A resistivity cave identification method based on squirrel search algorithm
The objective function is constructed through the squirrel search algorithm, and the potential distribution data is optimized, which solves the initial value and noise sensitivity problems of traditional resistivity cave detection methods, and realizes accurate identification and safety guarantee of underground caves in limestone areas.
Patent Information
- Application Number
- CN201911082591.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2019-11-07
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2039-11-07
AI Technical Summary
The traditional resistivity cave detection method is sensitive to initial values, noise-sensitive and easily trapped in local optimal solutions, resulting in poor identification results and the inability to accurately detect underground caves in limestone areas, posing hidden dangers to the safety of buildings.
The resistivity cave recognition method based on the squirrel search algorithm is used to construct the objective function through the potential distribution data, and the objective function is optimized by the squirrel search algorithm to identify the cave location without the initial value and gradient information, and is not sensitive to noise.
Accurate detection of underground caves is achieved, local optimal solutions are avoided, detection efficiency and accuracy are improved, and building safety is ensured.
Smart Images

Figure CN110850477B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a resistivity cave identification method based on a squirrel search algorithm, and belongs to the technical field of underground cave detection. Background Art
[0002] At present, in areas where underground limestone caves are developed, underground caves in construction sites are often located below the groundwater level. In order to ensure the safety of the building foundation, all parties involved in the construction have invested a lot of manpower, financial resources and time to conduct exploration of underground caves in construction sites in limestone areas. The traditional engineering geological drilling is only a partial view, or the traditional electrical exploration is used. Due to the large depth of the underground caves or the small scale of the underground caves, the resistivity of the underground caves is small and the resistivity of the surrounding limestone is not accurately detected, which leaves serious safety hazards to the buildings. Underground caves will cause certain harm to the construction of houses, subways, etc. Timely detection and treatment of underground caves are of great significance to the safety of structural construction and use.
[0003] The problem of detecting caves by resistivity method is an optimization problem in the identification research of inverse problems. Its basic idea is to determine the detection area and electrode scheme. The appearance of caves will cause changes in the measured potential distribution data, which will then be reflected in the changes in the apparent resistivity of the soil layer. These changes can be used to locate the location of the cave. That is, by defining an objective function about the cave structure, and then using various optimization methods to detect the location of the cave. The traditional resistivity cave detection method introduces a regularization method and uses sensitivity or gradient matrix to identify the location of the cave. It has the following disadvantages: it is sensitive to initial values, sensitive to noise, and easy to fall into the local optimal solution, resulting in poor recognition effect. Summary of the invention
[0004] In order to overcome the defects of the above-mentioned technology, the present invention provides a resistivity cave identification method based on the squirrel search algorithm which is practical, effective and accurate. The method adopts the potential distribution data difference to construct the objective function, and optimizes the objective function using the squirrel search algorithm. The location of various caves can be accurately detected without the need for initial value and gradient information, and the method is insensitive to noise and has good engineering application capabilities.
[0005] To achieve the above object, the technical solution of the present invention is as follows: a resistivity cave identification method based on the squirrel search algorithm, comprising the following steps:
[0006] Step 1: Establish a finite element model of the cave soil layer, determine the electrode scheme, and use the electrical impedance method to obtain the potential distribution data of the measurement point;
[0007] Step 2: Construct the objective function of the cave structure, that is, the optimization objective function, which is as follows:
[0008]
[0009] Among them, g(c) is the objective function, ||(.)|| represents the 2-norm of the vector, that is, c is the apparent resistivity of each unit, Su represents the data on the set of measured potential points, is the potential distribution data measured by the i-th group of electrode schemes, is the potential distribution data identified by the i-th group of electrode schemes, is a matrix consisting of s groups of measurement data, and R(c) is a matrix consisting of s groups of identification data;
[0010] Step 3: Use the squirrel search algorithm to continuously optimize the objective function. After the stop criteria are met, the identification position of the cave can be finally obtained;
[0011] The specific process of optimizing the objective function using the squirrel search algorithm in step 3 of the above method is as follows:
[0012] S1, random initialization: Assume there are n squirrels (FS), the position of the i-th squirrel can be represented by a vector, and the positions of all squirrels can be represented by the following matrix:
[0013]
[0014] Among them, FS i.j represents the jth dimension of the i-th squirrel, and uses uniform distribution to assign the initial position of each squirrel in the forest:
[0015] FS i =FS L +U(0,1)×(FS U -FS L ) (3)
[0016] Among them, FS L and FS U are the lower and upper bounds of the ith squirrel in the jth dimension, respectively, and U(0,1) is a uniformly distributed random number in the range [0,1];
[0017] S2, fitness evaluation: put the value of the decision variable (solution vector) into the user-defined fitness function, calculate the position fitness of each squirrel, and store the corresponding value in the following array:
[0018]
[0019] S3, Sorting, Declaration, and Random Selection: After storing the fitness value of each squirrel’s position, the array is sorted in ascending order;
[0020] S4, generate new position: the mathematical model of the new position has the following three cases:
[0021]
[0022]
[0023]
[0024] where d g is the random glide distance, R1, R2 and R3 are random numbers in the range of [0, 1], FS nt , FS at and FS ht are the positions of squirrels arriving at the common tree, acorn tree, and hickory tree respectively, t is the current iteration number, G c The value of P is 1.9, dp The value of is 0.1;
[0025] S5, seasonal monitoring conditions: seasonal monitoring conditions are introduced into the algorithm to prevent the algorithm from falling into the local optimal solution. Its modeling behavior involves the following steps:
[0026] (i) First, use equation (13) to calculate the seasonal constant S c
[0027]
[0028] Where t = 1, 2, 3;
[0029] (ii) Check seasonal monitoring, i.e. Where S min is the minimum value of the seasonal constant, and the calculation formula is:
[0030]
[0031] where t and t m They are the current iteration value and the maximum iteration value respectively;
[0032] (iii) if the seasonal monitoring condition is true, the squirrels are randomly relocated;
[0033] S6, random migration at the end of winter; the migration of squirrels is expressed by the following equation:
[0034]
[0035] Among them, the mathematical method of Lévy distribution is as follows:
[0036]
[0037] Where μ, γ>0. γ is the scale parameter, μ is the transfer parameter, and the Lévy flight mode is calculated by the following formula:
[0038] Lévy(x)=0.01×r a ×σ / |r b | 1 / β (16)
[0039] where r a and r b are two normally distributed random numbers in [0, 1]. In this work,
[0040] The β constant is 1.5, and σ is calculated as:
[0041]
[0042] Among them, Γ(x)=(x-1)!;
[0043] S7, stop criterion: define an allowable small threshold between the last two consecutive results, the maximum number of iterations is considered as the stop criterion, save the current optimal solution, determine whether the stop criterion is met, and output the optimal solution if it is met, otherwise, repeat steps S2-S6 until the algorithm ends. Compared with the prior art, the beneficial effects of the technical solution of the present invention are: the present invention constructs the objective function through the potential distribution data of the cave structure measurement point, uses the meta-heuristic algorithm to identify the cave location, is insensitive to the initial value, is insensitive to noise, and is not easy to fall into the local optimal solution, and has better efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 Flowchart for reducing cave identification problem to optimization problem;
[0045] Figure 2 Schematic diagram of the approximate model of the sliding behavior of the squirrel search algorithm;
[0046] Figure 3 This is a schematic diagram of the implementation process of the squirrel search algorithm;
[0047] Figure 4 is a finite element model of a single cave in Example 1 of the present invention;
[0048] Figure 5 It is a finite element model of multiple caves in Example 2 of the present invention;
[0049] Figure 6 The detection result of the method of the present invention in Example 1 of the present invention without noise;
[0050] Figure 7 The detection result of the method of the present invention in Example 1 of the present invention under the condition of 0.1% noise;
[0051] Figure 8The detection result of the particle swarm algorithm in the absence of noise in Example 1 of the present invention;
[0052] Fig. 9 The detection result of the particle swarm algorithm in Example 1 of the present invention with 0.1% noise;
[0053] Fig.10 This is a comparison diagram of algorithm convergence curves for a single cave in Example 1 of the present invention with 0.1% noise;
[0054] Fig.11 The detection result of the method of the present invention in Example 2 of the present invention without noise;
[0055] Fig.12 The detection result of the method of the present invention in Example 2 of the present invention under the condition of 0.1% noise;
[0056] Fig.13 The detection result of the particle swarm algorithm in the absence of noise in Example 2 of the present invention;
[0057] Fig.14 The detection result of the particle swarm algorithm in Example 2 of the present invention with 0.1% noise;
[0058] Fig.15 This is a comparison diagram of algorithm convergence curves of multiple caves in Example 2 of the present invention with 0.1% noise. DETAILED DESCRIPTION
[0059] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments.
[0060] See also Figure 1 and Figure 2 The resistivity cave identification method based on the squirrel search algorithm of the present invention is divided into two steps:
[0061] (1) Objective function
[0062] The cave model is simplified to a rectangular detection area. The current points and potential points are not consistent, but the number is equal. Using finite element to solve the direct problem, we have the following equation
[0063]
[0064] Among them, u=[u1;u2;…;u n ] includes all the measurement node potentials, f = [f1; f2; ...; f n ] represents the current vector. Different electrode schemes have different f. K is the stiffness matrix, which is related to the apparent resistivity of each unit c = [c1; c2; ...; cm ] linearly related, K j The element stiffness matrix when is unit resistivity.
[0065] When there is no cave in the soil layer, the apparent resistivity of all units is c0=10. When a cave appears, the apparent resistivity of the unit will decrease accordingly. In the present invention, the apparent resistivity of the unit with a cave is 3, that is, c0=10. j =3 (j=1, 2, 3, ..., m) means that a cave appears in the jth unit. According to the above finite element formula, there is a functional relationship between c and u. We can identify c and deduce the calculated potential distribution data.
[0066] The objective function constructed based on the potential distribution data is as follows:
[0067]
[0068] Where c is the apparent resistivity of each unit, Su represents the data on the set of measured potential points, is the potential distribution data measured by the i-th group of electrode schemes, is the potential distribution data identified by the i-th group of electrode schemes, is a matrix composed of s groups of measured data, and R(c) is a matrix composed of s groups of calculated data. When the measured potential distribution data is completely consistent with the identified potential distribution data, the value is minimum. The cave identification problem is equivalent to an optimization problem. The position of a certain individual squirrel is a distribution of apparent resistivity. When the objective function reaches the minimum value, the identified apparent resistivity c, that is, the optimal solution, can reflect the location of the cave.
[0069] (2) Use the squirrel search algorithm to optimize the objective function and obtain the recognition result.
[0070] See also Figure 3 , the squirrel search algorithm can be divided into the following 7 stages:
[0071] Phase 1: Random Initialization
[0072] The parameter settings include the initial population size n of the algorithm, the maximum number of iterations t, and the upper and lower bounds of all dimensions of the squirrels. The position of the i-th squirrel can be represented by a vector. The positions of all squirrels can be represented by the following matrix:
[0073]
[0074] Among them FS i.j represents the jth dimension of the ith squirrel. The initial position of each squirrel in the forest is assigned using uniform distribution (Formula (2)).
[0075] FS i =FSL +U(0,1)×(FS U -FS L ) (4)
[0076] Among them, FS L and FS U are the lower and upper bounds of the ith squirrel in the jth dimension, respectively. U(0, 1) is a uniformly distributed random number in the range [0, 1].
[0077] Phase 2: Fitness evaluation,
[0078] Put the values of the decision variables (solution vectors) into the user-defined fitness function, calculate the position fitness of each squirrel, and store the corresponding values in the following array:
[0079]
[0080] The fitness value of each squirrel's position describes the quality of the food source it searches for, i.e., optimal food source (pecan tree), normal food source (acorn tree), and no food source (squirrel in a normal tree), and therefore also describes their survival probability.
[0081] Phase 3: Sorting, assertion, and random selection
[0082] After storing the fitness value of each squirrel’s location, the array is sorted in ascending order. The squirrel with the minimum fitness value is considered to be in the pecan tree. The next three best squirrels are considered to be in the acorn tree and they are considered to be moving towards the pecan tree. The remaining squirrels are supposed to be in the normal tree. With further random selection, some squirrels are considered to be heading towards the pecan tree, assuming that they have already met their daily energy requirements. The remaining squirrels will go to the acorn nut tree (to meet their daily energy requirements). This foraging behavior of the squirrels is always affected by predators. Using the predator presence probability (P dp )’s position update mechanism models this natural behavior.
[0083] Phase 4: Generate new location
[0084] As mentioned earlier, there are three scenarios that may occur during a squirrel's dynamic foraging. In each scenario, it is assumed that in the absence of predators, the squirrel will glide and efficiently search for its favorite food throughout the forest, while the presence of a predator makes it cautious and forced to use small random walks to search nearby hiding places. The mathematical model of dynamic foraging behavior is as follows:
[0085] 1) A squirrel in an acorn nut tree might move to a walnut tree. In this case, the squirrel's new position can be found as follows:
[0086]
[0087] where d g is the random gliding distance, R1 is a random number in the range of [0, 1], FS ht is the position of the squirrel that has reached the pecan tree, and t is the current iteration number. In the mathematical model, the sliding constant G is used c Achieves a balance between exploration and development. Its value has a significant impact on the performance of the proposed algorithm. In the current work, G c The value of is considered to be 1.9, obtained after rigorous analysis.
[0088] 2) Squirrel FS on a normal tree nt They may move towards the acorn nut trees to meet their daily energy needs. In this case, the new positions of the squirrels can be obtained as follows:
[0089]
[0090] Wherein, R2 is a random number in the range of [0, 1].
[0091] 3) Some squirrels that have eaten acorn nuts in normal trees may fly to walnut trees to store walnuts that can be eaten when food is scarce. In this case, the new positions of the squirrels can be obtained as follows:
[0092]
[0093] where R3 is a random number in the range [0, 1]. In the present work, the probability of the presence of a predator P dp In all cases it was considered to be 0.1.
[0094] 4) The gliding mechanism of the squirrel is described by balanced gliding, in which the sum of lift (L) and drag (D) produces a force (R) equal to and opposite to the squirrel's weight (Mg). In the design of the optimization algorithm, the present invention adopts an approximate model of sliding behavior ( Figure 1 ). The squirrel glides at a steady speed and always descends at an angle to the horizontal and the lift-to-drag ratio or slip ratio is defined as follows:
[0095]
[0096] Squirrels can increase the length of their glide landings by using a small glide angle (φ), which increases lift proportionally to drag. Here, lift is the force generated by the downward deflection of air as it passes over the wing and is defined as:
[0097]
[0098] Where, ρ(=1.204kgm -3) is the air density, C L is the lift coefficient, V(=5.25ms -1 ) is the speed, S(=154cm 2 ) is the surface area of the limb, and the frictional resistance is defined as:
[0099]
[0100] Among them C D is the friction coefficient. At low speeds, this coefficient is very high, while at high speeds it becomes smaller. The steady-state glide angle is:
[0101]
[0102] Approximate glide distance d g Calculate as follows:
[0103]
[0104] where h g (=8m) is the height loss after gliding. Calculate d g All required parameter values, including C L and C D , are all considered from actual data. Therefore, the squirrel can change the glide path length or d simply by changing the lift-to-drag ratio according to the landing position. g . C was added during the simulation L At 0.675 ≤ C L Random changes within the range of ≤1.5, it is considered that C D Fixed at 0.60.
[0105] The horizontal gliding distance of a squirrel is usually 5 to 25 meters. The gliding distance of the model is 9 to 20 meters. g When the value is large, it may introduce large disturbances when generating new positions, resulting in unsatisfactory algorithm performance. g The value of is scaled down to achieve acceptable performance of the algorithm. g divided by a suitable non-zero value, called the scaling factor (sf), which is obtained through rigorous experimentation with benchmark functions. In the present work, sf = 18 provides sufficient d g Perturbation range, which helps to achieve the desired balance between exploration and exploitation phases.
[0106] Phase 5: Seasonal Monitoring Conditions
[0107] Seasonal changes significantly affect the foraging activities of squirrels. They suffer a lot of heat loss in low temperatures due to their high body temperature and small size, which makes foraging costly and risky due to the presence of active predators. Climatic conditions force them to be less active in winter compared to autumn. Therefore, the movement of squirrels is affected by weather changes and including this behavior may provide a more realistic approach to optimization. Therefore, a season monitoring condition is introduced in the algorithm to avoid the algorithm from getting stuck in a local optimum. Modeling the behavior involves the following steps:
[0108] (i) First calculate the seasonal constant S c
[0109]
[0110] Where t=1, 2, 3.
[0111] (ii) Check seasonal monitoring, i.e. Where S min is the minimum value of the seasonal constant, and the calculation formula is:
[0112]
[0113] where t and t m are the current iteration value and the maximum iteration value respectively. min This affects the exploration and development capabilities of this method. min The larger the value, the higher the algorithm's exploration ability. min The smaller the size, the more powerful the algorithm is at exploiting its capabilities. For any effective metaheuristic, there must be a proper balance between these two stages. c To maintain this balance, but in the iterative process by adaptively changing S min The value can be improved.
[0114] (iii) If the seasonal monitoring condition is true, squirrels that are unable to find the best winter food source in the forest will be randomly relocated.
[0115] Phase 6: Random relocation at the end of winter
[0116] As mentioned before, the end of winter makes squirrels active due to the lower foraging costs. Squirrels that survived the winter by not being able to find the best food source in the forest may forage in new directions. Incorporating this behavior into the model improves the exploration capabilities of the algorithm. Presumably, only squirrels that survived by not being able to find the pecan food source will move in a different direction to find a better food source. This squirrel migration is illustrated by the following equation:
[0117]
[0118] Lévy distribution encourages better and more efficient exploration of the search space. Lévy flight mode is a powerful mathematical tool used by researchers to improve the global search capabilities of various metaheuristic algorithms. Lévy flight mode helps to find new candidate solutions, far away from the current best solution. It is a random walk where the step size is given by the Lévy distribution. This distribution is often expressed as a power law formula L(s)~|s| -1-β , where 0<β<2 is an indicator. The mathematical method of Lévy distribution is as follows:
[0119]
[0120] Where μ, γ>0. γ is the scale parameter and μ is the transfer parameter. The Lévy flight mode is calculated by the following formula:
[0121] Lévy(x)=0.01×r a ×σ / |r b | 1 / β (16)
[0122] where r a and r b are two normally distributed random numbers in [0, 1]. In this work, the β constant is considered to be 1.5, and σ is calculated as:
[0123]
[0124] Among them, Γ(x)=(x-1)!.
[0125] Stage 7: Stopping Criteria
[0126] Function tolerance is a commonly used convergence criterion in which a small threshold is defined between the last two consecutive results. The maximum execution time is also sometimes used as a stopping criterion. In this study, the maximum number of iterations is considered as the stopping criterion.
[0127] Since the squirrel search algorithm is a swarm intelligence algorithm, it can be used to solve nonlinear problems. The resistivity method for detecting caves is a typical nonlinear problem. When combining the squirrel search algorithm, we first establish a finite element model for identifying caves using the resistivity method based on the detection area, calibrate the corresponding cave position with the apparent resistivity of the unit, set the input and output currents (electrode scheme), and solve the potential distribution of the detection area according to the formula of the direct problem. In actual engineering, we can only get the potential distribution, and it is a nonlinear problem to calculate the apparent resistivity distribution from the potential distribution. Therefore, we combine the squirrel search algorithm to solve the apparent resistivity distribution of the detection area, so as to know the location of the cave (the objective function is established by the potential distribution obtained from the direct problem, and the individual position of the squirrel is a kind of apparent resistivity distribution. The parameters such as the initial position and number of individuals of the squirrel are set, and then through the continuous iteration of the algorithm, after the stopping criteria are met, the best squirrel position finally obtained is the identified apparent resistivity distribution).
[0128] Example 1: Identification of the location of a single cave model
[0129] like Figure 4 As shown in the figure, a single cave finite element model is established, the rectangular detection area is divided into 64 units, and the underground cave model of 64 units is used for detection. The length and width of the rectangular detection area are both 4m*4m, and each side is evenly divided into 8 units; and the electrical impedance imaging method is used to detect underground caves. Assuming that the apparent resistivity of unit 28 is equivalent to 3, the parameter settings in the algorithm are: Maximum number of iterations Iter max , population size Np, and decision variable number n are 200, 250, and 64 respectively. The noise levels are 0% and 0.1% respectively.
[0130] from Figures 6 to 9 As shown in the figure, the squirrel search algorithm can detect the location of the cave, while the particle swarm algorithm gives many wrong identifications. The location of the cave can be accurately identified by the squirrel search algorithm, and the relative error between the actual data and the identified data is extremely small. Even under the influence of artificial noise, the relative error is within 1.3%, which is Fig.10 The convergence curve of the algorithm shows that the convergence speed of the squirrel search algorithm is significantly faster than that of the particle swarm algorithm, and the final convergence value is much smaller. The results show that this method is effective and accurate in detecting cave locations and is insensitive to noise.
[0131] Example 2: Identification of multiple cave model locations
[0132] like Figure 5 As shown, this embodiment uses the electrical impedance imaging method to detect underground caves for multi-unit cave detection. Assuming that the apparent resistivity of units 28, 36 and 37 is equivalent to 3, the parameter settings in the algorithm are: the maximum number of iterations Iter max, population size Np, and decision variable number n are 200, 250, and 64 respectively. The noise levels are 0% and 0.1% respectively.
[0133] Depend on Figures 11 to 14 It can be seen that in the absence of noise, the squirrel search algorithm can accurately identify the location of the cave. On the contrary, the particle swarm algorithm cannot detect effectively. Fig.15 The convergence curve of the algorithm shows that the convergence speed of the squirrel search algorithm is significantly faster than that of the particle swarm algorithm, and the final convergence value is much smaller. Although the number of recognition units is larger, the squirrel search algorithm is still feasible and accurate in cave detection and is insensitive to noise.
[0134] Obviously, the above embodiments of the present invention are only examples for clearly explaining the present invention. For those skilled in the art, different forms of changes can be made based on the above description, such as expanding the detection range, changing the location of the cave, etc., which are not exhaustively listed here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.
Claims
1. A resistivity cave identification method based on squirrel search algorithm, Features The following steps are involved: Step 1: Establish a finite element model of the cave soil layer, determine the electrode scheme, and use the electrical impedance method to obtain the potential distribution data of the measurement point; Step 2: Construct the objective function of the cave structure, which is as follows: Among them, g(c) is the objective function, ||(.)|| represents the 2-norm of the vector, that is, c is the apparent resistivity of each unit, Su represents the data on the set of measured potential points, is the potential distribution data measured by the x-th group of electrode schemes, is the potential distribution data identified by the x-th group of electrode schemes, x = 1, 2, ... s, is a matrix consisting of s groups of measurement data, and R(c) is a matrix consisting of s groups of identification data; Step 3: Use the squirrel search algorithm to continuously optimize the objective function. After the stop criteria are met, the identification position of the cave can be finally obtained; The specific process of optimizing the objective function using the squirrel search algorithm in step 3 is as follows: S1, random initialization: Assume that there are n squirrels FS, the position of the i-th squirrel can be determined by the vector, and the positions of all squirrels can be represented by the following matrix: Among them, FS n.d Represents the dth dimension of the nth squirrel, using uniform distribution to assign the initial position of each squirrel in the forest: FS i =FS L +U(0,1)×(FS U -FS L ) (3) Among them, FS L and FS U are the lower and upper bounds of the ith squirrel in the jth dimension, respectively. U(0,1) is a uniformly distributed random number in the range [0,1]. S2, fitness evaluation: put the values of the decision variables into the user-defined fitness function, calculate the position fitness of each squirrel, and store the corresponding values in the following array: S3, sorting, assertion, and random selection: f n represents the fitness function. After storing the fitness value of each squirrel's position, the array is sorted in ascending order; S4, generate new position: the mathematical model of the new position has the following three cases: 1) A squirrel in an acorn nut tree might move to a walnut tree, in which case the squirrel's new position is obtained as follows: 2) Squirrels in regular trees may move to acorn nut trees to meet their daily energy needs; in this case, the squirrels' new positions are obtained as follows: 3) Some squirrels that have eaten acorn nuts in normal trees may fly to walnut trees to store walnuts that can be eaten when food is scarce; in this case, the new position of the squirrels can be obtained as follows: where d g is the random glide distance, R 1 , R 2 and R 3 is a random number in the range [0,1], FS nt , FS at and FS ht are the positions of squirrels arriving at the common tree, acorn nut tree, and walnut tree respectively, t is the current iteration number, G c The value of P is 1.9, dp The value of is 0.1; S5, seasonal monitoring conditions: seasonal monitoring conditions are introduced into the algorithm to prevent the algorithm from falling into the local optimal solution. Its modeling behavior involves the following steps: (i) First, use equation (12) to calculate the seasonal constant Where t = 1, 2, 3; (ii) Check seasonal monitoring, Where S min is the minimum value of the seasonal constant, and the calculation formula is: where t and t m They are the current iteration value and the maximum iteration value respectively; (iii) if the seasonal monitoring condition is true, the squirrels are randomly relocated; S6, random migration at the end of winter; the migration of squirrels is expressed by the following equation: Among them, the mathematical method of Lévy distribution is as follows: Where μ,γ>0,γ is the scale parameter, s is the step size extracted from the Lévy distribution, μ is the transfer parameter, and the Lévy flight mode is calculated by the following formula: Lévy(x)=0.01×r a ×σ / |r b | 1 / β (16) where r a and r b are two normally distributed random numbers in [0,1], the β constant is 1.5, and σ is calculated as: Among them, Γ(x)=(x-1)!; S7, stopping criterion: define an allowable small threshold between the last two consecutive results, the maximum number of iterations is considered as the stopping criterion, save the current optimal solution, determine whether it meets the stopping criterion, and output the optimal solution if it meets it. Otherwise, repeat steps S2-S6 until the algorithm ends.
