Response data and model fused wind power plant grid-connected system oscillation parameter identification method

Abstract

The invention belongs to the field of power system stability analysis, and discloses a response data and model fused subsynchronous oscillation parameter identification method for a wind power plant grid-connected system. Firstly, the current of a grid-connected line is monitored, the energy is calculated, and an early warning is given out when the energy is continuously increased along with time. Then, self-adaptive VMD decomposition is carried out on the early warning signal, and the system state is preliminarily determined; then, proper band-pass filter parameters are set according to the decomposition result, and filtering operation is carried out on the signals. And solving a complete SSO current component from the filtering signal according to a derived mathematical relationship between related electrical quantities and parameters of the series compensation power transmission line, and further obtaining an oscillation frequency and a damping factor. And finally, carrying out a simulation test based on series compensation capacitor grid-connected data of the direct-driven wind power plant. The oscillation identification method provided by the invention has strong noise immunity and robustness, and can realize accurate and rapid identification of the dominant SSO mode.

Description

Technical Field

[0001] The present invention relates to the field of power system stability analysis, and in particular to a method for identifying oscillation parameters of a wind farm grid-connected system by combining response data and model data. Background Art

[0002] In order to achieve the goal of "carbon peak and carbon neutrality" as soon as possible, my country has actively promoted energy transformation and vigorously developed renewable energy. Wind power generation has developed rapidly both at home and abroad due to its unique advantages. Limited by the inverse distribution of my country's abundant wind energy resources and load centers, large-capacity wind power needs to be transmitted over long distances through technologies such as series compensation lines or high-voltage direct current transmission. The subsynchronous oscillation problem caused by its grid connection seriously restricts the development of wind power. SSO is mainly manifested as rapid divergence or continuous oscillation of power, current or voltage. Its hazards accumulate linearly over time. If it is not discovered in time and measures are not taken, it may threaten the safe and stable operation of the system and even damage the equipment. Therefore, there is an urgent need for a fast parameter identification method to realize effective monitoring and evaluation of the system status, which has certain engineering application value for ensuring the safe and stable operation of the power system.

[0003] Doubly-fed wind turbines and direct-drive wind turbines are the mainstream models of wind power generation. The methods for their oscillation parameter identification and stability analysis are mainly divided into two categories. One is based on mathematical models, and commonly used methods include impedance analysis, eigenvalue analysis, open-loop resonance, etc. These methods rely on accurate system models and data, but the "double high" characteristics of the current power system are becoming more and more obvious, the operation status of the power grid is complex and changeable, and its mathematical model presents complex characteristics such as high dimension and nonlinearity; and it is difficult to obtain detailed data of wind turbines, and it is quite difficult to establish an accurate research model. The limitations of model-based methods are becoming more and more obvious. The second is based on data-driven methods, mainly including fast Fourier transform, Prony analysis, wavelet transform, Hilbert-Huang transform (HHT), etc. In order to improve the accuracy of oscillation identification, related studies mostly improve and integrate multiple algorithms, or pre-process data to achieve noise reduction or avoid modal aliasing. In addition, the continuous development of machine learning, artificial intelligence and data mining capabilities has provided new ideas for oscillation research. However, most studies are based on training and learning based on simulation model data, and their applicability in practical applications remains to be verified. Therefore, it is necessary to combine models and data-driven to study oscillations.

[0004] On the basis of the above research, the present invention proposes an oscillation parameter identification method combining response data with line model. By performing adaptive VMD decomposition on data with oscillation risk, it is preliminarily determined whether SSO occurs in the signal and the distribution of oscillation frequency is clarified. Then, based on the decomposition results, a low-order digital bandpass filter is designed to filter the signal, retaining only the dominant subsynchronous oscillation component and the power frequency component. Then, based on the relationship between the derived line parameters, the complete subsynchronous current component is solved to achieve rapid and accurate identification of oscillation-related parameters. Summary of the invention

[0005] With the continuous development of SSO-specific detection equipment based on waveform data, this paper proposes an oscillation parameter identification method for wind farm series-compensated grid-connected system based on timestamp waveform data.

[0006] The present invention is achieved by adopting the following technical solutions:

[0007] First, the grid-connected line current is monitored. When its energy is detected to be increasing over time, the data is immediately subjected to adaptive VMD decomposition to determine whether SSO occurs in the system and obtain the distribution of the oscillation frequency. Then, a low-order digital filter is used to filter out other modes, retaining only the subsynchronous oscillation mode and the power frequency mode. Then, the relationship between the line parameters is derived according to the formula, and the system oscillation parameters are obtained by solution.

[0008] As a preferred embodiment of the present invention, the scheme comprises the following steps:

[0009] Step A1: For a series compensation transmission line containing capacitors, a time stamp waveform measurement unit is used to obtain a line current waveform and a voltage waveform at both ends;

[0010] Step A2: Monitor the current waveform in real time and calculate its RMS value for energy calculation. When it is detected that the energy continues to increase, it indicates that the system has an oscillation risk. It is necessary to further use the improved variational mode decomposition method to effectively decompose the signal to determine whether the signal has subsynchronous oscillation, the number of modal components contained and their frequency distribution.

[0011] Step A3: Calculate the voltage difference between the two ends of the line, and perform bandpass filtering on the current signal and the voltage difference signal to retain only the dominant subsynchronous oscillation component and the power frequency component;

[0012] Step A4: Based on the derived series compensation line voltage and current relationship formula, the complete sub-synchronous current component is calculated and separated, so that the oscillation parameters can be accurately identified within one or two cycles after the oscillation occurs.

[0013] Furthermore, the step A2 comprises:

[0014] Step A21: Calculate the mean square value of the current signal, which is a measurement method for the average power of the signal. The corresponding root mean square value is the square root of the mean square value. The calculation formula is:

[0015]

[0016] In the formula, x(t) is the current value, T is the length of the analysis data segment, and x rms is the RMS value of the current. When x is detected rms If it continues to increase, VMD decomposition of the current signal is immediately performed;

[0017] Step A22: Referring to the method for obtaining the effective modal number in the extended Prony algorithm, the acquired signal is discretized at the sampling interval time to obtain a discrete time function x(n). To achieve the minimum error energy calculation, the sample function is defined as follows:

[0018]

[0019] In the formula, N is the number of samples, P is the selected model order, and the corresponding extended order sample matrix R is as follows:

[0020]

[0021] In the formula, p e = [N / 2], perform singular value decomposition on the matrix R, and set the normalized ratio to determine the true model order P according to the matrix optimal approximation principle of the Frobenius norm and the singular value decomposition theory. z ; Since the parameters of the oscillation mode appear in pairs, considering the separation of the noise signal, the number of VMD decompositions K is taken as [P z / 2]+1, decompose the signal.

[0022] Furthermore, the step A4 comprises:

[0023] Step A41: According to the circuit principle, the auxiliary angle formula and the two-angle sum and difference formula are used to derive the formula, and the subsynchronous current component and the power frequency component are separated and completely solved by relying on the line related parameters;

[0024] Step A42: Input the acquired filtered voltage difference signal and current signal, construct an auxiliary voltage drop according to step S41 to eliminate the power frequency component, obtain a voltage drop signal containing only the SSO component, and use the zero-crossing method to solve the oscillation frequency, and then obtain a complete subsynchronous current signal to achieve the solution of the damping coefficient and the oscillation amplitude. According to the analysis of the oscillation mechanism of the direct-drive wind farm, the equivalent negative resistance part of the wind farm is prone to oscillation, and its impedance characteristics are generally affected by the wind turbine and its control parameters, but its oscillation range will not be very large. Therefore, according to the first VMD decomposition result, a wider bandpass filter range is set. In the subsequent system parameter identification process, the VMD decomposition step can be omitted and the filtering operation can be performed directly.

[0025] Compared with the existing technology, this technology has the following beneficial effects:

[0026] The oscillation parameter identification method provided by the present invention integrates the response data and the line model, monitors the grid-connected line current, calculates its root mean square value to realize oscillation signal early warning, combines adaptive VMD decomposition to realize preliminary identification of the oscillation signal, obtains a signal containing only SSO mode and power frequency mode after filtering, and finally realizes the parameter identification of SSO current through formula derivation. Compared with the identification analysis of a single electrical quantity, this paper conducts a comprehensive oscillation analysis on multiple electrical quantities of some known parameter lines of the system, uses the known series compensation line model as a "filter" to quickly extract the SSO component, and can more accurately identify the oscillation-related parameters, providing a correct data source for the oscillation characteristic analysis; at the same time, it provides a new direction for the identification of the oscillation propagation path and the oscillation source, which has certain guiding significance for the subsequent analysis and suppression of SSO. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solution and practical examples of the present invention, the used drawings are briefly introduced below.

[0028] Figure 1 The present invention proposes a general flow chart for implementing the method;

[0029] Figure 2 This is a simplified system diagram of a line with series compensation;

[0030] Figure 3 This is the topological structure diagram of the direct-drive wind farm grid-connected system through series capacitor compensation;

[0031] Figure 4 The voltage and current waveforms at both ends of the grid-connected line are collected;

[0032] Figure 5 The waveform of the effective components obtained by adaptive VMD decomposition of the current;

[0033] Figure 6 It is the FFT analysis result of some components;

[0034] Figure 7 The pure SSO line voltage drop and SSO current waveform obtained by calculation. DETAILED DESCRIPTION

[0035] The present invention provides an oscillation parameter identification method combining response data with a circuit model. To make the purpose, technical solution and effect of the present invention clearer, the specific implementation scheme of the present invention is described in detail below with reference to the accompanying drawings and examples. The specific examples described in the present invention are only used to explain the present invention and are not used to limit the present invention.

[0036] 1. Introduction to the specific embodiments of the invention

[0037] Figure 1 This is a general flow chart of the implementation of an oscillation parameter identification method combining response data with a circuit model provided by the present invention, wherein the dotted line box represents an unnecessary process. Figure 1 The present invention describes an oscillation parameter identification method that considers multiple electrical quantities and integrates data and line models, including:

[0038] Step A1: For a series compensation transmission line containing capacitors, a time stamp waveform measurement unit is used to obtain a grid-connected line current waveform and a voltage waveform at both ends;

[0039] Step A2: Monitor the current waveform in real time and calculate its RMS value for energy calculation. When it is detected that the energy continues to increase, it indicates that the system has an oscillation risk. It is necessary to further use the improved variational mode decomposition method to effectively decompose the signal to determine whether the signal has subsynchronous oscillation, the number of modal components contained and their frequency distribution.

[0040] Step A3: Calculate the voltage difference between the two ends of the line, and perform bandpass filtering on the current signal and the voltage difference signal to retain only the dominant subsynchronous oscillation component and the power frequency component;

[0041] Step A4: Based on the derived series compensation line voltage and current relationship formula, the complete sub-synchronous current component is calculated and separated, so that the oscillation parameters can be accurately identified within one or two cycles after the oscillation occurs.

[0042] In step A2, VMD decomposition is performed on the data with oscillation risk to preliminarily determine the system oscillation situation, including:

[0043] Step A21: Calculate the RMS value of the current data using the formula:

[0044]

[0045] In the formula, x(t) is the current value, T is the length of the analysis data segment, and x rmsIt is the RMS value of the current. When it is detected that its value continues to increase, an alarm signal is issued, and the system may have an oscillation risk.

[0046] Step A22: Perform adaptive VMD decomposition on the risk signal to determine whether the signal has subsynchronous oscillation, the number of modal components contained and their frequency distribution.

[0047] Referring to the method of obtaining the effective modal number in the extended Prony algorithm, the acquired signal is discretized at the sampling interval time to obtain the discrete time function x(n). In order to achieve the minimum error energy calculation, the sample function is defined as follows:

[0048]

[0049] In the formula, N is the number of samples, P is the selected model order, and the corresponding extended order sample matrix R is as follows:

[0050]

[0051] In the formula, p e =[N / 2], perform singular value decomposition on R as follows:

[0052]

[0053] We get S to be a diagonal matrix, with non-negative singular values ​​σ arranged in descending order. i , 1≤i≤p e According to the Frobenius norm matrix optimal approximation principle and singular value decomposition theory, larger singular values ​​represent pure signal characteristics, and smaller singular values ​​represent noise. Therefore, the signal can be separated from the noise by selecting the number of effective singular values. Generally, P is determined by setting the normalized singular value. z , the normalized ratio is defined as follows in the present invention:

[0054]

[0055] Since the parameters of the oscillation mode appear in pairs, considering the separation of the noise signal, the number of VMD decompositions M is taken as [P z / 2]+1. VMD decomposition is to obtain the unconstrained augmented Lagrangian function shown in formula (9):

[0056]

[0057] In the formula, u m is the modal component to be solved, ω mis the instantaneous angular frequency of the corresponding component, λ is the Lagrange multiplier; δ(t) is the unit pulse function, * represents convolution, and X(t) is the decomposed original signal. The multiplier alternating direction algorithm is used to solve equation (9) through cyclic iteration. The calculation formulas for updating each parameter are as follows:

[0058]

[0059] in, The IMF components and Lagrange multipliers representing the frequency domain state, τ is the noise tolerance parameter, and VMD decomposition can be achieved. By performing FFT analysis on each component, the approximate frequency distribution range of the signal can be obtained.

[0060] The step A3 divides the signal into frequency signals according to the information obtained in step A2, that is, obtains the required signal through bandpass filtering, including:

[0061] Step A31: For wind power grid-connected systems, the external characteristics of the unit are mainly determined by the converter and its control system. The second and super synchronous frequencies have strong coupling, resulting in the complementary second and super frequency oscillation components appearing in pairs after the oscillation occurs, and the latter often has a larger amplitude. In order to accurately analyze the SSO characteristics, it is necessary to filter the signal. The present invention uses a Butterworth bandpass filter for filtering, and the transfer function is as follows:

[0062]

[0063] Where L is the filter order, k is the gain, and p j According to the signal frequency distribution obtained by VMD in step A2, appropriate bandpass filter parameters are selected to complete the filtering and obtain a signal containing only SSO components and power frequency components.

[0064] Step A32: After filtering, the amplitude and phase information will be offset. In order to obtain the correct filtered waveform, the phase and amplitude need to be compensated. The parameter deviation can be adjusted by cascading an all-pass filter after the filter output signal, or the phase compensation can be achieved by using the filtfilt function in MATLAB.

[0065] It is worth noting that the present invention is mainly aimed at identifying subsynchronous oscillation signals. This is because in current practical projects, new energy stations often have oscillations within 100 Hz. When high-frequency oscillations occur in the wind power grid-connected system, the high-frequency mode to be analyzed can also be obtained by setting a low-pass filter, and the energy operator or other methods can be used to solve the oscillation parameters.

[0066] The step A4 combines the formula derivation result and calculates the SSO current component based on the line voltage drop and current signal and line parameters, including:

[0067] Step A41: According to the circuit principle, the auxiliary angle formula and the two-angle sum and difference formula are used to derive the formula, and the numerical correlation relationship between the line-related parameters is obtained. Figure 2 For a simplified system with series compensation, the voltages at both ends of the line are V1(t) and V2(t), the current is I1(t), and the line resistance, inductance and series capacitance are R, L and C respectively. When the system has subsynchronous oscillation, ideally, the voltage and current contain both power frequency and subsynchronous frequency components, and the oscillation source is equivalent to an AC voltage source whose amplitude change is affected by the system damping, then:

[0068]

[0069] Among them, V i0 , I i0 is the voltage and current amplitude of the fundamental frequency component, V iSS ,I iSS is the voltage and current amplitude of the subsynchronous component, α is the damping factor of the subsynchronous component, f0, f SS is the system power frequency and SSO frequency, θ i0 and γ i0 is the initial phase angle of the fundamental frequency component voltage and current, θ iSS and γ iSS is the initial phase angle of the subsynchronous component voltage and current.

[0070] According to the circuit principle and auxiliary angle formula, and based on the general α<<2πf SS , the voltage difference between the two ends of the line is simplified to:

[0071]

[0072] Among them, Z0, Z SS Equivalent to the equivalent impedance of the line at the fundamental frequency and subsynchronous frequency; η0, η ss As shown below:

[0073]

[0074] In order to obtain a signal containing only the subsynchronous component, an auxiliary voltage drop ΔV needs to be constructed. ass (t) is obtained by multiplying the line current by the equivalent impedance at the fundamental frequency and shifting it in time by a fundamental impedance angle. The pure SSO signal is as follows:

[0075]

[0076] in:

[0077]

[0078] Step A42: According to formula (15), ΔVSS (t) contains only subsynchronous frequency components, and A and θ are functions independent of time t, so the oscillation frequency f can be directly obtained SS =1 / 2(t2-t1), where t1 is the signal peak moment and t2 is the corresponding moment of the subsequent peak. SS After determining, ΔV SS (t) Carry out the following processing:

[0079]

[0080] The complete subsynchronous current component can be obtained. Solving equation (18) we can get the damping factor α:

[0081]

[0082] Among them, Δt peak is the interval between two adjacent peaks, 1 st peak is the subsynchronous current amplitude corresponding to the first peak, 2 nd peak is the sub-synchronous current amplitude corresponding to the next peak.

[0083] Although the original voltage and current components are time-shifted in the process of "filtering" and transformation, the transformation is small and the influence of this part can be ignored. And when there is a single integer harmonic distortion in the waveform, according to the principle that the harmonic remains constant within a cycle, the corresponding signal containing only the oscillation component is shifted by half a harmonic cycle to eliminate a certain harmonic component.

[0084] When the increase in line current energy is detected, VMD decomposition is performed to initially obtain modal information, and bandpass filter parameters are set to filter the line current and voltage drop at both ends. Then, based on the derivation calculation, the pure SSO current signal is obtained to achieve accurate identification of the oscillation parameters. When the acquired SSO current amplitude exceeds a certain threshold, it means that the system may have a local oscillation mode in a weakly damped state, and the operation and dispatching personnel need to take relevant measures to avoid the global occurrence and spread of oscillations. Once negative damping is present, no matter how small the oscillation amplitude is, emergency control measures should be taken to avoid the evolution into power angle instability.

[0085] 2. Technical feasibility verification of the present invention

[0086] Figure 3 The topological structure diagram of the direct-drive wind farm connected to the grid through series compensation used for simulation verification is shown in the figure. Figure 4 The voltage and current waveforms of the grid-connected lines are collected. The simulation results are as follows: Figure 5 , Figure 6 and Figure 7 As shown, the effectiveness of the present invention is verified.

[0087] Figure 4 In order to change the integral coefficient of the outer loop of the d-axis voltage of the grid-side controller in 3s to excite oscillation, the voltage and current waveforms at both ends of the line are obtained by the measurement unit. Figure 4 It can be seen that due to the low resonant frequency, the corresponding line impedance is small, and the voltage distortion caused is not obvious compared to the current change.

[0088] Figure 5 The component waveform diagram below 100 Hz obtained by VMD decomposition of the line current, the mode number obtained according to the solution is 14 / 2+1=8.

[0089] Figure 6 The FFT analysis results of the 5 components below 100Hz are combined with Figure 5 It can be seen that VMD can effectively decompose each component, and sub- / super-synchronous components appear in pairs and are complementary in frequency. Calculating the energy entropy of each component shows that IMF1, IMF2 and IMF7 are dominant. However, for super-synchronous components, the system often has positive damping characteristics and generally does not cause oscillation. Sub-synchronous components are prone to negative damping, thereby causing SSO. Therefore, sub-synchronous frequency components are often more concerned in engineering.

[0090] Figure 7 The pure SSO signal ΔV is obtained by calculating the line current and the voltage drop at both ends after bandpass filtering and compensation. SSO and the subsynchronous SSO current I 1SS The SSO component IMF2 obtained by VMD decomposition is compared with the SSO component I calculated by combining the voltage signal and RLC parameters. 1SS , we can see that the frequency difference between the two is small. However, due to the certain error in the VMD decomposition and reconstructed signal, the algorithm in this paper can further correct the oscillation signal, which is more conducive to the accurate identification of the oscillation parameters. Figure 7 It can be seen that the system is gradually dominated by the excited SSO oscillation mode starting from 3.05s, and the oscillation frequency is basically stable at 3.1s. Therefore, V SSO The first two peaks of the waveform are recorded to obtain the oscillation frequency f ss =1 / 2(3.1525-3.122)=32.79Hz, which is closer to the result of characteristic value analysis of the system (32.66Hz). Calculating the five peak data in the period of 3 to 3.18s, the damping factors are 19.658, 11.182, 4.836, and 2.193 respectively. It can be seen that as time increases, the attenuation factor continues to decrease. This is because the limiting link in the system works, causing the system to tend to equal amplitude oscillation.

[0091] The above simulation results show that the wind farm SSO identification method using the fusion response and model data of the present invention can accurately and quickly obtain system oscillation parameters and time-varying attenuation factors. The above simulation results verify the effectiveness and feasibility of the identification method of the present invention.

[0092] In summary, the SSO oscillation identification method integrating response data and model described in the present invention can accurately identify SSO in the early stage of oscillation. Compared with the identification and analysis of a single electrical quantity, the present invention performs a comprehensive oscillation analysis on multiple electrical quantities of some system circuits with known parameters, which can more accurately identify oscillation-related parameters and provide a correct data source for oscillation characteristic analysis.

[0093] Finally, it should be noted that the above examples of the present invention are merely examples for illustrating the present invention, and are not intended to limit the embodiments of the present invention. Although the applicant has described the present invention in detail with reference to the preferred embodiments, for those of ordinary skill in the art, other different forms of changes and modifications can be made based on the above description. It is impossible to list all the embodiments here. Any obvious changes or modifications derived from the technical solution of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for identifying oscillation parameters of a wind farm grid-connected system by integrating system response data with a line model, characterized in that: The following steps are involved: S1: For a series-compensated transmission line with capacitors, a time-stamped waveform measurement unit is used to obtain the line current waveform and the voltage waveform at both ends; S2: Monitor the current waveform in real time and calculate its RMS value for energy calculation. When it is detected that the energy continues to increase, it indicates that the system has an oscillation risk. It is necessary to further use the improved variational mode decomposition method to effectively decompose the signal to determine whether the signal has subsynchronous oscillation, the number of modal components contained and their frequency distribution. S3: Calculate the voltage difference between the two ends of the line, and perform bandpass filtering on the current signal and the voltage difference signal, retaining only the dominant subsynchronous oscillation component and the power frequency component; S4: Based on the derived series-compensated line voltage and current relationship formula, the complete sub-synchronous current component is calculated and separated, which can accurately identify the oscillation parameters within one or two cycles after the oscillation occurs.

2. The method according to claim 1, characterized in that: The step S2 comprises: S21: Calculate the mean square value of the current signal, which is a measurement method for the average power of the signal. The corresponding root mean square value is the square root of the mean square value. The calculation formula is: In the formula, x(t) is the current value, T is the length of the analysis data segment, and x rms is the RMS value of the current. When x is detected rms If it continues to increase, VMD decomposition of the current signal is immediately performed; S22: Referring to the method for obtaining the effective modal number in the extended Prony algorithm, the acquired signal is discretized at the sampling interval time to obtain the discrete time function x(n). In order to achieve the minimum error energy calculation, the sample function is defined as follows: In the formula, N is the number of samples, P is the order of the model to be analyzed, and the corresponding extended order sample matrix R is as follows: In the formula, p e = [N / 2], perform singular value decomposition on the matrix R, and set the normalized ratio to determine the true model order P according to the matrix optimal approximation principle of the Frobenius norm and the singular value decomposition theory. z ; Since the parameters of the oscillation mode appear in pairs, considering the separation of the noise signal, the number of VMD decompositions K is taken as [P z / 2]+1, decompose the signal.

3. The method according to claim 1, characterized in that The step S4 comprises: S41: According to the circuit principle, the auxiliary angle formula and the two angle sum and difference formula are used to derive the formula, and the separation and complete solution of the subsynchronous current component and the power frequency component are realized by relying on the line related parameters; S42: Input the obtained filtered voltage difference signal and current signal, construct an auxiliary voltage drop according to step S41 to eliminate the power frequency component, and obtain a voltage drop signal containing only the SSO component. The oscillation frequency can be solved by the zero-crossing method, and then the complete sub-synchronous current signal can be obtained to achieve the solution of the damping coefficient and the oscillation amplitude.

Images