Cross-layer closed-loop optimization method for strip steel hot continuous rolling process

Abstract

According to the cross-layer closed-loop optimization method for the strip steel hot continuous rolling process, cross-layer closed-loop multi-objective optimization is carried out on the strip steel hot continuous rolling process by using a nested model predictive control (MPC) frame, and then the requirements for high quality, low energy consumption and other production objectives in the hot continuous rolling process are met; and meanwhile, the optimization efficiency in the strip steel hot continuous rolling process is improved, and the optimization quality is improved. For a distributed manufacturing process, a cross-layer closed-loop optimization architecture based on cloud side end cooperation is designed by using a cloud side end cooperation mechanism. By adopting the method disclosed by the invention, the calculation efficiency can be improved, the global and local optimization targets of the strip steel outlet thickness and the system energy consumption in the finish rolling area in the strip steel hot continuous rolling process can be better considered, and a finer and more dynamic industrial process control strategy is realized.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent manufacturing technology and relates to a multi-objective optimization method for a strip hot rolling process, and specifically relates to a multi-objective cross-layer closed-loop optimization method for a strip hot rolling process based on a "cloud-edge-end" collaborative nested model predictive control framework. Background Art

[0002] With the rapid development of modern industrial systems, the complexity of industrial manufacturing processes continues to increase. Traditional single-objective optimization methods have been unable to meet the growing demand for high quality and low energy consumption in the manufacturing process. The finishing process is a key process in the hot rolling process of strip steel. After the rough rolling process and before the laminar cooling process, the finishing stand is the core component of the finishing process. The rough rolling intermediate billet will be continuously rolled through multiple finishing stands to make a strip of specified precision. Dynamically optimizing the roll gap and rolling speed of the finishing process, that is, the gap between the two rolls and the speed at which the strip passes through the finishing stand, is of great significance for improving product precision, strengthening the management and control of the manufacturing process, and realizing efficient and environmentally friendly manufacturing processes.

[0003] At present, a lot of research work has been carried out on multi-objective optimization methods for industrial manufacturing processes, but there are still deficiencies and challenges in practical applications. First, different industrial manufacturing processes focus on different optimization objectives. Existing multi-objective optimization methods often balance multiple conflicting objectives when solving the optimal solution, which may lead to the final solution being not ideal for some objectives; second, multi-objective optimization problems are often more computationally complex than single-objective optimization problems because of their higher data dimensions; finally, for complex systems that need to take into account both global and local objectives, existing multi-objective optimization methods for industrial manufacturing processes lack an effective combination of static and dynamic objectives for industrial manufacturing, and cannot adjust the optimization strategy in real time. These deficiencies make the existing methods have certain limitations in solving multi-objective optimization problems in the manufacturing process, and it is difficult to achieve effective optimization of the roll gap and rolling speed in the finishing process. Summary of the invention

[0004] In order to overcome the shortcomings of the above-mentioned prior art, the present invention proposes a cross-layer closed-loop optimization method for the hot strip rolling process. By using a nested model predictive control (MPC) framework, a cross-layer closed-loop multi-objective optimization is performed on the strip outlet thickness and the energy consumption of the finishing rolling process in the finishing rolling process area (F7 frame) of the hot strip rolling process, thereby meeting the requirements for production objectives such as high quality (steel thickness) and low energy consumption in the hot strip rolling manufacturing process, while improving the optimization efficiency and optimization quality in the manufacturing process.

[0005] For the distributed manufacturing process, the present invention utilizes the "cloud-edge-end" collaborative mechanism and proposes a cross-layer closed-loop optimization architecture based on the "cloud-edge-end" collaboration. According to the execution order, it is divided into: (1) "end" side: realize the collection of input and output signals of each subsystem and upload them to the edge side, and receive the control signal from the edge side to control the manufacturing process in real time; (2) "edge" side: for the optimization target of each subsystem, establish and solve the inner optimization framework (strip steel exit thickness system), calculate the inner optimal control sequence, and upload it to the cloud; at the same time, receive the optimal control signal from the cloud and send it to the end side; (3) "cloud" side: receive the inner optimal control sequence from each edge side, establish and solve the outer optimization framework (energy consumption system) for the global target of the whole system, and send the calculated optimal control signal to each edge side.

[0006] The technical solution provided by the present invention is:

[0007] A cross-layer closed-loop optimization method for a hot strip rolling process uses a nested model predictive control framework to perform cross-layer closed-loop multi-objective optimization on the strip outlet thickness and the energy consumption of the finishing rolling process in the hot strip rolling process, including the following steps:

[0008] a) Establishing an inner optimization model predictive control framework and an outer optimization model predictive control framework;

[0009] The outer optimization model predictive control framework is constructed as the dynamic performance model of the energy consumption system, and the optimization goal of the outer optimization model predictive control is to minimize energy consumption; the inner optimization model predictive control framework is constructed as the dynamic performance model of the strip steel outlet thickness system, and the optimization goal of the inner optimization model predictive control is to minimize the difference between the strip steel outlet thickness and the set expected value;

[0010] b) Establish the relationship between the internal and external control inputs and construct a cross-layer closed-loop optimization model;

[0011] The linear matrix is ​​introduced to optimize the outer optimization model predictive control while satisfying the optimization target of the inner optimization model predictive control, that is, the system energy consumption is minimized when the difference between the strip outlet thickness and the set expected value is minimized;

[0012] c) Convert the cross-layer closed-loop optimization into a solvable quadratic programming form: first solve the inner optimization problem, then solve the outer optimization problem, and obtain the solution of the cross-layer closed-loop optimization problem through the convex combination of the inner optimal control sequence through the linear matrix, thereby realizing the cross-layer closed-loop optimization of the multi-objective model predictive control problem.

[0013] In step b) of the cross-layer closed-loop optimization method for the hot strip rolling process, the cross-layer closed-loop optimization objective function is expressed as:

[0014]

[0015] stx 0|k =x k

[0016] x l+1|k =A u x l|k +B u u l|k

[0017] x l|k ∈X i ,x N|k ∈P i ,u l|k ∈U i ,l∈I 0:N-1 (Formula 7)

[0018]

[0019] ste 0|l =e 1|l-1

[0020]

[0021] Among them, l represents the current prediction step of the outer layer; L u (x l|k ) is the cost function of the outer stage; F u (x N|k ) is the outer terminal cost function; x l|k represents the predicted state of the energy consumption system at step l starting from the prediction time k; u l|k represents the prediction decision variable of the energy consumption system starting from the prediction time k for l steps; N is the outer prediction step length; A u is the outer state transfer matrix; B u is the outer control matrix; X i is the state range of the outer stage, P i is the outer terminal state range, U i is the range of decision variables, I 0:N-1 is an integer from 0 to N-1; Y is a linear matrix; are decision variables, namely, roll gap and roll linear speed; L w (e q|l ) is the cost function of the inner stage; F w (e P|l ) is the inner terminal cost function; E i is the state range of the inner stage, F i is the inner terminal state range; A w is the inner state transfer matrix, B wis the inner control matrix; e q|l It represents the prediction status of the strip steel exit thickness system in the q-step starting from the prediction time l; is the prediction decision variable of the strip exit thickness system for q steps starting from the prediction time l; P is the inner layer prediction step length.

[0022] Furthermore, the outer optimization model predictive control framework is expressed as:

[0023]

[0024] Among them, x l|k represents the predicted state of the energy consumption system at step l starting from the prediction time k, u l|k represents the prediction decision variable of the energy consumption system starting from the prediction time k, N is the outer prediction step length, X i is the state range of the outer stage, P i is the outer terminal state range, U i is the range of decision variables, I 0:N-1 is an integer from 0 to N-1; the decision variables are the roll gap and the roll linear speed, expressed as Outer stage cost function L u (x l|k ) and the outer terminal cost function F u (x N|k ) are respectively expressed as:

[0025]

[0026] Among them, Q u and P u are all positive definite matrices; x * is the expected value set, which is 0;

[0027] The inner optimization model predictive control framework is expressed as:

[0028]

[0029] where e q|l represents the prediction state of the strip steel exit thickness system at step q starting from the prediction time l, is the prediction decision variable of the strip steel exit thickness system for q steps starting from the prediction time l, P is the inner prediction step length, E i is the state range of the inner stage, F i is the inner terminal state range, U i is the range of decision variables, I 0:N-1 is an integer from 0 to N-1; the decision variable Inner stage cost function L w (eq|l ) and the inner terminal cost function F w (e P|l ) are respectively expressed as:

[0030]

[0031] Where E w and S w are all positive definite matrices; e * To set expectations.

[0032] Furthermore, in step c), the inner model predictive control framework is converted into a quadratic programming form, which is expressed as:

[0033]

[0034] Among them, J1 is the objective function of the inner layer model predictive control; by solving it, the inner layer optimal control sequence, that is, the outlet thickness optimal control sequence, can be obtained;

[0035] The outer model predictive control framework is transformed into a quadratic programming form, expressed as:

[0036]

[0037] Among them, J2 is the objective function of the outer model predictive control.

[0038] Furthermore, the inner optimal control sequence is passed through the linear matrix The convex combination obtains the solution to the cross-layer closed-loop optimization problem; the outer layer optimization problem is expressed as:

[0039]

[0040] in, The optimal control sequence is obtained for the inner optimization problem in each outer prediction step l=1,…,N-1. Solving (Equation 19) can minimize the system energy consumption when the difference between the strip outlet thickness and the set expected value is minimized.

[0041] For the distributed hot rolling process of strip steel, the present invention provides a cross-layer closed-loop optimization method for the distributed hot rolling process of strip steel, which is implemented according to the cloud-edge-end collaboration mechanism by utilizing the cross-layer closed-loop optimization method for the hot rolling process of strip steel described above; the execution order is as follows: 1) End side: realize the collection of input and output signals of each subsystem and upload them to the edge side, and at the same time receive the control signal from the edge side to control the manufacturing process in real time; 2) Edge side: for the optimization target of each subsystem, establish and solve the inner optimization framework, calculate the inner optimal control sequence, and upload it to the cloud; at the same time, receive the optimal control signal from the cloud and send it to the end side; 3) Cloud side: receive the inner optimal control sequence from each edge side, establish and solve the outer optimization framework for the global target of the entire system, and at the same time send the calculated optimal control signal to each edge side.

[0042] For the cross-layer closed-loop optimization of distributed hot strip rolling process, the cross-layer optimization function is expressed as:

[0043]

[0044]

[0045] in, The optimal control sequence obtained for the inner optimization problem in each outer prediction step l=1,…,N-1 of the i-th subsystem.

[0046] Compared with the prior art, the beneficial technical effects of the present invention are:

[0047] The present invention provides a cross-layer closed-loop optimization method for a hot strip rolling process. For the multi-objective optimization problem in the hot rolling process, on the basis of determining that the strip outlet thickness in the finishing process area (F7 frame) is a high-priority optimization target and the energy consumption of the process is a low-priority optimization target, the nested idea is used to take the optimal solution set of the inner-layer MPC problem as a constraint of the outer-layer MPC problem to achieve cross-layer closed-loop optimization. At the same time, the present invention proposes a "cloud-edge-end" collaborative cross-layer closed-loop optimization architecture for distributed manufacturing processes. The end side realizes the collection of input and output signals of each subsystem; the edge side establishes an inner-layer optimization framework and solves it for local optimization targets; the cloud side establishes an outer-layer optimization framework and solves it for the global target of the entire system. The method of the present invention can meet the requirements for production targets such as high quality and low energy consumption in the process of hot strip rolling, while improving the optimization efficiency and quality in the manufacturing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flowchart of the cross-layer closed-loop optimization process proposed in the present invention.

[0049] Figure 2A flowchart of the cross-layer closed-loop optimization process for “cloud-edge-end” collaboration in distributed manufacturing processes provided by the present invention.

[0050] Figure 3 This is a diagram showing an example simulation result of a hot strip rolling process based on the algorithm provided by the present invention.

[0051] Figure 4 It is a comparison chart of simulation results based on the cross-layer optimization framework (red line) and the linear weighted sum method (blue line) provided by the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described below by way of embodiments in conjunction with the accompanying drawings, but the scope of the present invention is not limited in any way.

[0053] The present invention relates to a multi-objective optimization method for a manufacturing process, and specifically provides a multi-objective cross-layer closed-loop optimization method for a manufacturing process based on a nested model predictive control framework and with "cloud-edge-end" collaboration. Figure 1 The figure shows the cross-layer closed-loop optimization process frame proposed by the present invention. For the multi-objective optimization control problem, on the basis of determining that the strip outlet thickness of the F7 frame in the finishing process area is a high-priority optimization target and the energy consumption of the finishing process is a low-priority optimization target, a cross-layer closed-loop optimization problem based on a nested MPC framework is established, a linear relationship between the internal and external control inputs is established, and the problem is converted into two quadratic programming problems for solution. To summarize the above discussion, we propose a cross-layer closed-loop optimization strategy, as shown in Algorithm 1.

[0054]

[0055] The specific implementation method described above realizes the solution of multi-objective cross-layer closed-loop optimization problems in the manufacturing process. Figure 2 A flowchart of the cross-layer closed-loop optimization process of "cloud-edge-end" collaboration provided by the present invention. For the distributed manufacturing process, a cross-layer closed-loop optimization method under the collaboration of "cloud-edge-end" is proposed, as shown in Algorithm 2.

[0056]

[0057] For distributed manufacturing processes, the cross-layer closed-loop optimization method of the manufacturing process coordinated by "cloud, edge and end" is an optimization algorithm based on the nested model predictive control (MPC) framework. In the specific implementation, the two objectives of the hot rolling process are optimized, and the strip outlet thickness of the F7 frame in the finishing process area and the energy consumption of the finishing process are taken as optimization targets, and the roll gap setting and the working roll speed are taken as decision variables. The dynamic performance models of the inner and outer system are established according to the priority and converted into corresponding MPC problems. Among them, the strip outlet thickness of the F7 frame in the finishing process area with high priority is taken as the optimization target of the inner system, and its purpose is to minimize the difference between the strip outlet thickness of the F7 frame in the finishing process area and the expected value of the outlet thickness setting; the energy consumption of the finishing process with low priority is taken as the optimization target of the outer system, and its purpose is to minimize the energy consumption of the process. The nested idea is used to take the optimal solution set of the inner MPC problem as the constraint of the outer MPC problem to realize cross-layer closed-loop optimization. It mainly includes the following steps:

[0058] a) Establish the inner and outer MPC framework according to modern control theory;

[0059] In the specific implementation of the present invention, two different optimization targets are involved in the manufacturing process. For the high-priority strip steel outlet thickness optimization target of the F7 stand in the finishing rolling area, an inner system dynamic performance model is established; for the low-priority energy consumption optimization target of the finishing rolling process, an outer system dynamic performance model is established. The decision variables of the two system dynamic performance models are the same, which are the roll gap and the roll linear speed. The most basic MPC framework can be constructed for these two dynamic performance models.

[0060] Dynamic performance model of inner strip steel outlet thickness system:

[0061] e q+1 =A w e q +B w u q

[0062] y w =C w e q +D w u q (Formula 1)

[0063] Where q represents the current prediction step of the inner layer, e q Indicates the current state of the inner layer, e q+1 is the state of the inner layer at the next moment, u q represents the current decision variables of the inner layer, namely the roll gap and the roller linear speed, y w is the inner layer output, i.e. the outlet thickness, A w is the inner state transfer matrix, B w is the inner control matrix, C wis the inner layer output matrix, D w Directly transfer the matrix to the inner layer.

[0064] Dynamic performance model of outer energy consumption system:

[0065] x l+1 =A u x l +B u u l

[0066] y u =C u x l +D u u l (Formula 2)

[0067] Where l represents the current prediction step of the outer layer, x l Indicates the current state of the outer layer, x l+1 is the outer state at the next moment, u l Represents the current decision variable in the outer layer, y u is the outer layer output, i.e. the energy consumption of this process, A u is the outer state transfer matrix, B u is the outer control matrix, C u is the outer output matrix, D u Directly transfer the matrix to the outer layer.

[0068] For the outer optimization target, the energy consumption of this process is set to be x in order to minimize the energy consumption. * = 0. Then its MPC framework can be expressed as:

[0069]

[0070] Among them, x l|k represents the predicted state of the energy consumption system at step l starting from the prediction time k, u l|k represents the prediction decision variable of the energy consumption system starting from the prediction time k, N is the outer prediction step length, X i is the state range of the outer stage, P i is the outer terminal state range, U i is the range of decision variables, I 0:N-1 is an integer from 0 to N-1; where roll gap and roll linear speed (decision variables) Outer stage cost function L u (x l|k ) and the outer terminal cost function F u (x N|k ) are respectively expressed as:

[0071]

[0072] Where Q u and P u are all positive definite matrices.

[0073] For the inner layer optimization target outlet thickness, the expected value is set to e * , then its MPC framework can be expressed as:

[0074]

[0075] where e q|l represents the prediction state of the strip steel exit thickness system at step q starting from the prediction time l, represents the prediction decision variable of the strip steel exit thickness system for q steps starting from the prediction time l, P is the inner prediction step length, E i is the state range of the inner stage, F i is the inner terminal state range, U i is the range of decision variables, I 0:N-1 is an integer from 0 to N-1; the decision variable Inner stage cost function L w (e q|l ) and the inner terminal cost function F w (e P|l ) are respectively expressed as:

[0076]

[0077] Where E w and S w are all positive definite matrices.

[0078] b) Establish the relationship between the internal and external control inputs and construct a cross-layer closed-loop optimization model;

[0079] The linear matrix Y is introduced to establish the relationship between the inner and outer control inputs, so that the outer optimization is performed while ensuring that the inner optimization target is met. That is, while ensuring that the strip outlet thickness of the high-priority finishing area (F7 stand) meets the minimum difference with the set expected value, the energy consumption of the low-priority process is minimized. Therefore, the cross-layer closed-loop optimization problem is constructed as follows:

[0080]

[0081] c) Convert the inner and outer optimization problems into quadratic programming form and obtain the objective function:

[0082] c1) In general, the more common quadratic programming problem is as follows:

[0083]

[0084] stAξ≤B (Equation 8) c2) Convert the inner MPC framework into a quadratic programming form. In the MPC framework, (Equation 5) can first be converted into:

[0085]

[0086] The objective function can be expressed as:

[0087]

[0088] in diag represents a diagonal matrix.

[0089] The linear constraints on state and control inputs can be expressed as:

[0090]

[0091] make as well as Then the linear constraint in (Equation 11) can be transformed into:

[0092]

[0093] in,

[0094]

[0095] Thus, the inner MPC framework is transformed into the general form of the quadratic constraint problem as shown in (Equation 8):

[0096]

[0097] By solving (Equation 14), the optimal control sequence of the inner layer, that is, the optimal control sequence of the outlet thickness, can be obtained.

[0098] c3) Transform the outer MPC framework:

[0099]

[0100] Similarly, on this basis, the objective function of outer optimization can be expressed as:

[0101]

[0102] in The linear constraints of the outer MPC can be transformed into the form of (Equation 12):

[0103]

[0104] Therefore, the outer optimization problem can be transformed into:

[0105]

[0106] c4) Using linear matrix Establish the relationship between inner and outer decision variables

[0107] For conventional MPC problems, the first item of the inner optimal control sequence obtained by solving the inner optimization problem is usually directly taken as the optimal solution. For nested MPC problems, in order to optimize the inner and outer objective functions at the same time, the outer objective function is optimized under the premise of satisfying the optimal inner objective function as much as possible. In the present invention, the solution to the cross-layer closed-loop optimization problem is the inner optimal control sequence (exit thickness) obtained by (Equation 14) after the linear matrix Convex combination is obtained. Therefore, combined with (Equation 18), (Equation 7) can be transformed into a solvable quadratic programming form, and the outer optimization problem can be expressed as:

[0108]

[0109] in, The optimal control sequence obtained by the inner optimization problem in each outer layer prediction step l = 1, ..., N-1. Solving (Equation 19) can achieve cross-layer closed-loop optimization and obtain the optimal control sequence for energy consumption

[0110] Through steps c1) to c4), the cross-layer closed-loop optimization problem (Formula 7) can be transformed into a solvable quadratic programming form: first solve the inner optimization problem (Formula 14), then solve the outer optimization problem (Formula 19), thereby achieving cross-layer closed-loop optimization of multi-objective MPC problems. This cross-layer closed-loop optimization method can improve computational efficiency and optimization results.

[0111] d) Cross-layer closed-loop optimization framework in distributed manufacturing process with “cloud-edge-end” collaboration

[0112] For the distributed hot rolling process, the present invention utilizes the "cloud-edge-end" collaborative mechanism and proposes a cross-layer closed-loop optimization architecture based on the "cloud-edge-end" collaboration. According to the execution order, it is divided into: (1) "End" side: realize the collection of input and output signals of each subsystem and upload them to the edge side, and receive the control signal from the edge side to control the manufacturing process in real time; (2) "Edge" side: for the optimization target of each subsystem, establish and solve the inner optimization framework, calculate the inner optimal control sequence, and upload it to the cloud; at the same time, receive the optimal control signal from the cloud and send it to the end side; (3) "Cloud" side: receive the inner optimal control sequence from each edge side, establish and solve the outer optimization framework for the global target of the entire system, and send the calculated optimal control signal to each edge side. Based on the above cross-layer closed-loop optimization method for the hot rolling process, considering the distributed manufacturing process, for each subsystem, the inner optimization problem (Formula 14) can be rewritten as:

[0113]

[0114] Considering the sum of the outer objective functions of each subsystem for global optimization, the outer optimization problem (Equation 19) can be rewritten as:

[0115]

[0116] in, The optimal control sequence (exit thickness) obtained for the inner optimization problem in each outer prediction step l=1,…,N-1 of the i-th subsystem.

[0117] Combining (Equation 20) and (Equation 21), the cross-layer optimization problem of the distributed manufacturing process can be expressed as:

[0118]

[0119] In the present invention, for the multi-objective optimization problem in the hot rolling process, on the basis of determining the outlet thickness as a high-priority optimization target and the energy consumption of the process as a low-priority optimization target, the nested idea is used to take the optimal solution set of the inner-layer MPC problem (exit thickness) as the constraint of the outer-layer MPC problem (energy consumption) to achieve cross-layer closed-loop optimization. At the same time, the present invention proposes a "cloud-edge-end" collaborative cross-layer closed-loop optimization architecture for distributed manufacturing processes. The end side realizes the collection of input and output signals of each subsystem; the edge side establishes an inner-layer optimization framework and solves it for local optimization targets; the cloud side establishes an outer-layer optimization framework and solves it for the global target of the entire system.

[0120] Example: The example of the present invention selects slabs produced by a steel plant, and after training the dynamic performance model of the strip exit thickness system of the F7 frame in the finishing rolling area and the dynamic performance model of the energy consumption system of the process, multi-objective cross-layer closed-loop optimization is performed according to corresponding steps.

[0121] The dynamic model parameters of the inner layer outlet thickness system are:

[0122]

[0123] C w =[-0.456 0.093 -0.063 0.072] D w =0

[0124] The dynamic model parameters of the outer energy consumption system are:

[0125]

[0126] C u =[-2.554 0.028 0.980 0.093 -0.094] D u =0

[0127] Figure 3 This is a diagram showing an example simulation result of a hot strip rolling process based on the algorithm provided by the present invention. Figure 4 The figure is a comparison chart of the simulation results of the cross-layer optimization algorithm (red line) and the linear weighted sum method (blue line) provided by the present invention. Figure 4 As shown in the figure, the system output (red line) obtained by the algorithm provided by the present invention has almost completed convergence when the number of simulation steps is about 100, while the system output (blue line) obtained by the linear weighted sum method has almost completed convergence when the number of simulation steps is nearly 400; at the same time, the final result shows that the algorithm provided by the present invention (red line) has a smaller tracking error in the optimization result of the inner layer outlet thickness, and also achieves lower energy consumption of the outer layer system process. The example results reflect that the optimization method in the present invention can accelerate the convergence speed and tracking accuracy of the system.

[0128] It should be noted that the purpose of publishing the embodiments is to help further understand the present invention, but those skilled in the art can understand that various substitutions and modifications are possible without departing from the scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the contents disclosed in the embodiments, and the scope of protection claimed by the present invention shall be subject to the scope defined in the claims.

Claims

1. A cross-layer closed-loop optimization method for hot strip rolling process, characterized in that: Using a nested model predictive control framework, a multi-objective cross-layer closed-loop optimization of the strip outlet thickness and the energy consumption of the finishing process in the hot strip rolling process is performed, including the following steps: a) Establishing an inner optimization model predictive control framework and an outer optimization model predictive control framework; The outer optimization model predictive control framework is constructed as the dynamic performance model of the energy consumption system, and the optimization goal of the outer optimization model predictive control is to minimize energy consumption; the inner optimization model predictive control framework is constructed as the dynamic performance model of the strip steel outlet thickness system, and the optimization goal of the inner optimization model predictive control is to minimize the difference between the strip steel outlet thickness and the set expected value; b) Establish the relationship between the internal and external control inputs and construct a cross-layer closed-loop optimization model; The linear matrix is ​​introduced to optimize the outer optimization model predictive control while satisfying the optimization target of the inner optimization model predictive control, that is, the system energy consumption is minimized when the difference between the strip outlet thickness and the set expected value is minimized; c) Convert the cross-layer closed-loop optimization into a solvable quadratic programming form: first solve the inner optimization problem, then solve the outer optimization problem, and obtain the solution of the cross-layer closed-loop optimization problem through the convex combination of the inner optimal control sequence through the linear matrix, thereby realizing the cross-layer closed-loop optimization of the multi-objective model predictive control problem.

2. The cross-layer closed-loop optimization method for hot strip rolling process according to claim 1, characterized in that: In step b), the cross-layer closed-loop optimization objective function is expressed as: Among them, l represents the current prediction step of the outer layer; L u (x l|k ) is the cost function of the outer stage; F u (x N|k ) is the outer terminal cost function; x l|k represents the predicted state of the energy consumption system at step l starting from the prediction time k; N is the outer prediction step length; A u is the outer state transfer matrix; B u is the outer control matrix; u l|k represents the prediction decision variable of the energy consumption system at step l starting from the prediction time k; the decision variables are the roll gap and the roll linear speed; X i is the state range of the outer stage, P i is the outer terminal state range, U i is the range of decision variables, I 0:N-1 is an integer from 0 to N-1; Y is a linear matrix; is the decision variable; L w (e q|l ) is the cost function of the inner stage; F w (e P|l ) is the inner terminal cost function; E i is the state range of the inner stage, F i is the inner terminal state range; A w is the inner state transfer matrix, B w is the inner control matrix; e q|l It represents the prediction status of the strip steel exit thickness system in the q-step starting from the prediction time l; is the prediction decision variable of the strip exit thickness system for q steps starting from the prediction time l; P is the inner layer prediction step length.

3. The cross-layer closed-loop optimization method for hot strip rolling process according to claim 2, characterized in that: The outer optimization model predictive control framework is expressed as: Among them, x i|k represents the predicted state of the energy consumption system at step l starting from the prediction time k, N is the outer prediction step length, X i is the state range of the outer stage, P i is the outer terminal state range, U i is the range of decision variables, I 0:N-1 is an integer from 0 to N-1; the decision variables are the roll gap and the roll linear speed, expressed as Stage cost function L u (x l|k ) and the terminal cost function F u (x N|k ) are respectively expressed as: Among them, Q u and P u are all positive definite matrices; x * is the expected value set, which is 0; The inner optimization model predictive control framework is expressed as: where e q|l represents the prediction state of the strip steel exit thickness system at step q starting from the prediction time l, P is the inner prediction step length, E i is the state range of the inner stage, F i is the inner terminal state range, U i is the range of decision variables, I 0:N-1 is an integer from 0 to N-1; the decision variable Stage cost function L w (e q|l ) and the terminal cost function F w (e P|l ) are respectively expressed as: Where E w and S w are all positive definite matrices; e * To set expectations.

4. The cross-layer closed-loop optimization method for hot strip rolling process according to claim 3, characterized in that: In step c), the inner model predictive control framework is converted into a quadratic programming form, which is expressed as: Among them, J1 is the objective function of the inner layer model predictive control; the inner layer optimal control sequence, that is, the outlet thickness optimal control sequence, is obtained by solving; The outer model predictive control framework is transformed into a quadratic programming form, expressed as: 0≤Y≤1,Y1=1 Y(i,i)≥δ Among them, J2 is the objective function of the outer model predictive control.

5. The cross-layer closed-loop optimization method for hot strip rolling process according to claim 4, characterized in that: The inner optimal control sequence passes through the linear matrix The convex combination obtains the solution to the cross-layer closed-loop optimization problem; the outer layer optimization problem is expressed in the form of quadratic programming: in, is the optimal control sequence obtained from the inner optimization problem in each outer prediction step l=1,…,N-1; diag is a diagonal matrix.

6. A cross-layer closed-loop optimization method for distributed hot strip rolling process, characterized in that: The cross-layer closed-loop optimization method for the hot strip rolling process described in claim 1 is implemented according to the cloud-edge-end collaboration mechanism; the execution order is: 1) Device side: collect the input and output signals of each subsystem and upload them to the edge side, and receive control signals from the edge side to control the manufacturing process in real time; 2) Edge: For each subsystem’s optimization goal, an inner optimization framework is established and solved, the inner optimal control sequence is calculated, and uploaded to the cloud. At the same time, the optimal control signal is received from the cloud and sent to the device. 3) Cloud side: Receive the inner optimal control sequence from each edge side, establish and solve the outer optimization framework for the global goal of the entire system, and send the calculated optimal control signal to each edge side.

7. The cross-layer closed-loop optimization method for distributed hot strip rolling process according to claim 6, characterized in that: The cross-layer optimization function of the distributed hot strip rolling process is expressed as: in, The optimal control sequence obtained for the inner optimization problem in each outer prediction step l=1,…,N-1 of the i-th subsystem.

Images