A reactive power allocation strategy based on sample entropy improved synge geometric mode decomposition

Abstract

The present application belongs to the field of wind power generation technology, aiming at the problem of voltage instability of wind farm grid-connected point caused by wind power uncertainty and grid-connected fault, a reactive power distribution strategy based on sample entropy improved symplectic geometry modal decomposition is proposed. The strategy first decomposes the total reference reactive power of the grid-connected point through symplectic geometry modal decomposition to obtain symplectic geometry components, calculates the sample entropy and similarity threshold of each component, and reconstructs the high, medium and low frequency components. Then, according to the difference in response time of the rotor side converter of the doubly-fed wind power plant and the static synchronous compensator, combined with the coordination of the shunt capacitor, the reference power of the three is distributed according to the frequency band, and the reactive power fluctuation is coordinated and suppressed. Simulation verification shows that this method avoids the defects of traditional algorithms, maintains voltage stability while greatly reducing the configuration capacity of reactive power compensation devices, and has effectiveness and economy.

Description

Technical Field

[0001] This invention belongs to the field of wind power generation technology and proposes a reactive power allocation strategy based on sample entropy-improved symplectic geometric mode decomposition. Background Technology

[0002] With the continuous increase in wind power capacity, the random fluctuations in wind power pose a threat to the voltage stability of the local power grid. Doubly fed induction generators (DFIGs) possess PQ decoupling control capabilities, and wind farms using DFIGs can provide voltage support to the grid. However, the reactive power output of DFIGs inevitably increases the current flowing through the converter, reducing the reliability of the unit's operation. Simultaneously, it lowers the power factor and increases active power losses in the collection system and within the unit. In actual operation, DFIGs typically operate at constant power, using reactive power compensation equipment installed at the wind farm's grid connection point to compensate for the system's reactive power requirements. This neglects the reactive power regulation capability of the DFIG itself. To fully utilize the reactive power regulation capability of DFIGs and improve the economic efficiency of wind farm operation, it is necessary to study the coordinated control between DFIGs and reactive power compensation equipment.

[0003] This paper proposes a reactive power allocation strategy based on sample entropy-improved symplectic geometric mode decomposition (SGMD). First, the reactive power at the grid connection point is used as the total reference power. Then, SGMD is used to decompose the total reference power, obtaining symplectic geometric components and calculating the sample entropy and corresponding similarity thresholds for each SGC (Symplocant Collector). Based on this, high, medium, and low frequency components are reconstructed. Next, considering the different reactive power response times of DFIG (Depleted Flow Injector) and STATCOM (Statistical Capacitor), the symplectic geometric mode decomposition algorithm is used to reconstruct the required reactive power adjustment at the wind farm grid connection point, determining the reference power for DFIG, STATCOM, and parallel capacitors, and coordinating the three to jointly mitigate reactive power fluctuations in the wind farm. Finally, a Simulink simulation model of a wind farm in a certain region is established to verify the effectiveness of the proposed power coordination control strategy. Summary of the Invention

[0004] The technical problem this invention aims to solve is that the uncertainty of wind power and short-circuit faults in the grid-connected system can cause voltage instability at the wind farm's grid connection point. To address the issue of mitigating wind power output fluctuations, this paper proposes a reactive power allocation strategy based on a sample entropy-improved symplectic geometric mode decomposition (SGMD) method. The reactive power signal required by the grid connection point system is decomposed using the SGMD method. Based on the different reactive power response times of DFIG and STATCOM, and considering the need for parallel capacitors to cooperate with STATCOM for timely and stable reactive power replenishment, a power coordination control strategy based on real-time reactive power frequency division at the wind farm's grid connection point is proposed. It includes the following steps, performed sequentially:

[0005] 1. Decompose the total reference power using SGMD.

[0006] The core of SGMD is to use symplectic geometric similarity transformation to solve for the eigenvalues ​​of the Hamiltonian matrix and to use its corresponding eigenvectors to construct symplectic geometric component signals.

[0007] Step 1: Assume that the time series of any initial signal is represented as: x = (x1, x2, ..., x...). n ), where n is the signal length. According to Takens' embedding theorem, the initial time series x is extended and reconstructed into a multidimensional trajectory signal matrix X:

[0008]

[0009] In the formula: d is the embedding dimension; λ is the stretching factor; m = n - (d - 1)λ.

[0010] Autocorrelation analysis of the trajectory matrix X yields a covariance symmetric matrix A:

[0011]

[0012] Construct the Hamiltonian matrix M from the symmetric matrix A:

[0013]

[0014] The symplectic orthogonal matrix Q is constructed using the Householder transformation as follows:

[0015]

[0016] In the formula, C is an upper triangular matrix, c ij =0 (i>j+1); N=M2.

[0017] Let the eigenvalues ​​of the upper triangular matrix C be λ1, λ2, ..., λ d Then the eigenvalues ​​of the symmetric matrix A are:

[0018]

[0019] Therefore, the eigenvectors corresponding to the eigenvalues ​​of matrix A are Q. i (i=1,2,…,d), according to the Householder transformation theory, matrix A is reconstructed to obtain the trajectory matrix Z:

[0020]

[0021]

[0022] In the formula, Z i Let Z be the initial subcomponent matrix of dimension m*d. Here, Z is defined. i The middle element is zij , if m < d, then z ij * = zij; otherwise z ij * = z ji .

[0023] Step 2. For any initial sub-component matrix Z i (i = 1, 2, …, n), perform average diagonalization to convert it into a set of time series with length n. Therefore, the sum of d sets of time series with length n is the initial time series x. The average diagonalization formula is:

[0024]

[0025] In the formula, d* = min(m, d), m* = max(m, d), n = m + (d - 1)λ.

[0026] Step 3. Convert Z i into a time series Y with length n i (i = 1, 2, …, n). The initial signal x is the superposition of d independent components, that is, d symplectic geometric mode components (SGCs), which are expressed as:

[0027]

[0028] 2. Obtain the symplectic geometric components and calculate the sample entropy of each SGC and the corresponding similarity threshold, and reconstruct the high, medium, and low frequency components based on this

[0029] Step 1. For any time series S = {S1, …, S i , …, S N}, construct an m-dimensional embedding vector S m :

[0030]

[0031] Then the sample entropy of the vector S i m can be calculated by the following formula:

[0032]

[0033] In the formula, r is the similarity limit value.

[0034] When v takes a finite value, SE can be calculated according to the following formula, that is:

[0035]

[0036] In the formula, O m(r) , O m+1(r) are the probabilities of the SGC matching m points and m + 1 points under r respectively.

[0037] Step 2: Calculate the SE value of each component and compare it with the corresponding similarity threshold to reconstruct the high, medium, and low frequency fluctuation components. The similarity threshold calculation expression is:

[0038]

[0039] 3. Determine the reference power of DFIG, STATCOM, and parallel capacitor based on the reconstructed high, medium, and low frequency components:

[0040] Step 1: In a doubly-fed induction generator (DFIG) wind farm, both the rotor-side converter (RSC) and the static synchronous compensator (STATCOM) exhibit continuous and smooth dynamic regulation of thermal characteristics, generating inductive and capacitive reactive power. However, the response time of RSC is in the order of seconds, while that of STATCOM is in the order of milliseconds. Literature review indicates that the frequency response range of the rotor-side converter in a DFIG wind turbine generator within this response time is 0–17 Hz, while the STATCOM responds rapidly and can cover the entire frequency band. In this paper, the 0–17 Hz frequency band of the reactive power signal is defined as the low-frequency band, and the frequency band above 17 Hz is defined as the high-frequency band. These are used as reference power for the DFIG and STATCOM, respectively, in coordinated control with the parallel capacitors.

[0041] Step 2: In an actual power grid, the reactive power capacity of the rotor-side converter of a wind turbine is fixed, and the rotor-side converter is used to supplement the reactive power of the low-frequency components after reconfiguration. Therefore, it is only necessary to calculate the energy ratio of the low, medium, and high-frequency components in the total reactive power, and then determine the compensation capacity of the parallel capacitors and STATCOM based on the reactive power capacity of the rotor-side converter of the wind turbine.

[0042] 4. Coordinate the efforts of all three parties to mitigate reactive power fluctuations in wind farms:

[0043] Step 1: Select a domestic wind power plant with an installed capacity of 100MW (66 1.5MW wind turbines) to establish a simulation test. Conduct simulation verification for voltage drop disturbances and short-circuit faults. Perform frequency division reconstruction on the reactive power required by the system calculated in real time at the grid connection point of the doubly-fed wind farm, and determine the capacity of each reactive power compensation based on the reconstructed high, medium and low frequency components.

[0044] Step 2: The SE-improvedSGMD method is used to decompose the reactive power required by the system after voltage disturbance and short-circuit fault. The decomposed SGC sub-mode components are obtained, and a total of 13 SGC sub-mode components are obtained using the proposed method. Furthermore, the proposed sample entropy similarity is used to reconstruct the high, medium and low frequency fluctuation components from the 13 decomposed SGC components.

[0045] Step 3: Given the configured capacity of the doubly-fed induction generator (RFG) RSC, and based on the energy ratio calculation results, considering a margin for reactive power compensation, the final configured capacity of the STATCOM is 11 Mvar. Traditionally, the reactive power compensation device installed in a doubly-fed wind farm is approximately 35% of the installed capacity of the RSG. Therefore, according to traditional calculations, the required reactive power compensation device capacity in this paper would be 35 Mvar. Compared to the method proposed in this paper, the reactive power compensation device capacity is significantly reduced, saving costs and verifying the economic viability of the proposed method.

[0046] Step four: Configure the reactive power compensation device according to the calculated capacity. The resulting voltage curve after a system fault can be obtained. It can be seen that the voltage fluctuation amplitude is significantly reduced, and the system voltage tends to stabilize shortly after the disturbance, further verifying the effectiveness of the proposed method.

[0047] Beneficial effects

[0048] 1. The SE-improvedSGMD model algorithm proposed in this invention effectively avoids the mode aliasing problem in EMD and MEEMD algorithms, and also avoids the subjective influence of manually setting the number of mode decompositions on the results in VMD algorithm, thus having obvious advantages in signal decomposition.

[0049] 2. While considering the joint compensation of reactive power by RSC and STATCOM, this invention also fully considers the stability when parallel capacitors are used in conjunction with STATCOM to supplement reactive power.

[0050] 3. The numerical examples show that the reactive power allocation strategy based on SE-improvedSGMD in the invention can greatly reduce the configuration capacity of reactive power compensation devices while maintaining the stability of the system grid connection voltage, which demonstrates the effectiveness and practicality of the method. Attached Figure Description

[0051] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0052] Figure 1 This is an SGC submode distribution diagram of a reactive power allocation strategy based on a sample entropy-improved symplectic geometric mode decomposition method according to the present invention.

[0053] Figure 2 The figure shows the calculation results of sample entropy and threshold for the reactive power allocation strategy of the improved symplectic geometric mode decomposition method based on sample entropy according to the present invention.

[0054] Figure 3 This is a reconstruction diagram of the high, medium, and low frequency wave sub-components of a reactive power allocation strategy based on a sample entropy-improved symplectic geometric mode decomposition method according to the present invention.

[0055] Figure 4 This is a comparison of voltage curves before and after configuring a reactive power compensation device in a system that uses a reactive power allocation strategy based on sample entropy-improved symplectic geometric mode decomposition method according to the present invention, after a voltage drop fault.

[0056] Figure 5 This is a comparison of STATCOM and RSC reactive power regeneration for a reactive power allocation strategy based on a sample entropy-improved symplectic geometric mode decomposition method according to the present invention.

[0057] Figure 6 This is a comparison of voltage curves before and after configuring a reactive power compensation device in a system that employs a reactive power allocation strategy based on a sample entropy-improved symplectic geometric mode decomposition method according to the present invention, after a short-circuit fault. Detailed Implementation

[0058] To better understand the objectives, technical solutions, and advantages of the embodiments of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. All other embodiments that can be implemented by those skilled in the art based on the embodiments of the present invention without inventive effort should be considered to be included within the protection scope of the present invention. The present invention proposes a reactive power allocation strategy based on a sample entropy-improved symplectic geometric mode decomposition method, aiming to reconstruct high, medium, and low frequency components; then, based on the response time difference between the rotor-side converter and the static synchronous compensator in a doubly-fed wind farm, and combined with the cooperative effect of parallel capacitors, the reference power of the three components is allocated according to frequency bands to coordinate and smooth reactive power fluctuations.

[0059] This invention discloses a strategy for allocating unbalanced power in a system jointly controlled by multiple reactive power compensation devices based on a sample entropy-improved symplectic geometric mode decomposition (SE-improvedSGMD) method. Figures 1 to 6 As shown, it includes the following steps:

[0060] 1. The unbalanced power is decomposed using SGMD to obtain a series of symplectic geometric components.

[0061] like Figure 1 As shown, the unbalanced power after a system fault is decomposed into symplectic geometric mode components using the symplectic geometric mode decomposition method.

[0062] Suppose that the time series of any initial signal is represented as: x = (x1, x2, ..., xn), where n is the signal length. According to Takens' embedding theorem, the initial time series x can be extended and reconstructed into a multidimensional trajectory signal matrix X:

[0063]

[0064] In the formula: d is the embedding dimension; λ is the stretching factor; m = n - (d - 1)λ.

[0065] Perform autocorrelation analysis on the trajectory matrix X to obtain the covariance symmetric matrix A:

[0066]

[0067] Construct the Hamiltonian matrix M from the symmetric matrix A:

[0068]

[0069] Construct the symplectic orthogonal matrix Q through Householder transformation, which is:

[0070]

[0071] In the formula, C is an upper triangular matrix, and c ij = 0 (i > j + 1); N = M2.

[0072] Let the eigenvalues of the upper triangular matrix C be λ1, λ2,..., λ d , then the eigenvalues of the symmetric matrix A are:

[0073]

[0074] Therefore, the eigenvectors corresponding to the eigenvalues of matrix A are Q i (i = 1, 2,..., d). According to the Householder transformation theory, reconstruct matrix A to obtain the trajectory matrix Z:

[0075]

[0076]

[0077] In the formula, Z i is the m*d-dimensional initial sub-component matrix. Among them, define the elements in Z i as z ij , if m < d, then z ij * = zij; otherwise z ij * = z ji .

[0078] Perform average diagonalization on any initial sub-component matrix Z i (i = 1, 2,..., n) to transform it into a set of time series with a length of n. Therefore, the sum of d sets of time series with a length of n is the initial time series x. The average diagonalization formula is:

[0079]

[0080] In the formula, d* = min(m, d), m* = max(m, d), n = m + (d - 1)λ.

[0081] Z i Transform into a time series Y of length n i (i=1,2,…,n). The initial signal x is a superposition of d independent components, i.e., d symplectic geometric mode components (SGCs), expressed as:

[0082]

[0083] 2. Based on the sample entropy and similarity x threshold of each decomposed SGC, the decomposed sub-components are reconstructed into high-frequency, medium-frequency, and low-frequency components.

[0084] like Figure 2 As shown, the high, medium, and low frequency components are reconstructed based on the calculated sample entropy of each SGC and its similarity x threshold.

[0085] For any time series S = {S1, ..., S2} i ,…,S N Construct an m-dimensional embedding vector S. m :

[0086]

[0087] Then vector S i m The sample entropy can be calculated using the following formula:

[0088]

[0089] In the formula, r is the similarity limit.

[0090] When v takes finite values, SE can be calculated according to the following formula:

[0091]

[0092] In the formula, O m(r) O m+1(r) Let r be the probability that SGC matches m points and m+1 points respectively.

[0093] A signal is decomposed into multiple SGC components using SGMD. The SE value of each component is calculated and compared with the corresponding similarity threshold to reconstruct high, medium, and low frequency fluctuation components. The similarity threshold calculation expression is as follows:

[0094]

[0095] 3. Capacity Selection

[0096] In a real power grid, the reactive power capacity of the rotor-side converter of a wind turbine is fixed, and the rotor-side converter is used to supplement the low-frequency reactive power after reconfiguration. Therefore, it is only necessary to calculate the energy ratio of low, medium, and high-frequency components in the total reactive power to determine the compensation capacity of the parallel capacitors and STATCOM based on the reactive power capacity of the rotor-side converter of the wind turbine.

[0097] 4. Simulation Experiment of Doubly Fed Wind Farm

[0098] This invention selects a 100MW wind power plant (66 1.5MW wind turbines) in China to establish a simulation test. Simulation verification is conducted for voltage dip disturbances and short-circuit faults. The reactive power required by the system, calculated in real time at the grid connection point of the doubly-fed wind farm, is reconstructed by frequency division, and the capacity of each reactive power compensation component is determined based on the reconstructed high, medium, and low frequency components.

[0099] The wind speed was set to 9 m / s, and the control objective was to maintain the bus voltage at 1.0 pu. Voltage dips and three-phase short-circuit faults were introduced at 2.0 s. The voltage dip fault lasted until the end of the simulation; the short-circuit fault recovered after 0.2 s. The voltage curve at the grid connection point after the fault is shown below. Figure 3 Show.

[0100] The SE-improvedSGMD method is used to decompose and reconstruct the reactive power required by the system after voltage disturbance. The capacity of each reactive power compensation device is calculated based on the reconstructed high, medium, and low frequency components. Simulation experiments are then conducted based on the calculated capacity.

[0101] Figure 4 Voltage curve after a voltage dip fault occurs in the system after configuring a reactive power compensation device

[0102] Figure 5 The reactive power supplemented by each reactive power compensation device at this time.

[0103] It can be seen that the voltage fluctuation amplitude is significantly reduced, and the system voltage tends to stabilize shortly after the disturbance, further verifying the effectiveness of the method proposed in this invention.

[0104] This method was used to conduct simulation tests on short-circuit faults.

[0105] Figure 6 Voltage curve of the system after a short circuit fault occurs after the reactive power compensation device is configured.

[0106] The calculation conditions, illustrations, etc. in the embodiments of this invention are only used to further illustrate the invention and are not exhaustive. They do not constitute a limitation on the scope of protection of the claims. Those skilled in the art, based on the inspiration gained from the examples of this invention, can conceive of other substantially equivalent alternatives without inventive effort, all of which are within the scope of protection of this invention.

Claims

1. A reactive power allocation strategy based on sample entropy-improved symplectic geometric mode decomposition, characterized in that, Includes the following steps: The total reference power is decomposed using SGMD, or symplectic geometric mode decomposition method. Obtain the symplectic geometric components and calculate the sample entropy and corresponding similarity threshold of each SGC (symplectic geometric modal component), and reconstruct the high, medium and low frequency components based on this. The reference power of DFIG (doubly fed induction generator) and STATCOM (static synchronous compensator and parallel capacitor) is determined based on the reconstructed high, medium and low frequency components. The three parties work together to mitigate reactive power fluctuations in wind farms.

2. The reactive power allocation strategy based on sample entropy-improved symplectic geometric mode decomposition according to claim 1, characterized in that, The SGMD method utilizes symplectic geometric similarity transformation to solve for the eigenvalues ​​of the Hamiltonian matrix and constructs symplectic geometric component signals using the corresponding eigenvectors of the Hamiltonian matrix. The decomposition of the total reference power using SGMD includes the following steps: Step 1: Assume that the time series of any initial signal is represented as: x = (x1, x2, ..., x...). n ), where n is the signal length; according to Takens' embedding theorem, the initial time series x is extended and reconstructed into a multidimensional trajectory signal matrix X: In the formula: d is the embedding dimension; λ is the stretching factor; m = n - (d - 1)λ; Autocorrelation analysis of the trajectory matrix X yields a covariance symmetric matrix A: A=X T X Construct the Hamiltonian matrix M from the symmetric matrix A: The symplectic orthogonal matrix Q is constructed using the Householder transformation as follows: In the formula, C is an upper triangular matrix, c ij =0, i>j+1; N=M2; Let the eigenvalues ​​of the upper triangular matrix C be λ1, λ2, ..., λ d Then the eigenvalues ​​of the symmetric matrix A are: Therefore, the eigenvectors corresponding to the eigenvalues ​​of matrix A are Q. i Let i = 1, 2, ..., d. Based on the Householder transformation theory, matrix A is reconstructed to obtain the trajectory matrix Z: Z i =Q i Q i T X i Z=Z1+Z2+…+Z d In the formula, Z i The initial subcomponent matrix is ​​m*d dimensional; Among them, define Z i The element in it is z ij , if m < d, then z ij * = zij; otherwise z ij * = z ji ; Step 2: For the initial subcomponent matrix Z i For i = 1, 2, ..., n, perform average diagonalization to make the initial subcomponent matrix Z i The time series is transformed into a set of time series of length n; the sum of the time series of length n is the initial time series x; the average diagonalization formula is: In the formula, d*=min(m, d), m*=max(m, d), n=m+(d-1)λ; Step 3: Convert the initial subcomponent matrix Z i Transform into a time series Y of length n i (i = 1, 2, ..., n); The initial signal x is a superposition of d independent components, i.e., d symplectic geometric modal components, expressed as: x=Y1+Y2+…+Y d 。 3. The reactive power allocation strategy based on sample entropy-improved symplectic geometric mode decomposition according to claim 1, characterized in that, The process of obtaining symplectic geometric components and calculating the sample entropy and corresponding similarity threshold of each SGC, and reconstructing high, medium, and low frequency components based on this, includes the following steps: Step 1: For any time series S = {S1, ..., S2} i ,…,S N Construct an m-dimensional embedding vector S. m : Then vector S i m The sample entropy can be calculated using the following formula: In the formula, r is the similarity limit; When v takes finite values, SE can be calculated according to the following formula: In the formula, O m(r) O m+1(r) Let r be the probabilities of SGC matching m points and m+1 points respectively; Step 2: Calculate the SE value of each component and compare it with the corresponding similarity threshold to reconstruct the high, medium, and low frequency fluctuation components; the similarity threshold calculation expression is:

4. The reactive power allocation strategy based on sample entropy-improved symplectic geometric mode decomposition according to claim 1, characterized in that, The determination of the reference power of DFIG, STATCOM and parallel capacitor based on the reconstructed high, medium and low frequency components includes the following steps: Step 1: The thermal regulation of the rotor-side converter RSC and static synchronous compensator STATCOM in the doubly fed wind farm is continuously and smoothly dynamically regulated, generating inductive and capacitive reactive power. The 0-17Hz frequency band of the reactive power signal is defined as the low frequency band, and the frequency band above 17Hz is defined as the high frequency band. These are used as reference power for the DFIG and STATCOM and the parallel capacitor to participate in coordinated control. Step 2: Calculate the energy ratio of low, medium and high frequency components in the total reactive power, and determine the compensation capacity of the parallel capacitor and STATCOM based on the reactive capacity of the rotor-side converter of the wind turbine.

5. The reactive power allocation strategy based on sample entropy-improved symplectic geometric mode decomposition according to claim 1, characterized in that, The coordination among the three parties to smooth out reactive power fluctuations in wind farms includes the following steps: Step 1: Select a wind farm to establish a simulation test; conduct simulation verification for voltage drop disturbance and short circuit fault conditions, perform frequency division reconstruction on the reactive power required by the system calculated in real time at the grid connection point of the doubly fed wind farm, and determine the capacity of each reactive power compensation based on the reconstructed high, medium and low frequency components; Step 2: Use the SE-improvedSGMD method to decompose the reactive power required by the system after voltage disturbance and short circuit fault to obtain each SGC submode component; Step 3: Given the configured capacity of the doubly fed wind turbine RSC, the configured capacity of STATCOM is finally obtained based on the energy ratio calculation results. Step 4: Configure the capacity of the reactive power compensation device to obtain the voltage curve of the system after a fault following the configuration and reactive power compensation.

Images