Distributed security fusion estimation method under random event attack

Abstract

The invention discloses a distributed security fusion estimation method under random event attack, and relates to the technical field of safe and reliable operation of a power system. The method is characterized by comprising the following steps: step a, initializing related parameters by a system; b, acquiring and compressing original measurement data of the bottom layer physical equipment; c, calculating an event trigger value; d, judging whether data at the current moment need to be transmitted or not; step e, executing an estimation process by a local estimator; step f, judging whether the diagonal elements of the # imgabs0 # and the matrix Piy (k) need to be transmitted or not; g, the fusion center carries out state compensation; and h, judging whether the information transmission at the current moment is finished or not, ending the flow after the transmission is finished, otherwise, returning to the step b. Through the distributed security fusion estimation method under the random event attack, the problem of fusion estimation misalignment caused by local estimation damage indirectly when a communication network is attacked by the random event is solved, the fusion estimation error can be ensured to be within an allowable range, and the security and adjustability of a power system are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of safe and reliable operation of power systems, and in particular to a distributed security fusion estimation method under random event attacks. Background Art

[0002] Traditional power systems are gradually moving towards smart grids by continuously integrating advanced information technology, control methods and communication means, with the goal of achieving global optimization of energy efficiency, economic efficiency and environmental benefits. Obviously, the cross-integration of many fields has brought major technical challenges to the wide-area monitoring, planning and control of smart grid networks. To meet these potential challenges, management and control departments need to monitor the operating characteristics of distributed power systems in real time, which mainly relies on basic and key state estimation processes. Specifically, it mainly includes the information transmission process based on wireless networks and the fusion processing of information measured by phase measurement units widely distributed in space. However, the perception measurement or data may be corrupted by attackers during the process of self-flow along the network, thereby affecting the data monitoring and acquisition system, and even misleading the energy management system. Therefore, considering the secure distributed fusion estimation under network attacks is crucial to ensure the safe, reliable and economical and efficient operation of the power system.

[0003] At present, there are some works that study the security fusion estimation method under network attacks:

[0004] Literature Zheng X, Zhang H, Wang Z, et al. Stochastic event-based distributed fusion estimation over sensor networks with fading channel [J]. IEEE Transactions on Circuits and Systems I: Regular Papers, 2022, 69 (4): 1741-1750. A two-step distributed fusion estimation scheme in a fading channel (equivalent to a denial of service attack) environment is proposed. The first step is to obtain a local estimate based on the measurements of itself and its neighbors, and the second step is to fuse the local estimates of each sensor. The data processing method in this scheme adopts a general matrix weighted fusion method. However, in actual power systems, this scheme may fail because some variables do not have physical meaning.

[0005] The Chinese invention patent with application number 202410371890.0 and application date March 29, 2024, and the patent name "A method for security fusion estimation of information-physical systems based on multi-rate sampling mechanism" records a technical solution, in which the method constructs an information security framework including three situations of denial of service attack, false data injection attack and no attack through two Bernoulli variables, ensuring the flexibility of the current security state estimation method. The defect of this solution is that it focuses on common denial of service attacks and false data injection attacks. However, with the application of advanced communication technologies and data transmission solutions, the power system will be invaded by new types of network attacks, and there is a lack of analysis of corresponding security strategies.

[0006] At the same time, in the existing technology, in order to ensure the accuracy of power system estimation, a large amount of information is transmitted at each moment, which brings a high burden to the wireless network. Although some data dimension reduction and data quantization methods are currently proposed to reduce the amount of data transmission at each moment as much as possible, these methods are all lossy data processing solutions.

[0007] In summary, most existing security fusion estimation schemes are designed for common types of network attacks, and are difficult to deal with new types of network attacks caused by the vulnerability of event-driven data transmission schemes. Therefore, designing a distributed security fusion estimation method under random event attacks to ensure that the energy management center can accurately grasp the operating status of each subsystem, so as to facilitate the subsequent stable operation and real-time regulation of the power system, has become an urgent problem to be solved in this field. Summary of the invention

[0008] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a solution to the problem of inaccurate fusion estimation caused by local estimation damage when the communication network is attacked by random events, which can ensure that the fusion estimation error is within the allowable range. The designed distributed security fusion estimation method under random event attacks can ensure the safety and adjustability of the power system.

[0009] The technical solution adopted by the present invention to solve the technical problem is: the distributed security fusion estimation method under the random event attack is characterized by comprising the following steps:

[0010] Step a, system initialization related parameters;

[0011] Step b: All N sensors obtain the original measurement data z of the underlying physical device i (k), and compress the original measurement data of the sensor into y i (k);

[0012] Step c, calculating the event trigger value;

[0013] Step d, determining whether the current moment data needs to be transmitted, if the current data needs to be transmitted, the data is transmitted and step e is executed; if the current data does not need to be transmitted, the sensor remains silent and step e is executed;

[0014] Step e, the local estimator is based on the currently available information set Perform the estimation process;

[0015] Step f, determine whether transmission is required and P i y If the diagonal elements of (k) need to be transmitted, they are transmitted and step g is executed; if the current data does not need to be transmitted, step g is directly executed;

[0016] Step g: The fusion center performs state compensation and performs diagonal matrix weighted fusion to obtain the fusion estimate and the fusion estimation error covariance P i y (k);

[0017] Step h, determine whether the information transmission is completed at the current moment. If the transmission is completed, then it ends. If the transmission is not completed, return to step b.

[0018] Preferably, in step b, all N sensors obtain the original measurement data z of the underlying physical device. i (k), the corresponding measurement equation is:

[0019] z i (k) = H i x(k)+υ i (k),i=1,2,…,N

[0020] in, Measurement Matrix x(k) represents the state variable at time k, υ i (k) represents the measurement noise at time k.

[0021] Preferably, in step b, the sensor sends the original measurement data Compress to Specifically:

[0022]

[0023] in, R i represents the measurement noise covariance of the ith sensor, H i represents the measurement matrix of the i-th sensor, there exists a full-rank decomposition transformation such that H i =M i O i , Mi and O i are full row rank and full column rank matrices respectively, M i The transposed matrix of i )=r i ≤m i ;

[0024] After equivalent transformation, we get:

[0025]

[0026] in, get The covariance satisfies Represents r i dimensional identity matrix.

[0027] Preferably, in step c, the sensor calculates the event trigger function value according to the following rules:

[0028]

[0029] Among them, η i (k) represents the trigger function value when there is no attack, represents the trigger function value under attack, β i represents the event trigger parameter, and the information fed back by the estimator satisfies in represents the local prior estimate of the state variable x(k), and represents the attack signal, generated by the random event attacker, y i (k) represents the compressed original measurement data.

[0030] Preferably, in step d, the event-triggered scheduler built into the sensor calculates the data y at the current moment. i (k) corresponds to the decision variable γ i (k)(or ), the calculation rule is:

[0031]

[0032] Among them, the random variable ε i (k) is a uniform distribution in the interval [0,1] and is independent of other variables, η i (k) represents the trigger function value when there is no attack, Indicates the trigger function value under attack;

[0033] When γ i (k) = 1 or Indicates that the current data y needs to be transmittedi (k).

[0034] Preferably, in step e, the local estimator is based on the currently available information set Perform the estimation process, specifically:

[0035]

[0036] P i y- (k) = AP i y (k-1)A T +Q

[0037]

[0038] P i y (k) = [IL i (k)C i ]P i y- (k)

[0039] Where A represents the system matrix of the linearized power system, B represents the input matrix of the linearized power system, represents the local estimate of sensor i at time k, P i y (k) represents the estimated error covariance of sensor i at time k, P i y- (k) represents the local prior estimate The error covariance of Q represents the covariance matrix of the system process noise, I represents the identity matrix with the same dimension as matrix A; u(k-1) represents the input variable at time k-1, P i y (k-1) represents the estimated error covariance of sensor i at time k-1, A T represents the transposed matrix of the system matrix A of the linearized power system, γ i (k) represents data y i (k) Corresponding decision variables; intermediate variables where ρ(k)=[β i +1-γ i (k)] / β i ,in, is the measurement matrix C i The transposed matrix of i Indicates event trigger parameters; available information collection And the local estimation process under random attack only needs to change the variable γ i (k) is replaced by That's it.

[0040] Preferably, in step f, the index variable λ is updated according to a preset rule. i (k), the specific preset rules are as follows:

[0041]

[0042] Among them, γ i (k) is the data y at time k i (k) corresponds to the decision variable, γ i (k-1) is the data y at the previous moment i (k-1) corresponding decision variables, λ i (0)=0.

[0043] When i When (k) = 0, the local estimate is transmitted and the matrix P i y (k) The diagonal elements of

[0044] Preferably, in step g, the state compensation scheme adopted by the fusion center is:

[0045] Define the vector Λ(k)=[λ1(k),...,λ N (k)] and the variable τ(k) = N - ‖Λ(k)‖0,

[0046] Among them, ‖·‖0 represents the zero norm of the vector;

[0047] Then, assuming that the measurement accuracy of sensor 1 is better than that of any other N-1 sensors, and resetting that λ1(k)=0 holds for any k>0, the information set to be fused is:

[0048]

[0049] The scheme of diagonal matrix weighted fusion is:

[0050]

[0051] Among them, D j =diag[d j1 ,d j2 ,...,d j(5n) ], n represents the number of synchronous motors in the current power system, represents [1,1,...,1], Denotes the cross covariance matrix P lj (k) the i-th diagonal element;

[0052] Calculate the fusion estimation error covariance P f (k) specifically:

[0053]

[0054] get

[0055] Among them, D i , D j is the diagonal weight matrix, P ij is the cross covariance matrix.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] Through this distributed secure fusion estimation method under random event attacks, the problem of inaccurate fusion estimation caused by local estimation damage when the communication network is attacked by random events is solved. It can ensure that the fusion estimation error is within the allowable range and ensure the safety and adjustability of the power system.

[0058] This distributed security fusion estimation method under random event attacks proposes a distributed security fusion estimation scheme, which can effectively resist the intrusion of wireless networks under random event attacks, ensure the accuracy of fusion estimation, and significantly improve the robustness of wireless networks when encountering malicious intrusions.

[0059] The state compensation method proposed in this distributed security fusion estimation method under random event attacks can improve the accuracy of fusion estimation.

[0060] This distributed security fusion estimation method under random event attacks uses full-rank decomposition transformation technology, combined with the precise setting of the estimator index variable, to greatly reduce the amount of data transmission in the network. By optimizing the communication requirements on the sensor and estimator sides, the load on the wireless network is reduced and the overall efficiency of the network is improved. Combined with the full-rank decomposition transformation technology, the system can save a lot of communication resources while maintaining high estimation accuracy, and is suitable for scenarios such as large-scale Internet of Things and smart grids.

[0061] This distributed secure fusion estimation method under random event attacks achieves a good trade-off between estimation performance and network throughput. By assigning appropriate weights to different data sources, this scheme reduces the negative impact of low-quality data on the overall estimation results, while optimizing the bandwidth consumption of data fusion and improving the throughput of the system when facing large-scale data streams. The diagonal matrix structure has high computational efficiency and can significantly reduce the computational and communication complexity while ensuring the estimation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1This is a flow chart of the distributed security fusion estimation method under random event attacks.

[0063] Figure 2 Schematic diagram of distributed security fusion estimation of power system under random event attack.

[0064] Figure 3 The figure is a comparison curve of the error covariance of power system fusion estimation under different scenarios. DETAILED DESCRIPTION

[0065] Figures 1 to 3 The best embodiment of the present invention is shown below in conjunction with the attached Figures 1 to 3 The present invention is further described.

[0066] like Figure 1 As shown, a distributed security fusion estimation method under random event attack includes the following steps:

[0067] Step 1001, start;

[0068] We started to implement the distributed security fusion estimation method under random event attacks, and further combined Figure 2 The schematic diagram of distributed security fusion estimation of power system under random event attack shown in the figure includes the underlying physical device layer, the middle information and network layer and the upper fusion center. In the information and network layer, SETM represents the random event triggered transmission mechanism adopted by the sensor, and FRDT represents the full rank decomposition transformation of the original measurement data.

[0069] The physical device layer consists of 5 synchronous motors and some loads connected via 8 buses. N sensors measure the relevant physical parameters and perform full-rank decomposition to compress the measured values. Then, the random event trigger mechanism is used to send the compressed information to the local estimator through the wireless network. However, the data transmission process may be corrupted by random event attacks (i.e., tampering with the feedback of the prior measurement information). ), causing the local estimate and the fusion estimate to deviate seriously from the true value.

[0070] Step 1002, the system initializes relevant parameters;

[0071] Initialization related parameters include: local estimation corresponding to all i sensors and its error covariance P i y (0), fusion estimation and its error covariance P f (0), specifically:

[0072] By selecting the rotor angle Δδ of each synchronous motor i , rotor angular velocity Δω i, q-phase transient voltage ΔE' qi , and the internal state z of the terminal voltage controller 2i ,z 1i As the system state variable, the input of the excitation system is the control input signal Δv of the system i , linearization near the system equilibrium point can establish the state space model of the continuous system. Furthermore, the discrete state space model can be obtained by using the reverse Euler method:

[0073] x(k+1)=Ax(k)+Bu(k)+ω(k)

[0074] Where x(k) represents the state variable at time k, u(k) represents the input variable at time k, ω(k) represents the process noise signal of the system at time k (obeying a Gaussian distribution with a mean of 0 and a variance of Q ≥ 0), A represents the system matrix of the linearized power system, B represents the input matrix of the linearized power system, and n represents the number of synchronous motors in the system. represents a 5n-dimensional Euclidean space.

[0075] Step 1003, the sensor acquires raw measurement data;

[0076] All N sensors obtain the raw measurement data z of the underlying physical devices i (k), the corresponding measurement equation is:

[0077] z i (k) = H i x(k)+υ i (k),i=1,2,…,N

[0078] in, in Indicates m i -dimensional Euclidean space. Measurement matrix Measurement noise i (Subject to 0 mean, variance R i >0). Assume (H i , A) is observable, the initial state variable x(0), the system process noise signal ω(k) at time k and the measurement noise υ at time k i (k) are independent of each other.

[0079] Step 1004, compressing the original measurement data acquired by the sensor;

[0080] To ease the communication burden, the sensor sends the raw measurement data Compress to Specifically:

[0081]

[0082] in, R i represents the measurement noise covariance of the ith sensor, H i represents the measurement matrix of the i-th sensor, there exists a full-rank decomposition transformation such that H i =M i O i , M i and O i are full row rank and full column rank matrices respectively, M i The transposed matrix of i )=r i ≤m i .

[0083] For the convenience of representation, the following equivalent transformation is given:

[0084]

[0085] Among them, the measurement matrix And it is not difficult to get The covariance satisfies Represents r i dimensional identity matrix.

[0086] Step 1005, calculating the event trigger value;

[0087] The sensor calculates the event trigger function value according to the following rules:

[0088]

[0089] Among them, η i (k) represents the trigger function value when there is no attack, represents the trigger function value under attack, β i represents the event trigger parameter, and the information fed back by the estimator satisfies in represents the local prior estimate of the state variable x(k), and represents the attack signal, generated by the random event attacker, y i (k) represents the compressed original measurement data.

[0090] Step 1006, determining whether the current time data needs to be transmitted;

[0091] The event-triggered scheduler based on the sensor calculates the current data y i (k) corresponds to the decision variable γ i (k)(or ), the calculation rule is:

[0092]

[0093] Among them, the random variable ε i (k) is a uniform distribution in the interval [0,1] and is independent of other variables, η i (k) represents the trigger function value when there is no attack, Represents the trigger function value under attack.

[0094] When γ i (k) = 1 or Indicates that the current data y needs to be transmitted i (k), and execute step 1007. Otherwise, execute step 1008.

[0095] Step 1007, keep silent, i.e. no information transmission;

[0096] The sensor remains silent, ie, does not transmit information, and step 1008 is executed.

[0097] Step 1008, the estimator performs an estimation process;

[0098] The local estimator is based on the currently available information set Perform the estimation process, specifically:

[0099]

[0100] P i y- (k) = AP i y (k-1)A T +Q

[0101]

[0102] P i y (k) = [IL i (k)C i ]P i y- (k)

[0103] Where A represents the system matrix of the linearized power system, B represents the input matrix of the linearized power system, represents the local estimate of sensor i at time k, P i y (k) represents the estimated error covariance of sensor i at time k, P i y- (k) represents the local prior estimate The error covariance of Q represents the covariance matrix of the system process noise, I represents the identity matrix with the same dimension as matrix A; u(k-1) represents the input variable at time k-1, P i y (k-1) represents the estimated error covariance of sensor i at time k-1, A T represents the transposed matrix of the system matrix A of the linearized power system, γ i (k) represents data y i (k) Corresponding decision variables; intermediate variables where ρ(k)=[β i +1-γ i (k)] / β i ,in, is the measurement matrix C i The transposed matrix of i Indicates event trigger parameters; available information collection And the local estimation process under random attack only needs to change the variable γ i (k) is replaced by That's it.

[0104] Step 1009, determine whether transmission is required and P i y (k) diagonal elements;

[0105] Update the index variable λ according to the preset rules i (k), the specific preset rules are as follows:

[0106]

[0107] Among them, γ i (k) is the data y at the current time (time k) i (k) corresponds to the decision variable, γ i (k-1) is the data y at the previous moment i (k-1) corresponding decision variables, λ i (0)=0.

[0108] When i When (k) = 0, the local estimate is transmitted and the matrix P i y (k) and execute step 1011, otherwise execute step 1010.

[0109] Step 1010, keep silent, i.e. no information transmission;

[0110] The local estimator remains silent, ie, does not transmit information, and step 1011 is executed.

[0111] Step 1011: The fusion center performs state compensation and performs diagonal matrix weighted fusion to obtain and P i y (0);

[0112] The fusion center uses the state compensation and diagonal matrix weighted fusion scheme to fuse the information set and obtain the fusion estimate And calculate the fusion estimation error covariance P f (k) Specifically:

[0113] The state compensation scheme adopted by the Fusion Center is as follows:

[0114] Define the vector Λ(k)=[λ1(k),...,λ N (k)] and the variable τ(k) = N - ‖Λ(k)‖0,

[0115] Here, ‖·‖0 represents the zero norm of the vector.

[0116] Then, it is assumed that the measurement accuracy of sensor 1 is better than that of any other N-1 sensors, and it is reset that λ1(k)=0 holds for any k>0.

[0117] Then the information set to be fused is:

[0118]

[0119] The scheme of diagonal matrix weighted fusion is:

[0120]

[0121] Among them, the diagonal weight matrix D j =diag[d j1 ,d j2 ,...,d j(5n) ],

[0122] n represents the number of synchronous motors in the current power system, represents [1,1,...,1], Denotes the cross covariance matrix P lj (k) is the i-th diagonal element, and

[0123]

[0124] Where A represents the system matrix of the linearized power system, A T represents the transposed matrix of the system matrix A of the linearized power system, Q represents the covariance matrix of the system process noise, γi (k) represents data y i (k) corresponds to the decision variable, γ j (k) represents data y j (k) corresponds to the decision variable, Denotes the compressed measurement matrix C j Transpose the matrix, Represents the intermediate variable L j The transposed matrix of (k), P ij (k-1) represents the cross-covariance matrix at time k-1.

[0125] Calculate the fusion estimation error covariance P f (k) specifically:

[0126]

[0127] get

[0128] Among them, D i , D j is the diagonal weight matrix, P ij is the cross covariance matrix.

[0129] Step 1012, determining whether the transmission is completed;

[0130] Determine whether the information transmission at the current moment is completed. If completed, execute step 1013; if not completed, return to step 1002.

[0131] Step 1013, end.

[0132] The following example verifies the effect of the distributed security fusion estimation method under random event attacks:

[0133] Refer to Figure 2 The power system model with 5 synchronous machines and 8 buses is shown in the figure. The sampling time Δt=10 -3 , the covariance of process noise is Q = 10 -3 I 25 Assume that the discretized state space model output is measured by four sensors with different precisions, where The corresponding measurement noises are: 10 - 5 I7, 4*10 -4 I7, 6*10 -2 I5, 2*10 -2 I6.

[0134] After full rank decomposition transformation, we get Where i = 1, 2, 3, 4. In addition, the event trigger parameter is set to β1 = 10 -7 ,β2=10 -3 ,β3=10 -1 ,β4=10 -1 .

[0135] After launching a random event attack on some feedback confirmation information of the corresponding estimators of the four sensors, the changes in the fusion estimation are analyzed. Figure 3 The solid line, long dashed line, short dashed line, and long and short dashed lines respectively show the change trajectories of the covariance matrix of the fusion estimation error in four different situations, namely, the method proposed in the present invention, with attack and without state compensation, with attack, and without attack, after 100 Monte Carlo experiments. Figure 3 The unit of sampling time of the horizontal axis is seconds, and the interval between two adjacent samplings is 0.02 seconds. Figure 3 It can be observed that the method proposed in the present invention has the smallest performance index, which verifies the effectiveness of the distributed security fusion estimation method proposed in the present invention.

[0136] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A distributed security fusion estimation method under random event attack, characterized by: The steps include: Step a, system initialization related parameters; Step b: All N sensors obtain the original measurement data z of the underlying physical device i (k), and compress the original sensor measurement data into y i (k); Step c, calculating the event trigger value; Step d, determining whether the current moment data needs to be transmitted, if the current data needs to be transmitted, the data is transmitted and step e is executed; if the current data does not need to be transmitted, the sensor remains silent and step e is executed; Step e, the local estimator is based on the currently available information set Perform the estimation process; Step f, determine whether transmission is required and If the diagonal elements of the data need to be transmitted, they are transmitted and step g is executed. If the current data does not need to be transmitted, step g is directly executed. Step g: The fusion center performs state compensation and performs diagonal matrix weighted fusion to obtain the fusion estimate and the fusion estimation error covariance P i y (k); Step h, determine whether the information transmission is completed at the current moment. If the transmission is completed, then it ends. If the transmission is not completed, return to step b.

2. The distributed security fusion estimation method under random event attack according to claim 1 is characterized in that: In step b, all N sensors acquire the raw measurement data z of the underlying physical device i (k), the corresponding measurement equation is: z i (k)=H i x(k)+υ i (k),i=1,2,…,N in, Measurement Matrix x(k) represents the state variable at time k, υ i (k) represents the measurement noise at time k.

3. The distributed security fusion estimation method under random event attack according to claim 2 is characterized in that: In step b, the sensor sends the raw measurement data Compress to Specifically: in, R i represents the measurement noise covariance of the ith sensor, H i represents the measurement matrix of the i-th sensor, there exists a full-rank decomposition transformation such that H i =M i O i , M i and O i are full row rank and full column rank matrices respectively, M i The transposed matrix of i )=r i ≤m i ; After equivalent transformation, we get: in, get The covariance satisfies Represents r i dimensional identity matrix.

4. The distributed security fusion estimation method under random event attack according to claim 1 is characterized in that: In step c, the sensor calculates the event trigger function value according to the following rules: Among them, η i (k) represents the trigger function value when there is no attack, represents the trigger function value under attack, β i represents the event trigger parameter, and the information fed back by the estimator satisfies in represents the local prior estimate of the state variable x(k), and represents the attack signal, generated by the random event attacker, y i (k) represents the compressed original measurement data.

5. The distributed security fusion estimation method under random event attack according to claim 1 is characterized in that: In step d, the event-triggered scheduler based on the sensor calculates the current data y i (k) corresponds to the decision variable γ i (k)(or ), the calculation rule is: Among them, the random variable ε i (k) is a uniform distribution in the interval [0,1] and is independent of other variables, η i (k) represents the trigger function value when there is no attack, Indicates the trigger function value under attack; When γ i (k) = 1 or Indicates that the current data y needs to be transmitted i (k).

6. The distributed security fusion estimation method under random event attack according to claim 1 is characterized in that: In step e, the local estimator is based on the currently available information set Perform the estimation process, specifically: Where A represents the system matrix of the linearized power system, B represents the input matrix of the linearized power system, represents the local estimate of sensor i at time k, P i y (k) represents the estimated error covariance of sensor i at time k, P i y- (k) represents the local prior estimate The error covariance of Q represents the covariance matrix of the system process noise, I represents the identity matrix with the same dimension as matrix A; u(k-1) represents the input variable at time k-1, P i y (k-1) represents the estimated error covariance of sensor i at time k-1, A T represents the transposed matrix of the system matrix A of the linearized power system, γ i (k) represents data y i (k) Corresponding decision variables; intermediate variables where ρ(k)=[β i +1-γ i (k)] / β i ,in, is the measurement matrix C i The transposed matrix of i Indicates event trigger parameters; available information collection And the local estimation process in the case of random attack will be variable γ i (k) is replaced by 7. The distributed security fusion estimation method under random event attack according to claim 1 is characterized in that: In step f, the index variable λ is updated according to the preset rules i (k), the specific preset rules are as follows: Among them, γ i (k) is the data y at time k i (k) corresponds to the decision variable, γ i (k-1) is the data y at the previous moment i (k-1) corresponding decision variables, λ i (0) = 0; When i When (k) = 0, the local estimate is transmitted and the matrix P i y (k) The diagonal elements of 8. The distributed security fusion estimation method under random event attack according to claim 1 is characterized by: In step g, the state compensation scheme adopted by the fusion center is: Define the vector Λ(k)=[λ1(k),...,λ N (k)] and the variable τ(k) = N - ‖Λ(k)‖0, Among them, ‖·‖0 represents the zero norm of the vector; Then, assuming that the measurement accuracy of sensor 1 is better than that of any other N-1 sensors, and resetting that λ1(k)=0 holds for any k>0, the information set to be fused is: The scheme of diagonal matrix weighted fusion is: Among them, D j =diag[d j1 ,d j2 ,...,d j(5n) ], n represents the number of synchronous motors in the current power system, represents [1,1,...,1], Denotes the cross covariance matrix P lj (k) the i-th diagonal element; Calculate the fusion estimation error covariance P f (k) specifically: get Among them, D i , D j is the diagonal weight matrix, P ij is the cross covariance matrix.

Images