ICOA-svm-based power transmission tower safety state prediction method

Abstract

The ICOA-SVM-based method for predicting the safety status of power transmission towers includes: initializing the raccoon optimization algorithm population using a Tent-Logistic-Cosine combined chaotic mapping strategy; improving the population update method during the exploration phase using an adaptive dynamic difference guidance strategy; introducing an SBOA attack strategy and adding a predator avoidance behavior strategy to improve the balance between global search and local exploitation, resulting in an improved ICOA algorithm; calculating the fitness value of each raccoon and its corresponding optimal position; establishing a finite element model of the power transmission tower, classifying the tower safety status levels, and forming a sample dataset; assigning the obtained penalty parameter c and kernel function parameter g to the SVM to construct a parameter-optimized SVM model; training the SVM model to obtain the ICOA-SVM prediction model for predicting the safety status of power transmission towers; inputting the wind speed, wind direction angle, and icing thickness of the tower to be evaluated into the trained ICOA-SVM prediction model, and outputting the safety status level of the tower. This method improves the accuracy of predicting the safety status of power transmission towers.

Description

Technical Field

[0001] This invention relates to the field of power transmission tower safety status assessment technology, specifically to a power transmission tower safety status prediction method based on ICOA-SVM. Background Technology

[0002] As a crucial carrier for power transmission, the safe operation of transmission towers is vital to social development and stability. However, due to their operational nature and environment, transmission towers are subjected to static and dynamic loads such as strong winds and heavy rains over extended periods, leading to frequent problems such as loose bolt connections, node corrosion, and component deformation. These issues reduce the load-bearing capacity of the towers and could potentially trigger serious power accidents such as tower deformation or even collapse. Therefore, conducting safety condition assessments of transmission towers is of great significance.

[0003] In the field of power transmission tower safety assessment, existing technologies mainly rely on three types of schemes: meteorological models, scoring analysis methods, and intelligent algorithms, but all of them have significant limitations. Traditional methods, such as the AHP-standard deviation method weight optimization proposed in reference [1], can achieve an accuracy of 82.3%, but the response time exceeds 30 minutes and does not consider the wind-ice coupling effect; the PLS-MOS-GIS multi-source fusion model developed in reference [2] can improve the accuracy to 85.7%, but it is still limited by the static assessment mode. The current best intelligent algorithm scheme, such as the landslide area power transmission tower state prediction model based on ISSA-BP neural network in reference [3], can improve the accuracy of power transmission tower operation state prediction to a certain extent, but it still has problems such as search efficiency and parameter optimization stability in complex environments. More importantly, existing algorithms generally have three major defects: strong initial solution randomness, easy to get trapped in local optima, and poor dynamic adaptability. For example, the parameter sensitivity coefficient of the original COA algorithm is as high as 0.82, while the small fluctuation of the SVM kernel function parameter g will cause a 15% accuracy deviation. These technical bottlenecks directly restrict the rapid and accurate assessment of the safety status of power transmission towers in complex environments.

[0004] Existing research on the safety assessment of power transmission towers relies on various data sources, including meteorological, geographical, and power grid data, and does not fully utilize intelligent algorithms. This results in insufficient accuracy of the assessment models, unsatisfactory prediction speed and accuracy, and an inability to rapidly assess the safety status of power transmission towers. Traditional support vector machines (SVMs) suffer from long training times and high computational complexity when dealing with large datasets, and the values ​​of penalty parameters and kernel function parameters are random. Existing power transmission tower assessment methods also suffer from strong subjectivity and low evaluation accuracy and efficiency.

[0005] References: Reference [1]: Zhao Ting. Evaluation of wind resistance performance of transmission line towers [D]. Baoding: North China Electric Power University, 2019. Reference [2]: Xiong Jun, Lin Han, Wang Qinghua, Zhang Jianguo, Zhuang Yulin. Research on regional power grid wind disaster early warning model based on GIS [J]. East China Electric Power, 2011, 39(8): 1248-1252. Reference [3]: Li Mengyuan, Dong Ruike, Wang Yanhai, Zhou Dongyang, Zou Mengjian. Prediction model of power transmission tower status in landslide area based on ISSA-BP neural network [J]. Electronic Measurement Technology, 2023, 46(11): 74-81. Summary of the Invention To address the aforementioned technical problems, this invention provides a method for predicting the safety status of power transmission towers based on ICOA-SVM. It utilizes a Tent-Logistic-Cosine combined chaotic mapping strategy to optimize population initialization, employs an adaptive dynamic difference guidance strategy to improve the population update method during the exploration phase, introduces the Secretary Bird Optimization Algorithm (SBOA) attack strategy to enhance local exploitation capabilities, and adds a predator avoidance behavior strategy to improve the balance between global search and local exploitation, effectively avoiding overlearning and underlearning states, thereby accelerating computation and making the model's classification results more accurate. Furthermore, based on an improved Raccoon Optimization Algorithm, the penalty parameters and kernel parameters of the support vector machine are optimized to improve the accuracy of predicting the safety status of power transmission towers.

[0006] The technical solution adopted in this invention is as follows: The method for predicting the safety status of transmission towers based on ICOA-SVM includes the following steps: Step 1: Initialize the population of the Raccoon Optimization Algorithm (COA) using the Tent-Logistic-Cosine combined chaotic mapping strategy and set the fitness function; Step 2: Adaptive dynamic differential guidance strategy is adopted to improve the population update method in the exploration phase. The Secretary Bird Optimization Algorithm (SBOA) attack strategy is introduced to enhance local exploitation capabilities. A predator avoidance behavior strategy is added to improve the balance between global search and local exploitation, resulting in the improved ICOA algorithm. Step 3: Following steps 1 and 2, update the raccoon population, calculate the fitness value of each raccoon and its corresponding optimal position. When the maximum number of iterations is reached, the optimal position corresponding to its minimum fitness value is taken as the parameter vector. , ],in: The corresponding penalty parameter c for Support Vector Machine (SVM) Corresponding kernel function parameter g; Step 4: Establish a finite element model of the transmission tower to simulate wind and ice load conditions under different wind angles, wind speeds and ice thicknesses; classify the safety status levels of the transmission tower, form a sample dataset, and divide the sample data into training and test sets after normalization.

[0007] Step 5: Assign the penalty parameter c and kernel function parameter g obtained in Step 3 to the Support Vector Machine (SVM) to construct the parameter-optimized SVM model; input the training set obtained in Step 4 into the SVM model for training to obtain the ICOA-SVM prediction model for predicting the safety status of power transmission towers.

[0008] Step 6: Input the wind speed, wind direction angle, and icing thickness of the transmission tower to be evaluated into the trained ICOA-SVM prediction model, and output the safety status level of the transmission tower.

[0009] In step 1, the Tent-Logistic-Cosine combined chaotic mapping combines the Tent mapping, Logistic mapping and cosine modulation function to enhance the randomness and ergodicity of the sequence.

[0010] The Tent mapping part uses piecewise linear mappings 2rx and 2r(1) x) Improve traversal, making the sequence cover the search space more evenly; The Logistic mapping part, through 4(1) r)x(1 x) Introducing nonlinear relationships prevents the sequence from falling into a periodic orbit and improves chaotic characteristics; In the cosine modulation function part, the cosine function is used to transform the linear and nonlinear terms, compress the result to the interval [-1,1], and normalize it to [0,1]. The periodic function is used to increase the complexity of the iteration and improve the randomness of the chaotic sequence. The mathematical expression for the Tent-Logistic-Cosine combinatorial chaotic mapping is as follows: (2); In formula (2): The initial value of the chaotic sequence is represented by rand, which is randomly generated and follows a uniform distribution of (0,1). Indicates the first Chaotic variables in the next iteration; The mixing coefficient represents the weight ratio of the Tent mapping and the Logistic mapping. It is randomly generated by rand and takes the value (0,1). The fixed parameters 2 and 4 represent the standard parameter extensions of the Tent mapping, which are used for the linear terms in the left and right halves and the chaotic parameters of the Logistic mapping, respectively, to ensure that the Logistic part is in a chaotic state when r = 0. 0.5 is the offset used to adjust the phase of the input to the cosine function.

[0011] In step 1, the mathematical expression for the fitness function is shown in equation (1): (1); In formula (1): Indicates the fitness value; Indicates the first The true category label of each sample; Indicates the first Predicted class labels for each sample; This represents the total number of samples.

[0012] In step 2, the adaptive dynamic differential guidance strategy incorporates the mutation mechanism of differential evolution (DE): 2.1: To address the varying search capability requirements of the Raccoon Optimization Algorithm at different optimization stages, a dynamic scaling factor F is introduced. A sine function is used to control the dynamic scaling factor F, which enhances the algorithm's exploration capability in the early stages and improves its local exploitation capability in the later stages, thereby accelerating the convergence speed while ensuring the breadth of the search. 2.2: To avoid the population becoming overly reliant on the current optimal solution, a random individual perturbation mechanism is introduced, involving three random individuals. , , By forming a difference vector, the population can have better diversity during the update process; 2.3: During the mutation process, add an elite guiding factor, that is, introduce the current best individual. This information is used to enhance global search capabilities while accelerating the convergence of the population towards the optimal solution.

[0013] (3); (4); In the above formula: This represents the current iteration number; This represents the maximum number of iterations. , , For randomly selected individuals, r1≠r2≠r3≠i, where i represents the ID of the individual to be updated; F is the dynamic scaling factor. This represents a random number within the interval (0,1); This is the current optimal value; This indicates the updated individual position.

[0014] In step 2, the Secretary Bird Optimization Algorithm (SBOA) attack strategy still adopts the random perturbation mode of the original Raccoon Optimization Algorithm (COA) in the case of 50% probability, in order to maintain the basic search capability and ensure the stability of the algorithm framework; in the other 50% probability case, the Secretary Bird attack strategy is introduced, which uses the contraction factor CF to dynamically reduce the search step size in the later stage of the iteration to enhance the local exploitation capability. At the same time, the Levy flight mechanism is adopted to enable individuals to make long-range jumps in the search space, thereby improving the ability to escape local optima. The Secretary Bird Optimization Algorithm (SBOA) attack strategy formula is as follows: (5); (6); (7); In the above formula, It is a contraction factor, which enhances exploration in the early stage and promotes algorithm development in the later stage. The Levi flight vector enables long-distance jumps through a heavy-tailed distribution. Represents the Hadamard product; This represents the current optimal solution; Indicates the updated position; Indicates the current position; The dimension is represented by t; t is the current iteration number. This represents the maximum number of iterations.

[0015] In step 2, the addition of predator avoidance strategies enables the raccoon optimization algorithm (COA) to dynamically adjust its risk avoidance methods according to different situations during the search process. One method is to simulate the mother bear leading her cubs to quickly evacuate, enhancing the target orientation of the search and improving the optimization efficiency, as shown in formula (8). Another method is to introduce a random cooperative mechanism, allowing individuals to escape irregularly in complex environments, increasing the randomness and diversity of the search, as shown in formula (9). (8); (9); In the above formula, This indicates the updated location after the predator avoidance behavior update; Indicates the group's historical best position; Indicates the current iteration number; Indicates the maximum number of iterations; and This represents a random vector located in the range [0,1]. For random vectors Dimensionally consistent vector; Indicates the current individual's location; This indicates an alertness coefficient that takes the value of 1 or 2. This indicates the position of a randomly selected group member.

[0016] Formula (8) describes targeted risk avoidance behavior, a strategy applicable when an individual perceives a single threat and is triggered with a 50% probability. The individual will move towards the group's historically optimal position. They move closer together to increase their chances of survival. This process simulates three biological characteristics: First, Representing the safe zone in collective memory, providing individuals with a clear escape route; secondly, the squared decay factor. This is used to adjust the search intensity, so that the threat's impact gradually weakens with iterations, ensuring the algorithm has strong exploration capabilities in the early stages and gradually converges in the later stages; finally, the perturbation term... Random offsets are generated within the interval [-1, 1] to enhance the diversity of escape paths, thereby avoiding the algorithm from getting trapped in local optima. ∈[0,1]D; D is the dimension of the search space.

[0017] Formula (9) corresponds to the cooperative escape strategy. When an individual perceives multiple threats, it is triggered with a 50% probability, no longer acting alone, but relying on group cooperation to improve its chances of survival. This mechanism randomly selects a group member. As a target to be followed, it simulates the phenomenon of collective escape to avoid the risks of isolated action. The parameter B∈{1,2} is used to adjust the individual's alertness to the current location, achieving dual-mode defense to adapt to search needs in different environments. Simultaneously, the step size random factor... Assigning ∈[0,1]D introduces nondeterministic characteristics to the escape behavior, further expanding the search space coverage. This improvement not only enhances the algorithm's adaptability to complex environments but also promotes information exchange between individuals, thereby strengthening the overall optimization performance.

[0018] In step 3, after obtaining the ICOA algorithm through steps 1 and 2, the fitness value of each raccoon individual at the current position is first calculated according to equation (1); then the position of each individual is updated according to equations (3) to (9), and the fitness values ​​before and after the update are compared. The position with better fitness is retained as the current optimal position of the individual; at the same time, the entire population is traversed, and the position of the individual with the smallest fitness is recorded as the current global optimal position.

[0019] During this process, continuous iteration and optimization are carried out. Each iteration updates the population based on the adaptive dynamic differential guidance strategy, the Secretary Bird Optimization Algorithm (SBOA) attack strategy, and the newly added predator avoidance behavior strategy.

[0020] The algorithm stops updating when it reaches the maximum number of iterations. The optimal position corresponding to the minimum fitness value obtained at this point is the parameter vector. , ],in: The corresponding penalty parameter c for support vector machines, This corresponds to the kernel function parameter g. Assigning this parameter vector to the SVM yields the SVM evaluation model with the best classification performance.

[0021] In step 4, a finite element model of the transmission tower is established to simulate wind and ice load conditions under different wind angles, wind speeds, and ice thicknesses. Based on the relationship between the component stress ratio ξ and the tower top displacement l and tower height h, safety status categories are divided into four types: safe, relatively safe, basically safe, and unsafe. A sample dataset with safety status labels is formed, and after normalization, the sample data is divided into training and test sets.

[0022] In step 5, the optimal penalty parameter c and kernel function parameter g obtained in step 3 are assigned to the support vector machine to construct the parameter-optimized SVM classification model. Then, the training set samples formed and preprocessed in step 4 are input into the SVM classification model for training, so that the model learns the mapping relationship between wind speed, wind direction angle and ice thickness and the safety status level of the transmission tower, thereby obtaining the ICOA-SVM prediction model for predicting the safety status of the transmission tower.

[0023] In step 6, the wind speed, wind direction angle, and icing thickness of the transmission tower to be evaluated are used as input features and input into the trained ICOA-SVM prediction model. The prediction model classifies and identifies the samples to be evaluated based on the mapping relationship between the input features and the safety status level established during the training process, and outputs the corresponding safety status level of the transmission tower, so as to realize the rapid prediction and evaluation of the operating status of the transmission tower.

[0024] This invention provides a method for predicting the safety status of power transmission towers based on ICOA-SVM, with the following technical advantages: (1) This invention uses the Tent-Logistic-Cosine combined chaotic mapping strategy to optimize the initial population, enhance sequence characteristics, and improve the algorithm's search and convergence capabilities; it introduces adaptive dynamic difference guidance and other strategies to improve the algorithm's stability, convergence accuracy and local exploitation capabilities, and adds a new strategy to balance global and local search.

[0025] (2) Compared with existing single improvement or static parameter optimization algorithms, this invention enables the algorithm to have stronger global exploration capabilities in the early stage and higher local development accuracy in the later stage through dynamic scaling factor, elite guidance mechanism and multi-strategy search method triggered by probability, thereby achieving a dynamic balance between search efficiency and solution accuracy and avoiding over-learning and under-learning phenomena.

[0026] (3) This invention utilizes the improved ICOA algorithm to adaptively optimize the support vector machine penalty parameter c and kernel function parameter g, avoiding the instability caused by manual setting or random selection of parameters in traditional SVM. Simulation results show that, compared with traditional SVM or other optimized SVM models, the model of this invention exhibits significant advantages in terms of root mean square error, average relative error, and average absolute error. Attached Figure Description

[0027] The present invention will be further described below with reference to the accompanying drawings and examples; Figure 1 This is the convergence curve of the COA-SVM iteration.

[0028] Figure 2 The convergence curve of the ICOA-SVM iteration.

[0029] Figure 3 This is the prediction result from SVM.

[0030] Figure 4 The results are from the COA-SVM prediction.

[0031] Figure 5 The results are predictions from the ICOA-SVM.

[0032] Figure 6 This is a flowchart of the ICOA-SVM model. Detailed Implementation

[0033] Principle Analysis: The Raccoon Optimization Algorithm is a novel metaheuristic intelligent optimization algorithm proposed by Mohammad et al. in 2023 based on the living habits of raccoons. This algorithm solves optimization problems by simulating the behaviors of raccoons attacking, hunting iguanas, and escaping predators.

[0034] To address the problems of slow convergence speed, easy getting trapped in local optima, sensitivity to parameter settings, and poor adaptability in dynamic environments, the basic raccoon optimization algorithm has been proposed. For example, reference [4] proposes a distribution network SOP optimization configuration method based on ICOA-KM cluster partitioning to address the problems of low new energy consumption rate, significant increase in network loss, and frequent voltage overruns in the distribution network caused by high proportion of new energy grid connection.

[0035] Reference [5] proposes a photovoltaic array fault diagnosis method based on an improved Longnose Raccoon optimization algorithm and an optimized Extreme Learning Machine to improve the accuracy of photovoltaic array fault diagnosis. To address the limitations of the Longnose Raccoon algorithm in terms of the randomness of its initial parameters and its global search capability, the algorithm is optimized using Circle chaotic mapping, Lévy flight, and dynamic refraction back-learning to improve both the accuracy and speed of the optimization process.

[0036] Support Vector Machines (SVMs) are a type of supervised learning model widely used in classification and regression analysis. Their core advantage lies in their ability to handle both linear and nonlinear data classification problems, and through the training process, they determine an optimal hyperplane that effectively separates data from different categories.

[0037] It is often combined with deep learning. Deep learning performs well in extracting complex features from massive data, while support vector machines are accurate in classification in feature space. The combination of the two complements the feature extraction capability of deep learning and the stable classification capability of SVM, forming an effective strategy. For example, reference [6] uses thermal images, stress distribution images and crack distribution maps of power generation equipment as monitoring indicators. It uses convolutional autoencoders to extract the deep feature indicators of each monitoring image, and then introduces the support vector machine algorithm to perform fault diagnosis on the extracted indicators, thereby realizing contactless monitoring and diagnosis of power generation equipment.

[0038] In addition, various intelligent optimization algorithms are used to optimize the parameters of SVM in order to find better kernel parameters, penalty factors, etc., thereby improving the performance and generalization ability of the model. For example, Reference [7] proposed a method for wind turbine bearing fault diagnosis based on the combination of adaptive noise complete set empirical mode decomposition (CEEMDAN) and improved sparrow search algorithm (ISSA) to optimize support vector machine (SVM) in order to solve the problem of difficulty in extracting fault features of wind turbine bearings under strong noise background. Reference [8] proposed an intrusion detection method based on Grey Wolf Optimization (GWO) to optimize DBN-SVM. In the GWO algorithm, the convergence speed is accelerated and the search range of the wolf pack is expanded by introducing adaptive hunting weight coefficient and improved alpha wolf position update formula. The optimal grey wolf individual adaptive perturbation strategy is added to avoid getting trapped in local optima. The improved GWO algorithm is further used to optimize DBN-SVM and applied to intrusion detection.

[0039] This invention presents an improved ICOA-SVM-based method for predicting the safety status of transmission towers. First, it optimizes the raccoon optimization algorithm (COA) population initialization using a Tent-Logistic-Cosine combined chaotic mapping strategy. Then, it improves the population update method during the exploration phase using an adaptive dynamic differential guidance strategy. A secretary bird optimization algorithm (SBOA) attack strategy is introduced to enhance local exploitation capabilities, and a predator avoidance strategy is added to improve the balance between global search and local exploitation, resulting in an improved ICOA algorithm. The raccoon population is updated using these strategies, calculating the fitness value and optimal position of each raccoon. The algorithm stops updating when it reaches the maximum number of iterations, and the optimal position corresponding to the minimum fitness value is assigned to the penalty parameter c and kernel function parameter g of the support vector machine (SVM). Sample data is then acquired and preprocessed. Taking the SG2-42 straight tower of a 500kV line as an example, a finite element model of the transmission tower is established using B31 beam elements. Considering factors such as wind direction, wind speed, and ice thickness, finite element analysis is performed using an equivalent static loading method to simulate various wind and ice load conditions. Based on the established safety status assessment standards for transmission towers, operating states are categorized into four levels: safe, relatively safe, basically safe, and unsafe. Simulation data is classified and proportionally divided into training and test sets, and then normalized. An SVM evaluation model is constructed using the optimal c and g parameters and the minimum fitness value obtained through iteration. The preprocessed training set data is input into the model for training, forming the final ICOA-SVM prediction model. Finally, real-time data such as wind speed, wind direction angle, and icing thickness of actual transmission towers are input into the trained ICOA-SVM model. The model calculates and outputs the safety status level, achieving a rapid and accurate assessment of the safety status of transmission towers.

[0040] The following explanation uses a straight-line tower of type SG2-42 in a certain 500kV line as an example.

[0041] Step 1: Initialize the population for the Raccoon Optimization Algorithm (COA) using a Tent-Logistic-Cosine combined chaotic mapping strategy; The Tent-Logistic-Cosine chaotic map is a typical composite chaotic system. Its core idea is to combine the Tent map, Logistic map, and cosine modulation function to enhance the randomness and ergodicity of the sequence. This map consists of three parts: the first part is the Tent map, which uses piecewise linear mappings 2rx and 2r(1...). x) Improve traversal to make the sequence cover the search space more evenly; the second part is the Logistic mapping part, through 4(1 r)x(1 x) Introducing nonlinear relationships prevents the sequence from falling into a periodic trajectory, improving chaotic characteristics; finally, the cosine modulation part uses the cosine function to transform the linear and nonlinear terms, compressing the result to the [-1,1] interval and normalizing it to [0,1]. The periodic function increases the complexity of the iteration, enhancing the randomness of the chaotic sequence. This design, while ensuring chaotic characteristics, improves the uniformity of population initialization, thus helping to improve the search ability and convergence performance of the Raccoon Optimization algorithm. Its mathematical expression is as follows: (2); In formula (2): represents the initial value of the chaotic sequence, randomly generated by rand, and follows a uniform distribution of (0,1); r is the mixing coefficient, which controls the weight ratio of the Tent mapping and the Logistic mapping, randomly generated by rand, and takes the value (0,1); the fixed parameters 2 and 4 represent the standard parameter extensions of the Tent mapping, used for the linear terms of the left and right halves and the chaotic parameters of the Logistic mapping, respectively, to ensure that the Logistic part is in a chaotic state when r = 0; 0.5 is the offset, which adjusts the phase of the input to the cosine function.

[0042] Step 2: An adaptive dynamic differential guidance strategy is adopted to improve the population update method in the exploration phase. The Secretary Bird Optimization Algorithm (SBOA) attack strategy is introduced to enhance local exploitation capabilities, and a predator avoidance behavior strategy is added to improve the balance between global search and local exploitation, resulting in an improved ICOA algorithm; An adaptive dynamic differential guidance strategy integrates the mutation mechanism of differential evolution (DE) to compensate for the shortcomings of the original algorithm in three ways. First, to address the varying search capability requirements of the raccoon optimization algorithm at different optimization stages, a dynamic scaling factor is introduced. A sine function is used to control the scaling factor F, enhancing the algorithm's exploration capability in the early stages and improving its local exploitation capability in the later stages, thereby accelerating convergence while maintaining search breadth. Second, to avoid the population's over-reliance on the current optimal solution, a random individual perturbation mechanism is added. By introducing three random individuals to form a difference vector, the population exhibits better diversity during updates. Third, during the mutation process, an elite guidance term is added, incorporating information from the current optimal individual to enhance global search capability and accelerate the population's convergence towards the optimal solution. This improved strategy improves the algorithm's stability and convergence accuracy while maintaining global search capability. This allows the algorithm to effectively alleviate premature convergence problems and enhance solution quality even in complex, high-dimensional, multi-peak optimization problems. The improved formula is shown below: (3); (4); In the formula: tThis represents the current iteration number. This represents the maximum number of iterations. , , Let r1 be a randomly selected individual (r1≠r2≠r3≠i), F be a dynamic scaling factor, and rand represent a random number in the interval (0,1). This is the current optimal value.

[0043] In the case of 50% probability, the Secretary Bird Optimization Algorithm (SBOA) still adopts the random perturbation mode of the original COA to maintain the basic search capability and ensure the stability of the algorithm framework. In the other 50% probability, the Secretary Bird attack strategy is introduced, which uses the contraction factor CF to dynamically reduce the search step size in the later stage of the iteration to enhance the local exploitation capability. At the same time, the Levy flight mechanism is adopted to enable individuals to make long-range jumps in the search space, thereby improving the ability to escape local optima.

[0044] Through the above improvements, COA can more effectively balance global exploration and local exploitation during the search process, ensuring population diversity, improving optimization accuracy and convergence speed, while avoiding a significant increase in computational complexity. The SBOA attack strategy formula is as follows: (5); (6); (7); In the formula: As a contraction factor, its main function is to enhance exploration in the early stages and promote algorithm development in the later stages. The Lévy flight vector enables long-distance jumps through a heavy-tailed distribution; ⊙ represents the Hadamard product; This represents the current optimal solution; Indicates the updated position; Indicates the current position. dim represents dimension; This represents the current iteration number. This represents the maximum number of iterations.

[0045] The addition of predator avoidance strategies enables the raccoon optimization algorithm to dynamically adjust its risk avoidance methods according to different situations during the search process. One approach is to simulate the mother bear leading her cubs to a rapid retreat, enhancing the target orientation of the search and improving optimization efficiency, as shown in formula (8). Another approach is to introduce a random cooperative mechanism, allowing individuals to escape irregularly in complex environments, increasing the randomness and diversity of the search, as shown in formula (9). The combination of these two strategies allows the algorithm to maintain a better balance between exploration and exploitation, which helps to improve the overall optimization performance.

[0046] (8); (9); Formula (8) describes targeted risk avoidance behavior, a strategy applicable when an individual perceives a single threat and is triggered with a 50% probability. The individual will move towards the group's historically optimal position. They move closer together to increase their chances of survival. This process simulates three biological characteristics: First, The first part represents the safe zone in the collective memory, providing individuals with a clear escape direction; secondly, the squared decay factor is used to adjust the search intensity, so that the threat's impact gradually weakens with iteration, ensuring that the algorithm has strong exploration capabilities in the early stage and gradually converges in the later stage; finally, the perturbation term (2 - I This generates random offsets within the interval [-1, 1], enhancing the diversity of escape paths and thus preventing the algorithm from getting trapped in local optima. ∈[0,1]D; D is the dimension of the search space.

[0047] Formula (9) corresponds to the cooperative escape strategy. When an individual perceives multiple threats, it is triggered with a 50% probability, no longer acting alone, but relying on group cooperation to improve its chances of survival. This mechanism randomly selects a group member. As a target to be followed, it simulates the phenomenon of collective escape to avoid the risks of isolated action. The parameter B∈{1,2} is used to adjust the individual's alertness to the current location, achieving dual-mode defense to adapt to search needs in different environments. Simultaneously, the step size random factor... Assigning ∈[0,1]D introduces nondeterministic characteristics to the escape behavior, further expanding the search space coverage. This improvement not only enhances the algorithm's adaptability to complex environments but also promotes information exchange between individuals, thereby strengthening the overall optimization performance.

[0048] Step 3: Following steps 1 and 2, update the raccoon population, calculating the fitness value and corresponding optimal position for each raccoon. During this process, iterative optimization is continuously performed. Each iteration updates the population based on the Adaptive Dynamic Differentiation Guidance strategy, the Secretary Bird Optimization (SBOA) attack strategy, and a newly added predator avoidance behavior strategy. The update stops when the algorithm reaches the maximum number of iterations. At this point, the minimum fitness value and its corresponding optimal position are assigned to the penalty parameter c and kernel function parameter g in the Support Vector Machine (SVM). These two parameters are crucial to the performance of the SVM model; the parameters optimized through the ICOA algorithm enable the SVM to perform better in subsequent modeling and prediction.

[0049] Step 4: Establish a finite element model and simulate different working conditions to obtain an evaluation dataset: Taking the SG2-42 straight-line tower as the research object, its tower height is 68.35m, nominal height is 42m, and the spans are 421m and 368m respectively. The main diagonal members of the tower body are Q420 and Q345 steel, and the remaining auxiliary materials are Q235 steel. A finite element model is established using B31 beam elements. Wind speeds of 0-30m / s (increasing by 2m / s sequentially) and ice thicknesses of 0-10mm (increasing by 2mm sequentially) are applied to the tower, with wind angles of 30°, 45°, 60°, and 90°, simulating 384 wind and ice load conditions. Based on risk assessment methods and referring to relevant research and regulations, a safety status assessment standard for transmission towers is formulated, classifying safety status categories according to the component stress ratio ξ (the ratio of the maximum stress location to its yield stress) and the relationship between the tower top displacement l and the tower height h. The simulation data are classified, with 75% of the sample data randomly selected as the training set and 25% as the test set.

[0050] In step 4, the dynamic response under different wind angles and loads is simulated. Wind speed, wind angle, and ice thickness are used as input data for the model. The operating status and safety level of the transmission tower are analyzed in combination with risk assessment methods, and the towers are classified into four categories: safe, relatively safe, basically safe, and unsafe, which are abbreviated as numbers 1, 2, 3, and 4. The relevant data obtained from the simulation are normalized and used as the analysis samples for the algorithm model.

[0051] Step 5: Construct an SVM evaluation model using the optimal c and g parameters and the minimum fitness value obtained through iteration. Input the experimental sample data obtained and classified under different simulated working conditions in Step 4 into the constructed SVM model for training. During training, the model continuously learns the features and patterns in the sample data, thus obtaining the trained ICOA-SVM evaluation model.

[0052] Step 6: In practical applications, real-time data such as wind speed, wind direction, and icing thickness of the actual transmission tower are input into the trained evaluation model (ICOA-SVM) for assessment. After receiving this data, the model performs calculations and analyses based on its learned knowledge and patterns, ultimately displaying the safety status level of the transmission tower. For example, if the model calculates a "safe" safety status level based on the input data, it indicates that the transmission tower is currently operating normally; if it is "unsafe," relevant personnel need to promptly inspect and maintain the tower to prevent potential safety accidents.

[0053] To verify the effectiveness of the ICOA-SVM algorithm in the parameter optimization process of support vector machines, the iterative convergence performance of the ICOA-SVM algorithm and the improved ICOA-SVM algorithm is compared and analyzed. The comparison results are shown in [the table below]. Figure 1 and Figure 2 .Depend on Figure 1 and Figure 2 It can be seen that both algorithms can converge within a finite number of iterations, but the ICOA-SVM algorithm converges faster and obtains a better fitness value. Specifically, the COA-SVM algorithm stabilizes after approximately the 18th iteration, with an optimal fitness value of approximately 0.0376388; the ICOA-SVM algorithm stabilizes after approximately the 13th iteration, with an optimal fitness value of approximately 0.0375155. This indicates that the improved strategy introduced in this invention effectively enhances the algorithm's global search capability, local exploitation capability, and optimization accuracy.

[0054] To verify the advantages of the ICOA-SVM model in predicting the safety status classification results of transmission towers, numerical experiments were conducted to compare and analyze this model with the SVM model and the COA-SVM model. The comparison results are shown below. Figures 3-5 .Depend on Figures 3 to 5 It can be seen that the prediction results of the ICOA-SVM model have the highest degree of fit with the real category sequence, the classification points are more concentrated, and the number of error points is the fewest. This indicates that it has a stronger ability to identify the category boundaries of the safety status of power transmission towers and has better classification stability and generalization performance.

[0055] When faced with the same set of transmission tower safety status data, it is evident that the ICOA-SVM model has a better prediction effect than the other two models, with the smallest error control, only 5 sets of error data, and a prediction accuracy of 94.7368%. In contrast, the COA-SVM model has 9 sets of error data, with a prediction accuracy of 90.5263%. The SVM model without population algorithm optimization has 14 sets of error data, with an accuracy of 85.4167%, indicating that the traditional SVM model has poor prediction performance for classification results when not optimized.

[0056] To better illustrate the higher accuracy of the ICOA-SVM model in predicting the safety status classification results of transmission towers, the prediction performance of the three models was compared by calculating the root mean square error (RMSE), mean absolute percentage error (MAPE), and mean absolute error (MAE). The calculation results are shown in Table 1.

[0057]

[0058] As shown in Table 1, the SVM model exhibits the highest error values ​​across all categories, with a MAPE of 0.0911 and a MAE of 0.2083, indicating significantly lower prediction accuracy compared to the other two algorithms. In contrast, the ICOA-SVM model demonstrates the highest accuracy in predicting the safety status of transmission towers, with RMSE of 0.2294, MAPE of 0.0368, and MAE of 0.0521, and the lowest accuracy across all three error categories. Compared to the SVM model, the ICOA-SVM model reduces these three errors by 62.5%, 59.6%, and 75.0%, respectively, demonstrating its ability to effectively predict the safety status of transmission towers and provide a reference for their safe operation.

[0059] This invention provides a method for predicting the safety status of transmission towers based on an improved strategy ICOA-SVM, which has the following characteristics: (1) When predicting the safety status of transmission towers, the safety status of transmission towers is comprehensively judged by using data such as wind speed, wind direction angle, ice thickness, stress and displacement of transmission towers under different working conditions.

[0060] (2) By combining chaotic mapping to optimize the initial population, introducing an adaptive differential guidance strategy to improve population updates, integrating the secretary bird attack strategy and adding a predator avoidance behavior strategy, the randomness, traversal and balance capabilities of the algorithm are enhanced from multiple dimensions such as population initialization, exploration and development mechanism and search strategy coordination, so as to improve search performance and optimization accuracy.

[0061] (3) The improved raccoon optimization algorithm (ICOA) was used to optimize the penalty factor c and kernel function parameter g in the support vector machine (SVM), which improved the accuracy and efficiency of SVM in predicting the safety status of transmission towers.

[0062] (4) An improved strategy-based ICOA-SVM method for predicting the safety status of transmission towers was proposed, which solved the problems of difficulty in assessing the safety status of existing transmission towers due to natural disasters and insufficient accuracy of existing assessment methods, and effectively achieved accurate prediction of the safety status of transmission towers.

[0063] References: Reference [4]: ​​Wang Yan, Zhou Fangzheng, Zhang Bo, Wang Jinmei, Wang Xiaoqing. Optimization configuration method of distribution network SOP based on ICOA-KM cluster partitioning [J / OL]. Journal of North China Electric Power University (Natural Science Edition), 2025 [2025-03-24]. Reference [5]: Zhang Jian, Zhao Mi, Huang Yi, Li Jingyun. Research on photovoltaic array fault diagnosis method based on CLD-COA-ELM [J]. Acta Energiae Solaris Sinica, 2025, 46(1): 632-640. Reference [6]: Zhong Weijin, Jiang Qiu, Zhu Haifeng, Liu Chunlin, Zheng Xubin. Fault feature identification and analysis of power generation equipment based on improved SVM [J]. Electronic Design Engineering, 2025, 33(6): 122-126. Reference [7]: Hu Chunxiang, Meng Fanyong, Luo Wenjiang, Zhou Shuang, Xu Xuan. A method for wind turbine bearing fault diagnosis based on CEEMDAN and ISSA optimized SVM [J]. Mechanical Design, 2025, 42(4): 109-119. Reference [8]: Peng Qingyuan, Wang Xiaofeng, Tang Ao, Wang Junxia, ​​Hua Yingying, He Fei. Intrusion detection method based on gray wolf algorithm to optimize DBN-SVM [J]. Journal of Nanjing University (Natural Science), 2025, 61(2): 270-282.

Claims

1. A power transmission tower safety state prediction method based on ICOA-SVM, characterized in that Includes the following steps: Step 1: Initialize the population of the Raccoon Optimization Algorithm (COA) using the Tent-Logistic-Cosine combined chaotic mapping strategy and set the fitness function; Step 2: Adaptive dynamic differential guidance strategy is adopted to improve the population update method in the exploration phase. The Secretary Bird Optimization Algorithm (SBOA) attack strategy is introduced to enhance local exploitation capabilities. A predator avoidance behavior strategy is added to improve the balance between global search and local exploitation, resulting in the improved ICOA algorithm. Step 3: Following steps 1 and 2, update the raccoon population, calculate the fitness value of each raccoon and its corresponding optimal position. When the maximum number of iterations is reached, the optimal position corresponding to its minimum fitness value is taken as the parameter vector. , ],in: The corresponding penalty parameter c for Support Vector Machine (SVM) Corresponding kernel function parameter g; Step 4: Establish a finite element model of the transmission tower to simulate wind and ice load conditions under different wind angles, wind speeds and ice thicknesses; classify the safety status levels of the transmission tower, form a sample dataset, and divide the sample data into training and test sets after normalization. Step 5: Assign the penalty parameter c and kernel function parameter g obtained in Step 3 to the support vector machine (SVM) to construct the parameter-optimized SVM model; input the training set obtained in Step 4 into the SVM model for training to obtain the ICOA-SVM prediction model for predicting the safety status of power transmission towers. Step 6: Input the wind speed, wind direction angle, and icing thickness of the transmission tower to be evaluated into the trained ICOA-SVM prediction model, and output the safety status level of the transmission tower.

2. The ICOA-SVM based power transmission tower safety state prediction method according to claim 1, characterized in that: In step 1, the Tent-Logistic-Cosine combined chaotic mapping combines the Tent mapping, Logistic mapping and cosine modulation function to enhance the randomness and ergodicity of the sequence. The Tent mapping part uses piecewise linear mappings 2rx and 2r(1) x) Improve traversal, making the sequence cover the search space more evenly; Logistic mapping part, by 4(1 r)x(1 x) introduces a nonlinear relationship, avoids the sequence into a periodic orbit, and improves the chaotic characteristics; In the cosine modulation function part, the cosine function is used to transform the linear and nonlinear terms, compress the result to the interval [-1,1], and normalize it to [0,1]. The periodic function is used to increase the complexity of the iteration and improve the randomness of the chaotic sequence. The mathematical expression for the Tent-Logistic-Cosine combinatorial chaotic mapping is as follows: （2）； In formula (2): The initial value of the chaotic sequence is represented by rand, which is randomly generated and follows a uniform distribution of (0,1). Indicates the first Chaotic variables in the next iteration; The mixing coefficient represents the weight ratio of the Tent mapping and the Logistic mapping, and is randomly generated by rand, taking values ​​(0,1). The fixed parameters 2 and 4 represent the standard parameter extensions of the Tent mapping, used for the linear terms in the left and right halves and the chaotic parameters of the Logistic mapping, respectively, to ensure that the Logistic part is in a chaotic state when r = 0. 0.5 is the offset used to adjust the phase of the input to the cosine function. The mathematical expression for the fitness function is shown in equation (1): （1）； In formula (1): Indicates the fitness value; Indicates the first The true category label of each sample; Indicates the first Predicted class labels for each sample; This represents the total number of samples.

3. The method for predicting the safety status of transmission towers based on ICOA-SVM according to claim 2, characterized in that: In step 2, the adaptive dynamic differential guidance strategy incorporates the mutation mechanism of differential evolution (DE): 2.1: Introduce a dynamic scaling factor F and use a sine function to control the dynamic scaling factor F, so that the algorithm can enhance its exploration ability in the early stage and improve its local development ability in the later stage. 2.2: To avoid the population becoming overly reliant on the current optimal solution, a random individual perturbation mechanism is introduced, involving three random individuals. , , Form a difference vector; 2.3: During the mutation process, add an elite guiding factor, that is, introduce the current best individual. This information is used to enhance global search capabilities while accelerating the convergence of the population towards the optimal solution; （3）； （4）； In the above formula: This represents the current iteration number; This represents the maximum number of iterations. , , For randomly selected individuals, r1≠r2≠r3≠i, where i represents the ID of the individual to be updated; F is the dynamic scaling factor. This represents a random number within the interval (0,1); This is the current optimal value; This indicates the updated individual position.

4. The method for predicting the safety status of transmission towers based on ICOA-SVM according to claim 3, characterized in that: In step 2, the Secretary Bird Optimization Algorithm (SBOA) attack strategy still adopts the random perturbation mode of the original Raccoon Optimization Algorithm (COA) in the case of 50% probability; in the other 50% probability, the Secretary Bird attack strategy is introduced, which uses the contraction factor CF to dynamically reduce the search step size in the later stage of the iteration to enhance the local exploitation capability. At the same time, the Levy flight mechanism is adopted to enable individuals to make long-range jumps in the search space, thereby improving the ability to escape local optima.

5. The method for predicting the safety status of transmission towers based on ICOA-SVM according to claim 4, characterized in that: The Secretary Bird Optimization Algorithm (SBOA) attack strategy formula is as follows: （5）； （6）； （7）； In the above formula, It is a contraction factor, which enhances exploration in the early stage and promotes algorithm development in the later stage. The Levi flight vector enables long-distance jumps through a heavy-tailed distribution. Represents the Hadamard product; This represents the current optimal solution; Indicates the updated position; Indicates the current position; Indicates dimension; t is the current iteration number; This represents the maximum number of iterations.

6. The method for predicting the safety status of transmission towers based on ICOA-SVM according to claim 5, characterized in that: The addition of predator avoidance behavior strategies enables the raccoon optimization algorithm (COA) to dynamically adjust its risk avoidance methods according to different situations during the search process. One method is to simulate the mother bear leading her cubs to quickly evacuate, which enhances the target orientation of the search and improves the optimization efficiency, as shown in formula (8). Another method is to introduce a random cooperative mechanism, which enables individuals to escape irregularly in complex environments, increasing the randomness and diversity of the search, as shown in formula (9). （8）； （9）； In the above formula, This indicates the updated location after the predator avoidance behavior update; Indicates the group's historical best position; Indicates the current iteration number; Indicates the maximum number of iterations; and This represents a random vector located in the range [0,1]. For random vectors Dimensionally consistent vector; Indicates the current individual's location; This indicates an alertness coefficient that takes the value of 1 or 2. Indicates the position of a randomly selected group member; Formula (8) describes the targeted risk avoidance behavior, a strategy applicable when an individual perceives a single threat and is triggered with a 50% probability; the individual will move to the group's historically best position. They move closer together to increase their chances of survival; this process simulates three biological characteristics: first, Representing the safe zone in collective memory, providing individuals with a clear escape route; secondly, the squared decay factor. This is used to adjust the search intensity, so that the threat's impact gradually weakens with iterations, ensuring the algorithm has strong exploration capabilities in the early stages and gradually converges in the later stages; finally, the perturbation term... Random offsets are generated within the interval [-1, 1] to enhance the diversity of escape paths, thereby avoiding the algorithm from getting trapped in local optima. ∈[0,1]D; D is the dimension of the search space; Formula (9) corresponds to the cooperative escape strategy. When an individual perceives multiple threats, it is triggered with a 50% probability. Instead of acting alone, it relies on group cooperation to improve its chances of survival. This mechanism randomly selects a group member. As a target to be followed, it simulates the phenomenon of collective escape to avoid the risks of isolated actions; the parameter B∈{1,2} is used to adjust the individual's vigilance level at the current location, realizing dual-mode defense to adapt to search needs in different environments; at the same time, the step size random factor The ∈[0,1]D makes the escape behavior nondeterministic, further increasing the coverage of the search space.

7. The method for predicting the safety status of transmission towers based on ICOA-SVM according to claim 6, characterized in that: In step 3, after obtaining the ICOA algorithm through steps 1 and 2, the fitness value of each raccoon individual at the current position is first calculated according to equation (1); then the position of each individual is updated according to equations (3) to (9), and the fitness values ​​before and after the update are compared. The position with better fitness is retained as the current optimal position of the individual; at the same time, the entire population is traversed, and the position of the individual with the smallest fitness is recorded as the current global optimal position. During this process, continuous iteration and optimization are carried out. Each iteration updates the population based on the adaptive dynamic differential guidance strategy, the Secretary Bird Optimization Algorithm (SBOA) attack strategy, and the newly added predator avoidance behavior strategy. The algorithm stops updating when it reaches the maximum number of iterations. The optimal position corresponding to the minimum fitness value obtained at this point is the parameter vector. , ],in: The corresponding penalty parameter c for support vector machines, Corresponding kernel function parameter g; By assigning this parameter vector to the SVM, we obtain the SVM evaluation model with the best classification performance.

8. The method for predicting the safety status of transmission towers based on ICOA-SVM according to claim 7, characterized in that: In step 4, a finite element model of the transmission tower is established to simulate the wind and ice load conditions under different wind angles, wind speeds and ice thicknesses; based on the relationship between the component stress ratio ξ and the tower top displacement l and tower height h, the safety status categories are divided into four categories: safe, relatively safe, basically safe and unsafe. A sample dataset with security status labels is generated, and the sample data is normalized and then divided into training and test sets.

9. The method for predicting the safety status of transmission towers based on ICOA-SVM according to claim 8, characterized in that: In step 5, the optimal penalty parameter c and kernel function parameter g obtained in step 3 are assigned to the support vector machine to construct the parameter-optimized SVM classification model. Then, the training set samples formed and preprocessed in step 4 are input into the SVM classification model for training, so that the model learns the mapping relationship between wind speed, wind direction angle and ice thickness and the safety status level of the transmission tower, thereby obtaining the ICOA-SVM prediction model for predicting the safety status of the transmission tower.

10. The method for predicting the safety status of transmission towers based on ICOA-SVM according to claim 9, characterized in that: In step 6, the wind speed, wind direction angle, and icing thickness of the transmission tower to be evaluated are used as input features and input into the trained ICOA-SVM prediction model. The prediction model classifies and identifies the samples to be evaluated based on the mapping relationship between the input features and the safety status level established during the training process, and outputs the corresponding safety status level of the transmission tower, so as to achieve rapid prediction and evaluation of the operating status of the transmission tower.

Images