CPO-improved parameter setting method for active disturbance rejection controller of permanent magnet synchronous motor
By using the improved Guanhaozhu optimization algorithm to tune the parameters of the active disturbance rejection controller for permanent magnet synchronous motors, the problem of insufficient disturbance rejection capability of traditional PID control methods in permanent magnet synchronous motors is solved, and automatic debugging and high-precision control are realized.
Patent Information
- Application Number
- CN202411172891.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2026-03-03
AI Technical Summary
Traditional PID control methods are insufficient in noise immunity and have weak ability to handle time-varying parameters in permanent magnet synchronous motors. Furthermore, parameter tuning relies on the engineer's experience, making the process cumbersome.
An improved CPO (Coral Porcupine Optimization) algorithm was used to tune the parameters of the active disturbance rejection controller, including optimal point set initialization, nonlinear dynamic weight adjustment, adaptive t-distribution disturbance mutation, and cyclic population technique. A specific fitness function was designed to optimize the control parameters.
Automatic tuning of the active disturbance rejection controller parameters was achieved, which improved control accuracy and optimization convergence speed, reduced labor costs, and achieved better control results.
Smart Images
Figure CN121602880A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of active disturbance rejection controller parameter tuning, specifically an improved CPO (Continuous Disturbance Rejection Controller) active disturbance rejection controller parameter tuning method for permanent magnet synchronous motors. Background Technology
[0002] Permanent magnet synchronous motors (PMSMs) are widely used in various industrial fields due to their high efficiency, small size, and low maintenance costs. PMSMs are complex nonlinear systems characterized by time-varying parameters and variations in internal and external disturbances. Currently, the traditional proportional-derivative-integral (PID) control method, while simple to implement, suffers from insufficient disturbance rejection capability, weak response to time-varying parameters, and a trade-off between fast response and overshoot. Therefore, research into advanced control algorithms is of great significance for achieving better control performance of PMSMs. Addressing the shortcomings of PID control, Han Jingqing et al. proposed Active Disturbance Rejection Control (ADRC) technology. ADRC, because it does not rely on the internal mechanisms and external disturbance laws of the controlled object, estimates the total disturbance of the system in real time and provides timely active compensation, exhibiting strong disturbance rejection capability, high accuracy, and fast response speed. It has been widely researched and applied, achieving significant results in the field of PMSM control. ADRC controllers have numerous parameters, typically requiring manual optimization by experienced engineers, a cumbersome process. Automatic ADRC parameter optimization technology enables intelligent tuning of controller parameters, significantly improving control system performance. Summary of the Invention
[0003] The active disturbance rejection controller (ADRC) for permanent magnet synchronous motors has numerous parameters, and the parameter tuning process is quite complicated, heavily reliant on engineers with extensive experience. This invention proposes a parameter tuning method for the ADRC of permanent magnet synchronous motors, enabling automatic tuning of control parameters to achieve optimal control performance.
[0004] The technical solution adopted by the present invention to achieve the above objectives is as follows:
[0005] An improved CPO (Continuous Disturbance Rejection Controller) parameter tuning method for permanent magnet synchronous motors includes the following steps:
[0006] 1) Initialize the control parameters of the CPO algorithm, initialize the population using the set of best points, and record the initial global optimal solution and the historical optimal solution of each individual;
[0007] 2) Perform nonlinear dynamic weight adjustment;
[0008] 3) Update individual positions based on the four defense mechanisms of the CPO algorithm;
[0009] 4) Adaptive t-distribution perturbation variation is applied to individuals;
[0010] 5) Construct a fitness function, use the value of the fitness function to evaluate the control effect of the current active disturbance rejection controller, and retain the historical best solution and the current global best solution of the population based on the value of the fitness function;
[0011] 6) Use cyclic population technology to dynamically adjust the population size based on the current search status;
[0012] 7) Determine if the termination condition has been met. If the termination condition has been met, retain the current global optimal solution and use it to update the parameters of the active disturbance rejection controller; otherwise, return to step 2.
[0013] The nonlinear dynamic weight adjustment is specifically as follows:
[0014]
[0015] Among them, T max α is the maximum number of iterations; t is the current iteration number; α(t) is the weight; and e is the natural constant.
[0016] The adaptive t-distribution perturbation variation of the individual specifically refers to:
[0017] x i,new =x i +f(x i ,v(t))
[0018] Where, x i,new For the new position after the perturbation, x i f(x) represents the current position of the individual. i ,v(t)) represents the adaptive t-distribution perturbation variation.
[0019] The fitness function is:
[0020]
[0021] Where T is the maximum running time, e(t) s ) represents the error present at the current time, θ(t) s ) represents the current location at the current time, θ * Given the target location.
[0022] The cyclic population technique specifically refers to:
[0023]
[0024] Among them, Tc The variable that determines the number of iterations; t is the current iteration number; T max N is the maximum number of iterations. min It is the minimum number of individuals in the new population, and N is the population size.
[0025] An improved CPO (Continuous Disturbance Rejection Controller) parameter tuning system for a permanent magnet synchronous motor includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the improved CPO parameter tuning method for a permanent magnet synchronous motor when the computer program is executed.
[0026] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the improved CPO permanent magnet synchronous motor active disturbance rejection controller parameter tuning method.
[0027] The present invention has the following beneficial effects and advantages:
[0028] 1. Compared with manual debugging methods, this method can realize automatic debugging of the control system, save labor costs, and achieve high control accuracy.
[0029] 2. Compared with the original CPO algorithm, the improved CPO algorithm in this paper has significantly improved the optimization convergence speed and convergence accuracy. Finally, when applied to the ADRC controller parameter tuning problem, the improved CPO can achieve better control performance. Attached Figure Description
[0030] Figure 1 Structure diagram of active disturbance rejection controller;
[0031] Figure 2 A control system block diagram for an improved CPO algorithm is proposed. Detailed Implementation
[0032] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.
[0033] This invention proposes a parameter tuning method for the active disturbance rejection controller (ADRC) of a permanent magnet synchronous motor (PMSM), which can automatically tune the control parameters to achieve optimal control performance. The Cinque Pig optimization algorithm is prone to getting trapped in local optima and its convergence speed needs improvement. We have made the following improvements: First, we use a best-point set to optimize the population initialization of individual position distribution, making the population distribution more uniform and increasing the early global search capability and stability. Second, we introduce nonlinear dynamic weights to better balance global and local searches. Finally, we introduce adaptive t-distribution disturbance mutation to increase development capability and avoid getting trapped in local optima. Regarding the parameter tuning problem of the ADRC in the PMSM control system, we have designed a specific fitness function for this problem, mainly considering the total system error and maximum overshoot. The optimal control parameters are obtained by inputting control parameters and evaluating the fitness function, and these optimal control parameters are retained for system control tasks.
[0034] The specific implementation steps of this invention are as follows:
[0035] Stage 1: Initialize the control parameters of the improved CPO algorithm (including population size, maximum number of iterations, balance factor, minimum population), and initialize the initial global optimal solution and the historical optimal solution of each individual;
[0036] Stage 2: Perform non-linear dynamic weight adjustment. As the algorithm progresses, the individual movement step size is gradually changed to balance global and local search capabilities. The weight adjustment expression is:
[0037]
[0038] In the formula, T max t is the maximum number of iterations; e is the current iteration algebra; in the early stage of iteration, the weights are larger and the individual movement step size is larger, resulting in a strong global search capability in the search space. In the later stage of iteration, the weights become smaller and the individual movement step size is smaller, resulting in a strong local development capability.
[0039] Stage 3: Update individual positions based on the four defense mechanisms of the original algorithm.
[0040] Stage 4: Adaptive t-distribution perturbation mutation is applied to individuals to prevent them from getting trapped in local optima.
[0041] Stage 5: Using the designed fitness function, evaluate the performance of the current individual parameter control effect, and retain the individual's historical best solution and the current global best solution.
[0042] Stage 6: Use cyclic population technology to dynamically adjust the population size based on the current search status.
[0043] Stage 7: Determine if the termination condition has been met. If the termination condition has been met, retain the current best result for direct use next time; otherwise, execute Stage 2.
[0044] An improved CPO optimization method, with the following main improvements:
[0045] S1. In the standard porcupine optimization algorithm, the population is initialized randomly. This method is prone to causing multiple individuals to cluster together, resulting in reduced population diversity and poor algorithm stability. This paper uses the optimal point set method to initialize the population in order to improve the uniformity of population distribution and algorithm stability.
[0046] S2. In the fourth defense mechanism of the original algorithm, the parameter α controls the individual's movement step size. The CPO algorithm, which uses a fixed speed, cannot adapt to the needs of different stages of the algorithm. This paper adopts a nonlinear dynamic weighting method to balance the global search capability and the local search capability.
[0047] S3. This paper adopts an adaptive degree-of-freedom parameter adjustment method. The adaptive t-distribution perturbation mutation dynamically adjusts the degree-of-freedom parameters, enabling the mutation process to adapt adaptively, thereby improving the diversity and exploration capability of the optimization algorithm. As the number of iterations increases, the degree-of-freedom parameters increase linearly to prevent the algorithm from getting trapped in local optima in the later stages.
[0048] like Figure 1 As shown, the active disturbance rejection controller consists of three parts: a differential tracker (DT), a nonlinear state error feedback (NLSEF), and an extended state observer (ESO).
[0049] Differential tracker: Responsible for arranging the transient response, providing appropriate control signals, and resolving the trade-off between response speed and overshoot. The expression is...
[0050]
[0051] In the formula, h0 is the filtering factor; r is the velocity factor, the value of which determines the approximation speed; fhan is a nonlinear function, the specific form of which is as follows:
[0052]
[0053] Where x1 and x2 are inputs, and f1 is the output; the expressions for a and d are as follows:
[0054]
[0055] Extended State Observer: The extended state observer is used to estimate the system's state and total disturbance in real time. ESO treats the system's disturbance as a new state variable, constructs an extended state-space model, and estimates the system's state and disturbance in real time by observing the system's input-output relationship. The expression is:
[0056]
[0057] In the formula, β 21 β 22 β 23 e1 is the observer gain; e1 is the observation error; α 21 α 22 α 23 δ 21 δ 22 δ 23 Let f be the parameter of the function fal. The expression for the nonlinear function fal is:
[0058]
[0059] Nonlinear state error feedback: A nonlinear state error feedback controller generates a control quantity to compensate for the total disturbance of the system by nonlinearly processing the state error, thereby achieving precise control of the system. The expression is as follows:
[0060]
[0061] In the formula, β 31 β 32 These represent the error gain; Z 23 is the total disturbance observation estimate of the system; b is the control gain.
[0062] As can be seen from the above structure and mathematical expression of the active disturbance rejection controller, there are more than ten control parameters of the active disturbance rejection controller. If the controller parameters are manually adjusted, the workload is huge and the personal skill level of the debugging personnel is high. This paper uses an improved hog optimization algorithm for parameter tuning of the active disturbance rejection controller.
[0063] The core of the hog optimization algorithm is to simulate the process of a hog herd collectively driving away intruding predators. The specific algorithm content is as follows:
[0064] First, the population is randomly initialized within the specified solution space.
[0065] The first defense mechanism is that when the crested porcupine becomes aware of a predator, it begins to flap its bristles to warn the intruder, expressed as follows:
[0066]
[0067] in, Let i be the position of the i-th crested porcupine in the t-th iteration. This represents the current optimal position for the crowned porcupine, where τ1 and τ2 are random numbers. For another randomly selected location of the crowned porcupine.
[0068] The second defense mechanism involves the porcupine making a sound to scare away intruders after they breach the first layer of defense. The closer the intruder gets, the louder the sound becomes, attempting to deter them. The mathematical expression for this is:
[0069]
[0070] In the formula, U1 is a random vector between 0 and 1; τ3 is a random number; Two crowned porcupine positions were randomly selected.
[0071] The third defense mechanism involves the crowned porcupine releasing gas against intruders who breach the first two layers of defense, preventing them from approaching. The expression for this is:
[0072]
[0073] In the formula, It is a fitness function; δ is a random vector; γ t It is a time-related factor; Three crowned porcupine positions were randomly selected.
[0074] The fourth defense mechanism: When an intruder breaches the first three defense mechanisms, the Crowned Porcupine will launch a physical attack on the intruder, expressed as follows:
[0075]
[0076] In the formula, α is the control parameter; τ4 and τ5 are random numbers; and δ is a random vector. This is a fitness-based factor.
[0077] Cyclic population reduction (CPO) is a novel method proposed by CPO that aims to accelerate convergence while maintaining population diversity. This strategy accelerates convergence by allowing some core members (CPs) to leave the population during optimization, and then reintegrating them to improve diversity and avoid getting trapped in local minima. The mathematical expression is:
[0078]
[0079] In the formula, T c The variable that determines the number of iterations; t is the current iteration number; T max N is the maximum number of iterations. min N is the minimum number of individuals in the new population; N is the population size; the population will initially maximize and then gradually decrease to N. min The periodic changes, completed in the entire algorithm process T c The loop continues.
[0080] Because the Crowned Porcupine optimization algorithm is prone to getting trapped in local optima, and its convergence speed and stability need improvement, the following improvements were made:
[0081] Population initialization is performed using a set of optimal points, a concept proposed by Hua Luogeng et al. Assume H... s It is a unit cube in s-dimensional Euclidean space, in which there exists a set of points.
[0082]
[0083] Its deviation satisfies
[0084] φ(n)=C(r,ε)n -1+ε
[0085] In the formula, C(r,ε) is a constant that depends only on r and ∈ (∈ is a decimal greater than 0), then P is called n Let (k) be the set of optimal points, and r be the optimal points. The value of the optimal point set r is...
[0086]
[0087] In the formula, p is the smallest prime number that satisfies (p-3) / 2≥s.
[0088] The population initialization method based on the best point set is as follows:
[0089]
[0090] In the standard porcupine optimization algorithm, the population is initialized randomly. This method is prone to causing multiple individuals to cluster together, resulting in reduced population diversity and poor algorithm stability. The optimal point set method is used to initialize the population to improve the uniformity of population distribution and algorithm stability.
[0091] The CPO algorithm requires strong global search capabilities in the early stages to explore as many possible solutions as possible within the search space, while in the later stages it requires strong local search capabilities to perform more refined searches within the possible solution regions. In the original algorithm's fourth defense mechanism, the parameter α controls the individual's step size. However, the fixed step size used in the CPO algorithm cannot meet the needs of different stages. This paper employs a nonlinear dynamic weighting method to balance global and local search capabilities, as expressed mathematically below.
[0092]
[0093] In the formula, T max t is the maximum number of iterations; e is the current iteration algebra; in the early stage of iteration, the weights are larger and the individual movement step size is larger, resulting in a strong global search capability in the search space. In the later stage of iteration, the weights become smaller and the individual movement step size is smaller, resulting in a strong local development capability.
[0094] The t-distribution, also known as the Student's t-distribution, is a symmetric probability distribution. Its specific mathematical expression is as follows:
[0095]
[0096] In the formula, Γ is the gamma function; υ is the degree of freedom parameter; when υ is large, the t-distribution is close to the dynamic distribution, and when t is small, the t-distribution has a thick tail. The algorithm is prone to getting trapped in local optima in the later stages of iteration. This paper adopts an adaptive degree of freedom parameter adjustment method. The adaptive t-distribution perturbation dynamically adjusts the degree of freedom parameter, allowing the mutation process to adapt adaptively, thereby improving the algorithm's diversity and exploration ability. min υ max T represents the maximum and minimum values of the degree of freedom parameter; max The maximum number of iterations is given. As the number of iterations increases, the degree-of-freedom parameter increases or decreases linearly. To prevent the algorithm from getting trapped in local optima later in the algorithm, an adaptive t-distribution perturbation mutation is applied to the individual positions, expressed as follows:
[0097] x i,new =x i +f(x i ,v(t))
[0098] In the formula, x i,new The new position after the perturbation; f(x) i ,v(t)) represents the adaptive t-distribution perturbation variation.
[0099] The population is initialized, and individual information is assigned to the ADRC controller. The ADRC position controller receives position commands and uses the feedback tracking error as input to the fitness function. An intelligent optimization algorithm iteratively searches for optimal ADRC controller parameters and then transmits these parameters to the ADRC controller to optimize the control performance. The fitness function used in this paper is...
[0100]
[0101] In the formula, T is the maximum running time, and t s e(t) represents the current running time. s ) represents the error present at the current time, θ(t) s ) represents the current location at the current time, θ * Given the target position, the fitness function mainly considers two performance indicators: total system error and maximum overshoot. The improved CPO algorithm optimizes the global optimal solution (i.e., the optimal individual) by minimizing the sum of these two indicators, thereby achieving optimal control of the control system.
[0102] The overall framework of the system is as follows Figure 2The designed system has two operating modes: optimization mode and regular operating mode. The system operates in optimization mode, where the optimal control parameters are updated based on actual conditions after each round of optimization before switching back to regular operating mode to achieve system control. If the control effect in the current regular operating mode is unsatisfactory, the initialization parameters of the proposed algorithm can be modified to re-obtain better control parameters using optimization mode.
[0103] This invention proposes a method for optimizing control parameters of an Active Disturbance Rejection Controller (ADRC). The Crested Porcupine Optimization (CPO) algorithm is a novel swarm intelligence optimization algorithm that significantly improves convergence speed and accuracy in multiple engineering applications compared to other optimization algorithms. However, it is prone to getting trapped in local optima, and its convergence speed requires further improvement. This invention improves the CPO algorithm by employing a set of optimal points, nonlinear dynamic weights, and an adaptive t-distribution disturbance mutation method. The improved CPO algorithm is then applied to the parameter optimization of an ADRC, and the optimized ADRC achieves good control performance.
Claims
1. An improved method for tuning parameters of the active disturbance rejection controller (ADC) for a permanent magnet synchronous motor, characterized in that, Includes the following steps: 1) Initialize the control parameters of the CPO algorithm, initialize the population using the set of best points, and record the initial global optimal solution and the historical optimal solution of each individual; 2) Perform nonlinear dynamic weight adjustment; 3) Update individual positions based on the four defense mechanisms of the CPO algorithm; 4) Adaptive t-distribution perturbation variation is applied to individuals; 5) Construct a fitness function, use the value of the fitness function to evaluate the control effect of the current active disturbance rejection controller, and retain the historical best solution and the current global best solution of the population based on the value of the fitness function; 6) Use cyclic population technology to dynamically adjust the population size based on the current search status; 7) Determine if the termination condition has been met. If the termination condition has been met, retain the current global optimal solution and use it to update the parameters of the active disturbance rejection controller; otherwise, return to step 2.
2. The improved CPO (Continuous Disturbance Rejection Controller) parameter tuning method for permanent magnet synchronous motors according to claim 1, characterized in that, The nonlinear dynamic weight adjustment is specifically as follows: Among them, T max α is the maximum number of iterations; t is the current iteration number; α(t) is the weight; and e is the natural constant.
3. The improved CPO (Continuous Disturbance Rejection Controller) parameter tuning method for permanent magnet synchronous motors according to claim 1, characterized in that, The adaptive t-distribution perturbation variation of the individual specifically refers to: x i,new =x i +f(x i ,v(t)) Where, x i,new For the new position after the perturbation, x i f(x) represents the current position of the individual. i ,v(t)) represents the adaptive t-distribution perturbation variation.
4. The improved CPO (Continuous Disturbance Rejection Controller) parameter tuning method for permanent magnet synchronous motors according to claim 1, characterized in that, The fitness function is: Where T is the maximum running time, e(t) s ) represents the error present at the current time, θ(t) s ) represents the current location at the current time, θ * Given the target location.
5. The improved CPO (Continuous Disturbance Rejection Controller) parameter tuning method for permanent magnet synchronous motors according to claim 1, characterized in that, The cyclic population technique specifically refers to: Among them, T c The variable used to determine the number of iterations; t is the current iteration number; T max N is the maximum number of iterations. min It is the minimum number of individuals in the new population, and N is the population size.
6. An improved CPO (Continuous Disturbance Rejection Controller) parameter tuning system for permanent magnet synchronous motors, characterized in that, It includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement, when executing the computer program, a parameter tuning method for an improved CPO permanent magnet synchronous motor active disturbance rejection controller as described in any one of claims 1-5.
7. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements an improved CPO (Continuous Disturbance Rejection Controller) parameter tuning method for a permanent magnet synchronous motor as described in any one of claims 1-5.