A temperature self-adaptive control method for a polyacrylonitrile reactor

Abstract

This invention discloses a temperature adaptive control method for a polyacrylonitrile reactor, belonging to the field of temperature control technology. The method includes the following steps: constructing a two-degree-of-freedom (DOF) temperature control system for the polyacrylonitrile reactor; constructing the error transfer function of the two-DOF temperature control system and performing performance analysis to obtain a dual PID parameter set; acquiring the operating state of the two-DOF temperature control system; constructing a two-DOF control BP neural network; constructing a dual-error fusion performance index; and synchronously and online adaptively optimizing the dual PID parameter set output by the two-DOF control BP neural network based on the operating state of the two-DOF temperature control system, corresponding to the outputs of the main controller and the feedforward compensator, thus completing the temperature adaptive control of the polyacrylonitrile reactor. This invention solves the problems of insufficient accuracy and robustness in temperature control of polyacrylonitrile reactors.

Description

Technical Field

[0001] This invention belongs to the field of temperature control technology, and particularly relates to an adaptive temperature control method for a polyacrylonitrile reactor. Background Technology

[0002] Polyacrylonitrile (PAC) is a core material in aerospace, military, and other fields, and its product quality indicators are highly dependent on the precise control of the polymerization reaction temperature. The PAC polymerization process is complex and potentially dangerous; temperature runaway can easily lead to safety accidents. Furthermore, the uniformity of the polymerization temperature directly affects the subsequent heat treatment effect and the core properties of the final product. Industrially, PAC is mainly synthesized through acrylonitrile free radical polymerization, with the reactor being a key piece of equipment. Due to factors such as uneven mixing of the reaction system, lag in heat and mass transfer, and the shared jacket for hot and cold media, the reactor exhibits significant hysteresis and strong coupling characteristics, making precise temperature control extremely challenging.

[0003] Currently, single-degree-of-freedom PID control or cascade control is widely used for temperature control in industrial polymerization reactors. While these methods can achieve basic setpoint tracking, they generally suffer from drawbacks such as response lag and large overshoot when faced with jacket temperature disturbances, cooling / heating medium flow fluctuations, and model parameter perturbations. Furthermore, single-degree-of-freedom PID controllers rely on offline parameter tuning, making real-time optimization difficult. To address the nonlinearity and parameter uncertainty of reactor temperature control, research has introduced intelligent methods such as fuzzy control and neural network control to improve system adaptability. However, the jacket temperature disturbance in polyacrylonitrile reactors has a significant impact, and intelligent control based on a single-loop feedback structure still struggles to overcome the mutual constraints between tracking and disturbance rejection performance, making it difficult to meet the requirements for high-precision temperature control. Two-degree-of-freedom control achieves structural performance decoupling by independently designing a main controller for setpoint tracking and a feedforward compensator for disturbance suppression, but it still fails to overcome the contradiction of online adaptive tuning of two-degree-of-freedom parameters, making it unsuitable for the dynamic operating conditions of polyacrylonitrile polymerization reactions. Moreover, changes in raw material ratios, load fluctuations, and the inherent nonlinear characteristics of the system significantly increase the difficulty of identification during the polymerization process. Summary of the Invention

[0004] To address the aforementioned shortcomings in the existing technology, this invention provides a temperature adaptive control method for a polyacrylonitrile reactor, which solves the problems of insufficient temperature control accuracy and robustness in polyacrylonitrile reactors.

[0005] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: The present invention provides a temperature adaptive control method for a polyacrylonitrile reactor, comprising the following steps: S1. Construct a two-degree-of-freedom temperature control system for a polyacrylonitrile reactor, wherein the two-degree-of-freedom temperature control system includes a main control loop and a feedforward compensation loop. S2. Construct the error transfer function of the two-degree-of-freedom temperature control system, and perform performance analysis on the two-degree-of-freedom temperature control system to obtain the dual PID parameter set; S3. Obtain the operating status of the two-degree-of-freedom temperature control system; S4. Construct a two-degree-of-freedom control BP neural network; S5. Construct a dual-error fusion performance index, and perform synchronous online adaptive optimization of the dual PID parameter group output by the two-degree-of-freedom control BP neural network according to the operating status of the two-degree-of-freedom temperature control system, and correspondingly control the output of the main controller and the feedforward compensator to complete the temperature adaptive control of the polyacrylonitrile reactor.

[0006] Furthermore, the calculation expression for the two-degree-of-freedom temperature control system in S1 is as follows: , in, This indicates the actual internal temperature value. Indicates the main controller. Indicates the controlled object, This indicates the preset internal temperature value. Indicates feedforward compensator. This represents the jacket temperature disturbance value, where, For Laplace variables.

[0007] Furthermore, the calculation expression for the error transfer function of the two-degree-of-freedom temperature control system in S2 is as follows: , in, This represents the error transfer function of a two-degree-of-freedom temperature control system.

[0008] Furthermore, the dual PID parameter group in S2 includes main controller PID parameters and feedforward compensator PID parameters, wherein the main controller PID parameters include a first proportional... First point and the first differential The feedforward compensator PID parameters include the second proportional... Second integral Second differential .

[0009] Furthermore, the operating state of the two-degree-of-freedom temperature control system in S3 includes the internal temperature tracking error, the rate of change of the internal temperature error, and the observed value of the jacket disturbance. ; The formula for calculating the internal temperature tracking error is as follows: , in, Indicates the first Internal temperature tracking error during the next iteration Indicates the first The preset internal temperature value for the next iteration Indicates the first The actual internal temperature value at the next iteration; The formula for calculating the rate of change of internal temperature error is as follows: , in, Indicates the first The rate of change of internal temperature error during each iteration. Indicates the first Internal temperature tracking error during the next iteration.

[0010] Furthermore, the two-degree-of-freedom control BP neural network in S4 adopts a three-layer neural network topology architecture, including an input layer, a hidden layer, and an output layer. The input layer has 3 nodes, the hidden layer has M nodes, and the output layer has 6 nodes. The input signals corresponding to the three nodes of the input layer are the operating status of the two-degree-of-freedom temperature control system. The hidden layer uses a sigmoid activation function for nonlinear feature mapping. The six nodes of the output layer correspond one-to-one with the PID parameters of the main controller and the PID parameters of the feedforward compensator in the dual PID parameter group.

[0011] Further, step S5 includes the following steps: S51. Construct dual-error fusion performance indicators; The calculation expression for the dual-error fusion performance index in S51 is as follows: , in, Indicates the first The performance metrics of dual-error fusion at the next iteration Indicates the tracking error weight. Indicates the weight of the disturbance error. Indicates the first Temperature deviation caused by perturbation during the next iteration; S52. Using the dual-error fusion performance index as the learning target, the operating state of the two-degree-of-freedom temperature control system is input into the two-degree-of-freedom control BP neural network to perform synchronous online adaptive optimization of the dual PID parameter group output by the two-degree-of-freedom control BP neural network, and the connection weights of the neurons in the two-degree-of-freedom BP neural network are updated using the gradient descent model. The calculation expression for the gradient descent model is as follows: , in, Indicates the first During the nth iteration Layer The update amount of the connection weights of each neuron Indicates the learning rate. Indicates the first During the next iteration for gradient, Indicates the first Layer The connection weights of each neuron Represents the momentum factor. Indicates the first During the nth iteration Layer The amount of update to the connection weights of each neuron; S53. Based on the dual PID parameter group output by the two-degree-of-freedom BP neural network, the outputs of the main controller and the feedforward compensator are controlled according to the incremental output model of the main controller and the incremental output model of the feedforward compensator, respectively, to complete the temperature adaptive control of the polyacrylonitrile reactor.

[0012] Furthermore, the calculation expressions for the incremental output model of the main controller and the incremental output model of the feedforward compensator in S53 are as follows: in, Indicates the first The main controller outputs an increment during the next iteration. Indicates the first The rate of change of internal temperature tracking error at each iteration Indicates the first The increment of the feedforward compensator output in the next iteration Indicates the first The increment of the observed value of the jacket perturbation at the next iteration Indicates the first The observed values ​​of the jacket perturbation at the next iteration. Indicates the first The increment of the rate of change of the perturbation at the next iteration. Indicates the first Rate of change of internal temperature error in each iteration Indicates the first The observed values ​​of the jacket perturbation at the next iteration. Indicates the first The increment of the observed value of the jacket perturbation at the next iteration.

[0013] The beneficial effects of this invention are as follows: The temperature adaptive control method for a polyacrylonitrile reactor provided by this invention achieves complete decoupling of tracking and anti-interference from a structural perspective through two-degree-of-freedom control, allowing the main controller to focus solely on setpoint tracking and optimization. The constructed two-degree-of-freedom control BP neural network senses the system's tracking dynamics in real time through the internal temperature tracking error and the rate of change of the internal temperature error in the input layer, and tunes the main controller's PID parameters online. Utilizing the nonlinear approximation capability of the two-degree-of-freedom control BP neural network, the response lag problem caused by the large lag characteristic of the polyacrylonitrile reactor is mitigated, achieving fast response, no overshoot, and high-precision tracking of the preset internal temperature value. This invention provides a series structure of the PID adjustment link of the feedforward compensator and a low-order inertial link, achieving disturbance advance compensation through reasonable selection of the order and time constant. The jacketed disturbance observation value introduced into the input layer of the two-degree-of-freedom control BP neural network enables early perception of disturbance change trends and online tuning of the feedforward compensator's PID parameters, compared to conventional two-degree-of-freedom control... This invention significantly shortens the disturbance recovery time, improves the speed and stability of disturbance suppression, and effectively solves the problem of frequent disturbances caused by the coupling of hot and cold media in the reactor. Through a fusion control method of "two-degree-of-freedom structural decoupling and BP network parameter coordination," this invention avoids the mutual constraint between tracking and disturbance rejection in traditional single-degree-of-freedom control. By leveraging the nonlinear approximation and online learning capabilities of the two-degree-of-freedom control BP neural network, it effectively compensates for the impact of model inaccuracies and time-varying parameters. This invention achieves online collaborative tuning of dual PID parameter groups through dual-error fusion performance indexes, dynamically adapting to changes in object characteristics. This ensures that the two-degree-of-freedom temperature control system of the polyacrylonitrile reactor maintains stable control performance and exhibits excellent robustness even under conditions of reactor model mismatch, frequent switching between hot and cold media, and large fluctuations in disturbance amplitude. This invention achieves significant improvements in three core dimensions: setpoint tracking, disturbance suppression, and robustness, enabling efficient and precise temperature control of the polyacrylonitrile reactor.

[0014] Other advantages of the present invention will be analyzed in more detail in the following embodiments. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1This is a flowchart illustrating the steps of a temperature adaptive control method for a polyacrylonitrile reactor according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the control process of the two-degree-of-freedom temperature control system for the polyacrylonitrile reactor in an embodiment of the present invention.

[0018] Figure 3 This is a block diagram of the two-degree-of-freedom temperature control system for the polyacrylonitrile reactor in an embodiment of the present invention.

[0019] Figure 4 This is a schematic diagram illustrating the principle of temperature adaptive control of the polyacrylonitrile reactor in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, in one embodiment of the present invention, the present invention provides a temperature adaptive control method for a polyacrylonitrile reactor, comprising the following steps: S1. Construct a two-degree-of-freedom temperature control system for a polyacrylonitrile reactor, wherein the two-degree-of-freedom temperature control system includes a main control loop and a feedforward compensation loop. like Figure 2 As shown, the two-degree-of-freedom temperature control system for the polyacrylonitrile reactor consists of a main control loop and a feedforward compensation loop: the main loop uses the temperature inside the reactor as the feedforward compensation loop. As the controlled variable, TC1 is the main controller that accurately tracks the preset internal temperature value; the feedforward loop uses the jacket temperature... The measurable disturbance is the input, and TC2 is the feedforward compensator. It completes the compensation in advance before the disturbance affects the controlled object. The two channels work independently and do not interfere with each other, and can be optimized to the optimal state respectively.

[0022] like Figure 3 As shown, the calculation expression for the two-degree-of-freedom temperature control system in S1 is as follows: , in, This indicates the actual internal temperature value. Indicates the main controller. Indicates the controlled object, This indicates the preset internal temperature value. Indicates feedforward compensator. This represents the jacket temperature disturbance value, where, For Laplace variables.

[0023] In this scheme, if only the preset internal temperature value is considered, and the jacket temperature disturbance value is 0, the calculation expression of the tracking transfer function can be found as follows: in, This represents the tracking transfer function; therefore, it can be seen that the tracking performance is solely determined by the main controller. The decision was made by adjusting the main controller. The parameters can achieve good tracking performance; If we only consider the input of the jacket temperature disturbance value, and the preset internal temperature is 0, the calculation expression of the disturbance transfer function can be found as follows: , in, This represents the disturbance transfer function; it can be seen that the disturbance suppression performance is related to both the main controller and the feedforward compensator. In this scheme, after the main controller PID parameters are tuned, only the feedforward compensator PID parameters need to be tuned to meet the disturbance suppression performance without changing the tracking performance. The two-degree-of-freedom temperature control system provided by this invention uses a two-degree-of-freedom control method. In this scheme, the preset internal temperature value tracking and disturbance suppression are divided into two independent channels, and corresponding controllers are designed for each channel. The parameters are adjusted independently, thus structurally removing performance constraints. This enables simultaneous fast and stable tracking and strong disturbance suppression, achieving complete decoupling of reactor internal temperature and jacket temperature control, as well as independent tuning of dual-path parameters.

[0024] S2. Construct the error transfer function of the two-degree-of-freedom temperature control system, and perform performance analysis on the two-degree-of-freedom temperature control system to obtain the dual PID parameter set; Analysis of the performance of a two-degree-of-freedom temperature control system for a polyacrylonitrile reactor reveals that the two-degree-of-freedom dual-loop structure requires simultaneous tuning of six PID parameters for both the main controller and the feedforward compensator. Traditional BP neural network PID controllers are only designed for single-loop control systems and cannot meet the optimization requirements of a two-degree-of-freedom dual-loop, dual-function, and dual-indicator system. Therefore, this solution specifically designs a 3-M-6 BP neural network for the two-degree-of-freedom temperature control system, achieving deep integration with the system from three levels: input signal, output structure, and learning objective. The calculation expression for the error transfer function of the two-degree-of-freedom temperature control system in S2 is as follows: , in, This represents the error transfer function of a two-degree-of-freedom temperature control system.

[0025] In this scheme, it can be seen from the calculation expression of the error transfer function of the two-degree-of-freedom temperature control system that the tracking error is eliminated by the main controller, the disturbance error is suppressed by the feedforward compensator, and the tracking error and disturbance error are completely decoupled. Here, the tracking error refers to the internal temperature tracking error, and the disturbance error refers to the temperature deviation caused by the disturbance.

[0026] In this scheme, the error transfer function of the two-degree-of-freedom temperature control system can be used to analyze the factors affecting the performance of the two-degree-of-freedom temperature control system of the polyacrylonitrile reactor. For target tracking performance, assuming the jacket temperature disturbance is 0, for a preset internal temperature value, to make the error transfer function of the two-degree-of-freedom temperature control system approach 0, the following requirements must be met. Very small, equivalent to loop transfer function It is large enough to be achieved by adjusting the parameters of the main controller. For the requirement of tracking the preset internal temperature value of the polyacrylonitrile reactor, such as temperature switching and steady-state maintenance, optimization can be used. The PID parameters are optimized to achieve a tracking effect of "no overshoot, fast response, and high precision".

[0027] Regarding anti-interference performance, the disturbance error is jointly determined by the main controller and the feedforward compensator. However, the parameters of the main controller are already tuned, and disturbance suppression can be achieved simply by optimizing the parameters of the feedforward compensator. The main controller's PID parameters include the first proportional... First point and the first differential The feedforward compensator PID parameters include the second proportional... Second integral Second differential Theoretically, let This achieves complete disturbance compensation, making the impact of the disturbance on the actual internal temperature value approach zero. However, such an ideal feedforward compensator is difficult to physically implement, and low-order inertial elements are often used to construct an engineered feedforward compensator. The calculation expression of the engineered feedforward compensator is as follows: , in, This refers to an engineered feedforward compensator. Represents the time constant. Indicates the order; in this embodiment, The value range is 1~2. The larger the value of the time constant, the faster the compensation speed. This is especially important for reactors with large hysteresis characteristics. By selecting a reasonably large value, jacket disturbances can be quickly suppressed while ensuring stability. In actual design, the feedforward compensator adopts a series structure of "PID control link combined with low-order inertial link". The BP neural network optimizes the parameters of the PID control link online. When the jacket temperature is disturbed, the feedforward compensation signal of the feedforward compensator can act in advance to offset the influence of the disturbance on the reactor internal temperature, so that the disturbance error decays quickly and avoids temperature fluctuations, thereby significantly improving the system's anti-interference capability and adapting to the situation of frequent disturbances in the reactor jacket.

[0028] Regarding robustness, for tracking error, optimizing the PID parameters of the main controller avoids significant fluctuations in tracking error due to parameter perturbations. For disturbance error, the low-order inertial element of the feedforward compensator can adapt to the large hysteresis characteristics of the reactor, effectively reducing the impact of pure hysteresis perturbations on disturbance error. Simultaneously, properly tuning the relevant parameters of the feedforward compensator can offset the error caused by gain perturbations of the controlled object. Compared to a single-degree-of-freedom system, a two-degree-of-freedom system, leveraging its error decoupling characteristics, can optimize the main controller and feedforward compensator separately, independently addressing the impact of parameter perturbations on tracking and disturbance errors. It maintains error stability even with model mismatch, exhibiting good robustness.

[0029] Two-degree-of-freedom control structurally resolves the coupling contradiction between tracking and anti-interference, and, in conjunction with feedforward compensation, achieves advance suppression of disturbances, effectively overcoming difficulties such as large reactor lag, thermal coupling, and model uncertainty.

[0030] The dual PID parameter group in S2 includes the main controller PID parameters and the feedforward compensator PID parameters, wherein the main controller PID parameters include a first proportional First point and the first differential The feedforward compensator PID parameters include the second proportional... Second integral Second differential .

[0031] S3. Obtain the operating status of the two-degree-of-freedom temperature control system; The operating status of the two-degree-of-freedom temperature control system in S3 includes the internal temperature tracking error, the rate of change of the internal temperature error, and the jacket disturbance observation value. ; The formula for calculating the internal temperature tracking error is as follows: , in, Indicates the first Internal temperature tracking error during the next iteration Indicates the first The preset internal temperature value for the next iteration Indicates the first The actual internal temperature value at the next iteration; The formula for calculating the rate of change of internal temperature error is as follows: , in, Indicates the first The rate of change of internal temperature error during each iteration. Indicates the first Internal temperature tracking error during the next iteration.

[0032] S4. Construct a two-degree-of-freedom control BP neural network; like Figure 4 As shown, in this scheme, the two-degree-of-freedom control BP neural network takes the state and disturbance observation results of the two-degree-of-freedom temperature control system as input, the dual PID parameter group of the main controller and the feedforward compensator as output, and the tracking error and disturbance error as learning targets to achieve synchronous online adaptive optimization of the PID parameters of the main controller and the feedforward compensator, so as to comprehensively improve the control system in terms of tracking, disturbance rejection and robustness. The two-degree-of-freedom control BP neural network in S4 adopts a three-layer neural network topology, including an input layer, a hidden layer, and an output layer. The input layer has 3 nodes, the hidden layer has M nodes, and the output layer has 6 nodes. The input signals corresponding to the three nodes of the input layer are the operating states of the two-degree-of-freedom temperature control system. In this scheme, the input signals of the BP neural network are set as the operating states of the two-degree-of-freedom temperature control system, including system dynamic errors and external disturbances. This enables the two-degree-of-freedom control BP neural network to actively sense disturbances and achieve advanced adjustment, which is highly compatible with the two-degree-of-freedom feedforward compensation mechanism. The hidden layer uses a sigmoid activation function for nonlinear feature mapping. In this scheme, through the multi-layer weight operation of the hidden layer, the two-degree-of-freedom control BP neural network can approximate the complex dynamic characteristics of the reactor, providing a basis for dual-loop parameter adaptation. The six nodes of the output layer correspond one-to-one with the PID parameters of the main controller and the PID parameters of the feedforward compensator in the dual PID parameter group.

[0033] In this scheme, the nodes of the output layer correspond to the dual PID parameter groups, realizing a one-to-one correspondence between the output structure and the two-degree-of-freedom decoupled architecture, ensuring independent adjustment and no interference between the main controller and the feedforward compensator.

[0034] S5. Construct a dual-error fusion performance index, and perform synchronous online adaptive optimization of the dual PID parameter group output by the two-degree-of-freedom control BP neural network according to the operating status of the two-degree-of-freedom temperature control system, and correspondingly control the output of the main controller and the feedforward compensator to complete the temperature adaptive control of the polyacrylonitrile reactor.

[0035] S5 includes the following steps: S51. Construct dual-error fusion performance indicators; The calculation expression for the dual-error fusion performance index in S51 is as follows: , in, Indicates the first The performance metrics of dual-error fusion at the next iteration Indicates the tracking error weight. Indicates the weight of the disturbance error. Indicates the first Temperature deviation caused by disturbance during the next iteration; In this scheme, minimizing tracking error and disturbance error are the dual learning objectives, so that the optimization direction of the two-degree-of-freedom control BP neural network is completely consistent with the two-degree-of-freedom control objective, realizing the deep integration of the two-degree-of-freedom document control system and the BP neural network; In this scheme, the tracking performance of the preset internal temperature value is optimized by the internal temperature tracking error, the performance of the known disturbance is optimized by the temperature deviation caused by the disturbance, and the weights of the tracking error and the disturbance error are used to balance the dual objective optimization weights; S52. Using the dual-error fusion performance index as the learning target, the operating state of the two-degree-of-freedom temperature control system is input into the two-degree-of-freedom control BP neural network to perform synchronous online adaptive optimization of the dual PID parameter group output by the two-degree-of-freedom control BP neural network, and the connection weights of the neurons in the two-degree-of-freedom BP neural network are updated using the gradient descent model. The calculation expression for the gradient descent model is as follows: , in, Indicates the first During the nth iteration Layer The update amount of the connection weights of each neuron Indicates the learning rate. Indicates the first During the next iteration for gradient, Indicates the first Layer The connection weights of each neuron Represents the momentum factor. Indicates the first During the nth iteration Layer The amount of update to the connection weights of each neuron; In this scheme, the gradient descent method combined with the momentum term is used to improve the network learning efficiency, and the connection weights of neurons in a two-degree-of-freedom BP neural network are corrected in real time to achieve online adaptive tuning of dual PID parameters without manual intervention and offline debugging. S53. Based on the dual PID parameter group output by the two-degree-of-freedom BP neural network, the outputs of the main controller and the feedforward compensator are controlled according to the incremental output model of the main controller and the incremental output model of the feedforward compensator, respectively, to complete the temperature adaptive control of the polyacrylonitrile reactor.

[0036] The calculation expressions for the incremental output model of the main controller and the incremental output model of the feedforward compensator in S53 are as follows: in, Indicates the first The main controller outputs an increment during the next iteration. Indicates the first The rate of change of internal temperature tracking error at each iteration Indicates the first The increment of the feedforward compensator output in the next iteration Indicates the first The increment of the observed value of the jacket perturbation at the next iteration Indicates the first The observed values ​​of the jacket perturbation at the next iteration. Indicates the first The increment of the rate of change of the perturbation at the next iteration. Indicates the first Rate of change of internal temperature error in each iteration Indicates the first The observed values ​​of the jacket perturbation at the next iteration. Indicates the first The increment of the observed value of the jacket perturbation at the next iteration.

[0037] This invention addresses the challenges of internal temperature control in polyacrylonitrile reactors, characterized by large hysteresis, strong coupling, and inaccurate models. It proposes a method combining a BP neural network with two-degree-of-freedom adaptive temperature control. This method leverages the complete decoupling between the two-degree-of-freedom preset internal temperature value tracking and the disturbance suppression structure, designing a 3-M-6 structured two-degree-of-freedom control BP neural network to achieve synchronous online self-tuning of six PID parameters in a dual-loop configuration. This deeply integrates the advantages of the two-degree-of-freedom structure with the adaptive characteristics of the BP neural network. Simulation comparisons show that this scheme, compared to traditional single-degree-of-freedom PID schemes and conventional two-degree-of-freedom control schemes, exhibits no overshoot, faster response, and superior disturbance rejection and robustness. It maintains ultra-high steady-state accuracy even under model mismatch, providing an effective solution for high-precision temperature control of reactors.

[0038] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for adaptive temperature control of a polyacrylonitrile reactor, characterized in that, Includes the following steps: S1. Construct a two-degree-of-freedom temperature control system for a polyacrylonitrile reactor, wherein the two-degree-of-freedom temperature control system includes a main control loop and a feedforward compensation loop. S2. Construct the error transfer function of the two-degree-of-freedom temperature control system, and perform performance analysis on the two-degree-of-freedom temperature control system to obtain the dual PID parameter set; S3. Obtain the operating status of the two-degree-of-freedom temperature control system; S4. Construct a two-degree-of-freedom control BP neural network; S5. Construct a dual-error fusion performance index, and perform synchronous online adaptive optimization of the dual PID parameter group output by the two-degree-of-freedom control BP neural network according to the operating status of the two-degree-of-freedom temperature control system, and correspondingly control the output of the main controller and the feedforward compensator to complete the temperature adaptive control of the polyacrylonitrile reactor.

2. The temperature adaptive control method for a polyacrylonitrile reactor according to claim 1, characterized in that, The calculation expression for the two-degree-of-freedom temperature control system in S1 is as follows: ， in, This indicates the actual internal temperature value. Indicates the main controller. Indicates the controlled object, This indicates the preset internal temperature value. Indicates feedforward compensator. This represents the jacket temperature disturbance value, where, For Laplace variables.

3. The temperature adaptive control method for a polyacrylonitrile reactor according to claim 2, characterized in that, The calculation expression for the error transfer function of the two-degree-of-freedom temperature control system in S2 is as follows: ， in, This represents the error transfer function of a two-degree-of-freedom temperature control system.

4. The temperature adaptive control method for a polyacrylonitrile reactor according to claim 3, characterized in that, The dual PID parameter group in S2 includes the main controller PID parameters and the feedforward compensator PID parameters, wherein the main controller PID parameters include a first proportional First point and the first differential The feedforward compensator PID parameters include the second proportional... Second integral Second differential .

5. The temperature adaptive control method for a polyacrylonitrile reactor according to claim 4, characterized in that, The operating status of the two-degree-of-freedom temperature control system in S3 includes the internal temperature tracking error, the rate of change of the internal temperature error, and the jacket disturbance observation value. ; The formula for calculating the internal temperature tracking error is as follows: ， in, Indicates the first Internal temperature tracking error during the next iteration Indicates the first The preset internal temperature value for the next iteration Indicates the first The actual internal temperature value at the next iteration; The formula for calculating the rate of change of internal temperature error is as follows: ， in, Indicates the first The rate of change of internal temperature error during each iteration. Indicates the first Internal temperature tracking error during the next iteration.

6. The temperature adaptive control method for a polyacrylonitrile reactor according to claim 5, characterized in that, The two-degree-of-freedom control BP neural network in S4 adopts a three-layer neural network topology, including an input layer, a hidden layer, and an output layer. The input layer has 3 nodes, the hidden layer has M nodes, and the output layer has 6 nodes. The input signals corresponding to the three nodes of the input layer are the operating status of the two-degree-of-freedom temperature control system. The hidden layer uses a sigmoid activation function for nonlinear feature mapping. The six nodes of the output layer correspond one-to-one with the PID parameters of the main controller and the PID parameters of the feedforward compensator in the dual PID parameter group.

7. The temperature adaptive control method for a polyacrylonitrile reactor according to claim 6, characterized in that, S5 includes the following steps: S51. Construct dual-error fusion performance indicators; The calculation expression for the dual-error fusion performance index in S51 is as follows: ， in, Indicates the first The performance metrics of dual-error fusion at the next iteration Indicates the tracking error weight. Indicates the weight of the disturbance error. Indicates the first Temperature deviation caused by perturbation during the next iteration; S52. Using the dual-error fusion performance index as the learning target, the operating state of the two-degree-of-freedom temperature control system is input into the two-degree-of-freedom control BP neural network to perform synchronous online adaptive optimization of the dual PID parameter group output by the two-degree-of-freedom control BP neural network, and the connection weights of the neurons in the two-degree-of-freedom BP neural network are updated using the gradient descent model. The calculation expression for the gradient descent model is as follows: ， in, Indicates the first During the nth iteration Layer The update amount of the connection weights of each neuron Indicates the learning rate. Indicates the first During the next iteration for gradient, Indicates the first Layer The connection weights of each neuron Represents the momentum factor. Indicates the first During the nth iteration Layer The amount of update to the connection weights of each neuron; S53. Based on the dual PID parameter group output by the two-degree-of-freedom BP neural network, the outputs of the main controller and the feedforward compensator are controlled according to the incremental output model of the main controller and the incremental output model of the feedforward compensator, respectively, to complete the temperature adaptive control of the polyacrylonitrile reactor.

8. The temperature adaptive control method for a polyacrylonitrile reactor according to claim 7, characterized in that, The calculation expressions for the incremental output model of the main controller and the incremental output model of the feedforward compensator in S53 are as follows: in, Indicates the first The main controller outputs an increment during the next iteration. Indicates the first The rate of change of internal temperature tracking error at each iteration Indicates the first The increment of the feedforward compensator output in the next iteration Indicates the first The increment of the observed value of the jacket perturbation at the next iteration Indicates the first The observed values ​​of the jacket perturbation at the next iteration. Indicates the first The increment of the rate of change of the perturbation at the next iteration. Indicates the first Rate of change of internal temperature error in each iteration Indicates the first The observed values ​​of the jacket perturbation at the next iteration. Indicates the first The increment of the observed value of the jacket perturbation at the next iteration.

Images