一种复杂微气候下覆冰导线舞动的智能预测方法

Abstract

This invention discloses an intelligent prediction method for conductor galloping under complex microclimates. The method constructs a finite element model of the conductor, identifies the three-component force coefficients of the conductor in a static state, analyzes the dynamic characteristics of the finite element model, and uses the harmonic superposition method to generate a forced vibration displacement time history signal to construct a forced vibration system. This forced vibration system forces the conductor finite element model to vibrate, obtaining the flutter derivative of the conductor during each forced vibration to obtain a training dataset. A long short-term memory deep learning network is constructed and trained using the training dataset to establish a nonlinear aerodynamic reduced-order model. Based on the conductor finite element model and the nonlinear aerodynamic reduced-order model, a coupled finite element model of the wind-conductor-transmission tower system is constructed to predict the displacement of icy conductors under complex microclimates and determine whether the conductor gallops. This invention achieves accurate prediction of conductor galloping behavior, laying the foundation for ensuring the safe operation of transmission lines under harsh weather conditions.

Description

Technical Field

[0001] This invention relates to the field of power transmission line analysis technology, specifically to an intelligent prediction method for the galloping of icy power transmission lines under complex microclimates. Background Technology

[0002] When the cross-section of a power transmission conductor is covered with ice, it forms a non-circular cross-section in shapes such as crescent, fan, or crown. This alters the conductor's aerodynamic characteristics, resulting in negative damping characteristics under the excitation of a certain wind speed and angle of attack. Consequently, the conductor exhibits low-frequency, high-amplitude galloping.

[0003] Currently, to prevent major power grid accidents caused by conductor galloping due to icing, real-time monitoring of conductors is mainly conducted using measurement methods such as video image recognition, inertial sensors, and fiber optic sensors. Based on the monitoring data, a measured database of conductor icing and galloping under different microclimate environments is established using methods such as correlation analysis, principal component analysis, and geographic weighted regression. A prediction model is trained using machine learning algorithms with supervised classification (such as support vector machines, artificial neural networks, Adaboost algorithm, Bayesian methods, decision trees, and radial basis function classification). The measured data, including route angle, wind direction, wind speed, rainfall, and icing thickness, are used as the input layer, and the occurrence of conductor galloping is used as the output layer, forming a prediction method for transmission line galloping.

[0004] However, the current methods for predicting power line galloping are essentially still a supervised classification problem. This simple machine learning not only fails to consider the fluid memory effect, three-dimensional effect, calm wind effect, aerodynamic nonlinearity, and geometric nonlinearity of aerodynamic forces, making it difficult to uncover the root causes of galloping of icy power lines, but also fails to accurately predict the galloping behavior of icy power lines under complex microclimate conditions. Summary of the Invention

[0005] This invention addresses the problem that existing technologies struggle to accurately predict the behavior of iced conductors in complex microclimates. It proposes an intelligent prediction method for the galloping of iced conductors under complex microclimates. By predicting the displacement of iced conductors in complex environments, this method achieves accurate prediction of conductor galloping behavior, laying the foundation for ensuring the safe operation of transmission lines in harsh weather conditions.

[0006] The present invention specifically adopts the following technical solution:

[0007] A smart prediction method for icy conductor galloping under complex microclimates includes the following steps:

[0008] Step 1: Construct a finite element model of the conductor using fluid dynamics simulation software and identify the three-component force coefficients of the conductor in a static state;

[0009] Step 2: Analyze the dynamic characteristics of the finite element model of the conductor, use the harmonic superposition method to generate the forced vibration displacement time history signal, and construct a forced vibration system for strong wave conductor vibration.

[0010] Step 3: Use the forced vibration system to induce forced vibration of the conductor finite element model multiple times, obtain the flutter derivative of the conductor during each forced vibration, obtain multiple sets of training data, and construct a training dataset;

[0011] Step 4: Construct a long short-term memory deep learning network and train the long short-term memory deep learning network using the training dataset to establish a nonlinear aerodynamic order reduction model.

[0012] Step 5: Based on the conductor finite element model and the nonlinear aerodynamic order reduction model, construct a coupled finite element model of the wind-conductor-transmission tower system to predict the displacement of the icing conductor under complex microclimates and determine whether the conductor will gallop.

[0013] Preferably, step 1 specifically includes the following steps:

[0014] Step 1.1: Construct a finite element model of the conductor using fluid dynamics simulation software, and set the model parameters and computational domain range of the conductor finite element model. The model parameters include the characteristic width of the conductor cross-section, the distance of the conductor cross-section from the fluid inlet and the fluid outlet, and the maximum torsional amplitude of the conductor cross-section in the vertical direction. The computational domain range includes the vertical height and horizontal length of the computational domain.

[0015] Step 1.2: Mesh the computational domain of the finite element model of the conductor. Divide the computational domain into a rigid mesh region, a dynamic mesh region, and a static mesh region. Structured meshes are used in both the dynamic and static mesh regions. Unstructured meshes are used at the junction of the dynamic and static mesh regions. In the rigid mesh region, except for the surface layer of the conductor cross-section which uses a boundary layer mesh, all other areas use quadrilateral unstructured meshes.

[0016] Step 1.3: Using a finite element model of the conductor, the stress on the conductor under static wind load is simulated, and the three-component force coefficients of the conductor in the static state are identified as follows:

[0017]

[0018]

[0019]

[0020] In the formula, F D F represents the static wind load drag on the conductor cross-section in the flow field. L Let F be the lift force experienced by the conductor cross-section in the flow field. MLet F be the torque experienced by the conductor cross-section in the flow field, and F be the static wind load drag experienced by the conductor cross-section in the flow field. D The lift force F experienced by the conductor cross-section in the flow field L The torque F experienced by the conductor cross-section in the flow field M The three forces that together make up the conductor; C D C is the drag coefficient. L C is the lift coefficient. M For torque coefficient and drag coefficient C D Lift coefficient C L and torque coefficient C M The three force coefficients that together constitute the conductor are: ρ, U, H, and B. ρ is the fluid density, U is the incoming flow velocity, H is the characteristic height of the conductor cross-section, and B is the characteristic width of the conductor cross-section.

[0021] Preferably, in step 1, the finite element model of the conductor is an SST k-ω turbulence model, the thickness of the first layer of the finite element model falls within the viscous sublayer, the thickness of the first layer of the mesh is 0.02 mm, the number of boundary layer meshes of the finite element model is 30, the expansion rate is 1.2, the structured mesh is an outwardly diffusing annular surface, and the skewness is less than 0.5.

[0022] Preferably, in step 2, when the harmonic superposition method is used to simulate the forced vibration of the conductor, the forced displacement signal of the conductor cross-section is:

[0023]

[0024] In the formula, x(t) is the forced displacement signal of the conductor cross-section, t is time, and A st A is the specified amplitude of the forced displacement signal for the conductor cross-section. max n represents the original amplitude of the superimposed signal. m A represents the number of superimposed harmonics. i For the randomly generated i-th amplitude, ω i Let be the randomly generated i-th circular frequency.

[0025] Preferably, step 3 specifically includes the following steps:

[0026] Step 3.1: Establish a coordinate system with the center of the conductor cross-section of the conductor finite element model as the origin. Dynamically update the position coordinates of the mesh in the conductor finite element model based on the spring smoothing method. Combined with the moving mesh sub-region method, divide the moving mesh region into multiple sub-regions, set each sub-region to move independently, and obtain the displacement S of each sub-region within the moving mesh region. q As shown in formula (5):

[0027] S q =q·S (5)

[0028] In the formula, S q is the motion vector of the sub-region in the moving mesh region, q is the diffusion coefficient, and S is the mesh motion vector of the guide section;

[0029] Step 3.2: Forced vibration is induced in the finite element model of the conductor using a forced vibration system. Based on the split-state forced vibration method, the vertical and torsional vibrations of the conductor are simulated separately to obtain the simulated aerodynamic self-excited lift data L corresponding to the vertical vibration of the conductor cross-section. testh Pneumatic self-excited torque simulation data M testh And the aerodynamic self-excited lift simulation data L corresponding to the torsional vibration of the conductor cross section testα Pneumatic self-excited torque simulation data M testα The flutter derivative of the conductor is calculated.

[0030] Step 3.3: The finite element model of the conductor is induced to vibrate repeatedly using a forced vibration system to obtain the simulated aerodynamic self-excited lift data L corresponding to the vertical vibration of the conductor cross-section. testh Pneumatic self-excited torque simulation data M testh And the aerodynamic self-excited lift simulation data L corresponding to the torsional vibration of the conductor cross section testα Pneumatic self-excited torque simulation data M testα The flutter derivative of the conductor is calculated, and the flutter derivative of the conductor during each forced vibration is obtained. Multiple sets of training data are obtained, and a training dataset is constructed.

[0031] Preferably, in step 3.2, based on the vertical displacement and torsional degrees of freedom of the conductor, eight flutter derivatives are used to describe the aerodynamic lift and aerodynamic torque acting on the conductor, as shown in formulas (6) and (7):

[0032]

[0033]

[0034] In the formula, L se M is the aerodynamic lift force acting on the conductor cross-section. se This refers to the aerodynamic torque acting on the conductor cross-section; This represents the contribution of aerodynamic damping caused by vertical vibration to the self-excited lift. The contribution of aerodynamic damping caused by torsional vibration to the self-excited lift. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited lift. This represents the combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited lift. This represents the contribution of aerodynamic damping caused by vertical vibration to the self-excited torque. The contribution of aerodynamic damping caused by torsional vibration to the self-excited torque. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited torque. This represents the combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited torque. and Together they constitute the flutter derivative of the conductor; B is the characteristic width of the conductor cross-section; K is the reduced frequency, K=ωB / U, where ω is the angular frequency of the conductor's vibration;

[0035] When the finite element model of the conductor is induced to vibrate vertically using a forced vibration system, the conductor cross-section undergoes simple harmonic motion in the flow field during the vertical vibration process. The displacement and velocity of the conductor cross-section during the simple harmonic motion are shown in formulas (8) and (9):

[0036] h(t) = h0sin(ω) h t) (8)

[0037]

[0038] In the formula, t is time, h is the displacement of the conductor cross-section in the vertical vibration, and h0 is the amplitude of the conductor cross-section in the vertical vibration. Let ω be the velocity of the vertical vibration of the conductor cross-section. h The angular frequency of the vertical vibration of the conductor cross-section;

[0039] When the vertical vibration of the conductor section in the flow field reaches a steady state, the aerodynamic self-excited lift and self-excited torque are as shown in equations (10) and (11):

[0040]

[0041]

[0042] Discretizing equations (10) and (11) in time respectively, we obtain:

[0043]

[0044]

[0045] In the formula, L seh M is the theoretical aerodynamic lift force when the conductor cross-section vibrates vertically in the flow field. seh S is the column vector of aerodynamic torque self-excited forces when the conductor cross-section vibrates vertically in the flow field; H and H 14 All of these are intermediate calculation quantities of the theoretical aerodynamic lift corresponding to the vertical vibration of the conductor cross-section; S A and A 14 These are all intermediate calculation quantities of the column vector of aerodynamic torque self-excited force corresponding to the vertical vibration of the conductor cross section;

[0046] Based on the least squares principle, the simulation data L of the aerodynamic self-excited lift corresponding to the vertical vibration of the conductor cross-section is used. testh Theoretical aerodynamic lift L when the substituted conductor cross-section vibrates vertically in the flow field seh Simulation data M of the aerodynamic self-excited torque corresponding to the vertical vibration of the conductor cross-section testh The aerodynamic torque self-excited force column vector M of the alternative conductor cross-section in the flow field during vertical vibration seh Solving for the problem yields:

[0047] H 14 =(S H T S H ) -1 S H T L testh (14)

[0048] A 14 =(S A T S A ) -1 S A T M testh (15)

[0049] Based on the intermediate calculation of the theoretical aerodynamic lift H corresponding to the vertical vibration of the conductor cross-section. 14 Intermediate calculation quantity A of the aerodynamic torque self-excited force column vector 14 The contribution of aerodynamic damping caused by vertical vibration to the self-excited lift was calculated. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to self-excited lift. The contribution of aerodynamic damping caused by vertical vibration to the self-excited torque The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited torque.

[0050] When the torsional vibration of the conductor finite element model is induced by a forced vibration system, the displacement and velocity of the conductor cross section during the torsional vibration process are shown in Equations (16) and (17):

[0051] α(t)=α0sin(ω α t) (16)

[0052]

[0053] In the formula, α is the displacement of the conductor cross-section under torsional vibration, and α0 is the amplitude of the conductor cross-section under torsional vibration. Let ω be the velocity of the torsional vibration of the conductor cross section. α The angular frequency of the torsional vibration of the conductor cross-section;

[0054] When the conductor cross-section reaches a steady state of torsional vibration in the flow field, the aerodynamic self-excited lift and self-excited torque are as shown in equations (18) and (19):

[0055]

[0056]

[0057] Discretizing equations (18) and (19) in time respectively, we obtain:

[0058]

[0059]

[0060] In the formula, L seα M is the theoretical aerodynamic lift force when the conductor cross-section undergoes torsional vibration in the flow field. seα Q is the column vector of aerodynamic torque self-excited forces when the conductor cross-section undergoes torsional vibration in the flow field; H and H 23 Q is an intermediate calculation quantity for the theoretical aerodynamic lift corresponding to the torsional vibration of the conductor cross-section; A and A 23 This is an intermediate calculation quantity for the column vector of aerodynamic torque self-excited force corresponding to the torsional vibration of the conductor cross section;

[0061] Based on the least squares principle, using the aerodynamic self-excited lift simulation data L corresponding to the torsional vibration of the conductor cross section testα Theoretical aerodynamic lift L when the alternative conductor cross-section undergoes torsional vibration in the flow field seα Simulation data M of the aerodynamic self-excited torque corresponding to the torsional vibration of the conductor cross section testα The aerodynamic torque self-excited force column vector M of the alternative conductor cross section during torsional vibration in the flow field seα Solving for the problem yields:

[0062] H 23 =(Q H T Q H ) -1 Q H T L testα (twenty two)

[0063] A 23 =(Q A T Q A ) -1 Q A T M testα (twenty three)

[0064] Based on the intermediate calculation of the theoretical aerodynamic lift H corresponding to the torsional vibration of the conductor cross-section 23 Intermediate calculation quantity A of the aerodynamic torque self-excited force column vector 23 The contribution of aerodynamic damping caused by torsional vibration to the self-excited lift was calculated. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to self-excited lift The contribution of aerodynamic damping caused by torsional vibration to the self-excited torque The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited torque.

[0065] Preferably, step 4 specifically includes the following steps:

[0066] Step 4.1: Construct a long short-term memory deep learning network. The long short-term memory deep learning network includes a forget gate, an input gate, and an output gate.

[0067] The forget gate is used to control the discarding of the previous time-state s. t-1 The formula for calculating useless information within is:

[0068] f t =σ(W f [h t-1 ,x t ]+b f ) (twenty four)

[0069] In the formula, f t For the Gate of Oblivion, W f For the weight of the forget gate, b f σ is the bias of the forget gate, and σ is the sigmoid activation function;

[0070] The input gate contains a sigmoid layer and a tanh layer to determine the current state s of the input pair. t The contribution is calculated using the following formula:

[0071] i t =σ(W i [h t-1 ,x t ]+b i (25)

[0072] c t =tanh(W c [h t-1 ,x t ]+b c (26)

[0073] s t =f t·s t-1 +i t ·c t (27)

[0074] In the formula, i t For the input gate, W i b represents the weights of the sigmoid layer. i For the bias of the input gate sigmoid layer, c t For the candidate values ​​obtained from the input gate tanh layer, W c b represents the weights of the input gate tanh layer. c The bias of the input gate tanh layer;

[0075] The output gate is used to extract the current cell state s. t The output value h t Output value h t Or the current cell state s t The output value h t It is directly used as the input value for the next time step, and the calculation formula is:

[0076] o t =σ(W o [h t-1 ,x t ]+b o (28)

[0077] h t =o t ·tanh(s t (29)

[0078] In the formula, o t For output gate, W o b represents the weights of the output gate sigmoid layer. o h is the bias of the output gate sigmoid layer. t This is the output value at the current moment;

[0079] Step 4.2: Train the Long Short-Term Memory (LSTM) deep learning network using the training dataset. Combine the root mean square propagation method and stochastic gradient descent, and add a momentum term to optimize the parameter vector in the LTM deep learning network. The update formula for the parameter vector is:

[0080]

[0081] in,

[0082]

[0083]

[0084] In the formula, Let α be the number of updates, θ be the learning rate, θ be the vector of parameters to be optimized, E(·) be the loss function, β1 be the first decay rate, and β2 be the second decay rate. and All of these are updates to the calculation parameters;

[0085] Step 4.3: Using the trained Long Short-Term Memory deep learning network, establish a nonlinear aerodynamic order reduction model and calculate the aerodynamic self-excited lift L of the conductor. Δt and aerodynamic self-excited torque M Δt .

[0086] Preferably, step 5 specifically includes the following steps:

[0087] Step 5.1: Based on the structure of the conductor to be predicted, obtain the number of conductor nodes, establish a nonlinear aerodynamic reduced-order model for each conductor node, initialize each nonlinear aerodynamic reduced-order model to obtain the conductor-transmission tower finite element model, determine the multiple of the aerodynamic force applied per unit length of each conductor node based on the conductor node spacing, calculate the aerodynamic force corresponding to each conductor node, and multiply the aerodynamic force of each conductor node by the set multiple to obtain the wind load applied to the conductor.

[0088] Step 5.2: Using finite element analysis software, establish a coupled finite element model of the wind-conductor-transmission tower system, and calculate the displacement D of the conductor under zero wind speed steady state. static And used as the initial displacement;

[0089] Step 5.3: Using finite element analysis software, perform transient analysis on the coupled finite element model of the wind-conductor-transmission tower system to obtain the displacement of each conductor node in the coupled finite element model of the wind-conductor-transmission tower system in real time;

[0090] Step 5.4: Based on the displacement of each conductor node in the coupled finite element model of the wind-conductor-transmission tower system, calculate the aerodynamic force of each conductor node using a nonlinear aerodynamic force reduction model and multiply it by the set multiplier to form wind load data.

[0091] Step 5.5: Using finite element analysis software, adjust the Rayleigh damping of the coupled finite element model of the wind-conductor-transmission tower system so that the structural modal damping ratio of the coupled finite element model of the wind-conductor-transmission tower system is greater than 1.

[0092] Step 5.6: Calculate the aerodynamic forces using a reduced-order nonlinear aerodynamic model, and calculate the displacement D of the conductor in the coupled finite element model of the wind-conductor-transmission tower system after one time step Δt. Δt ;

[0093] Step 5.7, based on the conductor displacement D obtained in step 5.6 Δt The data is input into a long short-term memory deep learning network, and the aerodynamic parameters of each conductor node are recalculated using a nonlinear aerodynamic order reduction model. The aerodynamic parameters of each conductor node are then multiplied by a set multiplier to update the wind load data.

[0094] Step 5.8: Using the reactivation method, repeat steps 5.6 and 5.7 to obtain the time t at which the static wind effect stops in the coupled finite element model of the wind-conductor-transmission tower system. sw The corresponding conductor displacement is used to calculate the aerodynamic parameters when the calm wind effect stops, and then the aerodynamic parameters are updated.

[0095] Step 5.9: Adjust the Rayleigh damping of the coupled finite element model of the wind-conductor-transmission tower system so that the structural modal damping ratio of the coupled finite element model of the wind-conductor-transmission tower system becomes a normal value;

[0096] Step 5.10: Using the reactivation method, after the calm wind effect stops, continue to repeat steps 5.6 and 5.7 to make the conductor in the wind-conductor-transmission tower coupled finite element model vibrate under the calm wind steady state. Continue to update the aerodynamic parameters and wind load data according to the displacement of the conductor in the wind-conductor-transmission tower coupled finite element model, obtain the response time history curve of the conductor, and proceed to step 5.11.

[0097] Step 5.11: Perform flutter time-domain analysis on the conductor based on the conductor's response time history curve. If the amplitude of the conductor's response time history curve suddenly increases or diverges, it is determined that the conductor has galloped; otherwise, it is determined that the conductor has not galloped.

[0098] Preferably, in step 5, the aerodynamic parameters include aerodynamic self-excited lift and aerodynamic self-excited torque.

[0099] The present invention has the following beneficial effects:

[0100] This invention proposes an intelligent prediction method for the galloping of icy power lines under complex microclimates. It comprehensively considers three-dimensional effects, calm wind effects, aerodynamic nonlinearity, geometric nonlinearity, and the fluid memory effect of aerodynamics. The method uses fluid dynamics simulation software to construct a finite element model of the power line, extracts the three-component force coefficients and flutter derivatives of the icy power line, and establishes a nonlinear aerodynamic reduced-order model by training a long short-term memory neural network. It then constructs a coupled finite element model of the wind-power line-transmission tower system and combines it with a reactivation method to achieve accurate prediction of the galloping behavior of power lines under complex microclimates.

[0101] This invention, by deeply exploring the galloping patterns and mechanisms of icing conductors, combines fluid dynamics simulation, long short-term memory neural networks, and reactivation methods to improve the prediction accuracy and computational efficiency of conductor galloping behavior. This is beneficial for analyzing conductor galloping behavior under complex microclimates, providing early warning of conductor galloping behavior, and guiding emergency decision-making, thus laying the foundation for ensuring the safe operation of transmission lines in harsh weather conditions. Attached Figure Description

[0102] Figure 1 This is a schematic diagram of the structure of a long short-term memory deep learning network.

[0103] Figure 2 This is a graph showing the input-output mapping relationship of a long short-term memory deep learning network. Detailed Implementation

[0104] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and specific examples:

[0105] A smart prediction method for icy conductor galloping under complex microclimates includes the following steps:

[0106] Step 1: Construct a finite element model of the conductor using fluid dynamics simulation software, and identify the three-component force coefficients of the conductor under static conditions. This includes the following steps:

[0107] Step 1.1: Construct a finite element model of the conductor using fluid dynamics simulation software, and set the model parameters and computational domain of the conductor finite element model. The model parameters include the characteristic width of the conductor cross-section, the distance of the conductor cross-section from the fluid inlet and the fluid outlet, and the maximum torsional amplitude of the conductor cross-section in the vertical direction. The computational domain includes the vertical height and horizontal length of the computational domain.

[0108] In this embodiment, the Fluent software is used to construct the finite element model of the conductor. The blockage rate of the conductor finite element model in the vertical direction is less than 3%. The maximum torsional amplitude of the conductor cross-section in the vertical direction is set to 30°. The center of the conductor cross-section is located on the central horizontal line. The ratio of the distance of the conductor cross-section from the fluid inlet to the fluid outlet in the horizontal direction is 1:2. In order to ensure the blockage rate requirement can still be guaranteed in extreme cases, this embodiment sets the vertical height of the calculation domain to 18B and the horizontal length to 30B, ensuring that the wake has enough space to develop fully and that the impact of the airflow on the conductor cross-section will not affect the boundary conditions.

[0109] Step 1.2: Mesh the computational domain of the conductor finite element model. To maintain mesh shape consistency, quadrilateral meshes are used throughout the computational domain. The computational domain is divided into a rigid mesh region, a moving mesh region, and a static mesh region. The rigid mesh follows the fluid boundary in a rigid motion to ensure the mesh quality around the conductor cross-section and to ensure that the mesh features always meet the requirements of the turbulence model. The deformation caused by the movement of the conductor cross-section is handled by the larger-scale moving mesh region. The moving mesh region surrounding the rigid mesh region uses regular boundaries (ring boundaries) to facilitate the construction of a structured mesh and ensure its quality in subsequent deformation updates. The outermost static mesh does not participate in mesh movement and deformation.

[0110] Structured meshes are used in both the moving and static mesh regions of the computational domain, while unstructured meshes are used at the junction of the moving and static mesh regions to adapt to changes in the boundary shape. In the rigid mesh region, except for the surface layer of the conductor cross-section which uses a boundary layer mesh, all other regions use quadrilateral unstructured meshes.

[0111] To meet the mesh requirements of the SST k-ω turbulence model, the mesh thickness of the first layer of the ductile finite element model is set to 0.02 mm in this embodiment, thus ensuring that the first layer mesh thickness falls within the viscous sublayer. The boundary layer mesh of the ductile finite element model has 30 layers and an expansion rate of 1.2. The structured mesh is an outwardly expanding annular surface. The entire mesh is optimized using the Winslow algorithm, and the skewness is less than 0.5.

[0112] Step 1.3: Using a finite element model of the conductor, the stress on the conductor under static wind load is simulated, and the three-component force coefficients of the conductor in the static state are identified as follows:

[0113]

[0114]

[0115]

[0116] In the formula, F D F represents the static wind load drag on the conductor cross-section in the flow field. L Let F be the lift force experienced by the conductor cross-section in the flow field. M Let F be the torque experienced by the conductor cross-section in the flow field, and F be the static wind load drag experienced by the conductor cross-section in the flow field. D The lift force F experienced by the conductor cross-section in the flow field L The torque F experienced by the conductor cross-section in the flow field M The three forces that together make up the conductor; C D C is the drag coefficient. L C is the lift coefficient. M For torque coefficient and drag coefficient C DLift coefficient C L and torque coefficient C M The three force coefficients that together constitute the conductor are: ρ, U, H, and B. ρ is the fluid density, U is the incoming flow velocity, H is the characteristic height of the conductor cross-section, and B is the characteristic width of the conductor cross-section.

[0117] Step 2: Analyze the dynamic characteristics of the finite element model of the conductor. Based on the results of the dynamic characteristics of the finite element model, design a forced vibration condition with large amplitude to excite the nonlinear characteristics of aerodynamic forces, thereby ensuring the effectiveness and generality of the training dataset. Use the harmonic superposition method to generate the forced vibration displacement time history signal and construct a forced vibration system for strong wave conductor vibration.

[0118] The forced displacement signal of the conductor cross section adopts the form of multiple positive harmonic superposition, as shown in formula (4):

[0119]

[0120] In the formula, x(t) is the forced displacement signal of the conductor cross-section, t is time, and A st A is the specified amplitude of the forced displacement signal for the conductor cross-section. max n represents the original amplitude of the superimposed signal. m A represents the number of superimposed harmonics. i For the randomly generated i-th amplitude, ω i Let be the randomly generated i-th circular frequency.

[0121] This embodiment employs the harmonic superposition method to generate the forced vibration displacement time history signal. This method ensures that the forced vibration displacement time history signal contains a sufficiently rich frequency range and a sufficiently large amplitude, while also guaranteeing the continuity and differentiability of the displacement time sequence, thus avoiding convergence difficulties in computational fluid dynamics numerical simulations. Furthermore, the harmonic superposition method for generating the forced vibration displacement time history signal fully considers the characteristics of the increasing amplitude of wind-induced vibration due to the increasingly flexible conductor and the increasingly pronounced nonlinear characteristics of the conductor's aerodynamic self-excitation force. This method effectively demonstrates the variation of the flutter derivative with amplitude and the existence of higher-order components in the self-excitation force.

[0122] Step 3: The finite element model of the conductor is repeatedly subjected to forced vibration using a forced vibration system. The flutter derivative of the conductor during each forced vibration is obtained, resulting in multiple sets of training data. A training dataset is then constructed, which includes the following steps:

[0123] Step 3.1: Establish a coordinate system with the center of the conductor cross-section of the conductor finite element model as the origin. Dynamically update the position coordinates of the mesh in the conductor finite element model based on the spring smoothing method, by equating the mesh in the initial state to a set of interconnected springs in equilibrium.

[0124] When the boundary moves, the displacement of the boundary is equivalent to generating an elastic force on the spring system. The magnitude of the elastic force is proportional to the displacement. The spring needs to adjust the position of the nodes so that each node returns to a state of equilibrium. The stiffness k of the spring... ij It is inversely proportional to the side length of each grid, as shown in formula (33):

[0125]

[0126] In the formula, k ij Let x be the stiffness of the spring between node i and node j. i y i ) and (x j y j ) represents the coordinates of two adjacent nodes.

[0127] After the update, the net spring force in each node should be zero. The iterative update process is as follows:

[0128]

[0129]

[0130] In the formula, Let x be the x-coordinate update value of the m-th node. This represents the x-coordinate update value of the (m+1)th node. Let be the ordinate update value of the m-th node. This is the update value of the ordinate of the (m+1)th node.

[0131] To enable more meshes to participate in ordered deformation and effectively alleviate mesh distortion, this embodiment, based on the spring-light multi-layer sub-region method, also employs a moving mesh sub-region method. This involves dividing the moving mesh region into multiple sub-regions and setting each boundary to move independently. This allows each sub-region to absorb only a portion of the boundary deformation, thus achieving the goal of gradually spreading the boundary movement. The displacement S of each sub-region within the moving mesh region... q As shown in formula (5):

[0132] S q =q·S (5)

[0133] In the formula, S q is the motion vector of the sub-region in the moving grid region, q is the diffusion coefficient with a value of 0 to 1, and S is the grid motion vector of the guide section.

[0134] Step 3.2: Forced vibration is induced in the finite element model of the conductor using a forced vibration system. Based on the split-state forced vibration method, the vertical and torsional vibrations of the conductor are simulated separately to obtain the simulated aerodynamic self-excited lift data L corresponding to the vertical vibration of the conductor cross-section. testh Pneumatic self-excited torque simulation data Mtesth And the aerodynamic self-excited lift simulation data L corresponding to the torsional vibration of the conductor cross section testα Pneumatic self-excited torque simulation data M testα The flutter derivative of the conductor is calculated.

[0135] Due to the two degrees of freedom of the conductor, namely vertical displacement and torsion, the aerodynamic lift and aerodynamic torque on the conductor are described by eight flutter derivatives, as shown in formulas (6) and (7):

[0136]

[0137]

[0138] In the formula, L se M is the aerodynamic lift force acting on the conductor cross-section. se This refers to the aerodynamic torque acting on the conductor cross-section; This represents the contribution of aerodynamic damping caused by vertical vibration to the self-excited lift. The contribution of aerodynamic damping caused by torsional vibration to the self-excited lift. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited lift. This represents the combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited lift. This represents the contribution of aerodynamic damping caused by vertical vibration to the self-excited torque. The contribution of aerodynamic damping caused by torsional vibration to the self-excited torque. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited torque. This represents the combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited torque. and Together they constitute the flutter derivative of the conductor; B is the characteristic width of the conductor cross-section; K is the reduced frequency, K=ωB / U, where ω is the angular frequency of the conductor's vibration.

[0139] When the finite element model of the conductor is induced to vibrate vertically using a forced vibration system, the conductor cross-section undergoes small simple harmonic motion in the flow field during the vertical vibration process. The displacement and velocity of the conductor cross-section during the simple harmonic motion are shown in equations (8) and (9):

[0140] h(t) = h0sin(ω) h t) (8)

[0141]

[0142] In the formula, t is time, h is the displacement of the conductor cross-section in the vertical vibration, and h0 is the amplitude of the conductor cross-section in the vertical vibration. Let ω be the velocity of the vertical vibration of the conductor cross-section. h It is the angular frequency of the vertical vibration of the conductor cross-section.

[0143] When the vertical vibration of the conductor section in the flow field reaches a steady state, the aerodynamic self-excited lift and self-excited torque are as shown in equations (10) and (11):

[0144]

[0145]

[0146] Discretizing equations (10) and (11) in time respectively, we obtain:

[0147]

[0148]

[0149] In the formula, L seh M is the theoretical aerodynamic lift force when the conductor cross-section vibrates vertically in the flow field. seh S is the column vector of aerodynamic torque self-excited forces when the conductor cross-section vibrates vertically in the flow field; H and H 14 All of these are intermediate calculation quantities of the theoretical aerodynamic lift corresponding to the vertical vibration of the conductor cross-section; S A and A 14 These are all intermediate calculation quantities of the aerodynamic torque self-excited force column vector corresponding to the vertical vibration of the conductor cross section.

[0150] Based on the least squares principle, the simulation data L of the aerodynamic self-excited lift corresponding to the vertical vibration of the conductor cross-section is used. testh Theoretical aerodynamic lift L when the substituted conductor cross-section vibrates vertically in the flow field seh Simulation data M of the aerodynamic self-excited torque corresponding to the vertical vibration of the conductor cross-section testh The aerodynamic torque self-excited force column vector M of the alternative conductor cross-section in the flow field during vertical vibration seh That is, let L seh =L testh M seh =M testh Solve for H 14 and A 14 The overdetermined system of equations yields:

[0151] H 14 =(S H T S H ) -1 S H T L testh (14)

[0152] A 14 =(S A T S A ) -1 S A T M testh (15)

[0153] Based on the intermediate calculation of the theoretical aerodynamic lift H corresponding to the vertical vibration of the conductor cross-section. 14 Intermediate calculation quantity A of the aerodynamic torque self-excited force column vector 14 The contribution of aerodynamic damping caused by vertical vibration to the self-excited lift was calculated. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to self-excited lift. The contribution of aerodynamic damping caused by vertical vibration to the self-excited torque The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited torque.

[0154] When the torsional vibration of the conductor finite element model is induced by a forced vibration system, the displacement and velocity of the conductor cross section during the torsional vibration process are shown in Equations (16) and (17):

[0155] α(t)=α0sin(ω α t) (16)

[0156]

[0157] In the formula, α is the displacement of the conductor cross-section under torsional vibration, and α0 is the amplitude of the conductor cross-section under torsional vibration. Let ω be the velocity of the torsional vibration of the conductor cross section. α Let be the angular frequency of the torsional vibration of the conductor cross section.

[0158] When the conductor cross-section reaches a steady state of torsional vibration in the flow field, the aerodynamic self-excited lift and self-excited torque are as shown in equations (18) and (19):

[0159]

[0160]

[0161] Discretizing equations (18) and (19) in time respectively, we obtain:

[0162]

[0163]

[0164] In the formula, L seαM is the theoretical aerodynamic lift force when the conductor cross-section undergoes torsional vibration in the flow field. seα Q is the column vector of aerodynamic torque self-excited forces when the conductor cross-section undergoes torsional vibration in the flow field; H and H 23 Q is an intermediate calculation quantity for the theoretical aerodynamic lift corresponding to the torsional vibration of the conductor cross-section; A and A 23 This is an intermediate calculation quantity for the column vector of aerodynamic torque self-excited force corresponding to the torsional vibration of the conductor cross section;

[0165] Based on the least squares principle, using the aerodynamic self-excited lift simulation data L corresponding to the torsional vibration of the conductor cross section testα Theoretical aerodynamic lift L when the alternative conductor cross-section undergoes torsional vibration in the flow field seα Simulation data M of the aerodynamic self-excited torque corresponding to the torsional vibration of the conductor cross section testα The aerodynamic torque self-excited force column vector M of the alternative conductor cross section during torsional vibration in the flow field seα L seα =L testα M seα =M testα Solve for H 23 and A 23 The overdetermined system of equations yields:

[0166] H 23 =(Q H T Q H ) -1 Q H T L testα (twenty two)

[0167] A 23 =(Q A T Q A ) -1 Q A T M testα (twenty three)

[0168] Based on the intermediate calculation of the theoretical aerodynamic lift H corresponding to the torsional vibration of the conductor cross-section 23 Intermediate calculation quantity A of the aerodynamic torque self-excited force column vector 23 The contribution of aerodynamic damping caused by torsional vibration to the self-excited lift was calculated. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to self-excited lift The contribution of aerodynamic damping caused by torsional vibration to the self-excited torque The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited torque.

[0169] Step 3.3: The finite element model of the conductor is induced to vibrate repeatedly using a forced vibration system to obtain the simulated aerodynamic self-excited lift data L corresponding to the vertical vibration of the conductor cross-section. testh Pneumatic self-excited torque simulation data M testh And the aerodynamic self-excited lift simulation data L corresponding to the torsional vibration of the conductor cross section testα Pneumatic self-excited torque simulation data M testα The flutter derivative of the conductor is calculated, and the flutter derivative of the conductor during each forced vibration is obtained. Multiple sets of training data are obtained, and a training dataset is constructed.

[0170] Step 4: Construct a Long Short-Term Memory (LSTM) deep learning network and train it using the training dataset to establish a nonlinear aerodynamic order reduction model. This includes the following steps:

[0171] Step 4.1: Construct a Long Short-Term Memory (LSTM) deep learning network. This network includes a forgetting gate, an input gate, and an output gate, such as... Figure 1 and Figure 2 As shown.

[0172] The forget gate is used to control the discarding of the previous time-state s. t-1 Useless information within the cell determines the cell state s of the previous time step. t-1 The forgetting rate is calculated using the following formula:

[0173] f t =σ(W f [h t-1 ,x t ]+b f ) (twenty four)

[0174] In the formula, f t For the Gate of Oblivion, W f For the weight of the forget gate, b f σ is the bias of the forget gate, and σ is the sigmoid activation function;

[0175] The input gate contains a sigmoid layer and a tanh layer to determine the current state s of the input pair. t Contribution, unit state s t The calculation formula is:

[0176] i t =σ(W i [h t-1 ,x t ]+b i (25)

[0177] c t =tanh(Wc [h t-1 ,x t ]+b c (26)

[0178] s t =f t ·s t-1 +i t ·c t (27)

[0179] In the formula, i t For the input gate, W i b represents the weights of the sigmoid layer. i For the bias of the input gate sigmoid layer, c t For the candidate values ​​obtained from the input gate tanh layer, W c b represents the weights of the input gate tanh layer. c The bias of the input gate tanh layer;

[0180] The output gate is used to extract the current cell state s. t The output value h t Output value h t Or the current cell state s t The output value h t It is directly used as the input value for the next time step, and the calculation formula is:

[0181] o t =σ(W o [h t-1 ,x t ]+b o (28)

[0182] h t =o t ·tanh(s t (29)

[0183] In the formula, o t For output gate, W o b represents the weights of the output gate sigmoid layer. o h is the bias of the output gate sigmoid layer. t This is the output value at the current moment.

[0184] Step 4.2: Train the Long Short-Term Memory deep learning network using the training dataset, combining the root mean square transfer method and stochastic gradient descent.

[0185] This not only further reduces the oscillations in the updates but also balances the update speed of each parameter, accelerating convergence. By adding a momentum term, the update formula for the parameter vector to be optimized in the Long Short-Term Memory deep learning network is obtained:

[0186]

[0187] in,

[0188]

[0189]

[0190] In the formula, Let α be the number of updates, θ be the learning rate, θ be the vector of parameters to be optimized, E(·) be the loss function, β1 be the first decay rate, and β2 be the second decay rate. and All of these are updates to the calculation parameters.

[0191] Step 4.3: Using the trained Long Short-Term Memory deep learning network, establish a nonlinear aerodynamic order reduction model and calculate the aerodynamic self-excited lift L of the conductor. Δt and aerodynamic self-excited torque M Δt .

[0192] Step 5: Based on the conductor finite element model and the nonlinear aerodynamic order reduction model, construct a coupled finite element model of the wind-conductor-transmission tower system to predict the displacement of the icing conductor under complex microclimates and determine whether the conductor will gallop.

[0193] This embodiment combines a coupled analysis strategy for the wind-conductor-transmission tower system with a reactivation method. Using the constructed coupled finite element model of the wind-conductor-transmission tower system, independent aerodynamic models are established at each node of the conductor. The aerodynamic forces calculated by these models are applied to each conductor node, and iterative calculations are used to achieve time-domain analysis of conductor flutter. Furthermore, this embodiment demonstrates the influence of multiple modes on flutter morphology in the coupled finite element model of the wind-conductor-transmission tower system. Moreover, the aerodynamic forces at each conductor node in the coupled finite element model are only related to the displacement time history of that node and are independent of the displacements of other nodes on the conductor.

[0194] This embodiment also employs a reactivation method to achieve iterative calculations of aerodynamic forces and structural responses at each time step, enabling time-domain analysis of conductor flutter. After completing an initial analysis, the reactivation method is used to modify the initial analysis and continue the analysis. The reactivation method can adjust time-varying loads in a timely manner during the full-method transient analysis, ensuring that after the ADPL script finishes running and exits, the previous analysis results are restarted and the wind loads at each node are updated to continue the calculation. In this embodiment, during the iterative update process, it is assumed that the aerodynamic forces at adjacent times are approximately equal within a short time step.

[0195] The specific steps for determining whether the conductor is galloping in this embodiment are as follows:

[0196] Step 5.1: Based on the structure of the conductor to be predicted, obtain the number of conductor nodes as Q. m One, corresponding to Q m Q was established at each conductor node. m A nonlinear aerodynamic force reduction model was developed and initialized to obtain a conductor-transmission tower finite element model. Then, based on the conductor node spacing, the multiple of the aerodynamic force applied per unit length of each conductor node was determined. After calculating the aerodynamic force corresponding to each conductor node, the aerodynamic force of each conductor node was multiplied by the set multiple to obtain the wind load to be applied to the conductor.

[0197] Step 5.2: Using finite element analysis software, call the ADPL script to establish a coupled finite element model of the wind-conductor-transmission tower system, and calculate the displacement D of the conductor under zero wind speed steady state. static And used as the initial displacement D initial .

[0198] Step 5.3: Using finite element analysis software, call the ADPL script to perform transient analysis on the coupled finite element model of the wind-conductor-transmission tower system, and obtain the initial displacement D. initial And after calculating for a sufficient amount of time, record the displacement D of the structural conductor. zeros This makes the initial displacement D initial Approaching 0 reduces the impact of initial displacement on aerodynamic prediction results, and allows for real-time acquisition of the displacement of each conductor node in the coupled finite element model of the wind-conductor-transmission tower system;

[0199] Step 5.4: Based on the displacement of each conductor node in the coupled finite element model of the wind-conductor-transmission tower system, calculate the aerodynamic forces (aerodynamic self-excited lift and aerodynamic self-excited torque) of each conductor node using a nonlinear aerodynamic force reduction model, and multiply it with the set multiplier to form wind load data.

[0200] Step 5.5: Using finite element analysis software, call the ADPL script to adjust the Rayleigh damping of the coupled finite element model of the wind-conductor-transmission tower system, so that the structural modal damping ratio of the coupled finite element model of the wind-conductor-transmission tower system is greater than 1.

[0201] Step 5.6: Calculate the aerodynamic forces using a reduced-order nonlinear aerodynamic model, and calculate the displacement D of the conductor in the coupled finite element model of the wind-conductor-transmission tower system after one time step Δt. Δt .

[0202] Step 5.7, based on the conductor displacement D obtained in step 5.6 Δt The data is input into a long short-term memory deep learning network, and the aerodynamic parameters of each conductor node are recalculated using a nonlinear aerodynamic order reduction model. The aerodynamic parameters of each conductor node are then multiplied by a set multiplier to update the wind load data.

[0203] Step 5.8: Using the reactivation method, repeat steps 5.6 and 5.7 to obtain the time t at which the static wind effect stops in the coupled finite element model of the wind-conductor-transmission tower system. sw The corresponding conductor displacement, using the still wind effect to stop at time t sw The aerodynamic parameters L when the static wind effect stops are obtained by calculating the conductor displacement. t_sw and M t_sw .

[0204] Step 5.9: Adjust the Rayleigh damping of the coupled finite element model of the wind-conductor-transmission tower system so that the structural modal damping ratio of the coupled finite element model of the wind-conductor-transmission tower system becomes a normal value.

[0205] Step 5.10: Using the reactivation method, after the calm wind effect stops, repeat steps 5.6 and 5.7 to make the conductor in the wind-conductor-transmission tower coupled finite element model vibrate under the calm wind steady state. Continue to update the aerodynamic parameters and wind load data according to the displacement of the conductor in the wind-conductor-transmission tower coupled finite element model, obtain the response time history curve of the conductor, and proceed to step 5.11.

[0206] Step 5.11: Perform flutter time-domain analysis on the conductor based on the conductor's response time history curve. If the amplitude of the conductor's response time history curve suddenly increases or diverges, it is determined that the conductor has galloped; otherwise, it is determined that the conductor has not galloped.

[0207] This embodiment updates the aerodynamic force based on the conductor displacement at the previous moment and applies it to the coupled finite element model of the wind-conductor-transmission tower system to calculate the displacement at the next moment. In other words, when the conductor in the coupled finite element model of the wind-conductor-transmission tower system begins to vibrate under a stable wind condition, the response time history curve of the conductor is recorded sequentially, and flutter time-domain analysis of the conductor is performed. Based on whether the amplitude of the conductor response time history suddenly increases or diverges, it is determined whether the conductor produces vertical or torsional galloping, thus achieving accurate prediction of conductor galloping behavior under complex microclimates.

[0208] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A smart prediction method for icy conductor galloping under complex microclimates, characterized in that, Specifically, the following steps are included: Step 1: Construct a finite element model of the conductor using fluid dynamics simulation software and identify the three-component force coefficients of the conductor in a static state; Step 2: Analyze the dynamic characteristics of the finite element model of the conductor, use the harmonic superposition method to generate the forced vibration displacement time history signal, and construct a forced vibration system for strong wave conductor vibration. Step 3: Use the forced vibration system to induce forced vibration of the conductor finite element model multiple times, obtain the flutter derivative of the conductor during each forced vibration, obtain multiple sets of training data, and construct a training dataset; Step 4: Construct a long short-term memory deep learning network and train the long short-term memory deep learning network using the training dataset to establish a nonlinear aerodynamic order reduction model. Step 5: Based on the conductor finite element model and the nonlinear aerodynamic order reduction model, construct a coupled finite element model of the wind-conductor-transmission tower system to predict the displacement of the icing conductor under complex microclimate and determine whether the conductor will gallop. Step 5 specifically includes the following steps: Step 5.1: Based on the structure of the conductor to be predicted, obtain the number of conductor nodes, establish a nonlinear aerodynamic reduced-order model for each conductor node, initialize each nonlinear aerodynamic reduced-order model to obtain the conductor-transmission tower finite element model, determine the multiple of the aerodynamic force applied per unit length of each conductor node based on the conductor node spacing, calculate the aerodynamic force corresponding to each conductor node, and multiply the aerodynamic force of each conductor node by the set multiple to obtain the wind load applied to the conductor. Step 5.2: Using finite element analysis software, establish a coupled finite element model of the wind-conductor-transmission tower system, and calculate the displacement D of the conductor under zero wind speed steady state. static And used as the initial displacement; Step 5.3: Using finite element analysis software, perform transient analysis on the coupled finite element model of the wind-conductor-transmission tower system to obtain the displacement of each conductor node in the coupled finite element model of the wind-conductor-transmission tower system in real time; Step 5.4: Based on the displacement of each conductor node in the coupled finite element model of the wind-conductor-transmission tower system, calculate the aerodynamic force of each conductor node using a nonlinear aerodynamic force reduction model and multiply it by the set multiplier to form wind load data. Step 5.5: Using finite element analysis software, adjust the Rayleigh damping of the coupled finite element model of the wind-conductor-transmission tower system so that the structural modal damping ratio of the coupled finite element model of the wind-conductor-transmission tower system is greater than 1. Step 5.6: Calculate the aerodynamic forces using a reduced-order nonlinear aerodynamic model, and calculate the displacement D of the conductor in the coupled finite element model of the wind-conductor-transmission tower system after one time step Δt. Δt ; Step 5.7, based on the conductor displacement D obtained in step 5.6 Δt The data is input into a long short-term memory deep learning network, and the aerodynamic parameters of each conductor node are recalculated using a nonlinear aerodynamic order reduction model. The aerodynamic parameters of each conductor node are then multiplied by a set multiplier to update the wind load data. Step 5.8: Using the reactivation method, repeat steps 5.6 and 5.7 to obtain the time t at which the static wind effect stops in the coupled finite element model of the wind-conductor-transmission tower system. sw The corresponding conductor displacement is used to calculate the aerodynamic parameters when the calm wind effect stops, and then the aerodynamic parameters are updated. Step 5.9: Adjust the Rayleigh damping of the coupled finite element model of the wind-conductor-transmission tower system so that the structural modal damping ratio of the coupled finite element model of the wind-conductor-transmission tower system becomes a normal value; Step 5.10: Using the reactivation method, after the calm wind effect stops, continue to repeat steps 5.6 and 5.7 to make the conductor in the wind-conductor-transmission tower coupled finite element model vibrate under the calm wind steady state. Continue to update the aerodynamic parameters and wind load data according to the displacement of the conductor in the wind-conductor-transmission tower coupled finite element model, obtain the response time history curve of the conductor, and proceed to step 5.

11. Step 5.11: Perform flutter time-domain analysis on the conductor based on the conductor's response time history curve. If the amplitude of the conductor's response time history curve suddenly increases or diverges, it is determined that the conductor has galloped; otherwise, it is determined that the conductor has not galloped.

2. The intelligent prediction method for icy conductor galloping under complex microclimates as described in claim 1, characterized in that, Step 1 specifically includes the following steps: Step 1.1: Construct a finite element model of the conductor using fluid dynamics simulation software, and set the model parameters and computational domain range of the conductor finite element model. The model parameters include the characteristic width of the conductor cross-section, the distance of the conductor cross-section from the fluid inlet and the fluid outlet, and the maximum torsional amplitude of the conductor cross-section in the vertical direction. The computational domain range includes the vertical height and horizontal length of the computational domain. Step 1.2: Mesh the computational domain of the finite element model of the conductor. Divide the computational domain into a rigid mesh region, a dynamic mesh region, and a static mesh region. Structured meshes are used in both the dynamic and static mesh regions. Unstructured meshes are used at the junction of the dynamic and static mesh regions. In the rigid mesh region, except for the surface layer of the conductor cross-section which uses a boundary layer mesh, all other areas use quadrilateral unstructured meshes. Step 1.3: Using a finite element model of the conductor, the stress on the conductor under static wind load is simulated, and the three-component force coefficients of the conductor in the static state are identified as follows: （1） （2） （3） In the formula, The static wind load drag experienced by the conductor cross-section in the flow field. Let be the lift force experienced by the conductor cross-section in the flow field. The torque experienced by the conductor cross-section in the flow field, and the static wind load drag experienced by the conductor cross-section in the flow field. The lift force experienced by the conductor cross section in the flow field Torque on the conductor cross section in the flow field The three forces that together make up the conductor; The drag coefficient, The lift coefficient, Torque coefficient, drag coefficient Lift coefficient and torque coefficient The three force coefficients that together make up the conductor; For fluid density, For the incoming flow velocity, The characteristic height of the conductor cross-section, The characteristic width of the conductor cross-section.

3. The intelligent prediction method for icy conductor galloping under complex microclimates as described in claim 2, characterized in that, In step 1, the finite element model of the conductor is the SST k-ω turbulence model. The thickness of the first layer of the finite element model falls within the viscous sublayer. The thickness of the first layer of the mesh is 0.02 mm. The boundary layer mesh of the finite element model has 30 layers and an expansion rate of 1.

2. The structured mesh is an outwardly diffused annular surface with a skewness of less than 0.

5.

4. The intelligent prediction method for icy conductor galloping under complex microclimates as described in claim 2, characterized in that, In step 2, when the harmonic superposition method is used to simulate the forced vibration of the conductor, the forced displacement signal of the conductor cross-section is: （4） In the formula, This is the forced displacement signal of the conductor cross-section. For time, The specified amplitude of the forced displacement signal for the conductor cross-section. The original amplitude of the superimposed signal. The number of superimposed harmonics, For the randomly generated i-th amplitude, Let be the randomly generated i-th circular frequency.

5. The intelligent prediction method for icy conductor galloping under complex microclimates as described in claim 4, characterized in that, Step 3 specifically includes the following steps: Step 3.1: Establish a coordinate system with the center of the conductor cross-section of the conductor finite element model as the origin. Dynamically update the position coordinates of the mesh in the conductor finite element model based on the spring smoothing method. Combined with the moving mesh sub-region method, divide the moving mesh region into multiple sub-regions, set each sub-region to move independently, and obtain the motion displacement of each sub-region within the moving mesh region. As shown in formula (5): （5） In the formula, The motion vector of the sub-region in the moving mesh region. The diffusion coefficient is... The grid motion vector for the guide section; Step 3.2: Forced vibration is induced in the finite element model of the conductor using a forced vibration system. Based on the split-state forced vibration method, the vertical and torsional vibrations of the conductor are simulated separately to obtain the simulated data of the aerodynamic self-excited lift force corresponding to the vertical vibration of the conductor cross-section. Aerodynamic self-excited torque simulation data And the aerodynamic self-excited lift simulation data corresponding to the torsional vibration of the conductor cross section. Aerodynamic self-excited torque simulation data The flutter derivative of the conductor is calculated. Step 3.3: The finite element model of the conductor is induced to vibrate repeatedly using a forced vibration system to obtain the aerodynamic self-excited lift simulation data corresponding to the vertical vibration of the conductor cross-section. Aerodynamic self-excited torque simulation data And the aerodynamic self-excited lift simulation data corresponding to the torsional vibration of the conductor cross section. Aerodynamic self-excited torque simulation data The flutter derivative of the conductor is calculated, and the flutter derivative of the conductor during each forced vibration is obtained. Multiple sets of training data are obtained, and a training dataset is constructed.

6. The intelligent prediction method for icy conductor galloping under complex microclimates as described in claim 5, characterized in that, In step 3.2, based on the vertical displacement and torsional degrees of freedom of the conductor, eight flutter derivatives are used to describe the aerodynamic lift and aerodynamic torque on the conductor, as shown in formulas (6) and (7): （6） （7） In the formula, This refers to the aerodynamic lift force acting on the conductor cross-section. This refers to the aerodynamic torque acting on the conductor cross-section; This represents the contribution of aerodynamic damping caused by vertical vibration to the self-excited lift. The contribution of aerodynamic damping caused by torsional vibration to the self-excited lift. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited lift. This represents the combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited lift. This represents the contribution of aerodynamic damping caused by vertical vibration to the self-excited torque. The contribution of aerodynamic damping caused by torsional vibration to the self-excited torque. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited torque. This represents the combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited torque. , , , , , , and Together they constitute the flutter derivative of the conductor; The characteristic width of the conductor cross-section; To calculate the frequency, K = ωB / U, where ω is the angular frequency of the conductor's vibration; When the finite element model of the conductor is induced to vibrate vertically using a forced vibration system, the conductor cross-section undergoes simple harmonic motion in the flow field during the vertical vibration process. The displacement and velocity of the conductor cross-section during the simple harmonic motion are shown in formulas (8) and (9): （8） （9） In the formula, For time, This represents the displacement of the conductor cross-section due to vertical vibration. The amplitude of the vertical vibration of the conductor cross-section. The velocity of the vertical vibration of the conductor cross-section. The angular frequency of the vertical vibration of the conductor cross-section; When the vertical vibration of the conductor section in the flow field reaches a steady state, the aerodynamic self-excited lift and self-excited torque are as shown in formulas (10) and (11): （10） （11） Discretizing equations (10) and (11) in time respectively, we obtain: （12） （13） In the formula, This represents the theoretical aerodynamic lift when the conductor cross-section vibrates vertically in the flow field. This is the column vector of aerodynamic torque self-excited forces when the conductor cross-section vibrates vertically in the flow field; and These are all intermediate calculation quantities of the theoretical aerodynamic lift corresponding to the vertical vibration of the conductor cross section; and These are all intermediate calculation quantities of the column vector of aerodynamic torque self-excited force corresponding to the vertical vibration of the conductor cross section; Based on the least squares principle, simulation data of aerodynamic self-excited lift corresponding to the vertical vibration of the conductor cross-section are used. Theoretical aerodynamic lift when the surrogate cross section vibrates vertically in the flow field Simulation data of aerodynamic self-excited torque corresponding to vertical vibration of conductor cross-section. Aerodynamic torque self-excited force column vector when the alternative conductor cross-section vibrates vertically in the flow field Solving for the problem yields: （14） （15） Intermediate calculation of theoretical aerodynamic lift based on vertical vibration of conductor cross-section. Intermediate calculation of aerodynamic torque self-excited force column vector The contribution of aerodynamic damping caused by vertical vibration to the self-excited lift was calculated. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to self-excited lift. The contribution of aerodynamic damping caused by vertical vibration to the self-excited torque. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by vertical vibration to the self-excited torque. ; When the torsional vibration of the conductor finite element model is induced by a forced vibration system, the displacement and velocity of the conductor cross section during the torsional vibration process are shown in formulas (16) and (17): （16） （17） In the formula, This represents the displacement due to torsional vibration of the conductor cross-section. The amplitude of the torsional vibration of the conductor cross section. The velocity of the torsional vibration of the conductor cross section. The angular frequency of the torsional vibration of the conductor cross-section; When the conductor cross-section reaches a steady state of torsional vibration in the flow field, the aerodynamic self-excited lift and self-excited torque are as shown in equations (18) and (19): （18） （19） Discretizing equations (18) and (19) in time respectively, we obtain: （20） （21） In the formula, This represents the theoretical aerodynamic lift when the conductor cross-section undergoes torsional vibration in the flow field. is the column vector of aerodynamic torque self-excited force when the conductor cross-section undergoes torsional vibration in the flow field; and This is an intermediate calculation quantity for the theoretical aerodynamic lift corresponding to the torsional vibration of the conductor cross-section; and This is an intermediate calculation quantity for the column vector of aerodynamic torque self-excited force corresponding to the torsional vibration of the conductor cross section; Based on the least squares principle, simulation data of aerodynamic self-excited lift corresponding to torsional vibration of the conductor cross-section are used. Theoretical aerodynamic lift of the alternative conductor section under torsional vibration in the flow field Simulation data of aerodynamic self-excited torque corresponding to torsional vibration of conductor cross section. Aerodynamic torque self-excited force column vector of the alternative conductor section during torsional vibration in the flow field Solving for the problem yields: （22） （23） Intermediate calculation of theoretical aerodynamic lift based on the torsional vibration of the conductor cross-section. Intermediate calculation of aerodynamic torque self-excited force column vector The contribution of aerodynamic damping caused by torsional vibration to the self-excited lift was calculated. The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to self-excited lift. The contribution of aerodynamic damping caused by torsional vibration to the self-excited torque The combined contribution of aerodynamic inertia and aerodynamic stiffness caused by torsional vibration to the self-excited torque. .

7. The intelligent prediction method for icy conductor galloping under complex microclimates as described in claim 5, characterized in that, Step 4 specifically includes the following steps: Step 4.1: Construct a long short-term memory deep learning network. The long short-term memory deep learning network includes a forget gate, an input gate, and an output gate. The forget gate is used to control the discarding of the previous time-state s. t-1 The formula for calculating useless information within is: （24） In the formula, For the Gate of Oblivion As for the weight of the Forgotten Gate, For the offset of the forget gate, It is the sigmoid activation function; The input gate contains a sigmoid layer and a tanh layer to determine the state of the cell at the current input. The contribution is calculated using the following formula: （25） （26） （27） In the formula, For input gate, The weights of the sigmoid layer. The bias of the input gate sigmoid layer. The candidate values ​​obtained from the input gate tanh layer are... The weights of the input gate tanh layer, The bias of the input gate tanh layer; The output gate is used to extract the cell state at the current time. Output value Output value Or the current cell state Output value It is directly used as the input value for the next time step, and the calculation formula is: （28） （29） In the formula, For output gate, The weights of the output gate sigmoid layer. The bias of the output gate sigmoid layer. This is the output value at the current moment; Step 4.2: Train the Long Short-Term Memory (LSTM) deep learning network using the training dataset. Combine the root mean square propagation method and stochastic gradient descent, and add a momentum term to optimize the parameter vector in the LTM deep learning network. The update formula for the parameter vector is: （30） in, （31） （32） In the formula, ℓ represents the number of updates. For learning rate, Let be the parameter vector to be optimized. For loss function, The first attenuation rate, The second attenuation rate, and All of these are updates to the calculation parameters; Step 4.3: Using the trained Long Short-Term Memory deep learning network, establish a nonlinear aerodynamic order reduction model and calculate the aerodynamic self-excited lift L of the conductor. Δt and aerodynamic self-excited torque M Δt .

8. The intelligent prediction method for icy conductor galloping under complex microclimates as described in claim 1, characterized in that, In step 5, the aerodynamic parameters include aerodynamic self-excited lift and aerodynamic self-excited torque.

Images