Liquid level safety control method based on unknown input observer

Abstract

The invention relates to a liquid level safety control method based on an unknown input observer. The method comprises the following steps: establishing a state equation of liquid level change under spoofing attack and a corresponding sensor measurement output equation; constructing a random network attack model by utilizing a Bernoulli variable; performing output transformation on the sensor spoofing attack signal; selecting a non-singular state transformation matrix for state transformation, and designing an interval observer; designing an unknown input observer based on an output signal of the interval observer, and estimating an actual liquid level state of the liquid level control system, an attack signal received by an actuator and an attack signal received by a sensor; and based on the estimation signal, designing a compensation controller for adjusting the liquid level state change and carrying out liquid level control. Compared with the prior art, unknown input reconstruction design is carried out for spoofing attacks, then the corresponding compensation controller is designed, the influence of an attack system on the system is greatly reduced, and therefore safe operation of the liquid level control system is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of safety control of information-physical systems, and in particular to a liquid level safety control method based on an unknown input observer. Background Art

[0002] The rapid development of computer technology and information technology has had a profound impact on people's production and lifestyle. A type of information-physical system consisting of a decision-making layer, a network layer, and a physical layer has attracted close attention from all walks of life due to its wide application in scenarios such as device interconnection, smart homes, robots, smart navigation, and smart factories. The decision-making layer of the information-physical system mainly generates control instructions based on pre-designed decision-making control rules and perception information of the physical world; the physical layer mainly includes devices such as actuators and sensors, which respectively realize the manipulation and real-time perception of physical objects; the network layer mainly realizes real-time communication of control instructions from the decision-making layer and perception information from the physical layer.

[0003] like Figure 2 As shown in the figure, the liquid level control system consists of a decision-making layer, a physical layer, and a network layer. Although the use of communication networks reduces the requirements for time and space for the implementation and deployment of information-physical systems, it also brings many problems to the security control of information-physical systems. For example, information is vulnerable to network attacks during network transmission. In the reaction vessels of industrial production, ensuring a safe liquid level is crucial for safe production. When the liquid level control system based on the information-physical system is attacked by a network, the actual liquid level of the reaction vessel will gradually move away from the expected safe production level, and it is easy for the production raw materials in the container to be insufficient or excessively overflowed, which will lead to reduced production efficiency and low product qualification rate, and even cause production accidents and casualties. Therefore, in order to ensure the safe production of the liquid level control system, reduce the economic losses caused by network attacks, and avoid safety accidents, it is of great significance to propose an effective liquid level safety control method. Summary of the invention

[0004] The purpose of the present invention is to provide a liquid level safety control method based on an unknown input observer to solve the problem of liquid level safety control under deception attacks.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] As a first aspect of the present invention, a liquid level safety control method based on an unknown input observer is provided, the steps comprising:

[0007] Collect data from the liquid level control system, establish the state equation of liquid level change under deception attack and the corresponding sensor measurement output equation;

[0008] Using Bernoulli random variables to build an actuator and sensor deception attack model;

[0009] Perform output transformation on the sensor deception attack signal to weaken the influence of the deception attack on the sensor measurement signal;

[0010] Select a non-singular state transformation matrix for state transformation and design an interval observer;

[0011] Based on the output signal of the interval observer, the algebraic reconstruction relationship between the unknown input and the liquid level state is established, and then the unknown input observer is designed to estimate the actual liquid level state of the liquid level control system, the attack signal of the actuator, and the attack signal of the sensor.

[0012] Based on the estimated signal, a compensation controller is designed to adjust the change of the liquid level state, and the liquid level control is performed based on the compensation controller.

[0013] As a preferred technical solution, the state equation of the liquid level change under the deception attack is expressed as:

[0014]

[0015] Where x(k) is an n-dimensional real vector, representing the deviation of the n reaction containers from their expected liquid levels at the kth moment; is the control input received by the m actuators from the network layer at the kth moment; w(k) is the unknown interference of the external environment on the actuator input at the kth moment, and its upper and lower bounds satisfy is the corresponding gain matrix.

[0016] As a preferred technical solution, the sensor measurement output equation is expressed as:

[0017] y(k)=Cx(k)+Ev(k)

[0018] Among them, y(k) represents the p measurement information of the liquid level of the reaction container in the physical layer sent by the sensor through the network layer and received by the decision layer at the kth moment; v(k) represents the q deception attack signals received by the sensor in the network layer at the kth moment; is the corresponding gain matrix.

[0019] As a preferred technical solution, the use of Bernoulli random variables to construct an actuator and sensor deception attack model includes:

[0020] Using Bernoulli variables to build a sensor attack model:

[0021]

[0022] Using Bernoulli variables to build an attack model for the actuator:

[0023]

[0024] in, and is the deception attack signal, and the expectations of α(k) and β(k) are and θ(k) satisfies the upper and lower limits. Unknown signal;

[0025] Then the state equation of the liquid level control system is rewritten as:

[0026] x(k+1)=Ax(k)+Bu(k)+Bd(k)

[0027] y(k)=Cx(k)+Ev(k)

[0028] Among them, the unknown input satisfy d = w , u and Obtained from the physical parameters of the actuator device itself.

[0029] As a preferred technical solution, the output transformation of the sensor spoofing attack signal is performed to weaken the influence of the spoofing attack on the sensor measurement signal, as follows:

[0030]

[0031] in, is the Moore-Penrose inverse of E, satisfying

[0032] Then the state equation of the liquid level control system is further rewritten as:

[0033]

[0034] As a preferred technical solution, the non-singular state transformation matrix is ​​selected to perform state transformation, and the specific steps are as follows:

[0035] Select a non-singular state transformation matrix The transformed state is Then the state equation of the liquid level control system is transformed into:

[0036]

[0037] As a preferred technical solution, the interval observer is designed as follows:

[0038]

[0039] in, and ζ (k) are the states of the upper bound interval observer and the lower bound interval observer, respectively. yes The i-th row and j-th column element of Represents the transformation matrix The inverse matrix of .

[0040] As a preferred technical solution, the designed unknown input observer specifically includes actual liquid level state observer, unknown input reconstruction and sensor deception attack signal estimation:

[0041]

[0042] in, and are the estimated values ​​of the actual liquid level state x(k) and the unknown input d(k), respectively; L is the unknown input observer gain matrix to be designed; γ(k) = [γ 1 (k)γ 2 (k)…γ p (k)], when Time i (k) = 1, otherwise γ i (k) = 0, express is the i-th element, i=1,2,…,p;

[0043] As a preferred technical solution, the compensation controller adopts PID control for adjusting the change of liquid level state, and the PID compensation control law based on the unknown input observer is:

[0044]

[0045] in, are the gain matrices of the proportional term, integral term, and differential term to be designed, and l is the length of the integral time interval of the integral term.

[0046] As a preferred technical solution, the liquid level control is performed based on the compensation controller, as shown in the figure below:

[0047] Under the PID compensation control law based on the unknown input observer, the following symbols are defined:

[0048]

[0049] Then, the augmented system is obtained as:

[0050]

[0051]

[0052] Select the Lyapunov function as:

[0053] V(ψ(k))=V 1 (ψ(k))+V 2 (ψ(k))

[0054] in,

[0055] According to the Lyapunov function, the stability condition of liquid level safety control is obtained:

[0056]

[0057] in, is the matrix to be calculated, is a positive definite symmetric matrix,

[0058] According to the stability condition of liquid level safety control, the gain matrix of the unknown input observer is calculated as:

[0059]

[0060] The gain matrix of the PID compensation controller based on the unknown input observer is:

[0061]

[0062] Among them, W 11 It is a reversible matrix calculated based on the stability conditions of liquid level safety control.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1) Aiming at the problem of safe control of liquid level information-physical system under deception attack, the present invention firstly performs unknown algebraic reconstruction on the unknown deception attack signal and establishes the relationship between the unknown deception attack signal and the actual liquid level state, and then designs an unknown input observer to obtain the unknown input reconstruction and sensor attack signal estimation. Based on the estimated information of the unknown input observer, a corresponding PID compensation controller is designed to reduce the impact of network attacks and unknown interference on the system, thereby ensuring the safe production of the liquid level control system.

[0065] 2) The present invention algebraically reconstructs the actuator deception attack signal and then performs compensation control to avoid irreversible accidents caused by deception attacks on the liquid level system. Compared with the traditional method of augmenting the state and sensor attack signal, the sensor attack signal is output transformed to obtain estimated information, which decouples the unknown input of the system without augmenting the system dimension, thereby reducing the complexity of the design. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a flow chart of a liquid level safety control method based on an unknown input observer in the present invention;

[0067] Figure 2 A schematic diagram of the relationship between the various layers in the liquid level information physical system of the present invention;

[0068] Figure 3 It is a technical framework diagram of the liquid level safety control method based on the unknown input observer in the present invention;

[0069] Figure 4 This is a diagram of a liquid level system simulation embodiment of the present invention;

[0070] Figure 5 A diagram showing the change in liquid level and its estimated change trajectory in a specific embodiment of the present invention;

[0071] Figure 6 Reconstructing a trajectory change graph for an unknown input in a specific embodiment of the present invention;

[0072] Figure 7 It is a diagram of the sensor attack signal and its estimated change trajectory in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0073] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0074] Example 1

[0075] Aiming at the problem of liquid level safety control under deception attack, the present invention proposes a liquid level safety control method based on unknown input observer. The flow chart of the method is as follows: Figure 1 shown. Figure 2 It is a liquid level information-physical system structure. The liquid level safety control system mainly includes an information layer and a physical layer. Figure 3 The liquid level safety control method based on unknown input observer provided by the present invention mainly realizes the safety control of liquid level by designing unknown input observer and PID compensation controller to ensure safe production. The specific implementation steps are as follows:

[0076] Step S1, analyzing the dynamic relationship of the liquid level change in the reaction container, establishing the state equation of the liquid level change under the deception attack and the corresponding sensor measurement output equation:

[0077] S1.1: Collect data from the liquid level system, analyze the dynamic relationship of the liquid level change in the reaction vessel, and establish the state equation:

[0078]

[0079] Where x(k) is an n-dimensional real vector, which represents the deviation of the n reaction containers from their expected liquid levels at the kth moment. is the control input received by the m actuators from the network layer at the kth time, and w(k) is the unknown interference from the outside world to the actuator input at the kth time, which can be called the upper and lower bounds on the control channel. and w Known and Bounded unknown input, is the corresponding gain matrix;

[0080] S1.2: According to the actual working conditions, the sensor measurement output equation is:

[0081] y(k)=Cx(k)+Ev(k) (2)

[0082] Among them, y(k) represents the p measurement information of the liquid level of the reaction container in the physical layer sent by the sensor through the network layer and received by the decision layer at the kth moment, and v(k) represents the q spoofing attack signals received by the sensor in the network layer at the kth moment. is the corresponding gain matrix, and both C and E can be obtained from the sensor layout on site.

[0083] Step S2: Use Bernoulli variables α(k) and β(k) to construct a random network attack model to describe the deceptive attack process on actuators and sensors:

[0084] S2.1: Introduce Bernoulli variables α(k) and β(k), including the attack model of the sensor:

[0085]

[0086] The attack model on the actuator is:

[0087]

[0088] in, is the deception attack signal, and the expectations of α(k) and β(k) are and θ(k) satisfies The unknown signal θ and is known;

[0089] S2.2: Order is the new unknown input of the system. Combining equations (1)-(2) in step S1 and equations (3)-(4) in step S2.1, the state equation of the liquid level control system can be rewritten as:

[0090]

[0091] Among them, the unknown input d(k) satisfies u and It can be obtained from the physical parameters of the actuator device itself.

[0092] S3, performing output transformation to weaken the influence of the deception attack on the sensor measurement signal;

[0093] S3.1: Perform output transformation to reduce the impact of cyber attacks on sensors:

[0094]

[0095] in, is the Moore-Penrose inverse of E, satisfying

[0096] S3.2: According to equation (6) in step S3.1, the rewritten state equation of the liquid level control system can be obtained:

[0097]

[0098] S4. Select a non-singular state transformation matrix Perform state transformation and design an interval observer; based on the output signal of the interval observer, establish the algebraic reconstruction relationship between the unknown input and the liquid level state, and then design the unknown input observer, which can estimate the actual liquid level state of the liquid level information-physical system and the attack signal of the actuator, and also obtain the attack signal of the sensor:

[0099] S4.1: Selecting a non-singular state transformation matrix The transformed state is Then the equation (7) in step S3.2 can be transformed into:

[0100]

[0101] S4.2: The interval observer is designed as:

[0102]

[0103] in, and ζ (k) are the states of the upper bound interval observer and the lower bound interval observer, respectively. yes The i-th row and j-th column element of Represents the transformation matrix The inverse matrix of

[0104] S4.3: According to formula (9) in step S4.2, we obtain:

[0105]

[0106] in, yes The i-th row and j-th column elements of , so we have:

[0107]

[0108] Among them, y ai (k) is y a The i-th element of (k), yes The i-th element of y ai (k) Yes y a The i-th element of (k),

[0109]

[0110] remember:

[0111] δ(k)=[δ 1 (k)δ 2 (k)…δ p (k)] T

[0112]

[0113] Thus:

[0114]

[0115] The function diag{χ} represents the conversion of vector χ into a diagonal matrix with the elements in χ as diagonal elements. Based on the above steps, the unknown input algebraic expression can be established:

[0116]

[0117] in,

[0118]

[0119] and

[0120]

[0121] Here, γ(k)=[γ 1 (k) γ 2 (k) … γ p (k)] T ,when Time i (k) = 1, otherwise γ i (k) = 0, express is the i-th element, i=1,2,…,p.

[0122] S4.4: Design the unknown input observer according to step S4.3, including the actual liquid level state observer, unknown input reconstruction and sensor deception attack signal estimation:

[0123]

[0124] in, and are the estimated values ​​of the actual liquid level state x(k) and the unknown input d(k), L is the observer gain matrix to be designed,

[0125] S5. Based on the estimated signal obtained in step S4, a PID (proportional, integral, differential) compensation controller is designed to adjust the change in the liquid level state:

[0126] S5.1: Design the PID compensation control law based on the unknown input observer as:

[0127]

[0128] in, are the gain matrices of the proportional term (P), integral term (I), and differential term (D) to be designed, and l is the length of the integral time interval of the integral term.

[0129] S5.2: Under the control law in step S5.1, first define the following symbols:

[0130]

[0131] Then, the augmented system is obtained as:

[0132]

[0133] in,

[0134]

[0135]

[0136] S5.3: Select the Lyapunov function as:

[0137] V(ψ(k))=V 1 (ψ(k))+V 2 (ψ(k))

[0138] in,

[0139] S5.4: According to step S5.3, the stability condition of liquid level safety control can be derived as follows:

[0140]

[0141] in, is the matrix to be calculated, is a positive definite symmetric matrix,

[0142] S5.5: According to step S5.4, the gain matrix of the unknown input observer is calculated as:

[0143]

[0144] The gain matrix of the PID compensation controller based on the unknown input observer is:

[0145]

[0146] Among them, W 11 is the reversible matrix calculated in step S5.4.

[0147] Apply the above-mentioned technical solution to the actual liquid level system, the simulation system is as follows Figure 4 As shown in the figure, the liquid level system consists of four liquid level tanks, where the pump is used to inject liquid into the liquid level tank, and the valve is used to adjust the amount of liquid in the liquid level tank. The compensation control effect of the liquid level control system under the deception attack is simulated with the help of MATLAB platform to verify the control effect of the method of the present invention. The trajectories of liquid level state change, unknown input reconstruction and sensor attack signal change are shown in Figure 1. Figure 5 to Figure 7 As shown in this embodiment, it can be seen that after adopting the unknown input observer method of the present invention, the unknown input and sensor attack signal can be reconstructed and compensated, thereby reducing the impact of attack and external interference on the liquid level system. The method designed by the present invention has a good control effect.

[0148] The present invention aims at the problem of liquid level safety control under deception attacks and provides a liquid level safety control method based on an unknown input observer. The method establishes a state space model of the cyber-physical system according to the number of liquid level states, performs unknown input reconstruction design for deception attacks, and then designs a corresponding compensation controller to greatly reduce the impact of the attack system on the system, thereby ensuring the safe operation of the liquid level control system and enabling safe production of the liquid level control system.

[0149] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A liquid level safety control method based on an unknown input observer, characterized in that the steps include: Collect data from the liquid level control system, establish the state equation of liquid level change under deception attack and the corresponding sensor measurement output equation; Using Bernoulli random variables to build an actuator and sensor deception attack model; Perform output transformation on the sensor deception attack signal to weaken the influence of the deception attack on the sensor measurement signal; Select a non-singular state transformation matrix for state transformation and design an interval observer; Based on the output signal of the interval observer, the algebraic reconstruction relationship between the unknown input and the liquid level state is established, and then the unknown input observer is designed to estimate the actual liquid level state of the liquid level control system, the attack signal of the actuator, and the attack signal of the sensor. Based on the estimated signal, a compensation controller is designed to adjust the change of the liquid level state, and the liquid level control is performed based on the compensation controller.

2. The liquid level safety control method based on unknown input observer according to claim 1 is characterized in that: The state equation of the liquid level change under the deception attack is expressed as: Where x(k) is an n-dimensional real vector, representing the deviation of the n reaction containers from their expected liquid levels at the kth moment; is the control input received by the m actuators from the network layer at the kth moment; w(k) is the unknown interference of the external environment on the actuator input at the kth moment, and its upper and lower bounds satisfy is the corresponding gain matrix.

3. The liquid level safety control method based on unknown input observer according to claim 1 is characterized in that: The sensor measurement output equation is expressed as: y(k)=Cx(k)+Ev(k) Among them, y(k) represents the p measurement information of the liquid level of the reaction container in the physical layer sent by the sensor through the network layer and received by the decision layer at the kth moment; v(k) represents the q deception attack signals received by the sensor in the network layer at the kth moment; is the corresponding gain matrix.

4. The liquid level safety control method based on unknown input observer according to claim 1 is characterized in that: The use of Bernoulli random variables to construct an actuator and sensor deception attack model includes: Using Bernoulli variables to build a sensor attack model: Using Bernoulli variables to build an attack model for the actuator: in, and is the deception attack signal, and the expectations of α(k) and β(k) are and Is to meet the set upper and lower limits Unknown signal; Then the state equation of the liquid level control system is rewritten as: x(k+1)=Ax(k)+Bu(k)+Bd(k) y(k)=Cx(k)+Ev(k) Among them, the unknown input satisfy u and Obtained from the physical parameters of the actuator device itself.

5. The liquid level safety control method based on unknown input observer according to claim 1 is characterized in that: The output transformation of the sensor deception attack signal is performed to weaken the influence of the deception attack on the sensor measurement signal, as follows: in, is the Moore-Penrose inverse of E, satisfying Then the state equation of the liquid level control system is further rewritten as: x(k+1)=Ax(k)+Bu(k)+Bd(k). y a (k)=C a x(k)。 6. The liquid level safety control method based on unknown input observer according to claim 1 is characterized in that: The specific steps of selecting a non-singular state transformation matrix for state transformation are as follows: Select a non-singular state transformation matrix The transformed state is Then the state equation of the liquid level control system is transformed into:

7. The liquid level safety control method based on unknown input observer according to claim 1 is characterized in that: The interval observer is designed as follows: in, and ζ(k) are the states of the upper bound interval observer and the lower bound interval observer, respectively. yes The i-th row and j-th column element of Represents the transformation matrix The inverse matrix of .

8. The liquid level safety control method based on unknown input observer according to claim 1 is characterized in that: The designed unknown input observer specifically includes the actual liquid level state observer, unknown input reconstruction and sensor deception attack signal estimation: in, and are the estimated values ​​of the actual liquid level state x(k) and the unknown input d(k), respectively; L is the unknown input observer gain matrix to be designed; when Time i (k) = 1, otherwise γ i (k) = 0, express is the i-th element, i=1,2,…,p; 9. The liquid level safety control method based on unknown input observer according to claim 1 is characterized in that: The compensation controller adopts PID control for adjusting the change of liquid level state, and the PID compensation control law based on the unknown input observer is: in, are the gain matrices of the proportional term, integral term, and differential term to be designed, and l is the length of the integral time interval of the integral term.

10. The liquid level safety control method based on unknown input observer according to claim 9 is characterized in that: The liquid level control is performed based on the compensation controller, as shown in the figure below: Under the PID compensation control law based on the unknown input observer, the following symbols are defined: Then, the augmented system is obtained as: Select the Lyapunov function as: V(ψ(k))=V1(ψ(k))+V2(ψ(k)) in, According to the Lyapunov function, the stability condition of liquid level safety control is obtained: in, is the matrix to be calculated, is a positive definite symmetric matrix, According to the stability condition of liquid level safety control, the gain matrix of the unknown input observer is calculated as: The gain matrix of the PID compensation controller based on the unknown input observer is: Among them, W 11 It is a reversible matrix calculated based on the stability conditions of liquid level safety control.

Images