A wind farm SVG reactive operation point dynamic setting method based on SVR

Abstract

This invention provides a dynamic tuning method for the operating point of a wind farm static var generator (SVG) based on support vector regression (SVR). The specific steps are as follows: First, acquire and preprocess historical data of the PCC voltage at the wind farm's grid connection point and the reactive power output of the SVG. Second, during the offline training phase, solve for the SVR model parameters based on the historical data and verify the reasonableness of the operating point range output by the SVR model. Finally, during the online tuning phase, monitor the PCC voltage and input it into the SVR model to calculate the optimal operating point of the SVG. This invention is applied to the field of reactive power and voltage control in wind farms. Compared with traditional methods that set fixed operating point targets for the SVG, it improves the SVG's adjustment capability when the PCC voltage is near its upper or lower limits, and has high engineering application value.

Description

Technical Field

[0001] This invention relates to reactive power control in wind farms, and particularly to a method for dynamically setting the reactive power operating point of a static var generator (SVG) in a wind farm based on support vector regression (SVR). Background Technology

[0002] Wind energy not only does not cause environmental pollution but is also inexhaustible, thus being regarded as an ideal energy source to help achieve the "dual carbon" goal. In my country, wind farms are typically located at the far end of the main power grid, and wind turbines are mainly connected to the grid via grid-connected converters (GFLs), lacking active voltage support capabilities. Therefore, they belong to typical weak grid systems, facing the risk of transient overvoltage under large disturbances. Simultaneously, wind power output is characterized by randomness and fluctuation, with changes in active power output causing rapid fluctuations in reactive power losses, making accurate control of the point of connection (PCC) voltage difficult. Currently, wind farms mainly use SVG (Static Var Generator) to achieve rapid control of the PCC voltage, but the configuration capacity of SVG is limited by its high cost, making it difficult to simultaneously achieve multiple objectives such as suppressing voltage fluctuations, executing automatic voltage control (AVC) voltage regulation commands, and high / low voltage ride-through. An effective means to address this problem is to fully utilize the idle capacity of wind turbines to output reactive power, ensuring that the SVG always maintains a certain dynamic reactive power margin to cope with potential voltage risks.

[0003] The implementation of SVG and wind turbine coordinated control includes PCC voltage control mode and reactive power cut-off control mode. In the latter, during the wind turbine reactive power output adjustment process, the SVG is controlled to track both the PCC voltage and the target reactive power value, achieving accurate alignment between the reactive power replacement process and the AVC target. The reactive power control process itself does not fundamentally change the PCC voltage; it only smoothly moves the SVG operating point to the set target by changing the overall reactive power output of the wind turbine. Specifically, some researchers prioritize the use of wind turbine reactive power when controlling the PCC voltage, only using the SVG to fill the gap when the wind turbine reactive power is insufficient; others use model predictive control methods to transform reactive power control into an optimization problem, using the minimum total reactive power output of the SVG in the prediction time domain as the objective function; still others have proposed the concept of "reactive power replacement," increasing the output of wind turbines or parallel reactors (or other chronic reactive power devices in the wind farm) after the SVG has acted to release the SVG margin.

[0004] Although the above methods have different implementation processes and effects, their core idea is to minimize the reactive power output of the SVG, that is, to always take the reactive power output of the SVG as 0 as the optimal operating point, at which time the inductive and capacitive reactive power reserves of the SVG are equal to its rated capacity. However, the actual demand of wind farms for SVG reactive power reserves will change continuously with the operating conditions, and the control objectives of the SVG during the reactive power replacement process should be dynamically adjusted accordingly.

[0005] Based on the above technical background, this invention proposes a dynamic setting method for the reactive power operating point of wind farms based on SVR, combining the characteristics of the current conventional static setting method for SVG reactive power operating point with the characteristics of the PCC voltage and AVC voltage regulation command of wind farms. Summary of the Invention

[0006] The purpose of this invention is to discover the correlation between the PCC voltage at the wind farm's grid connection point and the reactive power demand of the Static Var Generator (SVG) through historical operating data, and then dynamically adjust the optimal reactive power operating point of the SVG based on the real-time PCC voltage, providing practical guidance for reactive power control in wind farms. To achieve the above objective, this invention proposes a method for dynamically adjusting the reactive power operating point of the SVG in wind farms based on Support Vector Regression (SVR), comprising the following steps: Step 1: Enter the offline training phase and acquire historical data of PCC voltage and corresponding SVG voltage regulation reactive power output of the wind farm; Step 2: Standardize the historical data; Step 3: Remove historical data that has little impact on the tuning results; Step 4: Set constraint shrinkage variables; Step 5: Solve for the SVR model parameters; Step 6: Verify whether the reactive power operating point output by the SVR model might exceed the rated capacity range of the SVG when the PCC voltage is within the specified range. If so, reduce the constraint shrinkage variable and return to step 4 to solve again; if not, proceed to step 7. Step 7: Enter the online tuning stage, monitor the PCC voltage in real time and input it into the SVR model to calculate the optimal reactive power operating point of SVG.

[0007] Furthermore, in step 1, obtaining the historical data of the wind farm's PCC voltage and the corresponding SVG voltage regulation reactive power output specifically refers to: Acquiring wind farms M PCC voltage for each consecutive control cycle From 1 to M Number them and define them j =1, 2,…, M . j Moment The corresponding SVG voltage regulation reactive power It refers to j After the AVC updates the voltage command at +1 time, the SVG reactive power output... Q SVG Changes: (1) If missing Q SVGThe measurement data can also be based on Make approximate calculations: (2) In the formula The voltage-reactive power sensitivity of the PCC in a wind farm. for j Shike Wind Farm M The PCC voltage for each continuous control cycle. According to power system analysis theory, the voltage loss between the wind farm PCC and the equivalent infinite grid node can be represented by the longitudinal component of the voltage drop. U Approximate representation: (3) In the formula R and X These are the network-side equivalent resistance and reactance of the PCC port, respectively. P and Q These represent the active power and reactive power output from the wind farm to the grid, respectively. Further partial derivatives of the above equation can be used to calculate... : (4) Furthermore, the per-unit normalization of historical data in step 2 specifically refers to: If the rated PCC voltage is U N The rated capacity of SVG is S N Then calculate and per unit value U PCC and Q AVC The formulas are as follows: (5) (6) The purpose of data standardization is to balance and optimize the numerical scale of variables, improve the numerical characteristics of matrices and accelerate iterative convergence in subsequent solutions; eliminate dimensional interference, which facilitates the design of subsequent constraint shrinkage variables and makes the model universal under different power and voltage levels; separate the scaling effect, and correspond to relevant standards such as voltage upper and lower limits expressed in per-unit values, intuitively showing existing voltage limit exceedance problems and facilitating the visualization analysis of results.

[0008] Furthermore, removing data from historical data that has a relatively small impact on the tuning result in step 3 specifically refers to: To maximize the accuracy of dynamic tuning results, training with as much historical wind farm operation data as possible is recommended. However, a large amount of data also increases the complexity of the model solution, making it more time-consuming and placing certain demands on the computational resources of the wind farm configuration. Considering that in most scenarios, the SVG only performs small voltage adjustments to maintain it near the AVC voltage command value, the requirements for reactive power reserve are relatively low and... U PCC The correlation between these data points is weak, therefore this portion of the data will not have a significant impact on the training results. The SVR model primarily focuses on extreme data points. (Set all...) Q AVC The mean and standard deviation of the data are respectively μ and σ If it exists Then the Q AVC and corresponding U PCC Delete, thus filtering out Q AVC Concentrated μ For most nearby data points, computational efficiency is significantly improved. From 1 to T Renumber the remaining data and define... i =1, 2,…, T .

[0009] Furthermore, in step 4, setting the constraint contraction variable specifically refers to: Set constraint shrinkage variables Optimal reactive power operating point for SVG The corresponding reactive power reserve interval constraint is shrunk. The setting should be based on the fitting effect of the SVR model and should not exceed the rated capacity of the SVG. Specific setting principles are as follows: Smaller... Reducing the reactive power reserve range can improve the accuracy of fitting historical data, but it may also lead to overfitting problems. Follow Significant fluctuations due to changes increase the difficulty of reactive power control in wind farms and may make subsequent SVR problems unsolvable; conversely, larger fluctuations... This will expand the reactive power reserve range and improve the robustness of reactive power control, but the fitting accuracy will decrease accordingly, affecting performance under different operating scenarios. It is likely that the overall performance will remain essentially unchanged, and the final control result will be similar to the static setting of the SVG reactive power operating point, thus weakening the advantage of this invention in improving the economic efficiency of reactive power compensation equipment configuration in wind farms. Therefore, Specific needs should be considered, and a flexible selection should be made that comprehensively considers both security and economy. for per-unit value, This represents the optimal reactive power operating point of the SVG at a certain moment.

[0010] Furthermore, in step 5, solving for the SVR model parameters specifically refers to: Let the optimal reactive power operating point of the SVG at a certain moment be... Q op The reactive power demand for voltage regulation at the next moment is Q AVC To ensure that the reactive power margin of the SVG meets the AVC requirements, the following constraints should apply: (7) (8) (9) In the formula Q L and Q C These are the inductive reactive power reserve and capacitive reactive power reserve of the SVG, respectively. Q tighten The constraint contraction variable set in step 4, and Q tighten Not exceeding the rated capacity of the SVG.

[0011] Analysis of historical operating data from multiple near-shore wind farms in southern China reveals a significant correlation between the PCC voltage at a given moment and the reactive power demand for voltage regulation at the next moment, under extreme operating conditions where the SVG (Variable Reactive Power Control) needs to utilize substantial reactive power reserves to significantly regulate voltage. This is because, in actual dispatching, for safety reasons to avoid voltage exceeding limits, it is impossible to raise or lower voltage commands indefinitely. For example, when the PCC voltage approaches the upper limit of 1.07 pu, the AVC (Automatic Voltage Control) can only maintain or lower the voltage level at the next moment. Furthermore, even if power fluctuations occur, the possibility of the voltage exceeding the lower limit in a short period is extremely low. Therefore, it can be considered that there is almost no capacitive reactive power demand at this time. Traditional static reactive power setting strategies not only retain redundant capacitive reactive power but also limit inductive reactive power reserves. Similarly, when the voltage approaches the lower limit, only capacitive reactive power demand exists. Likewise, when... U PCC When approaching the lower limit, since the risk of exceeding the upper limit in a short period is low, it can be assumed that capacitive reactive power reserves should be primarily maintained at this point. Therefore, it can be considered that... Q SVG =f( U PCC ), and f( U PCCThe correlation is a monotonically increasing function, but unlike the voltage-power relationship in power flow, which can be explicitly characterized by a mathematical model, this correlation originates from the experience of dispatching experts. Without a complete dispatching process model, it is difficult to accurately fit this relationship regardless of the method used. Introducing overly complex algorithms not only incurs additional computational burdens but also carries the risk of overfitting. Therefore, a simple linear function is chosen to fit the relationship between the two: (10) Then seek ω and d The problem can be transformed into a standard SVR model. Based on the data obtained in step 3, solve the following linear programming problem to obtain the SVR parameters. ω and d : (11) (12) (13) (14) In the formula , These are slack variables representing the upper and lower bounds of reactive power, respectively. Their function is to tolerate a small number of over-limits by setting a "soft gap". C These are weighting coefficients, and their function is to correlate with... The importance of balancing the two objective functions of fitting accuracy and model complexity lies in minimizing reactive power exceeding limits while also avoiding... Q op The changes were too drastic, leading to frequent and significant shifts in the reactive power replacement targets of wind farms. and The first i The data includes voltage regulation reactive power demand and PCC voltage.

[0012] Furthermore, in step 6, when verifying that the PCC voltage is within the specified range, whether the reactive power operating point output by the SVR model might exceed the rated capacity range of the SVG specifically refers to: Since historical data cannot cover all scenarios, if Q tighten Setting it too low may lead to overfitting, causing issues with the SVR model output during online tuning. Q op It is possible that this exceeds the rated capacity range of the SVG and cannot be achieved in actual operation; therefore, the SVR model still needs to be verified. Assuming the wind farm's allowable... U PCC The range is [ U min , Umax ], then the optimal reactive power operating point of SVG Q op The range is [ ωU min + d , ωU max + d ],like( ωU min + d )<-1 or ( ωU max + d If )>1, then it is considered Q op This is unreasonable; the SVR model needs to be retrained.

[0013] Furthermore, in step 7, real-time monitoring of the PCC voltage and inputting it into the SVR model to calculate the optimal reactive power operating point of the SVG specifically refers to: Let the real-time value of the PCC voltage be... U rt First, normalize it, then perform online tuning based on the SVR model, that is, calculate it using the following formula. Q op : (15) Compared with the prior art, the present invention has the following advantages: Current wind farm SVG control technology does not set an operating point. The SVG is always in a constant voltage or constant reactive power control mode, adjusting reactive power according to AVC voltage commands and the real-time operating status of the wind farm. Under this control method, the capacitive reactive power of the wind farm's submarine cable, the reactive power loss of the collection network, and large voltage regulation actions can all consume a large amount of the SVG's dynamic reactive power reserve capacity, resulting in insufficient regulation capability. Subsequently, it will face risks such as PCC and turbine-end voltage exceeding limits, slower reactive power regulation speed, and weakened fault response capability, which may lead to penalties from dispatch assessments or even serious faults such as turbine disconnection from the grid. To ensure operational safety, wind farms are forced to configure large-capacity SVGs, which increases construction costs. Traditional static setting methods keep the SVG reactive power operating point at 0. Although this releases the SVG's dynamic reactive power margin to some extent, it ignores the impact of different spatiotemporal conditions on the wind farm's reactive power demand, limiting further optimization of the wind farm's SVG configuration capacity. This invention explores the impact of grid connection location and output characteristics on dispatching strategies based on the characteristics of historical reactive voltage data of different wind farms, dynamically adjusts the appropriate reactive operating point, and reduces the redundancy of capacitive or inductive reactive power reserves while fully utilizing the reactive power resources of wind turbines and ensuring that SVG can still meet the requirements of rapid voltage regulation. Furthermore, by setting different constraint contraction variables, a balance between safety and economy can be achieved, which has greater flexibility. Attached Figure Description

[0014] Figure 1 A schematic diagram illustrating the steps of a dynamic tuning method for the reactive power operating point of an SVR-based wind farm SVG; Figure 2 A comparison of SVG runtime point over-limit situations between the proposed dynamic tuning method and the traditional static tuning method; Figure 3 A comparison of the reactive power boundary of the proposed dynamic tuning method and the traditional static tuning method is presented. Detailed Implementation

[0015] The specific implementation of the present invention will be further described below with reference to the accompanying drawings and examples.

[0016] like Figure 1 As shown, this embodiment of the invention provides a step-by-step flowchart of a method for dynamically setting the reactive power operating point (SVG) of a wind farm based on SVR, including: Step 1: Enter the offline training phase and acquire historical data of the wind farm's PCC voltage and corresponding SVG voltage regulation reactive power output: Acquiring wind farms M PCC voltage for each consecutive control cycle From 1 to M Number them and define them j =1, 2,…, M . j Moment The corresponding SVG voltage regulation reactive power It refers to j After the AVC updates the voltage command at +1 time, the SVG reactive power output... Q SVG Changes: (1) If missing Q SVG The measurement data can also be based on Make approximate calculations: (2) In the formula The voltage-reactive power sensitivity of the PCC in a wind farm. for j Shike Wind Farm M The PCC voltage for each continuous control cycle. According to power system analysis theory, the voltage loss between the wind farm PCC and the equivalent infinite grid node can be represented by the longitudinal component of the voltage drop. U Approximate representation: (3) In the formula Rand X These are the network-side equivalent resistance and reactance of the PCC port, respectively. P and Q These represent the active power and reactive power output from the wind farm to the grid, respectively. Further partial derivatives of the above equation can be used to calculate... : (4) Step 2: Standardize the historical data: If the rated PCC voltage is U N The rated capacity of SVG is S N Then calculate and per unit value U PCC and Q AVC The formulas are as follows: (5) (6) Step 3: Remove historical data that has little impact on the tuning results: Set all Q AVC The mean and standard deviation of the data are respectively μ and σ If it exists Then the Q AVC and corresponding U PCC Delete, thus filtering out Q AVC Concentrated μ For most nearby data points, computational efficiency is significantly improved. From 1 to T Renumber the remaining data and define... i =1, 2, …, T .

[0017] in, To meet the reactive power requirements of voltage regulation, for per-unit value, For wind farm M PCC voltage for one consecutive control cycle M To represent the total number of continuous control cycles, use and They represent the first i The data includes voltage regulation reactive power demand and PCC voltage. iThis is used to reorder and renumber the remaining historical data points so that they can be indexed by this number in subsequent model calculation steps. T This represents the total number of remaining data.

[0018] Step 4: Set constraint shrinkage variables: Set constraint shrinkage variables Optimal reactive power operating point for SVG The corresponding reactive power reserve interval constraint is shrunk. The setting should be based on the fitting effect of the SVR model and should not exceed the rated capacity of the SVG. Specific setting principles are as follows: Smaller... Reducing the reactive power reserve range can improve the accuracy of fitting historical data, but it may also lead to overfitting problems. Follow Significant fluctuations due to changes increase the difficulty of reactive power control in wind farms and may make subsequent SVR problems unsolvable; conversely, larger fluctuations... This will expand the reactive power reserve range and improve the robustness of reactive power control, but the fitting accuracy will decrease accordingly, affecting performance under different operating scenarios. It is likely that the overall performance will remain essentially unchanged, and the final control result will be similar to the static setting of the SVG reactive power operating point, thus weakening the advantage of this invention in improving the economic efficiency of reactive power compensation equipment configuration in wind farms. Therefore, Specific needs should be considered, and a flexible selection should be made that takes into account both safety and economy.

[0019] in for per-unit value, This represents the optimal reactive power operating point of the SVG at a certain moment.

[0020] Step 5: Solve for the SVR model parameters: Let the optimal reactive power operating point of the SVG at a certain moment be... The reactive power demand for voltage regulation at the next moment is To ensure that the reactive power margin of the SVG meets the AVC requirements, the following constraints should apply: (7) (8) (9) In the formula and These are the inductive and capacitive reactive power reserves of the SVG, respectively. The constraint shrinkage variable set in step 4.

[0021] Based on the monotonically increasing relationship between PCC voltage and SVG reactive power output, a linear function is used as the regression function for the SVR model: (10) Then obtain the SVR model parameters. ω and d The problem can be transformed into a standard SVR model. Based on the data obtained in step 3, solve the following linear programming problem to obtain the SVR parameters. ω and d : (11) (12) (13) (14) In the formula , These are slack variables representing the upper and lower bounds of reactive power, respectively. Their function is to tolerate a small number of over-limits by setting a "soft gap". C These are weighting coefficients, and their function is to correlate with... The importance of balancing the two objective functions of fitting accuracy and model complexity lies in minimizing reactive power exceeding limits while also avoiding... Q op The changes were too drastic, leading to frequent and significant shifts in the reactive power replacement targets of wind farms. and The first i The data includes voltage regulation reactive power demand and PCC voltage.

[0022] Step 6: Verify whether the reactive power operating point output by the SVR model might exceed the rated capacity range of the SVG when the PCC voltage is within the specified range. If so, reduce the constraint shrinkage variable and return to step 4 to solve again; if not, proceed to step 7. Assuming the wind farm is permitted U PCC The range is [ U min , U max ], then the optimal reactive power operating point of SVG Q op The range is [ ωU min + d , ωU max + d ],like( ωU min + d )<-1 or ( ωU max + d If )>1, then it is considered Q opThis may exceed the rated capacity range of the SVG.

[0023] Step 7: Enter the online tuning stage, monitor the PCC voltage in real time and input it into the SVR model to calculate the optimal reactive power operating point of the SVG: Let the real-time value of the PCC voltage be... U rt First, normalize it, then perform online tuning based on the SVR model, that is, calculate it using the following formula. Q op : (15) in, This is the rated PCC voltage.

[0024] The following is an application example of the method of the present invention: A 600MW wind farm is connected to the grid via a 220kV submarine cable through a step-up substation. The substation is equipped with a voltage-controlled SVG. For a certain wind farm in the first half of 2023... U PCC and Q AVC Analyze actual data in different... Q tighten The following determinations are made based on the proposed dynamic target tuning method for reactive power substitution and the traditional static tuning method. Q op And calculate the number of historical running points that exceed the limit, i.e., the number of points that do not satisfy constraint (9). N The result is as follows Figure 2 As shown.

[0025] It can be seen that when Q tighten When the voltage is large, both methods can ensure that the SVG has sufficient reactive power regulation capability, but when... Q tighten As the limit is gradually reduced, the static method is the first to start exceeding the limit. Q tighten When the reactive power reserve is reduced to a certain extent, neither method can fully meet the requirements, but the proposed method can still effectively reduce the number of limit violations. To further quantitatively characterize the performance of the two methods, a definition is provided. N Minimum required for =0 Q tighten SVG critical capacity Q cr The calculation method is as follows: Initially... Q tighten Set to 0, and continuously increase in a certain step size. Q tighten Solve the SVR problem multiple times and calculate the corresponding N untilN It decreases until it reaches exactly 0, at which point Q tighten That is Q cr . Q cr The smaller the value, the smaller the SVG capacity required for a wind farm to achieve the same voltage regulation capability. S SVG The smaller it is, the lower its cost. And in the same... S SVG Below, smaller Q cr This allows SVG to retain more dynamic reactive power margin to cope with potential operational risks such as failures.

[0026] Table 1. Different tuning methods Q cr

[0027] As shown in Table 1, the proposed method... Q cr It reduced efficiency by approximately 15.3% compared to the static method. Figure 3 The black dashed line indicates Q tighten The reactive power operating point is set by the proposed method at 48.1 MVar. The black solid line and gray solid line represent the reactive power regulation capability boundaries of the two methods, respectively. It can be seen that the dynamic tuning method can enable all... Q AVC All are within the reactive power regulation capability range, but under the static method, multiple over-limit points as shown in the circles have appeared, which means that the voltage regulation command cannot be completed by a single action of SVG at this time.

[0028] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any modifications, alterations, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for dynamically setting the reactive power operating point of a wind farm based on SVR (Static Var Generator), characterized in that... Includes the following steps: Step 1: Enter the offline training phase and obtain the historical data of PCC voltage and corresponding SVG voltage regulation reactive power output at the grid connection point of the wind farm; Step 2: Standardize the historical data; Step 3: Remove historical data that has little impact on the tuning results; Step 4: Set constraint shrinkage variables; Step 5: Solve for the SVR model parameters; Step 6: Verify whether the reactive operating point output by the SVR model might exceed the rated capacity range of the SVG when the PCC voltage is within the specified range. If so, reduce the constraint shrinkage variable and return to step 4 to solve again; if not, proceed to step 7. Step 7: Enter the online tuning stage, monitor the PCC voltage in real time and input it into the SVR model to calculate the optimal reactive power operating point of SVG.

2. The method for dynamic setting of reactive power operating point of wind farm SVG based on SVR according to claim 1, characterized in that, The specific method for obtaining the historical data of PCC voltage and corresponding SVG voltage regulation reactive power output of the wind farm in step 1 is as follows: Historical PCC voltage data for multiple consecutive control cycles of the wind farm are acquired, and the change in SVG reactive power output after the automatic voltage control (AVC) updates the voltage command is recorded in each control cycle. This change is used as the SVG voltage regulation reactive power output data at the corresponding moment. If the measurement data of SVG reactive power output cannot be obtained directly, an approximate estimate is made based on the change in PCC voltage and the pre-acquired voltage-reactive power sensitivity.

3. The method for dynamic setting of reactive power operating point of wind farm SVG based on SVR according to claim 1, characterized in that, The specific method for standardizing historical data as described in step 2 is as follows: Using the rated PCC voltage as a reference value, the PCC voltage at each moment in the historical data is normalized to obtain dimensionless PCC voltage per-unit values; and using the rated capacity of the SVG as a reference value, the SVG voltage regulation reactive power output at each moment in the historical data is normalized to obtain dimensionless SVG voltage regulation reactive power per-unit values.

4. The method for dynamically setting the reactive power operating point of a wind farm based on SVR according to claim 1, characterized in that, The specific method for removing historical data that has a minor impact on the tuning result, as described in step 3, is as follows: Set all Q AVC The mean and standard deviation of the data are respectively μ and σ If it exists Then Q AVC and corresponding U PCC Delete, from 1 to T Renumber the remaining data and define... i =1, 2, …, T ; in, To meet the reactive power requirements of voltage regulation, for per-unit value, For wind farm M PCC voltage for one consecutive control cycle M To represent the total number of continuous control cycles, use and They represent the first i The data includes voltage regulation reactive power demand and PCC voltage. i This is used to reorder and renumber the remaining historical data points so that they can be indexed by this number in subsequent model calculation steps. T This represents the total number of remaining data.

5. The method for dynamically setting the reactive power operating point of a wind farm based on SVR according to claim 1, characterized in that, The setting of constraint contraction variables mentioned in step 4 specifically refers to: Set constraint shrinkage variables Optimal reactive power operating point for SVG The corresponding reactive power reserve interval constraint is shrunk. It should be set according to the fitting effect of the SVR model, and should not exceed the rated capacity of the SVG.

6. The method for dynamic setting of the reactive power operating point of a wind farm based on SVR according to claim 1, characterized in that, Step 5, which involves solving for the SVR model parameters, specifically refers to: Let the optimal reactive power operating point of the SVG at a certain moment be... The reactive power demand for voltage regulation at the next moment is To ensure that the reactive power margin of the SVG meets the AVC requirements, the following constraints should apply: （1） （2） （3） In the formula and These are the inductive and capacitive reactive power reserves of the SVG, respectively. The constraint shrinkage variable set in step 4.

7. The method for dynamic setting of the reactive power operating point of a wind farm based on SVR according to claim 1, characterized in that, The SVR model parameters are: ω and d， Based on the monotonically increasing relationship between PCC voltage and SVG reactive power output, a linear function is used as the regression function of the SVR model. The per-unit value of PCC voltage is used as the input feature of the SVR model, and the per-unit value of SVG voltage regulation reactive power is used as the output target of the SVR model. The parameters... and The problem of solving this problem is transformed into a standard SVR regression problem.

8. The method for dynamic setting of reactive power operating point of wind farm SVG based on SVR according to claim 1, characterized in that, The solution of SVR model parameters ω and d The specific method is as follows: Solve the following linear programming problem based on the data obtained in step 3 to obtain the SVR parameters. ω and d : （4） （5） （6） （7） In the formula , These are the slack variables representing the upper and lower bounds of reactive power, respectively. C These are the weighting coefficients.

9. The method for dynamic setting of reactive power operating point of wind farm SVG based on SVR according to claim 1, characterized in that, Step 6 involves verifying whether the reactive power operating point output by the SVR model might exceed the rated capacity range of the SVG when the PCC voltage is within the specified range. The specific method is as follows: Assuming the wind farm is permitted U PCC The range is [ U min , U max ], then the optimal reactive power operating point of SVG Q op The range is [ ωU min + d , ω U max + d ],like( ωU min + d )<-1 or ( ωU max + d If )>1, then it is considered Q op This may exceed the rated capacity range of the SVG.

10. The method for dynamic setting of reactive power operating point of wind farm SVG based on SVR according to claim 1, characterized in that, The method for real-time monitoring of the PCC voltage and inputting it into the SVR model to calculate the optimal reactive power operating point of the SVG, as described in step 7, is as follows: The PCC voltage at the grid connection point of the wind farm is collected in real time. The collected PCC voltage is normalized based on the rated voltage and then input into the SVR model whose model parameters have been obtained in the offline training stage. The SVR model calculates and outputs the corresponding SVG optimal reactive power operation point per unit value based on the current input PCC voltage per unit value. This per unit value is then used as the control command of the SVG for execution.

Images