A wind farm hybrid oscillation suppression strategy under different operating conditions

Abstract

The application discloses a wind farm hybrid oscillation suppression strategy under different operation conditions, comprising the following steps: step S1: determining the proportion of voltage source type wind turbine in the system through the short-circuit ratio of the system connected with the wind farm, the series compensation degree of the wind farm sending line and the real-time output of the wind farm; step S2: calculating the proportion of the voltage source type wind turbine in the system; step S3: selecting the wind farm internal unit for mode switching according to a certain strategy; the application calculates the proportion of the voltage source type wind turbine to avoid introducing new oscillation modes into the system; the system connected with the wind farm is predicted in real time to analyze and adjust the external electrical damping of the wind farm in real time; the grid-connected power of the wind turbine and the electrical distance of the wind turbine from the grid-connected point are used to select the appropriate wind turbine for control mode switching, thereby increasing the accuracy and adaptability of the mode switching.

Description

Technical Field

[0001] This invention relates to the field of power transmission, and in particular to a strategy for suppressing oscillations in wind farms under different operating conditions. Background Technology

[0002] Changing the energy structure and increasing the proportion of clean energy sources such as wind power is a crucial strategic measure to achieve "carbon peaking and carbon neutrality." However, due to resource constraints, my country's wind power resources are concentrated in the "Three Norths" region (Northeast, North, and Northwest China), far from load centers. Therefore, the wind power development model of "large-scale centralized development and long-distance high-voltage transmission" is adopted. Wind power is connected to the end of the power system, causing a gradual weakening of the local power grid. Furthermore, with increasing transmission distance, the system's transmission capacity decreases and stability margin diminishes. Fixed series compensation technology is mature and inexpensive, and is widely used in large-scale long-distance transmission systems to improve system transmission capacity and stability margin. However, with the increasing proportion of wind power, traditional current-source controlled wind turbines are equivalent to large inductors externally, easily forming capacitive-inductive resonant circuits with the series compensation capacitors in the line. This can cause sub- / super-synchronous oscillations in the transmission line, leading to large-scale wind turbine disconnection from the grid, seriously threatening the safe and stable operation of the power system, and causing irreversible mechanical damage to synchronous generators and wind turbines within the system.

[0003] Therefore, in order to suppress the oscillation of the wind farm transmitted through series compensation, there are currently two main methods:

[0004] (1) By changing the line parameters through the wind farm disconnection method or the thyristor-controlled series compensation (TCSC) device, the resonance point of subsynchronous / supersynchronous oscillation can be avoided. The wind farm disconnection method mainly uses online detection devices to monitor the harmonics of the wind farm grid connection point. When a wind farm experiences subsynchronous / supersynchronous oscillation, it is directly disconnected to avoid power system oscillation caused by the oscillation of a single wind farm. The TCSC can change the line impedance according to the changes in the auxiliary signal, thereby suppressing system oscillation.

[0005] (2) System oscillations can be suppressed by improving the electrical damping of the power system through optimized control strategies. The main methods include optimizing the control strategies of Flexible AC Transmission Systems (FACTS) and wind turbine control strategies. The former mainly includes optimizing and upgrading the control strategy of Static Var Generators (SVG), while the latter mainly includes (but is not limited to) using the virtual impedance method to provide positive damping to suppress oscillating current at different oscillation frequencies, and using an adaptive oscillation resonant controller to provide system oscillation damping.

[0006] The existing technology has the following two drawbacks: 1. The disconnection of wind farms will cause an impact on the power system, which will have an adverse effect on the system's safety and stability; 2. Using TCSC to suppress system oscillations will increase the construction and operation costs of the transmission system, and the addition of new hardware will increase additional failure points and reduce system reliability.

[0007] For example, a patent document published in China, titled "An Oscillation Suppression Strategy for a PIR-Based Flexible DC Transmission System," with publication number CN112103969A, addresses the issue that the electrical damping of wind farms cannot be adjusted in real time according to different external grid conditions, and that the disconnection of wind farms can cause an impact on the power system, thus adversely affecting the safety and stability of the system. Summary of the Invention

[0008] This invention aims to overcome the problems in the prior art where adjusting the control parameters of traditional current-source wind turbine generators (such as the parameters of the rotor current inner loop PI controller and the parameters of the power outer loop PI controller) can change the electrical damping of the wind farm, but improper parameter selection may lead to instability of the wind farm system, and the parameter selection method based on empirical analysis has poor robustness. This invention provides a hybrid oscillation suppression strategy for wind farms under different operating conditions.

[0009] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution:

[0010] A strategy for suppressing oscillations in a wind farm under different operating conditions includes the following steps:

[0011] Step S1: Determine the proportion of voltage-source wind turbines in the system by considering the short-circuit ratio of the wind farm's grid connection, the series compensation degree of the wind farm's transmission lines, and the real-time output of the wind farm. Step S2: Calculate the proportion of voltage-source wind turbines in the wind farm. Step S3: Select the internal units of the wind farm for mode switching according to a certain strategy. Through the above three steps, the proportion of voltage-source wind turbines is calculated, and the internal unit modes of the wind farm are switched to achieve the effect of oscillation suppression. This scheme can adjust the electrical damping of the wind farm in real time according to different external grid conditions without increasing the modes of the original system, thus improving the stability of the wind farm's grid connection.

[0012] Preferably, step S1 includes the following steps: Step S11: Calculate the power flow of the wind farm connected to the system and form the corresponding admittance matrix Y, and calculate the equivalent short-circuit ratio (ESCR) of the system; Step S12: Correct the input of series compensation of the power transmission line of the wind farm. If the line has series compensation, the original admittance matrix needs to be corrected and the series compensation degree of the power transmission line needs to be calculated. If there is no series compensation input, the series compensation degree of the line is directly calculated (k=0); The actual power output of the wind farm can be directly obtained from the wind farm energy management system; Calculate the power flow of the wind farm connected to the system and the series compensation input of the power transmission line to provide data support for the subsequent calculation of the proportion of voltage prototype wind turbine units, and improve the calculation accuracy.

[0013] As a preferred method, the following specific calculation method is obtained through step S11:

[0014]

[0015] Where ΔV represents the change in bus voltage, and the subscripts i and j represent the bus positions, respectively; z ji Z represents the mutual impedance between stations i and j. ii P represents the self-impedance of station i; i For the actual power output of the wind farm; S i The short-circuit capacity is denoted by pu; the subscript pu represents the per-unit output of the wind farm. When more than one renewable energy power plant is connected to a power system, the network strength in the area is significantly lower than the network short-circuit level calculated on the bus due to their electrical proximity. ECSR considers the impact of other nearby renewable energy power plants on the system short-circuit ratio and uses data to support real-time adjustment of the electrical damping of the wind farm, thereby improving the robustness of the parameter selection method and avoiding the introduction of new oscillation modes into the system.

[0016] As a preferred method, when the wind farm's transmission line is connected to series compensation, the system admittance is modified to calculate the series compensation degree of the transmission line, and the original transmission line admittance is changed from y il Become y il The corrected system admittance matrix is ​​Y'. Based on the corrected admittance matrix Y', the corrected system impedance matrix Z' = Y' is obtained. -1 Based on the modified line inductance x between busbars i and l il 'and the output line series capacitor resistance x' c Calculate the series complement k of the wind farm:

[0017]

[0018] The proportion of voltage source wind turbines in the system is determined by the short-circuit ratio of the wind farm's grid connection, the series compensation degree of the wind farm's transmission lines, and the real-time output of the wind farm. This provides real-time data support for the selection of strategy system parameters and improves the accuracy and adaptability of unit mode switching within the wind farm.

[0019] Preferably, step S2 includes the following steps:

[0020] Step S21: Train the wind farm damping assessment model based on historical wind farm operation data and monitor small disturbance processes in the wind farm;

[0021] Step S22: Obtain the system damping and dominant oscillation mode in real time by monitoring;

[0022] Step S23: Oscillation prediction employs a Broad Learning System (BLS) based on incremental learning. Its core principle is to update the weights of the original network structure using the previous calculation results and newly added data. Therefore, the sample input x of the prediction system is represented as:

[0023] x = [P] farm ,k,ESCR,K1,K2,K3,…,L1,L2,L3,…]

[0024] Among them, K i For the control mode of the i-th unit, K i =1 indicates that unit i adopts current source control mode, K i =0 indicates that unit i uses voltage source control mode, and the unit control mode can be obtained from the field control; P farm L represents the wind farm output power obtained from the wind farm energy management system. i Let be the electrical distance from the i-th turbine to the grid connection point of the wind farm. Multiple sets of data are used to accurately calculate the factors affecting the oscillation characteristics of the wind farm. Based on the wind farm output, different series compensation degrees of the wind farm's transmission system, the equivalent short-circuit ratio at the grid connection point, the control mode of the wind turbines within the wind farm, and the electrical distance between the wind turbines and the grid connection point, the system connected to the wind farm is predicted in real time. This allows for real-time analysis and adjustment of the external electrical damping of the wind farm, improving the accuracy and adaptability of turbine mode switching within the wind farm.

[0025] Preferably, the oscillation frequency ω of the system oscillation mode and the damping ratio ξ of the corresponding oscillation mode are used as the output of the width learning, i.e., Y = [ω, ξ]. The initial number of feature nodes Z and enhancement nodes H are n and m, respectively, and the network structure is represented as A. m =[Z n |H m The methods for obtaining each set of feature nodes and augmentation nodes are as follows:

[0026]

[0027] Where φ and f are function expressions; We i ,βe i Whi ,βh i The weights and biases from the feature nodes to the output and from the feature nodes to the augmentation nodes are randomly generated by the width learning system of incremental learning. The random generation of the weights and biases from the feature nodes to the augmentation nodes enables the strategy system to adjust the electrical damping of the wind farm in real time and accurately under different external environmental emergencies without increasing the modes of the original system, thereby improving the system's safety, stability and reliability.

[0028] As a preferred approach, reasonable judgment criteria and thresholds are set to determine the closeness of the oscillation frequency and damping ratio output by the width learning system to the oscillation frequency and damping ratio of the actual system. The judgment criterion adopts the root mean square error (RMSE), and the oscillation frequency and damping ratio thresholds are combined using a weighted approach, as specifically expressed in the following formula:

[0029]

[0030] Where k represents the number of key oscillation modes of the system; the superscript ^ represents the variable output by the width learning system; a and b represent the weights of oscillation frequency and damping ratio, and satisfy a+b=1; the above calculation formula comprehensively calculates the oscillation frequency and damping ratio, reduces the error of subsequent data integration, improves the real-time performance and accuracy of the data, and ensures the safety and stability of the system.

[0031] Therefore, the present invention has the following beneficial effects:

[0032] The hybrid oscillation control strategy for wind farms under different operating conditions comprehensively considers the series compensation degree of the wind farm's transmission system, the short-circuit ratio at the wind farm's grid connection point, and the wind farm's power output to determine the proportion of voltage source wind turbine units within the wind farm, in order to avoid introducing new oscillation modes into the system.

[0033] Based on the wind farm output, different series compensation degrees of the wind farm transmission system, the equivalent short-circuit ratio of the grid connection point, the control mode of the wind turbine units inside the wind farm, and the electrical distance between the wind turbine units and the grid connection point, the system connected to the wind farm is predicted in real time, so as to analyze and adjust the external electrical damping of the wind farm in real time.

[0034] Based on the proportion of voltage source wind turbines output by the field-level controller, the grid-connected power of the wind turbines, and the electrical distance between the wind turbines and the grid connection point, a suitable wind turbine is selected for control mode switching, thereby increasing the accuracy and adaptability of mode switching. Attached Figure Description

[0035] Figure 1 This is a flowchart of the present invention;

[0036] Figure 2 This is a flowchart of the decision-making process for the proportion of voltage sources in a wind farm;

[0037] Figure 3 This is a framework diagram for oscillation prediction in a width-learning system based on incremental learning;

[0038] Figure 4 This is a flowchart outlining the principles for determining and switching turbine units within a wind farm. Detailed Implementation

[0039] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0040] Example 1:

[0041] like Figure 1 As shown, the wind farm hybrid oscillation suppression strategy under different operating conditions mainly determines the proportion of voltage source wind turbine units in the system by the short-circuit ratio of the wind farm connected to the system, the series compensation degree of the wind farm's transmission line, and the real-time output of the wind farm. Then, based on the proportion of voltage source units calculated by the field control, the wind farm selects units within the wind farm for mode switching according to a certain strategy.

[0042] Calculate the power flow of the wind farm connected to the system and form the corresponding admittance matrix Y. Calculate the equivalent short-circuit ratio of the system. At the same time, determine whether the power transmission line of the wind farm has series compensation enabled. If the line has series compensation enabled, the original admittance matrix needs to be corrected and the series compensation degree of the power transmission line needs to be calculated. If there is no series compensation enabled, the series compensation degree of the power transmission line is directly calculated (k=0). The actual power output of the wind farm can be obtained directly from the wind farm energy management system.

[0043] Figure 1 In China, the interaction influence factor WPIF ji This ratio represents the impact of a small voltage change at bus j on the voltage at bus i. To a certain extent, this ratio measures the electrical distance between two renewable energy power stations j and i. When this value is close to 0, it indicates a relatively large electrical distance between the two stations; conversely, it indicates a very close electrical distance. The specific calculation formula is shown below:

[0044] In the above formula, ΔV represents the change in bus voltage, and the subscripts i and j represent the bus positions, respectively; z ji Z represents the mutual impedance between stations i and j. ii This represents the self-impedance of station i.

[0045] When more than one renewable energy power station is connected to a power system, due to their electrical proximity, the network strength in the area is significantly lower than the network short-circuit level calculated on the bus. The ECSR takes into account the impact of other nearby renewable energy power stations on the system short-circuit ratio. The specific calculation method for the ECSR of power station i is as follows:

[0046]

[0047] In the above formula, P i For the actual power output of the wind farm; S i This represents the short-circuit capacity of the wind power plant; the subscript pu represents the per-unit output value of the power station.

[0048] The system admittance matrix Y is shown below:

[0049]

[0050] When a wind farm's transmission line is connected to a series compensation, the system admittance needs to be revised to calculate the series compensation degree of the transmission line. The original transmission line admittance (with series compensation) is determined by y il Become y il (where i is the grid connection point bus of the wind farm, and l is the other end bus of the transmission line), the corrected system admittance matrix Y' can be expressed as

[0051]

[0052] Based on the corrected admittance matrix, the corrected system impedance matrix Z' = Y' can be obtained. -1 Based on the modified line inductance x between busbars i and l il 'and the output line series capacitor resistance x' c Calculate the series complement k of the wind farm:

[0053] like Figure 2 As shown, the calculation of the proportion of voltage-source wind turbines includes the following: determining the proportion of voltage-source wind turbines in the system based on the short-circuit ratio of the wind farm's grid connection, the series compensation degree of the wind farm's transmission lines, and the real-time output of the wind farm. The control flowchart is shown below. Figure 2 As shown, the wind farm damping assessment model is trained based on historical wind farm operation data, and the small disturbance process of the wind farm is monitored to obtain the system damping and dominant oscillation mode in real time. The damping analysis algorithm is not limited (it can be any one or a combination of machine learning and deep learning algorithms). The factors affecting the oscillation characteristics of the wind farm include the strength of the connected system, line series compensation, wind farm output and wind turbine control, etc. The strength of the system, line series compensation and wind farm output can be based on the previous calculation results. Wind turbine control mainly refers to current source control strategy and voltage source control strategy.

[0054] Oscillation prediction employs a Broad Learning System (BLS) based on incremental learning (not limited to this one). Its core principle is to update the weights of the original network structure using the previous calculation results and newly added data. Therefore, the sample input x of the prediction system is represented as:

[0055] x = [P] farm,k,ESCR,K1,K2,K3,…,L1,L2,L3,…] (5)

[0056] In the expression for the sample input x of the above prediction system, K i For the control mode of the i-th unit, K i =1 indicates that unit i adopts current source control mode, K i =0 indicates that unit i uses voltage source control mode, and the unit control mode can be obtained from the field control; P farm L represents the wind farm output power obtained from the wind farm energy management system. i Let be the electrical distance from the i-th unit to the grid connection point of the wind farm.

[0057] After obtaining the sample input x of the prediction system, the oscillation frequency ω of the system's oscillation mode and the corresponding damping ratio ξ of the oscillation mode are used as the output of the width learning, i.e., Y = [ω, ξ]. The specific process is as follows: Figure 3 As shown.

[0058] like Figure 3 As shown in the figure, the initial number of feature nodes Z and augmentation nodes H are n and m, respectively, and the network structure is represented as A. m =[Z n |H m The method for obtaining each set of feature nodes and enhancement nodes is as follows:

[0059]

[0060] In the above calculation formula, φ and f are function expressions, respectively; We i ,βe i Wh i ,βh i These are the weights and biases from the feature node to the output, and the weights and biases from the feature node to the augmentation node, respectively, which are randomly generated by the incremental learning width learning system.

[0061] Set reasonable judgment criteria and thresholds (the judgment criteria and thresholds are not limited to those mentioned in the patent) to judge the closeness of the oscillation frequency and damping ratio output by the width learning system to the oscillation frequency and damping ratio of the actual system: if e≤0.1, the accuracy requirement is met, and no additional enhancement nodes are needed; if the accuracy requirement is not met, it indicates that the initial model design has insufficient fitting ability, and enhancement nodes need to be added to improve the system fitting degree. At this time, the new network structure can be described as A m+1 =[A m |H m+1 The relationship between feature nodes and newly added enhancement nodes is shown in the following formula:

[0062]

[0063] The judgment criterion adopts the root mean square error (RMSE), and the oscillation frequency and damping ratio threshold are combined using a weighted approach. The specific calculation method after combination is as follows:

[0064]

[0065] In the above formula, k represents the number of key oscillation modes of the system; the superscript ^ represents the variable output by the width learning system; a and b represent the oscillation frequency and damping ratio weights, and satisfy a+b=1.

[0066] Real-time data from wind farms and the power grid are input into a pre-trained oscillation prediction model to predict key oscillation modes. The model then determines if the damping ratio of the key oscillation mode is too small (or negative): If the damping ratio of all key oscillation modes is greater than a threshold, the proportion of voltage-source turbines is output; if there are oscillation modes with a damping ratio less than the threshold (the threshold can be adjusted according to system requirements, but is recommended to be no less than 0.1), the proportion of voltage-source wind turbines needs to be adjusted based on the oscillation frequency of the oscillation mode. The damping judgment criteria are as follows:

[0067] ξ i ≥0.2 (7)

[0068] As shown in the above formula, ξ i The damping ratio represents oscillation mode i. Oscillation frequencies with damping ratios less than a threshold are extracted. When the oscillation frequency is less than or equal to 10 rad / s, the proportion of voltage sources within the wind farm is reduced to reflect the suppression of oscillations in the ultra-low and low frequency bands by current-source wind turbines. When the oscillation frequency is greater than 10 rad / s, the proportion of voltage sources within the wind farm is increased to reflect the suppression of oscillations in the mid- and high-frequency bands by voltage-source wind turbines. The specific dominant mode switching block diagram is shown below. Figure 2 As shown.

[0069] Example 2:

[0070] The working principle of unit selection and switching principles:

[0071] like Figure 4 As shown, when the wind farm controller detects that the proportion of voltage source turbines within the wind farm does not meet the proportion calculated by the wind farm controller, it is necessary to select specific turbines for mode switching to meet the proportion requirements of voltage source turbines within the wind farm. Since current source turbines exhibit greater negative resistance at subsynchronous frequencies under light loads and are more prone to instability, the selection principle for mode switching turbines can be expressed as follows: Figure 4 , where x t This represents the proportion of voltage-source wind turbine units calculated in real time by the field control system, x t-1This indicates the proportion of voltage source type wind turbines within the current wind farm.

[0072] The unit switching probability can be fitted using a multivariate function, as shown in the following expression:

[0073] PR i =K i+ b1L i+ b2P i (8)

[0074] In the above formula, K i Let b1 be the control mode for the i-th unit, and b2 be a constant term of 0 / 1; b1 and b2 are regression coefficients, where b1 is P. i When fixed, L i Each additional unit affects PR i The effect, namely L i PR i The partial regression system.

[0075] When the proportion of voltage sources obtained by the field-level controller is greater than the current proportion of the wind farm, the proportion of voltage source-type wind turbines within the wind farm should be increased, according to PR. i The mode switching of wind turbines is performed according to a descending order principle, prioritizing current-source wind turbines with light loads and greater electrical distance from the grid connection point for control mode switching; when the proportion of voltage sources obtained by the farm-level controller is less than the current proportion of the wind farm, the proportion of voltage-source wind turbines within the wind farm should be reduced, according to PR. i The mode switching of the units is carried out according to the ascending order principle, with priority given to voltage source wind turbine units that are heavily loaded and have a shorter electrical distance from the grid connection point for control mode switching.

[0076] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Other variations and modifications are possible without departing from the technical solutions described in the claims.

Claims

1. A strategy for suppressing oscillations in a hybrid wind farm under different operating conditions, characterized in that, Includes the following steps: Step S1: Calculate the short-circuit ratio of the wind farm connected to the system by forming the admittance matrix through the system power flow. If the outgoing line contains series compensation, then correct the admittance matrix to obtain the series compensation degree of the wind farm's outgoing line. Step S2: Using the real-time power output of the wind farm, the short-circuit ratio, and the series compensation degree as inputs, the oscillation frequency and damping ratio are output by the width learning system trained with historical data. The judgment standard adopts the root mean square error. The threshold judgments of oscillation frequency and damping ratio are combined in a weighted manner, and the proportion of voltage source type wind turbine units in the wind farm is obtained based on the comparison results with the threshold. Step S3: Select the internal units of the wind farm for mode switching according to a certain strategy. The certain strategy is: based on the proportion of voltage source units determined by the oscillation frequency and damping ratio threshold, select the units to switch between the current current source and voltage source control modes according to the comprehensive priority order of unit output and electrical distance to the grid connection point.

2. The wind farm hybrid oscillation suppression strategy under different operating conditions according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Calculate the power flow of the wind farm connected to the system and generate the corresponding admittance matrix. Y Calculate the equivalent short-circuit ratio (ESCR) of the system; Step S12: Correct the system admittance matrix by determining the series compensation input of the station's outgoing lines. Y '.

3. The wind farm hybrid oscillation suppression strategy under different operating conditions according to claim 2, characterized in that, Through step S11, the following specific calculation method is obtained: , where Δ V Indicates the change in bus voltage, subscript i , j These indicate the positions of the busbars; z ji Indicates station i , j mutual impedance between z ii Indicates station i Self-impedance; P i To provide actual power to the wind farm; S i This represents the short-circuit capacity of the wind farm; the subscript pu represents the per-unit output value of the station.

4. The wind farm hybrid oscillation suppression strategy under different operating conditions according to claim 3, characterized in that, When connecting the wind farm's transmission line to a series compensation system, the system admittance is modified to calculate the series compensation degree of the transmission line. The original transmission line admittance is changed from... y il become y il The corrected system admittance matrix is: Y '.

5. A wind farm hybrid oscillation suppression strategy under different operating conditions according to claim 4, characterized in that, Based on the corrected admittance matrix Y ', thus obtaining the corrected system impedance matrix Z '= Y ' -1 Based on the modified line inductance between busbars i and l x il 'and the output line series compensation capacitor x c Calculate the series complement of a wind farm k : 。 6. The wind farm hybrid oscillation suppression strategy under different operating conditions according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Train the wind farm damping assessment model based on historical wind farm operation data and monitor small disturbance processes in the wind farm; Step S22: Obtain the system damping and dominant oscillation mode in real time by monitoring; Step S23: Oscillation prediction employs a Broad Learning System (BLS) based on incremental learning to update the weights of the original network structure and predict the sample input of the system. x Represented as: x =[ P farm ， k ， ESCR ， K 1， K 2， K 3，…， L 1， L 2， L 3，…] in, K i For the first i Control method of the generator set. K i =1 indicates the unit i Current source control mode is adopted. K i =0 indicates the unit i The unit control mode can be obtained from the field control system when using voltage source control. P farm The output power of the wind farm obtained from the wind farm energy management system; L i For the first i The electrical distance from the turbine unit to the grid connection point of the wind farm; ESCR This is the system's equivalent short-circuit ratio; k For wind farm series compensation degree.

7. The wind farm hybrid oscillation suppression strategy under different operating conditions according to claim 1, characterized in that, Oscillation frequency using system oscillation mode ω and the damping ratio of the corresponding oscillation mode ξ As the output of width learning, i.e. Y =[ ω , ξ Feature nodes Z and enhanced nodes H The initial number of groups are respectively n and m The network structure is represented as A m =[ Z n | H m The methods for obtaining each set of feature nodes and augmentation nodes are as follows: ,in, φ and f They are function expressions; We i , βe i , Wh i , βh i These are the weights and biases from the feature node to the output, and the weights and biases from the feature node to the augmentation node, respectively, which are randomly generated by the incremental learning width learning system.

8. The wind farm hybrid oscillation suppression strategy under different operating conditions according to claim 6, characterized in that, To determine the closeness of the oscillation frequency and damping ratio output by the width learning system to the actual system, reasonable judgment criteria and thresholds are set. The judgment criterion uses root mean square error, and the oscillation frequency and damping ratio thresholds are combined using a weighted approach, as shown in the following formula: ,in, k Indicates the number of key oscillation modes of the system; the superscript ^ indicates the variable output by the width-learning system; a , b Represents the weights of the oscillation frequency and damping ratio, and satisfies a + b =1.

Images