Wind power plant wake flow cooperative control method based on double-layer architecture model

Abstract

The invention relates to a wind power plant wake flow cooperative control method based on a double-layer architecture model, and the method comprises the steps: taking the optimal energy capture efficiency of a divided region as a control target, constructing a wind power plant multi-station cooperative control model, and dynamically adjusting a target unit to operate according to an optimal attitude according to the real-time wind resource of a wind power plant and the operation state of each unit. And the wake flow influence between units is reduced. Firstly, according to unit position coordinates, a DBSCAN clustering method is utilized to cluster wind power plant fans, then a dynamic clustering algorithm based on the wind direction is provided, and the algorithm is based on a wind direction correlation curve, a distance threshold is given, long-distance sample points are removed in a self-adaptive mode, and therefore accurate segmentation and clustering of data are achieved. According to the method, wind power plant wake flow cooperative control with irregular layout can be realized, the fan dimension and the calculation difficulty are reduced, the wake flow control algorithm efficiency is improved, and the method is an innovation of the wind power plant wake flow cooperative control method.

Description

Technical Field

[0001] The present invention relates to the field of wind farm level control, and in particular to a wind farm wake coordinated control method based on a double-layer architecture model. Background Art

[0002] The research objects of wind farm wake coordinated control are mostly plain wind farms and offshore wind farms. There are two main methods used: one is the optimization method based on the mechanism wake model, which uses exhaustive method, particle swarm algorithm and genetic algorithm to find the optimal solution of the wind farm output power optimization problem, but the wake effect is extremely complex, and the mechanism-based wake analytical model is difficult to accurately describe it; the second method is based on data-driven optimization method, typical methods such as hill climbing algorithm, etc. This type of method has a slow convergence speed and is difficult to achieve fast online application. Therefore, its improvement and innovation are imperative. Summary of the invention

[0003] In view of the above situation, in order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a wind farm wake coordinated control method based on a two-layer architecture model. For wind farms with irregular layouts, the wake coordinated control can be effectively realized, the wind turbine dimension and calculation difficulty can be reduced, and the efficiency of the wake control algorithm can be improved.

[0004] The technical solution provided by the present invention is:

[0005] A wind farm wake coordinated control method based on a two-layer architecture model comprises the following steps:

[0006] S1. Extract the actual position coordinates of each unit according to the actual position distribution map of the wind farm;

[0007] S2. Set the wind turbine coordinate set to E = (x 1 ,x 2 ,...x m ), where x i is the coordinate of the i-th unit;

[0008] S3. Set parameters (ε, MinPts) to describe the compactness of the distribution of wind turbine units in the neighborhood, where ε is the neighborhood distance threshold of a certain unit, and MinPts describes the threshold of the number of units in the neighborhood with a distance of ε;

[0009] S4, clustering the wind turbines in the wind farm according to the location coordinates using the DBSCAN clustering method;

[0010] S5. According to the clustering results, the entire wind turbine is divided into n groups, where category -1 represents that the wind turbine is marked as an isolated point, and category 0-(n-2) represents n-1 groups of turbines that meet the clustering conditions;

[0011] S6. Construct the reference curve:

[0012] For a given set of coordinates, select the leftmost unit coordinate of the set as the starting point and generate a straight line with the same slope as the wind direction. Assume that the wind direction vector is v = [v x , v y ], then the curve equation can be expressed as Where (x start ,y start ) is the coordinate of the starting unit;

[0013] S7, Dynamic distance iteration:

[0014] Calculate the average distance D from the coordinates of the remaining units except the starting point of the unit to the reference curve. If D exceeds the distance threshold, iterate and perform the following steps:

[0015] a. Remove distant points: remove the unit coordinate point farthest from the reference curve;

[0016] b. Distance assessment: Recalculate the average distance from the remaining unit coordinates to the reference curve. The distance formula is:

[0017]

[0018] Where: N is the number of remaining unit coordinates, Pi is the coordinate of the ith unit of the remaining units, L is the reference curve, and D is the draw distance;

[0019] c. Classification marking: When the distance is less than the critical distance, mark all the remaining unit coordinates as the same category; if the distance is greater than the critical distance, continue to execute ac;

[0020] S8, starting point update: take the unmarked leftmost unit coordinate point as the new starting point, and repeat step S7 until all points are classified, that is, dynamic clustering based on wind direction is completed;

[0021] S9, coupling and classifying the DBSCAN clustering of all wind turbines in the field with the dynamic clustering results based on wind direction;

[0022] S10, for the same coupled cluster unit, in order to accurately determine whether the controlled fan is an upstream fan, the fan position is mapped to a coordinate system with the current wind direction as the x-axis;

[0023] S11, Assume that the coordinates of the fan P are (x p ,y p ), the current wind direction angle is θ P , the coordinates of the origin of the coordinate system are (0,0), then the coordinates of the fan P in the new coordinate system are:

[0024] X p =x p cosθp -y p sinθ p

[0025] Y p =x p sinθ p +y p cosθ p

[0026] S12, adopting a strategy of grouping fans one by one to establish the relationship between upstream and downstream fans; sorting the fans by size according to the transformed x-axis coordinate values, with smaller coordinate values ​​being upstream fans and larger coordinate values ​​being downstream fans, taking full account of the relative positions between fans to form an accurate fan grouping structure;

[0027] S13, the wind farm wake coordinated control model adopts the SCADA data based on the unit operation, and is executed with a single group of upstream and downstream wind turbines as the target, so that the grouped wind turbines can obtain the best energy capture efficiency as the control target;

[0028] S14. Data acquisition:

[0029] Collect the unit SCADA historical operation data and unit guaranteed power curve. The collected historical operation data include the output power, wind speed, wind angle, pitch angle, generator speed and grid connection information of the wind turbine unit;

[0030] S15. Data preprocessing

[0031] S16. Data Integration:

[0032] The operation data of the upstream and downstream wind turbines are integrated according to the time series, and the active power of the wind turbines is accumulated to form the overall time series:

[0033] P tot (t) = P up (t)+P down (t)

[0034] S17. Data warehouse:

[0035] After pre-processing, the operating data is divided into wind speed intervals at intervals of 0.5 m / s between the cut-in and cut-out wind speeds, and then the wind speed intervals are further divided into wind angle sub-intervals at 2° wind angle intervals;

[0036] S18. Find the optimal interval:

[0037] Calculate the power average of the operating data in a single compartment, and find the wind angle corresponding to the maximum power performance index at different wind speeds through clustering and fitting.

[0038] S19, upstream fan wind correction: For the same coupled cluster unit group, for two adjacent units, the upstream unit wind angle value when the sum of the power of the upstream unit and the downstream unit is optimal is calculated through S18, and the wind angle value is transmitted to the upstream unit main control system through communication. The wind angle obtained by the upstream unit main control system is subtracted from the wind angle value transmitted by communication.

[0039] For example, suppose there are three units A, B, and C, A is the upstream unit of B, B is the upstream unit of C, and when the sum of the power of units A and B is optimal, the wind angle value e1 of the upstream unit A is calculated by S18, and e1 is transmitted to the upstream main control system of unit A through communication, and the wind angle of the main control system of unit A is subtracted from e1; when the sum of the power of units B and C is optimal, the wind angle value e2 of the upstream unit B is calculated by S18, and e2 is transmitted to the upstream main control system of unit B through communication, and the wind angle of the main control system of unit B is subtracted from e2;

[0040] Preferably, the data preprocessing specifically includes:

[0041] a. Eliminate the operation data during downtime such as maintenance and failure;

[0042] b. Eliminate the operating data of wind turbines with limited power;

[0043] c. Eliminate the operation data within 120 seconds after grid connection and in non-grid connection state;

[0044] d. Eliminate the operating data outside the cabin position ±900°;

[0045] e. Eliminate the operating data other than the normal pitch angle;

[0046] f. Eliminate operating data below the cut-in wind speed and above the rated wind speed.

[0047] The present invention adopts a wind farm wake collaborative control method based on a two-layer architecture model, takes the optimal energy capture efficiency of the divided area as the control target, constructs a wind farm multi-machine collaborative control model, and dynamically adjusts the target unit to operate in the optimal posture according to the real-time wind resources of the wind farm and the operating status of each unit to reduce the wake impact between units. First, according to the unit position coordinates, the wind farm wind turbines are clustered using the DBSCAN clustering method, and then a dynamic clustering algorithm based on wind direction is proposed. The algorithm is based on the wind direction association curve and adaptively removes long-distance sample points by giving a distance threshold, thereby realizing accurate segmentation and clustering of data. The algorithm couples the wind direction with the clustering reference curve, dynamically adjusts the clustering criteria, has efficient and adaptive clustering capabilities, and is suitable for data sets with obvious wind direction association. Finally, for the same coupled clustering unit, a wind turbine grouping strategy is adopted to establish the relationship between upstream and downstream wind turbines. A single group of upstream and downstream wind turbines is executed as the target, so that the grouped wind turbines obtain the best energy capture efficiency. This method can realize coordinated control of wakes in wind farms with irregular layouts, reduce wind turbine dimensions and computational difficulty, and improve the efficiency of the wake control algorithm. It is an innovation in the coordinated control method of wind farm wakes. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the wind farm wake coordinated control process based on the double-layer architecture model of the present invention.

[0049] Figure 2 This is the DBSCAN clustering result of 20 wind turbines in a wind farm according to an embodiment of the present invention.

[0050] Figure 3 This is a dynamic clustering diagram based on wind direction of 20 wind turbines in a wind farm according to an embodiment of the present invention.

[0051] Figure 4 Schematic diagram of wind turbine coordinate transformation according to an embodiment of the present invention.

[0052] Figure 5 This is a coordinate transformation diagram based on wind direction for 20 wind turbines in a wind farm according to an embodiment of the present invention.

[0053] Figure 6 Schematic diagram of dynamic adjustment of a fan according to an embodiment of the present invention.

[0054] Figure 7 Schematic diagram of power generation improvement percentage according to an embodiment of the present invention.

[0055] Figure 8 This is a wind angle-power fitting diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0057] See also Figure 1 , a wind farm wake coordinated control method based on a two-layer architecture model in this embodiment includes the following steps:

[0058] S1. Taking a wind farm in Henan as an example, the wind farm is a hilly wind farm with irregular layout of units. The actual position coordinates of each unit are extracted;

[0059] S2. Set the wind turbine coordinate set to E = (x 1 ,x 2 ,...x m ), where x i is the coordinate of the i-th unit;

[0060] S3. Set parameters (ε, MinPts) to describe the compactness of the distribution of wind turbine units in the neighborhood. Calculate the distance from each sample point to its kth nearest neighbor and sort the distances. Select ε=300 and MinPts=2 based on the distance change.

[0061] S4, clustering the wind turbines in the wind farm using the DBSCAN clustering method according to the unit location coordinates;

[0062] S5. According to the clustering results, the DBSCAN clustering results for 20 wind turbines in a wind farm are as follows: Figure 2 As shown, the entire wind turbine is divided into 6 groups, where category -1 represents that the wind turbine is marked as an isolated point, and categories 0-4 represent 5 groups of turbines that meet the clustering conditions;

[0063] S6. Construct the reference curve: For a given set of coordinates, select the coordinate of the leftmost unit in the set as the starting point, and generate a straight line with the same slope as the wind direction. Assume that the wind direction vector is v = [v x , v y ], then the curve equation can be expressed as Where (x start ,y start ) is the unit coordinate; taking the wind direction slope 3 as an example, execute steps S7-S8, and the clustering results are shown in Figure 3 .

[0064] S7, calculating the average distance D from the coordinates of the remaining units except the starting point of the unit to the reference curve, if D exceeds the distance threshold, iteratively performing the following steps:

[0065] a. Remove distant points: remove the unit coordinate point farthest from the reference curve;

[0066] b. Distance assessment: Recalculate the average distance from the remaining unit coordinates to the reference curve. The distance formula is:

[0067]

[0068] Where: N is the number of remaining unit coordinates, Pi is the coordinate of the ith unit of the remaining units, L is the reference curve, and D is the draw distance;

[0069] c. Classification marking: When the distance is less than the critical distance, mark all the remaining unit coordinates as the same category; if the distance is greater than the critical distance, continue to execute ac;

[0070] S8, starting point update: take the unmarked leftmost unit coordinate point as the new starting point, and repeat step S7 until all points are classified; the clustering results of 20 wind turbines in a wind farm are as follows: Figure 3 As shown in the figure, the entire wind turbine is divided into 11 groups, among which category EJ is a single unit that does not meet the wind direction clustering condition.

[0071] S9. Couple and classify the DBSCAN clustering of all wind turbines in the field with the dynamic clustering results based on wind direction (as shown in Table 1); wind turbines 1-5 belong to the same coupling cluster 0-A, wind turbines 6-8 belong to the same coupling cluster 0-B, wind turbines 11-12 belong to the same coupling cluster 1-C, and wind turbines 13-14 belong to the same coupling cluster 2-D. Under the specified wind direction, the wind turbines in the same coupling cluster need to be coordinated with the wake control.

[0072] Table 1 Clustering of different fan coupling

[0073]

[0074]

[0075] S10. For the same coupled cluster unit, in order to accurately determine whether the controlled fan is the upstream fan, the fan position needs to be mapped to a coordinate system with the current wind direction as the x-axis, such as Figure 4 ;

[0076] S11, Assume that the coordinates of the fan P are (x p ,y p ), the current wind direction angle is θ P , the coordinates of the origin of the coordinate system are (0,0), then the coordinates of the fan P in the new coordinate system are:

[0077] X p =x p cosθ p -y p sinθ p

[0078] Y p =x p sinθ p +y p cosθp

[0079] S12. Adopt the strategy of grouping wind turbines one by one to establish the relationship between upstream and downstream wind turbines. Sort the wind turbines by size according to the transformed x-axis coordinate values, with smaller coordinate values ​​as upstream wind turbines and larger coordinate values ​​as downstream wind turbines. Take full account of the relative positions between wind turbines to form an accurate wind turbine grouping structure. The coordinate transformation results of 20 wind turbines in a wind farm are as follows: Figure 5 As shown, based on the result, the wind turbines under the same coupling cluster are further marked as a single group of wind turbine control groups (upstream wind turbine-downstream wind turbine). For example, under the current wind direction, wind turbine No. 5 is marked as the upstream wind turbine, and wind turbine No. 4 is the corresponding downstream wind turbine; and when wind turbine No. 4 is the upstream wind turbine, wind turbine No. 3 is the corresponding downstream wind turbine.

[0080] S13, the wind farm wake coordinated control model adopts the SCADA data based on the unit operation, and is executed with a single group of upstream and downstream wind turbines as the target, so that the grouped wind turbines can obtain the best energy capture efficiency as the control target;

[0081] S14, data acquisition. Collect the unit SCADA historical operation data and unit guaranteed power curve. The collected historical operation data should cover the output power, wind speed, wind angle, pitch angle, generator speed and grid connection information of the wind turbine unit;

[0082] S15. Data preprocessing: To ensure the objectivity and rationality of wind deviation analysis, the collected historical operation data of wind turbines should be preprocessed:

[0083] a. Eliminate the operation data during downtime such as maintenance and failure;

[0084] b. Eliminate the operating data of wind turbines with limited power;

[0085] c. Eliminate the operation data within 120 seconds after grid connection and in non-grid connection state;

[0086] d. Eliminate the operating data outside the cabin position ±900°;

[0087] e. Eliminate the operating data other than the normal pitch angle;

[0088] f. Eliminate the operating data below the cut-in wind speed and above the rated wind speed;

[0089] S16, data integration: integrating the operation data of the upstream and downstream wind turbines according to the time series, and accumulating the active power of the wind turbines to form an overall time series;

[0090] S17, data binning: the pre-processed operating data is divided into wind speed intervals at intervals of 0.5 m / s between the cut-in and cut-out wind speeds, and then the wind speed intervals are further divided into wind angle sub-intervals at 2° wind angle intervals;

[0091] S18, Find the optimal interval: Calculate the power average of the operating data in a single sub-bin, and find the wind angle corresponding to the maximum power performance index under different wind speeds through clustering, fitting, etc. Use the deterministic power curve to fit the power curve of each sub-data set, and use the cubic polynomial regression algorithm, such as Figure 8 .

[0092] S19, upstream wind turbine wind correction. Figure 5 Taking the coupled cluster 0-B unit 6-8 subgroup as an example, unit 6 is the upstream unit of unit 8, and unit 8 is the upstream unit of unit 7. After calculation, the wind angles are reduced by -3 degrees, 4 degrees, and 0 degrees respectively.

[0093] After the coordinated control of the wind farm wake, the coupled clustering of units 1-5 is analyzed as an example. At the cost of a 1.06% loss of power generation from unit 5, the power generation of unit 1 increased by 7.3%, and units 2-4 also increased to varying degrees. Figure 7 , effectively reducing the power loss of wind turbines caused by wake turbulence and increasing the overall power generation of the five units.

Claims

1. A wind farm wake coordinated control method based on a two-layer architecture model, characterized in that: The following steps are involved: S1. Extract the actual position coordinates of each unit according to the actual position distribution map of the wind farm; S2, set the wind turbine coordinate set to E = (x1, x2, ... x m ), where x i is the coordinate of the i-th unit; S3. Set parameters (ε, MinPts) to describe the compactness of the distribution of wind turbine units in the neighborhood, where ε is the neighborhood distance threshold of a certain unit, and MinPts describes the threshold of the number of units in the neighborhood with a distance of ε; S4, clustering the wind turbines in the wind farm according to the location coordinates using the DBSCAN clustering method; S5. According to the clustering results, the entire wind turbine is divided into n groups, where category -1 represents that the wind turbine is marked as an isolated point, and category 0-(n-2) represents n-1 groups of turbines that meet the clustering conditions; S6. Construct the reference curve: For a given set of coordinates, select the leftmost unit coordinate of the set as the starting point and generate a straight line with the same slope as the wind direction. Assume that the wind direction vector is v = [v x , v y ], then the curve equation can be expressed as Where (x start ,y start ) is the coordinate of the starting unit; S7, Dynamic distance iteration: Calculate the average distance D from the coordinates of the remaining units except the starting point of the unit to the reference curve. If D exceeds the distance threshold, iterate and perform the following steps: a. Remove distant points: remove the unit coordinate point farthest from the reference curve; b. Distance assessment: Recalculate the average distance from the remaining unit coordinates to the reference curve. The distance formula is: Where: N is the number of remaining unit coordinates, Pi is the coordinate of the ith unit of the remaining units, L is the reference curve, and D is the draw distance; c. Classification marking: When the distance is less than the critical distance, mark all the remaining unit coordinates as the same category; if the distance is greater than the critical distance, continue to execute ac; S8, starting point update: take the unmarked leftmost unit coordinate point as the new starting point, and repeat step S7 until all points are classified, that is, dynamic clustering based on wind direction is completed; S9, coupling and classifying the DBSCAN clustering of all wind turbines in the field with the dynamic clustering results based on wind direction; S10, for the same coupled cluster unit, in order to accurately determine whether the controlled fan is an upstream fan, the fan position is mapped to a coordinate system with the current wind direction as the x-axis; S11, Assume that the coordinates of the fan P are (x p ,y p ), the current wind direction angle is θ P , the coordinates of the origin of the coordinate system are (0,0), then the coordinates of the fan P in the new coordinate system are: X p =x p cosθ p -y p sinθ p Y p =x p sinθ p +y p cosθ p S12, adopting a strategy of grouping fans one by one to establish the relationship between upstream and downstream fans; sorting the fans by size according to the transformed x-axis coordinate values, with smaller coordinate values ​​being upstream fans and larger coordinate values ​​being downstream fans, taking full account of the relative positions between fans to form an accurate fan grouping structure; S13, the wind farm wake coordinated control model adopts the SCADA data based on the unit operation, and is executed with a single group of upstream and downstream wind turbines as the target, so that the grouped wind turbines can obtain the best energy capture efficiency as the control target; S14. Data acquisition: Collect the unit SCADA historical operation data and unit guaranteed power curve. The collected historical operation data include the output power, wind speed, wind angle, pitch angle, generator speed and grid connection information of the wind turbine unit; S15. Data preprocessing S16. Data Integration: The operation data of the upstream and downstream wind turbines are integrated according to the time series, and the active power of the wind turbines is accumulated to form the overall time series: P tot (t)=P up (t)+P down (t) S17. Data warehouse: After pre-processing, the operating data is divided into wind speed intervals at intervals of 0.5 m / s between the cut-in and cut-out wind speeds, and then the wind speed intervals are further divided into wind angle sub-intervals at 2° wind angle intervals; S18. Find the optimal interval: Calculate the power average of the operating data in a single compartment, and find the wind angle corresponding to the maximum power performance index at different wind speeds through clustering and fitting. S19, upstream fan wind correction: For the same coupled cluster unit group, for two adjacent units, the upstream unit wind angle value when the sum of the power of the upstream unit and the downstream unit is optimal is calculated through S18, and the wind angle value is transmitted to the upstream unit main control system through communication. The wind angle obtained by the upstream unit main control system is subtracted from the wind angle value transmitted by communication.

2. The wind farm wake coordinated control method based on the double-layer architecture model according to claim 1 is characterized in that: In order to ensure the objectivity and rationality of wind deviation analysis, the collected historical operation data of wind turbines should be preprocessed. The data preprocessing specifically includes: a. Eliminate the operation data during downtime such as maintenance and failure; b. Eliminate the operating data of wind turbines with limited power; c. Eliminate the operation data within 120 seconds after grid connection and in non-grid connection state; d. Eliminate the operating data outside the cabin position ±900°; e. Eliminate the operating data other than the normal pitch angle; f. Eliminate operating data below the cut-in wind speed and above the rated wind speed.

Images