An Adaptive Control Method Based on a High-Inertia Fan

Abstract

The present invention discloses an adaptive control method based on a large-inertia fan, belonging to the technical field of PID control optimization. The specific steps are as follows: S1. Construct a large-inertia fan control system; S2. Improve the mathematical model in the standard tornado optimization algorithm; S3. Use the improved tornado optimization algorithm to optimize the control parameters of the speed-loop PID controller in the large-inertia fan control system, and obtain a set of optimal PID control parameters Kp, Ki, and Kd through the optimization of the algorithm; S4. Apply the set of optimal PID control parameters obtained in S3 to the speed-loop PID controller in the large-inertia fan control system, output the corresponding control quantity, and realize the speed control of the large-inertia fan control system; By optimizing the speed-loop PID controller in the large-inertia fan control system with the improved tornado optimization algorithm, the adaptive ability of the PID controller is enhanced, the steady-state error during the operation of the fan is effectively reduced, the control accuracy is improved, and the robustness of the large-inertia fan control system is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of PID control optimization, and particularly relates to an adaptive control method based on a large-inertia fan. Background Art

[0002] The core characteristics of a large-inertia fan include high rotational inertia. Its impeller or rotor has a large mass and a long radius. When starting, it needs to overcome a huge inertia, and the acceleration time is relatively long. However, large-inertia fans have high stability and reliability. Their high-strength materials and anti-corrosion technologies can better adapt to harsh environments. They are usually controlled by asynchronous motors and are mainly concentrated in high-energy-consuming and large-scale industrial scenarios. In the energy field, they are mainly used in offshore wind farms. In the industrial field, industries such as steel, chemical, and power rely on large-inertia fans for flue gas treatment, material transportation, and cooling, etc. In terms of control, large-inertia fans have problems such as dynamic response lag, control parameter sensitivity, and high control complexity, which pose higher requirements for the performance of motor control. Future technological breakthroughs need to focus on intelligent control algorithms, lightweight materials, and digital twin-driven predictive maintenance to achieve a better balance among energy efficiency, lifespan, and economy.

[0003] The control of asynchronous motors usually adopts vector control FOC. Its core is to achieve decoupled control of torque and magnetic flux through a double closed-loop strategy of current and speed. The PID controller in the speed loop determines the dynamic response performance and steady-state accuracy of the system. Especially in scenarios with complex loads such as large-inertia fans, the design and optimization of its parameters are crucial. The PID controller is one of the most classic and widely used control algorithms in the industrial control field. Its core idea is to dynamically adjust the error of the system through the combination of three links: proportional (Kp), integral (Ki), and derivative (Kd). However, the design of traditional fixed parameters faces the disadvantages of slow dynamic response, weak anti-interference ability, and low control accuracy in the scenario of large-inertia fans. When facing a complex control environment or parameter changes due to equipment aging, traditional PID control is difficult to adapt to the changes, which may lead to problems such as large control errors and unstable performance. By introducing an intelligent optimization algorithm to adaptively adjust and optimize the parameters, the adaptability and performance of the control system can be effectively improved.

[0004] The Tornado Optimization Algorithm (TOC) is a new type of physically inspired intelligent optimization algorithm aimed at solving global optimization and constrained optimization problems. The inspiration for this algorithm comes from the formation and development process of tornadoes, combining natural phenomena such as storm airflow and Coriolis force, and forming an optimization search behavior through the interaction between individuals. TOC performs excellently in multiple engineering optimization problems. Compared with other optimization algorithms, it has better global optimization ability and stability. Summary of the Invention

[0005] The present invention aims to solve the problems that traditional control methods are difficult to cope with the response lag and poor adaptability caused by the large moment of inertia in the control system of high-inertia fans, as well as the problems that manual parameter tuning of traditional PID controllers is complex and difficult to optimize accurately. For this purpose, an adaptive control method based on high-inertia fans is proposed. By improving the tornado optimization algorithm, the parameters of the speed-loop PID controller in the high-inertia fan control system are optimized. This method can enhance the adaptive ability of the system, adapt to changes in different working conditions, effectively reduce the steady-state error during the operation of the fan, improve the control accuracy. With the optimized optimal PID control parameters, the fan can quickly respond to the target speed command, shorten the adjustment time of the system, and quickly reach the set target speed, thereby improving the overall dynamic performance and control stability of the system.

[0006] An adaptive control method based on high-inertia fans is as follows.

[0007] S1. Construct a high-inertia fan control system, which includes: a speed-loop PID controller, an improved tornado optimization algorithm, a q-axis current-loop PID controller, a d-axis current-loop PID controller, an IPARK transformation, a PARK transformation, a CLARK transformation, an SVPWM, a three-phase inverter, an AC asynchronous motor, a high-inertia fan, a flux observer, and a speed estimator.

[0008] S2. Improve the mathematical model in the standard tornado optimization algorithm, including two improvements, specifically as follows:

[0009] S21. Use a multi-stage adaptive search and perturbation enhancement strategy to improve the mathematical model of the stage where the storm evolves into a thunderstorm in the tornado optimization algorithm. This strategy is divided into two stages according to the iterative process of the algorithm. In the early stage of the algorithm, an adaptive memory factor λ1, λ2 and a reverse learning strategy are incorporated to update the population position. In the later stage of the algorithm, an adaptive inertia weight ψ(iter) and an adaptive Gaussian perturbation are incorporated to update the population position.

[0010] S22. Use a dynamic fitness value balance strategy that integrates population diversity to improve the Coriolis force factor f of the storm speed in the tornado optimization algorithm. This strategy dynamically adjusts the size of the factor according to the population diversity and the change of the fitness value of the current individual.

[0011] S3. Use the improved tornado optimization algorithm to optimize the control parameters of the speed-loop PID controller in the high-inertia fan control system, and obtain a set of optimal PID control parameters Kp, Ki, Kd through the optimization of the algorithm.

[0012] S4. Apply the set of optimal PID control parameters obtained in S3 to the speed-loop PID controller in the large-inertia fan control system, output the corresponding control quantity, and achieve the speed control of the large-inertia fan control system.

[0013] Preferably, in S1, the overall control process of the constructed large-inertia fan control system is as follows: Calculate the error between the given target speed and the actual speed obtained through the speed estimator to obtain the real-time error e(t). Input the real-time error e(t) into the speed-loop PID controller, optimize the parameters of the PID controller through the improved tornado optimization algorithm, apply the optimized parameters to the speed-loop PID controller, and give the corresponding control quantity U(t) according to the speed error. Use the q-axis current-loop PID controller and the d-axis current-loop PID controller to adjust the currents of the dq axes in the rotating coordinate system according to the q-axis current reference value and the d-axis current loop respectively, output the corresponding voltage commands under the dq coordinate axes, convert the dq-axis voltage commands into the αβ-axis voltage in the stationary coordinate system through the IPARK transformation, input the αβ-axis voltage into the SVPWM module, control the switches of the three-phase inverter through SVPWM to output three-phase alternating current to control the rotation of the AC asynchronous motor, and at the same time drive the rotation of the large-inertia fan. Collect the three-phase current of the motor in real time, convert the three-phase current of the motor into the αβ-axis current in the stationary coordinate system through the CLARK transformation, convert the αβ-axis current into the dq-axis in the rotating coordinate system through the PARK transformation, and feedback the dq-axis current to the corresponding current-loop PID controller. At the same time, obtain the magnetic flux signal, electrical angle, and real-time speed of the motor in real time through the magnetic flux observer and the speed estimator. The speed loop does not directly control the speed of the motor, but adjusts the q-axis current of the motor, and the q-axis current determines the electromagnetic torque of the motor. The change of speed is indirectly realized by adjusting the torque. The control of the large-inertia fan is achieved by the motor outputting electromagnetic torque to overcome the load torque of the fan, so that the fan reaches the target speed. The mathematical model of the speed-loop control is as follows:

[0014] (1);

[0015] In formula (1), J represents the moment of inertia of the system, K represents the gain coefficient, U(t) represents the control quantity as shown in formula (2), TL represents the load torque of the fan as shown in formula (3), B represents the damping coefficient, and ω represents the speed;

[0016] (2);

[0017] In formula (2), Kp represents the proportional gain, Ki represents the integral gain, Kd represents the derivative gain, and e(t) represents the real-time error;

[0018] (3);

[0019] In Equation (3), Cm represents the torque coefficient of the fan, ρ represents the air density, A represents the swept area of the fan blades, R represents the radius of the fan blades, ω represents the rotational speed, Bf represents the friction coefficient, M represents the mass of the fan blades, g represents the acceleration due to gravity, rg represents the eccentricity distance of the blade centroid relative to the rotation axis, and Co represents the additional mechanical torque of the system.

[0020] Preferably, in S21, a multi-stage adaptive search and perturbation enhancement strategy is used to improve the mathematical model in the stage where the storm evolves into a thunderstorm in the tornado optimization algorithm. The specific improved mathematical model is:

[0021] (4);

[0022] In Equation (4), y(iter + 1) represents the position of the updated individual, y(iter) represents the position of the current individual, iter represents the current iteration number, max_iter represents the maximum iteration number, rand represents a random number between [0, 1], represents the position of the individual with the second-best population fitness value, represents the position of the individual with the best fitness value in the population, represents the reverse learning position of the current individual, , lb represents the lower bound of the search space, ub represents the upper bound of the search space, λ1, λ2, λ3 represent adaptive memory factors, and the values of λ1 and λ2 are , and the value of λ3 is , ψ(iter) represents the adaptive inertia weight, , represents the adaptive Gaussian distribution, represents the adaptive standard deviation factor, and the calculation formula is as shown in Equation (5);

[0023] (5);

[0024] In Equation (5), represents the maximum standard deviation, represents the minimum standard deviation, iter represents the current iteration number, max_iter represents the maximum iteration number, represents the threshold for determining being trapped in a local optimum. When the algorithm starts the initial value of is 0. When the absolute value of the difference between the current optimal position in the population and the optimal position in the previous iteration is less than 0.01, the value of is incremented by one. When If the value is greater than (max_iter / 10)+1, it indicates that the algorithm has fallen into a local optimal solution. Then, Gaussian distribution perturbation is performed using the largest standard deviation. When the change range of the optimal position in the population exceeds the minimum value of the change, the value is cleared.

[0025] Preferably, a multi-stage adaptive search and perturbation enhancement strategy combines multiple adaptive adjustment mechanisms. By performing segmented execution in two stages, it can not only ensure good global search ability in the early stage of the algorithm but also ensure the local search ability of the algorithm in the later stage. By adding a reverse learning mechanism, the search range in the early stage of the algorithm can be increased, and the diversity of search individuals can be improved. In the second half of the algorithm iteration, adaptive Gaussian distribution is added for perturbation, which can improve the adaptability of the algorithm. When the algorithm stagnates in the local optimal solution, the update of the population is adjusted through Gaussian perturbation, and then it can quickly jump out to improve the optimization accuracy of the algorithm.

[0026] Preferably, in S22, a dynamic fitness value balance strategy that integrates population diversity is used to improve the Coriolis force factor f of the storm speed in the tornado optimization algorithm. The specific mathematical model after improvement is:

[0027] (6);

[0028] In formula (6), f represents the Coriolis force factor, rand represents a random number between [0, 1], β represents the perturbation amplitude factor, and its value range is [0.5, 2]. represents population diversity, and its calculation formula is shown in formula (7). represents the maximum diversity. lb represents the lower limit of the search space, ub represents the upper limit of the search space, and dim represents the problem dimension. represents the change weight of the individual fitness value, and its calculation formula is shown in formula (8);

[0029] (7);

[0030] In formula (7), N represents the population size, y(iter) represents the position of the current individual. represents the position of the individual with the best fitness value in the population;

[0031] (8);

[0032] In formula (8), fy represents the fitness value of the current individual, fbest represents the fitness value of the best individual in the population, and fworst represents the worst fitness value in the population.

[0033] Preferably, a dynamic fitness value balancing strategy that integrates population diversity combines the information of individual fitness values and population diversity. By dynamically adjusting the magnitude of the Coriolis force factor, it improves the global search ability of the algorithm in the initial stage, avoids premature convergence, improves the search accuracy of the algorithm in the later stage, and speeds up the convergence rate. By adjusting the value of f to optimize the change of the storm speed, the search process of the algorithm becomes more stable and the adaptability becomes stronger.

[0034] Preferably, in S3, an improved tornado optimization algorithm is used to optimize the control parameters of the speed loop PID controller in the large inertia fan control system. The specific steps are as follows:

[0035] S31. Initialize the parameters of the improved tornado optimization algorithm. The initialized parameters include: population size N, problem dimension dim, maximum number of iterations max_iter, upper bound ub of the search space, lower bound lb of the search space, and generate the individual positions in the initial population through the initialized parameters;

[0036] S32. Establish a mapping relationship between the improved tornado optimization algorithm and the speed loop PID controller in the large inertia fan control system. Convert the change of individual positions in the algorithm optimization process into the process of tuning the PID controller parameters. The fitness of the individual positions is calculated according to the error value of the actual control effect of the control system. The specific mapping relationship is: the position of the current individual y(iter)=[y1, y2, y3], and the PID control parameters are [Kp, Ki, Kd]. The value y1 of the individual in the first dimension corresponds to Kp, the value y2 in the second dimension corresponds to Ki, and the value y3 in the third dimension corresponds to Kd. Adjust the position of the individual y(iter) through the iteration update of the algorithm, so as to tune the parameters of the PID controller and finally obtain a set of optimal parameter combinations;

[0037] S33. Set the fitness value function of the improved tornado optimization algorithm. Calculate the fitness values of the individuals in the initial population through the fitness value function, and sort and select the positions of the top two individuals and the worst individual in terms of fitness value and the corresponding fitness value magnitudes. The specific fitness value function formula is as follows:

[0038] (9);

[0039] In formula (9), J represents the fitness value, T represents the total running time of the system, and e(t) represents the real-time speed error value calculated from the given target speed and the real-time speed obtained by the speed estimator;

[0040] S34. Update the positions of the individuals in the population through the mathematical model of the improved tornado optimization algorithm. The specific steps are as follows:

[0041] step1. Calculate the storm velocity V(iter) of the improved tornado optimization algorithm. The specific formula is as follows:

[0042] (10);

[0043] In formula (10), V(iter + 1) represents the updated storm velocity, η represents the contraction factor, τ represents the adaptive momentum, f represents the Coriolis force factor, and its calculation formula is as shown in formula (6). c represents a random number within a range, Rl represents the radius of curvature of the storm track in the Northern Hemisphere, Rr represents the radius of curvature of the storm track in the Southern Hemisphere, CFl represents the Coriolis force in the Northern Hemisphere, CFr represents the Coriolis force in the Southern Hemisphere, and its calculation formula is as shown in formula (11). rand represents a random number in the range of [0, 1].

[0044] step2. The improved tornado optimization algorithm enters the stage of evolving from a storm to a tornado, and updates the individual positions. The specific formula is as follows:

[0045] (11);

[0046] In formula (11), y(iter + 1) represents the updated position of the individual, y(iter) represents the current position of the individual, represents the position of the individual with the optimal fitness value in the population, α represents the evolution factor, represents the position of a randomly selected storm individual, and V(iter) represents the storm velocity, and its calculation formula is as shown in formula (10).

[0047] step3. The improved tornado optimization algorithm enters the stage of evolving from a storm to a thunderstorm, and updates the individual positions. The specific formula is as shown in formula (4).

[0048] step4. The improved tornado optimization algorithm enters the stage of evolving from a thunderstorm to a tornado, and updates the individual positions. The specific formula is as follows:

[0049] (12);

[0050] In formula (12), y(iter + 1) represents the updated position of the individual, y(iter) represents the current position of the individual, represents the position of the individual with the optimal fitness value in the population, α represents the evolution factor, represents the position of a randomly selected storm individual;

[0051] step5. The improved tornado optimization algorithm enters the stage of randomly generating storms, and updates the individual positions. The specific formula is as follows:

[0052] (13);

[0053] In Equation (12), y(iter + 1) represents the position of the updated individual, y(iter) represents the position of the current individual, ay represents the adaptability parameter, lb represents the lower bound of the search space, ub represents the upper bound of the search space, rand represents a random number in [0, 1], σ2 represents that the sign change factor is related to the update direction, and ||y(iter) - || represents the Euclidean distance between the current storm individual and the tornado individual, where the tornado individual is the position of the optimal individual in the population, and v represents an exponential parameter;

[0054] Step 6: Calculate the fitness value of the updated individual, and update the optimal position, the position with the second-best fitness value, and the worst position in the population;

[0055] S35: Determine whether the current iteration number has reached the maximum iteration number max_iter. If it has not reached the maximum iteration number, continue to execute S34 for algorithm optimization. If it has reached the maximum iteration number, end the algorithm optimization, and output the position of the optimal individual in the population as the optimal solution.

[0056] By adopting the above technical solutions, the advantages of the present invention are as follows: Combining the multi-stage adaptive search and perturbation enhancement strategy, and the dynamic fitness value balance strategy that integrates population diversity, the mathematical model of the tornado optimization algorithm is improved. This improvement significantly enhances the adaptability and optimization accuracy of the algorithm. The improved tornado optimization algorithm can more effectively cope with the application scenario of large-inertia fans driven by AC asynchronous motors. In the large-inertia fan control system, using the improved algorithm to optimize the parameters of the speed loop PID controller not only improves the response speed and stability of the speed loop, but also helps the fan quickly reach the target speed and maintain stable operation, significantly reducing the influence of errors in the adjustment process on the control system. Description of the Drawings

[0057] Figure 1 It is a flowchart of an adaptive control method based on a large-inertia fan.

[0058] Figure 2 It is a model diagram of a large-inertia fan control system.

[0059] Figure 3 It is a comparison chart of the responses of the standard tornado optimization algorithm and the improved tornado optimization algorithm for optimizing the speed loop PID controller.

[0060] Figure 4 It is a flowchart of the improved tornado optimization algorithm for optimizing the speed loop PID controller.

[0061] Figure 5It is a comparison chart of the fitness value changes during the optimization process of the speed loop PID controller optimized by the standard tornado optimization algorithm and the improved tornado optimization algorithm. Specific implementation mode

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0063] The present invention provides a technical solution: an adaptive control method based on a large inertia fan, which specifically includes the following steps, as Figure 1 shown.

[0064] S1. Construct a large inertia fan control system, and the control system includes: a speed loop PID controller, an improved tornado optimization algorithm, a q-axis current loop PID controller, a d-axis current loop PID controller, an IPARK transformation, a PARK transformation, a CLARK transformation, an SVPWM, a three-phase inverter, an AC asynchronous motor, a large inertia fan, a flux observer, and a speed estimator, as Figure 2 shown.

[0065] Further, in S1, the overall control process of the constructed large-inertia fan control system is as follows: Calculate the error between the given target speed and the actual speed obtained through the speed estimator to obtain the real-time error e(t). Input the real-time error e(t) into the speed-loop PID controller, optimize the parameters of the PID controller through the improved tornado optimization algorithm, apply the optimized parameters to the speed-loop PID controller, and give the corresponding control quantity U(t) according to the speed error. Use the q-axis current-loop PID controller and the d-axis current-loop PID controller to adjust the currents of the dq axes in the rotating coordinate system according to the q-axis current reference value and the d-axis current loop respectively, and output the voltage commands under the corresponding dq coordinate axes. Convert the dq-axis voltage commands into the αβ-axis voltages in the stationary coordinate system through the IPARK transformation, input the αβ-axis voltages into the SVPWM module, and control the switches of the three-phase inverter through SVPWM to output three-phase alternating current to control the rotation of the AC asynchronous motor, and at the same time drive the rotation of the large-inertia fan. Collect the three-phase currents of the motor in real time, convert the three-phase currents of the motor into the αβ-axis currents in the stationary coordinate system through the CLARK transformation, convert the αβ-axis currents into the dq-axis currents in the rotating coordinate system through the PARK transformation, and feedback the dq-axis currents to the corresponding current-loop PID controllers. At the same time, obtain the magnetic flux signal, electrical angle and real-time speed of the motor in real time through the magnetic flux observer and the speed estimator. The speed loop does not directly control the speed of the motor, but adjusts the q-axis current of the motor, and the q-axis current determines the electromagnetic torque of the motor. The change of speed is indirectly realized by adjusting the torque. The control of the large-inertia fan is to overcome the load torque of the fan through the electromagnetic torque output by the motor, so that the fan reaches the target speed. The mathematical model of the speed-loop control is as follows:

[0066] (1);

[0067] In formula (1), J represents the moment of inertia of the system, K represents the gain coefficient, U(t) represents the control quantity as shown in formula (2), TL represents the load torque of the fan as shown in formula (3), B represents the damping coefficient, and ω represents the speed;

[0068] (2);

[0069] In formula (2), Kp represents the proportional gain, Ki represents the integral gain, Kd represents the derivative gain, and e(t) represents the real-time error;

[0070] (3);

[0071] In Equation (3), Cm represents the torque coefficient of the fan, ρ represents the air density, A represents the swept area of the fan blades, R represents the radius of the fan blades, ω represents the rotational speed, Bf represents the friction coefficient, M represents the mass of the fan blades, g represents the acceleration due to gravity, rg represents the eccentricity distance of the blade centroid relative to the rotation axis, and Co represents the additional mechanical torque of the system.

[0072] S2. Improve the mathematical model in the standard tornado optimization algorithm, including two improvements, specifically as follows:

[0073] S21. Use a multi-stage adaptive search and perturbation enhancement strategy to improve the mathematical model in the stage where the storm evolves into a thunderstorm in the tornado optimization algorithm. This strategy is divided into two stages according to the iterative process of the algorithm. In the early stage of the algorithm, an adaptive memory factor λ1, λ2 and a reverse learning strategy are incorporated to update the population position. In the later stage of the algorithm, an adaptive inertia weight ψ(iter) and an adaptive Gaussian perturbation are incorporated to update the population position;

[0074] S22. Use a dynamic fitness value balance strategy that integrates population diversity to improve the Coriolis force factor f of the storm speed in the tornado optimization algorithm. This strategy dynamically adjusts the size of the factor according to the population diversity and the change of the fitness value of the current individual.

[0075] Furthermore, in S21, use a multi-stage adaptive search and perturbation enhancement strategy to improve the mathematical model in the stage where the storm evolves into a thunderstorm in the tornado optimization algorithm. The specific improved mathematical model is:

[0076] (4);

[0077] In Equation (4), y(iter + 1) represents the position of the updated individual, y(iter) represents the position of the current individual, iter represents the current iteration number, max_iter represents the maximum iteration number, rand represents a random number between [0, 1], represents the position of the individual with the second-best fitness value in the population, represents the position of the individual with the best fitness value in the population, represents the reverse learning position of the current individual, , lb represents the lower bound of the search space, ub represents the upper bound of the search space, λ1, λ2, λ3 represent adaptive memory factors, and the values of λ1 and λ2 are , and the value of λ3 is , ψ(iter) represents the adaptive inertia weight, , represents the adaptive Gaussian distribution, represents the adaptive standard deviation factor, and the calculation formula is as shown in Equation (5);

[0078] (5);

[0079] In formula (5), represents the maximum value of the standard deviation, represents the minimum value of the standard deviation, iter represents the current iteration number, max_iter represents the maximum iteration number, represents the threshold for judging being trapped in a local optimum. When the algorithm starts its initial value is 0. When the absolute value of the difference between the current optimal position in the population and the optimal position in the previous iteration is less than 0.01, the value of is incremented by one. When the value of is greater than (max_iter / 10)+1, it indicates that the algorithm is trapped in a local optimal solution. Then, Gaussian distribution perturbation is performed using the maximum standard deviation. When the change range of the optimal position in the population exceeds the minimum value of the change, the value of

[0080] is cleared. Furthermore, in S22, a dynamic fitness value balance strategy that integrates population diversity is used to improve the Coriolis force factor f of the storm speed in the tornado optimization algorithm. The specific mathematical model after improvement is:

[0081] (6);

[0082] In formula (6), f represents the Coriolis force factor, rand represents a random number between [0, 1], β represents the perturbation amplitude factor, and its value range is [0.5, 2]. represents the population diversity, and its calculation formula is as shown in formula (7). represents the maximum diversity, lb represents the lower limit of the search space, ub represents the upper limit of the search space, dim represents the problem dimension, represents the individual fitness value change weight, and its calculation formula is as shown in formula (8);

[0083] (7);

[0084] In formula (7), N represents the population size, y(iter) represents the position of the current individual, represents the position of the individual with the optimal fitness value in the population;

[0085] (8);

[0086] In formula (8), fy represents the fitness value of the current individual, fbest represents the fitness value of the optimal individual in the population, and fworst represents the worst fitness value in the population.

[0087] S3. Optimize the control parameters of the speed-loop PID controller in the large-inertia fan control system using the improved tornado optimization algorithm, and obtain a set of optimal PID control parameters Kp, Ki, and Kd through the optimization of the algorithm.

[0088] Further, in the above S3, when using the improved tornado optimization algorithm to optimize the control parameters of the speed-loop PID controller in the large-inertia fan control system, as Figure 4 shown, the specific steps are as follows:

[0089] S31. Initialize the parameters of the improved tornado optimization algorithm. The initialized parameters include: population size N, problem dimension dim, maximum number of iterations max_iter, upper bound ub of the search space, lower bound lb of the search space, and generate the individual positions in the initial population through the initialized parameters.

[0090] S32. Establish a mapping relationship between the improved tornado optimization algorithm and the speed-loop PID controller in the large-inertia fan control system, convert the change of individual positions in the algorithm optimization process into the process of tuning the PID controller parameters, and calculate the fitness value of the individual positions according to the error value of the actual control effect of the control system. The specific mapping relationship is: the position of the current individual y(iter)=[y1, y2, y3], the PID control parameters are [Kp, Ki, Kd], the value y1 of the individual in the first dimension corresponds to Kp, the value y2 in the second dimension corresponds to Ki, and the value y3 in the third dimension corresponds to Kd. Adjust the position of the individual y(iter) through the iteration update of the algorithm, so as to tune the parameters of the PID controller and finally obtain a set of optimal parameter combinations.

[0091] S33. Set the fitness value function of the improved tornado optimization algorithm, calculate the fitness values of the individuals in the initial population through the fitness value function, and sort and select the positions of the top two individuals and the worst individual in terms of fitness value and the corresponding fitness value magnitudes. The specific formula of the fitness value function is as follows:

[0092] (9);

[0093] In formula (9), J represents the fitness value, T represents the total running time of the system, and e(t) represents the real-time speed error value calculated from the given target speed and the real-time speed obtained by the speed estimator.

[0094] S34. Update the positions of the individuals in the population through the mathematical model of the improved tornado optimization algorithm. The specific steps are as follows:

[0095] step1. Calculate the storm speed V(iter) of the improved tornado optimization algorithm. The specific formula is:

[0096] (10);

[0097] In formula (10), V(iter + 1) represents the updated storm speed, η represents the contraction factor, τ represents the adaptive momentum, f represents the Coriolis force factor, and its calculation formula is as shown in formula (6). c represents a random number within a range, Rl represents the radius of curvature of the storm track in the Northern Hemisphere, Rr represents the radius of curvature of the storm track in the Southern Hemisphere, CFl represents the Coriolis force in the Northern Hemisphere, CFr represents the Coriolis force in the Southern Hemisphere, and its calculation formula is as shown in formula (11). rand represents a random number in the range of [0, 1];

[0098] Step 2: The improved tornado optimization algorithm enters the stage of storm to tornado evolution, and updates the individual positions. The specific formula is as follows:

[0099] (11);

[0100] In formula (11), y(iter + 1) represents the updated position of the individual, y(iter) represents the current position of the individual, represents the position of the individual with the optimal fitness value in the population, α represents the evolution factor, represents the position of a randomly selected storm individual, and V(iter) represents the storm speed, and its calculation formula is as shown in formula (10);

[0101] Step 3: The improved tornado optimization algorithm enters the stage of storm evolving into thunderstorm, and updates the individual positions. The specific formula is as shown in formula (4);

[0102] Step 4: The improved tornado optimization algorithm enters the stage of thunderstorm evolving into tornado, and updates the individual positions. The specific formula is as follows:

[0103] (12);

[0104] In formula (12), y(iter + 1) represents the updated position of the individual, y(iter) represents the current position of the individual, represents the position of the individual with the optimal fitness value in the population, α represents the evolution factor, represents the position of a randomly selected storm individual;

[0105] Step 5: The improved tornado optimization algorithm enters the stage of random generation of storms, and updates the individual positions. The specific formula is as follows:

[0106] (13);

[0107] In Equation (12), y(iter + 1) represents the position of the updated individual, y(iter) represents the position of the current individual, ay represents the adaptability parameter, lb represents the lower limit of the search space, ub represents the upper limit of the search space, rand represents a random number in [0, 1], σ2 represents that the sign change factor is related to the update direction, ||y(iter)- || represents the Euclidean distance between the current storm individual and the tornado individual, where the tornado individual is the position of the optimal individual in the population, and v represents an exponential parameter;

[0108] Step 6: Calculate the fitness of the updated individual, and update the optimal position, the position with the second-best fitness value, and the worst position in the population;

[0109] S35: Determine whether the current iteration number has reached the maximum iteration number max_iter. If it has not reached the maximum iteration number, continue to execute S34 for algorithm optimization. If it has reached the maximum iteration number, end the algorithm optimization and output the position of the optimal individual in the population as the optimal solution.

[0110] S4: Apply a set of optimal PID control parameters obtained in S3 to the speed-loop PID controller in the large-inertia fan control system, output the corresponding control quantity, and achieve the speed control of the large-inertia fan control system.

[0111] Furthermore, to verify the superiority of the present invention, Matlab and Simulink are used for simulation experiments. First, the mathematical model formula of the standard tornado optimization algorithm is improved through Matlab. The improved algorithm is defined as the function ITOC(N, max_iter, lb, ub, dim, fobj), and the algorithm before improvement is defined as TOC(N, max_iter, lb, ub, dim, fobj), where N represents the population size, max_iter represents the maximum iteration number, lb represents the lower limit of the search space, ub represents the upper limit of the search space, dim represents the problem dimension, fobj represents the fitness value calculation function, and fobj is designed with the corresponding code program according to the fitness value formula in Equation (9). The mathematical model of the speed-loop control is Laplace-transformed into the corresponding controlled object function , s represents the variable in the complex frequency domain, and U(s) represents the form of the control variable U(s) in the complex frequency domain. Set the initial experimental parameters: the moment of inertia of the system J = 0.4, the gain coefficient K = 20, the damping coefficient B = 0.03, the torque coefficient of the fan Cm = 0.1, the air density ρ = 1.225, the swept area of the fan blades A = 6.28, the radius of the fan blades R = 2, the friction coefficient Bf = 2, the mass of the fan blades M = 80, the acceleration due to gravity g = 9.81, the eccentricity distance rg = 0.07, and the additional mechanical torque of the system Co = 10. Design a simulation model program through Simulink to simulate the actual operating state of the fan. Initialize the parameters in the main program. The population size N = 100, the maximum number of iterations max_iter = 20, the search space range [lb, ub] = [0.001, 100], the problem dimension dim = 3. Run the improved and unimproved algorithm functions ITOC(N, max_iter, lb, ub, dim, fobj) and TOC(N, max_iter, lb, ub, dim, fobj). The parameters obtained by the algorithm optimization each time are passed into the simulation model program for operation, and the actual rotational speed output of the model is returned. Calculate the fitness value of the parameters through the fitness value function, which is used as the evaluation criterion for the algorithm optimization, and retain the results of each iteration. The optimal control parameters obtained by the improved algorithm are Kp = 0.08979, Ki = 4.3005, Kd = 4.4905, and the optimal parameters obtained by the unimproved algorithm are Kp = 0.28795, Ki = 1.2567, Kd = 89.9962. Visualize the change in the fitness value during the algorithm optimization process and the PID response curve of the control system to obtain Figure 3 and Figure 5 .

[0112] Furthermore, Figure 3 Figure Figure 3 is a comparison chart of the responses of the standard tornado optimization algorithm and the improved tornado optimization algorithm for optimizing the speed loop PID controller. Set the target value to 1 unit. It can be seen from the figure that the response curve of the speed loop PID controller optimized by the improved tornado optimization algorithm has better stability, smaller steady-state error, and stabilizes near the target value in a shorter time. Figure 5 Figure Figure 5 is a comparison chart of the changes in the fitness values during the optimization process of the standard tornado optimization algorithm and the improved tornado optimization algorithm for optimizing the speed loop PID controller. It can be seen from the figure that the improved tornado optimization algorithm reaches near the optimal solution with fewer iterations, and the fitness value of the finally obtained optimal solution is smaller and the accuracy is higher. The optimal fitness value obtained by the standard tornado optimization algorithm is 11.1169, and the optimal fitness value obtained by the improved tornado optimization algorithm is 11.1058.

Claims

1. An adaptive control method based on a large inertia fan, characterized in that: The specific steps are as follows: S1. Construct a large inertia fan control system, the control system comprising: a speed loop PID controller, an improved tornado optimization algorithm, a q-axis current loop PID controller, a d-axis current loop PID controller, an IPARK transformation, a PARK transformation, a CLARK transformation, a SVPWM, a three-phase inverter, an AC asynchronous motor, a large inertia fan, a flux observer, and a speed estimator; S2. Improve the mathematical model in the standard tornado optimization algorithm, including two improvements, as follows: S21. A multi-stage adaptive search and disturbance enhancement strategy is used to improve the mathematical model of the storm-to-thunderstorm stage in the tornado optimization algorithm. The strategy is divided into two stages according to the algorithm's iterative process. In the early stage of the algorithm, the adaptive memory factor and reverse learning strategy are integrated to update the population position. In the later stage of the algorithm, the adaptive inertia weight and adaptive Gaussian perturbation are integrated to update the population position. S22. A dynamic fitness value balancing strategy integrating population diversity is used to improve the Coriolis force factor f of the storm speed in the tornado optimization algorithm. The strategy dynamically adjusts the size of the factor according to the population diversity and the fitness value change of the current individual; S3. Use the improved tornado optimization algorithm to optimize the control parameters of the speed loop PID controller in the large inertia fan control system, and obtain a set of optimal PID control parameters Kp, Ki, and Kd through algorithm optimization; S4. Use a set of optimal PID control parameters obtained in S3 for the speed loop PID controller in the large inertia fan control system, output the corresponding control quantity, and realize the speed control of the large inertia fan control system.

2. The adaptive control method based on a large inertia fan according to claim 1, characterized in that: In the above S1, the overall control process of the constructed large inertia fan control system is as follows: the error between the given target speed and the actual speed obtained by the speed estimator is calculated to obtain a real-time error e(t), the real-time error e(t) is input into the speed loop PID controller, the parameters of the PID controller are optimized by the improved tornado optimization algorithm, the optimized parameters are applied to the speed loop PID controller, and the corresponding control quantity U(t) is given according to the speed error, the control quantity U(t) is used as the reference value of the q-axis current, and the q-axis current loop PID controller and the d-axis current loop PID controller are used to respectively adjust the rotating coordinate system according to the q-axis current reference value and the d-axis current loop. The current of the dq axis is adjusted, and the corresponding voltage command of the dq axis is output. The dq axis voltage command is converted into the αβ axis voltage in the stationary coordinate system through IPARK transformation, and the αβ axis voltage is input into the SVPWM module. The switch of the three-phase inverter is controlled by SVPWM to output three-phase AC power to control the rotation of the AC asynchronous motor, and at the same time drive the rotation of the large inertia fan. The three-phase current of the motor is collected in real time, and the three-phase current of the motor is converted into the current of the αβ axis in the stationary coordinate system through CLARK transformation. The current of the αβ axis is converted to the dq axis in the rotating coordinate system through PARK transformation, and the current of the dq axis is fed back to the corresponding current loop PID controller. At the same time, the flux observer and speed estimator can obtain the motor's flux signal, electrical angle and real-time speed in real time. The speed loop does not directly control the motor's speed, but adjusts the motor's q-axis current, which determines the motor's electromagnetic torque. The speed change is indirectly achieved by adjusting the torque. The control of the large inertia fan is to overcome the fan's load torque by the motor output electromagnetic torque, so that the fan reaches the target speed. The mathematical model of the speed loop control is as follows: (1)； In formula (1), J represents the moment of inertia of the system, K represents the gain coefficient, U(t) represents the control quantity, TL represents the load torque of the fan, B represents the damping coefficient, and ω represents the speed.

3. The adaptive control method based on a large inertia fan according to claim 2, characterized in that: In S21, a multi-stage adaptive search and disturbance enhancement strategy is used to improve the mathematical model of the storm evolving into the thunderstorm stage in the tornado optimization algorithm. The specific mathematical model after the improvement is: (4)； In formula (4), y(iter+1) represents the position of the updated individual, y(iter) represents the position of the current individual, iter represents the current iteration number, max_iter represents the maximum iteration number, and rand represents a random number between [0,1]. Indicates the position of the second-ranked individual in the population fitness value, represents the position of the individual with the best fitness value in the population, represents the reverse learning position of the current individual, , lb represents the lower limit of the search space, ub represents the upper limit of the search space, λ1, λ2, λ3 represent the adaptive memory factors, and the values ​​of λ1 and λ2 are , the value of λ3 is , ψ(iter) represents the adaptive inertia weight, , represents the adaptive Gaussian distribution, represents the adaptive standard deviation factor, and the calculation formula is shown in formula (5); (5)； In formula (5), represents the maximum standard deviation, It represents the minimum standard deviation, iter represents the current number of iterations, and max_iter represents the maximum number of iterations. Indicates the threshold for judging falling into the local optimum. When the algorithm starts The initial value of is 0. When the absolute value of the difference between the current optimal position in the population and the optimal position in the previous iteration is less than 0.01, The value of is increased by one, when If the value of is greater than (max_iter / 10)+1, it means that the algorithm falls into a local optimal solution. Then the maximum standard deviation is used for Gaussian distribution perturbation. When the optimal position in the population varies beyond the minimum value of the variation, The value of is cleared to zero.

4. The adaptive control method based on a large inertia fan according to claim 3 is characterized in that: In S22, a dynamic fitness value balance strategy integrating population diversity is used to improve the Coriolis force factor f of the storm speed in the tornado optimization algorithm. The specific mathematical model after the improvement is: (6)； In formula (6), f represents the Lilioolis force factor, rand represents a random number between [0, 1], and β represents the disturbance amplitude factor, which ranges from [0.5, 2]. represents population diversity, and the calculation formula is shown in formula (7): represents the maximum diversity, lb represents the lower limit of the search space, ub represents the upper limit of the search space, and dim represents the problem dimension. represents the weight of individual fitness value change, and the calculation formula is shown in formula (8); (7)； In formula (7), N represents the population size, y(iter) represents the position of the current individual, Indicates the position of the individual with the best fitness value in the population; (8)； In formula (8), fy represents the fitness value of the current individual, fbest represents the fitness value of the best individual in the population, and fworst represents the worst fitness value in the population.

5. The adaptive control method based on a large inertia fan according to claim 3 is characterized in that: In S3, the control parameters of the speed loop PID controller in the large inertia fan control system are optimized using the improved tornado optimization algorithm, and the specific steps are as follows: S31, initializing the parameters of the improved tornado optimization algorithm, the initialization parameters include: population size N, problem dimension dim, maximum number of iterations max_iter, search space upper limit ub, search space lower limit lb, and generating individual positions in the initial population through the initialization parameters; S32. Establish a mapping relationship between the improved tornado optimization algorithm and the speed loop PID controller in the large inertia fan control system, convert the individual position change in the algorithm optimization process into the PID controller parameter tuning process, and calculate the fitness of the individual position according to the error value of the actual control effect of the control system. The specific mapping relationship is: the current individual position y(iter)=[y1, y2, y3], the PID control parameters are [Kp, Ki, Kd], the individual value y1 in the first dimension corresponds to Kp, the value y2 in the second dimension corresponds to Ki, and the value y3 in the third dimension corresponds to Kd. The position of the individual y(iter) is adjusted through the iterative update of the algorithm, so as to tune the parameters of the PID controller and finally obtain a set of optimal parameter combinations; S33, setting the fitness value function of the improved tornado optimization algorithm, calculating the fitness value of individuals in the initial population through the fitness value function, and sorting and selecting the top two individuals in fitness value and the individual position with the worst fitness value and the corresponding fitness value size. The specific fitness value function formula is as follows: (9)； In formula (9), J represents the fitness value, T represents the total system operation time, and e(t) represents the real-time speed error value calculated by the given target speed and the real-time speed obtained by the speed estimator; S34, updating the positions of individuals in the population through the mathematical model of the improved tornado optimization algorithm; S35, determine whether the current number of iterations has reached the maximum number of iterations max_iter, if not, continue to execute S34 to perform algorithm optimization, if reached, end the algorithm optimization, and output the optimal individual position in the population as the optimal solution.

Images