A density control system for rapid heavy medium coal preparation based on QP-DMC

Abstract

The present invention provides a fast heavy medium coal preparation density control system based on QP-DMC, which is used in heavy medium coal preparation equipment, including a sub-range control module, a QP-DMC control model and a control actuator; and also includes a feedforward decoupling module, which decomposes a multivariable prediction task into multiple single variable prediction tasks by retaining important factors and performing feedforward compensation on influencing factors under different sub-ranges divided by the sub-range control module, and then combines the prediction results of each single variable to obtain an overall prediction result, and combines the QP-DMC control model to obtain a control instruction, and issues a control command to the control actuator to achieve rapid prediction and control, and reduce the sensitivity of the controller to disturbances and model mismatch. This system can simplify the algorithm calculation amount, improve the real-time performance of the control system, and speed up the control response speed, thereby realizing real-time online adjustment of the density of heavy medium coal preparation.

Description

Technical Field

[0001] The invention relates to the technical field of heavy medium coal preparation density control, and specifically provides a rapid heavy medium coal preparation density control system based on QP-DMC. Background Art

[0002] The realization of intelligentization of the heavy medium coal preparation process can increase the output of clean coal, reduce the amount of coal in the gangue, thereby improving production efficiency, and significantly improving the stability and accuracy of the heavy medium coal preparation system. The key to the intelligentization of heavy medium coal preparation is to realize the control of the density of the suspension. At present, simple PID control methods are usually used in industrial sites, but their control effect is not ideal. This is because the control of heavy medium density involves the mutual coupling of the density and liquid level of the suspension. Adjusting the opening of the diverter valve, the water supply valve and the medium addition valve will affect the density and liquid level of the suspension and cannot be considered separately. In addition, the control objects of the density and liquid level of the suspension belong to typical large inertia and large lag processes, and simple PID algorithms cannot control such objects well. Therefore, it is necessary to adopt a multivariable predictive control method for unified analysis and control to adapt to the actual characteristics of the heavy medium separation process. However, the common multivariable predictive control method currently used for density control of heavy medium coal preparation has problems such as large calculation amount, long calculation time and difficulty in realizing online real-time control. Therefore, it is very necessary to study a fast density control system for heavy medium coal preparation. Summary of the invention

[0003] In view of the problem in the prior art that ordinary predictive control algorithms in the heavy medium coal preparation process have large online calculation amount and long calculation time, the present invention provides a rapid heavy medium coal preparation density control system based on QP-DMC, a rapid heavy medium coal preparation density control system based on QP-DMC, which is used in heavy medium coal preparation equipment, at sampling times at preset time intervals, to obtain real-time suspension density data and real-time suspension liquid level data respectively through a densitometer and a liquid level meter arranged at preset positions, and adjust the diverter valve opening, the medium addition valve opening, and the water supply valve opening accordingly according to the real-time suspension density data and the real-time suspension liquid level data; it includes a split-range control module, a QP-DMC control model and a control actuator.

[0004] The control actuator is the component in the control system that is responsible for executing control commands. It receives control commands and converts them into corresponding control signals, thereby adjusting the opening of the diverter valve, water make-up valve, and medium make-up valve in the heavy medium coal preparation process to control the density and liquid level of the suspension to fluctuate within a predetermined range.

[0005] Specifically, the control actuator receives the control command from the split-range controller and converts it into a corresponding control signal, and then drives the corresponding actuator through the actuator to adjust the opening of the diverter valve, water supply valve, and medium supply valve, thereby achieving control of the suspension density and liquid level.

[0006] In the heavy medium coal preparation process, the flow, density and liquid level of the suspension can be controlled by adjusting the opening of the diverter valve, water supply valve and medium supply valve through the control actuator, thereby ensuring the stability and efficiency of the heavy medium coal preparation process. At the same time, the rapid response capability of the control actuator can also improve the response speed and accuracy of the control system, thereby further improving the control effect of the heavy medium coal preparation process.

[0007] The range control module divides different ranges based on the preset suspension density setting value, and for each range, according to the real-time suspension density data and real-time suspension level data, uses the QP-DMC control model to adjust the diverter valve opening, the medium valve opening and the water supply valve opening in real time, and issues control commands.

[0008] The different sub-processes are divided based on the preset suspension density setting value as follows:

[0009] When the real-time suspension density data y2 is less than the set value, the first sub-process is entered. According to the deviation between the real-time suspension level data y1 and the suspension level preset value, as well as the deviation between the real-time suspension density data y2 and the suspension density set value, the diverter valve opening u1 and the addition valve opening u2 are adjusted.

[0010] (2) When the real-time suspension density data y2 is greater than the set value, the second sub-process is entered. According to the deviation between the real-time suspension level data y1 and the suspension level preset value, as well as the deviation between the real-time suspension density data y2 and the suspension density set value, the diverter valve u1 and the water supply valve opening u3 are adjusted.

[0011] The QP-DMC control model is combined with the split-range control module, and a multivariable QP-DMC control model is obtained through system identification experiments, which specifically includes the following steps:

[0012] Select the diverter valve opening u1 as the system identification input signal, record the changes of the suspension level y1 and the suspension density y2; use the input and output data obtained from the system identification test to establish the transfer function G between the suspension level y1 and the diverter valve opening u1 11 , the transfer function G of the suspension density y2 and the diverter valve opening u1 21 ; Select the medium valve opening u2, act on the suspension level y1 and the suspension density y2 respectively, repeat the system identification test, and get the transfer function model of the control relationship corresponding to the first sub-range:

[0013]

[0014] Select other control factors in the heavy medium separation process, the water supply valve opening u3, act on the suspension level y1 object and the suspension density y2 object respectively, repeat the system identification test, and get the transfer function model of the control relationship corresponding to the second sub-process:

[0015]

[0016] According to the established transfer function models corresponding to the first sub-range and the second sub-range respectively, a multivariable split-range QP-DMC controller is designed. The multivariable split-range QP-DMC controller includes a multivariable first sub-range QP-DMC prediction model and a multivariable second sub-range QP-DMC prediction model, which are used to control the fluctuations of the heavy medium suspension density y2 and the suspension liquid level y1 within a predetermined range.

[0017] Among them, the first-range transfer function model is discretized to obtain a multivariable first-range QP-DMC prediction model to predict the changing trends of the heavy medium suspension level y1 and the suspension density y2 under the regulation of the diverter valve opening u1 and the addition valve opening u2.

[0018] The second-range transfer function model is discretized to obtain the multivariable second-range QP-DMC prediction model to predict the change trend of the heavy medium suspension level y1 and the suspension density y2 under the regulation of the diverter valve opening u1 and the water supply valve opening u3.

[0019] The multivariable QP-DMC control model is combined with a feedforward decoupling module. For each sub-range divided by the sub-range control module, the feedforward decoupling module is used to retain the important factors in the sub-range and perform feedforward compensation on the influencing factors, decompose the multivariable prediction task into multiple single-variable prediction tasks, and then combine the single-variable prediction results to obtain the overall prediction result, and combine the QP-DMC control model to obtain the control instruction, which is sent to the control actuator to achieve rapid prediction and control.

[0020] In order to obtain multiple single variable split-range QP-DMC fast controllers, the following steps are specifically included:

[0021] For the first sub-range, the diverter valve opening u1 is taken as the main factor to adjust the suspension level y1, the medium valve opening u2 is taken as the influencing factor to adjust the suspension level y1, the medium valve opening u2 is taken as the main factor to adjust the suspension density y2, and the diverter valve opening u1 is taken as the influencing factor to adjust the suspension density y2. The overall prediction model is:

[0022]

[0023]

[0024] Where y1(k) is the measured data of the suspension level y1 at time k, and the column vector is the predicted numerical sequence of the suspension level y1 at time k+1, ..., k+M, ..., k+P, column vector is the historical numerical sequence of the suspension level y1 before time k, P and M represent the prediction time domain and control time domain of the predictive control algorithm respectively, y2(k) is the measured data of the suspension density at time k, and the column vector is the predicted numerical sequence of the suspension density y2 at time k+1, ..., k+M, ..., k+P, column vector is the historical numerical sequence of the suspension density y2 at time k, u1(k) is the value of the diverter valve opening u1 at time k, Δu1(k) = u1(k) - u1(k-1) is the increment of the diverter valve opening u1 before time k, and the column vector is the control increment sequence of the diverter valve opening u1 at time k, ..., k+M, u2(k) is the value of the intermediate valve opening u2 at time k, Δu2(k) = u2(k) - u2(k-1) is the increment of the intermediate valve opening u2 at time k, and the column vector is the control increment sequence of the valve opening u2 at time k, ..., k+M; define the matrix A 11 is the dynamic matrix of the control of the diverter valve opening u1 variable on the suspension level y1,

[0025]

[0026] Where [a 11 (1), …, a 11 (P)] is the step response sequence of the suspension level y1 in response to the change of the diverter valve opening u1; the definition matrix A 22 is the dynamic matrix of the control of the suspension density y2 by the variable of the medium valve opening u2,

[0027]

[0028] Where [a 22 (1), …, a 22 (P)] is the step response sequence of the suspension density y2 in response to the change of the medium valve opening u2; the prediction model formula [a 12 (1), …, a 12 (P)] is the step response sequence of the suspension level y1 in response to the change of the medium valve opening u2, [a 21 (1), …, a 21 (P)] is the step response sequence of the suspension density y2 in response to the change of the diverter valve opening u1.

[0029] For the second sub-range, the diverter valve opening u1 is taken as the main factor to adjust the suspension level y1, the water supply valve opening u3 is taken as the influencing factor to adjust the suspension level y1, the water supply valve opening u3 is taken as the main factor to adjust the suspension density y2, and the diverter valve opening u1 is taken as the influencing factor to adjust the suspension density y2. The overall prediction model is:

[0030]

[0031]

[0032] Where u3(k) is the value of the water supply valve opening u3 at time k, Δu3(k) = u3(k) - u3(k-1) is the increment of the water supply valve opening u3 at time k, and the column vector is the control increment sequence of the water supply valve opening u3 at time k, ..., k+M; define the matrix A 23 is the dynamic matrix of the water supply valve opening u3 variable controlling the suspension density y2,

[0033]

[0034] Where [a 23 (1), …, a 23 (P)] is the step response sequence of the suspension density y2 in response to the change of the water supply valve opening u3; the prediction model formula [a 13 (1), …, a 13 (P)] is the step response sequence of the suspension level y1 in response to the change of the water supply valve opening u3.

[0035] Based on the QP-DMC problem, the QP-DMC problem includes performance indicators and the following contents: Multi-objective optimization problem: The QP-DMC problem is usually a multi-objective optimization problem, which requires optimizing multiple performance indicators at the same time, such as control speed, control accuracy, energy consumption, etc.

[0036] Nonlinear problems: Nonlinear problems in QP-DMC problems usually come from the nonlinear characteristics of the controlled object or the nonlinear characteristics of the control algorithm.

[0037] Robustness problem: The QP-DMC problem needs to have a certain degree of robustness, that is, it can still maintain good control performance in the presence of uncertainties or disturbances.

[0038] Real-time problem: QP-DMC problem needs to be solved under the condition of high real-time requirement, such as real-time control problem in industrial process control.

[0039] Large-scale problems: QP-DMC problems usually involve large-scale state spaces and control input spaces, requiring the use of efficient optimization algorithms and computational methods.

[0040] The present invention aims at multiple single-variable split-range QP-DMC fast controller performance indicators. In order to determine the setting ranges of the control variables diverter valve opening u1, the medium valve opening u2 and the water supply valve opening u3, thereby ensuring that the controlled variables suspension level and suspension density fluctuate within a predetermined range, multiple single-variable split-range QP-DMC fast controller performance indicators are constructed for each split-range process.

[0041] For the first sub-range, multiple single variable sub-range QP-DMC fast controllers are constructed with the following performance indicators:

[0042]

[0043]

[0044] Where: Matrix A′ 11 is the dynamic matrix A 11 The transpose of the matrix A′ 22 is the dynamic matrix A 22 The transpose of , column vector is the step response sequence of the suspension level y1 in response to the change of the medium valve opening u2, and the column vector is the step response sequence of the suspension density y2 in response to the change of the diverter valve opening u1, and the column vector is the target value sequence of the suspension level y1 at time k, column vector is the target value sequence of suspension density y2 at time k, matrix Q is the controlled variable error weighting matrix in multivariable predictive control, and matrix R is the manipulated variable incremental control weighting matrix in multivariable predictive control.

[0045] For the second split range, multiple single variable split range QP-DMC fast controller indicators are constructed as follows:

[0046]

[0047]

[0048] Where: Matrix A′ 23 is the dynamic matrix A 23 The transpose of , column vector It is the step response sequence of the suspension level y1 in response to the change of the water supply valve opening u3.

[0049] In one embodiment, the system identification test may be a step response test, or a pulse response test or other testing methods.

[0050] In the system identification test, in the process of establishing the transfer function model, the collected data needs to be processed, such as filtering, fitting and other operations, to obtain a more accurate model.

[0051] In the system identification test, in the process of designing a suitable multivariable predictive controller, other multivariable controllers such as DMC, GPC, MAC, etc. can also be used to simultaneously control the changes in the heavy medium suspension level and density.

[0052] Among them, the step-by-step prediction model is trained and optimized based on the system identification method.

[0053] The invention provides a fast heavy medium coal preparation density control system based on QP-DMC, which has the following beneficial effects:

[0054] This fast heavy medium coal preparation density control system based on QP-DMC can bring the following beneficial effects:

[0055] 1. Improve the efficiency of heavy medium coal separation equipment: By adjusting the opening of the diverter valve, the medium adding valve and the water supply valve in real time, the operating parameters of the equipment can be adjusted in time according to the suspension density data and the real-time suspension level data, thereby improving the separation efficiency and stability of the equipment.

[0056] 2. By real-time monitoring of suspension density and liquid level data, the operating parameters of the equipment can be adjusted in time to improve the quality and stability of the product. By introducing control actuators, the equipment can be automatically controlled, human intervention can be reduced, and production efficiency and safety can be improved.

[0057] 3. Through the design of the feedforward decoupling module, the uncertainty and interference in the multivariable prediction task can be reduced, thereby improving the reliability and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is the main step of the present invention;

[0059] Figure 2 The flowchart of the fast heavy medium suspension density control system based on the QP-DMC algorithm of the present invention;

[0060] Figure 3 This is the Matlab simulation diagram of the first subprocess;

[0061] Figure 4 The second step Matlab simulation diagram;

[0062] Figure 5 This is the simulation diagram of the suspension density control loop;

[0063] Figure 6 Simulation diagram of suspension level control loop. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] like Figure 1-2 As shown, in the heavy medium coal preparation equipment, at the sampling moment of the preset time interval, the diverter valve opening, the medium adding valve opening, and the water supply valve opening are adjusted accordingly according to the real-time suspension density data and the real-time suspension level data; it is characterized in that it includes a split-range control module, a QP-DMC control model and a control actuator.

[0066] The range control module divides different ranges based on the preset suspension density setting value, and for each range, according to the real-time suspension density data and real-time suspension level data, uses the QP-DMC control model to adjust the diverter valve opening, the medium valve opening and the water supply valve opening in real time, and issues control commands.

[0067] The control actuator receives the control command and converts it into a corresponding control signal, and makes real-time adjustments to the opening of the diverter valve, the medium valve, and the water supply valve.

[0068] The different sub-processes are divided based on the preset suspension density setting value as follows:

[0069] When the real-time suspension density data y2 is less than the set value, the first sub-process is entered. According to the deviation between the real-time suspension level data y1 and the suspension level preset value, as well as the deviation between the real-time suspension density data y2 and the suspension density set value, the diverter valve opening u1 and the addition valve opening u2 are adjusted.

[0070] When the real-time suspension density data y2 is greater than the set value, the second sub-process is entered. According to the deviation between the real-time suspension level data y1 and the suspension level preset value, as well as the deviation between the real-time suspension density data y2 and the suspension density set value, the size of the diverter valve u1 and the water supply valve opening u3 are adjusted.

[0071] The QP-DMC control model is combined with the split-range control module, and a multivariable QP-DMC control model is obtained through system identification experiments, which specifically includes the following steps:

[0072] Select the diverter valve opening u1 as the system identification input signal, record the changes of the suspension level y1 and the suspension density y2; use the input and output data obtained from the system identification test to establish the transfer function G between the suspension level y1 and the diverter valve opening u1 11 , the transfer function G of the suspension density y2 and the diverter valve opening u121 ; Select the medium valve opening u2, act on the suspension level y1 and the suspension density y2 respectively, repeat the system identification test, and get the transfer function model of the control relationship corresponding to the first sub-range:

[0073]

[0074] Select other control factors in the heavy medium separation process, the water supply valve opening u3, act on the suspension level y1 object and the suspension density y2 object respectively, repeat the system identification test, and get the transfer function model of the control relationship corresponding to the second sub-process:

[0075]

[0076] According to the established transfer function models corresponding to the first sub-range and the second sub-range respectively, a multivariable split-range QP-DMC controller is designed. The multivariable split-range QP-DMC controller includes a multivariable first sub-range QP-DMC prediction model and a multivariable second sub-range QP-DMC prediction model, which are used to control the fluctuations of the heavy medium suspension density y2 and the suspension liquid level y1 within a predetermined range.

[0077] Among them, the first-range transfer function model is discretized to obtain a multivariable first-range QP-DMC prediction model to predict the changing trends of the heavy medium suspension level y1 and the suspension density y2 under the regulation of the diverter valve opening u1 and the addition valve opening u2.

[0078] The second-range transfer function model is discretized to obtain the multivariable second-range QP-DMC prediction model to predict the change trend of the heavy medium suspension level y1 and the suspension density y2 under the regulation of the diverter valve opening u1 and the water supply valve opening u3.

[0079] The multivariable QP-DMC control model is combined with a feedforward decoupling module. For each sub-range divided by the sub-range control module, the feedforward decoupling module is used to retain important factors in the sub-range and perform feedforward compensation on the influencing factors, decompose the multivariable prediction task into multiple single-variable prediction tasks, and then combine the single-variable prediction results to obtain the overall prediction result, and combine the QP-DMC control model to obtain the control instruction, and send the control instruction to the control actuator to achieve rapid prediction and control.

[0080] Obtaining multiple single variable split-range QP-DMC fast controllers specifically includes the following steps:

[0081] For the first sub-range, the diverter valve opening u1 is taken as the main factor to adjust the suspension level y1, the medium valve opening u2 is taken as the influencing factor to adjust the suspension level y1, the medium valve opening u2 is taken as the main factor to adjust the suspension density y2, and the diverter valve opening u1 is taken as the influencing factor to adjust the suspension density y2. The overall prediction model is:

[0082]

[0083]

[0084] Where y1(k) is the measured data of the suspension level y1 at time k, and the column vector is the predicted numerical sequence of the suspension level y1 at time k+1, ..., k+M, ..., k+P, column vector is the historical numerical sequence of the suspension level y1 before time k, P and M represent the prediction time domain and control time domain of the predictive control algorithm respectively, y2(k) is the measured data of the suspension density at time k, and the column vector is the predicted numerical sequence of the suspension density y2 at time k+1, ..., k+M, ..., k+P, column vector is the historical numerical sequence of the suspension density y2 at time k, u1(k) is the value of the diverter valve opening u1 at time k, Δu1(k) = u1(k) - u1(k-1) is the increment of the diverter valve opening u1 before time k, and the column vector is the control increment sequence of the diverter valve opening u1 at time k, ..., k+M, u2(k) is the value of the intermediate valve opening u2 at time k, Δu2(k) = u2(k) - u2(k-1) is the increment of the intermediate valve opening u2 at time k, and the column vector is the control increment sequence of the valve opening u2 at time k, ..., k+M; define the matrix A 11 is the dynamic matrix of the control of the diverter valve opening u1 variable on the suspension level y1,

[0085]

[0086] Where [a 11 (1), …, a 11 (P)] is the step response sequence of the suspension level y1 in response to the change of the diverter valve opening u1; the definition matrix A 22 is the dynamic matrix of the control of the suspension density y2 by the variable of the medium valve opening u2,

[0087]

[0088] Where [a 22 (1), …, a 22(P)] is the step response sequence of the suspension density y2 in response to the change of the medium valve opening u2; the prediction model formula [a 12 (1), …, a 12 (P)] is the step response sequence of the suspension level y1 in response to the change of the medium valve opening u2, [a 21 (1), …, a 21 (P)] is the step response sequence of the suspension density y2 in response to the change of the diverter valve opening u1.

[0089] For the second sub-range, the diverter valve opening u1 is taken as the main factor to adjust the suspension level y1, the water supply valve opening u3 is taken as the influencing factor to adjust the suspension level y1, the water supply valve opening u3 is taken as the main factor to adjust the suspension density y2, and the diverter valve opening u1 is taken as the influencing factor to adjust the suspension density y2. The overall prediction model is:

[0090]

[0091]

[0092] Where u3(k) is the value of the water supply valve opening u3 at time k, Δu3(k) = u3(k) - u3(k-1) is the increment of the water supply valve opening u3 at time k, and the column vector is the control increment sequence of the water supply valve opening u3 at time k, ..., k+M; define the matrix A 23 is the dynamic matrix of the water supply valve opening u3 variable controlling the suspension density y2,

[0093]

[0094] Where [a 23 (1), …, a 23 (P)] is the step response sequence of the suspension density y2 in response to the change of the water supply valve opening u3; the prediction model formula [a 13 (1), …, a 13 (P)] is the step response sequence of the suspension level y1 in response to the change of the water supply valve opening u3.

[0095] Based on the QP-DMC problem, the QP-DMC problem includes performance indicators and the following contents: Multi-objective optimization problem: The QP-DMC problem is usually a multi-objective optimization problem, which requires optimizing multiple performance indicators at the same time, such as control speed, control accuracy, energy consumption, etc.

[0096] Nonlinear problems: Nonlinear problems in QP-DMC problems usually come from the nonlinear characteristics of the controlled object or the nonlinear characteristics of the control algorithm.

[0097] Robustness problem: The QP-DMC problem needs to have a certain degree of robustness, that is, it can still maintain good control performance in the presence of uncertainties or disturbances.

[0098] Real-time problem: QP-DMC problem needs to be solved under the condition of high real-time requirement, such as real-time control problem in industrial process control.

[0099] Large-scale problems: QP-DMC problems usually involve large-scale state spaces and control input spaces, requiring the use of efficient optimization algorithms and computational methods.

[0100] The present invention aims at multiple single-variable split-range QP-DMC fast controller performance indicators. In order to determine the setting ranges of the control variables diverter valve opening u1, the medium valve opening u2 and the water supply valve opening u3, thereby ensuring that the controlled variables suspension level and suspension density fluctuate within a predetermined range, multiple single-variable split-range QP-DMC fast controller performance indicators are constructed for each split-range process.

[0101] For the first sub-range, multiple single variable sub-range QP-DMC fast controllers are constructed with the following performance indicators:

[0102]

[0103]

[0104] Where: Matrix A′ 11 is the dynamic matrix A 11 The transpose of the matrix A′ 22 is the dynamic matrix A 22 The transpose of , column vector is the step response sequence of the suspension level y1 in response to the change of the medium valve opening u2, and the column vector is the step response sequence of the suspension density y2 in response to the change of the diverter valve opening u1, and the column vector is the target value sequence of the suspension level y1 at time k, column vector is the target value sequence of suspension density y2 at time k, matrix Q is the controlled variable error weighting matrix in multivariable predictive control, and matrix R is the manipulated variable incremental control weighting matrix in multivariable predictive control.

[0105] For the second split range, multiple single variable split range QP-DMC fast controller indicators are constructed as follows:

[0106]

[0107]

[0108] Where: Matrix A′ 23 is the dynamic matrix A23 The transpose of , column vector It is the step response sequence of the suspension level y1 in response to the change of the water supply valve opening u3.

[0109] Among them, in the system identification test, the system identification test can be a step response test, or a pulse response test or other testing methods.

[0110] In the system identification test, in the process of establishing the transfer function model, the collected data needs to be processed, such as filtering, fitting and other operations, to obtain a more accurate model.

[0111] In the system identification test, in the process of designing a suitable multivariable predictive controller, other multivariable controllers such as DMC, GPC, MAC, etc. can also be used to simultaneously control the changes in the heavy medium suspension level and density.

[0112] Among them, the step-by-step prediction model is trained and optimized based on the system identification method.

[0113] In one embodiment, through system identification test, the diverter valve opening u1, the medium adding valve opening u2, and the water supply valve opening u3 in the heavy medium coal separation process are used as step quantities to obtain the step response values ​​of the suspension real-time liquid level data y1 and the suspension real-time density data y2, and the heavy medium separation process transfer function is obtained as follows:

[0114]

[0115] (1) When the density is less than the set value, enter the first sub-stage. In this case, assuming that the initial density of the heavy medium is 1.2 kg / L, the required set density is 1.5 kg / L, the initial liquid level of the combined medium tank is 1.6 m, and the Matlab simulation results when the combined medium tank level is required to be stable at 2.5 m are as follows: Figure 3 shown.

[0116] (2) When the density is greater than the set value, the second sub-stage is entered. In this case, assuming that the initial density of the heavy medium is 1.8 kg / L, the required set density is 1.5 kg / L, the initial liquid level of the combined medium tank is 1.6 m, and the Matlab simulation results when the combined medium tank level is required to be stable at 2.5 m are as follows: Figure 4 shown.

[0117] In order to illustrate the control effect of rapid heavy medium coal preparation density control, a comparison was made with the traditional PID control algorithm. Figure 5 It is the Matlab simulation of the traditional PID algorithm in the density control loop and the liquid level control loop.

[0118] Depend on Figure 5 and Figure 6It can be seen that under the traditional PID control, the density and liquid level of the system have a large overshoot, and the adjustment time is about 300s, while under the control of the present invention, the adjustment time is at most 200s. It can be seen that the control effect of the control method of the present invention on the density and liquid level of the heavy medium suspension is better than that of the simple PID control effect.

[0119] The method of the present invention splits the rolling optimization in the ordinary multivariable predictive control into two decentralized optimizations, and the computational complexity changes from ○(2, M) to 2*O(1, M). The Matlab simulation computation time of the method of the present invention is 0.253 seconds, and the ordinary QP-DMC algorithm computation time is 1.147 seconds. The method of the present invention can simplify the computation process, shorten the computation time, meet the real-time requirements of the control system, and achieve the purpose of rapid control.

Claims

1. A fast heavy medium coal preparation density control system based on QP-DMC, which is used in heavy medium coal preparation equipment to adjust the diverter valve opening, the medium addition valve opening, and the water supply valve opening accordingly according to the real-time suspension density data and the real-time suspension level data at the sampling time of the preset time interval; characterized in that: It includes split-range control module, QP-DMC control model, feedforward decoupling module and control actuator; The range control module divides different ranges based on the preset suspension density setting value, and for each range, according to the real-time suspension density data and the real-time suspension level data, the QP-DMC control model is used to adjust the diverter valve opening, the medium adding valve opening and the water supply valve opening in real time, and issue a control command; the control actuator receives the control command and converts it into a corresponding control signal, and adjusts the diverter valve opening, the medium adding valve opening and the water supply valve opening in real time; The feedforward decoupling module is used for each sub-process to retain the important factors in the sub-process and perform feedforward compensation on the influencing factors, decompose the multivariable prediction task into multiple single variable prediction tasks, and then combine the prediction results of each single variable to obtain the overall prediction result, and combine the QP-DMC control model to obtain the control command, which is sent to the control actuator to achieve rapid prediction and control; First, the split-range control module is divided into the first split-range and the second split-range according to the size relationship between the real-time suspension density data and the set value, and then the transfer function models corresponding to the first split-range and the second split-range are constructed respectively; then, by discretizing the transfer function model, a multivariable first split-range QP-DMC prediction model and a multivariable second split-range QP-DMC prediction model are obtained, and combined into a multivariable split-range QP-DMC controller; then, the multivariable QP-DMC control model is combined with the feedforward decoupling module to obtain multiple single-variable split-range QP-DMC fast controllers; finally, for each split-range, multiple single-variable split-range QP-DMC fast controller performance indicators are constructed under the split-range, and then the setting ranges of the control variables diverter valve opening u1, medium valve opening u2 and water supply valve opening u3 under the split-range are determined; The QP-DMC control model is combined with the split-range control module, and a multivariable QP-DMC control model is obtained through system identification experiments, which specifically includes the following steps: Select the diverter valve opening u1 as the system identification input signal, record the changes of the suspension level y1 and the suspension density y2; use the input and output data obtained from the system identification test to establish the transfer function G between the suspension level y1 and the diverter valve opening u1 11 , the transfer function G of the suspension density y2 and the diverter valve opening u1 21 ; Select the medium valve opening u2, act on the suspension level y1 and the suspension density y2 respectively, repeat the system identification test, and get the transfer function model of the control relationship corresponding to the first sub-range: Select other control factors in the heavy medium separation process, the water supply valve opening u3, act on the suspension level y1 object and the suspension density y2 object respectively, repeat the system identification test, and get the transfer function model of the control relationship corresponding to the second sub-process: According to the established transfer function models corresponding to the first sub-range and the second sub-range, respectively, a multivariable sub-range QP-DMC controller is designed, wherein the multivariable sub-range QP-DMC controller includes a multivariable first sub-range QP-DMC prediction model and a multivariable second sub-range QP-DMC prediction model, which are used to control the density y2 of the heavy medium suspension and the liquid level y1 of the suspension to fluctuate within a predetermined range; The first-range transfer function model is discretized to obtain a multivariable first-range QP-DMC prediction model to predict the change trend of the heavy medium suspension level y1 and the suspension density y2 under the regulation of the diverter valve opening u1 and the addition valve opening u2; The second-range transfer function model is discretized to obtain a multivariable second-range QP-DMC prediction model to predict the change trend of the heavy medium suspension level y1 and the suspension density y2 under the regulation of the diverter valve opening u1 and the water supply valve opening u3; The multivariable QP-DMC control model is combined with a feedforward decoupling module to obtain multiple single variable split-range QP-DMC fast controllers, which specifically includes the following steps: For the first sub-range, the diverter valve opening u1 is taken as the main factor to adjust the suspension level y1, the medium valve opening u2 is taken as the influencing factor to adjust the suspension level y1, the medium valve opening u2 is taken as the main factor to adjust the suspension density y2, and the diverter valve opening u1 is taken as the influencing factor to adjust the suspension density y2. The overall prediction model is: Where y1(k) is the measured data of the suspension level y1 at time k, and the column vector is the predicted numerical sequence of the suspension level y1 at time k+1, ..., k+M, ..., k+P, column vector is the historical numerical sequence of the suspension level y1 before time k, P and M represent the prediction time domain and control time domain of the predictive control algorithm respectively, y2(k) is the measured data of the suspension density at time k, and the column vector is the predicted numerical sequence of the suspension density y2 at time k+1, ..., k+M, ..., k+P, column vector is the historical numerical sequence of the suspension density y2 at time k, u1(k) is the value of the diverter valve opening u1 at time k, Δu1(k) = u1(k) - u1(k-1) is the increment of the diverter valve opening u1 before time k, and the column vector is the control increment sequence of the diverter valve opening u1 at time k, ..., k+M, u2(k) is the value of the intermediate valve opening u2 at time k, Δu2(k) = u2(k) - u2(k-1) is the increment of the intermediate valve opening u2 at time k, and the column vector is the control increment sequence of the valve opening u2 at time k, ..., k+M; define the matrix A 11 is the dynamic matrix of the control of the diverter valve opening u1 variable on the suspension level y1, Where [a 11 (1), …, a 11 (P)] is the step response sequence of the suspension level y1 in response to the change of the diverter valve opening u1; the definition matrix A 22 is the dynamic matrix of the control of the suspension density y2 by the variable of the medium valve opening u2, Where [a 22 (1), …, a 22 (P)] is the step response sequence of the suspension density y2 in response to the change of the medium valve opening u2; the prediction model formula [a 12 (1), …, a 12 (P)] is the step response sequence of the suspension level y1 in response to the change of the medium valve opening u2, [a 21 (1), …, a 21 (P)] is the step response sequence of the suspension density y2 in response to the change of the diverter valve opening u1; For the second sub-range, the diverter valve opening u1 is taken as the main factor to adjust the suspension level y1, the water supply valve opening u3 is taken as the influencing factor to adjust the suspension level y1, the water supply valve opening u3 is taken as the main factor to adjust the suspension density y2, and the diverter valve opening u1 is taken as the influencing factor to adjust the suspension density y2. The overall prediction model is: Where u3(k) is the value of the water supply valve opening u3 at time k, Δu3(k) = u3(k) - u3(k-1) is the increment of the water supply valve opening u3 at time k, and the column vector is the control increment sequence of the water supply valve opening u3 at time k, ..., k+M; define the matrix A 23 is the dynamic matrix of the water supply valve opening u3 variable controlling the suspension density y2, Where [a 23 (1), …, a 23 (P)] is the step response sequence of the suspension density y2 in response to the change of the water supply valve opening u3; the prediction model formula [a 13 (1), …, a 13 (P)] is the step response sequence of the suspension level y1 in response to the change of the water supply valve opening u3.

2. According to claim 1, a rapid heavy medium coal preparation density control system based on QP-DMC is characterized in that: The different sub-ranges based on the preset suspension density setting value are as follows: (1) When the real-time suspension density data y2 is less than the set value, the first sub-range is entered, and the diverter valve opening u1 and the mediator valve opening u2 are adjusted according to the deviation between the real-time suspension liquid level data y1 and the preset suspension liquid level value, and the deviation between the real-time suspension density data y2 and the set suspension density value; (2) When the real-time suspension density data y2 is greater than the set value, the second sub-process is entered. According to the deviation between the real-time suspension level data y1 and the suspension level preset value, as well as the deviation between the real-time suspension density data y2 and the suspension density set value, the diverter valve u1 and the water supply valve opening u3 are adjusted.

3. According to the QP-DMC-based rapid heavy medium coal preparation density control system of claim 1, it is characterized in that: Based on the QP-DMC problem, including multiple single variable split-range QP-DMC fast controller performance indicators, in order to determine the setting range of the control variables diverter valve opening u1, the medium valve opening u2 and the water supply valve opening u3, so as to ensure that the controlled variables suspension level and suspension density fluctuate within the predetermined range, multiple single variable split-range QP-DMC fast controller performance indicators are constructed for each split process; for the first split, multiple single variable split-range QP-DMC fast controller performance indicators are constructed as follows: Where: Matrix A′ 11 is the dynamic matrix A 11 The transpose of the matrix A′ 22 is the dynamic matrix A 22 The transpose of , column vector is the step response sequence of the suspension level y1 in response to the change of the medium valve opening u2, and the column vector is the step response sequence of the suspension density y2 in response to the change of the diverter valve opening u1, and the column vector is the target value sequence of the suspension level y1 at time k, column vector is the target value sequence of suspension density y2 at time k, matrix Q is the controlled variable error weighting matrix in multivariable predictive control, and matrix R is the manipulated variable incremental control weighting matrix in multivariable predictive control; For the second split range, multiple single variable split range QP-DMC fast controller indicators are constructed as follows: Where: Matrix A′ 23 is the dynamic matrix A 23 The transpose of , column vector It is the step response sequence of the suspension level y1 in response to the change of the water supply valve opening u3.

Images