Power cable buffer layer stress analysis method and equipment

Abstract

The invention discloses a power cable buffer layer stress analysis method and equipment, and relates to the technical field of cable stress analysis, and the method comprises the following steps: embedding a shape memory material into a cable buffer layer, and obtaining excitation intervals of different cable sections through machine learning based on the obtained historical data of cable operation; self-adaptive excitation is obtained in combination with the excitation interval, the pre-constructed constitutive model and the pre-constructed proxy model; adaptive excitation is applied to the cable buffer layer, and an actual measurement response vector of the physical attribute of the shape memory material is obtained; and optimizing the stress state of the inversion cable buffer layer in combination with the actually measured response vector and the obtained theoretical response vector. The method is used for solving the problem that an existing cable buffer layer stress analysis technology is insufficient in accuracy.

Description

Technical Field

[0001] This invention relates to the field of cable stress analysis technology, and more specifically, to a method and equipment for stress analysis of the buffer layer of power cables. Background Technology

[0002] Power cables are the arteries of modern energy systems, and their operational reliability directly affects the safety and stability of the entire power grid. As a key structural component of the cable, the buffer layer, typically located between the metal sheath and the outer sheath, primarily performs multiple important functions, including mechanical buffering, electrical stress conduction, and longitudinal water blocking. However, during the cable's manufacturing, laying, and long-term operation, the buffer layer inevitably bears and accumulates various loads from bending, ground settlement, internal thermomechanical stress, and external compression. These changes in stress state are key indicators for assessing the cable's health and predicting its remaining lifespan.

[0003] Traditional stress analysis often relies on multiple independent specialized devices to test tensile, compressive, and bending forces separately. This analytical approach not only significantly reduces testing efficiency and increases operational complexity, but more importantly, it struggles to recreate and assess the complex coupled stress state of the buffer layer in real-world operating environments. The lack of a unified testing platform makes it impossible to effectively reveal the interrelationships between different stress forms, thus hindering a comprehensive and accurate assessment of the cable's overall mechanical condition. Furthermore, existing methods primarily focus on post-incident inspection; by the time visible damage or ablation occurs in the buffer layer, its performance has often severely deteriorated, missing the optimal maintenance window.

[0004] For example, the invention patent announcement CN115831273B discloses a numerical simulation method for shape memory polymers based on supervisoelastic constitutive models. It takes both the supervisoelastic constitutive model and the nonlinear finite element analysis method as starting points. First, a shape memory polymer constitutive model based on supervisoelastic theory is designed and implemented in the finite element analysis software ABAQUS. Then, finite element simulation is performed on the thermo-structure coupling behavior of the shape memory polymer. This method can, to a certain extent, provide a relatively accurate numerical simulation of the shape memory behavior of common structures such as beams, rods, plates, and shells, obtaining commonly used mechanical response quantities such as stress, strain, and displacement within the structure.

[0005] For example, the invention patent announcement CN118571382B discloses a method for inverting the constitutive model of shape memory alloy based on nanoindentation, which includes: S1, establishing a nanoindentation experimental model and a uniaxial tensile unloading experimental model; S2, extracting the force-displacement curves of the indenter under different material parameters; S3, extracting the stress-strain curves of the tensile surface under different material parameters; S4, extracting the force parameter at each point of the force-displacement curve and the stress parameter at each point of the stress-strain curve, which are used as the input and output of a fully connected neural network, respectively; S5, training the neural network to predict the stress-strain response of the uniaxial tensile unloading test model during the loading and unloading stages under different material parameters.

[0006] The above-disclosed technical solutions have at least the following technical problems: Existing methods for analyzing the stress on cable buffer layers involve complex model calibration processes, parameter errors are amplified by the model, and the model has poor generalization ability, making it difficult to adapt to complex and changing environments.

[0007] To address the above problems, this invention proposes a solution. Summary of the Invention

[0008] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method and apparatus for stress analysis of power cable buffer layers. By combining machine learning, constitutive models, and surrogate models, an adaptive excitation strategy is generated, and optimization inversion is performed based on the measured response and theoretical response under excitation, thereby solving the problem of insufficient accuracy of existing cable buffer layer stress analysis techniques.

[0009] To achieve the above objectives, the present invention provides the following technical solution: A stress analysis method for power cable buffer layers includes the following steps: embedding shape memory material into the cable buffer layer; obtaining excitation intervals for different cable sections based on historical data of cable operation obtained through machine learning; obtaining adaptive excitation by combining the excitation intervals, a pre-built constitutive model, and a pre-built surrogate model; applying adaptive excitation to the cable buffer layer to obtain the measured response vector of the physical properties of the shape memory material; and optimizing and inverting the stress state of the cable buffer layer by combining the measured response vector and the obtained theoretical response vector.

[0010] In a preferred embodiment, the step of obtaining the excitation intervals of different cable sections based on the acquired historical data of cable operation through machine learning specifically involves: constructing a unified discriminant causal forest model based on the acquired historical data of cable operation; obtaining the response effect matrix of different cable sections under different excitation levels based on the unified discriminant causal forest model; constructing an MCKP problem to allocate the optimal excitation to each cable section based on the response effect matrix; solving the MCKP problem; and converting the solution of the MCKP problem into the excitation intervals of different cable sections.

[0011] In a preferred embodiment, the construction of a unified discriminative causal forest model based on the acquired historical data of cable operation specifically involves: randomly dividing the historical data into a tree set and an estimation set; based on the tree set, using the maximization of the response differences of different cable sections to excitation as the splitting criterion, constructing several causal trees to form a causal forest; obtaining the rate of change of physical properties of cable sections falling into the same leaf node in the estimation set under different excitation levels, and using the mean difference of the rate of change as the conditional average treatment effect of that leaf node.

[0012] In a preferred embodiment, the adaptive excitation obtained by combining the excitation interval, the pre-built constitutive model, and the pre-built surrogate model specifically involves: pre-setting initial adaptive excitation parameters within the excitation interval, the adaptive excitation parameters including excitation intensity and excitation duration; obtaining predicted responses under different excitation parameters based on the pre-built surrogate model, and verifying the physical rationality of the predicted responses based on the pre-built constitutive model; constructing an excitation loss function based on the difference between the predicted response and the expected response preset based on historical data; and optimizing and updating the initial adaptive excitation parameters based on the excitation loss function to generate the final adaptive excitation parameters.

[0013] In a preferred embodiment, applying adaptive excitation to the cable buffer layer to obtain the measured response vector of the physical properties of the shape memory material specifically involves: preprocessing the measured response data to extract time-domain and frequency-domain features; performing curve fitting on the preprocessed measured response data to obtain a response curve and extracting the dynamic process features of the response curve; and integrating the time-domain features, frequency-domain features, and dynamic process features to obtain the response vector.

[0014] In a preferred embodiment, the method for obtaining the theoretical response vector specifically includes: obtaining initial theoretical response data through a surrogate model based on adaptive excitation and assumed stress state of the cable buffer layer; verifying the physical rationality of the initial theoretical response data through a constitutive model and correcting it to obtain theoretical response data; and converting the theoretical response data into a theoretical response vector.

[0015] In a preferred embodiment, the step of combining the measured response vector and the obtained theoretical response vector to optimize the stress state of the cable buffer layer specifically involves: constructing an inversion loss function based on the difference between the measured response vector and the theoretical response vector, with minimizing the inversion loss function as the optimization objective; solving for the optimization objective to obtain the initial optimal solution for the stress state of the cable buffer layer; and using the initial optimal solution for the stress state as a starting point, obtaining the final stress state of the cable buffer layer through local search optimization.

[0016] In a preferred embodiment, the step of solving the optimization objective to obtain the initial optimal solution for the stress state of the cable buffer layer specifically involves: generating an initial population of stress states of the cable buffer layer through Latin hypercube sampling; introducing a hyperbolic cosine adaptive function to adjust the inertial weights of particles in the stress state population; and, based on a differential mutation strategy, introducing the vector difference between particles in the stress state population as a perturbation term into the particle position update formula; iterating the initial stress state population; and when the iteration satisfies a preset global convergence condition, outputting the optimal particle position in the current stress state population as the initial optimal solution for the stress state.

[0017] In a preferred embodiment, the step of obtaining the final stress state of the cable buffer layer through local search optimization, starting from the initial optimal solution of the stress state, specifically involves: using the initial optimal solution of the stress state as the initial point of the local search and setting iteration parameters; during the iteration process, calculating the gradient of the inversion loss function corresponding to the current stress state of the cable buffer layer, and determining the local search direction based on the gradient of the inversion loss function; performing a linear search along the local search direction to obtain a trial solution, and calculating the corresponding trial inversion loss function value; updating the stress state of the cable buffer layer and iteration parameters through the Metropolis acceptance criterion; and iterating until the preset local convergence condition is met, outputting the final stress state of the cable buffer layer.

[0018] An electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the stress analysis method for the power cable buffer layer.

[0019] The technical effects and advantages of the stress analysis method and equipment for power cable buffer layers of this invention are as follows: This invention defines the range of subsequent adaptive excitation by determining the excitation intervals of different cable sections; it dynamically generates adaptive excitation for cable sections by combining constitutive and surrogate models to accurately guide shape memory materials to produce deformation recovery; and it achieves high-precision quantitative inversion of the stress state of cable buffer layers by comparing the measured physical response and theoretical response under excitation and optimizing the inversion, effectively solving the problem of insufficient accuracy in existing cable buffer layer stress analysis technology. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the stress analysis method for the buffer layer of a power cable provided in an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of the power cable buffer layer stress analysis device provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1, Figure 1 The present invention provides a method for stress analysis of the buffer layer of a power cable, comprising the following steps: S1, embed shape memory material into the cable buffer layer, and obtain the excitation range of different cable sections through machine learning based on the acquired historical data of cable operation; S2, combining the excitation interval, the pre-constructed constitutive model, and the pre-constructed surrogate model, yields adaptive excitation; S3, apply adaptive excitation to the cable buffer layer to obtain the measured response vector of the physical properties of the shape memory material; S4. Combining the measured response vector and the obtained theoretical response vector, the stress state of the cable buffer layer is optimized and inverted.

[0024] This embodiment defines the range of subsequent adaptive excitation by determining the excitation intervals of different cable sections; it dynamically generates adaptive excitation for cable sections by combining constitutive and surrogate models to accurately guide the shape memory material to generate deformation recovery; by comparing the measured physical response and theoretical response under excitation and optimizing the inversion, it achieves high-precision quantitative inversion of the stress state of the cable buffer layer, effectively solving the problem of insufficient accuracy of existing cable buffer layer stress analysis technology.

[0025] S1. Shape memory material is embedded in the cable buffer layer. Based on the historical data of cable operation, the excitation range of different cable sections is obtained through machine learning.

[0026] In this embodiment, obtaining the excitation intervals for different cable sections based on the acquired historical data of cable operation through machine learning includes: A unified discriminant causal forest model is constructed based on the acquired historical data of cable operation. The response effect matrix of different cable sections under different excitation levels is obtained based on the unified discriminant causal forest model. Based on the response effect matrix, we construct the MCKP problem of assigning optimal excitation to each cable segment; Solve the MCKP problem and convert the solution of the MCKP problem into excitation intervals for different cable sections.

[0027] In this embodiment, the construction of a unified discriminative causal forest model based on the acquired historical data of cable operation includes: Historical data is randomly divided into a tree-construction set and an estimation set; Based on the tree set, several causal trees are constructed to form a causal forest by using the maximization of the difference in response to excitation among different cable sections as the splitting criterion. Obtain the rate of change of physical properties of cable segments falling into the same leaf node under different excitation levels, and use the difference in the mean of the rate of change as the conditional averaging effect of that leaf node.

[0028] In this embodiment, obtaining the response effect matrix of different cable sections under different excitation levels based on the unified discriminant causal forest model specifically involves: Construct a dataset, which specifically includes feature variables, processing variables, and outcome variables; The characteristic variables include historical data of the cable section's operation, the processing variables are different excitation intensities and excitation durations, and the result variable is the rate of change of the physical properties of the shape memory material. Based on the dataset, the conditional average treatment effect of each cable segment under different excitation levels is obtained by using a unified discriminant causal forest model, and then converted into a matrix form to obtain the response effect matrix, with the cable segment as the row and the excitation level as the column.

[0029] In this embodiment, the MCKP problem of assigning optimal excitation to each cable segment based on the response effect matrix is ​​specifically as follows: The cable segments are mapped to items in the MCKP problem, and each stimulus level corresponds to a choice in the MCKP problem. The conditional average treatment effect of the cable section under the excitation level in the response effect matrix corresponds to the benefit in the MCKP problem. The cost in the MCKP problem is obtained by weighting the time cost and energy cost according to the preset weights. The optimization objective is to maximize the sum of the products of the values ​​of all segments and all choices with their decision variables.

[0030] In this embodiment, the decision variable is specifically defined as follows: if a certain cable section selects a certain incentive level, the corresponding decision variable is 1; otherwise, it is 0.

[0031] In this embodiment, solving the MCKP problem and converting the solution of the MCKP problem into excitation intervals for different cable segments specifically involves: By relaxing the cost constraint and introducing Lagrange multipliers, we obtain the Lagrange function and its dual problem. The excitation level for the cable segment is obtained by solving the Lagrange dual problem through binary search, and the excitation interval is obtained based on the selected excitation level.

[0032] In this embodiment, the Lagrange function is specifically formulated as follows:

[0033] In the formula, For Lagrange multipliers, To give the first The cable section is allocated the first Various incentive levels, Number of segments For section Optional number of incentive levels, For profit, For cost, Due to total cost constraints, It is a Lagrange function.

[0034] It should be noted that the Unified Discriminative Causal Forest is a machine learning model specifically designed to estimate the effects of heterogeneous treatments, i.e., the differences in the effects of a certain intervention among different individuals or groups. This model estimates the effects by constructing multiple causal trees and integrating their results, introducing a unified splitting framework that allows all different treatments to be compared in the same feature space. This overcomes the problem of effect estimation bias and computational complexity caused by traditional methods that require training a separate model for each treatment.

[0035] It should be noted that the MCKP problem is a multi-choice knapsack problem, which is an NP-hard combinatorial optimization problem. In this problem, the items to be chosen are divided into several non-overlapping groups. The goal is to select exactly one item from each group to put into the knapsack, maximizing the total value of the selected items while ensuring that the total weight of all items does not exceed the capacity limit of the knapsack. Because the MCKP model can handle resource allocation scenarios with grouping and mutually exclusive choices very well, it has important applications in many fields.

[0036] It should be noted that this embodiment transforms the incentive allocation problem into an MCKP problem by constructing a unified discriminative causal forest model and solving it to obtain the incentive interval. This improves the evaluation accuracy and comparability of incentive responses for different cable sections, ensures the optimal allocation of global incentive strategies under limited resource constraints, and provides a foundation for obtaining adaptive incentives in the future.

[0037] S2, combined with the excitation interval, the pre-built constitutive model and the pre-built surrogate model, yields adaptive excitation.

[0038] In this embodiment, the process of obtaining adaptive excitation by combining the excitation interval, the pre-built constitutive model, and the pre-built surrogate model includes: An initial adaptive excitation parameter is preset within the excitation range, and the adaptive excitation parameter includes excitation intensity and excitation duration; The predicted responses under different excitation parameters are obtained based on the pre-built surrogate model, and the physical rationality of the predicted responses is verified based on the pre-built constitutive model. An incentive loss function is constructed based on the difference between the predicted response and the expected response preset based on historical data. Based on the activation loss function, the initial adaptive activation parameters are optimized and updated to generate the final adaptive activation parameters.

[0039] In this embodiment, obtaining the initial adaptive excitation parameters based on the excitation interval specifically involves selecting the initial adaptive excitation parameters within the excitation interval based on expert experience.

[0040] In this embodiment, the optimization and update of the initial adaptive excitation parameters specifically employs a meta-gradient optimization algorithm, including an inner loop and an outer loop.

[0041] In this embodiment, the outer loop is specifically: Based on the excitation range and historical data of cable operation, an initial adaptive excitation parameter is generated using a meta-learner; The initial adaptive excitation parameters are input into the inner loop to obtain the loss of the inner loop; The meta-gradient is calculated based on the loss of the inner loop, and the meta-learner is optimized based on the meta-gradient.

[0042] In this embodiment, the inner loop is specifically: The initial adaptive excitation parameters are input into the surrogate model for prediction, and the physical rationality of the surrogate model's prediction is verified by the constitutive model to obtain the predicted response. The difference between the predicted response and the expected response is calculated to obtain the incentive loss function; The adaptive excitation parameters are iteratively updated based on the excitation loss function.

[0043] In this embodiment, the expected response is specifically obtained by pre-setting based on historical experience.

[0044] It should be noted that the surrogate model is a computationally efficient and sufficiently accurate approximate model built on finite element model data. By using experimental design methods, representative sample points are selected in the possible value space of the stress state of the cable buffer layer. A high-precision finite element model is run to simulate and calculate these sample points to obtain the theoretical physical response of the shape memory material under the corresponding stress state, thereby obtaining a set of sample data. Machine learning is then used to train this data to obtain a surrogate model that can quickly predict the theoretical response of the material under any given stress state. Its calculation speed far exceeds that of the original finite element model, which requires complex numerical calculations.

[0045] It should be noted that a constitutive model is a mathematical description of the inherent physical laws of a material. Through a set of thermodynamic constitutive equations, it accurately describes the intrinsic relationship between the macroscopic response of a material and its microscopic state variables and external conditions. It is the theoretical basis for ensuring that the data conforms to physical laws.

[0046] It should be noted that meta-gradient is a higher-order gradient used to optimize the learning process itself. By calculating the gradient of the loss function of a basic task with respect to the hyperparameters, it guides the automatic adjustment of these hyperparameters, enabling the model to quickly adapt to new tasks. Through a two-layer optimization structure, meta-gradient can capture the commonalities between different tasks, giving the model a strong ability to adapt quickly across tasks and generalize.

[0047] It should be noted that this embodiment replaces the computationally complex finite element simulation with a surrogate model, while using the constitutive model as a strict constraint on physical laws. Through the division of labor and cooperation between inner and outer loops, meta-gradient optimization is adopted to achieve rapid prediction of the theoretical response vector, ensuring that the predicted response of the surrogate model does not deviate from the actual behavior of the material.

[0048] S3 applies adaptive excitation to the cable buffer layer to obtain the measured response vector of the physical properties of the shape memory material.

[0049] In this embodiment, applying adaptive excitation to the cable buffer layer to obtain the measured response vector of the shape memory material's physical properties includes: Preprocessing of the measured response data yields time-domain and frequency-domain features; The preprocessed measured response data are curve fitted to obtain the response curve, and the dynamic process characteristics of the response curve are extracted. The response vector is obtained by integrating time-domain features, frequency-domain features, and dynamic process features.

[0050] In this embodiment, the process of obtaining the response curve by curve fitting the preprocessed measured response data specifically employs a symbolic regression algorithm: Randomly generate an initial fitted curve and calculate the fitness of the initial response curve; Based on the fitness of the initial response curve, the mathematical expression of the initial fitted curve is optimized using a genetic algorithm to obtain the response curve.

[0051] In this embodiment, the fitness is the mean square error between the fitted curve and the response curve.

[0052] In this embodiment, the extraction of the dynamic process characteristics of the response curve specifically includes: The first derivative of the fitted response curve is used to obtain the rate of change curve. The peak value of the extracted rate curve corresponds to the fastest instantaneous rate during the recovery process, and the zero-crossing point of the rate curve corresponds to the inflection point of the original response curve, thus obtaining the dynamic process characteristics.

[0053] It should be noted that this embodiment preprocesses the original data and extracts time-domain and frequency-domain features, uses a symbolic regression algorithm for curve fitting, quantifies the key parameters of the recovery dynamics of the fitted curve, and finally integrates the time-domain, frequency-domain, and dynamic features into a response vector, providing an information-dense and physically meaningful input for subsequent optimization and inversion algorithms.

[0054] S4. Combining the measured response vector and the obtained theoretical response vector, the stress state of the cable buffer layer is optimized and inverted.

[0055] In this embodiment, the method for obtaining the theoretical response vector includes: Based on adaptive excitation and assumed stress state of the cable buffer layer, initial theoretical response data are obtained through a surrogate model; The theoretical response data is obtained by verifying and correcting the physical rationality of the initial theoretical response data using a constitutive model. Convert theoretical response data into theoretical response vectors.

[0056] In this embodiment, the step of optimizing the stress state of the cable buffer layer by combining the measured response vector and the obtained theoretical response vector includes: An inversion loss function is constructed based on the difference between the measured response vector and the theoretical response vector, with the optimization objective being to minimize the inversion loss function. Solving for the optimization objective yields the initial optimal solution for the stress state of the cable buffer layer; Starting with the initial optimal solution of the stress state, the final stress state of the cable buffer layer is obtained through local search optimization.

[0057] In this embodiment, the step of solving the optimization objective to obtain the initial optimal solution for the stress state of the cable buffer layer includes: An initial population of the stress state of the cable buffer layer was generated by Latin hypercube sampling. A hyperbolic cosine adaptive function is introduced to adjust the inertial weights of particles in the force state population. Based on the differential mutation strategy, the vector difference between particles in the force state population is introduced as a perturbation term into the particle position update formula. The initial force state population is iterated. When the iteration satisfies the preset global convergence condition, the optimal particle position in the current force state population is output as the initial optimal solution of the force state.

[0058] In this embodiment, the hyperbolic cosine adaptive function is introduced to adjust the inertial weights of particles in the force-state population. The specific formula is as follows:

[0059] In the formula, This represents the current iteration number. The maximum number of iterations, The normalization ratio of the algorithm process, and The upper and lower limits are preset for the inertia weight. It is a hyperbolic cosine function. These are the preset adjustment parameters.

[0060] In this embodiment, the differential mutation strategy is specifically formulated as follows:

[0061] In the formula, For the generated perturbation term, For the preset mutation operator, and Let be a vector of two random individuals.

[0062] In this embodiment, the specific formula for solving the optimization problem and updating the global position is as follows:

[0063] In the formula, This refers to the stress state of the cable buffer layer. The current optimal individual position, The current fragrance concentration of the individual. This represents the inertia weight for the current iteration.

[0064] It should be noted that the method for solving the optimization problem dynamically adjusts the inertial weights of particles in the stress state population through hyperbolic cosine functions, so that the algorithm maintains a large weight in the early stage of iteration to enhance global exploration. The difference between the stress states of the cable buffer layer is introduced as a perturbation term through the differential mutation strategy, which effectively increases the diversity of the stress state population and helps the algorithm escape local optima. Latin hypercube sampling ensures that the population is evenly distributed in the parameter space, thereby improving the global search efficiency.

[0065] It should be noted that Latin hypercube sampling is an efficient stratified sampling technique for multidimensional spaces. It divides the range of values ​​for each variable into several equally probable intervals, randomly selects a sample point in each interval, and then randomly combines the values ​​of these points to ensure that the samples are uniformly distributed in the multidimensional space. It can efficiently cover the entire parameter space with a small number of samples, significantly improving the efficiency and accuracy of calculations such as Monte Carlo simulations.

[0066] In this embodiment, the step of obtaining the final stress state of the cable buffer layer by starting with the initial optimal solution of the stress state and optimizing through local search includes: The initial optimal solution of the stress state is used as the initial point of the local search, and the iteration parameters are set. During the iteration process, the gradient of the inversion loss function corresponding to the current stress state of the cable buffer layer is calculated, and the local search direction is determined based on the gradient of the inversion loss function. A one-dimensional line search is performed along the local search direction to obtain a trial solution, and the corresponding trial inversion loss function value is calculated. Based on the trial-and-error inversion loss function value, the stress state and iterative parameters of the cable buffer layer are updated using the Metropolis acceptance criterion. The output at which the preset local convergence condition is met is taken as the final stress state of the cable buffer layer.

[0067] In this embodiment, the specific formula for determining the local search direction based on the gradient of the inversion loss function is as follows:

[0068] In the formula, For the first Quasi-Newton matrix of the next iteration. The change in gradient, This represents the change in the stress state of the cable buffer layer.

[0069] It should be noted that the local search optimization method in this embodiment approximates the Hessian matrix information of the objective function by constructing and updating a quasi-Newton matrix, thereby enabling the use of the curvature information of the loss function to determine a more accurate search direction. This results in superlinear convergence speed during the local search phase. By using the Metropolis acceptance criterion, it accepts temporary inferior solutions with a certain probability, thus possessing the ability to escape local optima, enhancing global exploration performance, and effectively suppressing the excessive dependence of the inversion results on the initial values.

[0070] It should be noted that the Metropolis acceptance criterion is a probabilistic acceptance rule. When a new state is better than the current state, it is always accepted. When a new state is worse, it is not rejected directly, but rather the decision to accept it is based on an acceptance probability. This criterion helps the algorithm escape local optima and conduct global exploration by accepting inferior solutions with a higher probability at high temperatures. As the temperature T gradually decreases, the probability of accepting inferior solutions decreases, allowing for a more refined local search.

[0071] It should be noted that this embodiment obtains the theoretical response vector by combining the surrogate model and the constitutive model, solves the assumed stress state of the cable buffer layer to obtain the initial optimal solution, and then uses a local search optimization method to perform a more accurate search to obtain the final stress state of the cable buffer layer, thus achieving a balance between global exploration capability and local development accuracy.

[0072] Example 2: This example provides a computer electronic device, such as... Figure 2 As shown, the electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the stress analysis method for the power cable buffer layer.

[0073] The processor is the control core of the electronic device. It connects various components of the electronic device through various interfaces and lines. It performs various functions of the electronic device and processes data by running or executing programs or modules stored in the memory and calling data stored in the memory.

[0074] The figure only shows an electronic device with components. Those skilled in the art will understand that the structure shown in the figure does not constitute a limitation on the electronic device and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0075] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0076] The computer program stored in the memory of the electronic device is a combination of multiple instructions. When run in the processor, it can implement the steps in the above-mentioned power cable buffer layer stress analysis method.

[0077] Specifically, the implementation system of the processor for the above instructions can be referred to the description of the relevant steps in the corresponding embodiments of the accompanying drawings, which will not be repeated here.

[0078] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0079] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0080] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0081] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0082] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0083] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for analyzing the stress on a buffer layer of a power cable, characterized in that, Includes the following steps: Shape memory material is embedded in the cable buffer layer, and the excitation range of different cable sections is obtained through machine learning based on the acquired historical data of cable operation. Adaptive excitation is obtained by combining the excitation interval, the pre-constructed constitutive model, and the pre-constructed surrogate model; An adaptive excitation is applied to the cable buffer layer to obtain the measured response vector of the physical properties of the shape memory material; By combining the measured response vector and the obtained theoretical response vector, the stress state of the cable buffer layer is optimized and inverted.

2. The method for analyzing the stress on the buffer layer of a power cable according to claim 1, characterized in that, The excitation intervals for different cable sections are obtained through machine learning based on the acquired historical data of cable operation, including: A unified discriminant causal forest model is constructed based on the acquired historical data of cable operation. The response effect matrix of different cable sections under different excitation levels is obtained based on the unified discriminant causal forest model. Based on the response effect matrix, we construct the MCKP problem of assigning optimal excitation to each cable segment; Solve the MCKP problem and convert the solution of the MCKP problem into excitation intervals for different cable sections.

3. The method for analyzing the stress on the buffer layer of a power cable according to claim 2, characterized in that, The unified discriminative causal forest model constructed based on the acquired historical data of cable operation includes: Historical data is randomly divided into a tree-construction set and an estimation set; Based on the tree set, several causal trees are constructed to form a causal forest by using the maximization of the difference in response to excitation among different cable sections as the splitting criterion. Obtain the rate of change of physical properties of cable segments falling into the same leaf node under different excitation levels, and use the difference in the mean of the rate of change as the conditional averaging effect of that leaf node.

4. The method for analyzing the stress on the buffer layer of a power cable according to claim 3, characterized in that, The adaptive excitation obtained by combining the excitation interval, the pre-constructed constitutive model, and the pre-constructed surrogate model includes: An initial adaptive excitation parameter is preset within the excitation range, and the adaptive excitation parameter includes excitation intensity and excitation duration; The predicted responses under different excitation parameters are obtained based on the pre-built surrogate model, and the physical rationality of the predicted responses is verified based on the pre-built constitutive model. An incentive loss function is constructed based on the difference between the predicted response and the expected response preset based on historical data. Based on the activation loss function, the initial adaptive activation parameters are optimized and updated to generate the final adaptive activation parameters.

5. The method for analyzing the stress on the buffer layer of a power cable according to claim 4, characterized in that, The process of applying adaptive excitation to the cable buffer layer and obtaining the measured response vector of the physical properties of the shape memory material includes: Preprocessing of the measured response data yields time-domain and frequency-domain features; The preprocessed measured response data are curve fitted to obtain the response curve, and the dynamic process characteristics of the response curve are extracted. The response vector is obtained by integrating time-domain features, frequency-domain features, and dynamic process features.

6. The method for analyzing the stress on the buffer layer of a power cable according to claim 5, characterized in that, The method for obtaining the theoretical response vector includes: Based on adaptive excitation and assumed stress state of the cable buffer layer, initial theoretical response data are obtained through a surrogate model; The initial theoretical response data is physically validated using a constitutive model, and the theoretical response data is then corrected accordingly. Convert theoretical response data into theoretical response vectors.

7. The method for analyzing the stress on the buffer layer of a power cable according to claim 6, characterized in that, The optimization and inversion of the stress state of the cable buffer layer by combining the measured response vector and the obtained theoretical response vector includes: An inversion loss function is constructed based on the difference between the measured response vector and the theoretical response vector, with the optimization objective being to minimize the inversion loss function. Solving for the optimization objective yields the initial optimal solution for the stress state of the cable buffer layer; Starting with the initial optimal solution of the stress state, the final stress state of the cable buffer layer is obtained through local search optimization.

8. The method for analyzing the stress on the buffer layer of a power cable according to claim 7, characterized in that, The process of solving the optimization objective to obtain the initial optimal solution for the stress state of the cable buffer layer includes: An initial population of the stress state of the cable buffer layer was generated by Latin hypercube sampling. A hyperbolic cosine adaptive function is introduced to adjust the inertial weights of particles in the force state population. Based on the differential mutation strategy, the vector difference between particles in the force state population is introduced as a perturbation term into the particle position update formula. The initial force state population is iterated. When the iteration satisfies the preset global convergence condition, the optimal particle position in the current force state population is output as the initial optimal solution of the force state.

9. The method for analyzing the stress on the buffer layer of a power cable according to claim 8, characterized in that, The process of obtaining the final stress state of the cable buffer layer through local search optimization, starting from the initial optimal solution of the stress state, includes: The initial optimal solution of the stress state is used as the initial point of the local search, and the iteration parameters are set. Calculate the gradient of the inversion loss function corresponding to the current stress state of the cable buffer layer, and determine the local search direction based on the gradient of the inversion loss function. A linear search is performed along the local search direction to obtain a trial solution, and the corresponding trial inversion loss function value is calculated. Based on the trial-and-error inversion loss function value, the stress state and iterative parameters of the cable buffer layer are updated using the Metropolis acceptance criterion. When the preset local convergence condition is met during iteration, the final stress state of the cable buffer layer is output.

10. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the stress analysis method for a power cable buffer layer as described in any one of claims 1 to 9.

Images