Photovoltaic power station cleaning robot path planning method based on improved snake vulture optimization algorithm

Abstract

The invention discloses a photovoltaic power station cleaning robot path planning method based on an improved snake vulture optimization algorithm in the technical field of robots. Firstly, a fitness function considering heterogeneous topographic features is designed in combination with photovoltaic power station features, and the path length and the topographic safety are balanced by introducing a path reward and punishment mechanism; secondly, performing multi-strategy improvement on an original snake vulture optimization algorithm: introducing a double-population iteration mechanism combining elite initialization, Brownian motion and Weibull distribution, and fusing optimization strategies such as wavelet variable coefficients and quadratic interpolation in a development stage to enhance the optimization precision and convergence speed of the algorithm; finally, a simulation result shows that the optimization performance of the improved algorithm on a benchmark test function is obviously superior to that of a comparison algorithm, and the comprehensive routing effect in path planning simulation is optimal. Compared with a comparison algorithm, the improved snake vulture optimization algorithm provided by the invention has better global optimization capability and robustness, and the path planning capability of the algorithm is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robots, in particular to a photovoltaic power station cleaning robot path planning method improved by a secretary bird optimization algorithm. BACKGROUND

[0002] With the rapid development of photovoltaic power generation technology, the scale of photovoltaic power stations is continuously expanding, and the demand for cleaning and maintenance is increasing. Traditional manual cleaning methods are inefficient and costly, while automated cleaning robots are gradually becoming an important tool for photovoltaic power station operation and maintenance due to their high efficiency and precision. However, the environment of photovoltaic power stations is complex, and cleaning robots need to move efficiently in heterogeneous terrain (such as hardened roads represented by cement roads and non-hardened roads represented by grasslands), while avoiding obstacles such as photovoltaic modules and supports. Therefore, how to achieve efficient path planning for cleaning robots has become one of the hot issues in current research.

[0003] Path planning, as a core technology of robot autonomous navigation, aims to find the optimal or suboptimal path from the starting point to the target point. In recent years, swarm intelligence optimization algorithms have been widely applied in path planning due to their strong global search ability and robustness. For example, the particle swarm optimization algorithm can quickly converge to a better solution by simulating the foraging behavior of bird flocks; the ant colony algorithm is suitable for path search in complex environments by simulating the pheromone transmission mechanism in the foraging process of ants; the genetic algorithm has strong global optimization ability by simulating the biological evolution process. However, these traditional algorithms still have limitations such as slow convergence speed and easy to fall into local optimum when dealing with high-dimensional and multi-constrained path planning problems.

[0004] The secretary bird optimization algorithm (SBOA) is a new swarm intelligence optimization algorithm proposed in 2024, which is inspired by the efficient search strategy exhibited by secretary birds during hunting. By simulating the hunting behavior of secretary birds and combining global exploration and local exploitation capabilities, the algorithm has shown excellent performance in function optimization and engineering design. However, there is currently little research on the application of the secretary bird optimization algorithm in path planning, especially in the specific context of photovoltaic power station cleaning robots, which has not been systematically explored.

[0005] The present application is directed to the path planning problem of photovoltaic power station cleaning robots, and proposes a photovoltaic power station cleaning robot path planning method based on a multi-strategy improved secretary bird optimization algorithm (ISBOA). The main contributions of the research include: 1) a fitness function considering heterogeneous terrain features is designed, and through the introduction of a reward and punishment mechanism, the path length and terrain safety are balanced; 2) a path search strategy based on the improved secretary bird optimization algorithm is proposed, and a double-population iteration mechanism combining elite initialization, Brownian motion and Weibull distribution is introduced, and optimization strategies such as wavelet mutation coefficient and quadratic interpolation are developed in the development stage, combined with the prior knowledge of the photovoltaic power station environment, the convergence speed and path quality of the algorithm are improved; 3) simulation is carried out in the photovoltaic power station scene, and the effectiveness and robustness of the proposed method in various environments are verified. The research of the present application not only provides a new solution for the path planning of photovoltaic power station cleaning robots, but also provides theoretical support and practical reference for the application of swarm intelligence optimization algorithm in complex environment. SUMMARY

[0006] The present application aims to provide an improved secretary bird optimization algorithm for photovoltaic power station cleaning robot path planning method to improve the path planning ability of the improved algorithm.

[0007] To solve the above technical problems, the present application provides the following technical solutions: an improved secretary bird optimization algorithm for photovoltaic power station cleaning robot path planning method, comprising the following steps:

[0008] Step one: combine the environmental features and key parameters of the photovoltaic power station, the key parameters include geographic coordinates, terrain features and solar radiation, the photovoltaic modules adopt a southward arrangement strategy, are arranged in an array, and a grid map is drawn, marking the passable area and obstacle area, and determining the path starting point and path ending point;

[0009] Step two: introduce a reward and punishment mechanism to construct a fitness function that balances the evaluation of path quality;

[0010] Step three: generate an initial secretary bird population through the elite initialization method, and the generated initial secretary bird population is used as several initial paths of the photovoltaic power station cleaning robot;

[0011] Step four: the fitness function calculates the fitness value corresponding to each initial secretary bird population, finds the global optimal secretary bird position, and updates the secretary bird position in the exploration stage of the secretary bird optimization algorithm; the double-population iteration mechanism combining Brownian motion and Weibull distribution is introduced in the algorithm update stage;

[0012] Step five: compare the initial snake eagle population and the updated snake eagle population position, use the greedy strategy to retain the high-quality individuals of the exploration stage to form a new snake eagle population, enter the development stage of the snake eagle optimization algorithm to update the snake eagle position, and the wavelet mutation coefficient is fused when the algorithm is updated;

[0013] Step six: compare the latest position of the snake eagle with the last updated position, use the greedy strategy to retain the high-quality individuals of the development stage to form a new snake eagle population, update the global optimal snake eagle position of the snake eagle population, disturb the global optimal snake eagle position through quadratic interpolation method, compare the fitness values corresponding to the new and old global optimal snake eagle positions, retain the optimal fitness value, and enhance the ability of the algorithm to jump out of the local optimum;

[0014] Step seven: repeat steps four to six to update iteratively until the preset iteration number is reached, and output the global optimal snake eagle position as the optimal path.

[0015] Further, the fitness function in step two is set as follows:

[0016]

[0017]

[0018] In the formula: fitness is the comprehensive evaluation index; n is the total number of path nodes; alpha i is the terrain weight coefficient of the ith path (the path reward and punishment mechanism is realized through the terrain weight coefficient, for hardened roads, set alpha i = 1, that is, no additional penalty is applied; for non-hardened roads, set alpha i = beta (beta > 1) to increase the travel cost, and the specific value of beta can be adjusted according to actual engineering requirements, which is usually determined through experiments or experience); d i is the path length of the ith path; (x i , y i ) is the coordinate of the ith path node; (x i+1 , y i+1 ) is the coordinate of the i+1th path node.

[0019] Further, the specific method of generating the path in step three is to generate a plurality of non-obstacle area navigation points between the path starting point and the path ending point, and connect the path starting point, each non-obstacle area navigation point and the path ending point in turn, thereby generating a plurality of initial paths.

[0020] Further, the elite initialization method in step three is used to improve the diversity and quality of the initial population, and the specific steps are as follows:

[0021] S1, randomly generate N individuals according to a random initialization method, calculate the fitness value of each individual, take the first m individuals with smaller fitness values as elite individuals of the population, and require m to be a factor of N;

[0022] S2, randomly generate (N / m-1) new individuals around each elite individual;

[0023] S3, combine the m elite individuals and the (N-m) generated new individuals to form an elite population A, and randomly generate N individuals again to form a population B by the original random initialization method;

[0024] S4, calculate the fitness values of all individuals in populations A and B, sort all individuals in the two populations in order of fitness value from small to large, and the first N individuals with smaller fitness values form the initial population of the improved snake eagle optimization algorithm; wherein the calculation formula for generating new individuals from elite individuals is:

[0025] X e,l =X e ±σ×rand×(ub-lb)

[0026] X e,l =min(max(X e,l ,lb),ub)

[0027] Wherein, X e,l represents the position of the lth new individual generated near the e th elite individual, e = 1, 2, …, m, l = 1, 2, …, (N / m-1); X e represents the position of the e th elite individual; σ represents a coefficient; rand is a random number; ub and lb are the upper and lower bounds of the solution space.

[0028] Further, the step four combines the Brownian motion and Weibull distribution in the exploration stage of the snake eagle optimization algorithm to form a double-population iteration mechanism, which fully utilizes the advantages of different populations and further enhances population diversity to ensure the stability and global search ability of the algorithm. The specific implementation process is as follows: two different populations are obtained by Brownian motion and Weibull distribution, then the two populations are sorted uniformly, the first N individuals are selected according to the fitness value to form a new population for the next iteration, and the position update formula of the exploration stage of the double-population iteration mechanism is:

[0029]

[0030] In the formula: t is the current iteration number; T is the maximum iteration number; is the new position of the ith snake eagle individual in the exploration stage; X best is the global optimal position of the snake eagle; X iThe position of the ith snake eagle individual (where i = 1, 2,..., N, N is the population size); X random_1 and X random_2 is a random candidate solution in the exploration stage; R1 is a 1 x Dim-dimensional array randomly generated in the interval [0, 1], Dim is the dimension of the solution space; X best is the global optimal position of the snake eagle; RB is a random vector generated by the Brownian distribution; Rw is a random vector generated by the Weibull distribution; CF is an adaptive parameter; RL is a weighted Levy flight; f(x; λ; k) is the probability density function of the Weibull distribution; x is a random variable; λ > 0 is the scale parameter; k > 0 is the shape parameter; the present application sets the scale parameter of the Weibull distribution to 4 and the shape parameter to 0.3.

[0031] Further, in step five, the wavelet mutation coefficient is introduced into the development stage of the snake eagle optimization algorithm, so that the snake eagle can move with a larger step size in the later iteration stage to improve the local escape ability of the population. The position update formula of the development stage after the fusion of the wavelet mutation coefficient is:

[0032]

[0033] In the formula: is the new position of the ith snake eagle individual in the development stage; C1 and C2 are the first and second escape strategies respectively; η is the wavelet mutation coefficient; is a random number in [-2.5τ, 2.5τ], the stretching parameter τ decreases with the increase of the number of iterations, which inversely controls the mutation coefficient as the denominator, so that the wavelet mutation coefficient gradually decreases, thereby adaptively regulating the mutation degree of the population; δ is the shape parameter, δ = 0.5; g is the upper limit of τ, g = 100.

[0034] Further, in step six, the quadratic interpolation method is introduced to perturb the global optimal position to enhance the ability of the algorithm to jump out of the local optimum. The specific implementation process is as follows: two snake eagle individuals X a = (a1, a2,..., a Dim ) and X b = (b1, b2,..., b Dim ) are randomly selected from the snake eagle population, the global optimal position is X best = (c1, c2,..., c Dim ), and the corresponding fitness values are represented by F(X a ), F(X b ) and F(X best ) respectively. A new individual position information X new = (X new.1 , X new.2 ,..., X new.Dim), wherein the position information X of the new position in each dimension is generated by the following formula: new.j (j = 1, 2, …, Dim) is generated by the following formula:

[0035]

[0036] The beneficial effects of the present application are:

[0037] 1. The photovoltaic power station cleaning robot path planning method of the improved snake optimization algorithm provided by the present application proposes a fitness function considering heterogeneous terrain characteristics, effectively balances path length and terrain safety through a path reward and punishment mechanism, preferentially selects hardened roads with high stability and large friction coefficients, reduces the use of non-hardened roads, and reduces energy consumption and potential risks in the movement of the cleaning robot.

[0038] 2. The present application introduces an elite initialization method in the population initialization stage on the basis of the traditional snake optimization algorithm, improves the diversity and quality of the initial population, thereby accelerating the convergence speed of the algorithm and improving the convergence accuracy of the algorithm.

[0039] 3. The present application fuses a double-population iteration mechanism combining Brownian motion and Weibull distribution in the exploration stage on the basis of the traditional snake optimization algorithm, fully utilizes the advantages of different populations, further enhances population diversity, and ensures the stability and global search ability of the algorithm.

[0040] 4. The present application introduces a wavelet mutation coefficient in the development stage on the basis of the traditional snake optimization algorithm, so that the snake can move with a larger step in the later iteration stage to improve the local escape ability of the population.

[0041] 5. The present application introduces a quadratic interpolation method to perturb the global optimal position on the basis of the traditional snake optimization algorithm, enhancing the ability of the algorithm to jump out of the local optimum.

[0042] 6. The photovoltaic power station cleaning robot path planning method of the improved snake optimization algorithm provided by the present application verifies the optimization performance of the improved algorithm proposed by the present application on benchmark test functions and path planning experiments: the benchmark test function simulation results show that the optimization performance of the improved algorithm is significantly better than that of the comparative algorithm, the convergence accuracy is higher, the convergence speed is faster, the stability is strong, and the improved algorithm can effectively avoid falling into a local optimum, and is suitable for complex high-dimensional optimization problems; the path planning simulation results show that, in the 20x20 and 30x30 photovoltaic power station grid maps, compared with the comparative algorithm, the improved algorithm has the optimal comprehensive evaluation index and the best comprehensive pathfinding effect. The present application provides an efficient and reliable solution for the path planning of photovoltaic power station cleaning robots, and provides theoretical support and practical reference for the application of swarm intelligence optimization algorithms in complex environments. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 is a method flowchart in the embodiment of the application;

[0044] Figure 2 is a schematic diagram of a grid map and a robot eight-direction search method in the application;

[0045] Figure 3 is a comparison diagram of the convergence curves of various optimization algorithms on benchmark test functions in the application;

[0046] Figure 4 is a convergence curve diagram of various algorithms in path planning simulation in the application;

[0047] Figure 5 is a path planning effect diagram of the improved algorithm in the application. DETAILED DESCRIPTION

[0048] The following is further described in detail through specific embodiments:

[0049] The embodiment is basically as shown in the accompanying Figures 1-5 :

[0050] The photovoltaic power station cleaning robot path planning method of the improved snake eagle optimization algorithm, specifically includes the following steps:

[0051] Step one: combined with the environmental characteristics of the photovoltaic power station, the key parameters such as geographic coordinates, terrain features and solar radiation are comprehensively considered, so that the photovoltaic modules are usually arranged in a southward strategy, arranged in an array with an optimal inclination and a reasonable spacing, a grid map is drawn, the passable area and the obstacle area are marked, and the path starting point and the path ending point are determined;

[0052] Step two: the target function comprehensively considers two key factors of path length and terrain features, and a fitness function for balancing the quality of the path is constructed by introducing a reward and punishment mechanism;

[0053] Step three: an initial snake eagle population is generated by an elite initialization method, and the generated initial snake eagle population is used as a plurality of initial paths of the photovoltaic power station cleaning robot;

[0054] Step four: the fitness values corresponding to each snake eagle are calculated based on the above fitness function, the global optimal snake eagle position is found out, the snake eagle position is updated in the exploration stage of the snake eagle optimization algorithm, and the double-population iteration mechanism combining Brownian motion and Weibull distribution is fused in the algorithm improvement in the exploration stage;

[0055] Step five: the new and old positions of the snake eagle are compared, the high-quality individuals in the exploration stage are reserved to form a new snake eagle population by using a greedy strategy, the snake eagle position is updated in the development stage of the snake eagle optimization algorithm, and the wavelet mutation coefficient is fused in the algorithm improvement in the development stage;

[0056] Step six: compare the new and old positions of the snake eagle, use the greedy strategy to retain the high-quality individuals in the development stage to form a new snake eagle population, update the global optimal snake eagle position of the snake eagle population, perturb the global optimal snake eagle position by quadratic interpolation method, compare the fitness values corresponding to the new and old global optimal snake eagle positions, retain the optimal fitness value, and enhance the ability of the algorithm to jump out of the local optimum;

[0057] Step seven: repeat steps four to six to perform iterative updates until a preset iteration number is reached, and output the global optimal snake eagle position as the optimal path.

[0058] The embodiment enhances the optimization performance of the intelligent path planning method by fusing multiple improvement strategies, improves the path planning effect, and realizes accurate planning of the path of the photovoltaic power station cleaning robot. The photovoltaic power station cleaning robot path planning method based on the improved snake eagle optimization algorithm comprises the following steps:

[0059] S1: In combination with the environmental characteristics of the actual photovoltaic power station, the detailed description of the photovoltaic power station cleaning robot path planning method is as follows:

[0060] In the planning and design of photovoltaic power stations, key parameters such as geographic coordinates, terrain characteristics, and solar radiation need to be considered to maximize the use of light resources. Photovoltaic components are usually arranged in a southward strategy with an optimal inclination and reasonable spacing to form a photovoltaic square array. This standardized layout not only maximizes the use of site light resources, but also provides convenient conditions for inspection, maintenance, and cleaning operations. The present application considers the movement constraints in heterogeneous terrain environments for the path planning of photovoltaic power station cleaning robots. In the research scenario, the travel area of the cleaning robot includes two typical terrains: cement-hardened roads and grass-softened roads. Based on actual engineering experience and safety standards, cement-hardened roads are preferred as the main moving path for the cleaning robot due to their high load-bearing capacity and stable friction coefficient, while the feasibility of traversing grass-softened road areas when necessary is still retained.

[0061] S2: In the aspect of simulation environment modeling, the grid method is used to model and represent photovoltaic components and different path terrains in the photovoltaic power station.

[0062] The constructed grid map is as follows: Figure 2As shown in the figure, the white grid represents the passable area of the hardened road terrain represented by the cement road, the gray grid represents the passable area of the non-hardened road terrain represented by the grassland, and the black grid represents the obstacle area, and the size of the obstacle area is set based on the actual geometric size of the photovoltaic module, while considering the radius and safety margin of the cleaning robot body to effectively avoid the edge collision risk in the path planning process. The white, gray and black grids can be represented by the numbers "0", "1" and "2" respectively; and it is assumed that the cleaning robot uses an eight-direction search method to explore the path.

[0063] S3: The target function comprehensively considers two key factors of path length and terrain characteristics, and a fitness function for balancing the quality of the path is constructed by introducing a reward and punishment mechanism.

[0064] In the actual photovoltaic power station environment, the roads of different terrains have a significant impact on the safety and efficiency of the cleaning robot movement. In order to guide the algorithm to preferentially select safer and more efficient paths, a path reward and punishment mechanism based on terrain characteristics is designed in this paper. Specifically, the road is divided into two categories: hardened road represented by cement road and non-hardened road represented by grassland. The hardened road is suitable for the robot to move quickly due to its high stability and large friction coefficient, so it is given a lower penalty coefficient (or a higher reward coefficient); while the non-hardened road is given a higher penalty coefficient due to the risk of slipping and potential damage to vegetation.

[0065] In constructing the fitness function, the two key factors of path length and terrain characteristics are comprehensively considered: the path length evaluation function based on Euclidean distance and the road reward and punishment mechanism based on terrain characteristics. The fitness function constructed is as follows:

[0066]

[0067]

[0068] In the formula: fitness is the comprehensive evaluation index; n is the total number of path nodes; α i is the terrain weight coefficient of the ith path (the path reward and punishment mechanism is realized through the terrain weight coefficient, for the hardened road, α i =1, that is, no additional penalty is applied; for the non-hardened road, α i =β(β>1) is set to increase the cost of passing, and the specific value of β can be adjusted according to actual engineering requirements, and is usually determined through experiments or experience); d i is the path length of the ith path; (x i ,y i ) is the coordinate of the ith path node; (x i+1 ,y i+1) is the coordinate of the i+1th path node.

[0069] S4: The specific method of the intelligent path planning method for generating a robot path is as follows: a plurality of non-obstacle region navigation points are generated between a path starting point and a path ending point, and the path starting point, each non-obstacle region navigation point and the path ending point are sequentially connected, so as to generate a plurality of initial paths.

[0070] S5: In order to further enhance the performance of the algorithm, the improved secretary bird optimization algorithm (ISBOA) is proposed in the present application, which is improved on the basis of the traditional secretary bird optimization algorithm (SBOA) from four aspects of population performance, local fine development, convergence speed and jumping out of local optimum: 1) the elite initialization method is introduced to improve the quality of the initial population; 2) the double-population iteration mechanism combining Brownian motion and Weibull distribution is adopted to fully exert the advantages of different populations and further enhance the population diversity, so as to ensure the stability and global search ability of the algorithm; 3) the wavelet mutation coefficient is introduced in the escape stage, so that the secretary bird can move with a larger step in the later iteration stage, so as to improve the local escape ability of the population; 4) the optimal value of the interpolation polynomial is used to continuously reduce the gap of the optimal value of the target space, so as to achieve better solution accuracy.

[0071] S6: The traditional secretary bird optimization algorithm uses the rand function random generation method in the population initialization stage, which has the shortcomings of poor initial population diversity and low population quality, which affects the convergence speed of the algorithm. In view of this, the elite initialization method is introduced to generate the initial population, a plurality of new individuals are randomly generated in the vicinity of the better individual to replace the poor individual, which is used to generate a group of high-quality initial population in the given parameter space. The implementation steps of the elite initialization method are as follows:

[0072] Firstly, N individuals are randomly generated according to the random initialization method, the fitness value of each individual is calculated, and the first m individuals with smaller fitness values are taken as the elite individuals of the population (it is required that m is a factor of N); then (N / m-1) new individuals are randomly generated in the vicinity of each elite individual; secondly, the m elite individuals and the (N-m) new individuals generated together constitute the elite population A, and the original random initialization method randomly generates N individuals to constitute the population B; finally, the fitness values of all individuals in the population A and the population B are calculated, and all individuals in the two populations are sorted in order of fitness value from small to large, and the first N individuals with smaller fitness values constitute the initial population of the improved secretary bird optimization algorithm. The calculation formula of the new individual generated by the elite individual is:

[0073] X e,l =X e±σ x rand x (ub - lb)

[0074] X e,l = min(max(X e,l , lb), ub)

[0075] where X e,l represents the position of the lth new individual generated near the e th elite individual, e = 1, 2, …, m, l = 1, 2, …, (N / m - 1); X e represents the position of the e th elite individual; σ represents the coefficient; rand is a random number; ub and lb are the upper and lower bounds of the solution space.

[0076] S7: Based on the above elite initialization stage, the initial population generated enters the exploration stage of the snake-eagle optimization algorithm for snake-eagle position update. The exploration stage combines the Brownian motion and Weibull distribution combined double population iteration mechanism to fully exert the advantages of different populations and further enhance the population diversity, ensuring the stability and global search ability of the algorithm.

[0077] The exploration process of the snake-eagle is usually divided into three stages: finding prey, consuming prey, and attacking prey. The three stages are switched by changing the number of iterations: when t < T / 3, the algorithm executes the prey finding stage; when T / 3 < t < 2T / 3, the algorithm executes the prey consumption stage; when t > 2T / 3, the algorithm executes the prey attacking stage.

[0078] The specific implementation process of the double population iteration mechanism combining Brownian motion and Weibull distribution is as follows: in the prey consumption stage, two different populations are obtained according to Brownian motion and Weibull distribution, respectively, and then the two populations are uniformly sorted, and the top N individuals are selected according to the fitness size to form a new population entering the next iteration. The position update formula of the prey consumption stage combined with Weibull distribution is:

[0079]

[0080] where t is the current iteration number; T is the maximum iteration number; is the new position of the ith snake-eagle individual in the exploration stage; X best is the global optimal position of the snake-eagle; X i is the position of the ith snake-eagle individual (where i = 1, 2, …, N, N is the population size); X random_1 and X random_2 are random candidate solutions in the exploration stage; R1 is a 1 x Dim array randomly generated in the interval [0, 1], Dim is the dimension of the solution space; X bestThe global optimal position of the snake eagle; RB is a random vector generated by Brown distribution; Rw is a random vector generated by Weibull distribution; CF is an adaptive parameter; RL is a weighted Levy flight; f(x; lambda; k) is a probability density function of Weibull distribution; x is a random variable; lambda>0 is a scale parameter; k>0 is a shape parameter; the application sets the scale parameter of Weibull distribution to 4 and the shape parameter to 0.3.

[0081] S8: The new and old positions of the snake eagle in the exploration stage are compared, and a greedy strategy is used to retain high-quality individuals to form a new snake eagle population, and the specific calculation formula is:

[0082]

[0083] In the formula: is the fitness value of the new position of the ith snake eagle individual in the exploration stage; F i is the fitness value of the current position of the ith snake eagle individual.

[0084] S9: Based on the above-mentioned double-population iterative mechanism combining Brown motion and Weibull distribution, a new snake eagle population is formed through a greedy strategy to enter the development stage of the snake eagle optimization algorithm for snake eagle position updating. The development stage integrates the wavelet mutation coefficient when improving the algorithm, so that the snake eagle can move with a larger step size in the later iteration stage to improve the local escape ability of the population.

[0085] Morlet wavelet mutation is a single-frequency negative sine exponential function proposed by French scientist J. Morlet in 1974, and its main advantage is that it can adjust the amplitude value according to the number of iterations to achieve dynamic fine-tuning of the disturbance, thereby enhancing the ability of the algorithm to jump out of the local optimum. The position updating formula of the development stage integrating the wavelet mutation coefficient is:

[0086]

[0087]

[0088] In the formula: is the new position of the ith snake eagle individual in the development stage; C1 and C2 are the first and second escape strategies, respectively; η is the wavelet mutation coefficient; is a random number in [-2.5τ, 2.5τ], and the stretching parameter τ decreases with the increase of the number of iterations, which inversely controls the mutation coefficient as the denominator, so that the wavelet mutation coefficient gradually decreases, thereby adaptively adjusting the mutation degree of the population; δ is a shape parameter, and δ=0.5 is taken; g is the upper limit of τ, and g=100 is taken.

[0089] S10: The new and old positions of the snake eagle in the development stage are compared, and a greedy strategy is used to retain high-quality individuals to form a new snake eagle population, and the specific calculation formula is:

[0090]

[0091] In the formula: Fitness value of the new position of the ith snake eagle individual in the escape stage.

[0092] S11: Based on the development stage of the fusion wavelet variation coefficient, the new snake eagle population formed by the greedy strategy is updated to the global optimal snake eagle position.

[0093] The quadratic interpolation method is a method for finding the optimal value in a given initial value region, which is applied to the improvement of swarm intelligence algorithms as an improved method for enhancing the local development capability of the algorithm. The main idea is to use a quadratic polynomial to continuously approximate the target function in the target search space, and use the optimal value of the interpolation polynomial to continuously narrow the gap of the optimal value of the target space, so as to achieve better solving accuracy.

[0094] The specific implementation process of the quadratic interpolation method for perturbing the global optimal position is as follows: two snake eagle individuals X a =(a1,a2,…,a Dim ) and X b =(b1,b2,…,b Dim ) are randomly selected from the snake eagle population, the global optimal position is X best =(c1,c2,…,c Dim ), and the corresponding fitness values are represented by F(X a ), F(X b ) and F(X best ) respectively. A new individual position information X new =(X new.1 ,X new.2 ,…,X new.Dim ) is generated through the quadratic interpolation strategy, wherein the position information X new.j (j=1,2,…,Dim) of the new position in each dimension is generated by the following formula:

[0095]

[0096] S12: Compare the fitness values of the global optimal snake eagle position before and after perturbation by the quadratic interpolation method, and select the optimal one through the greedy strategy to ensure that the solution after perturbation is better than the original solution. The specific calculation formula is:

[0097]

[0098] In the formula: F(X new) is the fitness value of the new position of the globally optimal individual after disturbance.

[0099] S13: After this stage is completed, to verify the optimization performance of the improved algorithm proposed in the application, simulation tests are carried out on benchmark test functions and path planning experiments.

[0100] S14: Select some benchmark test functions for simulation testing. The relevant information is shown in the table. F1-F4 and F7 are unimodal functions, focusing on the local optimization ability of the algorithm; F9-F13 are multimodal functions, focusing on the global optimization ability of the algorithm; F20, F22 and F23 are fixed-dimensional composite test functions, focusing on the balance between local and global optimization of the algorithm.

[0101] Table 1: Information of benchmark test functions

[0102]

[0103] Particle Swarm Optimization (PSO), Harris Hawk Optimization (HHO), Horned lizard optimization algorithm (HLOA), Black-winged Kite Algorithm (BKA), Snake Hawk Optimization Algorithm (SBOA) and the improved ISBOA algorithm obtained in this paper are selected for comparison experiments. The above comparison algorithms are uniformly set: population size N = 30, maximum iteration number T = 500, and each algorithm is independently run 30 times to avoid the randomness of the optimization results of the algorithm. The average value and standard deviation are used to quantify the optimization ability and stability of the algorithm, respectively. The optimization results of each algorithm on the benchmark test functions are shown in Table 2.

[0104] Table 2: Comparison of optimization results on benchmark test functions

[0105]

[0106]

[0107] On unimodal test functions (F1-F4, F7), the ISBOA algorithm shows excellent local optimization ability. The average value and standard deviation of the ISBOA algorithm are better than those of PSO, HHO, HLOA, BKA and SBOA algorithms. Especially on F1-F4 functions, ISBOA can stably converge to the theoretical optimal value (standard deviation is 0), indicating that it has very high convergence accuracy and stability. This result verifies the efficiency and robustness of ISBOA algorithm in local search tasks.

[0108] On the multimodal test functions (F9~F12), the ISBOA algorithm further demonstrates its strong global optimization ability. The experimental results show that the average value and standard deviation of ISBOA on all multimodal functions are optimal compared with the comparative algorithms, especially on F9 and F11 functions, which can converge to the theoretical optimal value by 100%. This performance advantage shows that ISBOA has stronger global search ability and the ability to jump out of local optimum when dealing with complex multimodal problems, which benefits from its multi-strategy fusion design idea.

[0109] On the fixed dimension composite test functions (F20, F22, F23), the ISBOA algorithm also performs well. The average value obtained is very close to the theoretical optimal value, and the standard deviation is significantly lower than that of the comparative algorithm. This result shows that ISBOA can effectively balance global exploration and local development, thereby maintaining high search performance in complex high-dimensional optimization problems.

[0110] To further verify the dynamic convergence characteristics of the algorithm, this paper draws the convergence curves (as shown in Figure 3 ) of each algorithm. The experimental results show that ISBOA can quickly converge to the theoretical optimal value or close to the theoretical optimal value with fewer iterations on all test functions, and its convergence speed is significantly better than that of the comparative algorithm. Especially in the early stage of iteration, ISBOA shows a nearly linear fast convergence characteristic, which provides strong support for its efficiency in practical applications.

[0111] In summary, this paper verifies the significant advantages of ISBOA algorithm in convergence accuracy, convergence speed, global optimization ability and stability through benchmark testing. Compared with PSO, HHO, HLOA, BKA and SBOA and other comparative algorithms, ISBOA driven by multi-strategy fusion shows stronger comprehensive performance and can effectively deal with complex optimization problems.

[0112] S15: To verify the performance of the improved snake-eagle optimization algorithm proposed in the application applied to path planning of photovoltaic power station cleaning robot, simulation is carried out on 20×20 and 30×30 grid maps constructed after referring to actual photovoltaic power station two longitudinal and three longitudinal photovoltaic arrays, and the path planning effects of different algorithms are compared.

[0113] The comprehensive evaluation index is set for each algorithm, the independent running number is 10 times, the population size , and the iteration number . Table 3 is the comparison result of the evaluation index value obtained when each algorithm performs path planning, Figure 5 the convergence curve and planning effect in the path planning process of each algorithm.

[0114] Table 3 path planning results of each algorithm

[0115]

[0116] From the results of Table 3, in the 20x20 grid map, the average comprehensive evaluation index of the optimal path obtained by the ISBOA algorithm is reduced by 6.97% compared with the original SBOA algorithm, and is reduced by 17.16%, 21.76%, 17.99% and 17.22% compared with PSO, HHO, HLOA and BKA algorithms respectively; in the 30x30 grid map, the average comprehensive evaluation index of the optimal path obtained by the ISBOA algorithm is reduced by 8.84% compared with the original SBOA algorithm, and is reduced by 28.47%, 22.35%, 20.38% and 21.04% compared with PSO, HHO, HLOA and BKA algorithms respectively. In addition, the ISBOA algorithm has achieved the optimal average comprehensive evaluation index and the optimal comprehensive evaluation index in the two grid maps, which shows its significant performance advantage. Figure 4 The convergence curve shown further indicates that the ISBOA algorithm can quickly converge to the optimal path in a small number of iterations, which proves its superiority in convergence speed and optimization accuracy. The path planning effect of the ISBOA algorithm is shown in Figure 5 The path planning effect of the ISBOA algorithm is shown in

[0117] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application, therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A photovoltaic power station cleaning robot path planning method for improving the Viper optimization algorithm, characterized in that, The method comprises the following steps: Step 1: combining the environmental characteristics and key parameters of the photovoltaic power station, the key parameters including geographical coordinates, terrain characteristics and solar radiation, the photovoltaic components are arranged in a southward arrangement strategy, array arrangement is performed, and a grid map is drawn to mark the passable area and the obstacle area, and the path starting point and the path ending point are determined; Step 2: introducing a reward and punishment mechanism to build a fitness function for balancing the quality of the path; Step 3: generating an initial vulture-eagle population through an elite initialization method, and the generated initial vulture-eagle population is used as a plurality of initial paths of the photovoltaic power station cleaning robot; Step 4: the fitness function calculates the fitness values corresponding to each initial vulture-eagle population, finds the global optimal vulture-eagle position, and updates the vulture-eagle position in the exploration stage of the vulture-eagle optimization algorithm; the exploration stage combines the double-population iteration mechanism of Brown motion and Weibull distribution during algorithm updating; Step 5: comparing the initial vulture-eagle population and the updated vulture-eagle population position, using the greedy strategy to retain the high-quality individuals in the exploration stage to form a new vulture-eagle population, and updating the vulture-eagle position in the development stage of the vulture-eagle optimization algorithm; the development stage combines the wavelet mutation coefficient during algorithm improvement; Step 6: comparing the latest vulture-eagle position and the last updated position, using the greedy strategy to retain the high-quality individuals in the development stage to form a new vulture-eagle population, updating the global optimal vulture-eagle position of the vulture-eagle population, perturbing the global optimal vulture-eagle position through the quadratic interpolation method, comparing the fitness values corresponding to the new and old global optimal vulture-eagle positions, retaining the optimal fitness value, and enhancing the ability of the algorithm to jump out of the local optimum; Step 7: repeating steps 4 to 6 for iterative updating until a preset iteration number is reached, and outputting the global optimal vulture-eagle position as the optimal path.

2. The photovoltaic power station cleaning robot path planning method for improving the snake optimizer algorithm according to claim 1, characterized in that: The fitness function in step 2 is set as follows: In the formula: fitness is a comprehensive evaluation index; n is the total number of path nodes; α i is the terrain weight coefficient of the ith path segment (the path reward and punishment mechanism is realized through the terrain weight coefficient, for a hardened road, α i =1, that is, no additional penalty is applied; for a non-hardened road, α i =β (β>1) is set to increase the travel cost thereof, and the specific value of β can be adjusted according to actual engineering requirements, and is usually determined through experiments or experience); d i is the path length of the ith path segment; (x i ,y i ) is the coordinate of the ith path node; and (x i+1 ,y i+1 ) is the coordinate of the i+1th path node.

3. The photovoltaic power station cleaning robot path planning method for improving the snake optimizer algorithm according to claim 2, characterized in that: The method for generating the path in step 3 is: generating a plurality of non-obstacle area navigation points between the path starting point and the path ending point, and sequentially connecting the path starting point, each non-obstacle area navigation point and the path ending point to generate a plurality of initial paths.

4. The photovoltaic power station cleaning robot path planning method for improving the snake optimizer algorithm according to claim 3, characterized in that: The elite initialization method in step 3 is used to improve the diversity and quality of the initial population, and the specific steps are as follows: S1, randomly generating N individuals according to the random initialization method, calculating the fitness value of each individual, and taking the first m individuals with smaller fitness values as the elite individuals of the population, wherein m is a factor of N; S2, randomly generating (N / m-1) new individuals around each elite individual; S3, combining the m elite individuals and the (N-m) new individuals generated to form an elite population A, and randomly generating N individuals to form a population B by the original random initialization method; S4, calculating the fitness values of all individuals in population A and population B, sorting all individuals in the two populations in order of fitness value from small to large, and taking the first N individuals with smaller fitness values to form the initial population of the improved vulture-eagle optimization algorithm; wherein the calculation formula of the new individual generated by the elite individual is: X e,l = X e ± σ × rand × (ub - lb) X e,l = min(max(X e,l , lb), ub) where X e,l represents the position of the lth new individual generated near the e th elite individual, e = 1, 2, …, m, l = 1, 2, …, (N / m - 1); X e represents the position of the e th elite individual; σ represents a coefficient; rand is a random number; ub and lb are the upper and lower bounds of the solution space.

5. The photovoltaic power station cleaning robot path planning method for improving the snake optimizer algorithm according to claim 4, characterized in that: In the fourth step, the double-population iteration mechanism combining Brownian motion and Weibull distribution is fused in the exploration stage of the snake eagle optimization algorithm, the advantages of different populations are fully given, and the population diversity is further enhanced to ensure the stability and global search ability of the algorithm. The specific implementation process is as follows: two different populations are obtained by Brownian motion and Weibull distribution respectively, then the two populations are uniformly sorted, the first N individuals are selected according to the fitness size to form a new population into the next iteration, and the position update formula of the exploration stage of the double-population iteration mechanism is: where t is the current iteration number; T is the maximum iteration number; is the new position of the ith Viper individual in the exploration phase; X best is the global optimal position of the Viper; X i is the position of the ith Viper individual (where i = 1, 2, …, N, N is the population size); X random_1 and X random_2 is the random candidate solution in the exploration phase; R1 is a 1 × Dim dimensional array randomly generated in the interval [0, 1], Dim is the dimension of the solution space; X best is the global optimal position of the Viper; RB is a random vector generated by the Brownian distribution; Rw is a random vector generated by the Weibull distribution; CF is an adaptive parameter; RL is a weighted Levy flight; f(x; λ; k) is the probability density function of the Weibull distribution; x is a random variable; λ > 0 is the scale parameter; k > 0 is the shape parameter; the present application sets the scale parameter of the Weibull distribution to 4 and the shape parameter to 0.

3.

6. The photovoltaic power station cleaning robot path planning method for improving the snake optimizer algorithm according to claim 5, characterized in that: In the fifth step, the wavelet mutation coefficient is introduced in the development stage of the snake eagle optimization algorithm, so that the snake eagle can move with a larger step in the later iteration stage, and the local escape ability of the population is improved. The position update formula of the development stage of the fusion wavelet mutation coefficient is: In the formula: is the new position of the ith individual of the snake in the development stage; C1 and C2 are respectively the first and second escape strategies; η is a wavelet variation coefficient; is a random number within [-2.5τ, 2.5τ], and the scaling parameter τ decreases with the increase of the number of iterations, which inversely controls the variation coefficient as the denominator, so that the wavelet variation coefficient gradually decreases, thereby being able to adaptively control the variation degree of the population; δ is a shape parameter, and δ = 0.5 is taken; g is the upper limit of τ, and g = 100 is taken.

7. The photovoltaic power station cleaning robot path planning method for improving the snake optimizer algorithm according to claim 6, characterized in that: The step six introduces quadratic interpolation method to perturb the global optimal position, and enhances the ability of the algorithm to jump out of the local optimum. The specific implementation process is as follows: two snake eagle individuals X a =(a1,a2,…,a Dim ) and X b =(b1,b2,…,b Dim ) are randomly selected from the snake eagle population, the global optimal position is X best =(c1,c2,…,c Dim ), and the corresponding fitness values are represented by F(X a ), F(X b ) and F(X best ) respectively. A new individual position information X new =(X new.1 ,X new.2 ,…,X new.Dim ) is generated by quadratic interpolation strategy, wherein the position information X new.j (j=1,2,…,Dim) of the new position in each dimension is generated by the following formula:

Images