Electromechanical transient estimation method and system based on robust H infinite unscented particle filtering

Abstract

According to the electromechanical transient estimation method and system based on robust H infinite unscented particle filtering, on the basis of a dynamic estimation model of a generator, unscented particle filtering iteration solution is adopted; determining an estimation error covariance matrix by using a positive scalar parameter of a system uncertainty error defined based on a robust H infinity theory, a state prediction error covariance matrix and a cross covariance matrix of a predicted value and a measured value, and updating a Kalman gain; based on the state variable, estimating an error covariance matrix to establish suggested distribution, and resampling from the suggested distribution to obtain a new particle set; updating a dynamic estimation model of the generator by using the corrected system noise covariance matrix and the observation noise covariance matrix, and solving an obtained state variable estimation value by adopting unscented particle filter iteration; taking the state variable estimation value and the weighted mean value of the weight thereof as an electromechanical transient estimation result of the generator; the problems of accuracy and robustness caused by communication noise and uncertainty of model parameters in the dynamic estimation process of the generator are solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of generator electromechanical transient estimation, and particularly relates to an electromechanical transient estimation method and system based on robust H infinite unscented particle filtering. BACKGROUND

[0002] The phasor measurement unit (PMU) has been widely applied in real-time monitoring and control of power systems due to its rapidity and synchronicity of data measurement. However, due to the influence of measurement device failure and external interference in the measurement process, errors and bad data often appear in the measurement data. If these data are directly used for electromechanical transient analysis, it may lead to incorrect analysis results and control strategies, and further endanger the safety of the power grid. State estimation can effectively filter out the errors and bad data in the PMU, so it is of great significance to reasonably estimate the state of the generator in the electromechanical transient process to meet the dynamic demand of the power grid.

[0003] In recent years, a large number of researches have been conducted on the dynamic state estimation of generators. The traditional generator state estimation method based on linear Kalman filter (KF) ignores the nonlinear characteristics of the generator, and the linearization process of the extended Kalman filter (EKF) for estimating the state of the generator has a large truncation error and poor estimation accuracy. Compared with the EKF, the unscented Kalman filter (UKF) has improved accuracy, but it is easily limited by the state dimension and parameters, and it is difficult to apply to large-scale state estimation. In view of the limited filtering accuracy of the filtering method of the strong nonlinear non-Gaussian system under the Kalman framework, the PF does not consider the latest measurement information which easily leads to particle degradation, and causes poor model robustness when the model is uncertain. SUMMARY

[0004] In order to solve the problems in the prior art, the application provides an electromechanical transient estimation method and system based on robust H infinite unscented particle filtering, which solves the problems of accuracy and robustness caused by communication noise and model parameter uncertainty in the dynamic estimation process of the generator.

[0005] The application adopts the following technical solutions.

[0006] The application provides an electromechanical transient estimation method based on robust H infinite unscented particle filtering, which comprises the following steps. Step 1, a dynamic estimation model of the generator is established; Step 2: Based on the dynamic estimation model of the generator, the state variables are iteratively solved using unscented particle filtering, including a prediction process and an update process. During the prediction process, the estimation error covariance matrix is ​​determined using the positive scalar parameter of the system uncertainty error defined by the robust H-infinity theory, the state prediction error covariance matrix, and the cross-covariance matrix of the predicted and measured values. Based on the estimation error covariance matrix, the Kalman gain is updated. During the update process, a Gaussian distribution is established as the proposal distribution based on the state variables and the estimation error covariance matrix. A new particle set is obtained by resampling from the proposal distribution. Step 3: Based on the system noise, estimation error covariance matrix, positive scalar parameters, and the set observation noise covariance matrix, establish an adaptive adjustment model for the noise covariance matrix based on the Riccati equation; solve the adaptive adjustment model to obtain the equivalent system noise covariance matrix; use the weighted sum of the equivalent system noise covariance matrix and the system noise covariance matrix as the corrected system noise covariance matrix, and use the weighted sum of the set observation noise covariance matrix and the observation noise covariance matrix as the corrected observation noise covariance matrix; update the generator's dynamic estimation model using the corrected system noise covariance matrix and the observation noise covariance matrix. Step 4: Based on the updated dynamic estimation model of the generator, the estimated values ​​of the state variables are obtained by iterative solution using unscented particle filtering; the weighted average of the estimated values ​​of the state variables and their weights is used as the electromechanical transient estimation result of the generator.

[0007] Preferably, step 2 includes: Step 2.1, during the prediction process, calculate the state variables. The sensitivity coefficient matrix of the generator-injected power to the terminal voltage state; in terms of state variables. Jacobian matrix of the generator observation equation The sensitivity coefficient matrix for the generator injected power relative to the terminal voltage state is shown in the following equation:

[0008] In the formula, The observation equation; , They are time points +1 for state variables and control variables.

[0009] Step 2.2, based on the robust H-infinity theory, and according to the estimated state variables, estimation error covariance matrix, system noise covariance matrix, and observation noise covariance matrix obtained through iteration, define the positive scalar parameters of the system uncertainty error. As shown in the following formula:

[0010] In the formula, is an upper limit function, is a positive scalar parameter of system uncertainty error, is a two-norm function, and are the true value and the estimated value of the state variable at the initial time, respectively, and are the true value and the estimated value of the state variable at time , respectively, are the true value and the estimated value of the state variable at time , respectively, are the system noise and the observation noise at time , respectively, is the total number of times, are the inverse matrices of the estimated error covariance matrices at the initial time and at time , respectively, is the inverse matrix of the system noise covariance matrix at time , and is the inverse matrix of the observation noise covariance matrix at time .

[0011] Step 2.3, the state prediction error covariance matrix is revised by using the sensitivity coefficient matrix and the positive scalar parameter to obtain the observation noise error covariance matrix , as shown in the following formula:

[0012] In the formula, is the state prediction error covariance matrix obtained by iterative solution, is the sensitivity coefficient matrix, is the observation noise covariance matrix, is an identity matrix.

[0013] Step 2.4, the augmented covariance matrix is constructed by using the state prediction error covariance matrix, the mutual covariance matrix of the predicted value and the measured value, and the state prediction error covariance matrix is revised by using the augmented covariance matrix and the observation noise error covariance matrix to obtain the estimated error covariance matrix; based on the estimated error covariance matrix, the Kalman gain is updated. The estimated error covariance matrix is shown in the following formula:

[0014] In the formula, is the estimated error covariance matrix, is the state prediction error covariance matrix, is the mutual covariance square root matrix of the predicted value and the measured value; The Kalman gain is updated as shown in the following formula:

[0015] wherein, is the updated Kalman gain, is the pre-updated Kalman gain.

[0016] Step 2.5, in the updating process, a Gaussian distribution is established based on the state variable and the estimation error covariance matrix Step 2.6, resampling a new particle set from the proposal distribution.

[0017] Preferably, step 3 comprises: the set observation noise covariance matrix when the unit matrix is as shown in the following formula: ,

[0018] The system noise covariance matrix and the observation noise covariance matrix are corrected as shown in the following formula: ,

[0019] wherein, , are the corrected system noise covariance matrix and the observation noise covariance matrix respectively, , are weights.

[0020] The application further provides an electromechanical transient state estimation system based on robust H-infinity unscented particle filtering, comprising: a model establishing module configured to establish a dynamic estimation model of a generator; an algorithm fusion module configured to iteratively solve a state variable by using unscented particle filtering based on the dynamic estimation model of the generator, including a prediction process and an updating process; in the prediction process, a positive scalar parameter of system uncertainty error defined based on robust H-infinity theory, a state prediction error covariance matrix, a cross-covariance matrix of a predicted value and a measured value are used to determine an estimation error covariance matrix, and the estimation error covariance matrix is used to update a Kalman gain; in the updating process, a Gaussian distribution is established based on the state variable and the estimation error covariance matrix as a proposal distribution, and a new particle set is resampled from the proposal distribution. ​The model updating module is used for establishing a noise covariance matrix adaptive adjustment model based on a Riccati equation according to system noise, an estimated error covariance matrix, a positive scalar parameter and a set observation noise covariance matrix; an equivalent system noise covariance matrix is obtained by solving the adaptive adjustment model; a weighted sum of the equivalent system noise covariance matrix and the system noise covariance matrix is taken as a corrected system noise covariance matrix, and a weighted sum of the set observation noise covariance matrix and an observation noise covariance matrix is taken as a corrected observation noise covariance matrix; and the dynamic estimation model of the generator is updated by using the corrected system noise covariance matrix and the observation noise covariance matrix. The state estimation module is used for obtaining a state variable estimation value by iteratively solving the dynamic estimation model of the generator based on the updated dynamic estimation model of the generator by using an unscented particle filter; and a weighted mean of the state variable estimation value and a weight thereof is taken as an electromechanical transient estimation result of the generator.

[0021] The application also provides a terminal, which comprises a processor and a storage medium; the storage medium is used for storing instructions; and the processor is used for operating according to the instructions to execute the steps of the method.

[0022] The application also provides a computer readable storage medium, which stores a computer program; the program is executed by a processor to implement the steps of the method.

[0023] Compared with the prior art, the application has at least the following beneficial effects: the application provides a dynamic estimation method based on a robust H∞ unscented particle filter (HUPF), and the HUPF has higher precision and stronger robustness; a positive scalar parameter of system uncertainty error defined based on robust H∞ theory is used to update an estimated error covariance matrix affected by system uncertainty by combining robust H∞ control and an unscented particle filter; and the estimated error covariance matrix is used to establish a resampling suggestion distribution, thereby improving the reliability of resampling.

[0024] The application also provides an adaptive updating mechanism of a system noise covariance matrix and an observation noise covariance matrix, thereby updating the dynamic estimation model of the generator and effectively inhibiting the adverse effects of model uncertainty on state estimation.

[0025] The HUPF proposed in the application takes into account the characteristics of unscented particle filtering (UPF) to calculate an important density function by using unscented transformation and consider the latest measurement information, so that the particle distribution is closer to the posterior probability distribution of the real state, thereby improving the estimation precision; the uncertainty constraint is introduced into the UPF based on robust H∞, and the state error covariance matrix is updated in real time, which can significantly improve the robustness of the algorithm to noise and parameter uncertainty; the estimation precision of the algorithm is maintained, and the algorithm is more suitable for processing nonlinear problems, thereby well meeting the needs of actual application scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a flow chart of the robust H-infinity unscented particle filter-based electromechanical transient estimation method proposed in the present application. DETAILED DESCRIPTION

[0027] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. The embodiments described in the present application are only a part of the embodiments of the present application, but not all the embodiments. Based on the spirit of the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0028] The present application proposes a robust H-infinity unscented particle filter-based electromechanical transient estimation method, which is especially suitable for the communication noise and model parameter uncertainty scene in the dynamic estimation process of a generator, such as Figure 1 As shown in the figure, the method comprises the following steps. Step 1, based on the dynamic estimation model of the generator, the rotor absolute power angle , the angular velocity , the q-axis transient electromotive force and the d-axis transient electromotive force are obtained to constitute a state variable vector; the mechanical power , the stator excitation voltage , the active current and the reactive current are obtained to constitute a control variable vector; the rotor absolute power angle , the angular velocity , the active electromotive force and the reactive electromotive force are obtained to constitute an observation variable vector; and the dynamic estimation model of the generator is established based on the state variable vector, the control variable vector and the observation variable vector.

[0029] The dynamic estimation model of the generator usually comprises a state equation and an observation equation, as shown in the following formula:

[0030] In the formula, , and are the state variable, the observation variable and the control variable, respectively; and are the state equation and the observation equation of the nonlinear system, respectively; and respectively represent system noise and observation noise, in the embodiment, the system noise and the observation noise are both assumed to be zero mean, and the system noise covariance matrix and the observation noise covariance matrix are Q and R respectively; Since the sub-transient process of the generator is short, the existing PMU device is difficult to accurately obtain the measurement value in the sub-transient process, so the d-axis and q-axis windings and the stator process related to the sub-transient process are ignored, and after simplification, the four-order dynamic model of the generator is represented as:

[0031] In the formula, is the absolute power angle of the rotor of the generator; and are the angular velocity and the rated synchronous speed respectively; is the inertia time constant; is the damping coefficient; and are the mechanical power and the electromagnetic power of the generator respectively; and are the q-axis and d-axis transient electromotive force of the generator respectively; is the stator excitation voltage of the generator; and are the d-axis synchronous reactance and the transient reactance of the generator respectively; and are the d-axis and q-axis open-circuit transient time constants of the generator respectively; and are the stator currents of the d-axis and q-axis of the generator respectively; and are the synchronous reactance and the transient reactance of the q-axis of the generator respectively.

[0032] In the dynamic estimation process, it is usually assumed that the related parameters and the control input variables are known, and only the state variables are unknown; in order to have high redundancy, the observation variables of the generator dynamic estimation are set as follows:

[0033] By comparing the above formula, the state variables of the generator are ; the control variables are ; and the observation variables are .

[0034] Step 2, based on the dynamic estimation model of the generator, state variables are solved iteratively by using the unscented particle filter, including a prediction process and an updating process; in the prediction process, a positive scalar parameter of system uncertainty error defined based on robust H-infinity theory, a state prediction error covariance matrix, a mutual covariance matrix of a predicted value and a measured value are used to determine an estimation error covariance matrix, and a Kalman gain is updated based on the estimation error covariance matrix; in the updating process, a Gaussian distribution is established as a proposal distribution based on the state variable and the estimation error covariance matrix, and a new particle set is obtained by resampling from the proposal distribution.

[0035] Specifically, step 2 comprises: Step 2.1, in the prediction process, the state variable is calculated at the sensitivity coefficient matrix of the generator injection power to the terminal voltage state; Specifically, the state variable at the Jacobian matrix of the generator observation equation is taken as the sensitivity coefficient matrix of the generator injection power to the terminal voltage state, and is as follows:

[0036] In the formula, is the observation equation; , are the state variable and the control variable at time +1 respectively.

[0037] The Jacobian matrix of the generator observation equation describes the change of the generator injection power with the slight change of the terminal voltage state thereof, and represents the linearized sensitivity relationship of the power system near the current operating point.

[0038] Step 2.2, based on the robust H-infinity theory, the positive scalar parameter of the system uncertainty error is defined according to the iteratively obtained state variable estimation value, the estimation error covariance matrix, the system noise covariance matrix and the observation noise covariance matrix ; In an actual power system, the uncertainty of the model mainly includes two aspects: unknown noise and model parameter uncertainty, and these factors will inevitably reduce the filtering accuracy or cause divergence. In order to suppress the adverse effects of the uncertainty and to improve the system robustness, the positive scalar parameter of the system uncertainty error is determined based on the robust H-infinity theory:

[0039] In the formula, is a upper limit function, is the positive scalar parameter of the system uncertainty error, is a two-norm function, and are the true value and the estimated value of the state variable at initial time, respectively, are the true value and the estimated value of the state variable at time , respectively, are the system noise and the observation noise at time , respectively, is the total number of times, are the inverse matrix of the estimated error covariance matrix at initial time and at time , respectively, is the inverse matrix of the system noise covariance matrix at time , is the inverse matrix of the observation noise covariance matrix at time .

[0040] No matter the energy and spectrum of the system noise and the observation noise, the energy of the estimated error will never exceed times of the total energy of the noise, which guarantees the performance of the filtering estimation under the system uncertainty.

[0041] Step 2.3, the state prediction error covariance matrix is revised by using the sensitivity coefficient matrix and the positive scalar parameter to obtain the observation noise error covariance matrix , as shown in the following formula:

[0042] In the formula, is the state prediction error covariance matrix obtained by iterative solution, is the sensitivity coefficient matrix, is the observation noise covariance matrix, is the unit matrix.

[0043] Step 2.4, the augmented covariance matrix is constructed by using the state prediction error covariance matrix, the cross covariance matrix of the prediction value and the measurement value, and the state prediction error covariance matrix is revised by using the augmented covariance matrix and the observation noise error covariance matrix to obtain the estimated error covariance matrix; based on the estimated error covariance matrix, the Kalman gain is updated; The estimated error covariance matrix is shown in the following formula:

[0044] In the formula, is the estimated error covariance matrix, is the state prediction error covariance matrix, is the cross covariance square root matrix of the prediction value and the measurement value; ​​​The Kalman gain is updated as follows:

[0045] wherein, is the updated Kalman gain, is the Kalman gain before updating.

[0046] In step 2.5, during the updating process, the state variable , the estimation error covariance matrix is established as a Gaussian distribution is resampled from the proposal distribution to obtain a new particle set.

[0047] In step 3, based on the Riccati equation, an adaptive adjustment model of the noise covariance matrix is established according to the system noise, the estimation error covariance matrix, a positive scalar parameter and a set observation noise covariance matrix; an equivalent system noise covariance matrix is obtained by solving the adaptive adjustment model; a weighted sum of the equivalent system noise covariance matrix and the system noise covariance matrix is taken as a corrected system noise covariance matrix, and a weighted sum of the set observation noise covariance matrix and the observation noise covariance matrix is taken as a corrected observation noise covariance matrix; the dynamic estimation model of the generator is updated with the corrected system noise covariance matrix and the corrected observation noise covariance matrix.

[0048] Specifically, when the set observation noise covariance matrix is a unit matrix, the equivalent system noise covariance matrix obtained by solving the adaptive adjustment model is as follows: ,

[0049] The system noise covariance matrix and the observation noise covariance matrix are corrected as follows: ,

[0050] wherein, , are the corrected system noise covariance matrix and the observation noise covariance matrix respectively, , are weights. The present application realizes online adaptive adjustment of the system noise covariance matrix and the observation noise covariance matrix based on robust H-infinity theory, and fully considers the influence of system noise and observation noise caused by uncertain factors outside the system on the electromechanical transient estimation results.

[0051] The square root of the estimated error covariance matrix is taken as the innovation covariance; when the innovation covariance increases by a proportion greater than a set threshold, the innovation covariance is reduced by a set step , In the embodiment, the proportion of the increase of the innovation covariance is the ratio of the amount of the increase of the innovation covariance to the innovation covariance, the set threshold is 10%, and the set step is 0.1. In the application, the interference of uncertain factors in the system, such as the sudden increase of the innovation and the accelerated degradation of the particles, on the system is considered, the weight is automatically reduced, the corrected system noise covariance matrix and the observation noise covariance matrix are more biased towards the equivalent system noise covariance matrix and the set observation noise covariance matrix, and in the case of abnormality or strong interference, the method proposed in the application artificially increases the system noise covariance matrix and the observation noise covariance matrix based on the robust H-infinity theory, avoids the divergence problem of the particle filter under model mismatching or strong interference, and significantly improves the robustness of state estimation.

[0052] Step 4: based on the updated dynamic estimation model of the generator, the state variable estimation value obtained by iterative solving using the unscented particle filter; and the weighted mean of the state variable estimation value and the weight thereof is taken as the electromechanical transient estimation result of the generator.

[0053] The application further provides an electromechanical transient estimation system based on the robust H-infinity unscented particle filter, which comprises: a model establishment module configured to establish a dynamic estimation model of the generator; an algorithm fusion module configured to iteratively solve the state variable based on the dynamic estimation model of the generator using the unscented particle filter, including a prediction process and an updating process; in the prediction process, the positive scalar parameter of the system uncertainty error defined based on the robust H-infinity theory, the state prediction error covariance matrix, the mutual covariance matrix of the predicted value and the measurement value are used to determine the estimation error covariance matrix, and the Kalman gain is updated based on the estimation error covariance matrix; in the updating process, a Gaussian distribution is established as a proposal distribution based on the state variable and the estimation error covariance matrix, and a new particle set is obtained by resampling from the proposal distribution; a model updating module configured to establish a noise covariance matrix adaptive adjustment model based on the Riccati equation according to the system noise, the estimation error covariance matrix, the positive scalar parameter and the set observation noise covariance matrix; obtain the equivalent system noise covariance matrix by solving the adaptive adjustment model; take the weighted sum of the equivalent system noise covariance matrix and the system noise covariance matrix as the corrected system noise covariance matrix, and take the weighted sum of the set observation noise covariance matrix and the observation noise covariance matrix as the corrected observation noise covariance matrix; and update the dynamic estimation model of the generator by using the corrected system noise covariance matrix and the observation noise covariance matrix; The state estimation module is configured to obtain state variable estimation values based on the updated dynamic estimation model of the generator by using an unscented particle filter to iteratively solve, and to take a weighted mean of the state variable estimation values and their weights as an electromechanical transient estimation result of the generator.

[0054] The present disclosure can be a system, a method, and / or a computer program product. The computer program product can include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.

[0055] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or punched tape, a

[0056] The computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0057] Computer readable program instructions for carrying out operations of the present disclosure can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0058] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, but not to limit it. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced, and any modification or replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A method for power system electromechanical transient estimation based on robust H-infinity unscented particle filter, characterized in that, Comprising: Step 1, establishing a dynamic estimation model of the generator; Step 2, based on the dynamic estimation model of the generator, using the unscented particle filter to iteratively solve the state variable, including the prediction process and the update process; in the prediction process, using the positive scalar parameter of the system uncertainty error defined based on the robust H infinity theory, the state prediction error covariance matrix, the mutual covariance matrix of the predicted value and the measured value to determine the estimation error covariance matrix, and updating the Kalman gain based on the estimation error covariance matrix; In the update process, a Gaussian distribution is established as a proposal distribution based on the state variable and the estimation error covariance matrix, and a new particle set is obtained by resampling from the proposal distribution; Step 3, based on the Riccati equation, establishing a noise covariance matrix adaptive adjustment model according to the system noise, the estimation error covariance matrix, the positive scalar parameter and the set observation noise covariance matrix; solving the adaptive adjustment model to obtain an equivalent system noise covariance matrix; Taking the weighted sum of the equivalent system noise covariance matrix and the system noise covariance matrix as the corrected system noise covariance matrix, and taking the weighted sum of the set observation noise covariance matrix and the observation noise covariance matrix as the corrected observation noise covariance matrix; Updating the dynamic estimation model of the generator with the corrected system noise covariance matrix and the observation noise covariance matrix; Step 4, based on the updated dynamic estimation model of the generator, using the unscented particle filter to iteratively solve the state variable estimation value; Taking the weighted mean of the state variable estimation value and its weight as the electromechanical transient estimation result of the generator.

2. The electromechanical transient estimation method based on robust H infinity unscented particle filter according to claim 1, wherein Step 2 comprises: Step 2.1, in the prediction process, calculate state variables the sensitivity coefficient matrix of the generator injected power to the terminal voltage state; with the state variables the Jacobian matrix of the generator observation equation the sensitivity coefficient matrix of the generator injected power to the terminal voltage state, as shown in the following formula: wherein is the observation equation; , are the state variable and control variable at time +1, respectively.

3. The electromechanical transient estimation method based on robust H infinity unscented particle filter according to claim 2, wherein Step 2 further comprises: Step 2.2, based on robust H-infinity theory, according to the state variable estimation value obtained by iteration, the estimation error covariance matrix, the system noise covariance matrix and the observation noise covariance matrix, the positive scalar parameter of the system uncertainty error is defined ; as shown in the following formula: wherein is an upper limit function, is a positive scalar parameter of the system uncertainty error, is a two-norm function, and are the true value and the estimated value of the state variable at the initial time, respectively, and are the true value and the estimated value of the state variable at the time , respectively, and are the system noise and the observation noise at the time , respectively, is the total number of times, and are the inverse matrices of the estimated error covariance matrices at the initial time and at the time , respectively, is the inverse matrix of the system noise covariance matrix at the time , and is the inverse matrix of the observation noise covariance matrix at the time .

4. The electromechanical transient estimation method based on robust H infinity unscented particle filter according to claim 3, wherein Step 2 further comprises: Step 2.

3. Correct the state prediction error covariance matrix with the sensitivity coefficient matrix and the positive scalar parameter to obtain the observation noise error covariance matrix As shown in the following formula: wherein is the state prediction error covariance matrix obtained by iteration, is the sensitivity coefficient matrix, is the observation noise covariance matrix, is the identity matrix.

5. The electromechanical transient estimation method based on robust H infinity unscented particle filter according to claim 4, wherein Step 2 further comprises: Step 2.4, constructing an augmented covariance matrix by using the state prediction error covariance matrix, the prediction value and the mutual covariance matrix of the measurement value, correcting the state prediction error covariance matrix by using the augmented covariance matrix and the observation noise error covariance matrix to obtain an estimation error covariance matrix; updating the Kalman gain based on the estimation error covariance matrix ; The estimation error covariance matrix is as shown in the following formula: wherein is the state prediction error covariance matrix, is the state prediction error covariance matrix, is the cross-covariance square root matrix of the prediction and the measurement. The Kalman gain is updated as shown in the following formula: wherein is the updated Kalman gain, is the pre-updated Kalman gain.

6. The electromechanical transient estimation method based on robust H infinity unscented particle filter according to claim 5, wherein Step 2 further comprises: Step 2.5, update process, based on state variables , estimate error covariance matrix establish a Gaussian distribution as proposal distribution, resample from proposal distribution to get new set of particles.

7. The electromechanical transient estimation method based on robust H infinity unscented particle filter according to claim 6, wherein Step 3 comprises: Set observation noise covariance matrix When the set observation noise covariance matrix is an identity matrix, the equivalent system noise covariance matrix obtained by solving the adaptive adjustment model As shown in the following formula: ， The system noise covariance matrix and the observation noise covariance matrix are corrected as shown in the following formula: ， wherein , are the modified system noise covariance matrix and the observation noise covariance matrix, respectively, , are weights.

8. A robust H-infinity unscented particle filter based power system transient estimation system for implementing the robust H-infinity unscented particle filter based power system transient estimation method according to any one of claims 1-7, characterized in that, Comprising: A model establishing module for establishing a dynamic estimation model of the generator; The algorithm fusion module is configured to iteratively solve the state variable based on a dynamic estimation model of the generator using an unscented particle filter, including a prediction process and an update process. In the prediction process, a positive scalar parameter of a system uncertainty error defined based on a robust H-infinity theory, a state prediction error covariance matrix, a predicted value, and a mutual covariance matrix of the measurement value are used to determine an estimation error covariance matrix, and a Kalman gain is updated based on the estimation error covariance matrix. In the update process, a Gaussian distribution is established as a proposal distribution based on the state variable and the estimation error covariance matrix, and a new particle set is obtained by resampling from the proposal distribution. The model update module is configured to establish an adaptive adjustment model of a noise covariance matrix based on a Riccati equation according to a system noise, the estimation error covariance matrix, the positive scalar parameter, and a set observation noise covariance matrix, and to obtain an equivalent system noise covariance matrix by solving the adaptive adjustment model. A weighted sum of the equivalent system noise covariance matrix and the system noise covariance matrix is taken as a corrected system noise covariance matrix, and a weighted sum of the set observation noise covariance matrix and an observation noise covariance matrix is taken as a corrected observation noise covariance matrix. The state estimation module is configured to obtain a state variable estimation value by iteratively solving based on the updated dynamic estimation model of the generator using the unscented particle filter. A weighted mean of the state variable estimation value and a weight thereof is taken as an electromechanical transient estimation result of the generator.

9. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is configured to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method of any one of claims 1-7. The program is executed by the processor to implement the steps of the method of any one of claims 1-7. ​ 10. A computer-readable storage medium having stored thereon a computer program, characterized in that, ​

Images