Electric power system event triggering asynchronous safety control method

By modeling the power system as a non-homogeneous Markov jump system, combining event triggering mechanism and neural network technology, the asynchronous controller of event triggering neural networks is designed, and the problem of safety and reliability of power system under hybrid network attacks is solved, and the problem of power system safety and reliability under hybrid network attacks is achieved. Reduce network burden while ensuring system security and stability.

CN118889362APending Publication Date: 2024-11-01TIANJIN POLYTECHNIC UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410454444.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-04-16
Publication Date
2024-11-01

AI Technical Summary

Technical Problem

The existing power system security control technology is difficult to effectively ensure the safe and reliable operation of power systems under hybrid network attacks. Especially when using event triggering mechanisms, how to deal with the impact of asynchronous modal and hybrid network attacks is still a major challenge.

Method used

Model the power system as a non-homogeneous Markov jump system, construct the observer to estimate difficult-to-measure states, establish a hybrid network attack model, adopt event trigger mechanism and neural network technology, and design event trigger neural network asynchronous controllers to suppress hybrid networks The impact of the attack and ensure the normal operation of the power system.

Benefits of technology

Effectively reduce the burden on the network and ensure the safe and reliable operation of the non-homogeneous Markov jump power system under mixed network attacks. It removes the bounded assumption of spoofing attacks, relaxes the attack channel assumption, and ensures the system's probability of dependence. boundary.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118889362A_ABST
    Figure CN118889362A_ABST
Patent Text Reader

Abstract

The invention discloses a power system event triggering asynchronous safety control method, and relates to the technical field of power system safety control, and the method comprises the steps: building a power system model; constructing an observer to estimate a difficult-to-measure state; establishing a hybrid network attack model according to a denial of service attack occurring in a sensor-controller communication channel and a spoofing attack occurring in a controller-actuator channel; the network burden is relieved by adopting an event triggering mechanism; the neural network technology is adopted to approach unknown spoofing attack signals, and adverse effects of spoofing attacks on a power system are inhibited; according to the method, an asynchronous phenomenon between a power system and a controller is represented by using a hidden Markov method, and an event-triggered neural network asynchronous controller is designed based on an adopted event triggering mechanism and a neural network technology, so that safe and reliable operation of the power system under hybrid network attack is ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of power system safety control, and in particular to an event-triggered asynchronous safety control method for a power system. Background Art

[0002] The power system is a key infrastructure in modern society and plays an important role in economic and social development. Due to the complex dynamics of the power system and various uncertain factors such as load disturbances, environmental changes, and component failures, the parameters or structure of the power system may change suddenly. However, most existing technologies model the power system as a single-mode system, which is insufficient to characterize this random mutation behavior of the power system. Fortunately, the Markov jump system has a powerful ability to describe this random mutation behavior and can model the power system in the form of a Markov jump system. In addition, due to factors such as communication delay and packet loss, the system mode and the controller mode are usually asynchronous. Therefore, studying asynchronous control problems has important practical significance.

[0003] With the rapid development of network communication technology, the network is widely used in power system data transmission due to its advantages of low cost and high flexibility. However, due to the openness of the network, data transmission is vulnerable to network attacks. The attacked power system may have disastrous consequences for the economy and society. Therefore, the problem of power system security control needs to be solved urgently. However, most of the current technologies only consider denial of service attacks or only consider deception attacks. The problem of power system security control under hybrid network attacks still needs further research.

[0004] On the other hand, energy saving is a key issue in network transmission. Compared with the time-triggered mechanism, the event-triggered mechanism can effectively reduce the number of data transmissions and save network resources. Therefore, in order to reduce the network burden in the power system, researchers have proposed and improved a variety of event-triggered mechanisms suitable for the power system. However, how to ensure the safe and reliable operation of the power system under hybrid network attacks while adopting the event-triggered mechanism is still a major challenge. Summary of the invention

[0005] The purpose of the present invention is to provide an event-triggered asynchronous safety control method for a power system, which models the power system as a non-homogeneous Markov jump system, effectively suppresses the impact of asynchronous mode and hybrid network attacks on the non-homogeneous Markov jump power system while saving network resources, and ensures the normal operation of the non-homogeneous Markov jump power system.

[0006] In order to achieve the above object, the present invention provides a method for asynchronous safety control of a power system triggered by an event, comprising the following steps:

[0007] Step 1. Establish a power system model;

[0008] Step 2. Construct an observer to estimate the difficult-to-measure state;

[0009] Step 3. Establish a hybrid network attack model based on the denial of service attack occurring in the sensor-controller communication channel and the deception attack occurring in the controller-actuator channel;

[0010] Step 4. Use event triggering mechanism to relieve network burden;

[0011] Step 5. Use neural network technology to approximate unknown deception attack signals to suppress the adverse effects of deception attacks on the power system;

[0012] Step 6. Use the hidden Markov method to characterize the asynchronous phenomenon between the power system and the controller, and based on the adopted event triggering mechanism and neural network technology, design an event-triggered neural network asynchronous controller to ensure the safe and reliable operation of the power system under hybrid network attacks.

[0013] Furthermore, the establishment of the power system model in step 1 is specifically in the form of:

[0014]

[0015] Among them, δ, ω0, ω, T m , T e , M, T′ do , E fd , x d , x′ d , x e , E′ q , V, K E , E ref , V t , u and T E They are the angle between the quadrature axis and the infinite bus, the synchronous motor speed, the rotor speed, the mechanical torque, the electromagnetic torque, the inertia coefficient, the direct axis transient open circuit time constant, the generator excitation voltage, the direct axis synchronous reactance, the direct axis transient reactance, the external reactance, the quadrature axis voltage after the transient reactance, the infinite bus voltage, the excitation gain, the regulator reference voltage, the terminal voltage, the stabilization signal and the time constant;

[0016] For the operating point, the power system model can be linearized as:

[0017]

[0018] Among them, x T (t) = [Δδ Δω ΔE′ q ΔE fd] is the power system state, Δδ is the angle change, Δω is the speed change, ΔE′ q is the change in quadrature axis voltage after transient reactance, ΔE fd is the generator excitation voltage change, y(t) is the measured output, u(t) is the control input, t is the time, C=[0 1 0 0],

[0019]

[0020] k o (o=1, ..., 6) are linearized model constants of synchronous motor;

[0021] Various factors such as lightning strikes can cause the circuit breaker to trip immediately, resulting in structural changes in the power system. For this reason, the Markov chain is introduced to describe this random mutation phenomenon of the power system. Under the discretization of the sampling time T = 0.03s, the linearized power system model can be described in the form of a Markov jump system:

[0022]

[0023] Where x(k) is the power system state, y(k) is the measured output, and u(k) is the control input. and is the system matrix, is a non-homogeneous Markov chain, whose time-varying transition probability matrix Γ(k) = [γ ab (k)] is expressed as satisfying and 0≤γ ab γ with (k)≤1 ab (k) = Pr{θ k+1 =b|θ k =a}, l is the number of subsystems, and it is assumed that the time-varying transition probability matrix Γ(k) belongs to the following polyhedron:

[0024]

[0025] Among them, Γ h is a known constant transition probability matrix, Apparently,

[0026] Furthermore, the constructed observer in step 2 estimates the difficult-to-measure state, specifically in the form of:

[0027]

[0028] in, is the observer state, is the observer output, is the observer gain.

[0029] Furthermore, in step 3, a hybrid network attack model is established based on the denial of service attack occurring in the sensor-controller communication channel and the spoofing attack occurring in the controller-actuator channel, including:

[0030] The form of a periodic denial of service attack is described as follows:

[0031]

[0032] in, is the denial of service attack period, [T c , T′ c ) is the sleep interval of the cth denial of service attack, [T′ c , T c+1 ) is the activation interval of the cth denial of service attack;

[0033] The form of the spoofing attack is described as:

[0034]

[0035] in, For network attacks, is the latest transmitted observer state, k t is the latest trigger moment, Output of the controller.

[0036] Furthermore, the event triggering mechanism used in step 4 to alleviate the network burden includes:

[0037] In the observer-controller communication channel, the following event triggering mechanism is adopted:

[0038]

[0039] in, σ∈[0,1) is a given threshold, is the event trigger weight matrix;

[0040] When a denial of service attack occurs, the trigger data cannot be successfully transmitted to the controller, and the above event trigger mechanism is no longer applicable. A zero-order holder is used to hold the latest transmitted data until the next trigger data is successfully transmitted. The reconstruction event trigger condition is:

[0041]

[0042] in, s is the number of times the trigger data is blocked by the denial of service attack.

[0043] Furthermore, the step 5 uses a neural network technology to approximate the unknown deception attack signal to suppress the adverse effects of the deception attack on the power system, including:

[0044] Using neural network technology to approximate for:

[0045]

[0046] Among them, Θ is a compact set, and W satisfies ||W|| F ≤W * The ideal weight matrix is To satisfy The approximation error is To satisfy The basis function is for:

[0047]

[0048] Among them, v is the number of neural network nodes, c g Centered, w g is the width, and W * ,∈ * and ρ * is a known constant;

[0049] Design the neural network weight update law as:

[0050]

[0051] in, is the estimate of W, v>0 is the learning rate, and t>0 is the correction term.

[0052] Furthermore, in step 6, the asynchronous phenomenon between the power system and the controller is characterized by using the hidden Markov method, and based on the adopted event triggering mechanism and neural network technology, an event-triggered neural network asynchronous controller is designed to ensure the safe and reliable operation of the power system under hybrid network attacks. The specific form is:

[0053]

[0054] in, is the controller gain, is a hidden Markov chain, whose conditional transition probability matrix Π=[η ap ] indicates satisfaction and 0≤η ap η≤1 ap (k) = Pr{r k =p|θ k=a}, f is the number of controllers.

[0055] The beneficial effects of the present invention are as follows: the power system with random mutation behavior is modeled in the form of a non-homogeneous Markov jump system, which not only solves the problem that a single-mode power system is insufficient to describe the random mutation behavior, but also relaxes the restriction on the modal transfer probability of the homogeneous Markov jump power system; the hybrid network attack consisting of a denial of service attack occurring in a sensor-controller communication channel and a deception attack occurring in a controller-actuator channel relaxes the assumption in the prior art that the attack only occurs in a single channel; the neural network technology is used to approximate the unknown nonlinear deception attack signal, and the bounded assumption of the deception attack is removed; the designed observer-based event-triggered neural network asynchronous controller not only effectively reduces the network burden, but also ensures that the non-homogeneous Markov jump power system under the hybrid network attack is probabilistically bounded while considering the unavailable system state and modal asynchronous constraints. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a block diagram of asynchronous safety control triggered by power system events in the present invention;

[0057] Figure 2 A schematic diagram of a sequence of occurrence of a denial of service attack in the present invention;

[0058] Figure 3 This is a schematic diagram of the event interval time in the present invention;

[0059] Figure 4 It is a schematic diagram of the deception attack and neural network estimation curve in the present invention;

[0060] Figure 5 Schematic diagram of system mode and controller mode in the present invention;

[0061] Figure 6 Schematic diagram of the power system status trajectory under the hybrid network attack in the present invention. DETAILED DESCRIPTION

[0062] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings.

[0063] Example

[0064] like Figure 1 As shown, the present invention provides an asynchronous safety control method triggered by an electric power system event, and the specific implementation steps are as follows:

[0065] Step 1: Establish a power system model, the specific form is:

[0066]

[0067] Among them, δ, ω0, ω, T m , T e , M, T′ do , E fd , x d , x′ d , x e , E′ q , V, K E , E ref , V t , u and T E They are the angle between the quadrature axis and the infinite bus, the synchronous motor speed, the rotor speed, the mechanical torque, the electromagnetic torque, the inertia coefficient, the direct axis transient open circuit time constant, the generator excitation voltage, the direct axis synchronous reactance, the direct axis transient reactance, the external reactance, the quadrature axis voltage after the transient reactance, the infinite bus voltage, the excitation gain, the regulator reference voltage, the terminal voltage, the stabilization signal and the time constant;

[0068] For the operating point, the power system model can be linearized as:

[0069]

[0070] Among them, x T (t) = [Δδ Δω ΔE′ q ΔE fd ] is the power system state, Δδ is the angle change, Δω is the speed change, ΔE′ q is the change in quadrature axis voltage after transient reactance, ΔE fd is the generator excitation voltage change, y(t) is the measured output, u(t) is the control input, t is the time, C=[0 1 0 0],

[0071]

[0072] k o (o=1, ..., 6) are linearized model constants of synchronous motor;

[0073] Various factors such as lightning strikes can cause the circuit breaker to trip immediately, resulting in structural changes in the power system. For this reason, the Markov chain is introduced to describe this random mutation phenomenon of the power system. Under the discretization of the sampling time T = 0.03s, the linearized power system model can be described in the form of a Markov jump system:

[0074]

[0075] Where x(k) is the power system state, y(k) is the measured output, and u(k) is the control input. and is the system matrix, is a non-homogeneous Markov chain, whose time-varying transition probability matrix Γ(k) = [γ ab (k)] is expressed as satisfying and 0≤γ ab γ with (k)≤1 ab (k) = Pr{θ k+1 =b|θ k =a}, l is the number of subsystems, and it is assumed that the time-varying transition probability matrix Γ(k) belongs to the following polyhedron:

[0076]

[0077] Among them, Γ h is a known constant transition probability matrix, Apparently,

[0078] Let l = 2, H = 2, then, A1 and A2 respectively. They are B1 and B2 respectively. C1 and C2, Γ h are Γ 1 and Γ 2 , α h (k) are α1(k) and α2(k) respectively:

[0079]

[0080]

[0081] C1=C2=[0 1 0 0], α1(k)=|cos(k)|, α2(k)=1-α1(k),

[0082]

[0083] In addition, set the system initial value x(0) = [0.1 0.002 0.03 0.1] T .

[0084] Step 2: Construct an observer to estimate the difficult-to-measure state. The specific form is:

[0085]

[0086] in, is the observer state, is the observer output, is the observer gain.

[0087] Set the observer initial value Pick for:

[0088] L1=[2.6362 -1.0516 6.3730 -790.7486] T ,

[0089] L2=[2.4378 -1.0386 7.1206 -954.4672] T .

[0090] Step 3: Based on the denial of service attack occurring in the sensor-controller communication channel and the deception attack occurring in the controller-actuator channel, a hybrid network attack model is established, including:

[0091] The form of a periodic denial of service attack is described as follows:

[0092]

[0093] in, is the denial of service attack period, [T c , T′ c ) is the sleep interval of the cth denial of service attack, [T′ c , T c+1 ) is the activation interval of the cth denial of service attack,

[0094] make The sleep time is 5. Further, if Figure 2 As shown, when the curve value is 0, it means that there is no denial of service attack, and when the curve value is 1, it means that a denial of service attack has occurred;

[0095] The form of the spoofing attack is described as:

[0096]

[0097] in, For network attacks, is the latest transmitted observer state, k t is the latest trigger moment, is the controller output, let

[0098]

[0099] Step 4: Use event-triggered mechanisms to relieve network burden, including:

[0100] In the observer-controller communication channel, the following event triggering mechanism is adopted:

[0101]

[0102] in, σ∈[0,1) is a given threshold, is the event trigger weight matrix;

[0103] When a denial of service attack occurs, the trigger data cannot be successfully transmitted to the controller, and the above event trigger mechanism is no longer applicable. A zero-order holder is used to hold the latest transmitted data until the next trigger data is successfully transmitted. The reconstruction event trigger condition is:

[0104]

[0105] in, s is the number of times the trigger data is blocked by the denial of service attack.

[0106] Let s = 5, σ = 0.1, and further, Figure 3 As shown in the figure, after adopting the event trigger mechanism, the number of data transmissions in the time interval [0, 300] is 206. Compared with the time trigger mechanism that needs to transmit data 300 times, the network load is reduced by 31.33%, which effectively alleviates the network burden.

[0107] In step 5: neural network technology is used to approximate unknown deception attack signals to suppress the adverse effects of deception attacks on the power system, including:

[0108] Using neural network technology to approximate for:

[0109]

[0110] Among them, Θ is a compact set, and W satisfies ||W|| F ≤W * The ideal weight matrix is To satisfy The approximation error is To satisfy The basis function is selected for:

[0111]

[0112] Among them, v is the number of neural network nodes, c g Centered, w g is the width, and W * ,∈ * and ρ * is a known constant;

[0113] Design the neural network weight update law as:

[0114]

[0115] in, is the estimate of W, v>0 is the learning rate, and i>0 is the correction term.

[0116] Let v = 25, c g =15, i=0.95, v=0.49, further, as Figure 4 As shown, the neural network estimation value is close to the deception attack function value after a period of time, which means that the unknown deception attack signal is successfully approximated by the adopted neural network technology.

[0117] Step 6: Use the hidden Markov method to characterize the asynchronous phenomenon between the power system and the controller, and design an event-triggered neural network asynchronous controller based on the adopted event trigger mechanism and neural network technology to ensure the safe and reliable operation of the power system under hybrid network attacks. The specific form is:

[0118]

[0119] in, is the controller gain, is a hidden Markov chain, whose conditional transition probability matrix Π=[η ap ] indicates satisfaction and 0≤η ap η≤1 ap (k) = Pr{r k =p|θ k =a}, f is the number of controllers.

[0120] Let f = 2, Pick for:

[0121] K1=[-0.4145 32.1612 -0.4215 -0.0066],

[0122] K2=[-0.4535 29.7454 -0.4141 -0.0073],

[0123] Furthermore, if Figure 5 As shown, the system mode and the controller mode are asynchronous. In this asynchronous case, Figure 6 As shown, the state trajectory of the power system under the hybrid network attack is bounded, which indicates that the event-triggered neural network asynchronous controller effectively weakens the impact of the hybrid network attack on the power system and plays a good control effect.

[0124] It should be noted that the contents not described in detail in the specification of the present invention belong to the prior art known to professional and technical personnel in the field.

[0125] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformation or addition, subtraction or substitution made by any technician in the technical field using the contents of the present invention specification and drawings within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for asynchronous safety control triggered by power system events. It is characterized in that The following steps are involved: Step 1. Establish a power system model; Step 2. Construct an observer to estimate the difficult-to-measure state; Step 3. Establish a hybrid network attack model based on the denial of service attack occurring in the sensor-controller communication channel and the deception attack occurring in the controller-actuator channel; Step 4. Use event triggering mechanism to relieve network burden; Step 5. Use neural network technology to approximate unknown deception attack signals to suppress the adverse effects of deception attacks on the power system; Step 6. Use the hidden Markov method to characterize the asynchronous phenomenon between the power system and the controller, and based on the adopted event triggering mechanism and neural network technology, design an event-triggered neural network asynchronous controller to ensure the safe and reliable operation of the power system under hybrid network attacks.

2. The method for asynchronous safety control triggered by power system events according to claim 1, It is characterized in that The specific form of establishing the power system model in step 1 is: Among them, δ, ω 0 ,ω,T m , T e , M, T′ do , E fd , x d , x′ d , x e , E′ q , V, K E , E ref , V t , u and T E They are the angle between the quadrature axis and the infinite bus, the synchronous motor speed, the rotor speed, the mechanical torque, the electromagnetic torque, the inertia coefficient, the direct axis transient open circuit time constant, the generator excitation voltage, the direct axis synchronous reactance, the direct axis transient reactance, the external reactance, the quadrature axis voltage after the transient reactance, the infinite bus voltage, the excitation gain, the regulator reference voltage, the terminal voltage, the stabilization signal and the time constant; For the operating point, the power system model can be linearized as: Among them, x T (t) = [Δδ Δω ΔE′ q ΔE fd ] is the power system state, Δδ is the angle change, Δω is the speed change, ΔE′ q is the change in quadrature axis voltage after transient reactance, ΔE fd is the generator excitation voltage change, y(t) is the measured output, u(t) is the control input, t is the time, C=[0 1 0 0], k o (o=1, ..., 6) are linearized model constants of synchronous motor; Various factors such as lightning strikes can cause the circuit breaker to trip immediately, resulting in structural changes in the power system. For this reason, the Markov chain is introduced to describe this random mutation phenomenon of the power system. Under the discretization of the sampling time T = 0.03s, the linearized power system model can be described in the form of a Markov jump system: Where x(k) is the power system state, y(k) is the measured output, and u(k) is the control input. and is the system matrix, is a non-homogeneous Markov chain, whose time-varying transition probability matrix Γ(k) = [γ ab (k)] is expressed as satisfying and 0≤γ ab γ with (k)≤1 ab (k) = Pr{θ k+1 =b|θ k =a}, l is the number of subsystems, and it is assumed that the time-varying transition probability matrix Γ(k) belongs to the following polyhedron: Among them, Γ h is a known constant transition probability matrix, Apparently, 3. The method for asynchronous safety control triggered by power system events according to claim 1, It is characterized in that The constructed observer in step 2 estimates the difficult-to-measure state, and the specific form is: in, is the observer state, is the observer output, is the observer gain.

4. The method for asynchronous safety control triggered by an event in a power system according to claim 1, It is characterized in that In step 3, a hybrid network attack model is established based on the denial of service attack occurring in the sensor-controller communication channel and the deception attack occurring in the controller-actuator channel, including: The form of a periodic denial of service attack is described as follows: in, is the denial of service attack period, [T c , T′ c ) is the sleep interval of the cth denial of service attack, [T′ c , T c+1 ) is the activation interval of the cth denial of service attack; The form of the spoofing attack is described as: in, For network attacks, is the latest transmitted observer state, k t is the latest trigger moment, Output of the controller.

5. The method for asynchronous safety control triggered by power system events according to claim 1, It is characterized in that The event triggering mechanism in step 4 is used to alleviate the network burden, including: In the observer-controller communication channel, the following event triggering mechanism is adopted: in, σ∈[0,1) is a given threshold, is the event trigger weight matrix; When a denial of service attack occurs, the trigger data cannot be successfully transmitted to the controller, and the above event trigger mechanism is no longer applicable. A zero-order holder is used to hold the latest transmitted data until the next trigger data is successfully transmitted. The reconstruction event trigger condition is: in, s is the number of times the trigger data is blocked by the denial of service attack.

6. The method for asynchronous safety control triggered by power system events according to claim 1, It is characterized in that The step 5 uses a neural network technology to approximate an unknown deception attack signal to suppress the adverse effects of the deception attack on the power system, including: Using neural network technology to approximate for: Among them, Θ is a compact set, and W satisfies ||W|| F ≤W * The ideal weight matrix is To satisfy The approximation error is To satisfy The basis function is for: Among them, v is the number of neural network nodes, c g Centered, w g is the width, and W * ,∈ * and ρ * is a known constant; Design the neural network weight update law as: in, is the estimate of W, v>0 is the learning rate, and ι>0 is the correction term.

7. The method for asynchronous safety control triggered by power system events according to claim 1, It is characterized in that In step 6, the hidden Markov method is used to characterize the asynchronous phenomenon between the power system and the controller, and based on the event triggering mechanism and neural network technology adopted, an event-triggered neural network asynchronous controller is designed to ensure the safe and reliable operation of the power system under hybrid network attacks. The specific form is: in, is the controller gain, is a hidden Markov chain, whose conditional transition probability matrix Π=[η ap ] indicates satisfaction and 0≤η ap η≤1 ap (k) = Pr{r k =p|θ k =a}, f is the number of controllers.

Citation Information

Cited By

  • Elastic asynchronous sampling control method for resisting denial of service attack in networked system

    CN120729650A