A wind power cluster active power sub-cluster optimization scheduling method

Abstract

This invention proposes an optimized scheduling method for active power clusters. By analyzing the correlation between wind farms and considering factors such as the control capability score of wind turbine units, it achieves the goal of balancing the power generation of each wind farm and effectively reduces the number of wind farm adjustments. Cluster control is divided into two levels: the cluster level and the individual farm level. Individual farms within the cluster are grouped according to classification indicators. Several individual farms under the same indicator are grouped into one group, and within each group, several individual farms are further ranked according to their indicators. Finally, the wind farms are controlled in an orderly manner based on the grouping and ranking. The scheduling method of this invention reduces the number of wind farm adjustments. Unlike proportional allocation, which adjusts all wind farms and units when cluster demand changes (i.e., power increase or decrease), the proposed method only adjusts a portion of the wind farms, reducing the number and frequency of wind farm adjustments and effectively minimizing operational fatigue damage.

Description

Technical Field

[0001] This invention belongs to the field of wind power cluster optimization scheduling, and particularly relates to an active power optimization scheduling method based on cluster segmentation. Background Technology

[0002] Currently, there is no strict definition of indicators for wind farm clustering. Generally, it can be summarized as a group of several wind farms that are close to each other and have complementary relationships in terms of geographical location or grid structure. Integrating wind farm clusters into the main power grid allows for flexible control of wind power output within the cluster and full utilization of wind resources, enabling friendly scheduling and control of output power. To address the issue of bias caused by incomplete future output power information when grouping wind farms based on a single-point value at a certain moment, a dynamic grouping strategy based on power change trends is designed. This strategy fully considers that the output power information of a wind farm cluster within a fixed time window is not the output power of a single time period, and the dynamic grouping results are applicable to all wind farms within the selected time period. Furthermore, the wind farm load factor is defined. For wind farms within the same cluster, they are sorted according to their load factors, and scheduling is performed based on different load factors to balance the power generation of each wind farm. Summary of the Invention

[0003] Objective: To address the issue of bias caused by incomplete future output power information when grouping wind farms based on single-point values ​​at a given moment, this invention designs a dynamic wind farm grouping strategy based on power change trends. This strategy fully considers that the output power information of a wind farm cluster within a fixed time window is not the output power of a single time period. The dynamic grouping results are applicable to all wind farms within the selected time period. Furthermore, the invention defines wind farm load factors. For wind farms within the same cluster, they are sorted according to their load factors, and scheduling is performed based on different load factors to balance the power generation of each wind farm.

[0004] Technical Solution: To achieve the above-mentioned objectives, this invention proposes a method for optimized scheduling of active power clusters in wind power generation, which includes the following steps:

[0005] 1) Determine the control objectives and constraints, and impose constraints on wind farm output, wind farm output fluctuations, and cluster power tracking errors;

[0006] 2) Define the clustering index of wind farms and the scheduling order of wind farms within the same wind farm cluster. The power prediction value of wind farms from time t to the next 4 cycles is used as the basis for judging the power change trend of wind farms. Based on the change trend, i.e. the ramp rate, the wind farms are divided into three groups: the up-climbing group, the transition group, and the down-climbing group. Within each group, the wind farms are sorted according to the wind farm load rate. The number of wind farms is N.

[0007] 3) Based on the difference between the current output and the grid-connected power demand, combined with the cluster grouping order, the wind farm load rate ranking within the same cluster, and the wind farm power forecast, the cluster power increase / decrease is determined and scheduled according to the grouping.

[0008] Furthermore, step 1) specifically includes the following steps:

[0009] (1) Control objectives

[0010] Determine the cluster grid connection requirements P r The size serves as a target value to constrain cluster power tracking error and to guide the cluster in tracking grid connection requirements.

[0011] (2) Constraints

[0012] (2.1) Wind farm output constraints:

[0013] P i min <P i ref <P i for

[0014] Among them, P i min P i ref P i for Let P be the minimum dispatch power, dispatch power demand, and predicted power for wind farm i. i min =0.3P i for ;

[0015] (2.2) Wind farm output fluctuation constraints

[0016] Constraints on wind farm output fluctuations are applied by limiting the output fluctuation over a 10-minute period:

[0017] |P i real -P i ref |≤P i rate / 3

[0018] Among them, P i real For the actual power output of wind farm i, P i rate Let i be the installed capacity of wind farm i;

[0019] (2.3) Cluster power tracking error constraint: the actual cluster output error with Pr shall not exceed 1%Pr.

[0020] Furthermore, step 2) includes:

[0021] In the dynamic clustering process, the predicted power value of the wind farm from time t to the next four cycles is used as the basis for judging the power change trend of the wind farm. Based on the change trend, the wind farm ramping classification index value K is calculated. i The fields are grouped, but the groups are not fixed. They are regrouped according to the indicators each period.

[0022] The classification indexes for wind farm ramping are defined as follows:

[0023]

[0024] Where sign() is the sign function; The predicted power value of the wind farm in the nth future cycle; P i real This is the power output value of the wind farm at time t;

[0025] For K i ∈[-4,4], according to K i Wind farm groups are classified into uphill groups, transition groups, and downhill groups. Among them, the uphill group index K... i The value is 4, characterized by a continuous increase in predicted wind farm power; the transition group index K i The range is from -4 to 4, characterized by fluctuations in predicted wind farm power; the downhill group index K i The value is -4, characterized by a continuous decrease in the predicted power of wind farms;

[0026] Within the same wind farm cluster, the scheduling order of wind farms is determined based on their load factors, where the wind farm load factor δ is the load factor. i Defined as the ratio of the actual power generated by each wind farm to its installed capacity:

[0027]

[0028] In the formula, P i rate Let i be the installed capacity of wind farm i;

[0029] When it is necessary to increase the power of the wind farm cluster, priority should be given to increasing the power of wind farms with load rates below the first threshold within the cluster; when it is necessary to decrease the power of the wind farm cluster, priority should be given to reducing the power of wind farms with load rates above the second threshold within the cluster, so as to balance the power output of wind farms within the cluster.

[0030] Furthermore, step 3) includes the following steps:

[0031] (1) Compare the current cluster output power with the cluster demand P r Determine whether the wind farm needs to increase or decrease its power generation (DP) based on its current output.

[0032]

[0033] Among them, P i real Let n be the current power generation of each wind farm, and n be the number of wind farms in the current cluster.

[0034] (2) Determine whether the cluster needs to increase or decrease its power generation;

[0035] (2.1) If the cluster needs to increase power generation, i.e. DP<=0, then compare the current output power of each wind farm with the predicted power to determine the power increase potential or power reduction amount dP for each wind farm. i :

[0036] dP i =P i real -P i for

[0037] Because there are wind farms whose predicted output is less than the current output, i.e., dP i >0, assuming there are m wind farms, the required increase in power generation for the other Nm wind farms is:

[0038]

[0039] dP T To predict the increase in power generation required for a wind farm with an output not less than its current output, dP j This represents the power change of a wind farm whose predicted output is less than its current output.

[0040] Using the clustering index K in step 2 i The allocation is based on the load factor of the wind farms within the cluster. The order of grouping and upgrading at the wind farm cluster level is from the top-climbing group to the transition group, i.e., K. i From largest to smallest, wind farms within each group are sorted and allocated according to their load factor from lowest to highest:

[0041] Step 1: Starting from the first wind farm, allocate the power that needs to be cluster - enhanced to each wind farm in sequence from the upper - climbing group to the transition group according to the increasing order of the clustering index;

[0042] Step 2: If dP i > 0, at this time the predicted output of the wind farm decreases, then the scheduling demand P i ref of wind farm i = P i for ;

[0043] Step 3: If dP T <= dP i , the increased output predicted by the wind farm is greater than or equal to the output that the cluster needs to continue to increase, then the scheduling demand P i ref of wind farm i = P i real + dP T , jump to Step 6;

[0044] Step 4: If dP T > dP i , the increased output predicted by the wind farm is less than the output that the cluster needs to continue to increase, then the scheduling demand P i ref of wind farm i = P i for , and update dP T :

[0045] dP T = dP T - dP i

[0046] Step 5: Judge the size of i and N. If i < N, start from Step 1 again to allocate the next wind farm; if i > N, the allocation ends. All wind farms have been allocated. At this time, all wind farms reach the maximum power generation, and P i for are all less than Pr;

[0047] Step 5: For the remaining units, if dP i < 0, indicating that the predicted output of the wind farm decreases, then the scheduling demand P i ref of wind farm i = P i for , otherwise it remains unchanged P i ref = P i real ;

[0048] (2.2) If the cluster needs to reduce power generation, i.e. DP>0, compare the current output power of each wind farm with the predicted power to determine the power increase potential or power reduction amount dP for each wind farm. i :

[0049] dP i =P ireal -P ifor

[0050] Count all wind farms whose predicted output is less than their current output, i.e., dP i If the power is greater than 0, and there are m such units, the power reduction is:

[0051]

[0052] The power generation reduction required for the other Nm wind farms is as follows:

[0053]

[0054] If dP X If the power is >0, the cluster needs to further reduce its power to dP. X Using the clustering index K in step 2 i The allocation is based on the load factor of the wind farms within the cluster. The order of grouping and reducing wind farm clusters is from the bottom up group to the transition group and then to the top up group, i.e., K. i From smallest to largest, wind farms within each group are sorted and allocated according to their load factor from highest to lowest:

[0055] Step 1: Starting with the first wind farm, distribute the power that needs to be reduced to each wind farm in the order of decreasing cluster index, from the downhill group to the transition group and then to the uphill group.

[0056] Step 2: If dP i If the value is >0, the wind farm can continue to reduce its power margin by θ. i =dP i -P i min Otherwise dP i If the value is less than or equal to 0, the wind farm can continue to reduce its power output by a margin of θ. i =P i real -P i min If dP X <=θ i That is, the reduction margin of wind farm i is greater than or equal to the output that the cluster needs to further reduce, P i min =0.3P i for Then the wind farm i dispatch demand P iref =P i real -dP X Skip to step five;

[0057] Step 3: If dP X >θ i If the derating margin of wind farm i is less than the output that the cluster needs to further reduce, then the scheduling demand P of wind farm i is... i ref =P i min Update dP X The value is:

[0058] dP X =dP X -θ i

[0059] Step 4: Determine the relative values ​​of i and N. If i is less than N, start again from step 1 and allocate the next wind farm. If i is greater than N, the allocation ends and all wind farms have been allocated. At this point, all wind farms will operate at the minimum dispatch power.

[0060] Step 5: Remaining units, if dP i If the value is less than 0, meaning the predicted output of the wind farm decreases, then the dispatch demand P of wind farm i is... i ref =P i for Otherwise, keep P unchanged. i ref =P i real If dP X If the power reduction of each wind farm (m) is greater than the preset value, and the power reduction of the other Nm wind farms in the cluster needs to be increased to dP. X Improvement is carried out according to steps one through six in (2.1).

[0061] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0062] (1) The cluster allocation method proposed in this invention has more stable overall load rate fluctuations, reduces the number of load rate changes (i.e., the number of adjustments), alleviates fatigue load, and has certain significance in optimizing output stability and reducing fatigue load.

[0063] (2) The scheduling method described in this invention reduces the number of wind farms to be regulated. Unlike proportional allocation, when the cluster demand changes, i.e. when the power is increased or decreased, proportional allocation will adjust all wind farms and units. However, the method proposed in this paper only regulates some wind farms, reducing the number of wind farms to be regulated and effectively reducing fatigue damage. Attached Figure Description

[0064] Figure 1 This is a block diagram of the clustering and sorting scheduling method proposed in this paper.

[0065] Figure 2 Flowchart for active power optimization scheduling of wind farm clusters;

[0066] Figure 3 This invention provides a comparison of the power output of various wind farms before and after power reduction under different allocation methods.

[0067] Figure 4 This is a comparison chart of the power reduction of the field cluster under the cluster allocation method of the present invention and the traditional proportional allocation method;

[0068] Figure 5 This is a comparison chart of the power reduction of wind farms under the cluster allocation method of this invention and the traditional proportional allocation method;

[0069] Figure 6 This is a comparison chart of the load rates of various wind farms under two allocation methods according to the present invention. Detailed Implementation

[0070] The present invention will be further illustrated below with reference to the accompanying drawings and specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0071] like Figure 1 As shown, this invention proposes an active power clustering optimization scheduling method for wind power clusters, which includes the following steps:

[0072] 1) Determine the control objectives and constraints, and impose constraints on wind farm output, wind farm output fluctuations, and cluster power tracking errors;

[0073] 2) Define the clustering index of wind farms and the scheduling order of wind farms within the same wind farm cluster. The power prediction value of wind farms from time t to the next 4 cycles is used as the basis for judging the power change trend of wind farms. Based on the change trend, i.e. the ramp rate, the wind farms are divided into three groups: the up-climbing group, the transition group, and the down-climbing group. Within each group, the wind farms are sorted according to the wind farm load rate. The number of wind farms is N.

[0074] 3) Based on the difference between the current output and the grid-connected power demand, combined with the cluster grouping order, the wind farm load rate ranking within the same cluster, and the wind farm power forecast, the cluster power increase / decrease is determined and scheduled according to the grouping.

[0075] Furthermore, step 1) specifically includes the following steps:

[0076] (1) Control objectives

[0077] Determine the cluster grid connection requirements P r The size serves as a target value to constrain cluster power tracking error and to guide the cluster in tracking grid connection requirements.

[0078] (2) Constraints

[0079] (2.1) Wind farm output constraints:

[0080] P i min <P i ref <P i for

[0081] Among them, P i min P i ref P i for Let P be the minimum dispatch power, dispatch power demand, and predicted power for wind farm i. i min =0.3P i for ;

[0082] (2.2) Wind farm output fluctuation constraints

[0083] Constraints on wind farm output fluctuations are applied by limiting the output fluctuation over a 10-minute period:

[0084] |P i real -P i ref |≤P i rate / 3

[0085] Among them, P i real For the actual power output of wind farm i, P i rate Let i be the installed capacity of wind farm i;

[0086] (2.3) Cluster power tracking error constraint: the actual cluster output error with Pr shall not exceed 1%Pr.

[0087] Furthermore, step 2) includes:

[0088] In the dynamic clustering process, the predicted power value of the wind farm from time t to the next four cycles is used as the basis for judging the power change trend of the wind farm. Based on the change trend, the wind farm ramping classification index value K is calculated. iThe fields are grouped, but the groups are not fixed. They are regrouped according to the indicators each period.

[0089] The classification indexes for wind farm ramping are defined as follows:

[0090]

[0091] Where sign() is the sign function; The predicted power value of the wind farm in the nth future cycle; P i real This is the power output value of the wind farm at time t;

[0092] For K i ∈[-4,4], according to K i Wind farm groups are classified into uphill groups, transition groups, and downhill groups. Among them, the uphill group index K... i The value is 4, characterized by a continuous increase in predicted wind farm power; the transition group index K i The range is from -4 to 4, characterized by fluctuations in predicted wind farm power; the downhill group index K i The value is -4, characterized by a continuous decrease in the predicted power of wind farms;

[0093] Within the same wind farm cluster, the scheduling order of wind farms is determined based on their load factors, where the wind farm load factor δ is the load factor. i Defined as the ratio of the actual power generated by each wind farm to its installed capacity:

[0094]

[0095] In the formula, P i rate Let i be the installed capacity of wind farm i;

[0096] When it is necessary to increase the power of the wind farm cluster, priority should be given to increasing the power of wind farms with load rates below the first threshold within the cluster; when it is necessary to decrease the power of the wind farm cluster, priority should be given to reducing the power of wind farms with load rates above the second threshold within the cluster, so as to balance the power output of wind farms within the cluster.

[0097] Furthermore, step 3) includes the following steps:

[0098] (1) Compare the current cluster output power with the cluster demand P r Determine whether the wind farm needs to increase or decrease its power generation (DP) based on its current output.

[0099]

[0100] Among them, P i realLet n be the current power generation of each wind farm, and n be the number of wind farms in the current cluster.

[0101] (2) Determine whether the cluster needs to increase or decrease its power generation;

[0102] (2.1) If the cluster needs to increase power generation, i.e. DP<=0, then compare the current output power of each wind farm with the predicted power to determine the power increase potential or power reduction amount dP for each wind farm. i :

[0103] dP i =P i real -P i for

[0104] Because there are wind farms whose predicted output is less than the current output, i.e., dP i >0, assuming there are m wind farms, the required increase in power generation for the other Nm wind farms is:

[0105]

[0106] dP T To predict the increase in power generation required for a wind farm with an output not less than its current output, dP j This represents the power change of a wind farm whose predicted output is less than its current output.

[0107] Using the clustering index K in step 2 i The allocation is based on the load factor of the wind farms within the cluster. The order of grouping and upgrading at the wind farm cluster level is from the top-climbing group to the transition group, i.e., K. i From largest to smallest, wind farms within each group are sorted and allocated according to their load factor from lowest to highest:

[0108] Step 1: Starting with the first wind farm, allocate the power that needs to be clustered to each wind farm in the order of cluster index improvement, from the climbing group to the transition group.

[0109] Step 2: If dP i If the predicted output of the wind farm is greater than 0, then the dispatch demand P of wind farm i will decrease. i ref =P i for ;

[0110] Step 3: If dP T <= dP i If the predicted increase in power output of wind farm i is greater than or equal to the additional power output required by the cluster, then the scheduling demand P of wind farm i is [amount missing]. i ref =P i real+dP T , jump to step six;

[0111] Step 4: If dP T >dP i , the increased output of the wind farm prediction is less than the output that the cluster needs to continue to increase, then the scheduling demand P i ref = P i for , and update dP T :

[0112] dP T = dP T - dP i

[0113] Step 5: Judge the size of i and N. If i < N, start from step one again and allocate the next wind farm; if i > N, the allocation is over. All wind farms have been allocated. At this time, all wind farms reach the maximum power generation, and P i for are all less than Pr;

[0114] Step 5: For the remaining units, if dP i < 0, indicating that the predicted output of the wind farm has decreased, then the scheduling demand P i ref = P i for , otherwise keep it unchanged P i ref = P i real ;

[0115] (2.2) If the cluster needs to reduce the power generation, that is, DP > 0, compare the current output power and the predicted power of each wind farm to determine the power increase space or power reduction amount dP i :

[0116] dP i = P ireal - P ifor

[0117] Count all wind farms with predicted output less than the current output, that is, dP i > 0. Suppose there are m of them, and their power reduction amount is:

[0118]

[0119] The power generation that the other N - m wind farms need to reduce is:

[0120]

[0121] If dPX If the power is >0, the cluster needs to further reduce its power to dP. X Using the clustering index K in step 2 i The allocation is based on the load factor of the wind farms within the cluster. The order of grouping and reducing wind farm clusters is from the bottom up group to the transition group and then to the top up group, i.e., K. i From smallest to largest, wind farms within each group are sorted and allocated according to their load factor from highest to lowest:

[0122] Step 1: Starting with the first wind farm, distribute the power that needs to be reduced to each wind farm in the order of decreasing cluster index, from the downhill group to the transition group and then to the uphill group.

[0123] Step 2: If dP i If the value is >0, the wind farm can continue to reduce its power margin by θ. i =dP i -P i min Otherwise dP i If the value is less than or equal to 0, the wind farm can continue to reduce its power output by a margin of θ. i =P i real -P i min If dP X <=θ i That is, the reduction margin of wind farm i is greater than or equal to the output that the cluster needs to further reduce, P i min =0.3P i for Then the wind farm i dispatch demand P i ref =P i real -dP X Skip to step five;

[0124] Step 3: If dP X >θ i If the derating margin of wind farm i is less than the output that the cluster needs to further reduce, then the scheduling demand P of wind farm i is... i ref =P i min Update dP X The value is:

[0125] dP X =dP X -θ i

[0126] Step 4: Determine the relative values ​​of i and N. If i is less than N, start again from step 1 and allocate the next wind farm. If i is greater than N, the allocation ends and all wind farms have been allocated. At this point, all wind farms will operate at the minimum dispatch power.

[0127] Step 5: Remaining units, if dP i If the value is less than 0, meaning the predicted output of the wind farm decreases, then the dispatch demand P of wind farm i is... i ref =P i for Otherwise, keep P unchanged. i ref =P i real If dP X If the power reduction of each wind farm (m) is greater than the preset value, and the power reduction of the other Nm wind farms in the cluster needs to be increased to dP. X Improvement is carried out according to steps one through six in (2.1).

[0128] Example

[0129] The case study analysis is based on data from various wind farms in a large base. The wind speed data of each wind farm is obtained by fitting the wind speed at the hub height to the wind speed using Windograper software based on the wind measurement tower data of the wind farm. The wind farm power is determined by the wind speed, installed capacity and unit power curves of the wind farm.

[0130] A case study was used to analyze the effectiveness of the cluster optimization algorithm. The case study selected data from a specific base over three days of the same period, with an optimization cycle of 10 minutes and a total of 432 optimization cycles. A comparison was made between the cluster optimization and proportional allocation methods for demand tracking, with the following power requirements set for the base: free power generation for cycles 0-100; 20,000 MW for cycles 100-150; 30,000 MW for cycles 150-200; 40,000 MW for cycles 200-300; 20,000 MW for cycles 300-350; and 20,000 MW for cycles 450-432.

[0131] 1. Scheduling sequence analysis

[0132] (1) Sorting and scheduling analysis within the field group

[0133] With the wind speed of all wind farms set to a fixed 8 m / s, the clustering index of all wind farms in this base is located in (-4, 4), belonging to the transition group. The scheduling method within the group is analyzed based on the scheduling order of all wind farms in the transition group. The cluster power demand is reduced from the maximum free power output of 2900 MW to 2000 MW. A comparison of the output of each wind farm before and after the power reduction using the cluster allocation proposed in this invention and the traditional proportional allocation is shown below. Figure 3Before the power reduction, the load factor values ​​of each wind farm from 1 to 10 were 0.44252, 0.5441, 0.5757, 0.5453, 0.532, 0.532, 0.59375, 0.59375, 0.5255, and 0.5255, respectively.

[0134] Depend on Figure 3 (a) It can be seen that only five wind farms (2, 3, 4, 7, and 8) reduced their power output during the cluster allocation and scheduling process, and these reductions occurred sequentially from highest to lowest load factor, namely, farms 7, 8, 3, 4, and 2. This simulation result is consistent with the expectations of the method proposed in this paper. Figure 3 (b) Under the proportional allocation method, all wind farms took action. The comparison shows that the allocation method of sorting wind farms in the cluster according to the load rate proposed in this invention has the advantage of reducing the number of wind farms that take action during the scheduling process and reducing the fatigue load of the cluster as a whole, which proves the effectiveness of the sorting method proposed in this invention.

[0135] (2) Field group scheduling sequence analysis

[0136] The wind speeds of each wind farm were set to the actual predicted wind speeds. Based on clustering indices, each wind farm belonged to a different cluster, meaning different groups of wind farms had different ramp-up trends and different responses to cluster demand. Cluster data was analyzed when the cluster power demand decreased from 3700MW to 2000MW at the actual wind speed. Simulation results are shown below. Figure 4 and 5 .

[0137] Depend on Figure 4 It can be seen that when the cluster power is reduced, all three field groups are scheduled according to the proportional allocation, while under the cluster allocation method, only the downhill group and the transition group are scheduled, and the uphill group does not take any action; from Figure 5 It can be seen that the proportional allocation affected all wind farms, while wind farms 1, 3, and 5 in the cluster allocation were not involved in the scheduling. It can be observed that the cluster allocation mobilized fewer wind farms overall, resulting in lower fatigue loads due to the increased number of mobilizations, thus reducing the failure rate of units within the cluster.

[0138] 2. Load factor changes

[0139] The load factor variation curves of 10 wind farms were obtained from simulations based on actual wind speeds, as shown below. Figure 6In the figure, curve a represents the result of proportional allocation, and curve b represents the result of cluster allocation. The results show that the load rates under the two methods are not significantly different, exhibiting a uniform fluctuation trend, with changes mainly distributed in two phases: approximately 20-35 hours and 50-60 hours. During both changes, the blue curve quickly reaches the next stage load rate requirement with minimal fluctuations, requiring fewer adjustments and resulting in greater overall stability. In contrast, curve a experiences more fluctuations and a higher frequency of fluctuations, leading to increased fatigue load. The comparison demonstrates that the cluster allocation method proposed in this invention provides more stable overall load rate fluctuations, reduces the number of load rate changes (i.e., adjustments), and alleviates fatigue load, thus demonstrating significant value in optimizing output stability and reducing fatigue load.

Claims

1. A method for optimized scheduling of active power clusters in wind power generation, characterized in that, The method includes the following steps: 1) Determine the control objectives and constraints, and impose constraints on wind farm output, wind farm output fluctuations, and cluster power tracking errors; 2) Define the clustering index of wind farms and the scheduling order of wind farms within the same wind farm cluster. The power prediction value of wind farms from time t to the next 4 cycles is used as the basis for judging the power change trend of wind farms. Based on the change trend, i.e. the ramp rate, the wind farms are divided into three groups: the up-climbing group, the transition group, and the down-climbing group. Within each group, the wind farms are sorted according to the wind farm load rate. The number of wind farms is N. 3) Based on the difference between the current output and the grid-connected power demand, combined with the cluster grouping order, the wind farm load rate ranking within the same cluster, and the wind farm power forecast, the cluster power increase / decrease is determined and scheduled according to the grouping. Step 2) includes: In the dynamic clustering process, the predicted power value of the wind farm from time t to the next four cycles is used as the basis for judging the power change trend of the wind farm, and the wind farm ramp-up classification index value is calculated based on its change trend. The fields are grouped, but the groups are not fixed. They are regrouped according to the indicators each period. The classification indexes for wind farm ramping are defined as follows: ； Where sign() is the sign function; The predicted power value of the wind farm in the nth future cycle; The actual power output of wind farm i; for According to K i Wind farm groups are classified into uphill groups, transition groups, and downhill groups. Among them, the uphill group index K... i The value is 4, characterized by a continuous increase in predicted wind farm power; the transition group index K i The range is from -4 to 4, characterized by fluctuations in predicted wind farm power; the downhill group index K i The value is -4, characterized by a continuous decrease in the predicted power of wind farms; Within the same wind farm cluster, the scheduling order of wind farms is determined based on their load factors, where the wind farm load factor δ is the load factor. i Defined as the ratio of the actual power generated by each wind farm to its installed capacity: ； In the formula, Let i be the installed capacity of wind farm i; When it is necessary to increase the power of the wind farm cluster, priority should be given to increasing the power of wind farms with load rates below the first threshold within the cluster; when it is necessary to decrease the power of the wind farm cluster, priority should be given to reducing the power of wind farms with load rates above the second threshold within the cluster, so as to balance the power output of wind farms within the cluster.

2. The active power clustering optimization scheduling method for wind power clusters according to claim 1, characterized in that, Step 1) specifically includes the following steps: (1) Control objectives Determine the cluster grid connection requirements P r The size serves as a target value to constrain cluster power tracking error and to guide the cluster in tracking grid connection requirements. (2) Constraints (2.1) Wind farm output constraints: ； in, Let the minimum dispatch power, dispatch power demand, and predicted power be the values ​​for wind farm i, respectively. ; (2.2) Constraints on wind farm output fluctuations Constraints on wind farm output fluctuations are applied by limiting the output fluctuation over a 10-minute period: ； in, For the actual power output of wind farm i, Let i be the installed capacity of wind farm i; (2.3) Cluster power tracking error constraint: the actual cluster output error with Pr shall not exceed 1%Pr.

3. The active power cluster optimization scheduling method for wind power clusters according to claim 1, characterized in that, Step 3) includes the following steps: (1) Compare the current cluster output power with the cluster grid connection requirement P. r Determine whether the wind farm needs to increase or decrease its power generation (DP) based on its current output. ； in, The actual power output of wind farm i is given by n, where n is the number of wind farms in the current cluster. (2) Determine whether the cluster needs to increase or decrease its power generation; (2.1) If the cluster needs to increase power generation, i.e. DP<=0, then compare the current output power of each wind farm with the predicted power to determine the power increase potential or power reduction amount dP for each wind farm. i : ； Because there are wind farms whose predicted output is less than the current output, i.e., dP i >0, assuming there are m wind farms, the required power generation increase for the other Nm wind farms is: ； To predict the increase in power generation required for wind farms with an output no less than their current output, This represents the power change of a wind farm whose predicted output is less than its current output. Using the clustering indicators in step 2 The allocation is based on the load factor of the wind farms within the cluster. The order of grouping and upgrading at the wind farm cluster level is from the top-climbing group to the transition group, i.e. From largest to smallest, wind farms within each group are sorted and allocated according to their load factor from lowest to highest: Step 1: Starting with the first wind farm, allocate the power that needs to be clustered to each wind farm in the order of cluster index improvement, from the climbing group to the transition group. Step 2: If dP i If the predicted output of the wind farm is greater than 0, then the dispatch demand of wind farm i will decrease. ; Step 3: If dP T <=dP i If the predicted increase in power output of the wind farm is greater than or equal to the additional power output required by the cluster, then the scheduling demand of wind farm i is [not specified]. Jump to step six; Step 4: If dP T >dP i If the predicted increase in power output of the wind farm is less than the required increase in power output of the cluster, then the scheduling demand of wind farm i will be less. and update dP T : ； Step 5: Determine the magnitudes of i and N. If i < N, start from Step 1 again to allocate the next wind farm. If i > N, the allocation ends. All wind farms have undergone allocation, and at this time, all wind farms reach the maximum power generation capacity. All are less than Pr; Step 5: Remaining units, if dP i < 0 indicates a decrease in the predicted output of the wind farm, thus the dispatch demand of wind farm i is reduced. Otherwise, remain unchanged. ; (2.2) If the cluster needs to reduce power generation, i.e. DP>0, compare the current output power of each wind farm with the predicted power to determine the power increase space or power reduction amount dP for each wind farm. i : ； Count all wind farms whose predicted output is less than their current output, i.e., dP i If the power is greater than 0, and there are m such units, the power reduction is: ； The power generation reduction required for the other Nm wind farms is as follows: ； if If the power is >0, the cluster needs to continue reducing its power. Using the clustering index in step 2 The allocation is based on the load factor of the wind farms within the cluster. The order of grouping and reducing wind farm load factors at the cluster level is from the lower-climbing group to the transitional group and then to the upper-climbing group. From smallest to largest, wind farms within each group are sorted and allocated according to their load factor from highest to lowest: Step 1: Starting with the first wind farm, distribute the power that needs to be reduced to each wind farm in the order of decreasing cluster index, from the downhill group to the transition group and then to the uphill group. Step 2: If dP i >0, the wind farm can continue to reduce power margin to Otherwise dP i If the value is less than or equal to 0, the wind farm can continue to reduce its power output by a margin of [value missing]. ,if <=θ i In other words, the reduction margin of wind farm i is greater than or equal to the output that the cluster needs to further reduce. Then the wind farm i dispatch demand Skip to step five; Step 3: If >θ i If the derating margin of wind farm i is less than the output that the cluster needs to further reduce, then the scheduling demand of wind farm i is... ,renew The value is: ； Step 4: Determine the relative values ​​of i and N. If i is less than N, start again from step 1 and allocate the next wind farm. If i is greater than N, the allocation ends and all wind farms have been allocated. At this point, all wind farms will operate at the minimum dispatch power. Step 5: Remaining units, if dP i If the value is less than 0, meaning the predicted output of the wind farm decreases, then the dispatch demand for wind farm i is... Otherwise, remain unchanged. ;if If the power reduction of each wind farm (m) is greater than the preset value, and the power reduction of the other Nm wind farms in the cluster needs to be increased to [value missing]. Improvement is carried out according to the first to sixth steps in (2.1).

Images