Partial-format model-free adaptive disturbance compensation control method for measurable disturbances

Abstract

The present invention discloses a partial format model-free adaptive disturbance compensation control method for measurable disturbances. The method establishes a dynamic linearized data model of a controlled object under the action of a measurable disturbance; constructs and solves a cost function, optimizes and updates the pseudo-Jacobi input matrix and the pseudo-Jacobi disturbance matrix; designs a partial format model-free adaptive disturbance compensation control scheme for measurable disturbances; constructs and solves an energy function, optimizes and updates the partial format adaptive input matrix and the partial format adaptive disturbance matrix; and uses the control scheme of the present invention to control the controlled object under the action of a measurable disturbance. The control method of the present invention can significantly weaken the influence of the measurable disturbance on the actual value output by the controlled object system, achieve effective tracking of the expected value trajectory, and significantly improve the disturbance compensation control performance.

Description

Technical Field

[0001] The present invention belongs to the field of automatic control, and in particular relates to a partial format model-free adaptive disturbance compensation control method for measurable disturbances. Background Art

[0002] Disturbances are widespread in practical control systems, such as those affecting most controlled objects in industries like oil refining, petrochemicals, chemicals, pharmaceuticals, food, papermaking, water treatment, thermal power, metallurgy, cement, rubber, machinery, electrical engineering, transportation, and robotics. These include reactors, distillation towers, machinery, equipment, devices, production lines, workshops, factories, unmanned vehicles, unmanned ships, drones, and autonomous mobile robots. In fact, the presence of disturbances often degrades system control performance and, in severe cases, can cause instability throughout the system, potentially compromising system safety.

[0003] Existing partial-format model-free adaptive control methods were first proposed by Hou Zhongsheng and Jin Shangtai in their co-authored book, Model-Free Adaptive Control: Theory and Applications (Science Press, 2013, p. 104). Building on this foundation, inventions CN108107721A and CN108107722A proposed SISO-based decoupling methods to address the control challenges of strongly coupled multi-input, multi-output (MIMO) systems. Inventions CN108287470A and CN108287471A proposed neural network-based parameter self-tuning methods to address the time-consuming and labor-intensive parameter selection problem. Invention CN111522231A proposed a heterogeneous factor control method to address the control challenges of strongly nonlinear MIMO systems with varying control channel characteristics. Invention CN111522229A expanded on invention CN111522231A by proposing a heterogeneous factor control method with parameter self-tuning, further addressing the time-consuming and labor-intensive parameter tuning challenges of heterogeneous factors. It should be pointed out that the above-mentioned inventive methods have not yet considered the control difficulties of the controlled objects under disturbance.

[0004] For multi-input, multi-output controlled objects subject to measurable disturbances, the analysis and design of disturbance compensation control methods that efficiently utilize the real-time measured input and output data of the controlled objects without relying on any mathematical model information, and that can reduce the impact of the measurable disturbances on the actual output values ​​of the controlled object system and effectively track the expected output values ​​of the system, are of great industrial application value. To achieve this goal, the present invention proposes a partial-format model-free adaptive disturbance compensation control method for measurable disturbances. Summary of the Invention

[0005] In order to solve the problems existing in the background technology, the purpose of the present invention is to provide a partial format model-free adaptive disturbance compensation control method for measurable disturbances. The control method runs on a hardware platform to control a controlled object under the action of a measurable disturbance. The controlled object is a multi-input multi-output system including multiple control inputs and multiple system outputs. The control method is characterized by comprising the following steps:

[0006] Step (1): obtaining a measurable disturbance at k sampling moments, and establishing a dynamic linearized data model of the controlled object under the action of the measurable disturbance, wherein the dynamic linearized data model of the controlled object includes a pseudo-Jacobi input matrix θ(k) and a pseudo-Jacobi disturbance matrix χ(k);

[0007] Step (2): constructing a cost function and solving the cost function using a function extremum method, optimizing and updating the pseudo-Jacobian input matrix θ(k) and the pseudo-Jacobian perturbation matrix χ(k) in step (1);

[0008] Step (3): Using the measurable disturbance at the k sampling moment, based on the dynamic linearized data model of the controlled object after optimizing the pseudo-Jacobi input matrix θ(k) and the pseudo-Jacobi disturbance matrix χ(k) in step (2), a partial format model-free adaptive disturbance compensation control scheme for the measurable disturbance is designed, wherein the control scheme includes the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p (k);

[0009] Step (4): Construct an energy function and solve the energy function using the momentum gradient descent method to optimize and update the partial format adaptive input matrix π in step (3) p (k) and the partial format adaptive perturbation matrix ω p (k);

[0010] Step (5): Optimize the partial format adaptive input matrix π using step (4) p (k) and the partial format adaptive perturbation matrix ω p The control scheme after (k) controls the controlled object under the action of measurable disturbance, weakens the influence of measurable disturbance on the actual value of the controlled object system output, and realizes effective tracking of the expected value of the system output.

[0011] Furthermore, the measurable disturbance at k sampling moments is obtained in step (1), and the dynamic linearized data model of the controlled object under the action of the measurable disturbance is established as follows:

[0012] Δy(k+1)=θ(k)Δu(k)+χ(k)Δd(k)

[0013] Where k is the sampling time, k is a positive integer; y(k+1) is the actual value vector of the system output of the controlled object at the k+1 sampling time, y(k+1)=[y1(k+1),…,y n (k+1)] T , Δy(k+1)=y(k+1)-y(k); n is the total number of system outputs of the controlled object, n is an integer greater than 1; u(k) is the control input vector of the controlled object at sampling time k, u(k)=[u1(k),…,u m (k)] T , Δu(k)=u(k)-u(k-1); m is the total number of control inputs of the controlled object, and m is an integer greater than 1; d(k) is the measurable disturbance vector of the controlled object at sampling time k, d(k)=[d1(k),…,d q (k)] T , Δd(k)=d(k)-d(k-1); q is the total number of measurable disturbances to the controlled object, q is a positive integer; θ(k) is the pseudo-Jacobian input matrix at sampling time k, and χ(k) is the pseudo-Jacobian disturbance matrix χ(k) at sampling time k.

[0014] The process of constructing a cost function in step (2) and solving the cost function using a function extremum method, and optimizing and updating the pseudo-Jacobian input matrix θ(k) and the pseudo-Jacobian perturbation matrix χ(k) in step (1) mainly includes the following steps:

[0015] Step (2.1): Input matrix θ(k) to the pseudo-Jacobi and construct the cost function

[0016] J(θ(k))=||Δy(k)-θ(k)Δu(k-1)-χ(k-1)Δd(k-1)|| 2 +μ1||Δθ(k)|| 2

[0017] Among them, μ1 is the first weight factor;

[0018] Step (2.2): For the pseudo-Jacobi perturbation matrix χ(k), construct a cost function

[0019] J(χ(k))=||Δy(k)-θ(k-1)Δu(k-1)-χ(k)Δd(k-1)|| 2 +μ2||Δχ(k)|| 2

[0020] Among them, μ2 is the second weight factor;

[0021] Step (2.3): Use the function extremum method to solve the cost function described in step (2.1) and optimize and update the pseudo-Jacobi input matrix θ(k).

[0022]

[0023] Among them, α1 is the first step size factor;

[0024] Step (2.4): Use the function extremum method to solve the cost function described in step (2.2) and optimize and update the pseudo-Jacobi perturbation matrix χ(k).

[0025]

[0026] Among them, α2 is the second step size factor.

[0027] The measurable disturbance at the k sampling moment described in step (3) is used, based on the dynamic linearized data model of the controlled object after optimizing the pseudo-Jacobi input matrix θ(k) and the pseudo-Jacobi disturbance matrix χ(k) in step (2), to design a partial format model-free adaptive disturbance compensation control scheme for the measurable disturbance:

[0028] u(k)=u(k-1)+π p (k)ΔE(k)+ω p (k)ΔD(k)

[0029] Where, ΔE(k)=[-e(k) T ,Δe(k)…,Δe(k-L+2) T ] T ,ΔD(k)=[Δd(k) T ,…,Δd(k-L+1) T ] T ; e(k) is the system error vector of the controlled object at sampling time k, e(k) = y * (k)-y(k), e(k)=[e1(k),…,e n (k)] T , Δe(k)=e(k)-e(k-1); L is the linearization length constant; L is a positive integer; π p (k) is the adaptive input matrix of the partial format at sampling time k, ω p (k) is the adaptive perturbation matrix of the partial format at sampling time k.

[0030] The energy function described in step (4) is constructed and solved by the momentum gradient descent method, and the partial format adaptive input matrix π described in step (3) is optimized and updated. p (k) and the partial format adaptive perturbation matrix ω p(k), mainly including the following steps:

[0031] Step (4.1): Construct the energy function

[0032]

[0033] Among them, y * (k+1) is the system output expected value vector of the controlled object at the k+1 sampling time, λ is the penalty factor;

[0034] Step (4.2): Use the momentum gradient descent method to solve the energy function described in step (4.1) and optimize and update the partial format adaptive input matrix π p (k),

[0035]

[0036] Among them, σ1 is the first learning rate, η1 is the first momentum factor; Δπ p (k-1)=π p (k-1)-π p (k-2);

[0037] The energy function W is π p Partial derivative of (k-1);

[0038] Step (4.3): Use the momentum gradient descent method to solve the energy function described in step (4.1) and optimize and update the partial format adaptive perturbation matrix ω p (k),

[0039]

[0040] Among them, σ2 is the second learning rate, η2 is the second momentum factor; Δω p (k-1)=ω p (k-1)-ω p (k-2); The energy function W is the energy function of ω p (k-1) partial derivative.

[0041] The energy function W described in step (4.2) is p The formula for calculating the partial derivative of (k-1) is:

[0042]

[0043] The energy function W described in step (4.3) is p The formula for calculating the partial derivative of (k-1) is:

[0044]

[0045] described The mathematical formula is:

[0046] The step (5) described in step (4) is to optimize the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p The control scheme after (k) controls the controlled object under the action of the measurable disturbance and includes the following steps at each sampling time k:

[0047] Step (5.1): Obtain the measurable disturbance vector d(k) at the current sampling moment;

[0048] Step (5.2): Get the expected value vector y of the system output at the current sampling moment * (k), the system output actual value vector y(k), and the system error vector e(k) at the current sampling moment is calculated;

[0049] Step (5.3): Based on steps (5.1) and (5.2), use step (4) to optimize the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p (k) and the control scheme after calculation to obtain the control input vector u(k) at the current sampling moment;

[0050] Step (5.4): After the control input vector acts on the controlled object, the actual value vector of the system output of the controlled object at the next sampling moment is obtained.

[0051] Furthermore, the present invention adopts the following technical solutions:

[0052] A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the computer program implements the above-mentioned partial format model-free adaptive disturbance compensation control method for measurable disturbances.

[0053] Furthermore, the present invention adopts the following technical solutions:

[0054] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the processor implements the above-mentioned partial format model-free adaptive disturbance compensation control method for measurable disturbances.

[0055] Based on the theoretical foundations of existing partial-format model-free adaptive control methods, some inventive methods have made progress in addressing issues such as strong coupling of controlled objects, varying channel characteristics, and time-consuming and labor-intensive parameter tuning. However, these inventive methods have not yet considered the control issues of controlled objects under disturbances, which has restricted their widespread application. For multi-input, multi-output controlled objects under measurable disturbances, the present invention can efficiently utilize the input and output data measured in real time by the controlled objects, without relying on any mathematical model information to analyze and design disturbance compensation control methods. Furthermore, the designed control method can reduce the impact of measurable disturbances on the actual output value of the controlled object system, effectively tracking the expected value of the system output, and has important industrial application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a block diagram of the algorithm principle of the present invention;

[0057] Figure 2 This is a block diagram of the engineering application system of the present invention;

[0058] Figure 3 A schematic diagram of the hardware platform for running the present invention;

[0059] Figure 4 The control effect diagram of the output of the first system when the two-input and two-output system adopts the partial format model-free adaptive disturbance compensation control method for measurable disturbances proposed by the present invention and the comparative control method;

[0060] Figure 5 The control effect diagram of the second system output when the two-input and two-output system adopts the partial format model-free adaptive disturbance compensation control method for measurable disturbances proposed by the present invention and the comparative control method;

[0061] Figure 6 The first control input curve when the partial format model-free adaptive disturbance compensation control method for measurable disturbances and the comparative control method proposed in the present invention are used for a two-input two-output system;

[0062] Figure 7 The second control input curve when the partial format model-free adaptive disturbance compensation control method for measurable disturbances proposed by the present invention and the comparative control method are used for a two-input two-output system;

[0063] Figure 8 This is a flow chart of the refrigeration cycle of a vapor compression refrigeration system;

[0064] Figure 9 are the two measurable disturbance curves of the vapor compression refrigeration system;

[0065] Figure 10A control effect diagram of the first system output when the vapor compression refrigeration system adopts the partial format model-free adaptive disturbance compensation control method for measurable disturbances proposed by the present invention and the comparative control method;

[0066] Figure 11 A control effect diagram of the second system output when the vapor compression refrigeration system adopts the partial format model-free adaptive disturbance compensation control method for measurable disturbances proposed by the present invention and the comparative control method;

[0067] Figure 12 The first control input curve when the partial format model-free adaptive disturbance compensation control method for measurable disturbances proposed by the present invention and the comparative control method are used for a vapor compression refrigeration system;

[0068] Figure 13 The second control input curve when the partial format model-free adaptive disturbance compensation control method for measurable disturbances proposed by the present invention and the comparative control method are adopted for a vapor compression refrigeration system. DETAILED DESCRIPTION

[0069] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0070] Figure 1 The algorithm principle block diagram of the present invention is given. The present invention discloses a partial format model-free adaptive disturbance compensation control method for measurable disturbances. The method establishes a dynamic linearized data model of the controlled object under the action of measurable disturbances, and the model includes a pseudo-Jacobi input matrix and a pseudo-Jacobi disturbance matrix; constructs and solves a cost function, optimizes and updates the pseudo-Jacobi input matrix and the pseudo-Jacobi disturbance matrix; designs a partial format model-free adaptive disturbance compensation control scheme for measurable disturbances, and the scheme includes a partial format adaptive input matrix and a partial format adaptive disturbance matrix; constructs and solves an energy function, optimizes and updates the partial format adaptive input matrix and the partial format adaptive disturbance matrix; and adopts the control scheme of the present invention to control the controlled object under the action of measurable disturbances. Below, the implementation steps of the partial format model-free adaptive disturbance compensation control method for measurable disturbances provided by the present invention are further explained as follows:

[0071] The control method is run on a hardware platform to control a controlled object under a measurable disturbance, wherein the controlled object is a multi-input multi-output system including multiple control inputs and multiple system outputs. The control method is characterized by comprising the following steps:

[0072] Step (1): obtaining a measurable disturbance at k sampling moments, and establishing a dynamic linearized data model of the controlled object under the action of the measurable disturbance, wherein the dynamic linearized data model of the controlled object includes a pseudo-Jacobi input matrix θ(k) and a pseudo-Jacobi disturbance matrix χ(k);

[0073] Step (2): constructing a cost function and solving the cost function using a function extremum method, optimizing and updating the pseudo-Jacobian input matrix θ(k) and the pseudo-Jacobian perturbation matrix χ(k) in step (1);

[0074] Step (3): Using the measurable disturbance at the k sampling moment, based on the dynamic linearized data model of the controlled object after optimizing the pseudo-Jacobi input matrix θ(k) and the pseudo-Jacobi disturbance matrix χ(k) in step (2), a partial format model-free adaptive disturbance compensation control scheme for the measurable disturbance is designed, wherein the control scheme includes the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p (k);

[0075] Step (4): Construct an energy function and solve the energy function using the momentum gradient descent method to optimize and update the partial format adaptive input matrix π in step (3) p (k) and the partial format adaptive perturbation matrix ω p (k);

[0076] Step (5): Optimize the partial format adaptive input matrix π using step (4) p (k) and the partial format adaptive perturbation matrix ω p The control scheme after (k) controls the controlled object under the action of measurable disturbance, weakens the influence of measurable disturbance on the actual value of the controlled object system output, and realizes effective tracking of the expected value of the system output.

[0077] Furthermore, the measurable disturbance at k sampling moments is obtained in step (1), and the dynamic linearized data model of the controlled object under the action of the measurable disturbance is established as follows:

[0078] Δy(k+1)=θ(k)Δu(k)+χ(k)Δd(k)

[0079] Where k is the sampling time, k is a positive integer; y(k+1) is the actual value vector of the system output of the controlled object at the k+1 sampling time, y(k+1)=[y1(k+1),…,y n (k+1)] T , Δy(k+1)=y(k+1)-y(k); n is the total number of system outputs of the controlled object, n is an integer greater than 1; u(k) is the control input vector of the controlled object at sampling time k, u(k)=[u1(k),…,u m (k)] T, Δu(k)=u(k)-u(k-1); m is the total number of control inputs of the controlled object, and m is an integer greater than 1; d(k) is the measurable disturbance vector of the controlled object at sampling time k, d(k)=[d1(k),…,d q (k)] T , Δd(k)=d(k)-d(k-1); q is the total number of measurable disturbances to the controlled object, q is a positive integer; θ(k) is the pseudo-Jacobian input matrix at sampling time k, and χ(k) is the pseudo-Jacobian disturbance matrix χ(k) at sampling time k.

[0080] The process of constructing a cost function in step (2) and solving the cost function using a function extremum method, and optimizing and updating the pseudo-Jacobian input matrix θ(k) and the pseudo-Jacobian perturbation matrix χ(k) in step (1) mainly includes the following steps:

[0081] Step (2.1): Input matrix θ(k) to the pseudo-Jacobi and construct the cost function

[0082] J(θ(k))=||Δy(k)-θ(k)Δu(k-1)-χ(k-1)Δd(k-1)|| 2 +μ1||Δθ(k)|| 2

[0083] Among them, μ1 is the first weight factor;

[0084] Step (2.2): For the pseudo-Jacobi perturbation matrix χ(k), construct a cost function

[0085] J(χ(k))=||Δy(k)-θ(k-1)Δu(k-1)-χ(k)Δd(k-1)|| 2 +μ2||Δχ(k)|| 2

[0086] Among them, μ2 is the second weight factor;

[0087] Step (2.3): Use the function extremum method to solve the cost function described in step (2.1) and optimize and update the pseudo-Jacobi input matrix θ(k).

[0088]

[0089] Among them, α1 is the first step size factor;

[0090] Step (2.4): Use the function extremum method to solve the cost function described in step (2.2) and optimize and update the pseudo-Jacobi perturbation matrix χ(k).

[0091]

[0092] Among them, α2 is the second step size factor.

[0093] The measurable disturbance at the k sampling moment described in step (3) is used, based on the dynamic linearized data model of the controlled object after optimizing the pseudo-Jacobi input matrix θ(k) and the pseudo-Jacobi disturbance matrix χ(k) in step (2), to design a partial format model-free adaptive disturbance compensation control scheme for the measurable disturbance:

[0094] u(k)=u(k-1)+π p (k)ΔE(k)+ω p (k)ΔD(k)

[0095] Where, ΔE(k)=[-e(k) T ,Δe(k)…,Δe(k-L+2) T ] T ,ΔD(k)=[Δd(k) T ,…,Δd(k-L+1) T ] T ; e(k) is the system error vector of the controlled object at sampling time k, e(k) = y * (k)-y(k), e(k)=[e1(k),…,e n (k)] T , Δe(k)=e(k)-e(k-1); L is the linearization length constant; L is a positive integer; π p (k) is the adaptive input matrix of the partial format at sampling time k, ω p (k) is the adaptive perturbation matrix of the partial format at sampling time k.

[0096] The energy function described in step (4) is constructed and solved by the momentum gradient descent method, and the partial format adaptive input matrix π described in step (3) is optimized and updated. p (k) and the partial format adaptive perturbation matrix ω p (k), mainly including the following steps:

[0097] Step (4.1): Construct the energy function

[0098]

[0099] Among them, y * (k+1) is the system output expected value vector of the controlled object at the k+1 sampling time, λ is the penalty factor;

[0100] Step (4.2): Use the momentum gradient descent method to solve the energy function described in step (4.1) and optimize and update the partial format adaptive input matrix πp (k),

[0101]

[0102] Among them, σ1 is the first learning rate, η1 is the first momentum factor; Δπ p (k-1)=π p (k-1)-π p (k-2);

[0103] The energy function W is π p Partial derivative of (k-1);

[0104] Step (4.3): Use the momentum gradient descent method to solve the energy function described in step (4.1) and optimize and update the partial format adaptive perturbation matrix ω p (k),

[0105]

[0106] Among them, σ2 is the second learning rate, η2 is the second momentum factor; Δω p (k-1)=ω p (k-1)-ω p (k-2); The energy function W is the energy function of ω p (k-1) partial derivative.

[0107] The energy function W described in step (4.2) is p The formula for calculating the partial derivative of (k-1) is:

[0108]

[0109] The energy function W described in step (4.3) is p The formula for calculating the partial derivative of (k-1) is:

[0110]

[0111] described The mathematical formula is:

[0112] The step (5) described in step (4) is to optimize the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p The control scheme after (k) controls the controlled object under the action of the measurable disturbance and includes the following steps at each sampling time k:

[0113] Step (5.1): Obtain the measurable disturbance vector d(k) at the current sampling moment;

[0114] Step (5.2): Get the expected value vector y of the system output at the current sampling moment * (k), the system output actual value vector y(k), and the system error vector e(k) at the current sampling moment is calculated;

[0115] Step (5.3): Based on steps (5.1) and (5.2), use step (4) to optimize the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p (k) and the control scheme after calculation to obtain the control input vector u(k) at the current sampling moment;

[0116] Step (5.4): After the control input vector acts on the controlled object, the actual value vector of the system output of the controlled object at the next sampling moment is obtained.

[0117] Figure 2 This is the engineering application system block diagram of the present invention. Figure 2 The hardware platform in the engineering application system block diagram, Figure 3 A schematic diagram of the hardware platform for running the present invention is given in the figure. Specifically, the present invention adopts a non-transitory computer-readable storage medium on which a computer program is stored, and is characterized in that when the computer program is executed by a processor, the above-mentioned partial format model-free adaptive disturbance compensation control method for measurable disturbances is implemented; the present invention adopts an electronic device, including a memory, a processor and a computer program stored in the memory and capable of running on the processor, and is characterized in that when the processor executes the program, the above-mentioned partial format model-free adaptive disturbance compensation control method for measurable disturbances is implemented.

[0118] The following are two specific embodiments of the present invention. Specific embodiment 1:

[0120] The controlled object adopts a two-input and two-output nonlinear system:

[0121]

[0122]

[0123]

[0124]

[0125] y1(k+1)=x 11 (k+1)

[0126] y2(k+1)=x21 (k+1)

[0127] Here, a(k) = 1 + 0.1 sin(2πk / 1500) and b(k) = 1 + 0.1 cos(2πk / 1500) are two time-varying parameters; d1(k) = 0.15 sin(k / 10) and d2(k) = 0.15 sin(k / 10) are measurable disturbances. Therefore, the controlled object's two-input, two-output nonlinear system is a two-input, two-output nonlinear system under the action of a measurable disturbance.

[0128] System output expected value trajectory y * (k) as follows:

[0129]

[0130]

[0131] In specific embodiment 1, m=n=q=2.

[0132] In order to more clearly compare the control performance of the control method of the present invention and the comparative control method, the time-weighted integral absolute error (ITAE) is used as the control performance evaluation index:

[0133]

[0134] in, is the expected value of the j-th system output at the k-th sampling time, y j (k) is the actual value of the j-th system output at the k-th sampling time, j = 1, ..., n. ITAE(e j ) is smaller, indicating that the jth system outputs the actual value y j (k) and the expected value of the j-th system output The error is generally smaller, the control accuracy and speed are higher, and the control performance is better.

[0135] The hardware platform for running the control method of the present invention adopts an industrial control computer.

[0136] The control method of the present invention is used to control a two-input two-output system. The control method parameters of the present invention are set as follows: L = 3, θ(1) = [0.6, -0.05; 0.1, 0.6], χ(1) = [0.2, 0; 0, 0.2], π p (1)=[-0.35,0,0.1,0,0.08,0;0,-0.27,0,0.1,0,0.05],ω p(1)=[-0.35,0,0.01,0,0.01,0;0,-0.9,0,0.05,0,0.01], α1=0.5, α2=0.5, μ1=1, μ2=0.9, σ1=0.7, σ2=0.5, eta1=0.25, eta2=0.25, λ=2.

[0137] When the control method of the present invention is used to control a two-input two-output system under the action of a measurable disturbance, the following steps are included at each sampling time k: a) obtaining the measurable disturbance vector d(k) at the current sampling time; b) obtaining the system output expected value vector y at the current sampling time * (k), the system output actual value vector y(k), and the system error vector e(k) at the current sampling moment are calculated; c) based on steps a) and b), the step (4) is used to optimize the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p (k) The control scheme after calculation obtains the control input vector u(k) at the current sampling moment; d) After the control input vector acts on the two-input two-output system, the system output actual value vector of the two-input two-output system at the next sampling moment is obtained; e) Repeat a) to d) until the end of the sampling moment.

[0138] The control effects of the control method of the present invention and the existing PID control method (comparison control method) are compared as follows: Figure 4 This is a control effect diagram of the first system output when the control method of the present invention and the comparative control method are used. Figure 5 This is a control effect diagram of the second system output when the control method of the present invention and the comparative control method are used. Figure 6 is the first control input curve when the control method of the present invention and the comparative control method are used. Figure 7The second control input curve when the control method of the present invention and the comparative control method are used; from the control performance evaluation index, the ITAE (e1) output by the first system using the control method of the present invention is 12090, and the ITAE (e2) output by the second system is 10029, while the ITAE (e1) output by the first system using the comparative control method is 24067, and the ITAE (e2) output by the second system is 25998. The control performance evaluation index results are listed in Table 1; from the system output curve, the control method of the present invention can effectively suppress the influence of measurable disturbances on the actual output values ​​of the two-input and two-output system, and the control performance of the control method of the present invention is better than that of the comparative control method. Based on the above investigation, it is fully demonstrated that the partial format model-free adaptive disturbance compensation control method for measurable disturbances provided by the present invention can significantly weaken the influence of measurable disturbances on the actual output values ​​of the controlled object system, achieve effective tracking of the expected value trajectory, and significantly improve the disturbance compensation control performance.

[0139] Table 1 Comparison of control performance of two-input two-output system

[0140] Specific embodiment 2:

[0142] Vapor Compression Refrigeration Systems (VCRS) are the most common refrigeration cycle equipment, widely used in households (such as household refrigerators and air conditioners), businesses (such as building and car air conditioners, cold storage warehouses) and industries (such as petrochemical plants and natural gas processing plants). Figure 8 The two disturbances in the refrigeration cycle are the inlet temperature of the cooling medium and the inlet temperature of the cooled medium. With the widespread use of high-energy refrigeration equipment, achieving disturbance compensation control in vapor compression refrigeration systems is of great significance for promoting energy conservation and consumption reduction in my country and around the world.

[0143] The controlled vapor compression refrigeration system is a two-input, two-output nonlinear system. Its two control inputs, u1 and u2, are the compressor frequency (Hz) and valve opening (%), respectively. Its two system outputs, y1 and y2, are the superheat (°C) and the outlet temperature of the cooled medium (°C), respectively. The two disturbances, d1 and d2, to which the controlled vapor compression refrigeration system is subjected are the inlet temperature of the cooling medium (°C) and the inlet temperature of the cooled medium (°C), respectively. d1 and d2 are measurable disturbances, measured online using corresponding temperature sensors. Figure 9The two measurable disturbance curves for the vapor compression refrigeration system are shown in Figure 1. Therefore, the controlled vapor compression refrigeration system is a two-input, two-output nonlinear system under the influence of measurable disturbances. In Specific Example 2, m = n = q = 2. The hardware platform for running the control method of the present invention is an industrial control computer.

[0144] The initial operating conditions of the controlled object vapor compression refrigeration system are: u1(0) ​​= 36.45Hz, u2(0) = 48.79%, y1(0) = 14.65℃, y2(0) = -22.15℃. In order to meet the cooling demand of the cooled medium, the system outputs the expected value trajectory At the 2nd minute, the temperature was adjusted from 14.65℃ to 7.2℃ in steps. At the 9th minute, the temperature was adjusted from 7.2℃ to 22.2℃ in steps. Finally, at the 16th minute, the temperature was adjusted from 22.2℃ to 11.65℃ in steps. The system outputs the expected value trajectory. At the 2nd minute, the temperature was adjusted stepwise from -22.15°C to -22.65°C.

[0145] The control method of the present invention is used to control the vapor compression refrigeration system. The control method parameters of the present invention are set as follows: L = 2, θ (1) = [2, 0; 0, 0.1], χ (1) = [0.2, 0; 0, 0.2], π p (1)=[-2,0,0.01,0;0,-1,0,0],ω p (1)=[-1.3,0,0.2,0.3;0,-0.05,0.03,0], α1=0.5, α2=0.5, μ1=1, μ2=1, σ1=0.5, σ2=0.9, η1=0.2, η2=0.2, λ=0.1.

[0146] When the control method of the present invention is used to control a vapor compression refrigeration system under a measurable disturbance, the following steps are included at each sampling time k: a) obtaining the measurable disturbance vector d(k) at the current sampling time; b) obtaining the system output expected value vector y at the current sampling time * (k), the system output actual value vector y(k), and the system error vector e(k) at the current sampling moment are calculated; c) based on steps a) and b), the step (4) is used to optimize the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p (k) The control scheme after calculation obtains the control input vector u(k) at the current sampling moment; d) After the control input vector acts on the steam compression refrigeration system, the system output actual value vector of the steam compression refrigeration system at the next sampling moment is obtained; e) repeat a) to d) until the end of the sampling moment.

[0147] The control effects of the control method of the present invention and the existing PID control method (comparison control method) are compared as follows: Figure 10 This is a control effect diagram of the first system output when the control method of the present invention and the comparative control method are used. Figure 11 This is a control effect diagram of the second system output when the control method of the present invention and the comparative control method are used. Figure 12 is the first control input curve when the control method of the present invention and the comparative control method are used. Figure 13 The second control input curve when the control method of the present invention and the comparative control method are used; from the control performance evaluation index, the ITAE (e1) output by the first system using the control method of the present invention is 120374, and the ITAE (e2) output by the second system is 5609. The ITAE (e1) output by the first system using the comparative control method is 471529, and the ITAE (e2) output by the second system is 52531. The control performance evaluation index results are listed in Table 2; from the system output curve, the control method of the present invention can effectively suppress the influence of the measurable disturbance on the actual value output of the steam compression refrigeration system, and the control performance of the control method of the present invention is better than that of the comparative control method. Based on the above investigation, it is fully demonstrated that the partial format model-free adaptive disturbance compensation control method for measurable disturbances provided by the present invention can significantly weaken the influence of the measurable disturbance on the actual value output of the controlled object system, realize the effective tracking of the expected value trajectory, and significantly improve the disturbance compensation control performance.

[0148] Table 2 Comparison of control performance of vapor compression refrigeration system

[0149]

[0150] Furthermore, the following two points should be particularly noted:

[0151] (1) Disturbances are widely present in practical control systems, such as most controlled objects in industries such as oil refining, petrochemicals, chemicals, pharmaceuticals, food, papermaking, water treatment, thermal power, metallurgy, cement, rubber, machinery, electrical, transportation, and robotics, including reactors, distillation towers, machines, equipment, devices, production lines, workshops, factories, unmanned vehicles, unmanned ships, drones, and autonomous mobile robots. For example, a vapor compression refrigeration system is subject to the continuous and complex influence of two measurable disturbances: the cooling medium inlet temperature and the cooled medium inlet temperature. Specific Example 2 shows that the control method of the present invention can significantly reduce the influence of the measurable disturbance on the actual output value of the controlled object system, achieve effective tracking of the expected value trajectory, and thus significantly improve the disturbance compensation control performance. For example, unmanned boats are extremely susceptible to the influence of the wind field on the water surface during operation. Changes in wind speed and direction will not only affect the speed and heading of the unmanned boat, but may also cause the unmanned boat to capsize in severe cases. Online monitoring of the two measurable disturbances of wind speed and wind direction can be implemented based on wind speed sensors and wind direction sensors. The control method of the present invention can be used to compensate for the measurable disturbances and achieve smooth operation of the unmanned boat, which is of great significance to improving the safety and reliability of the unmanned boat.

[0152] (2) In the above-mentioned specific embodiments 1 and 2, the hardware platform for running the control method of the present invention is an industrial control computer; in actual application, according to specific circumstances, any one of a single-chip microcomputer controller, a microprocessor controller, a field programmable gate array controller, a digital signal processing controller, an embedded system controller, a programmable logic controller, a distributed control system, a fieldbus control system, an industrial Internet of Things control system, and an industrial Internet control system or any combination thereof can be selected as the hardware platform for running the control method of the present invention.

[0153] Through the description of the above embodiments, it will be clear to those skilled in the art that the implementation of the present invention can be implemented by means of software plus the necessary hardware platform. The embodiments of the present invention can be implemented using an existing processor, or by a dedicated processor used for this or other purposes for an appropriate system, or by a hard-wired system. The embodiments of the present invention also include non-transitory computer-readable storage media, which include machine-readable media for carrying or having machine-executable instructions or data structures stored thereon; such machine-readable media can be any available medium that can be accessed by a general-purpose or special-purpose computer or other machine with a processor. For example, such machine-readable media can include RAM, ROM, EPROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store the required program code in the form of machine-executable instructions or data structures and can be accessed by a general-purpose or special-purpose computer or other machine with a processor. When information is transmitted or provided to a machine via a network or other communication connection (hard-wired, wireless, or a combination of hard-wired and wireless), the connection is also considered a machine-readable medium.

[0154] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A partial format model-free adaptive disturbance compensation control method for measurable disturbances, wherein the control method operates on a hardware platform to control a controlled object under the action of a measurable disturbance, wherein the controlled object is a multi-input multi-output system including multiple control inputs and multiple system outputs. The control method is characterized by comprising the following steps: Step (1): obtaining a measurable disturbance at k sampling moments, and establishing a dynamic linearized data model of the controlled object under the action of the measurable disturbance, wherein the dynamic linearized data model of the controlled object includes a pseudo-Jacobi input matrix θ(k) and a pseudo-Jacobi disturbance matrix χ(k); Step (2): constructing a cost function and solving the cost function using a function extremum method, optimizing and updating the pseudo-Jacobian input matrix θ(k) and the pseudo-Jacobian perturbation matrix χ(k) in step (1); Step (3): Using the measurable disturbance at the k sampling moment, based on the dynamic linearized data model of the controlled object after optimizing the pseudo-Jacobi input matrix θ(k) and the pseudo-Jacobi disturbance matrix χ(k) in step (2), a partial format model-free adaptive disturbance compensation control scheme for the measurable disturbance is designed, wherein the control scheme includes the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p (k); Step (4): Construct an energy function and solve the energy function using the momentum gradient descent method to optimize and update the partial format adaptive input matrix π in step (3) p (k) and the partial format adaptive perturbation matrix ω p (k); Step (5): Optimize the partial format adaptive input matrix π using step (4) p (k) and the partial format adaptive perturbation matrix ω p The control scheme after (k) controls the controlled object under the action of measurable disturbance, weakens the influence of measurable disturbance on the actual value of the controlled object system output, and realizes effective tracking of the expected value of the system output.

2. The method for biased model-free adaptive disturbance compensation control for measurable disturbances according to claim 1, characterized in that: The measurable disturbance at k sampling moments described in step (1) is obtained, and the dynamic linearized data model of the controlled object under the action of the measurable disturbance is established as follows: Δy(k+1)=θ(k)Δu(k)+χ(k)Δd(k) Where k is the sampling time, k is a positive integer; y(k+1) is the actual value vector of the system output of the controlled object at the k+1 sampling time, y(k+1)=[y1(k+1),…,y n (k+1)] T , Δy(k+1)=y(k+1)-y(k); n is the total number of system outputs of the controlled object, n is an integer greater than 1; u(k) is the control input vector of the controlled object at sampling time k, u(k)=[u1(k),…,u m (k)] T , Δu(k)=u(k)-u(k-1); m is the total number of control inputs of the controlled object, and m is an integer greater than 1; d(k) is the measurable disturbance vector of the controlled object at sampling time k, d(k)=[d1(k),…,d q (k)] T , Δd(k)=d(k)-d(k-1); q is the total number of measurable disturbances to the controlled object, q is a positive integer; θ(k) is the pseudo-Jacobian input matrix at sampling time k, and χ(k) is the pseudo-Jacobian disturbance matrix χ(k) at sampling time k.

3. The method for biased model-free adaptive disturbance compensation control for measurable disturbances according to claim 1, characterized in that: The process of constructing a cost function in step (2) and solving the cost function using a function extremum method, and optimizing and updating the pseudo-Jacobian input matrix θ(k) and the pseudo-Jacobian perturbation matrix χ(k) in step (1) mainly includes the following steps: Step (2.1): Input matrix θ(k) to the pseudo-Jacobi and construct the cost function J(θ(k))=||Δy(k)-θ(k)Δu(k-1)-χ(k-1)Δd(k-1)|| 2 +μ1||Δθ(k)|| 2 Among them, μ1 is the first weight factor; Step (2.2): For the pseudo-Jacobi perturbation matrix χ(k), construct a cost function J(χ(k))=||Δy(k)-θ(k-1)Δu(k-1)-χ(k)Δd(k-1)|| 2 +μ2||Δχ(k)|| 2 Among them, μ2 is the second weight factor; Step (2.3): Use the function extremum method to solve the cost function described in step (2.1) and optimize and update the pseudo-Jacobi input matrix θ(k). Among them, α1 is the first step size factor; Step (2.4): Use the function extremum method to solve the cost function described in step (2.2) and optimize and update the pseudo-Jacobi perturbation matrix χ(k). Among them, α2 is the second step size factor.

4. The method for biased model-free adaptive disturbance compensation control for measurable disturbances according to claim 1, characterized in that: The measurable disturbance at the k sampling moment described in step (3) is used, based on the dynamic linearized data model of the controlled object after optimizing the pseudo-Jacobi input matrix θ(k) and the pseudo-Jacobi disturbance matrix χ(k) in step (2), to design a partial format model-free adaptive disturbance compensation control scheme for the measurable disturbance: u(k)=u(k-1)+π p (k)ΔE(k)+ω p (k)ΔD(k) Where, ΔE(k)=[-e(k) T ,Δe(k)…,Δe(k-L+2) T ] T ,ΔD(k)=[Δd(k) T ,…,Δd(k-L+1) T ] T ; e(k) is the system error vector of the controlled object at sampling time k, e(k) = y * (k)-y(k), e(k)=[e1(k),…,e n (k)] T , Δe(k)=e(k)-e(k-1); L is the linearization length constant; L is a positive integer; π p (k) is the adaptive input matrix of the partial format at sampling time k, ω p (k) is the adaptive perturbation matrix of the partial format at sampling time k.

5. The method for biased model-free adaptive disturbance compensation control for measurable disturbances according to claim 1, characterized in that: The energy function described in step (4) is constructed and solved by the momentum gradient descent method, and the partial format adaptive input matrix π described in step (3) is optimized and updated. p (k) and the partial format adaptive perturbation matrix ω p (k), mainly including the following steps: Step (4.1): Construct the energy function Among them, y * (k+1) is the system output expected value vector of the controlled object at the k+1 sampling time, λ is the penalty factor; Step (4.2): Use the momentum gradient descent method to solve the energy function described in step (4.1) and optimize and update the partial format adaptive input matrix π p (k), Among them, σ1 is the first learning rate, η1 is the first momentum factor; Δπ p (k-1)=π p (k-1)-π p (k-2); The energy function W is π p Partial derivative of (k-1); Step (4.3): Use the momentum gradient descent method to solve the energy function described in step (4.1) and optimize and update the partial format adaptive perturbation matrix ω p (k), Among them, σ2 is the second learning rate, η2 is the second momentum factor; Δω p (k-1)=ω p (k-1)-ω p (k-2); The energy function W is the energy function of ω p (k-1) partial derivative.

6. The method for biased model-free adaptive disturbance compensation control for measurable disturbances according to claim 5, characterized in that: The energy function W described in step (4.2) is p The formula for calculating the partial derivative of (k-1) is: The energy function W described in step (4.3) is p The formula for calculating the partial derivative of (k-1) is:

7. The method for biased model-free adaptive disturbance compensation control for measurable disturbances according to claim 6, characterized in that: described The mathematical formula is:

8. The method for partial format model-free adaptive disturbance compensation control for measurable disturbances according to claim 1, characterized in that: The step (5) described in step (4) is to optimize the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p The control scheme after (k) controls the controlled object under the action of the measurable disturbance and includes the following steps at each sampling time k: Step (5.1): Obtain the measurable disturbance vector d(k) at the current sampling moment; Step (5.2): Get the expected value vector y of the system output at the current sampling moment * (k), the system output actual value vector y(k), and the system error vector e(k) at the current sampling moment is calculated; Step (5.3): Based on steps (5.1) and (5.2), use step (4) to optimize the partial format adaptive input matrix π p (k) and the partial format adaptive perturbation matrix ω p (k) and the control scheme after calculation to obtain the control input vector u(k) at the current sampling moment; Step (5.4): After the control input vector acts on the controlled object, the actual value vector of the system output of the controlled object at the next sampling moment is obtained.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for biased format model-free adaptive disturbance compensation control for measurable disturbances according to any one of claims 1 to 8 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the partial format model-free adaptive disturbance compensation control method for measurable disturbances according to any one of claims 1 to 8 is implemented.

Images