An optimization design method, system, medium and program product of a switch cabinet structure

Abstract

The application provides an optimization design method, system, medium and program product of a switch cabinet structure, the method comprises the following steps: constructing a switch cabinet parameterized structure model, taking size and structure parameters as variables, taking material consumption minimization as a target, taking strength, temperature rise, insulation distance and assembly interference as constraints, and coding to generate an initial population; in iteration, the feasibility of the constraints is determined, individuals that do not violate the constraints are stored in a feasible solution set, a base vector and a differential individual are selected in the mutation stage to calculate a difference vector, the direction of the difference vector is guided by combining the feasible solution guide points, a mutation vector is generated and a candidate solution is obtained; the candidate solution is subjected to boundary adjustment, assembly correlation linkage adjustment and manufacturing specification rounding, a test individual is obtained, the objective function and constraint violation degree of the parent and the test individual are compared, and the optimization design result of the switch cabinet structure is output by iteration to convergence.

Description

Technical Field

[0001] This application belongs to the field of optimization, and in particular relates to an optimization design method, system, medium and program product for switch cabinet structure. Background Technology

[0002] With the rapid development of modern power systems and the continuous advancement of smart grid construction, switchgear, as a key piece of equipment in power transmission and distribution networks that plays a role in controlling, protecting, and distributing electrical energy, is experiencing a rapidly growing market demand. In actual operating environments, the structure of switchgear not only needs sufficient mechanical strength to withstand the electrodynamic impact of short circuits, but also must meet stringent temperature rise limits under high current conditions. Simultaneously, legally mandated insulation safety distances must be maintained between internal conductors and between live parts and the casing. Traditional switchgear structural layouts rely on engineers' experience through trial and error and repeated modifications. To absolutely guarantee equipment safety, extremely conservative structural parameters and excessive material margins are often adopted. This traditional development model not only leads to serious material redundancy and high manufacturing costs, but also results in inefficient manual adjustments due to the mutual constraints among multiple factors, making it impossible to achieve an optimal trade-off between structural safety and minimizing material usage.

[0003] Conventional difference algorithms suffer from highly arbitrary mutation operations, relying solely on the differences between random individuals to generate difference vectors, failing to identify and utilize the topological distribution of the feasible region. Under the pressure of multiple stringent constraints such as switchgear strength, temperature rise, and insulation, the feasible solution space is extremely narrow and nonlinearly distributed. Blind mutation leads to a large number of generated candidate individuals falling into infeasible regions, resulting in ineffective computational losses, low convergence efficiency, and a high susceptibility to local stagnation. Existing algorithms often optimize isolated variables within a purely mathematical continuous space, detached from the actual manufacturing and assembly processes of switchgear. The output continuous theoretical dimensions cannot directly match standard manufacturing specifications, rendering the optimization results unusable for production. Furthermore, neglecting the assembly topological relationships between structural components, adjustments to the dimensions of a single component fail to correlate with related variables, easily leading to severe assembly interference between components in three-dimensional space. This renders the optimal solution obtained with significant computational effort completely infeasible in practical engineering. Summary of the Invention

[0004] To address the problems of existing technologies, such as excessive material consumption, low convergence efficiency under multiple constraints, and difficulty in meeting actual manufacturing and assembly requirements.

[0005] In a first aspect, the present invention provides an optimized design method for a switch cabinet structure, comprising: A parameterized structural model of the switchgear is constructed, and the dimensions and structural parameters are extracted as variables. A function with the minimum material usage as the objective function is established, and a constraint function with strength, temperature rise, insulation distance and assembly interference as constraints is established. The variables are encoded to generate an initial population. A feasibility judgment loop is deployed in each generation evolution cycle. Individuals in the current population whose total degree of violation of any constraint is zero are stored in the feasible solution set. For each mutated individual, a basis vector individual and two difference individuals are selected to calculate the original difference vector. The angle between the original difference vector and the reference vector is calculated. The reference vector points from the current individual to the guiding point calculated based on the feasible solution set. When the angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector. The parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate a mutated vector constrained by the angle. The mutated vector is superimposed with the basis vector individuals to obtain the candidate solution. The candidate solutions are subjected to out-of-bounds projection adjustment, and the associated variables are adjusted in a linked manner based on the assembly association matrix to meet the assembly requirements. The dimensional parameters in the adjusted candidate solutions are rounded according to the manufacturing specifications to obtain the test individuals. The constraint violation degree and objective function value of the test individuals are compared with those of the parent individuals. The optimal individual is retained and iterated until the termination condition is met, and the switch cabinet structure optimization design results are output.

[0006] Furthermore, the establishment of a constraint function with the objective function of minimizing material usage and constraints of strength, temperature rise, insulation distance, and assembly interference includes: The objective function is to be minimized by the sum of the products of the volumes of all sheet metal parts and load-bearing frames and the corresponding physical material densities in the parameterized structural model of the switch cabinet. Extract the maximum nodal stress value generated by the skeleton under rated electrodynamic impact simulation, and set the maximum nodal stress value being greater than a preset stress threshold as a strength constraint violation; Extract the temperature peak at the connection of each phase bus inside the cabinet, and set the temperature peak exceeding the preset temperature threshold as a violation of the temperature rise constraint; A three-dimensional distance matrix is ​​established between the control points of the conductive busbar and the grounding box shell. The minimum value of the term in the three-dimensional distance matrix that is less than the preset insulation distance threshold is set as an insulation distance constraint violation. The total interference value is formed by accumulating the overlapping three-dimensional volumes of mutually intersecting components within the parameterized structural model. A total interference value greater than zero is defined as a violation of assembly interference constraints.

[0007] Furthermore, the reference vector points from the current individual to a guiding point calculated based on the feasible solution set, including: Within each generation of evolution, a feasibility determination loop is deployed, and individuals in the current population whose total degree of violation of any constraint is zero are stored in the set of feasible solutions. Read the number of feasible solutions contained in the feasible solution set. When the number of feasible solutions is greater than zero, sum the parameter values ​​of the feasible solutions in each dimension and divide by the number of feasible solutions to calculate the statistical average displacement in each independent dimension. When the number of feasible solutions is zero, the parameter value of the individual with the smallest constraint violation in the current population is used as the statistical average displacement in each independent dimension. The statistical average displacements of all dimensions are sequentially cascaded and integrated to form a unique guiding point in the multidimensional variable space; The displacement variable values ​​of the guiding points are extracted bit by bit and the corresponding dimension displacement variable values ​​of the current individual to be mutated are subtracted to generate a multidimensional difference vector, which is then used as a reference vector to drive evolution.

[0008] Further, when the included angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector, the parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate an angle-constrained variant vector, and the variant vector is superimposed with the individual basis vectors to obtain a candidate solution, including: Calculate the vector difference between the first and second difference individuals to produce a multi-dimensional original difference vector; Calculate the dot product of the original difference vector and the reference vector, calculate the product of the norm of the original difference vector and the norm of the reference vector, if the product is greater than zero, calculate the inverse cosine function value of the ratio of the two as the included angle, if the product is zero, set the included angle to zero; When the included angle is greater than the preset limit of ninety degrees, the original difference vector is decomposed into parallel and perpendicular components along the reference vector; The parallel component, which is opposite to the direction of the reference vector, is reversed to the component with the same direction as the reference vector to eliminate the back-inference factor, and then resynthesized with the vertical component to establish a completely new evolutionary mutation quantity. The evolutionary mutation amount is superimposed with the corresponding parameters of the basis vector individuals to generate an angle-constrained mutation vector, thereby obtaining the candidate solution.

[0009] Further, the step of adjusting the candidate solution for out-of-bounds projection includes: When it is detected that the variable value of any dimension in the candidate solution is less than the lower limit warning value, the bottom-level cross-interval scale is obtained by subtracting the variable value of the corresponding dimension of the basis vector individual from the lower limit warning value. A random decimal number between zero and one is generated and multiplied with the bottom-level cross-interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the out-of-bounds adjustment. When it is detected that the variable value of any dimension in the candidate solution is greater than the upper limit warning value, the upper limit warning value is subtracted from the variable value of the corresponding dimension of the basis vector individual to obtain the upper limit crossing interval scale. A random decimal number between zero and one is generated and multiplied with the upper limit crossing interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the over-limit adjustment.

[0010] Furthermore, the step of adjusting the associated variables in a coordinated manner based on the assembly association matrix to meet assembly requirements includes: Extract the component primitives that are fixedly attached to each other, establish an assembly association matrix with a dimension equal to the number of components, mark the corresponding array units of components with physical contact and overlapping hole mating as one, and mark the units with no matching relationship as zero. Monitor the change in the boundary position of the contact surface or mounting hole of the variable drive component with parameter fine-tuning during the dimensional adjustment process; Read the follower part number column marked as 1 in the assembly association matrix line by line, and implant the same position compensation amount as the boundary position change amount into the corresponding passive mating part to complete the linkage adjustment between the group hole flush and the locking plate.

[0011] Furthermore, comparing the constraint violation degree and objective function value between the experimental individual and the parent individual, and retaining the optimal individual, includes: The violation margin values ​​of the test individual and the parent individual exceeding the warning line in each constraint are calculated separately. After normalizing each violation margin value, they are combined and accumulated to obtain the total violation degree of the test individual and the total violation degree of the parent individual. If the total violation of the experimental individual is greater than zero and the total violation of the parent individual is zero, then the experimental individual is eliminated and the parent individual is retained and directly transitioned to the next generation of the population. If the total violation of the experimental individual is zero and the total violation of the parent individual is greater than zero, then the parent individual is eliminated and the experimental individual is retained and directly transitioned to the next generation of the population. If the total deviation of both the experimental individual and the parent individual is greater than zero, then the total deviation of the two individuals is compared, and the individual with the smallest total deviation enters the next generation of the population. If the total deviation of both the experimental individual and the parent individual is zero, then the objective function values ​​of the two individuals are compared, and the individual with the smallest objective function value enters the next generation of the population.

[0012] In a second aspect, the present invention also provides an optimized design system for switchgear structure, comprising: The construction module is used to build a parameterized structural model of the switchgear, extract dimensions and structural parameters as variables, establish a function with minimizing material usage as the objective function, establish a constraint function with strength, temperature rise, insulation distance and assembly interference as constraints, encode the variables to generate an initial population, deploy a feasibility judgment loop in each generation evolution cycle, and store individuals in the current population whose total degree of violation of any constraint is zero into the feasible solution set. The production module is used to select a basis vector individual and two difference individuals for the individual to be mutated to calculate the original difference vector, and to calculate the angle between the original difference vector and the reference vector. The reference vector points from the current individual to the guiding point calculated based on the feasible solution set. When the angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector, the parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate a mutation vector constrained by the angle, and the mutation vector is superimposed with the basis vector individual to obtain the candidate solution. The output module is used to perform out-of-bounds projection adjustment on the candidate solutions, and to perform linkage adjustment on the associated variables based on the assembly association matrix to meet the assembly requirements. The size parameters in the adjusted candidate solutions are rounded according to the manufacturing specifications to obtain the test individuals. The constraint violation degree and objective function value of the test individuals are compared with those of the parent individuals. The optimal individual is retained and iterated until the termination condition is met, and the switch cabinet structure optimization design results are output.

[0013] Preferably, the establishment of a constraint function with the objective function of minimizing material usage and constraints of strength, temperature rise, insulation distance, and assembly interference includes: The objective function is to be minimized by the sum of the products of the volumes of all sheet metal parts and load-bearing frames and the corresponding physical material densities in the parameterized structural model of the switch cabinet. Extract the maximum nodal stress value generated by the skeleton under rated electrodynamic impact simulation, and set the maximum nodal stress value being greater than a preset stress threshold as a strength constraint violation; Extract the temperature peak at the connection of each phase bus inside the cabinet, and set the temperature peak exceeding the preset temperature threshold as a violation of the temperature rise constraint; A three-dimensional distance matrix is ​​established between the control points of the conductive busbar and the grounding box shell. The minimum value of the term in the three-dimensional distance matrix that is less than the preset insulation distance threshold is set as an insulation distance constraint violation. The total interference value is formed by accumulating the overlapping three-dimensional volumes of mutually intersecting components within the parameterized structural model. A total interference value greater than zero is defined as a violation of assembly interference constraints.

[0014] Preferably, the reference vector points from the current individual to a guiding point calculated based on the feasible solution set, including: Within each generation of evolution, a feasibility determination loop is deployed, and individuals in the current population whose total degree of violation of any constraint is zero are stored in the set of feasible solutions. Read the number of feasible solutions contained in the feasible solution set. When the number of feasible solutions is greater than zero, sum the parameter values ​​of the feasible solutions in each dimension and divide by the number of feasible solutions to calculate the statistical average displacement in each independent dimension. When the number of feasible solutions is zero, the parameter value of the individual with the smallest constraint violation in the current population is used as the statistical average displacement in each independent dimension. The statistical average displacements of all dimensions are sequentially cascaded and integrated to form a unique guiding point in the multidimensional variable space; The displacement variable values ​​of the guiding points are extracted bit by bit and the corresponding dimension displacement variable values ​​of the current individual to be mutated are subtracted to generate a multidimensional difference vector, which is then used as a reference vector to drive evolution.

[0015] Preferably, when the included angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector, the parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate an angle-constrained variant vector, and the variant vector is superimposed with the individual basis vectors to obtain a candidate solution, including: Calculate the vector difference between the first and second difference individuals to produce a multi-dimensional original difference vector; Calculate the dot product of the original difference vector and the reference vector, calculate the product of the norm of the original difference vector and the norm of the reference vector, if the product is greater than zero, calculate the inverse cosine function value of the ratio of the two as the included angle, if the product is zero, set the included angle to zero; When the included angle is greater than the preset limit of ninety degrees, the original difference vector is decomposed into parallel and perpendicular components along the reference vector; The parallel component, which is opposite to the direction of the reference vector, is reversed to the component with the same direction as the reference vector to eliminate the back-inference factor, and then resynthesized with the vertical component to establish a completely new evolutionary mutation quantity. The evolutionary mutation amount is superimposed with the corresponding parameters of the basis vector individuals to generate an angle-constrained mutation vector, thereby obtaining the candidate solution.

[0016] Preferably, the step of adjusting the candidate solution by out-of-bounds projection includes: When it is detected that the variable value of any dimension in the candidate solution is less than the lower limit warning value, the bottom-level cross-interval scale is obtained by subtracting the variable value of the corresponding dimension of the basis vector individual from the lower limit warning value. A random decimal number between zero and one is generated and multiplied with the bottom-level cross-interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the out-of-bounds adjustment. When it is detected that the variable value of any dimension in the candidate solution is greater than the upper limit warning value, the upper limit warning value is subtracted from the variable value of the corresponding dimension of the basis vector individual to obtain the upper limit crossing interval scale. A random decimal number between zero and one is generated and multiplied with the upper limit crossing interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the over-limit adjustment.

[0017] Preferably, the step of adjusting the associated variables based on the assembly association matrix to meet the assembly requirements includes: Extract the component primitives that are fixedly attached to each other, establish an assembly association matrix with a dimension equal to the number of components, mark the corresponding array units of components with physical contact and overlapping hole mating as one, and mark the units with no matching relationship as zero. Monitor the change in the boundary position of the contact surface or mounting hole of the variable drive component with parameter fine-tuning during the dimensional adjustment process; Read the follower part number column marked as 1 in the assembly association matrix line by line, and implant the same position compensation amount as the boundary position change amount into the corresponding passive mating part to complete the linkage adjustment between the group hole flush and the locking plate.

[0018] Preferably, comparing the constraint violation degree and objective function value between the experimental individual and the parent individual, and retaining the optimal individual, includes: The violation margin values ​​of the test individual and the parent individual exceeding the warning line in each constraint are calculated separately. After normalizing each violation margin value, they are combined and accumulated to obtain the total violation degree of the test individual and the total violation degree of the parent individual. If the total violation of the experimental individual is greater than zero and the total violation of the parent individual is zero, then the experimental individual is eliminated and the parent individual is retained and directly transitioned to the next generation of the population. If the total violation of the experimental individual is zero and the total violation of the parent individual is greater than zero, then the parent individual is eliminated and the experimental individual is retained and directly transitioned to the next generation of the population. If the total deviation of both the experimental individual and the parent individual is greater than zero, then the total deviation of the two individuals is compared, and the individual with the smallest total deviation enters the next generation of the population. If the total deviation of both the experimental individual and the parent individual is zero, then the objective function values ​​of the two individuals are compared, and the individual with the smallest objective function value enters the next generation of the population.

[0019] This invention constructs a parameterized model that comprehensively considers multiple constraints such as strength, temperature rise, insulation distance, and assembly interference, thereby reducing the amount of materials used in switchgear while fully ensuring mechanical and electrical safety performance. During the evolutionary optimization process, this invention utilizes a difference vector angle constraint mechanism guided by the feasible solution set. By restricting and adjusting the angle between the original difference vector and the reference vector, components deviating from the feasible region are reversed and synthesized, guiding the population to evolve towards the feasible region. This improves the convergence speed and global optimization capability, avoiding meaningless computational losses. By combining the assembly correlation matrix to adjust the correlation variables of candidate solutions and performing manufacturing specification rounding, not only are the assembly interference risks between components eliminated, but the output parameters also ensure that they conform to actual industrial processing standards, improving the engineering practicality of the optimization results and achieving optimization of the switchgear structural layout. Attached Figure Description

[0020] Figure 1 A flowchart illustrating an optimized design method for a switchgear structure; Figure 2 A schematic diagram illustrating the optimization effect of the distribution range for the out-of-bounds projection adjustment mechanism. Figure 3 This is a schematic diagram comparing the convergence effect of total material mass. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Example

[0022] In Embodiment 1 of the present invention, an optimized design method for a switchgear structure is provided, such as... Figure 1 As shown, it includes: S1. Construct a parameterized structural model of the switchgear, extract dimensions and structural parameters as variables, establish a function with minimizing material usage as the objective function, and establish a constraint function with strength, temperature rise, insulation distance and assembly interference as constraints. Encode the variables to generate an initial population, deploy a feasibility judgment loop in each generation evolution cycle, and store individuals in the current population whose total degree of violation of any constraint is zero into the feasible solution set.

[0023] A 3D geometric model of the switchgear was established using the FreeCAD secondary development interface based on Python. The thickness of the busbars, the thickness of the cabinet panels, and the spacing of the insulation components were extracted as real variables using document object acquisition methods. An objective function was established as the summation function of the product of the volume of each component and its corresponding material density. The maximum equivalent stress was calculated using the transient dynamics analysis module of the ANSYS parametric modeling language interface to construct the strength constraint function. The steady-state temperature distribution was calculated using the thermal analysis module to construct the temperature rise constraint function. The spatial electric field gradient was calculated using the finite difference method to solve the Poisson equation to construct the insulation distance constraint function. The collision volume of components was calculated using a Boolean intersection algorithm based on an open-source boundary representation geometric kernel to construct the assembly interference constraint function. Using the SOBOb sequence generator in the NumPy numerical computing library, each variable is encoded as a real number within a given upper and lower bound space, generating an initial two-dimensional array of individuals. In each iteration of the differential evolution algorithm, conditional decision logic based on a loop control structure is deployed. Each individual in the population matrix is ​​traversed and substituted into four constraint functions for calculation. If the calculation result exceeds the safety threshold, the difference is calculated using the maximum value function as the constraint violation degree; otherwise, it is set to zero. The constraint violation degrees of each individual are accumulated, and the total violation degree is calculated. When the total violation degree is determined to be zero, the data row of that individual is pushed into a pre-established feasible solution set matrix using an array concatenation function.

[0024] The SOBOOL sequence generator is a deterministic algorithm for generating low-discrepancy sequences. Its core objective is to distribute the generated sample points as evenly as possible across the entire parameter space. It does not generate truly random numbers, but rather a quasi-random sequence, aiming to achieve more efficient search coverage than purely random numbers. In an alternative embodiment, the establishment of a constraint function with the objective function of minimizing material usage and constraints of strength, temperature rise, insulation distance, and assembly interference includes: The objective function is to be minimized by the sum of the products of the volumes of all sheet metal parts and load-bearing frames and the corresponding physical material densities in the parameterized structural model of the switch cabinet. Extract the maximum nodal stress value generated by the skeleton under rated electrodynamic impact simulation, and set the maximum nodal stress value being greater than a preset stress threshold as a strength constraint violation; Extract the temperature peak at the connection of each phase bus inside the cabinet, and set the temperature peak exceeding the preset temperature threshold as a violation of the temperature rise constraint; A three-dimensional distance matrix is ​​established between the control points of the conductive busbar and the grounding box shell. The minimum value of the term in the three-dimensional distance matrix that is less than the preset insulation distance threshold is set as an insulation distance constraint violation. The total interference value is formed by accumulating the overlapping three-dimensional volumes of mutually intersecting components within the parameterized structural model. A total interference value greater than zero is defined as a violation of assembly interference constraints.

[0025] The calculation of the objective function for the switchgear involves extracting the geometric volume variables of components such as the frame angle steel, side plates, and door panels in the model one by one, and multiplying them by the corresponding material density, such as the density of Q235 steel. kg / m 3 The density of the copper busbar is taken as follows kg / m 3 The mass of all frames and sheet metal parts is calculated and summed to obtain the total mass as the objective function value, where the sheet metal thickness variable range is selected between 1.5mm and 3.0mm. Simulation constraint parameters of the physical field are set to implement the evaluation calculation. The switchgear is imported into the finite element analysis software, and a rated short-time withstand current of 50kA for 1s is applied for electrodynamic simulation calculation. The Von Mises stress value of the node with the maximum deformation stress of the frame is extracted and compared with a preset stress threshold set according to the yield strength of Q235 divided by the safety factor. If it is greater than 180MPa, it is determined to be a violation of the strength constraint, and the excess difference violation degree is recorded. Steady-state temperature simulation node data of the busbar lap surface and incoming / outgoing terminals under a rated current of 4000A are extracted to obtain the highest absolute temperature value among the three phases. If the temperature value exceeds the threshold corresponding to the busbar temperature rise limit, it is considered a violation of the temperature rise constraint.

[0026] Regarding insulation distance, a control point array set is constructed by parametrically extracting the surface node coordinates between the three-phase busbars and the metal shell inside the cabinet. All Euclidean distance matrices between the conductive busbar node set and the shell node set are calculated. The KD-Tree algorithm is used to search for the minimum value of the distance matrix, and it is determined whether the minimum value is less than a set insulation distance threshold. If it is less, it is considered a violation of insulation constraints. For assembly interference detection, the CAD kernel is called to calculate the total spatial intersection volume formed by the mutual intrusion of all components under the assembly. If the overlap volume is greater than 0 mm... 3 This indicates that an assembly interference collision has occurred, and a multi-dimensional normalized constraint set is constructed accordingly for penalty determination calculation.

[0027] S2, select a basis vector individual and two difference individuals to calculate the original difference vector, and calculate the angle between the original difference vector and the reference vector. The reference vector points from the current individual to the guiding point calculated based on the feasible solution set. When the angle is greater than a preset limit value, decompose the original difference vector along the reference vector, reverse the parallel components with opposite directions and synthesize them with the perpendicular components to generate a mutation vector constrained by the angle. Superimpose the mutation vector with the basis vector individual to obtain the candidate solution.

[0028] For the individual to be mutated in the current evolutionary generation, the integer sampling function without replacement in NumPy's random number module is used to randomly select three integers not equal to the current individual's index from the population index, corresponding to the basis vector individual, the first difference individual, and the second difference individual, respectively. The original difference vector in multidimensional space is generated by subtracting the coordinate array of the second difference individual from the coordinate array of the first difference individual. The arithmetic mean center of all feasible individuals is calculated in the feasible solution set matrix as the guiding point. The coordinates of the guiding point are subtracted from the coordinates of the current individual to generate the reference vector. The Euclidean angle between the original difference vector and the reference vector is calculated using the formula of dividing the vector dot product by the product of their magnitudes, combined with the inverse cosine function. When this angle is determined to be greater than the set orthogonality limit of 90 degrees, the projection vector of the original difference vector onto the reference vector is calculated as the parallel component. The parallel component is subtracted from the original difference vector to obtain the perpendicular component. Since the included angle is greater than 90 degrees, the parallel component is opposite to the direction of the reference vector. The parallel component is multiplied by -1 to reverse it. The reversed parallel component is then combined with the perpendicular component by vector addition to generate a mutated vector constrained by the angle. A scaling factor is introduced, and this scaling factor is multiplied by the mutated vector constrained by the angle. Then, it is added element by element to the coordinate array of the individual basis vectors to generate the coordinate array of the candidate solution.

[0029] In one embodiment, the reference vector points from the current individual to a guiding point calculated based on the feasible solution set, including: Within each generation of evolution, a feasibility determination loop is deployed, and individuals in the current population whose total degree of violation of any constraint is zero are stored in the set of feasible solutions. Read the number of feasible solutions contained in the feasible solution set. When the number of feasible solutions is greater than zero, sum the parameter values ​​of the feasible solutions in each dimension and divide by the number of feasible solutions to calculate the statistical average displacement in each independent dimension. When the number of feasible solutions is zero, the parameter value of the individual with the smallest constraint violation in the current population is used as the statistical average displacement in each independent dimension. The statistical average displacements of all dimensions are sequentially cascaded and integrated to form a unique guiding point in the multidimensional variable space; The displacement variable values ​​of the guiding points are extracted bit by bit and the corresponding dimension displacement variable values ​​of the current individual to be mutated are subtracted to generate a multidimensional difference vector, which is then used as a reference vector to drive evolution.

[0030] In the decision loop of each generation's evolutionary cycle, the sum of constraint violations for each individual in a population of 50 is evaluated. All feasible solutions with a total violation of 0 are copied and stored in an independent external feasible solution archive. The capacity statistics algorithm is called to obtain the number of individuals K contained in this set. When the capacity K is greater than 0, the sum of the coordinates of the single-dimensional components of K individuals is calculated for a total of 25 dimensions, such as cabinet width, height, and cross-sectional area of ​​the mother row, and then divided by K to obtain the statistical mean center value of the coordinate system of each dimension. If the current stage is in the early stage of evolution and the number of feasible solutions K is equal to 0, the constraint violation values ​​of the current generation's 50 individuals are bubble sorted, and the coordinates of the parameters of the single best infeasible solution with the smallest constraint violation at the top of the list are selected as the provisional statistical mean displacement.

[0031] The calculated statistical average displacements across all 25 dimensions are concatenated to form the coordinates of the guide point in the current 25-dimensional real-valued optimization space. For example, assuming the guide point coordinates in a simplified 3D coordinate system are [800, 2200, 1500], and the coordinates of the individual to be mutated are [810, 2150, 1520], matrix subtraction is used to subtract the displacement variable matrix of the guide point element by element from the displacement variable matrix of the individual to be mutated, thus outputting a multidimensional directed line segment difference vector with a definite length and direction. This multidimensional difference vector possesses the geometric optimization characteristic of directly pointing to the center of the current constraint optimal solution or the local optimal solution region. Inputting this vector into subsequent operations serves as a reference vector guiding the population to move towards the feasible region and the optimal point, ensuring that the population converges globally and stably to the promising region.

[0032] In one embodiment, when the included angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector, the parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate an angle-constrained variant vector, and the variant vector is superimposed with the individual basis vectors to obtain a candidate solution, including: Calculate the vector difference between the first and second difference individuals to produce a multi-dimensional original difference vector; Calculate the dot product of the original difference vector and the reference vector, calculate the product of the norm of the original difference vector and the norm of the reference vector, if the product is greater than zero, calculate the inverse cosine function value of the ratio of the two as the included angle, if the product is zero, set the included angle to zero; When the included angle is greater than the preset limit of ninety degrees, the original difference vector is decomposed into parallel and perpendicular components along the reference vector; The parallel component, which is opposite to the direction of the reference vector, is reversed to the component with the same direction as the reference vector to eliminate the back-inference factor, and then resynthesized with the vertical component to establish a completely new evolutionary mutation quantity. The evolutionary mutation amount is superimposed with the corresponding parameters of the basis vector individuals to generate an angle-constrained mutation vector, thereby obtaining the candidate solution.

[0033] In differential mutation, a first and second differential individual, distinct from the individual to be evolved, are randomly selected from the current population. The coordinate components of their corresponding dimensions are subtracted to generate the original difference vector in multidimensional space. To determine the consistency of the evolutionary direction, the dot product between the original difference vector and the extracted reference vector is calculated, along with the scalar product of the corresponding Euclidean norms of the two vectors. If this norm product is greater than 0, it is substituted into the inverse cosine formula. Calculate the radian value and convert it to the angle in degrees. If the norm product is 0, it means that the vector has degenerated to zero. In order to ensure the continuity of the algorithm, the angle is forced to be 0 degrees.

[0034] A fixed evolutionary limit value is set. If the included angle is greater than a pre-set 90-degree limit threshold, it indicates that the mutation direction generated by the original difference vector deviates from the optimal path guiding the convergence of feasible solutions. At this time, according to the principle of geometric vector projection, the parallel projection component of the original difference vector in the direction of the reference vector is extracted, and the perpendicular component is separated using the orthogonal vector subtraction rule. For the parallel component that generates a regressive evolutionary trend, a scalar multiplication with a coefficient of -1 is applied in the mathematical domain to reverse the component by 180 degrees to form a parallel component in the same direction, thereby eliminating the mutation component that deviates from the feasible region. The reversed parallel component is recombined with the aforementioned unchanged perpendicular component using vector addition, and the reconstructed new evolutionary mutation amount is output. A scaling factor with a value preferably between 0.6 and 0.8 is introduced. The scaled new evolutionary mutation amount is superimposed on the coordinates of the pre-selected basis vector individual through matrix addition, generating a mutation vector with convergence guidance constraints and angle constraints. This candidate solution is output, thereby achieving a high balance between exploration capability and feasibility development efficiency.

[0035] S3, perform out-of-bounds projection adjustment on the candidate solution, and perform linkage adjustment on the correlation variables based on the assembly correlation matrix to meet the assembly requirements. Round the size parameters in the adjusted candidate solution to the manufacturing specifications to obtain the test individual. Compare the constraint violation degree and objective function value of the test individual with the parent individual. Retain the optimal individual and iterate until the termination condition is met. Output the switch cabinet structure optimization design result.

[0036] By comparing the candidate solution coordinates with the physical boundaries of the variables using a limit constraint function, values ​​below the lower boundary are adjusted to the basis vector value plus a random proportion multiplied by (lower boundary - basis vector value); values ​​above the upper boundary are adjusted to the basis vector value plus a random proportion multiplied by (upper boundary - basis vector value). This achieves out-of-bounds projection adjustment. An assembly association matrix is ​​constructed, representing the topological connections of components using Boolean values. By traversing the column elements of the row containing the current driving component in the assembly association matrix, the follower component labels marked as "1" are extracted to identify the association variables where the overlapping area is suspended or overlapped due to size changes. A position compensation amount equal to the boundary position change is implanted into the corresponding passive mating component to meet the assembly connection requirements. The floor function and the floor function are called to match the plate thickness and aperture size parameters in the compensated candidate solutions to the closest discrete numerical sequence in the standard material specification library to obtain the test individual. The Deb feasibility rule is used for selection. The sum of the deviations of the experimental individual and its parent individuals is calculated. If both are greater than zero, the individual with the smaller deviation is retained. If one is equal to zero and the other is greater than zero, the individual with the zero deviation is retained. If both are equal to zero, the individual with the smallest objective function value is retained. This process is placed in an outer loop, and a counter continuously increments the iteration count until the maximum preset iteration count is reached, at which point the algorithm stops running. The data row with the minimum material usage in the set is decoded and restored to output a 3D drawing of the switchgear.

[0037] The Deb feasibility rule is a criterion used to compare the merits of two solutions in constrained optimization problems. The idea is that during the evolutionary process, feasible solutions take precedence over infeasible solutions; only when both solutions are feasible are the objective function values ​​compared.

[0038] In one embodiment, the out-of-bounds projection adjustment of the candidate solution includes: When it is detected that the variable value of any dimension in the candidate solution is less than the lower limit warning value, the bottom-level cross-interval scale is obtained by subtracting the variable value of the corresponding dimension of the basis vector individual from the lower limit warning value. A random decimal number between zero and one is generated and multiplied with the bottom-level cross-interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the out-of-bounds adjustment. When it is detected that the variable value of any dimension in the candidate solution is greater than the upper limit warning value, the upper limit warning value is subtracted from the variable value of the corresponding dimension of the basis vector individual to obtain the upper limit crossing interval scale. A random decimal number between zero and one is generated and multiplied with the upper limit crossing interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the over-limit adjustment.

[0039] A candidate solution boundary detection procedure is set up to scan the multidimensional coordinate sequence generated by the differential algorithm. Assuming that for the busbar copper bus width dimension, the specified lower limit warning value is set to 40mm, and the upper limit warning value is set to 120mm, when a sudden change in the value of this variable in a candidate solution to 30mm, which is less than the lower limit warning value of 40mm, results in a lower boundary crossing, the valid normal value of the corresponding dimension of the basis vector individual providing the position reference in the preceding evolution is retrieved. Arithmetic subtraction is used to calculate the bottom-level crossing interval scale as -10mm. The system's pseudo-random generator calls the rand() function to generate a random decimal within a continuous uniform distribution [0,1]. This random decimal is multiplied by the bottom-level crossing interval scale to obtain a repair increment of -4.5mm. This increment is applied to the reference coordinates to generate an out-of-bounds adjustment variable of 45.5mm, thereby pulling the out-of-bounds parameter back into the valid domain and ensuring the search diversity of the bottom layer.

[0040] When a candidate variable, such as the depth dimension of the switch cabinet, suddenly increases to 1600mm, exceeding the set upper limit warning value of 1500mm for physical parameters, the upper bound overflow adjustment mechanism is triggered. The difference between the upper limit warning value of 1500mm and the corresponding variable of the baseline individual is calculated, establishing the upper bound crossing interval scale as 100mm. At this point, a uniform random number generator is invoked to extract another independent random number within the interval zero to one, and a random displacement value of 60mm is obtained through floating-point multiplication. This calculated random increment is then fed back and accumulated into the base vector coordinate system of 1400mm, resulting in a new compliant dimension parameter of 1460mm after the over-boundary adjustment. This, through interval random interpolation projection, completely eliminates invalid over-boundary samples exceeding the allowable range of physical geometry. The optimization effect distribution of the over-boundary projection adjustment mechanism is as follows: Figure 2 As shown, after projection adjustment, all invalid samples of out-of-bounds mutations that were originally below the 40mm lower limit warning value were adjusted to compliant parameters.

[0041] In one embodiment, the step of adjusting the associated variables based on the assembly association matrix to meet the assembly requirements includes: Extract the component primitives that are fixedly attached to each other, establish an assembly association matrix with a dimension equal to the number of components, mark the corresponding array units of components with physical contact and overlapping hole mating as one, and mark the units with no matching relationship as zero. Monitor the change in the boundary position of the contact surface or mounting hole of the variable drive component with parameter fine-tuning during the dimensional adjustment process; Read the follower part number column marked as 1 in the assembly association matrix line by line, and implant the same position compensation amount as the boundary position change amount into the corresponding passive mating part to complete the linkage adjustment between the group hole flush and the locking plate.

[0042] The B-Rep topological geometry information within the switchgear 3D software is analyzed. The total number N of all independent components in the structure tree that are fixedly attached to each other or have mating surfaces is scanned, and a 150×150 Boolean assembly association matrix is ​​initialized based on this. The assembly constraint attributes of the 3D model are read through the underlying interface. It iterates between any two components to check for physical contact features such as face overlap or concentric alignment of shafts and holes. If the assembly association feature is confirmed to be effective, the value 1 is written to the array cell in the i-th row and j-th column of the matrix. For those components that are suspended or completely independent without matching relationships, the corresponding cross array coordinate points are set to 0.

[0043] When optimization iteration causes a sudden change in the length of the underlying load-bearing main beam component, which acts as the driving force for variation, the central reference coordinate axis of the mounting contact surface or bolt mounting through hole connected to the main beam will undergo a corresponding translational displacement. The corresponding driving row of the assembly association matrix is ​​checked, and the association label column marked as "1" within that matrix row is extracted column by column to locate the follower component. The +5mm three-dimensional geometric position compensation vector is translated and transferred as is, and modified and rewritten into the origin absolute coordinate attribute of the passive mating component. This drives the corresponding corner piece and bolt to simultaneously follow the same dimensional compensation translation within the parametric model, thereby ensuring synchronous displacement of all assembly matrix response nodes. This achieves flush alignment of multiple holes and seamless linkage of various locking pad components during length and width variation adjustments in complex switchgear, avoiding interference and disconnection defects.

[0044] In one embodiment, comparing the constraint violation and objective function value between the test individual and the parent individual, and retaining the optimal individual, includes: The violation margin values ​​of the test individual and the parent individual exceeding the warning line in each constraint are calculated separately. After normalizing each violation margin value, they are combined and accumulated to obtain the total violation degree of the test individual and the total violation degree of the parent individual. If the total violation of the experimental individual is greater than zero and the total violation of the parent individual is zero, then the experimental individual is eliminated and the parent individual is retained and directly transitioned to the next generation of the population. If the total violation of the experimental individual is zero and the total violation of the parent individual is greater than zero, then the parent individual is eliminated and the experimental individual is retained and directly transitioned to the next generation of the population. If the total deviation of both the experimental individual and the parent individual is greater than zero, then the total deviation of the two individuals is compared, and the individual with the smallest total deviation enters the next generation of the population. If the total deviation of both the experimental individual and the parent individual is zero, then the objective function values ​​of the two individuals are compared, and the individual with the smallest objective function value enters the next generation of the population.

[0045] The penalty function method was used to statistically analyze the degree of violation of four engineering constraints, including skeleton stress, busbar temperature rise, air insulation clearance, and spatial assembly interference, by the experimental individuals and the original parent individuals produced by differential evolution. By extracting the non-negative violation margin values ​​outside the physical safety warning lines, and using the preset tolerance range or the maximum span parameter of each constraint theory for division, the violation margins with different dimensions were uniformly mapped to the interval [0,1] to achieve normalization. Then, the linear accumulation was performed item by item to obtain the real numerical index of quantified infeasibility: the total violation degree of the experimental individuals and the total violation degree of the parent individuals.

[0046] The optimal replacement mechanism is implemented according to the feasibility criterion priority rule: if the violation degree of the experimental individual is detected to be greater than 0 and the violation degree of the parent generation is equal to 0, the infeasible experimental solution is discarded, and the original parent generation is directly retained in place in the next generation population; conversely, if the experimental individual successfully enters the zero-constraint region and the total violation degree of the parent generation is greater than 0 and has physical defects, the parent generation individual is discarded, and the experimental individual with safe and reliable characteristics is accepted to enter the next evolutionary cycle. When the total violation degree of both compared solutions is greater than zero, indicating that both are outside the feasible region, the absolute values ​​of the two violation degrees are compared, and the individual closer to the feasible region and with a smaller real value of violation is selected for offspring replacement. When the safety assessment of both the experimental solution and the parent solution has no violation records, the ultimate optimization comparison mode is initiated, the objective function evaluation results of each recorded solution are extracted for competition, and the individual with smaller total material consumption and better lightweight effect is selected as the new winning solution to cover the original node, thereby completing the update iteration of individuals in each generation of the population from inferior to superior and from infeasible to feasible.

[0047] This comparative experiment was conducted based on a parametric model of a 12kV high-voltage switchgear of the same specifications. The experimental platform employed a high-performance graphics workstation coupled with finite element analysis software and 3D computational auxiliary software. The population size of the differential evolution algorithm was set to 50, with a maximum of 200 iterations. Two sets of comparative models were set up: the first set used a standard differential evolution algorithm model employing a conventional differential mutation mechanism and a basic penalty function; the second set incorporated the complete algorithm model of this invention, integrating an angle-constrained mutation strategy, an out-of-bounds projection adjustment mechanism, and an assembly-related linkage adjustment mechanism. Both sets of experiments aimed to minimize the total material mass of the switchgear frame and sheet metal parts as the single optimization objective, while simultaneously being constrained by four engineering constraints: strength, temperature rise, insulation, and assembly interference. The convergence curves of the total material mass of the two models are shown below. Figure 3 As shown, the convergence speed and weight reduction effect of the complete algorithm model are both better than the basic standard model, verifying the lightweight optimization efficiency of the method of the present invention.

[0048] The basic standard model only reached convergence in generation 165, outputting a total switchgear material mass of 520 kg, and the intermediate solutions generated during the evolutionary optimization process contained 12% of assembly interference violations. In contrast, the complete algorithm model using the scheme of this invention rapidly achieved global stable convergence in generation 98, reducing the total material mass of the optimized switchgear to 475 kg. Furthermore, throughout the entire iteration process, the assembly interference violation rate of the candidate solutions output by the complete scheme model was reduced to 0, while the steady-state temperature of the three-phase busbar and the maximum nodal stress were strictly maintained within the set safety thresholds of 105 degrees Celsius and 180 MPa, achieving full feasibility approval across all dimensions.

[0049] The complete solution utilizes reference vectors and employs angular decomposition to eliminate evolutionary regression components, generating angle-constrained mutation vectors with clear convergence guidance. This improves population optimization efficiency, increasing convergence speed by 40.6% compared to the basic model. The coordinated adjustment of the assembly correlation matrix and the out-of-bounds projection mechanism completely avoids potential size out-of-bounds and group hole disconnection defects during mutation, eliminating physical collision and interference problems at the source. Overall, this technical solution reduces the overall material consumption of the switchgear by 8.7% while ensuring absolute safety in electrical performance and structural strength, achieving a dual leap in lightweight structural construction and algorithm feasibility development capabilities under multi-physics field constraints. Example

[0050] Embodiment 2 of the present invention provides an optimized design system for switchgear structure, comprising: The construction module is used to build a parameterized structural model of the switchgear, extract dimensions and structural parameters as variables, establish a function with minimizing material usage as the objective function, establish a constraint function with strength, temperature rise, insulation distance and assembly interference as constraints, encode the variables to generate an initial population, deploy a feasibility judgment loop in each generation evolution cycle, and store individuals in the current population whose total degree of violation of any constraint is zero into the feasible solution set. The production module is used to select a basis vector individual and two difference individuals for the individual to be mutated to calculate the original difference vector, and to calculate the angle between the original difference vector and the reference vector. The reference vector points from the current individual to the guiding point calculated based on the feasible solution set. When the angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector, the parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate a mutation vector constrained by the angle, and the mutation vector is superimposed with the basis vector individual to obtain the candidate solution. The output module is used to perform out-of-bounds projection adjustment on the candidate solutions, and to perform linkage adjustment on the associated variables based on the assembly association matrix to meet the assembly requirements. The size parameters in the adjusted candidate solutions are rounded according to the manufacturing specifications to obtain the test individuals. The constraint violation degree and objective function value of the test individuals are compared with those of the parent individuals. The optimal individual is retained and iterated until the termination condition is met, and the switch cabinet structure optimization design results are output.

[0051] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0052] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An optimized design method for switchgear structure, characterized in that, include: A parameterized structural model of the switchgear is constructed, and the dimensions and structural parameters are extracted as variables. A function with the minimum material usage as the objective function is established, and a constraint function with strength, temperature rise, insulation distance and assembly interference as constraints is established. The variables are encoded to generate an initial population. A feasibility judgment loop is deployed in each generation evolution cycle. Individuals in the current population whose total degree of violation of any constraint is zero are stored in the feasible solution set. For each mutated individual, a basis vector individual and two difference individuals are selected to calculate the original difference vector. The angle between the original difference vector and the reference vector is calculated. The reference vector points from the current individual to the guiding point calculated based on the feasible solution set. When the angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector. The parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate a mutated vector constrained by the angle. The mutated vector is superimposed with the basis vector individuals to obtain the candidate solution. The reference vector points from the current individual to a guiding point calculated based on the feasible solution set, including: Within each generation of evolution, a feasibility determination loop is deployed, and individuals in the current population whose total degree of violation of any constraint is zero are stored in the set of feasible solutions. Read the number of feasible solutions contained in the feasible solution set. When the number of feasible solutions is greater than zero, sum the parameter values ​​of the feasible solutions in each dimension and divide by the number of feasible solutions to calculate the statistical average displacement in each independent dimension. When the number of feasible solutions is zero, the parameter value of the individual with the smallest constraint violation in the current population is used as the statistical average displacement in each independent dimension. The statistical average displacements of all dimensions are sequentially cascaded and integrated to form a unique guiding point in the multidimensional variable space; The displacement variable values ​​of the guiding points are extracted one by one and the corresponding dimension displacement variable values ​​of the current individual to be mutated are subtracted to generate a multidimensional difference vector, which is used as a reference vector to drive evolution. The candidate solutions are subjected to out-of-bounds projection adjustment, and the associated variables are adjusted in a linkage manner based on the assembly association matrix to meet the assembly requirements. The dimensional parameters in the adjusted candidate solutions are rounded to the manufacturing specifications to obtain the test individuals. The constraint violation degree and objective function value of the test individuals are compared with those of the parent individuals. The optimal individual is retained and iterated until the termination condition is met, and the switch cabinet structure optimization design results are output. The step of adjusting the candidate solution by out-of-bounds projection includes: When it is detected that the variable value of any dimension in the candidate solution is less than the lower limit warning value, the bottom-level cross-interval scale is obtained by subtracting the variable value of the corresponding dimension of the basis vector individual from the lower limit warning value. A random decimal number between zero and one is generated and multiplied with the bottom-level cross-interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the out-of-bounds adjustment. When it is detected that the variable value of any dimension in the candidate solution is greater than the upper limit warning value, the upper limit warning value is subtracted from the variable value of the corresponding dimension of the basis vector individual to obtain the upper limit crossing interval scale. A random decimal number between zero and one is generated and multiplied with the upper limit crossing interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the over-limit adjustment. The step of adjusting the associated variables based on the assembly association matrix to meet assembly requirements includes: Extract the component primitives that are fixedly attached to each other, establish an assembly association matrix with a dimension equal to the number of components, mark the corresponding array units of components with physical contact and overlapping hole mating as one, and mark the units with no matching relationship as zero. Monitor the change in the boundary position of the contact surface or mounting hole of the variable drive component with parameter fine-tuning during the dimensional adjustment process; Read the follower part number column marked as 1 in the assembly association matrix line by line, and implant the same position compensation amount as the boundary position change amount into the corresponding passive mating part to complete the linkage adjustment between the group hole flush and the locking plate.

2. The method according to claim 1, characterized in that, The establishment of the constraint function, with minimizing material usage as the objective function and strength, temperature rise, insulation distance, and assembly interference as constraints, includes: The objective function is to be minimized by the sum of the products of the volumes of all sheet metal parts and load-bearing frames and the corresponding physical material densities in the parameterized structural model of the switch cabinet. Extract the maximum nodal stress value generated by the skeleton under rated electrodynamic impact simulation, and set the maximum nodal stress value being greater than a preset stress threshold as a strength constraint violation; Extract the temperature peak at the connection of each phase bus inside the cabinet, and set the temperature peak exceeding the preset temperature threshold as a violation of the temperature rise constraint; A three-dimensional distance matrix is ​​established between the control points of the conductive busbar and the grounding box shell. The minimum value of the term in the three-dimensional distance matrix that is less than the preset insulation distance threshold is set as an insulation distance constraint violation. The total interference value is formed by accumulating the overlapping three-dimensional volumes of mutually intersecting components within the parameterized structural model. A total interference value greater than zero is defined as a violation of assembly interference constraints.

3. The method according to claim 1, characterized in that, When the included angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector, the parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate an angle-constrained variant vector, and the variant vector is superimposed with the individual basis vectors to obtain a candidate solution, including: Calculate the vector difference between the first and second difference individuals to produce a multi-dimensional original difference vector; Calculate the dot product of the original difference vector and the reference vector, calculate the product of the norm of the original difference vector and the norm of the reference vector, if the product is greater than zero, calculate the inverse cosine function value of the ratio of the two as the included angle, if the product is zero, set the included angle to zero; When the included angle is greater than the preset limit of ninety degrees, the original difference vector is decomposed into parallel and perpendicular components along the reference vector; The parallel component, which is opposite to the direction of the reference vector, is reversed to the component with the same direction as the reference vector to eliminate the back-inference factor, and then resynthesized with the vertical component to establish a completely new evolutionary mutation quantity. The evolutionary mutation amount is superimposed with the corresponding parameters of the basis vector individuals to generate an angle-constrained mutation vector, thereby obtaining the candidate solution.

4. The method according to claim 1, characterized in that, The comparison of constraint violation and objective function value between the experimental individual and the parent individual, and the retention of the optimal individual, includes: The violation margin values ​​of the test individual and the parent individual exceeding the warning line in each constraint are calculated separately. After normalizing each violation margin value, they are combined and accumulated to obtain the total violation degree of the test individual and the total violation degree of the parent individual. If the total violation of the experimental individual is greater than zero and the total violation of the parent individual is zero, then the experimental individual is eliminated and the parent individual is retained and directly transitioned to the next generation of the population. If the total violation of the experimental individual is zero and the total violation of the parent individual is greater than zero, then the parent individual is eliminated and the experimental individual is retained and directly transitioned to the next generation of the population. If the total deviation of both the experimental individual and the parent individual is greater than zero, then the total deviation of the two individuals is compared, and the individual with the smallest total deviation enters the next generation of the population. If the total deviation of both the experimental individual and the parent individual is zero, then the objective function values ​​of the two individuals are compared, and the individual with the smallest objective function value enters the next generation of the population.

5. An optimized design system for switchgear structure, characterized in that, include: The construction module is used to build a parameterized structural model of the switchgear, extract dimensions and structural parameters as variables, establish a function with minimizing material usage as the objective function, establish a constraint function with strength, temperature rise, insulation distance and assembly interference as constraints, encode the variables to generate an initial population, deploy a feasibility judgment loop in each generation evolution cycle, and store individuals in the current population whose total degree of violation of any constraint is zero into the feasible solution set. The production module is used to select a basis vector individual and two difference individuals for the individual to be mutated to calculate the original difference vector, and to calculate the angle between the original difference vector and the reference vector. The reference vector points from the current individual to the guiding point calculated based on the feasible solution set. When the angle is greater than a preset limit value, the original difference vector is decomposed along the reference vector, the parallel components with opposite directions are reversed and synthesized with the perpendicular components to generate a mutation vector constrained by the angle, and the mutation vector is superimposed with the basis vector individual to obtain the candidate solution. The reference vector points from the current individual to a guiding point calculated based on the feasible solution set, including: Within each generation of evolution, a feasibility determination loop is deployed, and individuals in the current population whose total degree of violation of any constraint is zero are stored in the set of feasible solutions. Read the number of feasible solutions contained in the feasible solution set. When the number of feasible solutions is greater than zero, sum the parameter values ​​of the feasible solutions in each dimension and divide by the number of feasible solutions to calculate the statistical average displacement in each independent dimension. When the number of feasible solutions is zero, the parameter value of the individual with the smallest constraint violation in the current population is used as the statistical average displacement in each independent dimension. The statistical average displacements of all dimensions are sequentially cascaded and integrated to form a unique guiding point in the multidimensional variable space; The displacement variable values ​​of the guiding points are extracted one by one and the corresponding dimension displacement variable values ​​of the current individual to be mutated are subtracted to generate a multidimensional difference vector, which is used as a reference vector to drive evolution. The output module is used to perform out-of-bounds projection adjustment on the candidate solutions, and to perform linkage adjustment on the correlation variables based on the assembly correlation matrix to meet the assembly requirements. The size parameters in the adjusted candidate solutions are rounded according to the manufacturing specifications to obtain the test individuals. The constraint violation degree and objective function value of the test individuals are compared with those of the parent individuals. The optimal individual is retained and iterated until the termination condition is met, and the switch cabinet structure optimization design results are output. The step of adjusting the candidate solution by out-of-bounds projection includes: When it is detected that the variable value of any dimension in the candidate solution is less than the lower limit warning value, the bottom-level cross-interval scale is obtained by subtracting the variable value of the corresponding dimension of the basis vector individual from the lower limit warning value. A random decimal number between zero and one is generated and multiplied with the bottom-level cross-interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the out-of-bounds adjustment. When it is detected that the variable value of any dimension in the candidate solution is greater than the upper limit warning value, the upper limit warning value is subtracted from the variable value of the corresponding dimension of the basis vector individual to obtain the upper limit crossing interval scale. A random decimal number between zero and one is generated and multiplied with the upper limit crossing interval scale. The product result is superimposed on the variable value of the corresponding dimension of the basis vector individual as the variable after the over-limit adjustment. The step of adjusting the associated variables based on the assembly association matrix to meet assembly requirements includes: Extract the component primitives that are fixedly attached to each other, establish an assembly association matrix with a dimension equal to the number of components, mark the corresponding array units of components with physical contact and overlapping hole mating as one, and mark the units with no matching relationship as zero. Monitor the change in the boundary position of the contact surface or mounting hole of the variable drive component with parameter fine-tuning during the dimensional adjustment process; Read the follower part number column marked as 1 in the assembly association matrix line by line, and implant the same position compensation amount as the boundary position change amount into the corresponding passive mating part to complete the linkage adjustment between the group hole flush and the locking plate.

6. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program, when executed by a processor, implements the method as described in any one of claims 1-4.

7. A computer program product, said computer program product comprising computer code or computer program, characterized in that, When computer code or a computer program is run on a computer, the operations in the method provided by any one of claims 1-4 are performed.

Images