Converter interface power supply time-varying virtual inertia robust estimation method based on maximum entropy criterion chaotic polynomial Kalman filtering

Through the chaotic polynomial Kalman filtering method based on the maximum entropy criterion, the problems of low inertia estimation accuracy and poor robustness in the power system are solved, and the rapid and accurate tracking and estimation of time-varying virtual inertia of asynchronous power is achieved, which improves the system noise resistance.

CN118937791APending Publication Date: 2024-11-12SOUTHEAST UNIV +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410933301.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing inertia estimation methods of power systems have problems such as large data demand, poor robustness, low accuracy in normal and small disturbance states, and the inability to reflect the contribution of diverse virtual inertia resources to system inertia, especially in asynchronous power supplies. Tracking and estimation of time-varying inertia is difficult to accurately perform.

Method used

The chaotic polynomial Kalman filtering method based on the maximum entropy criterion is adopted to predict and correct the system state through discrete state space model and fourth-order Longkuta method, and the state and noise covariance matrix are updated in combination with the maximum entropy criterion to achieve real-time estimation and tracking. Time-varying virtual inertia of the power supply of the inverter interface.

Benefits of technology

It realizes rapid and accurate estimation and tracking of time-varying virtual inertia of the converter interface power supply, with high noise immunity, and is suitable for real-time operation environments of asynchronous power supply.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118937791A_ABST
    Figure CN118937791A_ABST
Patent Text Reader

Abstract

The invention relates to a converter interface power supply time-varying virtual inertia robust estimation method based on maximum entropy criterion chaos polynomial Kalman filtering, and the method comprises the following steps: A, constructing a discrete state space model of a power system, and carrying out the discretization through employing a fourth-order Runge-Kutta method, and obtaining the discrete state space model of the power system; b, taking the system state empirical value and the process noise covariance matrix empirical value as input of a state estimator, and calculating a current system state and a covariance matrix predicted value according to a system state equation and a system state estimated value at a previous moment; and C, updating the filtering gain, the system state variable estimation value and the covariance matrix prediction value according to the updated system measurement noise and the covariance matrix prediction value. And D, repeating the steps B and C to obtain the real-time virtual inertia estimation value of the converter interface power supply. According to the method, the virtual inertia can be estimated online, and the time-varying virtual inertia can be tracked; process and measurement noise statistical information are not needed, change tracking is fast, estimation precision is high, and anti-noise performance is high.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method for robustly estimating time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering, and belongs to the field of frequency stability analysis and control of power systems. Background Art

[0002] With the continuous depletion of fossil energy and increasingly stringent carbon emission requirements, the power system is gradually changing from being dominated by conventional synchronous power sources to being dominated by non-synchronous power sources such as wind power and solar energy. In the power system, inertia plays a vital role. Inertia can provide the system with anti-disturbance and recovery capabilities, and has a key impact on maintaining the system frequency within a stable range. Traditional synchronous generators provide inertia support through their rotational kinetic energy, allowing the system to operate stably in the face of external disturbances. However, non-synchronous power sources such as wind power and photovoltaics lack rotational kinetic energy and therefore cannot directly provide the same level of inertia support. In the future, the number and capacity of non-synchronous power sources such as wind power and photovoltaics connected through the power electronic converter interface will continue to grow rapidly. Non-synchronous power sources are decoupled from the grid frequency and cannot actively provide inertia support, which will lead to a decrease in the inertia level of the power system, thereby affecting the system's anti-disturbance ability and stability characteristics. In recent years, system stability problems caused by insufficient inertia levels have been common. Therefore, accurately estimating the system's inertia level and its changing trend is of great significance for improving the load limit of non-synchronous power sources and ensuring stable operation of the system under low inertia conditions.

[0003] Power system inertia estimation has become a hot topic in current research, and many domestic and foreign research institutions have made relevant research progress. Inertia estimation can be divided into two methods: offline estimation and online estimation. The main difference is whether it is instantaneous. Offline estimation refers to the use of the system equivalent swing equation to calculate the system inertia after a disturbance event occurs. Offline estimation is more dependent on large disturbances, so it is often used to study and analyze inertia and summarize experience when an accident occurs. By reviewing and analyzing historical data, the inertia characteristics of the system can be estimated. With the popularization and improvement of power grid wide-area measurement systems, online estimation of power system inertia has gradually become a reality. Compared with offline estimation, online estimation has a slight advantage in real-time performance. Online estimation can estimate and track the inertia of the system through real-time measurement data and advanced algorithms. By obtaining the latest data and information in a timely manner, the changes and dynamic characteristics of the system inertia can be more accurately reflected. The online estimation method can combine real-time measurement data and advanced signal processing technology to obtain real-time information on the system status using existing sensors and measurement equipment. This information can be used to establish mathematical models and algorithms to estimate the inertia of the system. By continuously updating and optimizing the algorithm, the accuracy and robustness of the online estimation can be improved, making it more suitable for the actual power system operating environment. However, there are some problems with the existing inertia estimation methods, such as large data requirements, poor robustness, low accuracy in normal and small disturbance states, and inability to reflect the contribution of diverse virtual inertia resources to the system inertia. In order to accurately and quickly estimate the inertia within a synchronous generator or regional system, some methods propose to use the measurement data of the phasor measurement unit. In addition, in order to alleviate the frequency instability caused by the converter interface power supply, some converters are equipped with inertia simulation control, namely virtual synchronous control, which may introduce time-varying inertia. However, the dynamic estimators currently developed for tracking the time-varying inertia of VSGs have the problem of numerical oscillation. In addition, there are studies that use environmental measurements to estimate the inertia constant of non-synchronous generators, but this method assumes knowledge of the generator rotor speed and angle, which is difficult to obtain in practice. In addition, due to the changes in control parameters, this method cannot track the time-varying inertia of the converter interface power supply. Summary of the invention

[0004] The present invention provides a method for robustly estimating the time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering. The method can estimate the time-varying virtual inertia of the converter interface power supply in real time, has fast change tracking, high estimation accuracy and strong anti-noise performance.

[0005] The present invention is implemented by the following technical solution: a method for robustly estimating time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering, the method comprising the following steps:

[0006] A. Construction of discrete state space model of power system

[0007] The synchronous machine in the power system adopts the classic second-order model:

[0008]

[0009]

[0010] Where: δ, ω and ω0 are the rotor angle, speed and rated speed of the generator respectively; H is the inertia constant of the generator; P m and P e are the mechanical power and electromagnetic power of the generator respectively; D is the damping coefficient.

[0011] The inverter interface power supply adopts the grid-following model:

[0012]

[0013]

[0014] θ PLL =ω PLL +ω0

[0015]

[0016] ω PLL =K p v q +K i x PLL

[0017] Where: P v and P vs are the active power of CIG and the active power reference value without virtual inertia control; K v is the proportional coefficient used for the virtual inertial control link, where K v =2H v , H v is the virtual inertia constant. is the introduced intermediate variable; T v and T p is the equivalent time constant; ω PLL is the measured value of the system frequency; θ is the bus voltage phase angle. Under ideal conditions, the phase-locked loop output θ PLL =θ; x PLL is the introduced intermediate variable; v q is the q-axis voltage; K p and K i They are the proportional coefficient and integral coefficient of PLL respectively.

[0018] Choose δ, ω, θ PLL, P v 、x v 、x PLL , H v As the system state quantity, select P e , Q e 、V mag 、V ang , P v For system quantity measurement, a continuous state space model of the power system is established; the fourth-order Runge-Kutta method is used for discretization to obtain a discrete state space model of the power system

[0019] x k =f(x k-1 )+w k

[0020] z k =h(x k )+v k

[0021] B. System status prediction

[0022] First, the configuration points CPs are calculated as the coefficients of the chaotic polynomial. After PCE dimensionality reduction, the proposed PCKF only needs 2N+1 polynomial chaotic coefficients, that is, only 2N+1 suitable CPs combinations are needed. For this purpose, a (N)×(2N+1) matrix can be used:

[0023]

[0024] Where: The i-th column of Ξ represents the random variable vector ξ i , represents the i-th combination of CPs.

[0025] Substitute the collocation points into the matrix get:

[0026]

[0027]

[0028] The empirical value of the system state and the empirical value of the process noise covariance matrix are used as the input of the state estimator, and the current system state and covariance matrix prediction value are calculated according to the system state equation and the estimated value of the system state at the previous moment.

[0029]

[0030]

[0031]

[0032] Where: w is the standard deviation of process noise; is the estimated value of the state variable at time k-1; Predict the state vector for the i-th CP. is the state prediction matrix obtained by aggregation.

[0033] according to We can get:

[0034]

[0035] Define A1 as a matrix Transpose the first row and you get:

[0036]

[0037] Will The remaining matrix is ​​defined as A2, and the state variable estimation error covariance matrix is ​​calculated as:

[0038]

[0039] Where: Q is the process noise.

[0040] C. System measurement correction

[0041] The current system state estimate is calculated based on the system measurement equation, the current system measurement value and the system state prediction value.

[0042] C1. Measurement error covariance and cross-covariance matrix update:

[0043] Use the measurement equation to propagate the configuration points and calculate the measurement prediction value:

[0044]

[0045]

[0046] Where: is the measurement prediction vector calculated for the i-th CP, is the aggregated measurement prediction matrix.

[0047] The PC coefficient matrix of the measurement model is:

[0048]

[0049] Take the first row of matrix W and transpose W1 to get the measured prediction value

[0050]

[0051]

[0052]

[0053] C2. System state quantity and its covariance matrix update

[0054] The state equation and measurement equation considering the maximum entropy criterion can be written as:

[0055]

[0056] Where the measurement matrix for:

[0057]

[0058]

[0059] Multiply both sides of the equation by S k -1 have to

[0060] D k =W k x k +e k

[0061] Where:

[0062]

[0063]

[0064]

[0065] Defined by the maximum entropy formula Available

[0066]

[0067]

[0068] Where e l,k Yes k The kth element of .

[0069] The state covariance matrix prediction value, measurement noise covariance matrix and gain estimation value are updated to obtain:

[0070]

[0071]

[0072]

[0073] The system state quantity and its covariance matrix estimate are calculated based on the state covariance matrix prediction value, the measurement noise covariance matrix and the gain estimate value:

[0074]

[0075]

[0076] C3. System status output

[0077] The iterative update of the system state estimation value is completed, and the estimated value of the time-varying virtual inertia of the current converter interface power supply is output.

[0078] D. Repeat steps B and C to obtain the real-time virtual inertia estimation value of the converter interface power supply. The scheme adopts the maximum entropy criterion and the chaotic polynomial Kalman filter, can estimate the virtual inertia online, and can track the time-varying virtual inertia; the present invention does not require process and measurement noise statistical information, has fast change tracking, high estimation accuracy, and strong noise resistance.

[0079] A device for robustly estimating time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering, comprising:

[0080] Memory for storing computer programs;

[0081] A processor is used to execute the computer program to implement the steps of the method for robustly estimating the time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering.

[0082] A computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of a method for robustly estimating time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering.

[0083] Compared with the prior art, the present invention has the following beneficial effects:

[0084] Compared with the traditional method for estimating the virtual inertia of the converter interface power supply, the present invention adopts a robust estimation method for the time-varying virtual inertia of the converter interface power supply based on the maximum entropy criterion chaotic polynomial Kalman filter, and verifies through simulation analysis and comparison that this method can estimate the time-varying virtual inertia of the converter interface power supply in real time, with fast change tracking, high estimation accuracy and strong anti-noise performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 Schematic diagram of the inertia estimation curve of the simulation experiment of the traditional unscented Kalman filter method (UKF);

[0086] Figure 2 Schematic diagram of inertia estimation curve of simulation experiment of this method. DETAILED DESCRIPTION

[0087] Example: See Figure 1 , Figure 2 ,

[0088] The present invention provides a method for robustly estimating time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering, and the steps are as follows:

[0089] A. Construction of discrete state space model of power system

[0090] The synchronous machine in the power system adopts the classic second-order model:

[0091]

[0092]

[0093] Where: δ, ω and ω0 are the rotor angle, speed and rated speed of the generator respectively; H is the inertia constant of the generator; P m and P e are the mechanical power and electromagnetic power of the generator respectively; D is the damping coefficient.

[0094] The inverter interface power supply adopts the grid-following model:

[0095]

[0096]

[0097] θ PLL =ω PLL +ω0

[0098]

[0099] ω PLL =K p v q +K i x PLL

[0100] Where: P v and P vs are the active power of CIG and the active power reference value without virtual inertia control; K v is the proportional coefficient used for the virtual inertial control link, where K v =2H v , H v is the virtual inertia constant. is the introduced intermediate variable; T v and T p is the equivalent time constant; ω PLL is the measured value of the system frequency; θ is the bus voltage phase angle. Under ideal conditions, the phase-locked loop output θ PLL =θ; xPLL is the introduced intermediate variable; v q is the q-axis voltage; K p and K i They are the proportional coefficient and integral coefficient of PLL respectively.

[0101] Choose δ, ω, θ PLL , P v 、x v 、x PLL , H v As the system state quantity, select P e , Q e 、V mag 、V ang , P v For system quantity measurement, a continuous state space model of the power system is established; the fourth-order Runge-Kutta method is used for discretization to obtain a discrete state space model of the power system

[0102] x k =f(x k-1 )+w k

[0103] z k =h(x k )+v k

[0104] B. System status prediction

[0105] First, the configuration points CPs are calculated as the coefficients of the chaotic polynomial. After PCE dimensionality reduction, the proposed PCKF only needs 2N+1 polynomial chaotic coefficients, that is, only 2N+1 suitable CPs combinations are needed. For this purpose, a (N)×(2N+1) matrix can be used:

[0106]

[0107] Where: The i-th column of Ξ represents the random variable vector ξ i , represents the i-th combination of CPs.

[0108] Substitute the collocation points into the matrix get:

[0109]

[0110]

[0111] The empirical value of the system state and the empirical value of the process noise covariance matrix are used as the input of the state estimator, and the current system state and covariance matrix prediction value are calculated according to the system state equation and the estimated value of the system state at the previous moment.

[0112]

[0113]

[0114]

[0115] Where: w is the standard deviation of process noise; is the estimated value of the state variable at time k-1; Predict the state vector for the i-th CP. is the state prediction matrix obtained by aggregation.

[0116] according to We can get:

[0117]

[0118] Define A1 as a matrix Transpose the first row and you get:

[0119]

[0120] Will The remaining matrix is ​​defined as A2, and the state variable estimation error covariance matrix is ​​calculated as:

[0121]

[0122] Where: Q is the process noise.

[0123] C. System measurement correction

[0124] The current system state estimate is calculated based on the system measurement equation, the current system measurement value and the system state prediction value.

[0125] C1. Measurement error covariance and cross-covariance matrix update:

[0126] Use the measurement equation to propagate the configuration points and calculate the measurement prediction value:

[0127]

[0128]

[0129] Where: is the measurement prediction vector calculated for the i-th CP, is the aggregated measurement prediction matrix.

[0130] The PC coefficient matrix of the measurement model is:

[0131]

[0132] Take the first row of matrix W and transpose W1 to get the measured prediction value

[0133]

[0134]

[0135]

[0136] C2. System state quantity and its covariance matrix update

[0137] The state equation and measurement equation considering the maximum entropy criterion can be written as:

[0138]

[0139] Where the measurement matrix for:

[0140]

[0141]

[0142] Multiply both sides of the equation by S k -1 have to

[0143] D k =W k x k +e k

[0144] Where:

[0145]

[0146]

[0147]

[0148] Defined by the maximum entropy formula Available

[0149]

[0150]

[0151] Where e l,k Yes k The kth element of .

[0152] The state covariance matrix prediction value, measurement noise covariance matrix and gain estimation value are updated to obtain:

[0153]

[0154]

[0155]

[0156] The system state quantity and its covariance matrix estimate are calculated based on the state covariance matrix prediction value, the measurement noise covariance matrix and the gain estimate value:

[0157]

[0158]

[0159] C3. System status output

[0160] The iterative update of the system state estimation value is completed, and the estimated value of the time-varying virtual inertia of the current converter interface power supply is output.

[0161] D. Repeat steps B and C to obtain the real-time virtual inertia estimation value of the converter interface power supply.

[0162] Simulation experiment:

[0163] In order to verify the effectiveness of the present invention, a simulation experiment was carried out and compared with the traditional unscented Kalman filtering method.

[0164] The simulation experiment adopts a modified IEEE-39 node system, in which there are 10 synchronous power supplies and 2 asynchronous power supplies, which are installed at node 5 and node 9 respectively, and the capacity is 200MW. During the simulation process, a line break fault is applied to nodes 15 and 16. According to the above experimental conditions, the method of the present invention and the traditional unscented Kalman filter method are used to estimate the inertia of the power system. Figure 1 It is the simulation experiment inertia estimation curve of the method of the present invention and the traditional unscented Kalman filtering method when there are abnormal values ​​in the noise.

[0165] from Figure 1 It can be seen that (a) UKF inertia estimation result, (b) this method inertia estimation result,

[0166] When the system is disturbed, the system virtual inertia changes. When the noise measurement value shows an abnormal value, the estimation result has low accuracy. The method of the present invention can estimate the time-varying virtual inertia of the converter interface power supply in real time, with fast change tracking, high estimation accuracy and strong noise resistance.

[0167] It should be noted that the above embodiments are not intended to limit the protection scope of the present invention, and equivalent changes or substitutions made on the basis of the above technical solutions all fall within the protection scope of the claims of the present invention.

Claims

1. A method for robust estimation of time-varying virtual inertia of converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filter, characterized in that: The method comprises the following steps: A. Construction of discrete state space model of power system, B. System status prediction, C. System measurement correction, D. Repeat steps B and C to obtain the real-time virtual inertia estimation value of the converter interface power supply.

2. The method for robustly estimating time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering according to claim 1 is characterized in that: Step A: construct the discrete state space model of the power system, as follows: The synchronous machine in the power system adopts the classic second-order model: Where: δ, ω and ω0 are the rotor angle, speed and rated speed of the generator respectively; H is the inertia constant of the generator; P m and P e are the mechanical power and electromagnetic power of the generator respectively; D is the damping coefficient, The inverter interface power supply adopts the grid-following model: i PLL =ω PLL +ω0 ω PLL =K p v q +K i x PLL Where: P v and P vs are the active power of CIG and the active power reference value without virtual inertia control; K v is the proportional coefficient used for the virtual inertial control link, where K v =2H v , H v is the virtual inertia constant, is the introduced intermediate variable; T v and T p is the equivalent time constant; ω PLL is the measured value of the system frequency, θ is the bus voltage phase angle, and under ideal conditions, the phase-locked loop output θ PLL =θ; x PLL is the introduced intermediate variable; v q is the q-axis voltage; K p and K i are the proportional coefficient and integral coefficient of PLL respectively, Choose δ, ω, θ PLL , P v 、x v 、x PLL , H v As the system state quantity, select P e , Q e 、V mag 、V ang , P v For system quantity measurement, a continuous state space model of the power system is established; the fourth-order Runge-Kutta method is used for discretization to obtain a discrete state space model of the power system. x k =f(x k-1 )+w k z k =h(x k )+v k , Where f(x) is the state space variable function; h(x k ) is the measured variable function; w k and v k All are zero-mean Gaussian noise.

3. The method for robust estimation of time-varying virtual inertia of converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filter according to claim 2 is characterized in that: Step B: System status prediction, as follows: First, the configuration points CPs are calculated as the coefficients of the chaotic polynomial. After PCE dimension reduction, the proposed PCKF only needs 2N+1 polynomial chaotic coefficients, that is, only 2N+1 suitable CPs combinations are needed. For this purpose, a (N)×(2N+1) matrix is ​​used: Where: The i-th column of Ξ represents the random variable vector ξ i , represents the i-th combination of CPs, Substitute the collocation points into the matrix get: The empirical value of the system state and the empirical value of the process noise covariance matrix are used as the input of the state estimator, and the current system state and covariance matrix prediction value are calculated according to the system state equation and the estimated value of the system state at the previous moment. Where: w is the standard deviation of process noise; is the estimated value of the state variable at time k-1; is the predicted state vector of the i-th CP, is the aggregated state prediction matrix, according to We can get: Define A1 as a matrix Transpose the first row and you get: Will The remaining matrix is ​​defined as A2, and the state variable estimation error covariance matrix is ​​calculated as: Where: Q is the process noise.

4. The method for robust estimation of time-varying virtual inertia of converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filter according to claim 3 is characterized in that: Step C: System measurement and correction, as follows: Calculate the current system state estimate based on the system measurement equation, the current system measurement value and the system state prediction value. C1. Measurement error covariance and cross-covariance matrix update: Use the measurement equation to propagate the configuration points and calculate the measurement prediction value: Where: is the measurement prediction vector calculated for the i-th CP, is the aggregated measurement prediction matrix, The PC coefficient matrix of the measurement model is: Take the first row of matrix W and transpose W1 to get the measured prediction value In the formula, and are all error covariance matrices; R is zero-mean Gaussian white noise, C2, system state quantity and its covariance matrix update, The state equation and measurement equation considering the maximum entropy criterion can be written as: Where the measurement matrix for: Multiply both sides of the equation by S k -1 have to D k =W k x k +e k Where: Defined by the maximum entropy formula Available Where e l,k Yes k The kth element of The state covariance matrix prediction value, measurement noise covariance matrix and gain estimation value are updated to obtain: The system state quantity and its covariance matrix estimate are calculated based on the state covariance matrix prediction value, the measurement noise covariance matrix and the gain estimate value: C3, system status output, The iterative update of the system state estimation value is completed, and the estimated value of the time-varying virtual inertia of the current converter interface power supply is output.

5. A device for robust estimation of time-varying virtual inertia of converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filter, characterized in that: include: Memory for storing computer programs; A processor is used to execute the computer program to implement the steps of a method for robustly estimating time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering as described in any one of claims 1 to 4.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which is executed by a processor to implement the steps of a method for robustly estimating time-varying virtual inertia of a converter interface power supply based on maximum entropy criterion chaotic polynomial Kalman filtering as described in any one of claims 1 to 4.