A Generator State Estimation Method Applicable to Non-Gaussian Noise

Abstract

The present invention discloses a generator state estimation method applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion, comprising the following steps: (1) initializing each parameter for generator state estimation; (2) constructing a DOS network attack model to determine whether a DOS network attack occurs; (3) calculating the state prediction value and the error covariance matrix; (4) calculating the measurement prediction value and the cross-covariance matrix; (5) calculating the state gain, the state estimation value, and the state estimation error covariance; (6) performing iterative calculations for the next moment according to steps (3)-(5) until the loop ends, and outputting the state estimation result. This method combines the Cauchy kernel correlation entropy loss criterion (CMC) with the cubature Kalman filter (CKF) framework to make up for the defects of traditional state estimation methods in non-Gaussian noise environments and DOS network attacks, such as low estimation accuracy and weak robustness.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system analysis and control. Specifically, it relates to a generator state estimation method applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion. Background Art

[0002] Due to the rapidity and synchronization of data measurement, the synchronized phasor measurement unit (PMU) based on the wide area measurement system (WAMS) has been widely used in the field of power system real-time monitoring and control in recent years. However, in the actual measurement process, due to the influence of measurement equipment failures and external interferences, there are often errors and bad data in the measurement data. If these data are directly used for electromechanical transient analysis, it may lead to incorrect analysis results and control strategies, thus seriously threatening the safe and stable operation of the power system. State estimation can effectively filter out the errors and bad data existing in the PMU; therefore, in order to meet the requirements of power grid analysis and control, it is necessary to perform dynamic state estimation on the generator during the electromechanical transient process.

[0003] Currently, the relatively classical methods for power system dynamic state estimation mainly focus on the extended Kalman filter (EKF), unscented Kalman filter (UKF), particle filter (PF), and their corresponding improved methods. However, it should be noted that these methods all assume that the system noise is Gaussian white noise, that is, the noise variance is a constant; in the actual operation and analysis of the power system, it is very difficult to accurately obtain the statistical characteristics of the system noise and they are dynamically changing. Assuming that the system noise covariance matrix is a constant will cause the set value of the system noise covariance matrix to not match the true value, thus seriously affecting the dynamic state estimation results and reducing the state estimation accuracy. Summary of the Invention

[0004] The purpose of the present invention is to reduce the influence of non-Gaussian noise in the measured quantities on dynamic state estimation, overcome the deficiencies of traditional filtering methods, improve the accuracy of generator dynamic state estimation, and provide solid data information for the safe and stable operation of the power grid.

[0005] The first aspect of the present invention provides a generator state estimation method applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion, including the following steps:

[0006] (1) Initialize each parameter of the generator state estimation;

[0007] Select a fourth-order generator model as the generator dynamic state estimation model;

[0008] The state equation and measurement equation of the generator model for dynamic state estimation are expressed as:

[0009]

[0010] In the formula, φ(·) represents the generator state equation, h(·) represents the measurement equation, x represents the state variable, u represents the control variable, z represents the measurement vector, k and k + 1 represent time instants; w represents the system noise, v is the measurement noise, and it is assumed that w and v respectively satisfy the Gaussian distributions of w ~ N(0, Q) and v ~ N(0, R), where Q and R respectively represent the covariance matrices satisfied by the system noise and the measurement noise, w and v are independent of each other and independent of the state variable;

[0011] The form of the generator dynamic state estimation model is as follows:

[0012]

[0013]

[0014]

[0015]

[0016] In the formula, δ represents the absolute power angle of the generator rotor; ω and ω0 are the electrical angular velocity and its initial value respectively; T j represents the inertia constant, K D represents the damping coefficient; T m and T e respectively represent the mechanical power and electromagnetic power of the generator; e′ q and e′ d respectively represent the q-axis and d-axis transient electromotive forces of the generator; E fd is the excitation voltage of the generator stator; xd and x′ d are the synchronous reactance and transient reactance of the generator d-axis respectively; T d ′0 and T q ′0 are the open-circuit transient time constants of the generator d-axis and q-axis respectively; i d and i q respectively represent the stator currents on the d-axis and q-axis of the generator; x q and x′ q respectively represent the synchronous reactance and transient reactance on the q-axis;

[0017] Set the state variable of the generator as x k = [δ ω e' q e' d T , and take the mechanical power of the generator, the stator excitation voltage, and the currents i R and i I on the R-axis and I-axis of the stator as the control input quantities, and the input variable is u k = [T m E​fd i R i I T The measurement variable is set to z k = [δ ω e R e I T ;

[0018] When k = 0, the initial state variable Estimation error covariance System noise covariance Q k-1 and measurement noise covariance R k , kernel width β and threshold ε;

[0019] (2) Construct a DOS network attack model to determine whether a DOS network attack has occurred;

[0020] In the DOS attack, the measurement loss uses the data set u k (k = 1, 2,..., m) represents the data loss, which is specifically expressed as follows:

[0021]

[0022] var(u k = 1) = P s (u k = 0).P s (u k = 1)

[0023] = ρ.(1 - ρ)

[0024] where ρ represents the probability of measurement loss, 0 ≤ ρ ≤ 1, and P s represents the covariance of the data set u k ;

[0025] Then the new measurement after the measurement loss is Z' k = u k × Z k k = 1, 2,..., m;

[0026] (3) Calculate the state prediction value and the error covariance matrix P k∣k-1 ;

[0027] When k > 0, calculate the state variable prediction value at time k and the error covariance matrix P k∣k-1 , and the calculation formulas are as follows:

[0028]

[0029] ​​

[0030]

[0031]

[0032]

[0033] Among them, S k-1∣k-1 is obtained by performing Cholesky decomposition on the covariance P k-1∣k-1 , X i,k-1∣k-1 and are the i-th cubature point at the (k - 1)-th moment of the generator state variables and their propagated points at the k-th moment, n is the number of state variables, and the cubature point set ξ i is expressed as follows:

[0034]

[0035] (4) Calculate the measurement prediction value and the cross-covariance matrix P xz,k∣k-1 ;

[0036] Calculate the measurement prediction value and the cross-covariance matrix P xz,k∣k-1 at the k-th moment. The calculation formulas are as follows:

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] Among them, S k∣k-1 is obtained from the Cholesky decomposition of the covariance P k∣k-1 , X i,k∣k-1 and Z i,k∣k-1 are the i-th cubature point at the (k - 1)-th moment of the generator state variables and their propagated points Z i,k∣k-1 .

[0043] (5) Calculate the state gain the state estimate value and the state estimation error covariance P k∣k ;

[0044] At the k-th moment, let t = 1, calculate the state gain the state estimate value and the state estimation error covariance P k∣k, the calculation formula is as follows:

[0045]

[0046]

[0047]

[0048] Among them, the estimated error covariance is multiplied by the gain and adjusted to obtain whose calculation formula is:

[0049] The measurement error covariance has the following calculation formula:

[0050] Among them, the measurement noise covariance is multiplied by the gain and adjusted to obtain whose calculation formula is:

[0051] Among them, the Cauchy entropy gain and have the following calculation formula:

[0052]

[0053]

[0054] In the above formula, C β (·) represents the kernel function of the Cauchy kernel correlation entropy, which is specifically as follows:

[0055]

[0056]

[0057]

[0058] In the formula: B p,k∣k-1 and B r,k respectively represent the Cholesky decomposition of P k∣k-1 and R k at time step k, U represents the Cauchy entropy gain, and β represents the kernel width of the Cauchy entropy;

[0059] Compare the state estimates at time step t and t - 1 and

[0060] If is satisfied, where ε is a preset sufficiently small positive number, then let And execute step (6); otherwise, let t = t + 1, and return to step (5) to continue the loop.

[0061] (6) Perform iterative calculations for the next moment according to steps (3)-(5) until the loop ends, and output the state estimation result.

[0062] The second aspect of the present invention provides a generator state estimation system applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion, including a parameter initialization module;

[0063] A DOS attack judgment module for constructing a DOS network attack model and judging whether a DOS network attack occurs;

[0064] For calculating the state prediction value and the error covariance matrix P k∣k-1 of the first calculation module;

[0065] For calculating the measurement prediction value and the cross-covariance matrix P xz,k∣k-1 of the second calculation module;

[0066] For the state gain state estimation value and the state estimation error covariance P k∣k of the third calculation module;

[0067] An output module for outputting the state result;

[0068] The parameter initialization module, the DOS attack judgment module, the first calculation module, the second calculation module, the third calculation module, and the output module are connected in sequence to complete the described generator state estimation method applicable to non-Gaussian noise based on the adaptive Cauchy kernel correlation entropy loss criterion.

[0069] The third aspect of the present invention provides a generator controller, including:

[0070] A memory; and

[0071] A processor coupled to the memory, and the processor is configured to perform dynamic state estimation on the generator in the electromechanical transient process by using the described generator state estimation method applicable to non-Gaussian noise based on the adaptive Cauchy kernel correlation entropy loss criterion.

[0072] The present invention has prominent substantive features and remarkable progress. Specifically, in order to reduce the impact of unknown system noise on dynamic state estimation and overcome the deficiencies of traditional filtering methods, the present invention proposes a generator state estimation method applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion. This method combines the Cauchy kernel correlation entropy loss criterion (CMC) with the cubature Kalman filter (CKF) framework. It can not only use the CKF to solve strong model nonlinear problems, but also use the CKL developed in information-theoretic learning (ITL) to handle non-Gaussian noise interference problems, making up for the defects of traditional state estimation methods in non-Gaussian noise environments and DOS network attacks, such as low estimation accuracy and weak robustness. It has good superiority and practicality for the prediction of generator states. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is a flowchart of the method according to an embodiment of the present invention.

[0074] Figure 2 It is a structure diagram of the IEEE 10-machine 39-bus system.

[0075] Figure 3 It is a comparison diagram of the power angle estimation results of different methods for generator G2 under Gaussian noise.

[0076] Figure 4 It is a comparison diagram of the angular velocity estimation results of different methods for generator G2 under Gaussian noise.

[0077] Figure 5 It is a comparison diagram of the dynamic estimation of the q-axis transient electromotive force of different methods for generator G2 under Gaussian noise.

[0078] Figure 6 It is a comparison diagram of the dynamic estimation of the d-axis transient electromotive force of different methods for generator G2 under Gaussian noise.

[0079] Figure 7 It is a comparison diagram of the power angle estimation results of different methods for generator G2 under DOS network attack.

[0080] Figure 8 It is a comparison diagram of the angular velocity estimation results of different methods for generator G2 under DOS network attack. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0081] The present invention will be further clarified below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. After reading the present invention, various equivalent forms of modification by those skilled in the art fall within the scope defined by the appended claims of this application.

[0082] Embodiment 1

[0083] As Figure 1 shown, a generator state estimation method applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion is characterized by the following steps:

[0084] (1) Initialize each parameter of the generator state estimation;

[0085] Select a fourth-order generator model as the generator dynamic state estimation model;

[0086] The state equation and measurement equation of the generator model for dynamic state estimation are expressed as:

[0087]

[0088] where φ(·) represents the generator state equation, h(·) represents the measurement equation, x represents the state variable, u represents the control variable, z represents the measurement vector, k and k + 1 represent moments; w represents the system noise, v is the measurement noise, and it is assumed that w and v respectively satisfy the Gaussian distributions of w ∼ N(0, Q) and v ∼ N(0, R), where Q and R respectively represent the covariance matrices satisfied by the system noise and the measurement noise, w and v are independent of each other and independent of the state variable;

[0089] The form of the generator dynamic state estimation model is as follows:

[0090]

[0091]

[0092]

[0093]

[0094] where δ represents the absolute power angle of the generator rotor; ω and ω0 are respectively the electrical angular velocity and its initial value; T j represents the inertia constant, K D represents the damping coefficient; T m and T e respectively represent the mechanical power and electromagnetic power of the generator; e′ q and e′ d respectively represent the q-axis and d-axis transient electromotive forces of the generator; E fd is the excitation voltage of the generator stator; x d and x′ d are respectively the synchronous reactance and transient reactance of the generator d-axis; T′ d0 and T′ q0 are respectively the open-circuit transient time constants of the generator d-axis and q-axis; i d and i qrespectively represent the stator currents on the d-axis and q-axis of the generator; x q and x′ q respectively represent the synchronous reactance and transient reactance on the q-axis;

[0095] Set the state variables of the generator as x k =[δ ω e' q e' d T , take the mechanical power of the generator, the stator excitation voltage, and the currents i R and i I on the R-axis and I-axis of the stator as the control input variables, and the input variable is u k =[T m E fd i R i I T , and set the measurement variable as z k =[δ ω e R e I T ;

[0096] When k = 0, the initial state variables estimation error covariance system noise covariance Q k-1 , measurement noise covariance R k , kernel width β and threshold ε;

[0097] (2) Construct a DOS network attack model to determine whether a DOS network attack occurs;

[0098] In the DOS attack, the measurement loss uses the data set u k (k = 1, 2,..., m) represents the data loss, and the specific representation is as follows:

[0099]

[0100] var(u k =1)=P s (u k =0).P s (u k =1)=ρ.(1 - ρ)

[0101] where ρ represents the probability of measurement loss, 0 ≤ ρ ≤ 1, and P s represents the covariance of the data set u k ;

[0102] Then the new measurement after the measurement loss is Z' k =u k ×Z k ​​​k = 1, 2,..., m;

[0103] (3) Calculate the state prediction value and the error covariance matrix P k∣k-1 ;

[0104] When k > 0, calculate the predicted value of the state variable at time k and the error covariance matrix P k∣k-1 , and the calculation formulas are as follows:

[0105]

[0106]

[0107]

[0108]

[0109]

[0110] where S k-1∣k-1 is obtained by performing Cholesky decomposition on the covariance P k-1∣k-1 , X i,k-1∣k-1 and are the i-th cubature point of the generator state at time k - 1 and its propagated point at time k, n is the number of state variables, and the cubature point set ξ i is expressed as follows:

[0111]

[0112] (4) Calculate the measurement prediction value and the cross-covariance matrix P xz,k∣k-1 ;

[0113] Calculate the predicted value of the measurement at time k and the cross-covariance matrix P xz,k∣k-1 , and the calculation formulas are as follows:

[0114]

[0115]

[0116]

[0117]

[0118]

[0119] where S k∣k-1 is obtained from the Cholesky decomposition of the covariance P k∣k-1 , Xi,k∣k-1 and Z i,k∣k-1 is the i-th volume point and its propagation point Z at the (k - 1)-th moment of the generator state quantity i,k∣k-1 .

[0120] (5) Calculate the state gain State estimate value and the state estimation error covariance P k∣k ;

[0121] At the k-th moment, let t = 1, Calculate the state gain State estimate value and the state estimation error covariance P k∣k , and the calculation formulas are as follows:

[0122]

[0123]

[0124]

[0125] Among them, the estimation error covariance is adjusted by the gain to obtain and its calculation formula is:

[0126] The measurement error covariance has the following calculation formula:

[0127] Among them, the measurement noise covariance is adjusted by the gain to obtain and its calculation formula is:

[0128] Among them, the Cauchy entropy gain and have the following calculation formulas:

[0129]

[0130]

[0131] In the above formula, C β (·) represents the kernel function of the Cauchy kernel correlation entropy, specifically as follows:

[0132]

[0133]

[0134]

[0135] where: B p,k∣k-1 , B r,k respectively represent the Cholesky decomposition of P k∣k-1 and R k at time k, U represents the Cauchy entropy gain, and β represents the kernel width of the Cauchy entropy;

[0136] Compare the state estimates at step t and step t - 1 and

[0137] If it satisfies where ε is a preset sufficiently small positive number, then let and execute step (6); otherwise, let t = t + 1, and return to step (5) to continue the loop.

[0138] (6) Perform iterative calculations for the next moment according to steps (3)-(5) until the loop ends, and output the state estimation result.

[0139] To verify the effectiveness and practicality of the method in this embodiment, the IEEE 10-machine 39-bus standard system is used as the test system to verify the algorithm performance, and the system structure diagram is as shown in Figure 2 shown.

[0140] During the algorithm verification, generator G2 in the system is taken as the research object, and the fourth-order generator model is used for the experiment during the simulation. The generator inertia time constant T J is taken as 30.3. The damping coefficient D is taken as 2, and the system noise covariance Q = 10 - 8 I 4×4 , and the measurement noise covariance R = 10 -5 I 4×4 . It is assumed that a three-phase metallic short circuit occurs at the outlet of the Bus16 - Bus21 line in the system, and the system transitions to a new steady state after 0.12 s.

[0141] The PSCAD / EMTDC software is used to simulate the PMU device to collect measurement data, and the true values of the generator operation are obtained. The measurement data values are formed by adding random noise to the true values. During the simulation experiment, to verify the effectiveness of the proposed algorithm, the CKF, MCCKF, and the method proposed in this embodiment are respectively used to estimate the state of generator G2. The initial value of the state estimation is taken as the steady-state value of the previous moment, and the sampling frequency of the PMU is 50 samples per second. Then, 200 independent Monte Carlo simulations are performed for each case study to obtain more reliable statistical characteristics. The comparison of the state estimation results of generator G2 by different methods under Gaussian noise is shown in Table 1 and Figures 3 - 6 shown. When the packet loss probability ρ = 0.05, the comparison of the first two state estimation results of each method for generator G2 during the DOS network attack is asFigures 7 - 8 as shown

[0142] In order to conduct a comparative analysis of the estimation results between different algorithms, the overall performance index E x and the root mean square error (RMSE) are used as indicators to compare the performance between algorithms.

[0143]

[0144] In the formula, N MC and N T respectively represent the total number of Monte Carlo runs and the total simulation time; X i,k and respectively represent the true value and the estimated value of the state.

[0145] Table 1 Comparison of the overall performance of different methods

[0146]

[0147] It can be clearly seen from the data in Table 1 and the comparison curves in the figure that the method proposed in this embodiment uses the CMC optimal criterion and cubature Kalman filter, and dynamically adjusts the error covariance and noise covariance matrices through the correlation entropy gain. It can not only greatly reduce the sensitivity to non-Gaussian noise in the measurement information, but also effectively solve the strong nonlinear problem, thereby further improving the estimation accuracy. In addition, the Cauchy kernel can not only prevent the matrix from having singular value phenomena, but is also insensitive to the kernel width, and can more accurately track the dynamic state changes of the generator, and the estimation accuracy is much higher than that of the CKF and MCCKF methods. It can be seen that the method proposed in this embodiment has more accurate estimation accuracy and stronger robustness, and has better applicability.

[0148] Embodiment 2

[0149] This embodiment provides a generator state estimation system applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion, including a parameter initialization module;

[0150] a DOS attack judgment module for constructing a DOS network attack model and judging whether a DOS network attack occurs;

[0151] a first calculation module for calculating the state prediction value and the error covariance matrix P k∣k-1 ;

[0152] a second calculation module for calculating the measurement prediction value and the cross-covariance matrix P xz,k∣k-1 ;

[0153] for the state gain the state estimated value and the state estimation error covariance P k∣k The third calculation module;

[0154] An output module for outputting a state result;

[0155] The parameter initialization module, the DOS attack determination module, the first calculation module, the second calculation module, the third calculation module, and the output module are connected in sequence to complete the generator state estimation method applicable to non-Gaussian noise based on the adaptive Cauchy kernel correlation entropy loss criterion described in Embodiment 1.

[0156] Embodiment 3

[0157] This embodiment provides a generator controller, including:

[0158] A memory; and

[0159] A processor coupled to the memory, the processor being configured to perform dynamic state estimation on a generator in an electromechanical transient process by using the generator state estimation method applicable to non-Gaussian noise based on the adaptive Cauchy kernel correlation entropy loss criterion described in Embodiment 1 based on instructions stored in the memory.

[0160] Among them, the memory may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory stores, for example, an operating system, application programs, a boot loader (Boot Loader), and other programs.

[0161] The generator controller may further include an input / output interface, a network interface, a storage interface, etc. These interfaces and the memory and the processor may be connected through a bus, for example. Among them, the input / output interface provides a connection interface for input / output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface provides a connection interface for various networking devices. The storage interface provides a connection interface for external storage devices such as an SD card and a USB flash drive.

[0162] Those skilled in the art should understand that the embodiments disclosed in the present invention may be provided as a method, a system, or a computer program product. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer non-transitory readable storage media (including but not limited to a magnetic disk memory, a CD-ROM, an optical memory, etc.) containing computer program code.

[0163] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems) according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one or more flows and / or blocks Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in one or more blocks.

[0164] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in one or more flows and / or blocks Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in one or more blocks.

[0165] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more flows and / or blocks Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in one or more blocks.

[0166] The above are only the preferred embodiments disclosed by the present invention, and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A generator state estimation method applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion, characterized in that, It includes the following steps: (1) Initialize each parameter of the generator state estimation; Select a fourth-order generator model as the generator dynamic state estimation model; The forms of the state equation and the measurement equation for the dynamic state estimation of the generator model are expressed as: In the formula, φ(·) represents the generator state equation, h(·) represents the measurement equation, x represents the state variable, u represents the control variable, z represents the measurement vector, k and k + 1 represent moments; w represents the system noise, v is the measurement noise, and it is assumed that w and v respectively satisfy the Gaussian distributions of w~N(0,Q) and v~N(0,R), where Q and R respectively represent the covariance matrices satisfied by the system noise and the measurement noise, w and v are independent of each other and independent of the state variable; The form of the generator dynamic state estimation model is as follows: where δ represents the absolute power angle of the generator rotor; ω and ω0 are the electrical angular velocity and its initial value respectively; T j represents the inertia constant, K D represents the damping coefficient; T m and T e represent the mechanical power and electromagnetic power of the generator respectively; e′ q and e′ d represent the transient electromotive forces of the q-axis and d-axis of the generator respectively; E fd is the excitation voltage of the generator stator; x d and x′ d are the synchronous reactance and transient reactance of the d-axis of the generator respectively; T′ d0 and T′ q0 are the open-circuit transient time constants of the d-axis and q-axis of the generator respectively; i d and i q represent the stator currents on the d-axis and q-axis of the generator respectively; x q and x′ q are the synchronous reactance and transient reactance on the q-axis respectively; Set the state variables of the generator to x k = [δ ω e' q e' d T , take the mechanical power of the generator, the stator excitation voltage, and the currents i R and i I on the stator R-axis and I-axis as control input variables, and the input variable is u k = [T m E fd i R i I T , and set the measurement variable to z k = [δ ω e R e I T ;​​​ When k = 0, the initial state variable Estimation error covariance System noise covariance Q k-1 , measurement noise covariance R k , kernel width β and threshold ε; (2) Construct a DOS network attack model and judge whether a DOS network attack occurs; Measurement loss in DOS attack uses dataset u k (k = 1, 2, …, m) represents data loss, which is specifically expressed as follows: var(u k = 1) = P s (u k = 0).P s (u k = 1) =ρ.(1 - ρ) where ρ represents the probability of measurement loss, 0 ≤ ρ ≤ 1, P s represents the covariance of the dataset u k ; The new measurement after measurement loss is Z'. k = u k × Z k k = 1, 2, ..., m; (3) Calculate the predicted state value and the error covariance matrix P k∣k-1 ; When k > 0, calculate the predicted value of the state variable at time k and the error covariance matrix P k∣k-1 , and the calculation formulas are as follows: where S k-1∣k-1 is obtained by performing Cholesky decomposition on the covariance P k-1∣k-1 , X i,k-1∣k-1 and are the i-th volume point at the (k - 1)-th moment of the generator state quantity and its propagation point at the k-th moment, n is the number of state variables, and the volume point set ξ i is expressed as follows: (4) Calculate the measured prediction value and the cross-covariance matrix P xz,k∣k-1 ; Calculate the predicted value of the measurement at time k and the cross-covariance matrix P xz,k∣k-1 , and the calculation formula is as follows: where S k∣k-1 is obtained from the Cholesky factorization of the covariance P k∣k-1 , X i,k∣k-1 and Z i,k∣k-1 are the i-th volume point and its propagated point Z at the (k-1)-th moment of the generator state variables i,k∣k-1； (5) Calculate the state gain State estimate and the state estimation error covariance P k∣k ; At time k, let t = 1, Calculate the state gain State estimate And the state estimation error covariance P k∣k , and the calculation formulas are as follows: Among them, the estimated error covariance is multiplied by the gain and then adjusted to obtain whose calculation formula is as follows: Measurement error covariance The calculation formula is as follows: Among them, the measurement noise covariance is obtained by being gain-adjusted after adjustment and its calculation formula is: Among them, the Cauchy entropy gain and The calculation formula is as follows: In the above formula, C β (·) represents the kernel function of the Cauchy kernel correlation entropy, which is specifically as follows: Where: B p,k∣k-1 , B r,k respectively represent the Cholesky decomposition of P k∣k-1 and R k at time k, U represents the Cauchy entropy gain, and β represents the kernel width of the Cauchy entropy; Compare the state estimates at step t and step t-1 and If the following condition is satisfied where ε is a preset positive number small enough, then let and execute step (6); otherwise, let t = t + 1, and return to step (5) to continue the loop; (6) Perform iterative calculations for the next moment according to steps (3)-(5) until the loop ends, and output the state estimation result.

2. A generator state estimation system applicable to non-Gaussian noise based on an adaptive Cauchy kernel correlation entropy loss criterion, characterized in that: It includes a parameter initialization module; A DOS attack judgment module for constructing a DOS network attack model and judging whether a DOS network attack occurs; A first calculation module for calculating a state prediction value and an error covariance matrix P k∣k-1 ; For calculating a measured predicted value and a cross-covariance matrix P xz,k∣k-1 of the second calculation module; For state gain State estimate value and state estimation error covariance P k∣k The third calculation module; An output module for outputting the state result; The parameter initialization module, the DOS attack judgment module, the first calculation module, the second calculation module, the third calculation module, and the output module are connected in sequence to complete the generator state estimation method applicable to non-Gaussian noise based on the adaptive Cauchy kernel correlation entropy loss criterion described in claim 1.

3. A generator controller, characterized in that: It includes: A memory; And A processor coupled to the memory, the processor being configured to perform dynamic state estimation of the generator in the electromechanical transient process by using the generator state estimation method applicable to non-Gaussian noise based on the adaptive Cauchy kernel correlation entropy loss criterion described in claim 1 based on the instructions stored in the memory.

Images