A Prediction Method and System for the Aerodynamic Characteristics of Stay Cables Based on PSO-GRU

Abstract

The present invention discloses a method and system for predicting the aerodynamic characteristics of stay cables based on PSO-GRU. The method includes: cyclically optimizing the key hyperparameters of the GRU network according to the particle swarm algorithm and calculating the fitness function to obtain the optimal parameters, and constructing a PSO-GRU model according to the optimal parameters; obtaining a plurality of stay cable data, and performing weighted fusion on the plurality of stay cable data according to a preset multi-channel information weighted fusion model to obtain target stay cable data, inputting the target stay cable data into the PSO-GRU model, and using the quantile regression algorithm to obtain the point prediction probability intervals at different quantiles. The PSO-GRU model outputs the aerodynamic characteristics of the stay cables corresponding to the target stay cable data, which can effectively predict the aerodynamic characteristics of stay cables under complex working conditions and improve the prediction accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of cable-stayed cable detection, and in particular relates to a method and system for predicting aerodynamic characteristics of a cable based on PSO-GRU. Background Art

[0002] As the core load-bearing component of a cable-stayed bridge, the cable is responsible for transferring the bridge deck load to the main tower. Its performance is directly related to the overall stability and safety of the cable-stayed bridge. The cable itself has a large slenderness ratio, which makes it relatively slender in geometry. When subjected to external loads, it is easy to produce large deformations. The cable has a small stiffness, which means that its ability to resist deformation is relatively weak. A small external force may cause a significant displacement of the cable. The cable has low damping, which results in a limited ability of the cable to consume vibration energy when vibration occurs. The vibration decays slowly, and it is easy to produce large-amplitude and long-duration vibrations, posing a potential threat to the safety of the bridge structure. In the research of cable-stayed bridges, In research and engineering practice, the aerodynamic characteristics of cables under complex working conditions is a very challenging and critical research field. The climatic conditions of the geographical location of cable-stayed bridges are complex and diverse, which adds many variables to the study of aerodynamic characteristics of cables, such as seasonal temperature differences and icing conditions. Changes in wind direction will change the relative angle between the cables and the airflow, thereby affecting the size and direction of aerodynamic forces. Fluctuations in wind speed will not only change the amplitude of aerodynamic forces, but may also induce different types of aerodynamic vibrations, such as vortex-induced vibrations and flutter. Different wind field characteristics, such as uniform flow and turbulent flow, also have completely different effects on the aerodynamic characteristics of cables. These complex aerodynamic conditions are intertwined, making it extremely difficult to accurately grasp the aerodynamic characteristics of cables. Summary of the invention

[0003] The present invention provides a method and system for predicting aerodynamic characteristics of cables based on PSO-GRU, which overcomes several shortcomings of traditional aerodynamic characteristics prediction, improves the accuracy of aerodynamic characteristics prediction of cables and the precision of classification, and is conducive to predicting aerodynamic characteristics of cables under complex working conditions.

[0004] In a first aspect, the present invention provides a method for predicting aerodynamic characteristics of a cable based on PSO-GRU, comprising:

[0005] According to the particle swarm algorithm, the key hyperparameters of the GRU network are cyclically optimized and the fitness function is calculated to obtain the optimal parameters, and a PSO-GRU model is constructed according to the optimal parameters;

[0006] Acquire multiple cable data, and perform weighted fusion on the multiple cable data according to a preset multi-channel information weighted fusion model to obtain target cable data;

[0007] Input the target cable data into the PSO-GRU model, and use the quantile regression algorithm to obtain the point prediction probability intervals at different quantiles. The PSO-GRU model outputs the aerodynamic characteristics of the cable corresponding to the target cable data.

[0008] Second, the present invention provides a cable aerodynamic characteristic prediction system based on PSO-GRU, including:

[0009] An optimization module configured to cyclically optimize the key hyperparameters of the GRU network according to the particle swarm algorithm, calculate the fitness function, obtain the optimal parameters, and construct a PSO-GRU model according to the optimal parameters;

[0010] A fusion module configured to obtain multiple cable data, and perform weighted fusion on the multiple cable data according to a preset multi-channel information weighted fusion model to obtain target cable data;

[0011] An output module configured to input the target cable data into the PSO-GRU model, use the quantile regression algorithm to obtain the point prediction probability intervals at different quantiles, and the PSO-GRU model outputs the aerodynamic characteristics of the cable corresponding to the target cable data.

[0012] Third, an electronic device is provided, which includes: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the steps of the cable aerodynamic characteristic prediction method based on PSO-GRU according to any embodiment of the present invention.

[0013] Fourth, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the program instructions are executed by a processor, the processor is enabled to execute the steps of the cable aerodynamic characteristic prediction method based on PSO-GRU according to any embodiment of the present invention.

[0014] The cable aerodynamic characteristic prediction method and system based on PSO-GRU of the present application obtain multiple cable data, perform weighted fusion on the multiple cable data according to a preset multi-channel information weighted fusion model to obtain target cable data, input the target cable data into the PSO-GRU model, use the quantile regression algorithm to obtain the point prediction probability intervals at different quantiles, and the PSO-GRU model outputs the aerodynamic characteristics of the cable corresponding to the target cable data, which can effectively predict the aerodynamic characteristics of the cable under complex working conditions and improve the prediction accuracy. Description of the Drawings

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0016] Figure 1 It is a flowchart of a cable aerodynamic characteristic prediction method based on PSO-GRU provided by an embodiment of the present invention;

[0017] Figure 2 It is a structural block diagram of a cable aerodynamic characteristic prediction system based on PSO-GRU provided by an embodiment of the present invention;

[0018] Figure 3 It is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0020] Please refer to Figure 1 , which shows a flowchart of a cable aerodynamic characteristic prediction method based on PSO-GRU of the present application.

[0021] As Figure 1 shown, the cable aerodynamic characteristic prediction method based on PSO-GRU specifically includes the following steps:

[0022] Step S101, circularly optimize the key hyperparameters of the GRU network according to the particle swarm algorithm and calculate the fitness function to obtain the optimal parameters, and construct a PSO-GRU model according to the optimal parameters.

[0023] In this step, during the iteration process, each particle will adjust its travel path according to the best position it identifies and the best position recognized as a whole. As the iteration progresses, the particle will update its own coordinates according to its momentum. Use to represent the best position found by the th particle, and to represent the best position among all particles. represents the number of iterations. Then, in the rd iteration, the th particle is at the The expression of the velocity change in the dimension is as follows:

[0024] ,

[0025] In the formula, represents the velocity of the -th particle in the -th dimension at the -th iteration. is the inertia weight, which affects the continuity of the particle velocity. and are the individual learning factor and the group learning factor respectively, and their values are in the range of (0, 2). and are two independent random numbers with values in the range of [0, 1]. represents the position of the -th particle in the -th dimension at the -th iteration. The expression for updating the particle position is as follows:

[0026]

[0027] In the formula, is the position of the +1-th particle in the -th dimension at the -th iteration. is the position of the -th particle in the -th dimension at the -th iteration. is the position of the -th particle in the -th dimension at the -th iteration;

[0028] Set the minimum value of the fitness function , that is:

[0029] ,

[0030] In the formula, is the aerodynamic characteristic weight. When the fitness value is lower, the solution to this problem is closer to the optimal solution.

[0031] It should be noted that according to the parametric biharmonic optimization decision equation, the key hyperparameter mapping relationship of the GRU network is constructed. Considering complex working condition factors and constraint conditions, parameters related to the decision are set to obtain the optimal parameters. The expression of the parametric biharmonic optimization decision equation is as follows:

[0032] ,

[0033] ,

[0034] ,

[0035] ,

[0036] ,

[0037] wherein, is the biharmonic operator, is the feasible region projection operator, is the sensitivity operator, is the parametric biharmonic optimization decision equation, is the parametric modulus threshold, is the parametric modulus accuracy, is the parametric modulus bias term, is the parametric weight, is the parametric sequence resolution, is the parametric sequence local resolution, is the parametric sequence displacement resolution, is the hyperparameter, is the hyperparameter prediction component, is the hyperparameter actual component, is the hyperparameter mixing component, is the hyperparameter weight, is the hyperparameter predicted forgetting data, is the hyperparameter actual forgetting data, is the hyperparameter mixing forgetting data, is the hyperparameter data sequence, is the hyperparameter prediction error, is the hyperparameter actual error, is the hyperparameter mixing error, is the parametric modulus threshold weight, is the parametric modulus threshold correction parameter, is the parametric modulus threshold sensitivity, is the parametric modulus threshold mean square error, is the parametric modulus threshold correlation coefficient, is the parametric modulus threshold smoothing factor, is the parametric modulus threshold correction factor, is the parametric modulus threshold high-order mode, is the parametric modulus threshold low-order mode, is the parametric modulus threshold hole width, is the parametric modulus threshold mixed-order mode, is the parametric modulus accuracy time entropy sequence, is the parameter modulus precision energy entropy sequence, is the weight of the parameter modulus precision sub-signal time entropy sequence, is the parameter modulus precision angle sequence, is the weight of the parameter modulus precision mixed signal time entropy sequence, is the parameter modulus precision numerical sequence, is the discrete coefficient of the parameter modulus offset term, is the discrete weight of the parameter modulus offset term, is the discrete angle of the parameter modulus offset term, is the discrete gradient of the parameter modulus offset term, is the discrete boundary threshold of the parameter modulus offset term, is the node value of the parameter modulus offset term, is the boundary value of the parameter modulus offset term, is the error of the parameter modulus offset term, is the main index of the parameter modulus offset term, is the total index of the parameter modulus offset term, is the first-order index of the parameter modulus offset term, is the second-order index of the parameter modulus offset term.

[0038] Specifically, constructing the PSO-GRU model according to the optimal parameters includes:

[0039] Extracting the deep correlation relationship between different working conditions and the aerodynamic characteristics of the cable according to the multifractal nonlinear cable aerodynamic characteristic correlation strategy, and training the GRU network configured with the optimal parameters according to the deep correlation relationship to obtain the PSO-GRU model. The expression of the multifractal nonlinear cable aerodynamic characteristic correlation strategy is:

[0040] ,

[0041] In the formula, is the multifractal nonlinear cable aerodynamic characteristic correlation strategy, is the aerodynamic feature extractor, is the nonlinear mapping function, is the lower bound of the label threshold, is the upper bound of the label threshold, is the lower bound of the domain discriminator, is the upper bound of the domain discriminator;

[0042] Among them, ,

[0043] ,

[0044] ,

[0045] In the formula, represents the objective function for pneumatic feature extraction, represents the single interval of the function, represents the mixed interval of the function, represents the pneumatic feature weight, represents the long vector, represents the non - linear feature mapping coefficient, represents the non - linear pneumatic feature coefficient, represents the linear feature mapping coefficient, represents the linear pneumatic feature coefficient, represents the adversarial bias term, represents the multi - branch bias term, represents the reverse bias term, represents the superposition bias term, represents the non - linear pneumatic feature regularization coefficient, represents the non - linear pneumatic feature Boolean coefficient, represents the non - linear pneumatic feature significance coefficient, represents the non - linear pneumatic feature original coefficient, represents the non - linear pneumatic feature prediction coefficient, represents the non - linear pneumatic feature density coefficient, represents the non - linear pneumatic feature random variable coefficient, represents the non - linear pneumatic feature reliability coefficient, represents the linear pneumatic feature bounding coefficient, represents the linear pneumatic feature distribution vector, represents the linear pneumatic feature singular vector, represents the linear pneumatic feature one - dimensional standard vector, represents the linear pneumatic feature state vector, is the torque, is the drag coefficient.

[0046] Step S102: Obtain a plurality of cable data, and perform weighted fusion on the plurality of cable data according to a preset multi - channel information weighted fusion model to obtain target cable data.

[0047] In this step, a plurality of similar vibration sensors collect data simultaneously to construct a data set. The data collected by each sensor is normalized to dimensionless data, providing conditions for subsequent weighted fusion. The expression of the dimensionless data is:

[0048] ,

[0049] In the formula, is the dimensionless data, Data collected by sensor i within time T; is the mean value of, is the maximum value of, is the minimum value of;

[0050] Calculate the correlation coefficient value of and during this time period, then the correlation support degree between the two time data series can be defined as:

[0051] ,

[0052] In the formula, is the Pearson correlation coefficient of variables X and Y; is the covariance of variables X and Y; represents the variance of variable X; represents the variance of variable Y. Further construct the correlation support degree matrix S, and the expression is:

[0053] ,

[0054] In the formula, is the correlation support degree of the m-th row and m-th column;

[0055] Therefore, within the time period T, the support degree of all sensors except sensor i to sensor i is:

[0056] ,

[0057] This support degree represents the correlation degree between each sensor data and other sensor data. Within the entire detection time period T, it is necessary to evaluate the reliability of the vibration sensor's own node to ensure the accuracy and stability of the detection result. Then, the mean value and standard deviation of the data of the i-th sensor node within the detection time period T are defined as:

[0058] ,

[0059] ,

[0060] Define the variation factor as:

[0061] The expression of the final weighted fusion is:

[0062] ,

[0063] ,

[0064] In the formula, is the final weighted fusion expression, is the weighted value corresponding to the time series data collected by sensor node i within the detection time period T. Data with high support and high self-reliability should be given a greater weight, is the number of sensor nodes, is dimensionless data, is the support degree of all sensors except sensor i for sensor i, is the mutation factor, is the mean value of the data of the i-th sensor node within the defined detection time period T, is the standard deviation, is the number of acquisition sequences, is the data collected by sensor i within time T, is the sensor node data, is the correlation support degree between two time data sequences.

[0065] An anomaly data filtering framework (ADFF) guided by topological information is established. By suppressing the anomaly data collected by sensors, using topological information to clean the anomaly data, and using an anomaly judgment and filtering module, the Euclidean structure features with enhanced correlation and the graph topological features of non-Euclidean structures are feature-fused, effectively improving the accuracy of anomaly recognition for complex data forms and dynamically optimizing the high-quality output. The expression of the topological information-guided anomaly data filtering framework is:

[0066] ,

[0067] ,

[0068] ,

[0069] ,

[0070] ,

[0071] In the formula, is the anomaly data filtering equation guided by topological information, is the data margin, is the strong impact data component, is the band energy, is the third-order fault data, is the single data component, is the interference data weight, is the interference data component, is the fuzzy data weight, is the fuzzy data component, is the data component sequence, is the peak value of strong impact data, is the effective value of strong impact data, is the weight value of the data energy envelope component, is the modal mixing component, is the envelope transition component, is the modal aliasing weight, is the number of classifications, is the data entropy, is the data scatter entropy, is the modal overlap component, is the number of overlaps, is the number of data, is the coarse-grained sequence, is the probability average value, is the embedding component, and are both fault information, is the numerical accuracy of the fault data, is the numerical singular value of the fault data, is the extended fault numerical energy ratio, is the data density, is the neighborhood density.

[0072] Step S103, input the target cable data into the PSO-GRU model, and use the quantile regression algorithm to obtain the point prediction probability interval at different quantiles. The PSO-GRU model outputs the cable aerodynamic characteristics corresponding to the target cable data.

[0073] In this step, quantile regression is to regress the independent variable based on the conditional quantile of the dependent variable to obtain the regression model at the given quantile. Assume the independent variable , the dependent variable , the expression form of the linear quantile regression model is:

[0074] ,

[0075] In the formula, is the th conditional quantile of the dependent variable, ranges from (0, 1), is the i-dimensional vector, is the vector of regression coefficients. When the training data of the model is known, find the vector of regression coefficients at different quantile points; The problem of

[0076] ,

[0077] In the formula, is an absolute value function of inclination, and its calculation formula is:

[0078] ,

[0079] ,

[0080] Since the linear quantile regression model is only applicable to studying the linear relationship between independent variables and dependent variables, while the actual wind power data has more non-linear relationships:

[0081] ,

[0082] In the formula, and are regularization parameters, v and w are weight parameters; i is the node of the network input layer, k is the node of the hidden layer, and the non-linear quantile regression loss function of the PSO-GRU model is obtained as:

[0083] ,

[0084] In the formula, is the non-linear quantile regression loss function, is the weight parameter, is the quantile point, is the element-wise minimum value, are different quantile points, is the response correction amount, is the parameter model compensation, is the fuzzy existence quantification.

[0085] In summary, the method of this application obtains multiple cable data, performs weighted fusion on the multiple cable data according to a preset multi-channel information weighted fusion model to obtain target cable data, inputs the target cable data into the PSO-GRU model, and uses the quantile regression algorithm to obtain the point prediction probability interval at different quantiles. The PSO-GRU model outputs the cable aerodynamic characteristics corresponding to the target cable data, which can effectively predict the cable aerodynamic characteristics under complex working conditions and improve the prediction accuracy.

[0086] Please refer to Figure 2 , which shows the structural block diagram of a cable aerodynamic characteristic prediction system based on PSO-GRU of this application.

[0087] As Figure 2 shown, the cable aerodynamic characteristic prediction system 200 includes an optimization module 210, a fusion module 220, and an output module 230.

[0088] Among them, the optimization module 210 is configured to cyclically optimize the key hyperparameters of the GRU network according to the particle swarm algorithm, calculate the fitness function, obtain the optimal parameters, and construct a PSO-GRU model according to the optimal parameters;

[0089] The fusion module 220 is configured to obtain multiple cable data and perform weighted fusion on the multiple cable data according to a preset multi-channel information weighted fusion model to obtain target cable data;

[0090] The output module 230 is configured to input the target cable data into the PSO-GRU model, and use the quantile regression algorithm to obtain the point prediction probability intervals at different quantiles. The PSO-GRU model outputs the cable aerodynamic characteristics corresponding to the target cable data.

[0091] It should be understood that Figure 2 the modules described in Figure 1 correspond to the respective steps in the method described in Figure 2 Therefore, the operations, features, and corresponding technical effects described above for the method also apply to

[0092] In some other embodiments, the embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored. When the program instructions are executed by a processor, the processor is caused to execute the PSO-GRU-based cable aerodynamic characteristic prediction method in any of the above method embodiments;

[0093] As an implementation manner, the computer-readable storage medium of the present invention stores computer-executable instructions, and the computer-executable instructions are set as:

[0094] Cyclically optimize the key hyperparameters of the GRU network according to the particle swarm algorithm, calculate the fitness function, obtain the optimal parameters, and construct a PSO-GRU model according to the optimal parameters;

[0095] Obtain multiple cable data and perform weighted fusion on the multiple cable data according to a preset multi-channel information weighted fusion model to obtain target cable data;

[0096] Input the target cable data into the PSO-GRU model, and use the quantile regression algorithm to obtain the point prediction probability intervals at different quantiles. The PSO-GRU model outputs the cable aerodynamic characteristics corresponding to the target cable data.

[0097] A computer-readable storage medium may include a storage program area and a storage data area. Among them, the storage program area can store an operating system and application programs required for at least one function; the storage data area can store data created according to the use of the cable aerodynamic characteristic prediction system based on PSO-GRU, etc. In addition, the computer-readable storage medium may include high-speed random access memory, and may also include a memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the computer-readable storage medium may optionally include a memory remotely provided with respect to the processor, and these remote memories can be connected to the cable aerodynamic characteristic prediction system based on PSO-GRU through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0098] Figure 3 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention, as Figure 3 shown, the device includes: a processor 310 and a memory 320. The electronic device may further include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330, and the output device 340 may be connected through a bus or other means, Figure 3 taking the connection through the bus as an example. The memory 320 is the above-mentioned computer-readable storage medium. The processor 310 executes various functional applications and data processing of the server by running non-volatile software programs, instructions, and modules stored in the memory 320, that is, implements the cable aerodynamic characteristic prediction method based on PSO-GRU in the above method embodiment. The input device 330 can receive input digital or character information, and generate key signal inputs related to user settings and function controls of the cable aerodynamic characteristic prediction system. The output device 340 may include a display device such as a display screen.

[0099] The above electronic device can execute the method provided by the embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method. For technical details not described in detail in this embodiment, reference can be made to the method provided by the embodiment of the present invention.

[0100] As an implementation manner, the above electronic device is applied to a cable aerodynamic characteristic prediction system based on PSO-GRU and is used for a client, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0101] Circularly optimize the key hyperparameters of the GRU network according to the particle swarm optimization algorithm, calculate the fitness function, obtain the optimal parameters, and construct a PSO-GRU model according to the optimal parameters;

[0102] Obtain multiple cable data, and perform weighted fusion on the multiple cable data according to a preset multi-channel information weighted fusion model to obtain target cable data;

[0103] Input the target cable data into the PSO-GRU model, and use the quantile regression algorithm to obtain the point prediction probability intervals at different quantiles. The PSO-GRU model outputs the cable aerodynamic characteristics corresponding to the target cable data.

[0104] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course also by hardware. Based on this understanding, the above technical solutions, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiments.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A method for predicting aerodynamic characteristics of cables based on PSO-GRU, characterized in that: include: According to the particle swarm algorithm, the key hyperparameters of the GRU network are cyclically optimized and the fitness function is calculated to obtain the optimal parameters, and the PSO-GRU model is constructed according to the optimal parameters, wherein the key hyperparameters of the GRU network are cyclically optimized and the fitness function is calculated according to the particle swarm algorithm to obtain the optimal parameters including: According to the parameter double-harmonic optimization decision equation, the key hyperparameter mapping relationship of the GRU network is constructed, and the parameters related to the decision are set considering the complex working conditions and constraints to obtain the optimal parameters. The expression of the parameter double-harmonic optimization decision equation is: In the formula, is a biharmonic operator, is the feasible domain projection operator, is the sensitivity operator, is the parameter biharmonic optimization decision equation, is the parameter modulus threshold, is the parameter modulus accuracy, is the parameter modulus bias term, a 4 is the parameter weight, is the parameter sequence resolution, b 4 is the local resolution of the parameter sequence, c 4 is the parameter sequence displacement resolution, U is the hyperparameter, is the hyperparameter prediction component, is the actual component of the hyperparameter, is the hyperparameter mixture component, E is the hyperparameter weight, Δω m1 (k) is the hyperparameter predicting the forgotten data, Δω m2 (k) is the actual forgotten data of the hyperparameter, Δω m12 (k) is the hyperparameter mixed forgetting data, Δω(k) is the hyperparameter data sequence, R1 is the hyperparameter prediction error, R2 is the hyperparameter actual error, α λ is the parameter modulus threshold weight, A T is the parameter modulus threshold correction parameter, λ τ is the parameter modulus threshold sensitivity, R τ is the parameter modulus threshold mean square error, n τ is the parameter modulus threshold correlation coefficient, T s is the parameter modulus threshold smoothing factor, j τ is the parameter modulus threshold correction factor, M λ is the parameter modulus threshold high-order mode, k λ is the parameter modulus threshold hole width, is the parameter modulus precision time entropy sequence, is the parameter modulus precision energy entropy sequence, γ1 is the parameter modulus precision sub-signal time entropy sequence weight, is the parameter modulus precision angle sequence, γ2 is the parameter modulus precision mixed signal time entropy sequence weight, is the numerical sequence of parameter modulus precision, T hw is the parameter modulus bias term dispersion coefficient, J w is the parameter modulus bias discrete weight, θ w is the discrete angle of the parameter modulus bias term, is the discrete gradient of the parameter modulus bias term, C h is the discrete boundary threshold of the parameter modulus bias term, θ mr is the parameter modulus bias node value, is the parameter modulus bias term boundary value, G f is the parameter modulus bias error, T w,s is the main index of the parameter modulus bias term, T w,c is the total index of parameter modulus bias term, α w is the first-order index of the parameter modulus bias term, B w is the second-order index of the parameter modulus bias term; The constructing of the PSO-GRU model according to the optimal parameters comprises: According to the multi-fractal nonlinear cable aerodynamic characteristic association strategy, the deep correlation between different working conditions and the cable aerodynamic characteristics is extracted, and the GRU network configured with the optimal parameters is trained according to the deep correlation to obtain the PSO-GRU model. The expression of the multi-fractal nonlinear cable aerodynamic characteristic association strategy is: Where Δ1 is the multi-fractal nonlinear cable aerodynamic characteristics correlation strategy, is the aerodynamic feature extractor, is a nonlinear mapping function, is the lower bound of the label threshold, N i,1 is the upper bound of the label threshold, is the lower bound of the domain discriminator, is the upper bound of the domain discriminator; in Where D ψ Represents a single interval of function, L ψ represents the function mixed interval, A sin represents the aerodynamic characteristic weight, R ψ represents a long vector, μ0 represents the nonlinear feature mapping coefficient, L vac represents the nonlinear aerodynamic characteristic coefficient, μ1 represents the linear characteristic mapping coefficient, L met represents the linear aerodynamic characteristic coefficient, l ψ represents the adversarial bias term, e r represents the multi-branch bias term, L a represents the reverse bias term, r ψ represents the superimposed bias term, θ p represents the random variable coefficient of nonlinear aerodynamic characteristics, Z ω represents the reliability coefficient of nonlinear aerodynamic characteristics, represents the linear aerodynamic characteristic limit coefficient, c (j) represents the linear aerodynamic characteristic distribution vector, λ χ represents the linear aerodynamic characteristic singular vector, w μ(jk) Represents the one-dimensional standard vector of linear aerodynamic characteristics, w σ(jk) represents the linear aerodynamic characteristic state vector, K is the moment, p is the drag coefficient, ω ψ Represents the original coefficient of nonlinear aerodynamic characteristics; Acquire multiple cable data, and perform weighted fusion on the multiple cable data according to a preset multi-channel information weighted fusion model to obtain target cable data; The target cable data is input into the PSO-GRU model, and the point prediction probability intervals under different quantiles are obtained by using the quantile regression algorithm. The PSO-GRU model outputs the aerodynamic characteristics of the cable corresponding to the target cable data.

2. The method for predicting aerodynamic characteristics of a cable based on PSO-GRU according to claim 1, characterized in that: in, The expression of weighted fusion is: In the formula, is the final weighted fusion expression, ω i (T) is the weighted value corresponding to the time series data collected by sensor node i within the detection time period T. Data with high support and high reliability should be given a larger weight. m is the number of sensor nodes. i (T) is dimensionless data, γ i (T) is the support of all sensors except sensor i to sensor i, mufa i (T) is the variation factor, is the mean value of the data of the i-th sensor node in the detection time period T, σ i (T) is the standard deviation, n is the number of acquisition sequences, x i (t) is the data collected by sensor i within time T, z i (T) is the sensor node data, S ij is the correlation support between two time data series.

3. The method for predicting aerodynamic characteristics of a cable based on PSO-GRU according to claim 1, characterized in that: Before weighted fusion of the plurality of cable data is performed according to a preset multi-channel information weighted fusion model, the method further includes: The plurality of cable data are processed according to a preset abnormal data filtering framework guided by topology information, wherein the expression of the abnormal data filtering framework guided by topology information is: min f CN (M)+q, In the formula, f CN (M) is the abnormal data filtering equation guided by topological information, q is the data margin, y is the ij is the strong impact data component, α is the frequency band energy, β is the third-order fault data, r j is a single data component, α j is the interference data weight, is the interference data component, α k is the fuzzy data weight, is the fuzzy data component, y is the data component sequence, is the peak value of strong impact data, is the effective value of strong impact data, γ is the weight of data energy envelope component, is the modal mixing component, is the envelope transition component, μ is the modal aliasing weight, r is the number of classifications, U j is the data entropy, L j is the data spreading entropy, is the modal overlap component, r1 is the number of overlaps, δ is the number of data, and f j is a coarse-grained sequence, θ i is the probability mean, e j is the embedded component, I1 and I2 are both fault information, is the numerical accuracy of fault data, ω β is the numerical singular value of the fault data, A is the energy ratio of the extended fault value, n is the data density, and p″ is the neighborhood density.

4. The method for predicting aerodynamic characteristics of a cable based on PSO-GRU according to claim 1, characterized in that: The expression of the nonlinear quantile regression loss function of the PSO-GRU model is: In the formula, L(τ i ) is the nonlinear quantile regression loss function, W(τ i ) is the weight parameter, n is the number of quantiles, γ z (τ i ) is the element-wise minimum, τ i For different quantiles, y i is the response correction, f(x i ,W(τ i )) is the parameter model compensation, x i Quantify fuzzy existence.

5. A cable aerodynamic characteristics prediction system based on PSO-GRU according to the method of claim 1, characterized in that: include: The optimization module is configured to perform cyclic optimization on the key hyperparameters of the GRU network according to the particle swarm algorithm and calculate the fitness function to obtain the optimal parameters, and construct the PSO-GRU model according to the optimal parameters, wherein the cyclic optimization on the key hyperparameters of the GRU network according to the particle swarm algorithm and calculate the fitness function to obtain the optimal parameters include: According to the parameter double-harmonic optimization decision equation, the key hyperparameter mapping relationship of the GRU network is constructed, and the parameters related to the decision are set considering the complex working conditions and constraints to obtain the optimal parameters. The expression of the parameter double-harmonic optimization decision equation is: In the formula, is a biharmonic operator, is the feasible domain projection operator, is the sensitivity operator, is the parameter biharmonic optimization decision equation, is the parameter modulus threshold, is the parameter modulus accuracy, is the parameter modulus bias term, a 4 is the parameter weight, is the parameter sequence resolution, b 4 is the local resolution of the parameter sequence, c 4 is the parameter sequence displacement resolution, U is the hyperparameter, is the hyperparameter prediction component, is the actual component of the hyperparameter, is the hyperparameter mixture component, E is the hyperparameter weight, Δω m1 (k) is the hyperparameter predicting the forgotten data, Δω m2 (k) is the actual forgotten data of the hyperparameter, Δω m12 (k) is the hyperparameter mixed forgetting data, Δω(k) is the hyperparameter data sequence, R1 is the hyperparameter prediction error, R2 is the hyperparameter actual error, α λ is the parameter modulus threshold weight, A T is the parameter modulus threshold correction parameter, λ τ is the parameter modulus threshold sensitivity, R τ is the parameter modulus threshold mean square error, n τ is the parameter modulus threshold correlation coefficient, T s is the parameter modulus threshold smoothing factor, j τ is the parameter modulus threshold correction factor, M λ is the parameter modulus threshold high-order mode, k λ is the parameter modulus threshold hole width, is the parameter modulus precision time entropy sequence, is the parameter modulus precision energy entropy sequence, γ1 is the parameter modulus precision sub-signal time entropy sequence weight, is the parameter modulus precision angle sequence, γ2 is the parameter modulus precision mixed signal time entropy sequence weight, is the numerical sequence of parameter modulus precision, T hw is the parameter modulus bias term dispersion coefficient, J w is the parameter modulus bias discrete weight, θ w is the discrete angle of the parameter modulus bias term, is the discrete gradient of the parameter modulus bias term, C h is the discrete boundary threshold of the parameter modulus bias term, θ mr is the parameter modulus bias node value, is the parameter modulus bias term boundary value, G f is the parameter modulus bias error, T w,s is the main index of the parameter modulus bias term, T w,c is the total index of parameter modulus bias term, α w is the first-order index of the parameter modulus bias term, B w is the second-order index of the parameter modulus bias term; A fusion module is configured to obtain a plurality of cable data, and perform weighted fusion on the plurality of cable data according to a preset multi-channel information weighted fusion model to obtain target cable data; The output module is configured to input the target cable data into the PSO-GRU model, use the quantile regression algorithm to obtain the point prediction probability interval under different quantiles, and the PSO-GRU model outputs the cable aerodynamic characteristics corresponding to the target cable data.

6. An electronic device, characterized in that: include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images