Parameter increment correction method for analytic wake flow model based on working condition scene driving

Abstract

The invention discloses an analytic wake flow model parameter increment correction method based on working condition scene driving, and mainly relates to the field of wind power plant wake flow engineering modeling. Comprising the following steps: S1, establishing a wind power plant whole wake flow model; the method specifically comprises the steps that modeling of a whole-field analysis wake flow model of the wind power plant is achieved through a single-machine wake flow model, a yaw wake flow model and a wake flow superposition model in sequence; s2, generating a wind power plant working condition scene; s3, on-line acquisition and processing of SCADA data are carried out; and S4, adaptively correcting parameters of the analysis wake flow model. The method has the advantages that adaptive correction of the analysis wake flow model in different working condition scenes can be achieved, the depiction deviation of the model on the wake flow diffusion characteristic and the wind speed loss rule in different operation scenes is effectively reduced, the analysis wake flow model can reflect complex and changeable actual working conditions more accurately, and the prediction precision is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of wind farm wake engineering modeling, specifically a method for incremental correction of analytical wake model parameters based on operating scenario-driven conditions. Background Technology

[0002] Wind power has become a key way to alleviate the contradiction between continuously growing energy demand and increasing environmental pressure, and is considered the cornerstone of renewable energy. In recent years, the wind power industry has experienced accelerated growth. However, behind this rapid growth, the levelized cost of wind power remains high. Among the factors hindering the overall economic viability and large-scale development of wind farms is the wake effect between turbines, which leads to increased power generation losses and fatigue loads in downstream turbines. The wake effect, generated by upstream turbines capturing wind energy, causes significant velocity losses and a surge in turbulence intensity in the downstream flow field. Accurately describing the wake effect between turbines is the first step in implementing measures to mitigate the wake effect, such as optimizing wind farm layout and implementing active wake control.

[0003] Regarding wake modeling, methods can be broadly categorized into: Computational Fluid Dynamics (CFD) models, data-driven models, and analytical wake models. CFD models are computationally extremely complex and require substantial resources. Even with the rapid advancements in computational science and significant increases in computing power, building CFD models remains time-consuming, making them unsuitable for research such as wind farm layout optimization and wake control. Data-driven models rely heavily on machine learning or deep learning algorithms based on large datasets, resulting in relatively low physical interpretability and a lack of general applicability. Therefore, analytical wake models have gained significant attention. Analytical wake models primarily utilize wind tunnel experiments or CFD model results, combined with mathematical models built based on certain assumptions, to quickly obtain the velocity and turbulence distributions within the wake field.

[0004] However, existing analytical wake models typically rely on fixed or empirically set model parameters. The parameter values ​​are mostly based on statistical results under ideal operating conditions or limited experimental conditions, making it difficult to reflect the time-varying and scenario-specific differences in operating conditions such as wind speed, wind direction, atmospheric stability, and unit operating status. This results in significant deviations in the model's characterization of wake diffusion characteristics and wind speed deficit patterns under different operating scenarios. Existing parameter correction methods mostly adopt holistic or static adjustment approaches, lacking technical means to perform scenario-specific and incremental corrections as operating conditions change. This makes it difficult to adapt to complex and ever-changing actual operating conditions while maintaining the computational efficiency and structural simplicity of analytical models, thus limiting the effectiveness of analytical wake models in refined analysis and online applications. Summary of the Invention

[0005] The purpose of this invention is to provide an incremental correction method for analytical wake model parameters based on operating conditions. This method can adaptively correct analytical wake models under different operating conditions, effectively reduce the deviation of the model in describing wake diffusion characteristics and wind speed loss patterns under different operating scenarios, and enable analytical wake models to more accurately reflect complex and changing actual operating conditions, thereby significantly improving prediction accuracy.

[0006] To achieve the above objectives, the present invention employs the following technical solution: A method for incremental parameter correction of an analytical wake model driven by operating conditions includes the following steps: S1. Establish a full-field wake model for the wind farm; Specifically, this includes: sequentially using a single-unit wake model, a yaw wake model, and a wake superposition model to achieve full-field analytical wake model modeling for the wind farm; S2. Generate wind farm operating scenario; S3. Online acquisition and processing of SCADA data; S4. Adaptive correction of wake model parameters.

[0007] Furthermore, step S1 is specifically expressed as follows: ; In the formula, The first term, calculated by the analytical wake model, represents the... Inlet air velocity of the kilowatt unit, function The specific form is determined by the corresponding single-aircraft wake model, yaw wake model, and wake superposition model. These represent the inflow wind speed, inflow wind direction, and turbulence intensity of the wind farm, respectively. Indicates the axial thrust coefficient. This represents the model parameters of the corresponding analytical wake model.

[0008] Furthermore, step S2 specifically includes: S21. Based on the analysis of factors affecting wake formation between units, preliminary screening of SCADA data items is conducted to establish an initial feature pool; S22. Sensitivity analysis is used to extract features from the initial feature pool. By quantifying the sensitivity of each environmental parameter and unit operating status parameter to the wake evolution characteristics, key environmental feature parameters and key unit operating status feature parameters generated in the operating scenario are screened and determined. S23. Construct a sample set by collecting key feature SCADA data; S24. The DBSCAN algorithm is used to generate the working condition scenario. The generated scenario is represented as follows: .

[0009] Furthermore, step S3 specifically includes: S31. Real-time acquisition of key feature data via SCADA system. ; S32. Perform routine cleaning to screen outliers in the data collected in step S31, mainly including removing blank values ​​and situations where the rated power exceeds the normal range. S33. Further online cleaning using a Hample filter: First, the real-time acquired data... Build length is A sliding time window is used to calculate the median of the data within the window. , is represented as: , In the formula, median represents the operator for calculating the median of a sequence; S34. Based on the median in step S33 The metrics for the Hample filter are calculated using the following formula. : ; S35. If the current time data A value is considered an outlier if it meets the following conditions: , In the formula, The threshold is preset for the Hample filter; if this condition is not met, it is judged as a normal value.

[0010] Furthermore, the specific process of step S4 includes: S41. Input new SCADA observation data ; S42. Regarding the new observation data in S41 To perform working condition scenario identification, the specific steps are as follows: based on existing working condition scenarios... Calculate the distance between it and the center of the existing working condition scenario, expressed as: ; S43. Determine if a working condition scenario exists. The following neighborhood conditions are true: , In the formula, MinPts is the preset neighborhood radius; If so, then the new observation data will be... Included in working conditions If not, proceed to step S45; otherwise, determine it as a candidate working condition sample and proceed to step S44. S44. Use the DBSCAN algorithm to monitor the candidate working condition samples to determine whether a new working condition scenario can be generated, and re-execute step S41. S45. Determine the working condition scenario If a certain amount of new observation data has been obtained, Bayesian inference will be used to adaptively correct the wake model parameters; specifically as follows: The SCADA observation data will be represented as: , In the formula, For model error, For variance; If not, then repeat step S41.

[0011] Furthermore, the parameters of the analytical wake model are treated as random variables, and their posterior probability can be calculated by the following formula: , In the formula, These are the prior probabilities of the model parameters. The prior probability distribution is obtained through historical experience or other experimental observations and statistics. Under conditions of insufficient confidence, a uniform distribution is adopted according to the "uninformative Bayesian prior criterion," i.e. The likelihood distribution is given by the following formula: ; Finally, the corrected wake model parameters for this operating scenario are obtained through maximum a posteriori estimation, and are expressed as: , In the formula, These are the corrected wake model parameters.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes an incremental correction method for analytical wake model parameters based on operating condition scenarios. This method can automatically identify key operating condition features for different wind conditions and realize incremental and differentiated updates and corrections of model parameters, avoiding the limitations of traditional "one-size-fits-all" parameter calibration. Through the dynamic adjustment and correction of model parameters by this invention, the analytical wake model can more accurately reflect complex and changing actual operating conditions, significantly improving the prediction accuracy of wake velocity field and power loss. Moreover, the method of this invention does not increase model complexity and maintains the high computational efficiency of analytical models, making it suitable for online application scenarios such as real-time wind farm control, power prediction, and layout optimization. Attached Figure Description

[0013] Appendix Figure 1 This is the overall flowchart of the present invention.

[0014] Appendix Figure 2This is a flowchart of the wind farm full-field analytical wake model modeling of the present invention.

[0015] Appendix Figure 3 This is a flowchart of the wind farm operating scenario generation method of the present invention.

[0016] Appendix Figure 4 This is a flowchart of the online SCADA data acquisition and processing of the present invention.

[0017] Appendix Figure 5 This is a flowchart of the adaptive correction of analytical wake model parameters in this invention. Detailed Implementation

[0018] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined in this application.

[0019] Example: A method for incremental parameter correction of an analytical wake model driven by operating conditions, comprising the following steps: S1. Establish a full-field wake model for the wind farm; Combined with appendix Figure 2 As shown, this specifically includes: modeling the entire wind farm's analytical wake model sequentially through a single-unit wake model, a yaw wake model, and a wake superposition model; represented as: , In the formula, The first term, calculated by the analytical wake model, represents the... Inlet air velocity of the kilowatt unit, function The specific form is determined by the corresponding single-aircraft wake model, yaw wake model, and wake superposition model. These represent the inflow wind speed, inflow wind direction, and turbulence intensity of the wind farm, respectively. Indicates the axial thrust coefficient. This represents the model parameters of the corresponding analytical wake model.

[0020] S2. Generate wind farm operating scenario; Combined with appendix Figure 3 As shown, step S2 specifically includes: S21. Based on the analysis of the factors affecting the formation of wake between units, the SCADA data items are initially screened and an initial feature pool is established. The pool mainly includes environmental characteristic parameters such as wind farm inflow wind speed and direction, ambient temperature, and ambient turbulence intensity, as well as unit inflow wind speed, unit output power, and unit pitch angle. S22. Sensitivity analysis is used to extract features from the initial feature pool. By quantifying the sensitivity of each environmental parameter and unit operating status parameter to the wake evolution characteristics, key environmental feature parameters and key unit operating status feature parameters generated in the operating scenario are screened and determined. S23. Construct a sample set by collecting key feature SCADA data; S24. The DBSCAN algorithm is used to generate the working condition scenario. The generated scenario is represented as follows: .

[0021] S3. Online acquisition and processing of SCADA data; Combined with appendix Figure 4 As shown, step S3 specifically includes: S31. Real-time acquisition of key feature data via SCADA system. The data mainly includes environmental characteristic data such as inflow wind speed and direction, ambient temperature, and ambient turbulence intensity of the wind farm, as well as unit operating status characteristic data such as inflow wind speed, unit output power, and unit pitch angle. S32. Perform routine cleaning to screen outliers in the data collected in step S31, mainly including removing blank values ​​and situations where the rated power exceeds the normal range. S33. Further online cleaning using a Hample filter: First, the real-time acquired data... Build length is A sliding time window is used to calculate the median of the data within the window. , is represented as: , In the formula, median represents the operator for calculating the median of a sequence; S34. Based on the median in step S33 The metrics for the Hample filter are calculated using the following formula. : ; S35. If the current data A value is considered an outlier if it meets the following conditions: , In the formula, The threshold is preset for the Hample filter; if this condition is not met, it is judged as a normal value.

[0022] S4. Adaptive correction of wake model parameters; Combined with appendix Figure 5 As shown, the specific process of step S4 includes: S41. Input new SCADA observation data ; S42. Regarding the new observation data in S41 To perform working condition scenario identification, the specific steps are as follows: based on existing working condition scenarios... Calculate the distance between it and the center of the existing working condition scenario, expressed as: , S43. Determine if a working condition scenario exists. The following neighborhood conditions are true: , In the formula, MinPts is the preset neighborhood radius; If so, then the new observation data will be... Included in working conditions If not, proceed to step S45; otherwise, determine it as a candidate working condition sample and proceed to step S44. S44. Use the DBSCAN algorithm to monitor the candidate working condition samples to determine whether a new working condition scenario can be generated, and re-execute step S41. S45. Determine the working condition scenario If a certain amount of new observation data has been obtained, Bayesian inference will be used to adaptively correct the wake model parameters; specifically as follows: The SCADA observation data will be represented as: , In the formula, For model error, For variance; If not, repeat step S41; Furthermore, if we treat the parameters of the analytical wake model as random variables, their posterior probability can be calculated using the following formula: , In the formula, These are the prior probabilities of the model parameters. The prior probability distribution is obtained through historical experience or other experimental observations and statistics. Under conditions of insufficient confidence, a uniform distribution is adopted according to the "uninformative Bayesian prior criterion," i.e. The likelihood distribution is given by the following formula: , Finally, the corrected wake model parameters for this operating scenario are obtained through maximum a posteriori estimation, and are expressed as: , In the formula, These are the corrected wake model parameters.

[0023] This invention proposes an incremental correction method for analytical wake model parameters based on operating conditions. This method can automatically identify key operating condition features for different wind conditions and realize incremental and differentiated updates and corrections of model parameters. By dynamically adjusting and correcting model parameters through this invention, the analytical wake model can more accurately reflect complex and changing actual operating conditions, significantly improving the prediction accuracy of wake velocity field and power loss. It is suitable for online application scenarios such as real-time wind farm control, power prediction, and layout optimization.

Claims

1. A method for incremental parameter correction of an analytical wake model based on operating condition scenario-driven approach, characterized in that: Includes the following steps: S1. Establish a full-field wake model for the wind farm; Specifically, this includes: sequentially using a single-unit wake model, a yaw wake model, and a wake superposition model to achieve full-field analytical wake model modeling for the wind farm; S2. Generate wind farm operating scenario; S3. Online acquisition and processing of SCADA data; S4. Adaptive correction of wake model parameters.

2. The method for incremental correction of analytical wake model parameters based on working condition scenario driving according to claim 1, characterized in that: Step S1 is specifically represented as follows: ； In the formula, The first term calculated by the analytical wake model represents the... Inlet air velocity of the kilowatt unit, function The specific form is determined by the corresponding single-aircraft wake model, yaw wake model, and wake superposition model. These represent the inflow wind speed, inflow wind direction, and turbulence intensity of the wind farm, respectively. Indicates the axial thrust coefficient. This represents the model parameters of the corresponding analytical wake model.

3. The method for incremental correction of analytical wake model parameters based on working condition scenario driving according to claim 1, characterized in that: Step S2 specifically includes: S21. Based on the analysis of factors affecting wake formation between units, preliminary screening of SCADA data items is conducted to establish an initial feature pool; S22. Sensitivity analysis is used to extract features from the initial feature pool. By quantifying the sensitivity of each environmental parameter and unit operating status parameter to the wake evolution characteristics, key environmental feature parameters and key unit operating status feature parameters generated in the operating scenario are screened and determined. S23. Construct a sample set by collecting key feature SCADA data; S24. The DBSCAN algorithm is used to generate the working condition scenario. The generated scenario is represented as follows: 。 4. The method for incremental correction of analytical wake model parameters based on working condition scenario driving according to claim 1, characterized in that: Step S3 specifically includes: S31. Real-time acquisition of key feature data via SCADA system. ; S32. Perform routine cleaning to screen outliers in the data collected in step S31, mainly including removing blank values ​​and situations where the rated power exceeds the normal range. S33. Further online cleaning using a Hample filter: First, the real-time acquired data... Build length is A sliding time window is used to calculate the median of the data within the window. , is represented as: ， In the formula, median represents the operator for calculating the median of a sequence; S34. Based on the median in step S33 The metrics for the Hample filter are calculated using the following formula. : ； S35. If the current time data A value is considered an outlier if it meets the following conditions: ， In the formula, The threshold is preset for the Hample filter; if this condition is not met, it is judged as a normal value.

5. The method for incremental correction of analytical wake model parameters based on working condition scenario driving according to claim 1, characterized in that: The specific process of step S4 includes: S41. Input new SCADA observation data ; S42. Regarding the new observation data in S41 To perform working condition scenario identification, the specific steps are as follows: based on existing working condition scenarios... Calculate the distance between it and the center of the existing working condition scenario, expressed as: ； S43. Determine if a working condition scenario exists. The following neighborhood conditions are true: ， In the formula, MinPts is the preset neighborhood radius; If so, then the new observation data will be... Included in working conditions If not, proceed to step S45; otherwise, determine it as a candidate working condition sample and proceed to step S44. S44. Use the DBSCAN algorithm to monitor the candidate working condition samples to determine whether a new working condition scenario can be generated, and re-execute step S41. S45. Determine the working condition scenario If a certain amount of new observation data has been obtained, Bayesian inference will be used to adaptively correct the wake model parameters; specifically as follows: The SCADA observation data will be represented as: ， In the formula, For model error, For variance; If not, repeat step S41.

6. The method for incremental correction of analytical wake model parameters based on working condition scenario driving according to claim 5, characterized in that: Furthermore, if we treat the parameters of the analytical wake model as random variables, their posterior probability can be calculated using the following formula: ， In the formula, These are the prior probabilities of the model parameters. The prior probability distribution is obtained through historical experience or other experimental observations and statistics. Under conditions of insufficient confidence, a uniform distribution is adopted according to the "uninformative Bayesian prior criterion," i.e. The likelihood distribution is given by the following formula: ； Finally, the corrected wake model parameters for this operating scenario are obtained through maximum a posteriori estimation, and are expressed as: ， In the formula, These are the corrected wake model parameters.

Images