CN105759607B - The design method of PAC controllers based on intelligent control algorithm - Google Patents

Abstract

The invention discloses a design method of a PAC controller based on an intelligent control algorithm. On the basis of a traditional PAC controller, the invention adopts an improved PID control algorithm, a variable universe fuzzy PID control algorithm, a neural network-based PID control algorithm, etc. Intelligent control algorithm can improve control precision, improve control response, and meet higher control requirements. In terms of network communication, on the basis of traditional serial ports 232 and 485, communication methods such as Ethernet and CAN bus are added, and Modbus‑RTU, Modbus‑TCP, CANopen, custom protocols, etc. are used to realize network interconnection. On the basis of conventional PID, the algorithm of preventing integral saturation, differentiating the control variable, and differential first are added, which can significantly improve the effect of PID, improve the response time, and reduce the control overshoot.

Description

Design Method of PAC Controller Based on Intelligent Control Algorithm

technical field

The invention belongs to the field of industrial automatic control, and relates to the design of an embedded PAC controller, high-precision temperature control, intelligent control algorithm and the like.

Background technique

In the production process of industrial manufacturing, the demand for control is getting higher and higher. PAC controllers have some notable advantages in control performance, information processing and network communication capabilities. PAC controller combines the inherent reliability, robustness and distribution characteristics of PLC. Compared with PC control, PAC uses a real-time operating system, which has incomparable advantages in processing performance such as real-time and deterministic PC. However, the traditional PAC controller is relatively simple in the control algorithm. When it comes to some complex algorithms, the method often used is combined control with a PC, which also exposes some defects, such as poor control effect and communication delay. sex, cost increases.

Contents of the invention

According to the problems proposed in the above-mentioned prior art, the purpose of the present invention is to: improve the control performance of the control system, adopt the improved PID control algorithm, the variable domain fuzzy PID control algorithm, the PID control algorithm based on neural network, and can improve the control accuracy , to improve control response. In addition, the scalability of the acquisition signal is improved, and the analog input signal can be a voltage signal, a current signal, or a resistance signal. In terms of network communication, on the basis of traditional serial ports 232 and 485, communication methods such as Ethernet and CAN bus are added, and Modbus-RTU, Modbus-TCP, CANopen, custom protocols, etc. are used to realize network interconnection.

The present invention is realized based on an intelligent control device, and the intelligent control device mainly includes the following parts:

1. Sensor sampling circuit, which includes a sampling signal conditioning circuit for voltage, current, and resistance signals, in order to achieve high-precision acquisition of sensor signals, such as pressure transmitters, PT100 temperature sensors, angle sensors, 4-20mA current sensors, etc. .

2. Control output circuit, including isolation unit, using solid state relay to control high voltage and high current. Using PWM output signal, the control output has high precision.

3. Microcontrollers used to participate in calculations, as well as Ethernet communication circuits, 232, 485, CAN bus communication circuits, reset circuits, and watchdog circuits.

4. A power module circuit that ensures a stable power supply.

The sensor interfaces with the microcontroller through an input signal conditioning circuit.

The present invention proposes some intelligent control algorithms and its realization, mainly including the following parts:

Step 1: PID self-tuning algorithm is designed. Common industrial control objects have factors such as nonlinearity, time-varying and uncertainty, which cause PID parameters to be time-consuming by manual tuning, and the tuning effect is relatively poor. In the present invention The algorithm of PID self-tuning is added in it, which can automatically determine the PID parameters of equipment operation, saves time and improves the control effect.

Step 2: In addition, on the basis of traditional PID control, algorithms such as integral saturation prevention, differentiation of control variables, and differential advance are added, which can significantly improve the effect of PID, improve response time, and reduce control overshoot.

Step 3: Due to the nonlinear, time-varying and uncertain factors of the control object, the control effect is relatively poor if only PID control is used. On the basis of improving the conventional PID control method, this method adds a fuzzy control algorithm, and performs fuzzy control operations by establishing a fuzzy rule table imitating human knowledge language and a membership function, using a combination of fuzzy control and conventional PID control. The method can improve the control effect of the controller when the control object has nonlinear, time-varying and uncertain factors.

Step 4: In fuzzy control, most of the fuzzy rules and membership functions are obtained from human experience, and the control effect is also closely related to the selection of fuzzy rules and membership functions. However, the neural network has the advantages of strong nonlinear fitting ability, learnable and self-adaptive ability. Combining the neural network with PID control can use the self-learning ability of the neural network to calculate the optimal parameters of the PID control to achieve the optimal Excellent control, the effect is remarkable.

The realization of the PID self-tuning mentioned in the step 1 is specifically divided into the following steps:

1.1, using the ZN (Ziegler-Nicels law) relay feedback setting method to achieve, compared with the ZN method has a good advantage, that is, using the relay method to generate the oscillation link, in which the description equation of the relay characteristic : In the formula, M is the amplitude of the relay characteristic, and X is calculated from the measured output peak difference. When argG(jω)=-π is satisfied, In the formula, A is obtained by measuring the maximum and minimum values of the output, K _u is the critical oscillation proportional gain, and d is the amplitude of the divided symmetrical relay characteristic.

1.2. Obtain the above K _u and the critical oscillation period T _u through the generated oscillation curve. According to the calculation formula of Ziegler-Nichols self-tuning method, as shown in Table 1, calculate the tuned PID parameters according to the required performance requirements, and complete PID self-tuning process of relay feedback.

Table 1 Z-N self-tuning PID parameters (fast performance)

The implementation of the improved conventional PID mentioned in step 2 is specifically divided into the following steps:

2.1. If the conventional PID algorithm is used without modification, it will expose the shortcomings of easy overshoot, long system stabilization period, and oscillation phenomenon when it is stable.

2.2, the common form of PID is:

The anti-integral saturation algorithm is used to limit the overshoot phenomenon caused by the integral, and it can quickly improve the rise time. Among them, Umin≤u(t)≤Umax, when , let u ₁ (t)=Umax, where let The phenomenon of integral windup is limited. In addition to improving the variation of the differential, due to Where e(t)=r(t)-y(t), where r(t) is the setting value and y(t) is the sensor sampling value, however, when r(t) is adjusted, it will inevitably lead to the instantaneous Changes will increase the instability of the system, so make

The realization of the variable universe fuzzy control PID mentioned in the step three is specifically divided into the following steps:

3.1. The variable universe adaptive fuzzy control uses the error e and the error change rate ec as the input of the system, and automatically adjusts the PID parameters according to the difference between e and ec at different times. The fuzzy rules online adjust the Kp, Ki, and Kd for modification. Fuzzy self-tuning PID is based on the PID algorithm, by calculating the current system error e and error change rate ec, using fuzzy rules for fuzzy reasoning, querying the fuzzy rule table for parameter adjustment.

3.2. The establishment of the fuzzy control table is not unique, and it is different for different control systems, but for common systems, the following fuzzy control tables can meet the control requirements:

ΔKP fuzzy rule table:

ΔKi fuzzy rule table:

ΔKd fuzzy rule table:

3.3 The selection of the domain of discourse and the determination of the membership function are determined by manual experience to ensure as accurate as possible. In addition, by monitoring the changes of e and ec, the scope of the domain of discourse is continuously optimized and narrowed to achieve the problem of self-tuning of the domain of discourse. The realization of the neural network PID mentioned in the step 4 is specifically divided into the following steps:

4.1 This method adopts the single-neuron neural network PID control algorithm, which can avoid the problem of large amount of calculation like using BP neural network, genetic algorithm, and particle swarm algorithm, and can also ensure real-time control and continuous correction online.

4.2 As shown in Figure 4, the neuron PID structure model is a structure with three inputs and one output, of which three inputs are e(k), The output of the neuron is u(k), and the weight of the neuron is the three coefficients of PID, namely, Kp, Ki, and Kd, which are proportional, integral, and differential.

In the same way,

Where μ is the learning rate, 0<μ<1;

Kp(k+1)=Kp(k)+ΔKp(k+1),

Ki(k+1)=Ki(k)+ΔKi(k+1),

Kd(k+1)=Kd(k)+ΔKd(k+1).

4.5 The initial value of the weight of the neural network uses the value calculated by the PID self-tuning method involved in step 1 as the initial value, which can speed up the speed of the PID self-tuning of the neural network.

In the design of the conventional PAC controller, the present invention improves the acquisition accuracy on the hardware, improves the network interconnection and intercommunication, uses CANopen, Modbus-TCP, and Modbus-RTU as common communication protocols, and adopts the improved conventional PID control method, Variable universe fuzzy self-adaptive method, neural network PID algorithm, more accurate and good control of industrial control field equipment, can be better applied in oil pipeline electric heating system, cotton textile industry temperature and humidity control, boiler temperature control, etc. occasion.

Description of drawings

Fig. 1 system hardware structure block diagram;

Figure 2 PID relay self-tuning block diagram;

Fig. 3 block diagram of variable domain fuzzy control PID structure;

Fig. 4 neural network PID structural block diagram;

Detailed ways

According to the following detailed description of the specific implementation of the present invention in conjunction with the accompanying drawings, those skilled in the art will be more aware of the above advantages and features of the present invention.

The first is the hardware structure, as shown in Figure 1.

The microcontroller, STM32F207VET6 from STMicroelectronics, has a large storage space and high-performance computing speed, which can ensure the operation of some control algorithms.

The watchdog and reset circuit, choose SP706, which is a special watchdog reset chip, which can improve the stability and anti-interference of the system.

The power supply module adopts switching power supply 220v to 24v outside the equipment, and LM2576 and SPX1117-3.3 are used for voltage conversion in the system equipment to meet the voltage requirements of the system module. In addition, the ADC sampling module needs an accurate reference voltage, and TL431 is selected as the reference voltage source.

The sensor and input signal conditioning circuit, in order to ensure that the 4-20mA current type, voltage type, resistance value change type sensors, etc. are connected to the system, the corresponding jumping selection switch can be set on the same port according to the selection of sensors. For example, the temperature sensor uses PT100, and the signal conditioning circuit adopts the bridge method to convert the resistance value changed by PT100 into a corresponding voltage form and send it to the ADC of the controller for sampling. After conversion, the actual temperature value is obtained, and the collected temperature is sent to the microprocessor for processing. Other sensor circuits are also processed similarly.

The output signal conditioning circuit and actuator, the digital control quantity obtained by the controller operation, can convert the output mode into switching quantity, PWM output, etc. according to the demand. There is an electrical isolation module inside to prevent external interference, and a solid-state relay is built in for high-power output control.

The data storage module is FM24CL04, which can store a small amount of data and save it when power off. Can be used to store some key parameter information.

The communication module adopts MAX3485, MAX3232, PCA82C250, DP83848, etc. as data processing chips and physical layer transceiver circuits for RS485, RS232, CAN communication, and Ethernet communication respectively. The choice of multiple communication methods can carry out more compatible information exchange with different devices.

The PID relay self-tuning method is shown in Figure 2.

Through the oscillation curve generated by the relay link, K _u and the critical oscillation period T _u (see above) are obtained. The calculation formula of the Ziegler-Nichols self-tuning method is shown in Table 2. According to the required performance requirements, Calculate the adjusted PID parameters and complete the PID self-tuning process of relay feedback.

Table 2 Relay feedback type Z-N self-tuning PID parameters (fast performance)

After calculation, PID parameters can be obtained, which can be used as the initial value of fuzzy control and neural network control.

The variable domain fuzzy control method is shown in Figure 3. It is mainly composed of fuzzy control and PID controller.

The input of the digital PID controller is the acquisition value r(t) obtained through sensor sampling and the target value y(t) set manually, and the digital output control quantity u(t) is obtained by improving the PID control algorithm. The control quantity u(t ) through D/A change or output control through digital output PWM wave. And then control the operation of the equipment. To achieve the function of controlling the output at the target value.

The variable universe fuzzy control PID algorithm is shown in Fig.3.

1. Determine the domain of discourse and the membership function. Determine the domain of discourse according to the deviation e of the input signal and the deviation change rate ec input range. The determination of the domain of discourse is generally based on artificial experience. The fuzzy set of the domain of discourse is divided into {negative large, negative medium, negative small, zero, positive small, positive, Zhengda}. The specific scope of discourse needs to be determined in combination with the usage scenarios. The membership function is to determine the proportion of the range of the input signal value. In embedded devices, the curve composed of straight lines is generally used as the membership function model. In order to facilitate the calculation of the embedded processor, it is prepared for the following fuzzy rules and defuzzification.

2. Determine the fuzzy rule table. The establishment of the fuzzy rule table should be as accurate as possible, because the quality of the control algorithm is closely related to the accuracy of the fuzzy rules. Generally, the range of e and ec is divided into seven segments, namely: {negative large, negative medium, negative small, zero, positive small, positive medium, positive large}. Therefore, a complete set of 49 fuzzy rules can be established. Taking the temperature rise curve as an example, when the temperature is rising, Kp should be set to a relatively large value to ensure quick response and reduce the rise time, and Ki should be relatively small to prevent integral saturation and integral overshoot. , Kd takes a relatively large value, in order to prevent overshoot, reduce the output of the system, and control the rate of rise.

So Rule 1: When e is "PB" and ec is "Z0", ΔKp is "PM", ΔKi is "PS", and ΔKd is "PM".

Rule 2: When e is "PB" and ec is "NS", ΔKp is "PM", ΔKi is "PS", and ΔKd is "PM".

Rule 3: When e is "PB" and ec is "NM", ΔKp is "PS", ΔKi is "Z0", and ΔKd is "PM".

Based on this, the specific rule list is as follows:

ΔKP fuzzy rule table:

ΔKi fuzzy rule table:

ΔKd fuzzy rule table:

3. In order to obtain the real PID controller parameters Kp, Ki and Kd, make the following changes:

Kp=Kp0+ΔKp; Ki=Ki0+ΔKi; Kd=Kd0+ΔKd (Kp0, Ki0, Kd0 represent the initial value of the PID coefficient respectively);

The neuron PID structure model is shown in Figure 4,

1. The structure is a structure with three inputs and one output, where three inputs are e(k), The output of the neuron is u(k), and the weight of the neuron is the three coefficients (Kp, Ki, Kd) of the PID ratio, integral, and differential.

(where μ is the learning rate, 0<μ<1)

In the same way,

4. The initial value of the neural network weight can use the value calculated by the PID self-tuning method mentioned in step 1 of this method as the initial value, which can speed up the speed of the neural network PID self-tuning.

Claims (1)

1. the design method of the PAC controllers based on intelligent control algorithm, this method are realized based on a kind of intelligent controlling device, The intelligent controlling device includes mainly following components：

1) senser sampling circuit, which includes the sampled signal modulate circuit of voltage, electric current, resistance signal, to realize The signal high precision of sensor acquires；

2) control output circuit controls big voltage high-current containing isolated location using solid-state relay；It is exported and is believed using PWM Number, control output accuracy is high；

3) be used to participate in the microcontroller and Ethernet communication circuit of operation, 232,485, CAN bus communicating circuit, reset Circuit, watchdog circuit；

4) ensure the power module circuitry of power good；

Sensor is interacted by input signal conditioning circuit with microcontroller；

It is characterized in that：This approach includes the following steps：

Step 1：PID Self-tuning System algorithms are designed, common Industry Control object has non-linear, time variation and uncertainty Factor is caused pid parameter to be compared using the method manually adjusted and expends the time, and the effect adjusted is also poor, in the method The algorithm of PID Self-tuning Systems is increased, the pid parameter of equipment operation can be automatically determined, save the time, improves control effect Fruit；

Step 2：In addition on the basis of traditional PID control, increase prevent integral saturation algorithm, to control variable carry out it is micro- Divide, differential forward algorithm, the effect of PID can be significantly improved, improve the response time, reduces control overshoot；

Step 3：Due to the non-linear of control object, time variation and uncertain factor, PID control, control effect is only used only Fruit is poor；This method increases FUZZY ALGORITHMS FOR CONTROL on the basis of improving regulatory PID control method, by establishing mould Fuzzy reasoning table and the membership function of human knowledge language are imitated to carry out fuzzy control operation, using fuzzy control and often The mode that is combined of rule PID control, can improve controller control object have non-linear, time variation and it is uncertain because The control effect of element；

Step 4：In fuzzy control, the acquisition of fuzzy rule and membership function greatly mostly from the experience of people, control effect Selection of the quality also with fuzzy rule and membership function is closely bound up；However neural network has very strong nonlinear fitting energy Power, can learn with adaptive ability advantage, neural network is combined with PID control, can utilize neural network self study Ability calculates the optimized parameter of PID control, is optimal control, significant effect；

The realization for the PID Self-tuning Systems that the step 1 is mentioned is specifically divided into following steps：

1.1, realize have very compared to Z-N methods using Z-N, that is, Ziegler-Nichols rule relay feedback formula setting method Good advantage, i.e., generate oscillation element, the wherein descriptive equation of relay characteristics using relay method： In formula, M is relay characteristics amplitude, and X is to measure output peak difference calculating to acquire；When meeting argG (j ω)=- π,A is acquired by measuring the maximum value exported and minimum value in formula, K_uFor threshold oscillation proportional gain, d is The amplitude of the symmetrical relay characteristics divided；

1.2, by the oscillating curve of generation, obtain the above K_uAnd critical period of the oscillation T_u, by Ziegler-Nichols Self-tuning Systems The calculation formula of method, as shown in table 1, performance requirement as needed calculate the pid parameter adjusted, and complete relay feedback PID Self-tuning System processes；

1 Z-N selftuning PID parameters of table

The realization for the improvement Traditional PID that the step 2 is mentioned is specifically divided into following steps：

2.1, normal PID lgorithm if use without modification, can expose easy overshoot, system time stable period it is long, The shortcomings that oscillatory occurences is had when stablizing；

2.2, PID common type is：

Over control is led to limit integral using anti-windup saturation algorithm, and can quickly be improved the rise time；Wherein, Umin≤ U (t)≤Umax, whenWhen, enable u₁(t)=Umax, wherein enablingLimit integral saturated phenomenon；In addition the variable quantity for improving differential, due toWherein e (t)=r (t)-y (t), wherein r (t) are setting value, and y (t) is sensor sample value, however works as r (t) when adjustment, necessarily lead to the instantaneous variation of differential term, the unstability of system can be increased, therefore enable

The realization for the Varied scope fuzzy control PID that the step 3 is mentioned is specifically divided into following steps：

3.1, theory-region changed adaptive-fuzzy control is using error e and error rate ec as the input of system, in different moments Pid parameter is automatically adjusted according to the difference of e and ec, fuzzy rule online modifies to Kp, Ki, Kd of pid parameter； Fuzzy selftuning PID is on the basis of pid algorithm, by calculating current system error e and error rate ec, using fuzzy Rule carries out fuzzy reasoning, and inquiry fuzzy reasoning table carries out parameter adjustment；

3.2, the foundation of fuzzy control table is not unique, for different control systems difference, but is directed to system, use with Lower fuzzy control table can meet control and require：

Δ KP fuzzy reasoning tables:

Δ Ki fuzzy reasoning tables:

Δ Kd fuzzy reasoning tables:

The selection of 3.3 domains and the determination of membership function are determined by artificial experience, it is ensured that accurate as possible；Additionally by prison The variation for surveying e and ec, the problem of continuing to optimize and reduce the range of domain, reach variable universe Self-tuning System；

The realization for the Neural network PID that the step 4 is mentioned is specifically divided into following steps：

4.1 this method use single neuron Neural Network PID, can avoid as being calculated using BP neural network, heredity The computationally intensive problem of method, particle cluster algorithm, can also ensure real-time control, online constantly to correct；

4.2 Neuron PID structural models are 3 and input the structure singly exported, wherein 3 inputs are e (k),e(k)-e (k-1), the output of neuron is u (k), wherein the weights of neuron be the ratio of PID, integral, three coefficient, that is, Kp of differential, Ki、Kd；

The output of 4.3 neurons isWherein Ts is sampling time, error Function is e (k)=r_in(k)-y_out(k), performance indicator is takenUsing steepest descent method, the weights tune of neuron It is whole as follows：

It can similarly obtain,

Wherein μ be learning rate, 0<μ<1；

4.4 enablingIt can then be acquired successively by step 3：

Kp (k+1)=Kp (k)+Δ Kp (k+1),

Ki (k+1)=Ki (k)+Δ Ki (k+1),

Kd (k+1)=Kd (k)+Δ Kd (k+1)；

The initial value of 4.5 neural network weights using the calculated value of PID automatic setting methods that step 1 is related to as initial value, It can accelerate the speed of Neural network PID Self-tuning System.