An optimization method for water flow rate control of a fishery experimental pool

Abstract

This invention discloses a method for optimizing water flow velocity control in a fishery experimental pond, belonging to the technical field of control optimization. The method includes the following steps: constructing a PID control system for water flow velocity in the fishery experimental pond that incorporates an improved cosmic evolution optimization algorithm module; tuning and applying the improved cosmic evolution optimization algorithm to the PID control parameters for water flow velocity in the fishery experimental pond. This invention focuses on optimizing water flow velocity control in fishery experimental ponds, introducing multiple improvement strategies to enhance the cosmic evolution optimization algorithm. The optimized PID parameter combination obtained through the improved cosmic evolution optimization algorithm effectively reduces overshoot and settling time in the water flow velocity control process, while strengthening the suppression capability against load disturbances and system parameter perturbations such as sudden water flow changes. This improves the dynamic performance and robustness of the overall control system, effectively ensuring the accuracy and reliability of water flow velocity detection and regulation in the fishery experimental pond.

Description

Technical Field

[0001] This invention belongs to the technical field of control optimization, and in particular relates to a method for controlling and optimizing the water flow velocity in a fishery experimental pond. Background Technology

[0002] Fisheries experimental ponds serve as the basic culture medium for aquatic animals in fisheries experiments, playing a crucial role in maintaining water flow and optimizing the culture environment. Water flow velocity control is fundamental to ensuring the healthy growth of aquatic animals and the efficiency of experiments. During operation, the experimental pond experiences continuous flow velocity fluctuations due to pump operation and changes in the external environment. These fluctuations are controlled by flow regulation devices. Water flow velocity directly affects the behavior of aquatic animals and water quality parameters. Excessively high flow velocities may lead to stress and increased energy consumption in aquatic animals, while excessively low flow velocities may cause water quality deterioration and pathogen proliferation. Currently, the commonly used solutions are manual adjustment or simple PID control methods, maintaining stable flow velocity through preset parameters or manual intervention. However, these methods suffer from limitations when dealing with complex hydrodynamics and nonlinear systems, including insufficient control precision and slow response. For example, they rely on operator experience; the adjustment process is cumbersome; and they cannot adapt to environmental changes. Currently, some researchers are attempting to use metaheuristic optimization algorithms such as genetic algorithms and particle swarm optimization to optimize flow velocity control parameters. Compared with traditional methods, these algorithms have shown some improvement. They search for optimal solutions by simulating the natural evolutionary process, thereby improving control performance and stability. However, standard metaheuristic optimization algorithms still face problems such as being prone to getting trapped in local optima, slow convergence speed, and unstable performance when dealing with highly nonlinear systems. Further optimization requires the development of more efficient and robust control strategies to achieve better flow velocity control effects in fishery experimental water bodies. Summary of the Invention

[0003] To overcome the technical problems described in the background section, this invention provides an optimization method for water flow velocity control in a fishery experimental pond. Focusing on the optimization of water flow velocity control in the fishery experimental pond, it introduces multiple improved strategies to enhance the cosmic evolution optimization algorithm. The improved PID parameter combination obtained through the optimized cosmic evolution optimization algorithm effectively reduces overshoot and settling time in the water flow velocity control process. Simultaneously, it strengthens the suppression of load disturbances and system parameter perturbations such as sudden water flow changes, improving the dynamic performance and robustness of the overall control system. This effectively ensures the accuracy and reliability of water flow velocity detection and regulation in the fishery experimental pond.

[0004] The technical solution of this invention is: a method for optimizing water flow velocity control in a fishery experimental pond, comprising the following steps: S1. Construct a PID control system for water flow velocity in a fishery experimental pond, including a water flow velocity error calculation module, a water flow velocity PID controller module, an improved cosmological evolution optimization algorithm module, a water flow velocity regulation module, and a water flow velocity monitoring module. S2. Introduce an improved cosmological evolution optimization algorithm. The specific improvement strategy is as follows: S21. Introduce a Torus sequence initialization strategy during the population initialization stage to achieve the effects of obtaining a uniform initial coverage in any dimension, eliminating clusters and voids, and improving the diversity of the initial population. S22. After Torus sequence initialization, a quasi-oppositional learning initialization strategy is introduced to achieve the effect of covering the center to the edge interval and improving the initial diversity and starting point quality. S23. When the stagnation triggering condition is met, the Cauchy heavy-tailed mutation escape strategy is introduced to achieve the effect of using heavy-tailed large step-length jumps to make up for the lack of global jumps after the expansion speed collapse and to escape the local optimum by crossing the deep potential well. S24. At the end of each generation of the main iteration loop, a stagnation detection adaptive triggering strategy is introduced to achieve the effect of turning blind perturbation into targeted escape and adaptive response to stagnation state. S25. After each generation of individual position updates, a Hooke-Jeeves pattern search local refinement strategy is introduced to eliminate low-order fractional oscillations at the end, improve convergence accuracy and accelerate convergence in high-potential regions. S3. The improved cosmological evolution optimization algorithm is used to tune the PID control parameters of the experimental pond water flow velocity in the PID control system of the fishery experimental pond, and the optimal control parameters are obtained through optimization. S4. The three optimal control parameters obtained by using the improved cosmological evolution optimization algorithm are set as the parameters of the PID controller for the water flow velocity in the experimental pond in the fishery experimental pond water flow velocity PID control system, thereby optimizing the water flow velocity regulation and control performance of the experimental pond.

[0005] Furthermore, in the PID control system for water flow velocity in the fishery experimental pond constructed in step S1, an improved cosmological evolution optimization algorithm module is run offline using historical operating data or a dynamic identification model to globally optimize the PID control parameters of the water flow velocity PID controller module. The obtained optimal parameters are then written into the controller. During operation, the experimental pond water flow velocity monitoring module collects the actual flow velocity in real time and transmits it to the experimental pond water flow velocity error calculation module. The error calculation module receives the set target flow velocity and compares and calculates the target flow velocity with the actual flow velocity, outputting a flow velocity deviation signal to the experimental pond water flow velocity PID controller module. Then, the controller, which has been written with the optimal parameters, calculates the precise control quantity based on the deviation signal and sends it to the experimental pond water flow velocity adjustment module to execute the flow velocity adjustment operation.

[0006] Furthermore, the Torus sequence initialization strategy in step S21 includes the following steps: Step S211: Generate irrational basis according to dimension index In the formula Indicates the first The irrational basis of dimensionality Indicates the first A prime number, This indicates the floor function; Step S212: By individual serial number With irrational basis Generate Torus low-dissimilarity sequence coordinate components In the formula Indicates the first The individual The coordinate components of the Torus sequence in dimensionality; Step S213: Convert the Torus sequence coordinate vector Mapping to the search space forms the initial population of Torus. In the formula Indicates the first The initial position vectors of each Torus individual. , Let represent the lower boundary vector of the search space and the upper boundary vector of the search space, respectively. This indicates element-wise multiplication.

[0007] Furthermore, the quasi-oppositional learning initialization strategy in step S22 includes the following steps: Step S221: Calculate the center vector of the search space As a reference center for quasi-oppositional learning, among which , Let these represent the lower boundary vector of the search space and the upper boundary vector of the search space, respectively. Step S222: Using the center of the search space as a reference, perform quasi-opposites learning on the initial Torus population to generate a quasi-opposites population. In the formula Indicates the first The position vectors of quasi-opposites Indicates the interval The numbers above follow a uniform distribution. Indicates the first The initial position vectors of each Torus individual. This represents element-wise multiplication. , Let these represent the lower boundary vector of the search space and the upper boundary vector of the search space, respectively. Step S223: After merging the initial Torus population with the quasi-opposing population, sort them in ascending order of fitness and select the top... Each individual is used as the initial population, and the position of the globally optimal individual is recorded. With global optimal fitness .

[0008] Furthermore, the Cauchy heavy-tailed mutation escape strategy in step S23 includes the following steps: Step S231: Determine the number of elite individuals involved in the escape. And select the first one after sorting. An elite individual, in the formula This indicates the rounding up operation. Indicates population size; Step S232, for the first An elite individual executes the Cauchy heavy-tail perturbation. In the formula Indicates the first The escape candidate position vectors of elite individuals Indicates the first The current position vector of an elite individual. Describes the Cauchy distribution scale vector and is equal to , Indicates the interval A random vector that follows a uniform distribution. Represents pi (π). This represents element-wise multiplication; Step S233: When the fitness of the escape candidate position is less than the first... When the current fitness of an elite individual is such that the escape candidate position is replaced by the first one... The current position of each elite individual is determined, and the position of the globally optimal individual is updated synchronously when the fitness of the escape candidate position is less than the globally optimal fitness. With global optimal fitness After execution is complete, the stall count will be reset to zero.

[0009] Furthermore, the stagnation detection adaptive triggering strategy in step S24 includes the following steps: Step S241: Calculate the stagnation determination threshold With stag window length In the formula This represents the globally optimal fitness. Indicates the maximum number of iterations. This indicates the rounding up operation; Step S242: When the decrease in the global optimal fitness between two adjacent generations is less than the stagnation threshold. Increment the stall count by one if the stall occurs, otherwise set the stall count to zero. Step S243: When the stall count reaches the stall window length At that time, trigger the sorting of the previous... The Cauchy heavy-tail perturbation escapes for each elite individual, and the stall count is reset to zero after the escape is completed.

[0010] Furthermore, the local refinement strategy for the Hooke-Jeeves pattern search in step S25 includes the following steps: Step S251: Perform bidirectional probing along each coordinate dimension for the globally optimal individual. In the formula Indicates along the first Candidate position vectors obtained by detecting in the positive or negative direction. This represents the globally optimal individual position vector. Indicates the first The detection step size of the dimension, Indicates the first A unit direction vector of dimension; Step S252: When the fitness of the candidate probe position is less than the global optimal fitness, replace the global optimal individual position with the candidate probe position and update the global optimal fitness. Step S253: When a round of detection in each dimension fails to improve the global optimal fitness, reduce the detection step size. In the formula Represents the probe step size vector. This represents the step size contraction factor.

[0011] Furthermore, in step S3, the improved cosmological evolution optimization algorithm is used to tune the PID control parameters of the experimental pond water flow velocity in the PID control system, including the following steps: Step S301: Perform parameter initialization and enhanced population initialization. Generate the initial Torus population using the low-dissimilarity Torus sequence and construct a quasi-opposite population with the center of the search space as a reference. After merging the initial Torus population and the quasi-opposite population, sort them in ascending order of fitness and select the top few individuals with a number equal to the population size as the formal initial population. Record the position of the globally optimal individual and the globally optimal fitness. Step S302: Enter the main iteration loop and update the subpopulation structure and adaptive evolution control parameters. Sort the current population in ascending order of fitness and select the top few individuals equal to the number of subpopulations as the centers of each subpopulation. Calculate the average radius of each subpopulation based on the number of individuals participating in the radius calculation and the Euclidean norm to the subpopulation center. Calculate the current generation's cosmic expansion rate, alignment coefficient, collision trigger probability, and orbital resonance trigger probability. The cosmic expansion rate decreases with iterations by both linear and exponential decay terms. The alignment coefficient increases linearly with iterations. The collision trigger probability decreases with iterations. The orbital resonance trigger probability decreases exponentially with iterations. Step S303: Perform global perturbation and multi-subgroup comprehensive guidance force calculation for subgroup collision. Calculate the mean of the average radius of all subgroups as the collision perturbation step size. When the uniform random number is less than the collision trigger probability, generate collision perturbation candidate positions by scaling and superimposing a standard normal random vector according to the collision step size and replacing the current position when the fitness is smaller. Calculate the guidance force of each subgroup on the current individual in the direction pointing to the subgroup center, with the intensity decreasing exponentially with distance. Weight the guidance force of each subgroup according to the weight obtained by exponential mapping between the individual fitness and the subgroup center fitness to obtain the comprehensive guidance force. Step S304: Perform individual position update and optimal alignment and orbital resonance local perturbation. Calculate the alignment displacement of the current individual pointing to the global optimal individual. Update the individual candidate position by combining the cosmic expansion displacement, the combined guiding force and the alignment displacement. If the fitness of the candidate position is smaller, accept the position update. Otherwise, apply a normal resonance perturbation with an amplitude base of 0.01 to the unimproved individual when the uniform random number is less than the orbital resonance trigger probability and accept it when the fitness is smaller. When the fitness of the updated individual is less than the global optimal fitness, update the global optimal individual position and the global optimal fitness simultaneously. Step S305: Perform local refinement of the global best individual and stagnation-triggered heavy-tailed escape. Perform bidirectional mode search and detection along each coordinate dimension for the global best individual. If the detection of candidate positions improves the global best fitness, the individual is accepted for replacement. If the detection of each dimension in the whole round does not improve the fitness, the detection step size is reduced by the step size contraction factor. The stagnation determination threshold is used to compare the cumulative decrease in global best fitness between adjacent generations or to clear the stagnation count. When the stagnation count reaches the stagnation window length, Cauchy distribution heavy-tailed perturbation is applied to the top few elite individuals whose number after sorting is equal to the number of elite individuals participating in the escape. The individuals are accepted when their fitness is smaller. After the execution is completed, the stagnation count is set to zero. Step S306: Record convergence information and determine termination conditions. Record the global optimal fitness and global optimal individual position of the current generation into the convergence curve vector and the historical optimal solution trajectory matrix. When the number of iterations reaches the maximum number of iterations, terminate the algorithm and output the global optimal individual position, global optimal fitness, convergence curve and historical optimal solution trajectory. If the maximum number of iterations has not been reached, increment the number of iterations and return to the subgroup structure update stage of the main iteration loop to continue the next generation iteration.

[0012] The beneficial effects of this invention due to the adoption of the above-mentioned technologies are as follows: This invention focuses on the control and optimization problem of water flow velocity in a fishery experimental pond, and proposes a solution for control and optimization using an improved cosmological evolution optimization algorithm. First, a Torus sequence initialization strategy is introduced during the population initialization stage to achieve uniform initial coverage in any dimension, eliminate clusters and voids, and improve the initial population diversity. Second, a quasi-oppositional learning initialization strategy is introduced after the Torus sequence initialization to achieve coverage from the center to the edge region and improve the initial diversity and starting quality. Third, when the stagnation trigger condition is met, a Cauchy heavy-tailed mutation escape strategy is introduced to compensate for the global jump loss after the expansion velocity collapse by using a large step-size jump with a heavy tail, and to escape the local maximum by crossing the deep potential well. The system achieves superior performance and introduces a stagnation detection adaptive triggering strategy at the end of each generation of the main iteration loop. This strategy transforms blind disturbances into targeted escapes and adaptive responses to stagnation states. Furthermore, after each generation's individual position update, a Hooke-Jeeves pattern search local refinement strategy is introduced to eliminate low-order fractional oscillations at the end of the process, improve convergence accuracy, and accelerate convergence in high-potential regions. The PID parameter combination optimized by the improved cosmological evolution optimization algorithm effectively reduces overshoot and settling time in the water flow velocity control process. At the same time, it strengthens the ability to suppress load disturbances and system parameter perturbations such as sudden water flow changes, thereby improving the dynamic performance and robustness of the overall control system. This effectively ensures the accuracy and reliability of water flow velocity detection and regulation in the fishery experimental pond. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the present invention.

[0014] Figure 2 This is a comparison of the best fitness convergence curves of the cosmological evolution optimization algorithm of this invention and the improved cosmological evolution optimization algorithm.

[0015] Figure 3 This is a comparison chart of the system response curves of the cosmological evolution optimization algorithm of this invention and the improved cosmological evolution optimization algorithm.

[0016] Figure 4 This is a comparison chart of the PID control parameter Kp optimization curves of the cosmological evolution optimization algorithm of this invention and the improved cosmological evolution optimization algorithm.

[0017] Figure 5 This is a comparison chart of the PID control parameter Ki optimization curves of the cosmological evolution optimization algorithm of this invention and the improved cosmological evolution optimization algorithm.

[0018] Figure 6 This is a comparison chart of the PID control parameter Kd optimization curves of the cosmological evolution optimization algorithm of this invention and the improved cosmological evolution optimization algorithm. Detailed Implementation

[0019] Example 1: As Figure 1 As shown, the present invention provides a method for optimizing water flow velocity control in a fishery experimental pond, comprising the following steps: S1. Construct a PID control system for the water flow velocity in a fishery experimental pond, including a water flow velocity error calculation module, a water flow velocity PID controller module, an improved cosmic evolution optimization algorithm module, a water flow velocity regulation module, and a water flow velocity monitoring module. Utilizing historical operating data or a dynamic identification model, the improved cosmic evolution optimization algorithm module runs offline to globally optimize the PID control parameters of the water flow velocity PID controller module and writes the obtained optimal parameters into the controller. During operation, the water flow velocity monitoring module collects the actual flow velocity in real time and transmits it to the water flow velocity error calculation module. The error calculation module receives the set target flow velocity, compares and calculates the target flow velocity with the actual flow velocity, and outputs a flow velocity deviation signal to the water flow velocity PID controller module. Then, the controller, which has the optimal parameters written in, calculates the precise control quantity based on the deviation signal and sends it to the water flow velocity regulation module to execute the flow velocity regulation operation. S2. Introduce an improved cosmological evolution optimization algorithm. The specific improvement strategy is as follows: S21. In the population initialization phase, a Torus sequence initialization strategy is introduced, using the fractional part of the square root of the prime numbers in each dimension. For an irrational basis, generate low-discrepancy sequence components recursively according to individual index. It linearly maps to a deterministic tiling structure in the search space, achieving the effects of obtaining a uniformly filled initial cover in any dimension, eliminating clusters and voids, and improving the diversity of the initial population. S22. After Torus sequence initialization, a quasi-oppositional learning initialization strategy is introduced, using a search space center... As a reference, quasi-opposite populations are generated by random sampling between the center and the opposing point. The optimal bilateral expansion structure, which merges with the initial Torus population, achieves coverage from the center to the edge, further enhancing initial diversity and starting quality. S23. When the stagnation trigger condition is met, introduce the Cauchy heavy-tailed mutation escape strategy, and adopt the strategy of sorting the first... An elite individual exerts The scale vector, after inverse transformation An elite perturbation structure is generated by Cauchy heavy-tailed random perturbation and greedily selected for replacement, so as to make up for the lack of global jump after the expansion velocity collapse by using heavy-tailed large step jumps and escape the local optimum by crossing the deep potential well. S24. At the end of each iteration of the main iteration loop, an adaptive triggering strategy for stagnation detection is introduced, using a threshold-based approach. Determine the decrease in global optimal fitness between two adjacent generations, using the stagnation window. The closed-loop monitoring structure that accumulates stall counts and switches the Cauchy heavy tail escape accordingly achieves the effect of turning blind disturbances into targeted escape and adaptive response to stall states. S25. After each generation of individual position updates, a Hooke-Jeeves pattern search local refinement strategy is introduced, employing a method along each coordinate dimension to refine the globally optimal individual. Two-way detection, accept only when fitness improves, and press when there is no improvement in the entire round of detection. The deterministic coordinate direction search structure with shrinking step size achieves the effects of eliminating low-order fractional oscillations at the end, improving convergence accuracy, and accelerating convergence in high-potential regions. S3. The improved cosmological evolution optimization algorithm is used to tune the PID control parameters of the experimental pond water flow velocity PID control system, and the optimal control parameters are obtained through optimization. , , This includes the following steps: S301. Perform parameter initialization and enhanced population initialization. Generate the initial Torus population using the low-difference Torus sequence and construct a quasi-opposite population with the center of the search space as a reference. After merging the initial Torus population and the quasi-opposite population, sort them in ascending order of fitness and select the top few individuals with a number equal to the population size as the formal initial population. Record the position of the globally optimal individual and the globally optimal fitness. S302. Enter the main iteration loop and update the subpopulation structure and adaptive evolution control parameters. Sort the current population in ascending order of fitness and select the top few individuals equal to the number of subpopulations as the centers of each subpopulation. Calculate the average radius of each subpopulation based on the number of individuals participating in the radius calculation and the Euclidean norm to the subpopulation center. Calculate the current generation's cosmic expansion rate, alignment coefficient, collision trigger probability, and orbital resonance trigger probability. The cosmic expansion rate decreases with iterations by both linear and exponential decay terms. The alignment coefficient increases linearly with iterations. The collision trigger probability decreases with iterations. The orbital resonance trigger probability decreases exponentially with iterations. S303. Perform global perturbation calculation for subgroup collisions and comprehensive guidance force calculation for multiple subgroups. Calculate the mean of the average radius of all subgroups as the collision perturbation step size. When the uniform random number is less than the collision trigger probability, generate collision perturbation candidate positions by scaling and superimposing a standard normal random vector according to the collision step size and replacing the current position when the fitness is smaller. Calculate the guidance force of each subgroup on the current individual in the unit direction pointing to the subgroup center, with the intensity decreasing exponentially with distance. Weight the guidance force of each subgroup according to the weight obtained by exponential mapping of the difference between the individual fitness and the subgroup center fitness to obtain the comprehensive guidance force. S304. Perform individual position update and optimal alignment and orbital resonance local perturbation. Calculate the alignment displacement of the current individual pointing to the global optimal individual. Update the individual candidate position by combining the cosmic expansion displacement, the comprehensive guiding force and the alignment displacement. If the fitness of the candidate position is smaller, accept the position update. Otherwise, apply a normal resonance perturbation with an amplitude base of 0.01 to the unimproved individual when the uniform random number is less than the orbital resonance trigger probability and accept it when the fitness is smaller. When the fitness of the updated individual is less than the global optimal fitness, update the global optimal individual position and the global optimal fitness simultaneously. S305. Perform local refinement of the global best individual and stagnation-triggered heavy-tailed escape. Perform bidirectional pattern search and detection along each coordinate dimension for the global best individual. If the detection of candidate positions improves the global best fitness, it is accepted for replacement. If the detection of each dimension in the whole round does not improve, the detection step size is reduced by the step size shrinkage factor. The stagnation judgment threshold is used to compare the cumulative decrease in global best fitness between adjacent generations or to clear the stagnation count. When the stagnation count reaches the stagnation window length, Cauchy distribution heavy-tailed perturbation is applied to the top few elite individuals whose number after sorting is equal to the number of elite individuals participating in the escape. They are accepted when the fitness is smaller. After the execution is completed, the stagnation count is set to zero. S306. Record convergence information and determine termination conditions. Record the global optimal fitness and global optimal individual position of the current generation into the convergence curve vector and the historical optimal solution trajectory matrix. When the number of iterations reaches the maximum number of iterations, terminate the algorithm and output the global optimal individual position, global optimal fitness, convergence curve and historical optimal solution trajectory. If the maximum number of iterations has not been reached, increment the number of iterations and return to the subgroup structure update stage of the main iteration loop to continue the next generation iteration. S4. The three optimal control parameters obtained by using the improved cosmological evolution optimization algorithm are set as the parameters of the PID controller for the water flow velocity in the experimental pond in the fishery experimental pond water flow velocity PID control system, thereby optimizing the water flow velocity regulation and control performance of the experimental pond.

[0020] For the flow velocity control in the fishery experimental pond, due to the significant physical fluid inertia of water and the significant dynamic transmission delay in the physical pipeline of the circulating water tank, the transfer function of the controlled object in the PID control system for the flow velocity in the fishery experimental pond is characterized as a second-order continuous time-delay system containing a pure input time delay element. Its transfer function is: , In the formula, This indicates that the controlled object passes a function. This represents the physical gain of the controlled object, reflecting the energy conversion efficiency between the frequency modulation signal of the variable frequency pump and the steady-state water flow velocity. This represents the first inertial time constant of the controlled object. This represents the second inertial time constant of the controlled object. This represents the pure time delay of the controlled object input, which maps to the time delay of the water flow in the physical channel and the electrical delay of the frequency converter between the issuance of the control signal and the measurement of the water flow change by the high-precision flow velocity sensor. This represents the Laplace operator, after system identification. , , , The values ​​are 1, 8, 5, and 3.5 in sequence. After substituting the parameters, the corresponding transfer function is: ; The system exhibits large inertia and long time delay characteristics. To achieve closed-loop adaptive correction control of the flow velocity, the unity negative feedback closed-loop transfer function is constructed as follows: , In the formula, Represents the closed-loop transfer function. This represents the transfer function of the PID controller. This indicates that the controlled object passes a function.

[0021] In order to simulate the flow velocity switching scenario caused by the behavioral observation needs of different experimental fish species in fisheries experiments, the step amplitude is defined as: , In the formula, Indicates the step amplitude. This represents the target output, which is the baseline target flow rate set in the fisheries experiment. This represents the initial output of the system, that is, the initial static state of the experimental pool or the basic background flow velocity. , , The values ​​are 1, 1, and 0 in sequence; To meet the accuracy requirements of discretized control simulation and algorithm iteration, the simulation time series is as follows: , In the formula, Represents the simulated time series. Indicates the sampling step size. This indicates the simulation termination time, ensuring full-time coverage of the complete physical process of the water body moving from a static to a steady state. , The values ​​are 0.1 and 100 respectively. To prevent improper control parameter configuration from causing positive feedback oscillations in the flow field, which could lead to severe standing waves or reverse vortex water rolls in the experimental pool and thus disrupt flow field stability, the formula used to determine the divergence of the closed-loop response is as follows: , In the formula, This represents the threshold for determining closed-loop response divergence. Once the real-time flow velocity exceeds this physical threshold, it is determined that the flow field has experienced malignant resonance turbulence. Indicates the target output. The value is 2; The control error is defined as follows: , In the formula, Indicates time The control error dynamically reflects the degree of deviation between the current transient flow field and the experimentally set standard flow velocity. Indicates the target output. Indicates time The actual output flow rate of the system; The integral time absolute error (ITAE) index is as follows: , In the formula, Indicates the absolute error in integration time. To represent a time variable, due to the introduction of a time weighting factor. This index imposes a very high mathematical penalty on the residual error in the later stages of the system response. Its physical significance in fisheries experiments lies in the constraint algorithm's ability to select parameter combinations that can quickly overcome the 3.5s pure lag wake of the water body, eliminate low-frequency reactive oscillations at the end of the flow velocity, and avoid secondary interference from long-term flow field fluctuations on water quality parameters (such as dissolved oxygen stratification and suspended uneaten food). Indicates time The control error, Indicates the simulation termination time; The overshoot ratio is defined as follows: , In the formula, Indicates the overshoot ratio. This represents the maximum output value during the closed-loop response process. Indicates the target output. Indicates the step amplitude; To prevent panic stress, mechanical abrasions, or even death from severe physiological stress in the experimental fish caused by instantaneous high flow velocity overshoot, an overshoot and overlimit penalty is introduced, the corresponding mathematical expression of which is: , In the formula, This indicates the penalty for overshooting or exceeding limits. Indicates the overshoot ratio. This indicates the maximum allowable overshoot ratio, suppressing transient bounces in the flow velocity response at the source and locking the maximum allowable overshoot ratio at an extremely low threshold. The value is 0.02; Among them, the actual adjustment time The determination is based on the degree of approximation between the time-domain response output and the target output, and the determination formula is as follows: , In the formula, Indicates the actual adjustment time; Indicates the time-based search variable; Indicates time The actual output flow rate of the system; Indicates the target output; Indicates the step amplitude; This represents the threshold of the settling time range, used to define the physical boundary by which the actual flow velocity enters and stabilizes within the error band of the target value. The value is 0.02, meaning the flow rate is allowed to be within the target value. Slight fluctuations within a certain range are considered, in fisheries ecological behavior, to indicate that the flow field has reached the steady-state standard required for the experiment. The penalty for exceeding the adjustment time limit is as follows: , In the formula, This indicates the penalty for exceeding the adjustment time limit; This indicates that the aforementioned stable band threshold Determine the actual adjustment time generated; This indicates the maximum allowable adjustment time. This is to avoid overheating and wear of the frequency converter caused by high-frequency control signal modulation, and to prevent slight fluctuations in the experimental pool water temperature due to pump heat dissipation, which could interfere with the quantitative assessment of fish behavior. It also ensures efficient and stable sampling by the dynamic measurement equipment. The value is set to 50; when the actual adjustment time cannot be determined due to continuous system oscillation, that is, the actual flow velocity is not completely locked within the entire time domain cycle. Within the specified stability band, to maintain algorithm robustness and apply a maximum power consumption penalty, the actual settling time is... Direct value The corresponding value is 150; In summary, the fitness function of the water flow velocity control system in the fishery experimental pond is composed of the ITAE basic index, the stress overshoot penalty term, and the experimental efficiency adjustment time penalty term. The mathematical expression of the corresponding fitness function is: , In the formula, This represents the fitness function value. Indicates the absolute error in integration time. This indicates the overshoot and overlimit penalty weight. In order to establish the biological control bottom line of sacrificing a gentle ascent speed but never allowing the water flow to rush over and harm the experimental fish, an extremely strict weight suppression is imposed on this item. This indicates the penalty for overshooting or exceeding limits. This indicates the weight of the penalty for exceeding the adjustment time limit. This indicates a penalty for exceeding the adjustment time limit, used to restrict the transition process time to ensure sampling efficiency; , The values ​​are 2000 and 200, respectively. When the PID control parameters are not finite, the PID controller fails to build, or the closed-loop simulation crashes, corresponding to extreme hardware disasters such as the control signal exceeding hardware limits causing the actual frequency converter overload protection to trip and cut off power or the valve to jam, in order to ensure the absolute safety of system operation, the fitness function triggers an electrical overload prevention mechanism, the value of which is: , In the formula, This represents the penalty value for invalid solutions, used as the highest level of hardware protection barrier. When the maximum output value of the closed-loop response exceeds the closed-loop response divergence threshold, it indicates that the control parameters have caused severe standing waves or reverse eddies in the flow field, resulting in adverse fluid dynamic responses. The algorithm then triggers the flow field disaster recovery protection mechanism, and the fitness function takes the following value: , In the formula, This represents the penalty value for divergent solutions. By directly classifying the evolutionary branch of this parameter as a poor solution and discarding it, the algorithm is forced to converge to a highly stable and safe region that conforms to the laws of the physical flow field.

[0022] In the PID control system for water flow velocity in the fishery experimental pond, since the water is a continuously distributed fluid medium with significant mass inertia and shear viscous resistance, and is physically constrained by the nonlinear characteristics of the circulating water pump during variable frequency speed regulation, the frequency domain transfer function of the PID controller for water flow velocity in the fishery experimental pond is as follows in order to accurately characterize the dynamic gain compensation characteristics of the controller in the frequency domain: , In the formula, This represents the transfer function of a PID controller; This represents the proportionality coefficient, which in flow field control is essentially the static gain of transient error, determining the system's response start-up speed to flow velocity deviation. It represents the integral coefficient, used to eliminate static error in the flow field and overcome the reactive power loss caused by pool wall friction resistance and fluid viscosity; This represents the differential coefficient, used to introduce advanced predictive damping to suppress the kinetic energy surge generated by the water body under the large inertial drive of time-delay water. Represents the Laplace operator; In order to describe the closed-loop dynamic driving relationship between the control signal and the flow velocity error in the frequency domain, the frequency domain control law is as follows: , In the formula, This represents the Laplace transform of the controller output signal, which maps the frequency domain distribution of the inverter input control voltage. This represents the transfer function of the PID controller. Represents the Laplace transform of the control error signal; Substituting the PID controller transfer function, the frequency domain control law can be written as: , In the formula, This represents the Laplace transform of the controller output signal. This represents the proportionality coefficient. Represents the integral coefficient. Denotes the differential coefficient. Represents the Laplace operator. Represents the Laplace transform of the control error signal; To guide real-time time-domain physical control of physical hardware actuators such as frequency converters and circulating water pumps, the aforementioned frequency-domain control law is transformed into a time-domain description through inverse Laplace transform. The time-domain control law of the PID controller for water flow velocity in the fishery experimental pond is as follows: , In the formula, Indicates time The controller output signal corresponds to the frequency modulation control voltage of the frequency-controlled water pump at the physical level; Indicates the proportionality coefficient; Indicates time The control error, and its corresponding proportional action term This constitutes the basic driving force for time-domain control. When the flow velocity in the experimental pool changes abruptly due to the dense backflow of fish or external disturbances, this item directly adjusts the pump speed proportionally according to the current error amplitude to ensure the agility of flow velocity tracking. Indicates the term of action of integration. The term represents the integral variable. By performing full-time accumulation on the historical flow velocity error, its physical significance in the fishery experiment scenario is to continuously compensate for the inherent flow velocity hierarchy attenuation (i.e., physical static error) caused by viscous shear stress when water passes through the guide grid and boundary pool wall, ensuring that the water flow velocity in the core observation area of ​​the experimental pool can be locked at the target standard value set by the experiment without static error at the end of the full-cycle simulation. This represents the differential action term, which makes predictive adjustments based on the changing trend of the flow velocity error. Its purpose is to address the large inertia of the water body (the first inertial time constant). Second inertial time constant ) and long time delay in pipeline transmission (pure time delay) The response wake caused by the flow velocity rapidly approaches the target value and the error rate reverses sharply, providing deterministic physical reverse damping, suppressing the output power of the circulating water pump in advance, thereby forcibly blocking the transient bounce and surge of the flow velocity at the physical flow field level. This avoids overshoot caused by kinetic energy accumulation in traditional control, fundamentally constructing a behavioral safety physical boundary for experimental fish to prevent panic stress, and avoiding mechanical vibration and coil electrical overheating losses caused by violent oscillation of the control signal in the variable frequency water pump.

[0023] The improved cosmological evolution optimization algorithm was used to tune the PID control parameters of the experimental pond's water flow velocity in the PID control system. The detailed steps are as follows: Step 1: Perform parameter initialization and storage pre-allocation; Step 101: Set basic control parameters, including the number of subgroups. The value is 3, which is the expansion attenuation factor. The initial value of the alignment coefficient is 0.1. The value is 0.7, representing the base collision probability. The value is 0.2, and it is a very small positive number that is not divisible by zero. Values Step length contraction factor The value is 0.5, and the population size is... The value is 30; Step 102: Pre-allocate storage space for convergence curves and historical optimal solutions. ,and , In the formula, Indicates length is The convergence curve vector, Indicates size is The historical optimal solution trajectory matrix, This represents the maximum number of iterations and has a value of 1000. This represents the dimension of the problem and has a value of 30.

[0024] Step 2: Perform enhanced population initialization using Torus sequences and quasi-oppositional learning; Step 201: Generate irrational basis by dimension number. , In the formula, Indicates the first The irrational basis of dimensionality Indicates the first A prime number, This indicates the floor function.

[0025] Step 202: Generate Torus low-discrepancy sequence coordinate components based on individual serial numbers and the irrational basis. , In the formula, Indicates the first The individual The coordinate components of the Torus sequence in dimensionality. Indicates the individual serial number. Indicates the first The irrational basis of dimensionality.

[0026] Step 203: Map the Torus sequence coordinates to the search space to form the initial Torus population. , In the formula, Indicates the first The initial position vectors of each Torus individual. This represents the lower boundary vector of the search space, and its value is [0,0,0,]. This represents the boundary vector of the search space, with values ​​[30, 30, 30]. Indicates the first The Torus sequence coordinate vector of each individual, This indicates element-wise multiplication.

[0027] Step 204: Using the center of the search space as a reference, perform quasi-oppositional learning on the initial Torus population to generate a quasi-oppositional population. , In the formula, Indicates the first The position vectors of quasi-opposites Describes the center vector of the search space and equals , Indicates the interval The numbers above follow a uniform distribution. Indicates the first The initial position vectors of each Torus individual. Represents the lower boundary vector of the search space. This represents the boundary vector in the search space.

[0028] Step 205: After merging the initial Torus population with the quasi-opposing population, sort them in ascending order of fitness, and select the top... Each individual is used as the initial population, and the position of the globally optimal individual is recorded. With global optimal fitness .

[0029] Step 3: Enter the main iteration loop and update the subgroup structure parameters; Step 301: Sort the current population in ascending order of fitness, and select the top [number] [seeds] after sorting. Each individual serves as the center of its respective subgroup; Step 302: Determine the number of individuals participating in the subgroup radius calculation. , In the formula, Indicates participation in the The number of individuals calculated from the radius of a subgroup. Indicates population size, This represents the number of subgroups, and the constant 2 represents the lower limit of the number of individuals participating in the subgroup radius calculation.

[0030] Step 303: Calculate the average radius of each subgroup based on its distance to the center of the subgroup. , In the formula, Indicates the first The average radius of each subgroup Indicates participation in the The number of individuals calculated from the radius of a subgroup. Indicates pressing the number After sorting the central distances of the subgroups in ascending order, the... The position vector of each individual Indicates the first The center position vectors of each subgroup This represents the Euclidean norm.

[0031] Step 4: Calculate the adaptive evolution control parameters; Step 401, Calculate the... The rate of expansion of the universe in that era, , In the formula, Indicates the first The rate of expansion of the universe in that era, This represents the expansion attenuation factor and has a value of 0.1. Indicates the current iteration number. The maximum number of iterations is represented by 0.001, the constant 0 represents the linear decay coefficient of the expansion rate, and the constant 4 represents the exponential decay rate of the expansion rate.

[0032] Step 402, Calculate the... Alignment factor of generation, , In the formula, Indicates the first Alignment factor of generation, This represents the initial value of the alignment factor, which is 0.7. Indicates the current iteration number. This represents the maximum number of iterations, and the constant 0.0005 represents the linear growth coefficient of the alignment coefficient with each iteration.

[0033] Step 403, Calculate the... The probability of collision triggering in a generation, , In the formula, Indicates the first The probability of collision triggering in a generation, This represents the base collision probability and has a value of 0.2. Indicates the current iteration number. This represents the maximum number of iterations, and the constant 0.0005 represents the linear decay coefficient of the collision trigger probability with each iteration.

[0034] Step 404, Calculate the... The probability of orbital resonance triggering in a generation, , In the formula, Indicates the first The probability of orbital resonance triggering in a generation, Indicates the current iteration number. The maximum number of iterations is represented by 0.1, the constant 0.1 represents the base probability of orbital resonance, and the constant 3 represents the exponential decay rate of the orbital resonance trigger probability.

[0035] Step 5: Perform a global perturbation of subgroup collisions on each individual; Step 501: Calculate the mean of the average radii of all subgroups. , In the formula, This represents the mean of the average radii of all subgroups. Indicates the number of subgroups. Indicates the first The average radius of each subgroup.

[0036] Step 502, when random number Less than the collision trigger probability At that time, candidate collision perturbation locations are generated. , In the formula, This represents the candidate position vector after the collision perturbation. Indicates the first The current position vector of each individual. This represents the mean of the average radii of all subgroups. Let represent a random vector that follows a standard normal distribution. Indicates the interval The above are random numbers that follow a uniform distribution.

[0037] Step 503: When the fitness of the candidate collision perturbation position is less than the first... When an individual has its current fitness, the candidate position for collision perturbation is replaced. The current position of each individual; otherwise, retain the position of the first individual. The current location of each individual.

[0038] Step 6: Calculate the combined guiding force of multiple subgroups; Step 601: Calculate the pairwise values ​​of each subgroup. The direction and intensity of guidance for each individual. ,and , In the formula, Indicates the first Subgroups for the first The guiding force vector of each individual, Indicates by the first Individuals point to the first The unit direction vector of the center of each subgroup Indicates the first The center position vectors of each subgroup Indicates the first The current position vector of each individual. Indicates the first The average radius of each subgroup.

[0039] Step 602: Calculate the weight of each subgroup and the sum of the weights. ,and , In the formula, Indicates the first The weights of each subgroup, Indicates the first The fitness of an individual Indicates the first Fitness of subgroup centers This represents the globally optimal fitness. Represents the smallest positive number that is protected against division by zero and has a value of 0. , This represents the sum of the weights of all subgroups. This indicates the number of subgroups.

[0040] Step 603: Weighted normalization of the guiding force of each subgroup according to its weights to obtain the comprehensive guiding force. , In the formula, Indicates the action on the first The combined guiding force vector of each individual Indicates the first The weights of each subgroup, Indicates the first Subgroups for the first The guiding force vector of each individual, This represents the sum of the weights of all subgroups. This indicates the number of subgroups.

[0041] Step 7: Perform individual position updates and optimal alignment; Step 701, Calculate the... The reference direction and alignment displacement of each individual pointing to the globally optimal individual. , In the formula, Indicates the action on the first Alignment displacement vector of each individual, Indicates the first Alignment factor of generation, This represents the globally optimal individual position vector. Indicates the first The current position vector of each individual. Let represent a random vector that follows a standard normal distribution. This indicates element-wise multiplication.

[0042] Step 702, Update the integrated cosmic expansion displacement, integrated guiding force, and alignment displacement. Individual position, , In the formula, Indicates the first The updated candidate position vectors for each individual. Indicates the first The current position vector of each individual. Indicates the first The rate of expansion of the universe in that era, Let represent a random vector that follows a standard normal distribution. Represents the boundary vector in the search space. Represents the lower boundary vector of the search space. Indicates the action on the first The combined guiding force vector of each individual Indicates the action on the first Alignment displacement vector of each individual, This indicates element-wise multiplication.

[0043] Step 703: When the fitness of the updated candidate position is less than that of the first... When an individual has its current fitness, replace the candidate position with the updated position. Each individual is currently positioned and proceeds to step 9; otherwise, the position is retained. The individual is currently at its current position and proceeds to step 8.

[0044] Step 8: Perform local perturbation of orbital resonance on individuals that have not been improved; Step 801, when random number Less than the probability of orbital resonance triggering At that time, candidate locations for resonance perturbation are generated. , In the formula, This represents the candidate position vector after the resonance perturbation. Indicates the first The updated candidate position vectors for each individual. Let represent a random vector that follows a standard normal distribution. Represents the boundary vector in the search space. Represents the lower boundary vector of the search space. Indicates the interval The numbers above follow a uniform distribution. This indicates element-wise multiplication, and the constant 0.01 represents the amplitude base of the orbital resonance perturbation.

[0045] Step 802: When the fitness of the candidate resonant perturbation position is less than the first... When an individual has its current fitness, the candidate position of the resonant perturbation is replaced. The current position of each individual; otherwise, retain the position of the first individual. The current position of each individual; when the first individual... When the fitness of an individual's updated position is less than the global optimal fitness, the position of the first individual is used as the starting point. After each individual updates its position, it replaces the position of the globally optimal individual and updates the globally optimal fitness.

[0046] Step 9: Perform a Hooke-Jeeves pattern search for local refinement on the globally optimal individual; Step 901: Perform bidirectional probing along each coordinate dimension to find the globally optimal individual. , In the formula, Indicates along the first Candidate position vectors obtained by detecting in the positive or negative direction. This represents the globally optimal individual position vector. Indicates the first The detection step size of the dimension, Indicates the first A unit direction vector of dimension.

[0047] Step 902: When the fitness of a candidate probe position is less than the global optimal fitness, replace the global optimal individual position with the candidate probe position and update the global optimal fitness; when a round of probing in each dimension fails to improve the global optimal fitness, reduce the probe step size. , In the formula, Represents the probe step size vector. This represents the step size contraction factor and has a value of 0.5.

[0048] Step 10: Perform stall detection and Cauchy heavy tail escape; Step 1001: Calculate the stagnation threshold and the stagnation window length. ,and , In the formula, Indicates the threshold for determining stagnation. This represents the globally optimal fitness. Indicates the length of the stall window. Indicates the maximum number of iterations. Indicates the round-up operation, constant. The scaling factor represents the stagnation threshold. The constant 0.05 represents the proportion of the stagnation window to the maximum number of iterations, and the constant 5 represents the lower limit of the stagnation window length.

[0049] When the decrease in the global optimal fitness between two adjacent generations is less than the stagnation threshold Increment the stall count by one when the stall occurs; otherwise, set the stall count to zero.

[0050] Step 1002: When the stall count reaches the stall window length At that time, for the sorted first An elite individual executes the Cauchy heavy-tail perturbation. , In the formula, Indicates the first The escape candidate position vectors of elite individuals Indicates the first The current position vector of an elite individual. Describes the Cauchy distribution scale vector and is equal to , Indicates the interval A random vector that follows a uniform distribution. This represents the number of elite individuals who participated in the escape and is equal to... , This indicates element-wise multiplication, and the constant 0.5 indicates multiplying the interval... Uniform random number translation is used to construct the central translation constant of the inverse Cauchy transform. It represents pi (π).

[0051] When the fitness of the escape candidate position is less than the first When the current fitness of an elite individual is such that the escape candidate position is replaced by the first one... The system calculates the current position of each elite individual and updates the global optimal individual position and global optimal fitness simultaneously when the fitness of the escape candidate position is less than the global optimal fitness. After execution, the stall count is set to zero.

[0052] Step 11: Record convergence information and determine the termination condition; Step 1101, Record the... The global optimal fitness and the global optimal individual position of each generation. ,and , In the formula, Indicates the first The convergence curve records the values ​​of the generation. This represents the globally optimal fitness. The historical optimal solution trajectory matrix represents the first... OK, This represents the globally optimal individual position vector.

[0053] Step 1102, when the iteration count... Reaching the maximum number of iterations The algorithm terminates at time and outputs the globally optimal individual position. Global optimal fitness Convergence curve Trajectory of Historical Optimal Solution If the maximum number of iterations has not been reached, then let Then return to step 3 to continue executing the next iteration.

[0054] The random numbers mentioned in the above steps, unless otherwise specified, refer to those within the interval [missing information]. Random sampled values ​​that follow a continuous uniform distribution are called random numbers, while random numbers in vector form refer to random numbers whose components in each dimension independently follow the above distribution.

[0055] Simulations in Matlab showed that the optimal fitness value of the Cosmic Evolution Optimization Algorithm (UEO) was 276.062525, while the optimal fitness value of the Improved Cosmic Evolution Optimization Algorithm (IUEO) was 31.395985. Since the calculation of the optimal fitness value is based on the performance evaluation function of the control system, a lower optimal fitness value indicates that the integral error index of the closed-loop system transient process is smaller. The optimal fitness value obtained by IUEO is lower than that of UEO, indicating that the control parameters obtained by IUEO optimization can produce a lower cumulative error in the control process of the controlled system. And such Figure 2 As shown, the optimal fitness value of UEO drops to around 21400 in the 5th iteration and remains constant between the 5th and 53rd iterations, exhibiting a long-term horizontal stagnation. In contrast, the optimal fitness value of IUEO drops to around 450 in the 6th iteration, further drops to around 60 in the 11th iteration, and stabilizes at 31.395985 after the 53rd iteration. This indicates that the long-term horizontal stagnation of the UEO curve suggests that UEO is prone to getting trapped in local optima, and population diversity is lost in the early stages of iteration. The rapid and step-like decline of the IUEO curve in the early stages of iteration indicates that the improvement mechanism introduced by IUEO enhances the algorithm's ability to escape local extrema and reduces the number of iterations required to find the optimal solution. like Figure 3 As shown, the system output driven by UEO first reaches the target value of 1.0 at 18 seconds, with a maximum overshoot of 1.1. It reaches its peak at 23 seconds, experiences a downshoot of 0.96 at 41 seconds, and finally enters and maintains a steady state at the target value of 1.0 at 60 seconds. The system output driven by IUEO first reaches the target value of 1.0 at 10 seconds, with a maximum value of only 1.01. It fully enters a steady state at 15 seconds, and the transition process is oscillating, indicating that the differential coefficients obtained by IUEO optimization are superior. The value of 6.8025 is higher than the differential coefficient obtained by UEO optimization. With a value of 1.6739, the high differential coefficient provides stronger advance damping control, thereby suppressing overshoot during the transient process and eliminating oscillations in the controlled system. Simultaneously, the proportional coefficient optimized by IUEO... The value is 1.9355, which is higher than the proportional coefficient of UEO. The value is 1.3453, which enhances the system's response speed and shortens the rise time; like Figures 4-6 As shown, after the 80th iteration, UEO corresponds to IUEO. , and The optimization trajectories are all perfectly horizontal straight lines without any fluctuations, and the final states all converge to fixed values. This indicates that the horizontal trajectory of the parameter optimization later stage has no fluctuation characteristics, and that both algorithms can stably converge to a specific solution space in the later stage of the search. This shows that the entire optimization system has numerical stability and convergence determinism. Therefore, it can be seen that in the current process of controlling water flow velocity in fishery experimental ponds, IUEO has significant advantages over UEO, mainly including: 1. The final optimization performance index is lower, that is, the optimal fitness value obtained by IUEO is 31.395985, which is 244.666540 lower than the optimal fitness value of UEO 276.062525. This indicates that IUEO has stronger global optimization accuracy. 2. It converges faster and can effectively escape local extrema. That is, the fitness value of IUEO has dropped to around 60 by the 11th iteration, while UEO has been stuck around 21400 for a long time during the 5th to 53rd iterations. This shows that the improvement mechanism adopted by IUEO has improved the global search efficiency. 3. The dynamic response quality of the water flow velocity control system in the fishery experimental pond has been improved. Specifically, under the drive of IUEO optimized parameters, the rise time of the closed-loop control system has been shortened from 18 seconds to 10 seconds, the settling time from 60 seconds to 15 seconds, the maximum overshoot has been reduced from 1.1 to 1.01, and the oscillation and downshoot phenomena during the transition process have been eliminated.

Claims

1. A method for optimizing water flow velocity control in a fishery experimental pond, characterized in that: Includes the following steps: S1. Construct a PID control system for water flow velocity in a fishery experimental pond, including a water flow velocity error calculation module, a water flow velocity PID controller module, an improved cosmological evolution optimization algorithm module, a water flow velocity regulation module, and a water flow velocity monitoring module. S2. Introduce an improved cosmological evolution optimization algorithm. The specific improvement strategy is as follows: S21. Introduce a Torus sequence initialization strategy during the population initialization stage to achieve the effects of obtaining a uniform initial coverage in any dimension, eliminating clusters and voids, and improving the diversity of the initial population. S22. After Torus sequence initialization, a quasi-oppositional learning initialization strategy is introduced to achieve the effect of covering the center to the edge interval and improving the initial diversity and starting point quality. S23. When the stagnation triggering condition is met, the Cauchy heavy-tailed mutation escape strategy is introduced to achieve the effect of using heavy-tailed large step-length jumps to make up for the lack of global jumps after the expansion speed collapse and to escape the local optimum by crossing the deep potential well. S24. At the end of each generation of the main iteration loop, a stagnation detection adaptive triggering strategy is introduced to achieve the effect of turning blind perturbation into targeted escape and adaptive response to stagnation state. S25. After each generation of individual position updates, a Hooke-Jeeves pattern search local refinement strategy is introduced to eliminate low-order fractional oscillations at the end, improve convergence accuracy and accelerate convergence in high-potential regions. S3. The improved cosmological evolution optimization algorithm is used to tune the PID control parameters of the experimental pond water flow velocity in the PID control system of the fishery experimental pond, and the optimal control parameters are obtained through optimization. S4. The three optimal control parameters obtained by using the improved cosmological evolution optimization algorithm are set as the parameters of the PID controller for the water flow velocity in the experimental pond in the fishery experimental pond water flow velocity PID control system, thereby optimizing the water flow velocity regulation and control performance of the experimental pond.

2. The method for optimizing water flow velocity control in a fishery experimental pond according to claim 1, characterized in that: In the PID control system for water flow velocity in the fishery experimental pond constructed in step S1, an improved cosmological evolution optimization algorithm module is run offline using historical operating data or a dynamic identification model to globally optimize the PID control parameters of the water flow velocity PID controller module. The obtained optimal parameters are then written into the controller. During operation, the experimental pond water flow velocity monitoring module collects the actual flow velocity in real time and transmits it to the experimental pond water flow velocity error calculation module. The error calculation module receives the set target flow velocity and compares and calculates the target flow velocity with the actual flow velocity, outputting a flow velocity deviation signal to the experimental pond water flow velocity PID controller module. Then, the controller, which has been programmed with the optimal parameters, calculates the precise control quantity based on the deviation signal and sends it to the experimental pond water flow velocity adjustment module to execute the flow velocity adjustment operation.

3. The method for optimizing water flow velocity control in a fishery experimental pond according to claim 1, characterized in that, The Torus sequence initialization strategy in step S21 includes the following steps: Step S211: Generate irrational basis according to dimension index In the formula Indicates the first The irrational basis of dimensionality Indicates the first A prime number, This indicates the floor function; Step S212: By individual serial number With irrational basis Generate Torus low-dissimilarity sequence coordinate components In the formula Indicates the first The individual The coordinate components of the Torus sequence in dimensionality; Step S213: Convert the Torus sequence coordinate vector Mapping to the search space forms the initial population of Torus. In the formula Indicates the first The initial position vectors of each Torus individual. , Let represent the lower boundary vector of the search space and the upper boundary vector of the search space, respectively. This indicates element-wise multiplication.

4. The method for optimizing water flow velocity control in a fishery experimental pond according to claim 1, characterized in that, The quasi-oppositional learning initialization strategy in step S22 includes the following steps: Step S221: Calculate the center vector of the search space As a reference center for quasi-oppositional learning, among which , Let these represent the lower boundary vector of the search space and the upper boundary vector of the search space, respectively. Step S222: Using the center of the search space as a reference, perform quasi-opposites learning on the initial Torus population to generate a quasi-opposites population. In the formula Indicates the first The position vectors of quasi-opposites Indicates the interval The numbers above follow a uniform distribution. Indicates the first The initial position vectors of each Torus individual. This represents element-wise multiplication. , Let these represent the lower boundary vector of the search space and the upper boundary vector of the search space, respectively. Step S223: After merging the initial Torus population with the quasi-opposing population, sort them in ascending order of fitness and select the top... Each individual is used as the initial population, and the position of the globally optimal individual is recorded. With global optimal fitness .

5. The method for optimizing water flow velocity control in a fishery experimental pond according to claim 1, characterized in that, The Cauchy heavy-tailed mutation escape strategy in step S23 includes the following steps: Step S231: Determine the number of elite individuals involved in the escape. And select the first one after sorting. An elite individual, in the formula This indicates the rounding up operation. Indicates population size; Step S232, for the first An elite individual executes the Cauchy heavy-tail perturbation. In the formula Indicates the first The escape candidate position vectors of elite individuals Indicates the first The current position vector of an elite individual. Describes the Cauchy distribution scale vector and is equal to , Indicates the interval A random vector that follows a uniform distribution. Represents pi (π). This represents element-wise multiplication; Step S233: When the fitness of the escape candidate position is less than the first... When the current fitness of an elite individual is such that the escape candidate position is replaced by the first one... The current position of each elite individual is determined, and the position of the globally optimal individual is updated synchronously when the fitness of the escape candidate position is less than the globally optimal fitness. With global optimal fitness After execution is complete, the stall count will be reset to zero.

6. The method for optimizing water flow velocity control in a fishery experimental pond according to claim 1, characterized in that, The stagnation detection adaptive triggering strategy in step S24 includes the following steps: Step S241: Calculate the stagnation determination threshold With stag window length In the formula This represents the globally optimal fitness. Indicates the maximum number of iterations. This indicates the rounding up operation; Step S242: When the decrease in the global optimal fitness between two adjacent generations is less than the stagnation threshold. Increment the stall count by one if the stall occurs, otherwise set the stall count to zero. Step S243: When the stall count reaches the stall window length At that time, trigger the sorting of the previous... The Cauchy heavy-tail perturbation escapes for each elite individual, and the stall count is reset to zero after the escape is completed.

7. The method for optimizing water flow velocity control in a fishery experimental pond according to claim 1, characterized in that, The Hooke-Jeeves pattern search local refinement strategy in step S25 includes the following steps: Step S251: Perform bidirectional probing along each coordinate dimension for the globally optimal individual. In the formula Indicates along the first Candidate position vectors obtained by detecting in the positive or negative direction. This represents the globally optimal individual position vector. Indicates the first The detection step size of the dimension, Indicates the first A unit direction vector of dimension; Step S252: When the fitness of the candidate probe position is less than the global optimal fitness, replace the global optimal individual position with the candidate probe position and update the global optimal fitness. Step S253: When a round of detection in each dimension fails to improve the global optimal fitness, reduce the detection step size. In the formula Represents the probe step size vector. This represents the step size contraction factor.

8. The method for optimizing water flow velocity control in a fishery experimental pond according to claim 1, characterized in that, Step S3 involves tuning the PID control parameters for the water flow velocity in the fishery experimental pond using an improved cosmological evolution optimization algorithm. This includes the following steps: Step S301: Perform parameter initialization and enhanced population initialization. Generate the initial Torus population using the low-dissimilarity Torus sequence and construct a quasi-opposite population with the center of the search space as a reference. After merging the initial Torus population and the quasi-opposite population, sort them in ascending order of fitness and select the top few individuals with a number equal to the population size as the formal initial population. Record the position of the globally optimal individual and the globally optimal fitness. Step S302: Enter the main iteration loop and update the subpopulation structure and adaptive evolution control parameters. Sort the current population in ascending order of fitness and select the top few individuals equal to the number of subpopulations as the centers of each subpopulation. Calculate the average radius of each subpopulation based on the number of individuals participating in the radius calculation and the Euclidean norm to the subpopulation center. Calculate the current generation's cosmic expansion rate, alignment coefficient, collision trigger probability, and orbital resonance trigger probability. The cosmic expansion rate decreases with iterations by both linear and exponential decay terms. The alignment coefficient increases linearly with iterations. The collision trigger probability decreases with iterations. The orbital resonance trigger probability decreases exponentially with iterations. Step S303: Perform global perturbation and multi-subgroup comprehensive guidance force calculation for subgroup collision. Calculate the mean of the average radius of all subgroups as the collision perturbation step size. When the uniform random number is less than the collision trigger probability, generate collision perturbation candidate positions by scaling and superimposing a standard normal random vector according to the collision step size and replacing the current position when the fitness is smaller. Calculate the guidance force of each subgroup on the current individual in the direction pointing to the subgroup center, with the intensity decreasing exponentially with distance. Weight the guidance force of each subgroup according to the weight obtained by exponential mapping between the individual fitness and the subgroup center fitness to obtain the comprehensive guidance force. Step S304: Perform individual position update and optimal alignment and orbital resonance local perturbation. Calculate the alignment displacement of the current individual pointing to the global optimal individual. Update the individual candidate position by combining the cosmic expansion displacement, the combined guiding force and the alignment displacement. If the fitness of the candidate position is smaller, accept the position update. Otherwise, apply a normal resonance perturbation with an amplitude base of 0.01 to the unimproved individual when the uniform random number is less than the orbital resonance trigger probability and accept it when the fitness is smaller. When the fitness of the updated individual is less than the global optimal fitness, update the global optimal individual position and the global optimal fitness simultaneously. Step S305: Perform local refinement of the global best individual and stagnation-triggered heavy-tailed escape. Perform bidirectional mode search and detection along each coordinate dimension for the global best individual. If the detection of candidate positions improves the global best fitness, the individual is accepted for replacement. If the detection of each dimension in the whole round does not improve the fitness, the detection step size is reduced by the step size contraction factor. The stagnation determination threshold is used to compare the cumulative decrease in global best fitness between adjacent generations or to clear the stagnation count. When the stagnation count reaches the stagnation window length, Cauchy distribution heavy-tailed perturbation is applied to the top few elite individuals whose number after sorting is equal to the number of elite individuals participating in the escape. The individuals are accepted when their fitness is smaller. After the execution is completed, the stagnation count is set to zero. Step S306: Record convergence information and determine termination conditions. Record the global optimal fitness and global optimal individual position of the current generation into the convergence curve vector and the historical optimal solution trajectory matrix. When the number of iterations reaches the maximum number of iterations, terminate the algorithm and output the global optimal individual position, global optimal fitness, convergence curve and historical optimal solution trajectory. If the maximum number of iterations has not been reached, increment the number of iterations and return to the subgroup structure update stage of the main iteration loop to continue the next generation iteration.

Images