid in the pores of the buffer layer.

[0010] Preferably, step S121 specifically includes: Step S1211: Based on the Arrhenius equation and the degree of cure correlation, construct a coupled model of the viscosity of the repair material as a function of temperature and degree of cure: , in, To achieve the desired curing degree With temperature The viscosity of the material is as follows. It is a viscosity-dependent activation energy. This is the gas constant, typically taking the value of [value missing]. , This is the correlation coefficient between viscosity and degree of cure, with a value range of [value range missing]. ; Step S1212: The basic autocatalytic curing kinetic equation is combined with the Arrhenius equation to obtain the fully coupled curing kinetic equation, which is used to predict the curing process of the repair material in the pores of the buffer layer in real time and to characterize the change of curing degree over time. , in, The curing reaction rate, The order of the autocatalytic reaction is... This represents the stage of the curing reaction process.

[0011] Preferably, the equivalent permeability model in step S122 is: ; , in, The equivalent permeability of the buffer layer is expressed in units of... , This is the pore morphology correction factor, with a value range of [value range missing]. , The correlation is an empirical fitting formula for pore tortuosity and porosity, which is applicable to the pore characteristics of cable buffer layers made of non-woven fabric and semi-conductive tape.

[0012] Preferably, the dynamic seepage equation in step S123 is: Considering the dynamic characteristics of the viscosity of the repair material changing with the degree of curing, the curing kinetics equation is integrated with Darcy's law to construct a dynamic seepage equation: , in, For instantaneous injection of flow, which varies over time, The equivalent permeability of the buffer layer, The coupling model is updated in real time to achieve dynamic coupling between the injection process and the curing reaction.

[0013] Preferably, the inference parameters in step S13 specifically include: Real-time viscosity of repair material , curing degree Equivalent penetration rate Pore ​​filling rate pore filling uniformity Injection efficiency .

[0014] Preferably, step S13 specifically includes: Step S131: Based on the fully coupled curing kinetic equation, when the curing reaction rate is less than or equal to a preset curing reaction rate threshold, the repair material is determined to have reached a fully cured state, and the corresponding time is the full curing time. At this point, by integrating the fully coupled solidification kinetic equation (when... and (When), solve the relationship between the degree of cure and time, and obtain the degree of cure at time t. : , in, The integral constant is determined by the initial degree of curing. Decision, that is hour , Typically ≤0.05; Step S132: Based on the dynamic seepage equation, the formula for calculating the injection efficiency is derived, and the permeation rate of the repair material in the pores is calculated. , in, Injection efficiency, expressed as a percentage, reflects the permeability of the repair material per unit pressure difference and unit area. Its value needs to be... This ensures the uniformity of pore filling; Step S133: Calculate the pore filling uniformity. This is to compensate for the limitations of a single fill rate metric. , in, The pore filling uniformity is defined as follows: When the pore filling uniformity When the preset pore filling uniformity threshold is exceeded, it is determined that the repair material is evenly distributed in the pores to avoid local insufficient filling or accumulation. The cumulative injection amount is calculated as follows: ; The volume of pores that the buffer layer can fill is calculated as follows: ,in, The total volume of the buffer layer is calculated as follows: ,in, The cross-sectional area of ​​the buffer layer, The axial length of the repair section; Step S134: Calculate the pore filling rate : , in, This represents the percentage of uncured repair material, used to correct the matching relationship between the injected repair material volume and pore filling. Indicates the density of the material; Step S135: Compare the porosity in the inference parameters with the collected parameters. Injection pressure difference Instantaneous injection of traffic Combine them to obtain the model training input parameters; The model training input parameters include: Input parameters for material properties group, structural parameters group, and injection process dynamic parameters group; The input parameters for the material property group include: real-time viscosity of the repair material. , curing degree ; The structural parameter set input parameters include: buffer layer porosity. Equivalent penetration rate ; The input parameters of the dynamic parameter set for the injection process include: injection pressure difference. Pore ​​filling rate pore filling uniformity Injection efficiency Instantaneous injection of traffic .

[0015] Preferably, the PINN model in step S2 includes: an input layer, a hidden layer, a physical constraint layer, and an output layer; The input layer is used to input the model training input parameters, and the buffer layer is used for volume resistivity. The measured values ​​of partial discharge quantity (PD) were used for training the PINN model. The hidden layer consists of four fully connected layers, each containing 80 neurons. To avoid gradient vanishing or exploding during training, the weights of each fully connected layer are initialized using a He normal distribution. The hidden layer is connected to the input layer using the ReLU activation function, and the hidden layer is connected to the physical constraint layer using the Tanh activation function, which can improve the model's nonlinear fitting ability and numerical stability. The physical constraint layer is used to embed domain prior knowledge into the data-driven model, and corrects and limits the output of the PINN model through the following three physical constraints: (1) Dynamic seepage equation constraint: Based on Darcy's law, constrain the instantaneous flow rate change rate of the repair fluid in the pores of the buffer layer. With injection pressure difference With material viscosity rate of change of the ratio The seepage mechanics conservation relationship must be satisfied; (2) Injection and solidification coupling constraint: The constraint is based on the deviation between the injection flow rate predicted by the dynamic seepage equation and the measured injection flow rate. The error must be less than or equal to the preset coupling error threshold to ensure that the coupling model of the curing process and the percolation behavior is adaptable to the field. (3) Pore filling constraint: Constraining the uniformity of pore filling in the buffer layer by the repair material. and pore filling rate The preset repair process quality requirements must be met to ensure the final restoration effect of electrical performance; The output layer is responsible for outputting the model training result parameters, which include: buffer layer volume resistivity. Partial discharge quantity Curing uniformity .

[0016] Preferably, the hybrid loss function in step S2 transforms the injection process into a physical constraint term to balance data fitting accuracy, consistency of physical laws, and adaptability of the injection process. The specific form is as follows: , in, For data loss The weighting coefficients, For physical loss The weighting coefficients, Loss during the injection process The weighting coefficients; this hybrid loss function ensures that the model simultaneously follows the laws of solidification kinetics, Darcy flow, and injection process characteristics, thus improving the reliability of the evaluation; The data loss expression is: , in, The parameter sample size corresponding to the data loss. Let be the measured value of the partial discharge of the i-th sample. Let be the partial discharge quantity of the i-th sample predicted by the PINN model. Let be the measured value of the buffer layer volume resistivity for the i-th sample. Let be the volume resistivity of the i-th sample predicted by the PINN model; The physical loss expression is as follows: , in, The sample size of the parameters corresponding to physical loss. To predict the curing reaction rate using the PINN model for the j-th sample point, denoted as the predicted degree of curing by the PINN model, where m and n are the autocatalytic reaction orders. To inject instantaneous flow for the prediction of the PINN model for the j-th sample point, Real-time viscosity of the material; The loss expression for the injection process is as follows: , in, The sample size of the parameters corresponding to the loss during the injection process. Let be the measured value of the pore filling uniformity of the r-th sample. Let be the pore filling uniformity predicted by the PINN model for the r-th sample. Let be the measured value of the injection efficiency for the r-th sample. Let represent the injection efficiency of the r-th sample predicted by the PINN model.

[0017] Preferably, the preset accuracy standard for training the PINN model in step S2 includes: Buffer layer volume resistivity Prediction error ≤4%, partial discharge amount The prediction error is ≤5%, and the prediction error of the injection process parameters pore filling uniformity and injection efficiency is ≤3%.

[0018] Preferably, step S3 involves determining the level of repair effectiveness of the high-voltage cable buffer layer, including: The output parameters and injection process parameters of the PINN model are used as evaluation indicators for repair effectiveness. Combined with the operating standards for high-voltage cable buffer layers, a three-level evaluation system is established: The first level (superior level) means that all the evaluation indicators of the repair effect meet the requirements of complete restoration of electrical performance and structural integrity, ensuring that the cable can operate stably for a long time. Level 2 (Intermediate): The evaluation indicators of the repair effect all meet the requirements that the buffer layer can operate normally, but some indicators do not reach the level of complete recovery. It is necessary to monitor the partial discharge and resistivity changes in the injection area regularly to prevent performance degradation. Level 3 (Poor): Some repair effect evaluation indicators exceed the allowable range of operation. The repair effect is judged to be substandard. The injection process parameters need to be adjusted and the repair needs to be carried out again, and the evaluation should be carried out again.

[0019] Furthermore, this invention also provides a non-invasive evaluation system for the repair effect of high-voltage cable buffer layers based on multi-parameter fusion, comprising: The model input parameter acquisition module contains: Collect the characteristic parameters of the repair material, the pore structure parameters of the buffer layer, and the parameters of the repair fluid injection process; substitute the collected parameters into the prior knowledge base constructed based on solidification kinetics and seepage mechanism to obtain the model training input parameters; The model training module contains: A PINN model and a hybrid loss function based on physical constraints of the injection process are constructed, and the PINN model is trained using measured data of cable repair until the preset accuracy standard is met, thus obtaining a well-trained repair effect evaluation model. The model application module contains: The real-time collected data of the cable to be evaluated is input into the trained repair effect evaluation model, and the predicted value of the repair effect evaluation index is output. Based on the predicted value, the repair effect of the high-voltage cable buffer layer is judged and the injection process and curing parameters of the PINN model are optimized.

[0020] The beneficial effects of this invention are as follows: First, it achieves non-destructive and live-line application, requiring no external high voltage or pulse voltage excitation. Evaluation can be completed simply by collecting data on the characteristics of the repair material, pore structure, and injection process, making it particularly suitable for live cables, thus avoiding power outage losses and operational risks, and meeting the actual needs of on-site operation and maintenance. Second, it can construct a full-process coupled model based on the curing kinetic equation of the repair material through multi-parameter coupling and high-precision modeling, supplementing multiple correlation formulas to quantify the interaction between temperature, degree of curing, pore structure, and injection parameters. Compared with traditional single-index evaluation, it not only improves accuracy but also reduces the probability of misjudgment. In addition, the addition of physical constraints enhances the reliability of the model. Integrating physical equations such as curing kinetics and Darcy flow into the PINN model reduces dependence on massive amounts of measured data and improves the model's generalization ability under different voltage levels and different repair material systems, laying the foundation for its widespread application. Finally, this method has dynamic evaluation and prediction capabilities. It can not only evaluate the final effect after repair but also dynamically optimize the injection process through real-time parameter feedback, thereby improving the repair success rate.

[0021] Therefore, it is evident that the present invention has outstanding substantive features and significant progress compared with the prior art, and the beneficial effects of its implementation are also obvious. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0023] Figure 1 This is a flowchart of a non-invasive evaluation method for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion, provided by the present invention.

[0024] Figure 2This is a schematic diagram of a non-invasive evaluation system for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion, provided by the present invention.

[0025] Figure 3 This is a fitting curve of the fully coupled curing kinetic equation of the curing degree-time-temperature of the repair material provided by the present invention.

[0026] Figure 4 This is a schematic diagram of X-CT imaging and parameter extraction of the three-dimensional pore structure of the buffer layer provided by the present invention.

[0027] Figure 5 This is a block diagram of the multi-parameter fusion and effect evaluation structure based on the PINN model provided by the present invention.

[0028] Among them, 1-model input parameter acquisition module, 2-model training module, and 3-model application module. Detailed Implementation

[0029] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following implementation methods.

[0030] Example 1: like Figure 1 As shown in the figure, this embodiment provides a non-invasive evaluation method for the repair effect of high-voltage cable buffer layers based on multi-parameter fusion, which includes the following steps: Step S1: Collect the characteristic parameters of the repair material, the pore structure parameters of the buffer layer, and the parameters of the repair fluid injection process; substitute the collected parameters into the prior knowledge base constructed based on the curing kinetics and seepage mechanism to obtain the model training input parameters; Step S1 specifically includes: Step S11: Collect data on the characteristics of the repair material, the pore structure of the buffer layer, and the parameters of the repair fluid injection process; Step S12: Construct a prior knowledge base that can describe the entire repair process based on solidification kinetics and seepage mechanism; Step S13: Substitute the collected parameters into the prior knowledge base to obtain the inference parameters, and combine the inference parameters with some of the collected parameters to obtain the model training input parameters.

[0031] The specific steps in step S11, including collecting the characteristics of the repair material, the pore structure of the buffer layer, and the parameters of the repair fluid injection process, include: Step S111: Select a conductive repair material. The repair material adopts a silicone-based, epoxy-based, or polyurethane-based conductive repair system, and collect the characteristic parameters of the repair material. Step S112: A three-dimensional pore structure image of the buffer layer of the high-voltage cable to be repaired is obtained using X-ray computed tomography (X-CT) technology; the three-dimensional pore structure image is processed through image segmentation and morphological analysis to extract the pore structure parameters of the buffer layer; such as... Figure 4 As shown, the left side of the image is the original 3D image of the buffer layer obtained by X-CT scan, the middle side is the pore contour map after image segmentation and morphological processing, and the right side is the key parameter porosity. Average aperture Degree of curvature The annotation and calculation results can intuitively present the distribution pattern and quantitative indicators of the pores in the buffer layer, providing data support for the construction of the equivalent permeability model; Step S113: Real-time acquisition of repair fluid injection process parameters using high-precision sensors; The repair material characteristic parameters in step S111 include: the initial viscosity of the material. Activation energy of curing reaction Frequency factor Reaction order ambient temperature , curing degree Material density ; It should be noted that the initial viscosity of the material is... Need to be Data collected under ambient temperature conditions; the ambient temperature at the site. The range is generally in The degree of curing Data are collected using differential scanning calorimetry (DSC) or a rheometer. The pore structure parameters of the buffer layer in step S112 include: porosity. Average aperture Maximum aperture Pore ​​tortuosity Pore ​​connectivity ; It should be noted that the porosity The ratio of pore volume to the total volume of the buffer layer; the pore tortuosity The ratio of the actual seepage path length to the straight-line distance; the pore connectivity rate. The ratio of the volume of connected pores to the total pore volume; The parameters for the repair fluid injection process in step S113 include: injection pressure difference. Instantaneous injection of traffic Injection time seepage cross-sectional area seepage length ; It should be noted that the injection pressure difference The axial pressure difference between the injection end and the far outlet of the buffer layer; the seepage cross-sectional area The effective radial seepage area of ​​the buffer layer; the seepage length To repair the axial length of the segment.

[0032] The prior knowledge base describing the entire repair process in step S12 includes: Coupled model of viscosity of repair material with temperature and degree of cure, fully coupled curing kinetic equation describing the curing process of repair material, equivalent permeability model applicable to porous media of buffer layer, dynamic seepage equation coupling curing process and seepage behavior. Step S12 specifically includes: Step S121: Based on the collected characteristic parameters of the repair material, the Arrhenius equation and the autocatalytic curing kinetic theory are used to construct a coupled model of the change of the viscosity of the repair material with temperature and degree of curing, and a fully coupled curing kinetic equation describing the curing process of the repair material. Step S122: Based on the obtained pore structure parameters of the buffer layer, the Kozeny-Carman equation is used to construct an equivalent permeability model suitable for porous media in the buffer layer, which is used to quantify the seepage capacity of the remediation fluid in the pores. Step S123: Based on Darcy's law, the fully coupled curing kinetic equation and the equivalent permeability model are coupled to construct a dynamic seepage equation describing the seepage behavior of the repair fluid in the pores of the buffer layer.

[0033] Step S121 specifically includes: Step S1211: Based on the Arrhenius equation and the degree of cure correlation, construct a coupled model of the viscosity of the repair material as a function of temperature and degree of cure: , in, To achieve the desired curing degree With temperature The viscosity of the material is as follows. It is a viscosity-dependent activation energy. This is the gas constant, typically taking the value of [value missing]. , This is the correlation coefficient between viscosity and degree of cure, with a value range of [value range missing]. ; Step S1212: Combine the basic autocatalytic curing kinetic equation with the Arrhenius equation, as follows: Figure 3 As shown, a fully coupled curing kinetic equation is obtained, which is used to predict the curing process of the repair material in the pores of the buffer layer in real time and to characterize the change of the degree of curing over time. , in, The curing reaction rate, The order of the autocatalytic reaction is... This represents the stage of the curing reaction process.

[0034] The equivalent permeability model in step S122 is: ; , in, The equivalent permeability of the buffer layer is expressed in units of... , This is the pore morphology correction factor, with a value range of [value range missing]. , The correlation is an empirical fitting formula for pore tortuosity and porosity, which is applicable to the pore characteristics of cable buffer layers made of non-woven fabric and semi-conductive tape.

[0035] The dynamic seepage equation in step S123 is: Considering the dynamic characteristics of the viscosity of the repair material changing with the degree of curing, the curing kinetics equation is integrated with Darcy's law to construct a dynamic seepage equation: , in, For instantaneous injection of flow, which varies over time, The equivalent permeability of the buffer layer, The coupling model is updated in real time to achieve dynamic coupling between the injection process and the curing reaction.

[0036] The inference parameters in step S13 specifically include: Real-time viscosity of repair material , curing degree Equivalent penetration rate Pore ​​filling rate pore filling uniformity Injection efficiency .

[0037] Step S13 specifically includes: Step S131: Based on the fully coupled curing kinetic equation, when the curing reaction rate is less than or equal to a preset curing reaction rate threshold, the repair material is determined to have reached a fully cured state, and the corresponding time is the full curing time. At this point, by integrating the fully coupled solidification kinetic equation (when... and (When), solve the relationship between the degree of cure and time, and obtain the degree of cure at time t. : , in, The integral constant is determined by the initial degree of curing. Decision, that is hour , Typically ≤0.05; It should be noted that, through real-time monitoring and The variation curve can verify the coupling consistency between the dynamic seepage equation and the solidification kinetic equation: if the coupling error Less than or equal to the preset coupling error threshold This indicates that the dynamic seepage equation and the solidification kinetic equation are well-matched and can be used as the physical constraint basis for the PINN model. To predict the injection flow rate based on the dynamic seepage equation, The actual injection flow rate is measured; if coupling error... If the error exceeds the preset coupling error threshold, it indicates that the predicted injection flow rate of the dynamic seepage equation deviates too much from the measured injection flow rate. The currently constructed "solidification-seepage" coupling model fails to truly reflect the actual injection process and cannot be used as the physical constraint basis for PINN. Parameter correction is required.

[0038] Step S132: Based on the dynamic seepage equation, the formula for calculating the injection efficiency is derived, and the permeation rate of the repair material in the pores is calculated. , in, Injection efficiency, expressed as a percentage, reflects the permeability of the repair material per unit pressure difference and unit area. Its value needs to be... This ensures the uniformity of pore filling; Step S133: Calculate the pore filling uniformity. This is to compensate for the limitations of a single fill rate metric. , in, The pore filling uniformity is defined as follows: When the pore filling uniformity When the preset pore filling uniformity threshold is exceeded, it is determined that the repair material is evenly distributed in the pores to avoid local insufficient filling or accumulation. The cumulative injection amount is calculated as follows: ; The volume of pores that the buffer layer can fill is calculated as follows: ,in, The total volume of the buffer layer is calculated as follows: ,in, The cross-sectional area of ​​the buffer layer, The axial length of the repair section; Step S134: Calculate the pore filling rate : , in, This represents the percentage of uncured repair material, used to correct the matching relationship between the injected repair material volume and pore filling. This indicates the density of the material.

[0039] Step S135: Compare the porosity in the inference parameters with the collected parameters. Injection pressure difference Instantaneous injection of traffic Combine them to obtain the model training input parameters; The model training input parameters include: Input parameters for material properties group, structural parameters group, and injection process dynamic parameters group; The input parameters for the material property group include: real-time viscosity of the repair material. , curing degree ; The structural parameter set input parameters include: buffer layer porosity. Equivalent penetration rate ; The input parameters of the dynamic parameter set for the injection process include: injection pressure difference. Pore ​​filling rate pore filling uniformity Injection efficiency Instantaneous injection of traffic .

[0040] Step S2: Construct the PINN model and a hybrid loss function based on the physical constraints of the injection process, and train the PINN model using measured data of cable repair until the preset accuracy standard is met, thus obtaining a well-trained repair effect evaluation model. It should be noted that the measured cable repair data used for model training includes data covering different porosities, injection pressure differences, and temperature conditions. Group The actual cable repair data was used as the training set, and 30 sets of new cable repair working condition data were used as the test set to test the model's generalization ability. The training process of the PINN model was configured as follows: 10,000 iterations, gradient descent using the Adam optimizer, initial learning rate of 0.001, and the learning rate was reduced to 0.0001 after 5,000 iterations to ensure training convergence. During training, the model was iteratively updated using the training set. After training was completed, the model performance was evaluated using the test set until the model accuracy judgment criteria were met, i.e., the model was deemed to have passed training.

[0041] like Figure 5 As shown, the PINN model in step S2 includes: an input layer, a hidden layer, a physical constraint layer, and an output layer; The input layer is used to input the model training input parameters, and the buffer layer is used for volume resistivity. The measured values ​​of partial discharge quantity (PD) were used for training the PINN model. The hidden layer consists of four fully connected layers, each containing 80 neurons. To avoid gradient vanishing or exploding during training, the weights of each fully connected layer are initialized using a He normal distribution. The hidden layer is connected to the input layer using the ReLU activation function, and the hidden layer is connected to the physical constraint layer using the Tanh activation function, which can improve the model's nonlinear fitting ability and numerical stability. The physical constraint layer is used to embed domain prior knowledge into the data-driven model, and corrects and limits the output of the PINN model through the following three physical constraints: (1) Dynamic seepage equation constraint: The seepage behavior of the repair material in the pores of the buffer layer is constrained to satisfy the Darcy flow conservation law: ,in, The instantaneous flow rate change rate To inject differential pressure, Where K is the viscosity of the material, and K is the equivalent permeability of the buffer layer. For seepage cross-sectional area, For pore tortuosity, The seepage length; (2) Injection and solidification coupling constraint: The constraint is based on the deviation between the injection flow rate predicted by the dynamic seepage equation and the measured injection flow rate. The error must be less than or equal to the preset coupling error threshold to ensure that the coupling model of the curing process and the percolation behavior is adaptable to the field. (3) Pore filling constraint: Constraining the uniformity of pore filling in the buffer layer of the repair material. and pore filling rate To meet the preset repair process quality requirements and ensure the final restoration effect of electrical performance; It should be noted that by using a physical constraint layer, the dynamic characteristics of the injection process are transformed into constraints for training the PINN model, which can avoid the overfitting problem of a purely data-driven model. The output layer is responsible for outputting the model training result parameters, which include: buffer layer volume resistivity. Partial discharge quantity Curing uniformity ; It should be noted that the curing uniformity is... It can reflect the spatial uniformity of the curing of the repair material.

[0042] The hybrid loss function in step S2 transforms the injection process into a physical constraint term to balance data fitting accuracy, consistency of physical laws, and adaptability of the injection process. Its specific form is as follows: , in, For data loss The weighting coefficients, For physical loss The weighting coefficients, Loss during the injection process The weighting coefficients; this hybrid loss function ensures that the model simultaneously follows the laws of solidification kinetics, Darcy flow, and injection process characteristics, thus improving the reliability of the evaluation; The data loss expression is: , in, The parameter sample size corresponding to the data loss. Let be the measured value of the partial discharge of the i-th sample. Let be the partial discharge quantity of the i-th sample predicted by the PINN model. Let be the measured value of the buffer layer volume resistivity for the i-th sample. Let be the volume resistivity of the i-th sample predicted by the PINN model; The physical loss expression is as follows: , in, The sample size of the parameters corresponding to physical loss. To predict the curing reaction rate for the j-th sample point using the PINN model, denoted as the predicted degree of curing by the PINN model, where m and n are the autocatalytic reaction orders. To inject instantaneous flow for the prediction of the PINN model for the j-th sample point, Real-time viscosity of the material; The loss expression for the injection process is as follows: , in, The sample size of the parameters corresponding to the loss during the injection process. Let be the measured value of the pore filling uniformity of the r-th sample. Let be the pore filling uniformity predicted by the PINN model for the r-th sample. Let be the measured value of the injection efficiency for the r-th sample. Let represent the injection efficiency of the r-th sample predicted by the PINN model.

[0043] The preset accuracy standard for training the PINN model in step S2 includes: Buffer layer volume resistivity Prediction error ≤4%, partial discharge amount The prediction error is ≤5%, and the prediction error of the injection process parameters pore filling uniformity and injection efficiency is ≤3%.

[0044] Step S3: Input the real-time collected data of the cable to be evaluated into the trained repair effect evaluation model, output the predicted value of the repair effect evaluation index, and perform grade determination of the repair effect of the high-voltage cable buffer layer and optimize the injection process and curing parameters of the PINN model based on the predicted value. Step S3 involves determining the repair effect level of the high-voltage cable buffer layer, including: Using the output parameters and injection process parameters of the PINN model as evaluation indicators of repair effectiveness, and combining them with the operating standards of high-voltage cable buffer layers, a three-level evaluation system was established: First grade (superior): Volume resistivity Partial discharge quantity pore filling rate , degree of solidification pore filling uniformity Injection efficiency When this level is met, the electrical performance and structural integrity of the buffer layer are fully restored, and it can operate stably for a long time. Level 2 (Intermediate): , , , , , When this level is met, the buffer layer can operate normally, but the partial discharge and resistivity changes in the injection area need to be monitored every 3 months to prevent performance degradation. Third level (poor): , , , , , The injection process needs to be optimized (adjusting the injection pressure difference ΔP and controlling the ambient temperature T to optimize). After re-injecting the repair material, a re-evaluation will be conducted. It should be noted that the evaluation results based on the repair effectiveness level can optimize the injection process and curing parameters. For example, if the evaluation is medium and If it is too low, the injection pressure differential can be increased. No more than To avoid damaging the buffer layer and improve the uniformity of material penetration; if The temperature is too low; you can adjust the ambient temperature to [lower temperature]. Reduce material viscosity To optimize penetration efficiency, a closed-loop system of "injection modeling, PINN evaluation, and process optimization" is formed.

[0045] Example 2: like Figure 2 As shown in the figure, this embodiment provides a non-invasive evaluation system for the repair effect of high-voltage cable buffer layers based on multi-parameter fusion, comprising: Model input parameter acquisition module 1, in which: Collect the characteristic parameters of the repair material, the pore structure parameters of the buffer layer, and the parameters of the repair fluid injection process; substitute the collected parameters into the prior knowledge base constructed based on solidification kinetics and seepage mechanism to obtain the model training input parameters; Model training module 2, in which: A PINN model and a hybrid loss function based on physical constraints of the injection process are constructed, and the PINN model is trained using measured data of cable repair until the preset accuracy standard is met, thus obtaining a well-trained repair effect evaluation model. Model application module 3, in which: The real-time collected data of the cable to be evaluated is input into the trained repair effect evaluation model, and the predicted value of the repair effect evaluation index is output. Based on the predicted value, the repair effect of the high-voltage cable buffer layer is judged and the injection process and curing parameters of the PINN model are optimized.

[0046] The above-disclosed embodiments are merely preferred embodiments of the present invention, but the present invention is not limited thereto. Any non-creative variations that can be conceived by those skilled in the art, as well as any improvements and modifications made without departing from the principles of the present invention, should fall within the protection scope of the present invention.

Claims

1. A non-invasive evaluation method for the repair effect of high-voltage cable buffer layers based on multi-parameter fusion, characterized in that, Includes the following steps: Step S1: Collect the characteristic parameters of the repair material, the pore structure parameters of the buffer layer, and the parameters of the repair fluid injection process; The collected parameters are substituted into the prior knowledge base built on solidification kinetics and seepage mechanism to obtain the model training input parameters; The prior knowledge base includes: a coupled model of the viscosity of the repair material as a function of temperature and degree of curing; a fully coupled curing kinetic equation describing the curing process of the repair material; an equivalent permeability model applicable to porous media in buffer layers; and a dynamic seepage equation that couples the curing process with seepage behavior. Step S2: Construct the PINN model and a hybrid loss function based on the physical constraints of the injection process, and train the PINN model using measured data of cable repair until the preset accuracy standard is met, thus obtaining a well-trained repair effect evaluation model. Step S3: Input the real-time collected data of the cable to be evaluated into the trained repair effect evaluation model, output the predicted value of the repair effect evaluation index, and perform grade determination of the repair effect of the high-voltage cable buffer layer and optimize the injection process and curing parameters of the PINN model based on the predicted value. The PINN model in step S2 includes: an input layer, a hidden layer, a physical constraint layer, and an output layer; The input layer is used to input the model training input parameters, and the buffer layer is used for volume resistivity. And the measured value of partial discharge quantity PD; The hidden layer consists of four fully connected layers, each containing 80 neurons. To avoid gradient vanishing or exploding during training, the weights of each fully connected layer are initialized using a He normal distribution. The hidden layer is connected to the input layer using the ReLU activation function, and the hidden layer is connected to the physical constraint layer using the Tanh activation function, which can improve the model's nonlinear fitting ability and numerical stability. The physical constraint layer is used to embed domain prior knowledge into the data-driven model, and corrects and limits the output of the PINN model through the following three physical constraints: (1) Dynamic seepage equation constraint: The seepage behavior of the repair material in the pores of the buffer layer is constrained to satisfy the Darcy flow conservation law: ,in, The instantaneous flow rate change rate To inject differential pressure, Where K is the viscosity of the material, and K is the equivalent permeability of the buffer layer. For seepage cross-sectional area, For pore tortuosity, The seepage length; (2) Injection and solidification coupling constraint: The constraint is based on the deviation between the injection flow rate predicted by the dynamic seepage equation and the measured injection flow rate. It must be less than or equal to the preset coupling error threshold; (3) Pore filling constraint: Constraining the uniformity of pore filling in the buffer layer of the repair material. Greater than or equal to the preset pore filling uniformity threshold and pore filling rate Greater than or equal to the preset pore filling rate threshold; The output layer is responsible for outputting the model training result parameters, which include: buffer layer volume resistivity. Partial discharge quantity Curing uniformity ; The hybrid loss function in step S2: The injection process is transformed into physical constraint terms, in the following form: ， in, For data loss The weighting coefficients, For physical loss The weighting coefficients, Loss during the injection process Weighting coefficients; The data loss The calculation expression is: ， in, The parameter sample size corresponding to the data loss. Let be the measured value of the partial discharge of the i-th sample. Let be the partial discharge quantity of the i-th sample predicted by the PINN model. Let be the measured value of the buffer layer volume resistivity for the i-th sample. Let be the volume resistivity of the i-th sample predicted by the PINN model; The physical loss The calculation expression is: ， in, The sample size of the parameters corresponding to physical loss. To predict the curing reaction rate using the PINN model for the j-th sample point, denoted as the predicted degree of curing by the PINN model, where m and n are the autocatalytic reaction orders. To inject instantaneous flow for the prediction of the PINN model for the j-th sample point, Real-time viscosity of the material; The loss expression for the injection process is: ， in, The sample size of the parameters corresponding to the loss during the injection process. Let be the measured value of the pore filling uniformity of the r-th sample. Let be the pore filling uniformity predicted by the PINN model for the r-th sample. Let be the measured value of the injection efficiency for the r-th sample. Let represent the injection efficiency of the r-th sample predicted by the PINN model.

2. The non-invasive evaluation method for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion according to claim 1, characterized in that, Step S1 specifically includes: Step S11: Collect data on the characteristics of the repair material, the pore structure of the buffer layer, and the parameters of the repair fluid injection process; Step S12: Construct a prior knowledge base that can describe the entire repair process based on solidification kinetics and seepage mechanism; Step S13: Substitute the collected parameters into the prior knowledge base to obtain the inference parameters, and combine the inference parameters with some of the collected parameters to obtain the model training input parameters. The inference parameters in step S13 specifically include: Real-time viscosity of repair material , curing degree Equivalent penetration rate Pore ​​filling rate pore filling uniformity Injection efficiency .

3. The non-invasive evaluation method for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion according to claim 2, characterized in that, The specific steps in step S11, including collecting the characteristics of the repair material, the pore structure of the buffer layer, and the parameters of the repair fluid injection process, include: Step S111: Select a conductive repair material. The repair material adopts a silicone-based, epoxy-based, or polyurethane-based conductive repair system, and collect the characteristic parameters of the repair material. Step S112: Obtain a three-dimensional pore structure image of the buffer layer of the high-voltage cable to be repaired using X-ray computed tomography (CT) technology; process the three-dimensional pore structure image through image segmentation and morphology to extract the pore structure parameters of the buffer layer. Step S113: Real-time acquisition of repair fluid injection process parameters using high-precision sensors; The repair material characteristic parameters in step S111 include: the initial viscosity of the material. Activation energy of curing reaction Frequency factor Reaction order ambient temperature , curing degree Material density ; The pore structure parameters of the buffer layer in step S112 include: porosity. Average aperture Maximum aperture Pore ​​tortuosity Pore ​​connectivity ; The parameters for the repair fluid injection process in step S113 include: injection pressure difference. Instantaneous injection of traffic Injection time seepage cross-sectional area seepage length .

4. The non-invasive evaluation method for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion according to claim 2, characterized in that, Step S12 specifically includes: Step S121: Using the Arrhenius equation and the autocatalytic curing kinetics theory, a coupled model of the viscosity of the repair material as a function of temperature and degree of curing is constructed, along with a fully coupled curing kinetic equation describing the curing process of the repair material. Step S122: Using the Kozeny-Carman equation, construct an equivalent permeability model suitable for porous media in the buffer layer; Step S123: Based on Darcy's law, the fully coupled curing kinetic equation and the equivalent permeability model are coupled to construct a dynamic seepage equation describing the seepage behavior of the repair fluid in the pores of the buffer layer. Step S121 specifically includes: Step S1211: Based on the Arrhenius equation and the degree of cure correlation, construct a coupled model of the viscosity of the repair material as a function of temperature and degree of cure: ， in, To achieve the desired curing degree With temperature The viscosity of the material is as follows. It is a viscosity-dependent activation energy. The gas constant is This is the correlation coefficient between viscosity and degree of cure; Step S1212: Combine the basic autocatalytic curing kinetic equation with the Arrhenius equation to obtain the fully coupled curing kinetic equation: ， in, The curing reaction rate, The order of the autocatalytic reaction is... The order of the curing reaction process; The equivalent permeability model in step S122 is: ； ， in, The equivalent permeability of the buffer layer is expressed in units of... , This is a pore morphology correction factor. The correlation is an empirical fitting formula for pore tortuosity and porosity; The dynamic seepage equation in step S123 is: By integrating the solidification kinetic equation with Darcy's law, a dynamic seepage equation is constructed: ， in, To inject traffic instantly, The equivalent permeability of the buffer layer is expressed in units of... , The coupled model is updated in real time.

5. The non-invasive evaluation method for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion according to claim 2, characterized in that, Step S13 specifically includes: Step S131: Based on the fully coupled curing kinetic equation, when the curing reaction rate is less than or equal to a preset curing reaction rate threshold, the repair material is determined to have reached a fully cured state, and the corresponding time is the full curing time. At this point, by integrating the fully coupled curing kinetic equation, the relationship between the degree of curing and time is solved, and the degree of curing at time t is obtained. : ， in, The integral constant is determined by the initial degree of curing. Decision, that is hour ; Step S132: Based on the dynamic seepage equation, the formula for calculating the injection efficiency is derived, and the permeation rate of the repair material in the pores is calculated. ， in, Injection efficiency, expressed in % %. Step S133: Calculate the pore filling uniformity. : ， in, For pore filling uniformity; when pore filling uniformity When the preset pore filling uniformity threshold is exceeded, it is determined that the repair material is evenly distributed in the pores to avoid local insufficient filling or accumulation. The cumulative injection amount is calculated as follows: ; The volume of pores that the buffer layer can fill is calculated as follows: ,in, The total volume of the buffer layer is calculated as follows: ,in, The cross-sectional area of ​​the buffer layer, The axial length of the repair section; Step S134: Calculate the pore filling rate : ， in, The percentage of uncured repair material. Indicates the density of the material; Step S135: Compare the porosity in the inference parameters with the collected parameters. Injection pressure difference Instantaneous injection of traffic Combine them to obtain the model training input parameters; The model training input parameters include: Input parameters for material properties group, structural parameters group, and injection process dynamic parameters group; The input parameters for the material property group include: real-time viscosity of the repair material. , curing degree ; The structural parameter set input parameters include: buffer layer porosity. Equivalent penetration rate ; The input parameters of the dynamic parameter set for the injection process include: injection pressure difference. Pore ​​filling rate pore filling uniformity Injection efficiency Instantaneous injection of traffic .

6. The non-invasive evaluation method for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion according to claim 1, characterized in that, The preset accuracy standard for training the PINN model in step S2 includes: Buffer layer volume resistivity Prediction error, partial discharge amount The prediction errors, as well as the prediction errors of the injection process parameters pore filling uniformity and injection efficiency, are all less than the corresponding preset thresholds.

7. The non-invasive evaluation method for the repair effect of high-voltage cable buffer layer based on multi-parameter fusion according to claim 1, characterized in that, Step S3 involves determining the repair effect level of the high-voltage cable buffer layer, including: The output parameters and injection process parameters of the PINN model are used as evaluation indicators for repair effectiveness. Combined with the operating standards for high-voltage cable buffer layers, a three-level evaluation system is established: The first level means that all the evaluation indicators of the repair effect meet the requirements of complete restoration of electrical performance and structural integrity, ensuring that the cable can operate stably for a long time. The second level is where all the evaluation indicators of the repair effect meet the requirements that the buffer layer can operate normally, but some indicators do not meet the requirements of complete recovery. It is necessary to monitor the partial discharge and resistivity changes in the injection area regularly to prevent performance degradation. The third level indicates that some repair effect evaluation indicators are outside the allowable range of operation, and the repair effect is deemed unsatisfactory. The injection process parameters need to be adjusted and the repair needs to be carried out again, and the evaluation should be conducted again.

8. A non-invasive evaluation system for the repair effect of high-voltage cable buffer layers based on multi-parameter fusion, characterized in that, include: The model input parameter acquisition module contains: Collect parameters of the repair material properties, the pore structure of the buffer layer, and the injection process parameters of the repair fluid. The collected parameters are substituted into the prior knowledge base built on solidification kinetics and seepage mechanism to obtain the model training input parameters; The model training module contains: A PINN model and a hybrid loss function based on physical constraints of the injection process are constructed, and the PINN model is trained using measured data of cable repair until the preset accuracy standard is met, thus obtaining a well-trained repair effect evaluation model. The model application module contains: The real-time collected data of the cable to be evaluated is input into the trained repair effect evaluation model, and the predicted value of the repair effect evaluation index is output. Based on the predicted value, the repair effect of the high-voltage cable buffer layer is judged and the injection process and curing parameters of the PINN model are optimized. The hybrid loss function transforms the injection process into a physical constraint term to balance data fitting accuracy, consistency of physical laws, and adaptability of the injection process. Its specific form is as follows: ， in, For data loss The weighting coefficients, For physical loss The weighting coefficients, Loss during the injection process The weighting coefficients; this hybrid loss function ensures that the model simultaneously follows the laws of solidification kinetics, Darcy flow, and injection process characteristics, thus improving the reliability of the evaluation; The data loss expression is: ， in, The parameter sample size corresponding to the data loss. Let be the measured value of the partial discharge of the i-th sample. Let be the partial discharge quantity of the i-th sample predicted by the PINN model. Let be the measured value of the buffer layer volume resistivity for the i-th sample. Let be the volume resistivity of the i-th sample predicted by the PINN model; The physical loss expression is: ， in, The sample size of the parameters corresponding to physical loss. To predict the curing reaction rate using the PINN model for the j-th sample point, denoted as the predicted degree of curing by the PINN model, where m and n are the autocatalytic reaction orders. To inject instantaneous flow for the prediction of the PINN model for the j-th sample point, Real-time viscosity of the material; The loss expression for the injection process is: ， in, The sample size of the parameters corresponding to the loss during the injection process. Let be the measured value of the pore filling uniformity of the r-th sample. Let be the pore filling uniformity predicted by the PINN model for the r-th sample. Let be the measured value of the injection efficiency for the r-th sample. Let represent the injection efficiency of the r-th sample predicted by the PINN model.

Images