New energy station equivalent method and its application based on improved k-means algorithm
Through the improved k-means algorithm and particle swarm algorithm, the equivalent value model of new energy station faults is simplified, the problem of insufficient accuracy and computing efficiency in the existing technology is solved, and efficient equal value grouping of new energy power supplies is realized, which improves the improvement of Computation speed and model simplification.
Patent Information
- Application Number
- CN202210126799.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-02-10
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2042-02-10
AI Technical Summary
The existing new energy station fault equivalent methods have insufficient model simplification and computing efficiency, especially in the operating state of multiple units, with poor accuracy and large calculation amount, which cannot accurately characterize the fault characteristics of the station.
Using the improved k-means algorithm, the relationship between the equal value error and the electrical quantity of the station is established by deducing the multi-machine short-circuit current analytical expression and phase conversion of new energy power supplies, and the equivalent voltage is obtained by using the particle swarm algorithm to find the best results, and the The equal value error is used as the particle spacing, and iteratively grouping equal values is performed to simplify the complexity of the model and improve the calculation efficiency.
It has achieved simplification of the new energy power supply model while ensuring accuracy, reducing the calculation amount and simulation time, improving the calculation speed and iterative computing efficiency, and reducing the number of nodes. It is suitable for the new energy station fault analysis system.
Smart Images

Figure BDA0003500289840000031 
Figure BDA0003500289840000032 
Figure BDA0003500289840000041
Abstract
Description
Technical Field
[0001] The present invention relates to a method and application thereof, and in particular to a new energy station equivalence method based on an improved k-means algorithm and application thereof. Background Art
[0002] With the large-scale centralized grid connection of wind power and photovoltaic power, the fault characteristics of new energy power sources are quite different from those of synchronous power sources, which greatly affects the performance of protection action. It is urgent to study accurate fault equivalence methods for new energy sites.
[0003] At present, the equivalent method for renewable energy station faults mainly adopts single-machine equivalent method or multi-machine equivalent method. The single-machine equivalent method has a large degree of model simplification and a small amount of calculation. However, when performing equivalent modeling on a large number of renewable energy power sources with different operating conditions, the accuracy of the single-machine equivalent model is poor and it cannot accurately characterize the fault characteristics of the station.
[0004] The current multi-machine equivalent method of renewable energy power sources mostly adopts clustering algorithm. The existing research on renewable energy power source equivalent needs to consider a large number of related factors: low voltage ride-through characteristics, output, collection lines, and fault occurrence, etc. The influence of these factors has a certain ambiguity. In addition, the grouping problem involved in dynamic equivalent also has a certain ambiguity. Therefore, the key to establishing a dynamic voltage equivalent model for wind farms lies in how to describe these factors and their interactions to form a reasonable clustering result.
[0005] Traditional clustering algorithms use the vector distance of particle coordinates as the distance between the particle and the cluster center, usually using the Euclidean distance and other distance formulas for description, and use the mean value of the coordinates of particles of the same type as the cluster center. However, since the clustering problem is actually a nonlinear problem, its objective function is not uniform and monotonic, and the partial derivatives at each coordinate point are different, so the proximity of the particle and the cluster center in the spatial coordinate system is not equivalent to a more optimized objective function after clustering. The present invention uses the equivalent problem as an optimization problem, uses the particle swarm algorithm to optimize the equivalent results, and uses the equivalent error obtained as the particle spacing. At this time, the close particle spacing indicates that the equivalent error is small and the objective function is better, and the equivalent voltage is used as the cluster center and brought into the algorithm for iteration. Summary of the invention
[0006] In order to solve the deficiencies in the existing technology, we proposed a new energy station equivalent method based on the improved k-means algorithm, and its technical solution includes the following steps:
[0007] Step 1. According to the national standard requirements and the voltage at each station, the analytical expression of the short-circuit current of multiple machines of the new energy power source is derived and the phase conversion is performed to establish the analytical relationship between the equivalent error and the electrical quantity of the station;
[0008] Step 2. Based on the voltage-controlled current source characteristics of the new energy power source, the optimization algorithm is used to calculate the equivalent voltage and impedance of the units divided into the same category. The equivalent error is used as the distance between each particle, and the equivalent voltage is used as the clustering center. The improved k-means algorithm is used to group the new energy units, and the final number of groups is determined according to the silhouette coefficient to complete the grouping of the units.
[0009] The present invention also discloses a new energy station equivalence method based on an improved k-means algorithm applied to a new energy station fault analysis system.
[0010] Beneficial Effects
[0011] The present invention uses the equivalent error as the particle spacing of the clustering algorithm and adopts the improved k-means algorithm for clustering equivalents, which simplifies the model complexity of the renewable energy power supply and ensures a certain degree of accuracy, reduces the amount of calculation and simulation time, and reduces the number of nodes in the process of calculating the network current containing renewable energy power supplies, thereby improving the calculation speed and iterative calculation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] Figure 1 A detailed model topology diagram of the research site of the present invention;
[0013] Figure 2 The topological diagram of the equivalent model of the research station of the present invention;
[0014] Figure 3 This is a flow chart of the new energy station equivalence method based on the improved k-means algorithm of the present invention. DETAILED DESCRIPTION
[0015] The present invention discloses a new energy station equivalent method based on an improved k-means algorithm, which is described in detail below in conjunction with the accompanying drawings.
[0016] Step 1: According to the national standard requirements and the voltage at each station, derive the analytical expression of the short-circuit current of multiple machines of the new energy power supply and perform phase conversion to establish the analytical relationship between the equivalent error and the electrical quantity of the station:
[0017] At present, my country's grid-connected new energy power sources are mainly divided into two categories: full-power inverter power sources represented by doubly-fed wind turbines and partial-power inverter power sources represented by photovoltaic and permanent magnet direct-drive wind turbines.
[0018] According to the Technical Regulations for Photovoltaic Power Station Access to the Power System, and considering that the inverter generally adopts the d-axis voltage orientation strategy, the short-circuit current output by the photovoltaic station after a fault is obtained as
[0019]
[0020]
[0021] Among them, id and iq are the dq axis currents output by the full-power inverter energy source, Us is the voltage at the station, IN is the rated current of the station, Imax is the maximum current allowed to be output by the inverter, K1 and K2 are the reactive current support coefficients of the low-through control stage 1 and 2, UN is the rated value of the voltage at the station, and P0 is the active power of the new energy source;
[0022] Different from the photovoltaic voltage ride-through strategy, according to the Technical Regulations for Wind Farm Access to the Power System, the wind turbine is disconnected from the grid when the voltage drops below 0.2pu. At this time, the short-circuit current output by the wind turbine after the fault can be obtained as
[0023]
[0024]
[0025] Among them, id and iq are the dq axis currents output by the permanent magnet wind farm respectively, Us is the per unit value of the voltage at the station end, UN is the rated value of the voltage at the station end, IN is the rated current of the station, Imax is the maximum current allowed to be output by the inverter, K1 is the reactive current support coefficient of low-through control, and P0 is the active power of the new energy power source.
[0026] For doubly fed wind turbines, unlike permanent magnet direct-drive wind turbines, the short-circuit current provided during a fault is mainly composed of the stator short-circuit current and the short-circuit current of the grid-side converter. Since the capacity of the grid-side converter is small, it is approximately considered that the stator current output by the doubly fed wind turbine is equal to the short-circuit current it provides. The doubly fed wind turbine adopts different control methods according to the degree of voltage drop, and its fault characteristics are also different. They are divided into three categories according to the control method: (1) When the voltage drops ΔU<0.2pu, the doubly fed wind turbine does not enter the low-breakthrough interval, and the RSC maintains outer loop control. At this time, the fault characteristics of the doubly fed wind turbine are consistent with the fault characteristics of the full-power inverter power supply, and will not be repeated; (2) When the voltage drops 0.2pu<ΔU<αp.u., it enters the low-breakthrough control, the outer loop is disconnected, and the RSC single-loop control is performed according to the national standard; (3) When the voltage drop degree α pu<ΔU<0.8pu, the rotor current is overcurrent. In order to prevent the rotor side converter from being damaged by overcurrent, the rotor converter is cut off and the crowbar resistor is put into control.
[0027] a. Crowbar resistor input stage, 0.2pu <U<αp.u.
[0028] In the two-phase rotating coordinate system, the stator and rotor voltage flux equations of the doubly fed wind turbine are:
[0029] Stator and rotor voltage equations:
[0030]
[0031] Stator and rotor flux equations:
[0032]
[0033] Where: u sd 、u sq 、u rd 、u rq are the dq axis voltages of the stator and rotor, i sd 、i sq 、i rd , i rq are the dq axis currents of the stator and rotor, ψ sd , sq , rd , rq are the dq axis flux of the stator and rotor, L m , L s , L r They are the stator-rotor coaxial equivalent winding mutual inductance, stator self-inductance and rotor self-inductance, R s , R r are the stator and rotor resistances respectively, ω is the synchronous speed, s=ω-ω1, ω1 is the rotor speed.
[0034] When the terminal voltage drops, considering its fault steady state, the stator flux can be expressed as:
[0035]
[0036] After the fault, the crowbar is instantly put into operation, and the crowbar resistor will bypass the rotor converter, and the rotor voltage u rdq =0, the rotor resistance becomes the sum of the original rotor resistance and the crowbar resistance R re =R r +R c ,R c is the crowbar resistor.
[0037] From the stator and rotor flux equation, the relationship between the stator and rotor currents and flux can be obtained:
[0038]
[0039] By combining equations (11)-(14), it can be obtained that the steady-state short-circuit current output by the doubly fed wind turbine after the crowbar resistor is put into operation is:
[0040]
[0041] Where: σ = 1 - L m 2 / (LrLs), L m is the mutual inductance between the stator and the rotor.
[0042] b. During the RSC control stage, α p.u. < U < 0.8 p.u.
[0043] According to the "Technical Regulations for Wind Farms Connected to the Power System", the reactive current output by the doubly-fed wind turbine during the RSC control stage can be obtained as follows:
[0044] i sq = K(0.9 - U s )I N α < U s < 0.8 (16)
[0045] Where: i sq is the stator q-axis current, U s is the per-unit value of the stator voltage, I N is the rated current.
[0046] Considering that the doubly-fed wind turbine adopts stator voltage-oriented control, based on the stator flux linkage equation of the wind turbine, the rotor q-axis current can be obtained as follows:
[0047]
[0048] In order to prevent the rotor current from overcurrent and damaging the rotor-side converter during the fault process, it is necessary to limit the rotor current. Considering the active power limit of the wind turbine, the rotor d-axis current can be obtained as follows:
[0049]
[0050] Where I rmax is the limit value of the rotor converter, i rdmax is the rotor current considering the active power limit of the wind turbine, and its expression is:
[0051]
[0052] Where P0 is the maximum active power that the wind turbine can output.
[0053] In summary, during the RSC control stage when the voltage is α p.u. < U < 0.8 p.u., the short-circuit current output by the doubly-fed wind turbine is:
[0054]
[0055] The short-circuit current of the doubly-fed wind turbine in different control stages is:
[0056]
[0057]
[0058] Since the new energy power source adopts voltage-oriented control, the dq axis current output by each station uses the output voltage of its own phase-locked loop as the reference system. Therefore, the dq axis short-circuit current output by each station is converted to the phase of the system synchronous power supply:
[0059]
[0060] Among them, i dΣ 、i qΣ is the dq axis current of each station converted to the reference system phase; i di 、i qi are the dq axis current output by station i, θ i is the difference between the voltage at terminal i of the station and the reference phase.
[0061] For the equivalent model:
[0062]
[0063] Among them, id Σeq 、iq Σeq are the dq axis currents after the equivalent station reduction; id eq 、iq eq are the dq axis currents before equivalent station reduction, θ eq is the phase difference between the terminal voltage of the equivalent unit i and the reference system.
[0064] From the above analysis, after a fault, the new energy power supply presents the characteristics of a voltage-controlled current source. Its steady-state fault current is determined by different control methods, voltage drop degree, active power generated during the fault, and the phase of the phase-locked loop output, that is, the phase of the wind turbine outlet. Therefore, the relevant electrical quantities that affect the steady-state characteristics of the fault of the new energy power supply are selected as grouping indicators. The site topology before and after the equivalent value is as follows: Figure 1 , 2 shown.
[0065] Step 2: Based on the voltage-controlled current source characteristics of the new energy power source, the optimization algorithm is used to calculate the equivalent voltage and impedance of the units classified into the same category; the obtained equivalent error is used as the distance between each particle, and the equivalent voltage is used as the clustering center. The improved k-means algorithm is used to group the new energy units, and the final number of groups is determined according to the silhouette coefficient to complete the grouping of the units.
[0066] Traditional clustering algorithms use the vector distance of particle coordinates as the distance between the particle and the cluster center, usually using the Euclidean distance and other distance formulas for description, and use the mean value of the coordinates of particles of the same type as the cluster center. However, since the clustering problem is actually a nonlinear problem, its objective function is not uniform and monotonic, and the partial derivatives at each coordinate point are different, so the proximity of the particle and the cluster center in the spatial coordinate system is not equivalent to a more optimized objective function after clustering. The present invention uses the equivalent problem as an optimization problem, uses the particle swarm algorithm to optimize the equivalent results, and uses the equivalent error obtained as the particle spacing. At this time, the close particle spacing indicates that the equivalent error is small and the objective function is better, and the equivalent voltage is used as the cluster center and brought into the algorithm for iteration.
[0067] 1. Establish the distance between particles and cluster centers
[0068] The clustering algorithm first initializes the cluster center, and the initial center is randomly assigned to a particle. The present invention uses the equivalent error as the particle spacing, and assigns the particles to the center with the smallest equivalent error as a cluster. The error generated by the equivalent error is used as the objective function, and the clustering problem is transformed into an optimization problem.
[0069] The error resulting from the equalization is given by:
[0070] Q=(i dΣ -i dΣeq ) 2 +(i qΣ -i qΣeq ) 2 (25)
[0071] The optimization algorithm is used to calculate the equivalent error. First, particles are randomly generated, and the equivalent voltage is used as the position of the particle. For a certain equivalent voltage, according to the topology of the equivalent model, there are:
[0072]
[0073] Where: ΔU and δU are the longitudinal and transverse components of the voltage drop, respectively, and U eq is the equivalent voltage randomly generated by the optimization algorithm, U f is the fault point voltage, P eq , Q eq are the active and reactive power output by the equivalent units, r and x are the resistance and reactance per unit length of the line, and l is the equivalent line length.
[0074] According to the above formula, for a certain voltage drop, the equivalent model is solvable. According to formula (26), a certain line impedance length can be solved, and the voltage phase at this time can be solved. According to formula (24), the dq axis current output by the equivalent station after the reduction can be calculated. According to (25), the error calculation is performed, and the objective function Q is optimized to obtain l eq , U eq And the error E at this time, the error E is used as the distance d between the particle and the cluster center of the clustering algorithm ij , complete the establishment of algorithm space distance.
[0075] 2. Determination of cluster centers during iteration
[0076] The clustering algorithm first initializes the cluster center and assigns the particles to the closest cluster according to the particle spacing. According to the clustering results obtained in the previous step, for the fans classified into the same category, the objective function remains unchanged and the equivalent voltage is randomly generated to search for the best again. At this time, the equivalent voltage U eq The equivalent voltage at this time is used as the cluster center, the distance between the particle and the cluster center is recalculated, and the particles are redistributed based on the principle of the closest distance, and the next iteration is performed until the algorithm converges and outputs the clustering result. The specific process is shown in Figure 3.
[0077] 3. Determine the number of clusters based on the silhouette coefficient
[0078] The silhouette coefficient formula is as follows:
[0079]
[0080] Where a(i) represents the cohesion of the sample point, and the calculation formula is as follows
[0081]
[0082] Where j represents other sample points in the same class as sample i, and distance represents the distance between samples ij. The smaller a(i), the closer the samples are.
[0083] The calculation method of b(i) is similar to that of a(i). It is just necessary to traverse other clusters to obtain multiple values {b(1), b(2), ..., b(n)} and select the smallest value as the final result. Dividing the renewable energy power source into more categories will not significantly improve the error and evaluation index, but will significantly increase the required simulation time. Therefore, the number of clusters can be set within a reasonable range. The equivalent process is as follows Figure 3 shown.
[0084] Example 1
[0085] Taking the actual topology and parameters of multiple wind farms in a certain area of Jilin as an example, the following Figure 1 The detailed model of the wind farm shown (including 10×100MW permanent magnet, double-fed wind farm). The wind turbines are connected to the grid connection point via a box transformer (0.69kV / 35kV), and connected to the external grid via an overhead line through a main transformer (35kV / 220kV). The short-circuit impedance of the box transformer is 6.39%, and the short-circuit impedance of the main transformer is 13.54%. The distance between each station is 5km.
[0086] 1. When the grid connection point voltage is about 90%, since the permanent magnet and double-fed steady-state characteristics are consistent before entering the low-breakthrough interval, only the permanent magnet wind farm is simulated and verified. According to the voltage information of each station, the current after the conversion of each wind farm can be obtained according to step one. According to the equivalent method described in step two, the optimization algorithm is used to generate the equivalent voltage, find the optimal solution, and obtain the equivalent error between samples for initial classification; the stations classified into the same category are obtained through the optimization algorithm, and the equivalent voltage is obtained, and the distance between the equivalent voltage and each sample is calculated again, and the classification is completed again, and it is continuously iterated until the algorithm converges.
[0087] Table 1 Wind farm grouping results
[0088]
[0089] The current errors generated before and after the wind farm adopts the equivalent model proposed by the present invention are shown in Table 2. It can be seen that the fault current output by the equivalent model and the detailed model has a good fit. The current error of the equivalent model is 1.1% with high accuracy. The model can be simplified while ensuring the accuracy of the model.
[0090] Table 2 Comparison of the proposed equivalent method and the single machine equivalent error
[0091]
[0092] 2 The grid voltage drops by 60%
[0093] When the voltage at the grid connection point drops to 60%, it is first divided into permanent magnet and double-fed according to the wind farm type, and then classified according to the voltage information of each station.
[0094] Table 3 Wind farm grouping results
[0095]
[0096] The current errors generated before and after the wind farm adopts the equivalent model proposed by the present invention are shown in Table 4. It can be seen that the fault current output by the equivalent model and the detailed model has a good fit. The current error of the equivalent model is 1.34% with high accuracy. The model can be simplified while ensuring the accuracy of the model.
[0097] Table 4 Comparison of the proposed equivalent method and the single machine equivalent error
[0098]
[0099] The grid voltage drops by 30%
[0100] When the voltage at the grid connection point drops to 30%, the RSC of the doubly fed unit is cut off and the crowbar is put into use. The permanent magnet wind farm enters low-through control. It is first classified according to the model and then grouped according to the terminal voltage information.
[0101] Table 5 Wind farm grouping results
[0102]
[0103] The current errors generated before and after the wind farm adopts the equivalent model proposed by the present invention are shown in Table 6. It can be seen that the fault current output by the equivalent model and the detailed model has a good fit. The current error of the equivalent model is 2.7% with high accuracy. The model can be simplified while ensuring the accuracy of the model.
[0104] Table 6 Comparison of the proposed equivalent method and the single machine equivalent error
[0105]
[0106] Compared with the detailed model, the fault current error of the proposed equivalent method is less than 3% when combined with Tables 2, 4, and 6. The final equivalent method of new energy power supply suitable for engineering practical application has a high degree of simplification and ensures a certain degree of accuracy, reduces the amount of calculation and simulation time, and reduces the number of nodes in the current calculation process of the network containing new energy power supply, and improves the calculation speed and iterative calculation efficiency.
[0107] In theory, clustering algorithms can be used to achieve multi-unit clustering through the fault characteristics and electrical quantity characteristics of new energy power sources. The traditional clustering algorithm uses the vector distance of particle coordinates as the distance between the particle and the cluster center, but because the clustering problem is actually a nonlinear problem, its objective function is not uniform and monotonic, and the partial derivatives at each coordinate point are different, so the particle and the cluster center are close in the spatial coordinate system, which is not equivalent to a more optimized objective function after clustering. The present invention uses the equivalent problem as an optimization problem, and uses a particle swarm algorithm to optimize the equivalent result to obtain an equivalent voltage, and uses the obtained equivalent error as the particle spacing. At this time, the particle spacing is close, indicating that the equivalent error is small, the objective function is better, and the equivalent voltage is used as the cluster center, and brought into the algorithm for iteration. The present invention uses the equivalent error as the particle spacing of the clustering algorithm, and uses the improved k-means algorithm for clustering equivalence, which simplifies the model complexity of the new energy power supply, and ensures a certain accuracy, reduces the amount of calculation and simulation time, and reduces the number of nodes in the network current calculation process containing new energy power supplies, and improves the calculation speed and iterative calculation efficiency.
[0108] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.
Claims
1. New energy station equivalent method based on improved k-means algorithm, Its characteristics are: The steps include: Step 1: Derive the analytical expression of the short-circuit current of multiple machines of the new energy power source according to the voltage at each station end and perform phase conversion to establish the analytical relationship between the equivalent error and the electrical quantity of the station; Step 2: Based on the voltage-controlled current source characteristics of the new energy power source, the optimization algorithm is used to calculate the equivalent voltage and impedance of the units classified into the same category. The equivalent error obtained is used as the distance between each particle, and the equivalent voltage is used as the clustering center. The improved k-means algorithm is used to group the new energy units, and the final number of groups is determined according to the silhouette coefficient to complete the grouping of units; The step 1 further includes the following contents: Considering that the inverter adopts the d-axis voltage orientation strategy, the short-circuit current output by the photovoltaic station after the fault is obtained as follows: Among them, i d 、i q are the dq axis current output by the photovoltaic power station, U s The per unit value of the station voltage, U N is the rated value of the station terminal voltage, I N is the rated current of the station, I max K is the maximum current allowed to be output by the inverter. 1 , K 2 are the reactive current support coefficients of low-through control stages 1 and 2, respectively, 0 is the active power of the new energy source; when the voltage of wind power drops below 0.2pu, the wind turbine is disconnected from the grid. At this time, the short-circuit current output by the wind turbine after the fault is: Among them, i d 、i q are the dq axis current output by the permanent magnet wind farm, U s The per unit value of the station voltage, U N is the rated value of the station terminal voltage, I N is the rated current of the station, I max K is the maximum current allowed to be output by the inverter. 1 is the reactive current support factor of low wear control, P 0 is the active power of the new energy source; the doubly fed wind turbine adopts different control methods according to the degree of voltage drop, and its fault characteristics are also different. According to the control method, it can be divided into three categories: (1) When the voltage drops ΔU<0.2pu, the doubly fed wind turbine does not enter the low-through interval, and the RSC maintains outer loop control. At this time, the fault characteristics of the doubly fed wind turbine are consistent with the fault characteristics of the full-power inverter power supply; (2) When the voltage drops 0.2pu<ΔU<αp.u., it enters the low-through control, the outer loop is disconnected, and the RSC single-loop control is performed according to the national standard; (3) When the voltage drops to αp.u.<ΔU<0.8pu, the rotor current is overcurrent. In order to prevent the rotor side converter from being damaged by overcurrent, the rotor converter is cut off and the crowbar resistor is put into control; α is the voltage boundary between the crowbar resistor control and the low-through control; The crowbar resistance control further includes the following contents: a. During the crowbar resistor input stage, when the voltage drops by 0.2pu<ΔU<αp.u., the steady-state short-circuit current output by the doubly-fed wind turbine after the crowbar resistor is input is: Where: σ=1-L m 2 / (LrLs), L m Ls is the mutual inductance of the stator and rotor, Lr is the self-inductance of the stator and rotor, R re =Rcb+Rr, Rcb is the crowbar resistance, Rr is the rotor resistance, s is wn-ws, wn is the synchronous speed, ws is the stator speed, U is the terminal voltage, ω=2πf, where f is the grid frequency, which is 50Hz; b. RSC control stage αp.u. <U<0.8p.u. The short-circuit current output by the doubly-fed wind turbine is: Among them, isd and isq are the stator dq axis currents output by the doubly fed wind turbine respectively, U s Station terminal voltage, I N is the rated current of the station, I rmax is the maximum current allowed to be output by the rotor converter, i rq is the rotor q-axis current, K is the reactive current support coefficient of low wear control, P 0 is the active power of the new energy power source; The short-circuit current of the doubly-fed wind turbine in different control stages is obtained as follows: Among them, K d , K q are the dq axis current and voltage ratio coefficients of the doubly fed wind turbine in the crowbar control stage; i rd is the rotor d-axis current.
2. According to the new energy station equivalent method based on the improved k-means algorithm according to claim 1, Its characteristics are: due to The new energy power source adopts voltage-oriented control. The dq axis current output by each station uses the output voltage of its own phase-locked loop as the reference system. Therefore, the dq axis short-circuit current output by each station is converted to the phase of the system synchronous power supply: Among them, i dΣ 、i qΣ is the dq axis current of each station converted to the reference system phase; i di 、i qi are the dq axis current output by station i, θ i is the difference between the voltage at terminal i of the station and the reference phase; For the equivalent model: Among them, id Σeq 、iq Σeq are the dq axis currents after the equivalent station reduction; id eq 、iq eq are the dq axis currents before equivalent station reduction, θ eq is the phase difference between the terminal voltage of the equivalent unit i and the reference system.
3. According to the new energy station equivalent method based on the improved k-means algorithm of claim 1, Its characteristics are: The step 2 further includes the following contents: (1) establishing the distance between the particle and the cluster center; (2) Determination of cluster centers during the iteration process; (3) Determine the number of clusters based on the silhouette coefficient.
4. According to the new energy station equivalent method based on the improved k-means algorithm of claim 3, Its characteristics are: The distance between the particles and the cluster center is further established by: using the equal value error as the particle distance, assigning the particles to the center with the smallest equal value error as a cluster, using the error generated by the equal value as the objective function, and converting the clustering problem into an optimization problem; The method of determining the number of groups according to the silhouette coefficient further includes the following contents: for wind turbines classified into the same category, the objective function remains unchanged and the equivalent voltage is randomly generated to search for the best again. At this time, the equivalent voltage U eq , and use the equivalent voltage at this time as the cluster center, recalculate the distance between the particle and the cluster center at this time, and redistribute the particles based on the principle of the closest distance, and perform the next iteration until the algorithm converges and outputs the clustering result; The silhouette coefficient formula is as follows: Where a(m) represents the cohesion of the sample points, and the calculation formula is as follows: Where n represents other sample points in the same class as sample m, and distance represents the distance between samples mn; the smaller a(m), the closer the samples are; b(m) is the minimum value of the distance between sample m and other different cluster samples n with inter-cluster dissimilarity. The calculation formula is as follows: Where n represents other sample points that are not in the same class as sample m. The larger b(m) is, the more discrete the samples of different classes are.
