False data injection attack reconstruction method for fan system

Abstract

The invention discloses a false data injection attack reconstruction algorithm for a fan system, and aims to solve the problem of insufficient detection under the combined action of multi-type attacks (actuator attacks and sensor attacks) and unknown disturbance in the prior art. A system state and a sensor attack variable are fused into an augmented state by constructing an augmented descriptor system, and a proportional differential estimation algorithm based on a sliding-mode observer is designed to realize robust reconstruction of an attack signal. The algorithm has the characteristics of high sensitivity to high-frequency attack signals, strong anti-interference capability and good real-time performance, and is suitable for network security protection of the wind generating set.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial control system network security, and in particular relates to a detection and reconstruction method for false data injection attacks in wind turbines, which can simultaneously handle actuator attacks, sensor attacks and unknown disturbances. Background Art

[0002] With the rapid advancement of technology, today's wind turbines are becoming increasingly intelligent. Many physical objects in the system operate in coordination, with decisions made by a central control center, and data is constantly being transmitted and exchanged. This results in a larger and more complex system, but also more vulnerable to external influences and malicious attacks. A cyberattack on a wind turbine can degrade the performance of the entire system, or even lead to system failure or paralysis. Therefore, timely and accurate detection of malicious cyberattacks injected into wind turbine systems is crucial to maintaining normal operation.

[0003] Existing technologies mostly focus on linear systems, considering only actuator or sensor attacks and often ignoring the impact of unknown disturbances. Estimation is suboptimal for high-frequency signals, and most rely on augmented methods to estimate state and attack signals, lacking robustness to nonlinear system dynamics and disturbances. Summary of the Invention

[0004] The present invention aims to overcome the deficiencies of the prior art and provide a wind turbine system false data injection attack reconstruction method that is compatible with multiple attack types, has enhanced robustness, and is sensitive to high-frequency attack signals.

[0005] To solve the above problems, the present invention is achieved as follows:

[0006] A method for reconstructing a wind turbine system from a false data injection attack comprises the following steps:

[0007] (1) Establish a dynamic model of the wind turbine system, which includes the actuator attack signal f a (t), sensor attack signal f s (t) and unknown disturbance d(t), whose expressions are:

[0008]

[0009] Where x(t)∈R n is the state vector, u(t)∈R m is the control input vector, Φ(x(t)) represents the known nonlinear dynamics acting on the system, and y(t)∈R p is the output vector that can be measured, d(t)∈R d is an unknown but bounded external disturbance vector, fa (t)∈R q and f s (t)∈R r Respectively represent the actuator attack signal and sensor attack signal injected into the system; A, B, G, C, B a ,B d and D s are known matrices with appropriate dimensions;

[0010] (2) Building an augmented descriptor system: Definition The equivalent augmented generalized system is obtained:

[0011]

[0012] in:

[0013]

[0014] (3) Design of augmented sliding mode observer:

[0015]

[0016] in is an estimate of the system state ξ(t), is the augmented state x a (t)∈R n+r Estimate of ν, the sliding mode term to be designed, L∈R (n+r)×p is the observer gain to be solved;

[0017] (4) The observer parameters are solved through Lyapunov stability analysis and linear matrix inequality to ensure the robust stability of the error dynamic system.

[0018] Furthermore, the step (2) defines The following augmented system can be obtained:

[0019]

[0020] in: C a =[CD s ],C a1 =[C 0].

[0021] Furthermore, the step (2) defines The following equivalent augmented system can be obtained:

[0022]

[0023] Among them: A e =S +(A a -KC a1 ),B e =S + B ua ,G e =S + G a ,

[0024] B fe =S + B fa ,B de =S + B da ,K e =S + K;

[0025] Furthermore, the matrix S of the augmented generalized system in step (2) is E+KC a ; S and its pseudo-inverse S + satisfy:

[0026]

[0027] Furthermore, in step (3), the sliding mode gain ρ≥ρ0+α, where ρ0 is the upper bound of the disturbance and α is the upper bound of the attack signal amplitude.

[0028] Furthermore, in step (4), the linear matrix inequality condition is solved:

[0029]

[0030] Determine the observer gain L = P -1 Y, where P, μ, r, and Y are the parameters to be optimized.

[0031] A wind turbine system attack detection device includes a processor and a memory, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the program.

[0032] This method establishes a descriptor dynamic system equivalent to the original regular dynamic system by forming an extended state vector composed of the system state and sensor attacks. This system can achieve good estimation of high-frequency attack signals and ensure the dynamic robustness of the error estimation equation using linear matrix inequality techniques. The algorithm is designed offline and implemented online, resulting in excellent real-time performance.

[0033] The present invention uses an augmented descriptor system to construct a system state x(t) and a sensor attack variable f s (t) Fusion into augmented state x a(t), expanding the state space for joint estimation; establishing a descriptor dynamic system equivalent to the original regular dynamic system by forming an extended state vector consisting of the system state and sensor attack; introducing a sliding mode term through the design of a proportional-differential sliding mode observer to suppress attack signals and disturbances, improving robustness to high-frequency noise; and optimizing observer parameters through matrix inequality constraints to ensure the global stability of the error dynamic system. The present invention constructs an augmented descriptor system, fuses the system state and sensor attack variables into an augmented state, and designs a proportional-differential sliding mode observer to achieve robust reconstruction of the attack signal. The algorithm of the present invention adopts offline design and online implementation, and has good real-time performance.

[0034] Compared with the prior art, the present invention has the following characteristics:

[0035] (1) Compatible with multiple attack types: can simultaneously detect actuator attacks, sensor attacks, and unknown disturbances;

[0036] (2) High-frequency response capability: The sliding mode design reduces the algorithm's detection delay for sudden attack signals to less than 2ms;

[0037] (3) Enhanced robustness: The present invention has been verified to be highly efficient in simulation and wind turbine scale-down tests, with an attack signal reconstruction error of less than 5% and a real-time response delay of less than 2ms, making it suitable for network security protection of wind turbines. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention will be described in detail below through specific examples. These examples are provided in order to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art. As mentioned throughout the specification and claims, "including" or "comprising" is an open-ended term and is interpreted as "including but not limited to". The subsequent description of the specification is a preferred embodiment of the present invention, but the description is based on the general principles of the specification and is not intended to limit the scope of the present invention. The scope of protection of the present invention shall be determined by the appended claims.

[0039] Figure 1 This is a schematic diagram of the false data injection attack reconstruction algorithm for the fan system of the present invention. DETAILED DESCRIPTION

[0040] See also Figure 1 As shown, for the wind turbine system that is simultaneously affected by actuator attacks, sensor attacks, and unknown input interference in the following forms:

[0041]

[0042] Where x(t)∈R n is the state vector, u(t)∈R mis the control input vector, Φ(x(t)) represents the known nonlinear dynamics acting on the system, and y(t)∈R p is the output vector that can be measured, d(t)∈R d is an unknown but bounded external disturbance vector, f a (t)∈R q and f s (t)∈R r Represent the actuator attack and sensor attack injected into the system respectively. a ,B d and D s are known matrices with appropriate dimensions.

[0043] Construction of augmented descriptor system:

[0044] The original state of the system and the sensor attack variable are augmented into a new state variable, thereby establishing an equivalent generalized system, which provides conditions for estimating the system state and the time-varying sensor attack variable. Specifically, define Then we can get the following augmented system:

[0045]

[0046] in:

[0047]

[0048] C a =[CD s ],C a1 =[C 0].

[0049] According to the second equation of equation (1), the augmented generalized system can be simplified as:

[0050]

[0051] Let S = E + KC a ,but:

[0052]

[0053] Add both sides of equation (3) You can get:

[0054]

[0055] Depend on The left pseudo-inverse of S can be obtained as:

[0056]

[0057] Multiply S on both sides of equation (5) + , we can get:

[0058]

[0059] remember:

[0060] A e =S + (A a -KC a1 ),B e =S + B ua ,G e =S + G a ,

[0061] B fe =S + B fa ,B de =S + B da ,K e =S + K------(7)

[0062] Then, equation (6) can be simplified to:

[0063]

[0064] remember:

[0065] ξ(t)=x a (t)-K e y(t)------(9)

[0066] Then equation (8) can be rewritten as:

[0067]

[0068] x a (t)=ξ(t)+K e Substituting y(t) into equation (10) yields:

[0069]

[0070] In summary, by using generalized system theory and mathematical transformation, an equivalent augmented system in the form of equation (11) can be obtained.

[0071] Augmented Sliding Mode Observer:

[0072] For the augmented system (11), a proportional differential sliding mode observer estimation algorithm of the following form is proposed, which can realize the estimation of the system state and the reconstruction of the attack signal.

[0073]

[0074] in is an estimate of the system state ξ(t), is the augmented state x a (t)∈R n+r Estimate of, v is the sliding mode term to be designed, L∈R (n+r)×p is the observer gain to be solved. From equation (15), we can get:

[0075]

[0076] remember:

[0077]

[0078] where ρ is the sliding mode gain to be designed, F∈R q×p is the observer gain to be solved, e y is the output estimation error, and

[0079] Subtracting equation (13) from equation (8) yields:

[0080]

[0081] Stability analysis:

[0082] Lemma 1: For any positive scalar μ and real constant matrices x,y∈R n , the following inequality holds:

[0083]

[0084] Lemma 2: For a given symmetric matrix If and only if S 22 <0 and When S<0.

[0085] Theorem 1: For system (2), if there exists a positive definite matrix P, positive scalars μ and r, matrix Y and a given constant γ, such that the following equalities and inequalities hold:

[0086]

[0087] Then, there exists an augmented sliding mode observer in the form of equation (12) that makes the estimation error of equation (17) robust and stable, and satisfies the robust performance index The observer gain can be calculated by L = P -1 Y calculation,

[0088] Proof: The Lyapunov function of the error dynamics system (17) is defined as:

[0089] V(e a )=e a T Pe a (twenty one)

[0090] Taking the derivative of the above formula and substituting equation (17) into it, we can get:

[0091]

[0092] From equations (16) and (18), and ‖f a ‖≤α,ρ≥ρ0+α, we can get:

[0093]

[0094] From Lemma 1, we can get:

[0095]

[0096] Substituting equations (23) and (24) into equation (22) yields:

[0097] In order to suppress the influence of disturbance, performance indicators are introduced:

[0098]

[0099] From (25) and (26), we can get:

[0100]

[0101] in:

[0102]

[0103] From Lemma 2, Ω can be written as:

[0104]

[0105] At zero initial condition e a (0)=0, we have:

[0106]

[0107] Since Ω<0, It can be obtained that Γ<0, which means Proof completed.

[0108] Example 1: Simulation Verification

[0109] (1) Simulation environment: Simulink, wind turbine model parameters;

[0110] (2) Attack injection: sensor attacks are steps, ramps, periodic signals, etc., and actuator attacks are steps, ramps, periodic signals, etc.;

[0111] (3) Results: A good estimate of the attack signal can be achieved .

[0112] Explanation of relevant terms:

[0113] (1) False data injection attack

[0114] Misleading system operation by sending false data packets directly to the target node or injecting false data into the original data packet is called a false data injection attack. When the attack occurs on the sensor-to-controller channel, it is called a sensor attack; when the attack occurs on the controller-to-actuator channel, it is called an actuator attack.

[0115] (2) Sliding mode observer

[0116] Sliding mode variable structure control is a nonlinear system analysis and design method that can force nonlinear systems to perform small, high-frequency movements along a pre-defined state trajectory. It boasts strong robustness to system modeling uncertainty and external disturbances, and is easily implemented in engineering. The sliding mode observer, an application of sliding mode variable structure control, is an observation technique used in control engineering to estimate unknown system variables. Its basic concept is to use the system's input and output signals to estimate states or parameters that cannot be directly measured. By observing the system's input and output, the sliding mode observer uses sliding modes to infer unknown states or parameters within the system. This observation method is robust and insensitive to noise, making it widely used in many engineering applications.

[0117] The above description is only a specific embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent transformation made using the present invention, or directly or indirectly applied in other related technical fields, is also included in the patent protection scope of the present invention.

Claims

1. A wind turbine system false data injection attack reconstruction method, characterized in that: The following steps are involved: (1) Establish a dynamic model of the wind turbine system, which includes the actuator attack signal f a (t), sensor attack signal f s (t) and unknown disturbance d(t), whose expressions are: Where x(t)∈R n is the state vector, u(t)∈R m is the control input vector, Φ(x(t)) represents the known nonlinear dynamics acting on the system, and y(t)∈R p is the output vector that can be measured, d(t)∈R d is an unknown but bounded external disturbance vector, f a (t)∈R q and f s (t)∈R r Respectively represent the actuator attack signal and sensor attack signal injected into the system; A, B, G, C, B a ,B d and D s are known matrices with appropriate dimensions; (2) Building an augmented descriptor system: Definition The equivalent augmented generalized system is obtained: in: (3) Design of augmented sliding mode observer: in is an estimate of the system state ξ(t), is the augmented state x a (t)∈R n+r Estimate of ν, the sliding mode term to be designed, L∈R (n+r)×p is the observer gain to be solved; (4) The observer parameters are solved through Lyapunov stability analysis and linear matrix inequality to ensure the robust stability of the error dynamic system.

2. The wind turbine system false data injection attack reconstruction method according to claim 1 is characterized by: Defined in step (2) The following augmented system is obtained: in: C a =[CD s ],C a1 =[C 0].

3. The wind turbine system false data injection attack reconstruction method according to claim 2 is characterized by: Defined in step (2) The following equivalent augmented system is obtained: Among them: A e =S + (A a -KC a1 ),B e =S + B ua ,G e =S + G a , B fe =S + B fa ,B de =S + B da ,K e =S + K。 4. The wind turbine system false data injection attack reconstruction method according to claim 3 is characterized by: The matrix of the augmented generalized system in step (2) is S = E + KC a ; S and its pseudo-inverse S + satisfy:

5. The wind turbine system false data injection attack reconstruction method according to claim 4 is characterized by: In step (3), the sliding mode gain ρ ≥ ρ0 + α, where ρ0 is the upper bound of the disturbance and α is the upper bound of the attack signal amplitude.

6. The wind turbine system false data injection attack reconstruction method according to claim 5 is characterized in that: In step (4), the linear matrix inequality condition is solved: Determine the observer gain L = P -1 Y, where P, μ, r, and Y are the parameters to be optimized.

7. A wind turbine system attack detection device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and when the processor executes the program, the method implements the steps of any one of claims 1 to 6.

Images