Quarary wind energy assessment method considering seasonal wind speed simulation

Abstract

A seasonal wind energy assessment method considering seasonal wind speed simulation comprises the steps that a Canopy clustering algorithm is optimized based on a particle swarm optimization (PSO) algorithm, after two optimal parameters of the Canopy clustering algorithm are solved, the optimal Canopy clustering algorithm is used for conducting coarse clustering on daily wind speed samples, and the optimal clustering number and the initial clustering center are obtained; dividing the daily wind speed into clusters with different wind speed characteristics by using the clustering cluster number and the initial clustering center in combination with a K-means clustering model, and combining a state transition Markov chain meter and wind speed seasonality between the seasonal clusters; performing wind speed simulation on each cluster to generate a seasonal simulated wind speed sample; and calculating a wind energy evaluation index by using the simulated wind speed sample, and evaluating wind energy resources of the region. According to the method, after the wind speed is accurately simulated by considering the seasonal property of the wind speed, the quarterly wind energy resource evaluation index is calculated, the quarterly wind energy resources of the region are further accurately evaluated, and a quantitative decision basis is provided for investment planning and operation of a wind power plant.

Description

Technical Field

[0001] The present invention relates to the field of wind energy assessment in a wind power system, and in particular to a quarterly wind energy assessment method taking seasonal wind speed simulation into account. Background Art

[0002] As a clean energy, wind energy has the characteristics of no pollution, renewable, and low power generation cost. It has attracted widespread attention from all over the world, and the total installed capacity of wind power has grown rapidly. Accurately describing the statistical characteristics of wind speed can quantify wind energy resources, provide a reference for accurately evaluating wind energy resources in a certain area, and further guide the investment and construction of wind power stations in the area. Affected by the revolution of the earth, the natural wind generated by the uneven heating of the earth's surface by the sun has a certain autocorrelation with the natural wind generated at previous moments, and has different statistical characteristics such as quarterly mean and standard deviation with seasonal changes. It is called the seasonality of wind speed. Affected by the seasonality of wind speed, wind energy resources also show seasonal variation characteristics; at the same time, wind speed is also volatile and intermittent, which makes wind energy resources show strong uncontrollability. Therefore, establishing a reasonable and accurate wind speed simulation model is a prerequisite for seasonal wind energy resource assessment in the region.

[0003] At present, many scholars at home and abroad have conducted extensive research on wind speed models based on the seasonal characteristics of wind speed. However, the wind speed model cannot simultaneously consider the randomness and seasonality of wind speed, which makes the simulated wind speed unable to better characterize the seasonality of wind energy resources. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a quarterly wind energy assessment method taking into account seasonal wind speed simulation. After accurately simulating the wind speed taking into account the seasonality of the wind speed, the quarterly wind energy resource assessment index is calculated, and then the quarterly wind energy resources in the area are accurately assessed, providing a quantitative decision-making basis for wind farm investment planning and operation.

[0005] The technical solution adopted by the present invention is:

[0006] A seasonal wind energy assessment method taking into account seasonal wind speed simulation comprises the following steps:

[0007] Step 1: Optimize the Canopy clustering algorithm based on the particle swarm optimization algorithm (PSO). After obtaining the two optimal parameters of the Canopy clustering algorithm, use the optimal Canopy clustering algorithm to perform coarse clustering on the daily wind speed samples to obtain the optimal number of clusters and initial cluster centers;

[0008] Step 2: Using the number of clusters and the initial cluster centers in step 1, the K-means clustering model is combined to divide the daily wind speed into clusters with different wind speed characteristics, and the state transition Markov chain between clusters in each season is combined to calculate the seasonality of wind speed;

[0009] Step 3: Perform wind speed simulation on each cluster in step 2 to generate simulated wind speed samples that take seasonality into account;

[0010] Step 4: Using the simulated wind speed samples in step 3, calculate the wind energy assessment index and evaluate the wind energy resources in the area.

[0011] The step 1 comprises the following steps:

[0012] S1.1: Collect the measured wind speed samples of a certain observation station, divide the measured wind speed samples into 24 hours a day, and obtain the daily wind speed samples;

[0013] S1.2: Apply the PSO optimization algorithm to optimize the two distance thresholds in the Canopy clustering algorithm to obtain the optimal Canopy clustering algorithm;

[0014] S1.3: Use the optimal Canopy algorithm of S1.2 to roughly cluster the daily wind speed samples and obtain the optimal number of clusters and initial cluster centers.

[0015] In S1.1, the wind speed samples measured over a long period of time at the meteorological observation station are collected and recorded as v1, v2, …v N is N groups of measured wind speed samples, which are separated by one hour. The measured wind speed samples are divided into units of days to obtain the daily wind speed matrix V = [V1, V2, …, V M ],V1,V2,…,V M Respectively represent the matrix composed of 24-hour wind speed samples on the mth day, where V m =[v 24(m-1)+1 ,v 24(m-1)+2 ,…,v 24(m-1)+24 ] is the 24-hour measured wind speed sample on the mth day, v 24(m-1)+1 ,v 24(m-1)+2 ,…,v 24(m-1)+24 They represent the measured wind speed samples from the 1st hour to the 24th hour of the mth day, respectively, m=1, 2, …, M, M represents the number of days in a year, usually 365 days.

[0016] In S1.2, x is the position of a single particle on a one-dimensional coordinate axis, and its range of variation is limited to [X min ,X max ], X min ,X max They represent the lower and upper limits of the position of a single particle, respectively. v is the velocity of a single particle, and its range of variation is limited to [Vmin ,V max ],V min ,V max They represent the lower and upper limits of the speed of a single particle, respectively, where X min and X max The following steps are performed:

[0017] (1) Find the mean:

[0018] The daily wind speed average P is calculated according to the following formula: [p1, p2, ..., p 24 ]; p1, p2, …, p 24 Respectively represent the mean wind speed of all daily wind speed samples from the 1st to the 24th hour:

[0019] P=(V1+V2+...+V M ) / M

[0020] (2) Calculate the Euclidean distance:

[0021] Calculate the wind speed sample V on the mth day according to the following formula m The Euclidean distance L with P constitutes the distance matrix L = [L1, L2, .., L M ],L1,L2,..,L M Respectively represent the wind speed sample V on the mth day m Euclidean distance from P:

[0022]

[0023] (3) Calculate X min and X max , the specific calculation formula is as follows:

[0024] X max =(L max -L min )×0.6

[0025] X min =L min ×0.35

[0026] Where: L max is the maximum value in L, L min is the minimum value in L.

[0027] In S1.2, the PSO optimization algorithm sets two groups of particle populations, each group of populations consists of K particles. Taking the kth particle in the two groups of populations as an example, k = 1, 2, ..., K, K represents the number of particles in the two groups of populations, and the position coordinate vector is X k =[x 1k ,x 2k ], x 1k, x2 represent the position of the kth particle in particle population 1 and the position of the kth particle in population 2, respectively, where x 2k <x 1k , the fitness value is f k , f k The calculation process is as follows:

[0028] 1) The number of cluster centers Q = 1, and a day is randomly selected from the daily wind speed matrix V as the first cluster center C q , C=[C1,C2,..,C Q ], mark the wind speed of that day as Q and delete the data of that day from V;

[0029] 2) Traverse V and calculate the distance between the daily wind speed sample in V and the Q cluster centers in C. If the wind speed sample on that day is close to the cluster center C q The distance is less than x 1k , q=1,2,..,Q, Q represents the total number of cluster centers;

[0030] The wind speed of the day is marked as q and the sample data of the day is deleted from V. Then the cluster center C is updated with the mean of the wind speed samples marked as q. q ; If the distance to each cluster center is greater than x 2k , then Q = Q + 1, and the wind speed sample of that day is used as the new cluster center C Q , mark the wind speed of that day as Q, and then delete the data of that day from V;

[0031] 3) If V is not empty, repeat step 2); otherwise, end the algorithm;

[0032] 4) Calculate the Davidson-Boulding Index (DBI) as the fitness value f k , the calculation formula is as follows:

[0033]

[0034] in: is the average Euclidean distance from the i-th class to its class center, i = 1, 2, ... Q, Q represents the total number of cluster centers; represents the average Euclidean distance from the jth class to its class center, w i represents the Euclidean distance of the center of the i-th class, w j Represents the Euclidean distance of the center of the i-th class.

[0035] The S1.2 comprises the following steps:

[0036] Step 1. Initialization: The maximum number of iterations is N; in the particle position range [X min ,X max ] Randomly generate initial particle positions represents the initial position of the kth particle in particle population 1, represents the initial position of the kth particle in particle population 2, in the particle velocity range [V min ,V max ] Randomly generate initial particle speed represents the initial motion speed of the kth particle in particle population 1, represents the initial motion speed of the kth particle in particle population 2, particle self-learning factor s1, group learning factor s2, and maximum inertia weight W max , minimum inertia weight W min ;

[0037] Step 2. Find the fitness value f k :If the initial particle position According to the fitness value calculation process, the fitness value fk of the kth particle in the two groups of particle populations is calculated in turn, where k = (1, 2, ..., K), otherwise the fitness value f k is a larger number, which is set to e in the present invention 10 , two groups of particle populations K constitute the vector F = [f1,f2,…,f K ],f1,f2,…,f K They represent the fitness values ​​calculated for the kth particle in the two populations respectively.

[0038] Step 3. Optimize individuals: Select the minimum fitness value f from the calculated result F bn , and record the minimum fitness value f during the nth iteration bn , forming a matrix F2 = [f b1 ,f b2 ,…f bn ],f b1 ,f b2 ,…f bn They represent the minimum fitness value in the vector F formed by the nth iteration, and the corresponding X bn Construct vector X B =[X b1 ,X b2 ,…X bn ],X b1 ,X b2 ,…X bn They respectively represent the position coordinates corresponding to the minimum fitness value in the vector F constructed in the nth iteration.

[0039] Step 4. Position update: Taking population 1 as an example, apply the following formula to update the particle position:

[0040]

[0041] Where: r1, r2 are random numbers between [0,1]; f m is the minimum value in the matrix F2; s1 represents the self-learning factor of the particle; s2 represents the group learning factor of the particle.

[0042] Step 5. Loop: Repeat the process from Step 2 to Step 4 until the maximum number of iterations is reached, then stop the loop and the minimum value f in F2 is bmin Corresponding X bmin That is, the optimal threshold of the Canopy clustering algorithm.

[0043] In S1.3, according to the optimal threshold of the Canopy clustering algorithm obtained in step S1.3, the daily wind speed samples are roughly clustered using the process described in S1.3 to obtain the optimal number of clusters Q and the initial cluster center C.

[0044] In step 2, the initial cluster center in step S1 is used as the initial randomly selected cluster center for K-means clustering, and K-means clustering is performed on the daily wind speed samples to divide the daily wind speed samples into wind speed clusters with different characteristics; based on the clustering results, the daily wind speed state transition Markov chain of the four quarters is calculated. The following steps are included:

[0045] S2.1: Use the optimal number of clusters Q and the initial cluster center C obtained in step S1 as the initial clustering input parameters of K-means clustering;

[0046] S2.2: Calculate the daily wind speed D on the mth day in the daily wind speed sample m The Euclidean distance to each cluster center Cq constitutes the matrix L m =[L1,L2,…,L Q ],L1,L2,…,L Q represents the daily wind speed D on the mth day respectively. m To the Euclidean distance from the 1st cluster center to the Qth cluster center, find L m Minimum value L min The corresponding cluster number is q, and the wind speed on that day is marked as q. Apply formula C q =(C q +V m ) / 2 Update the cluster center C q ; C q represents the qth cluster center, V m represents the daily wind speed on the mth day;

[0047] S2.3: Repeat step S2.2 until all wind speed samples are divided;

[0048] S2.4: Use Markov chain to describe the random transfer relationship of daily wind speed in spring, summer, autumn and winter, P1, P2, P3 and P4, and then take into account the seasonality of wind speed. Taking spring as an example, the transfer probability of the daily wind speed sample of cluster i (i = 1, 2, ..., Q) to the daily wind speed sample of cluster j (j = 1, 2, ..., Q) in spring can be calculated by the following formula:

[0049]

[0050] Where: p 1(i,j) represents the probability of the i-th wind speed sample transferring to the j-th wind speed sample in spring, K represents the number of daily wind speed sample types, N 1(i,j) is the number of times the measured spring daily wind speed samples are transferred from the i-th cluster wind speed to the j-th cluster wind speed, so the spring state transition probability matrix P1 is:

[0051]

[0052] Where: p 1(1,1) It represents the probability of the first type of wind speed sample in spring transferring to the second type of wind speed sample, p 1(1,2) It represents the probability of the second type of wind speed sample in spring transferring to the second type of wind speed sample, p 1(1,K) It represents the probability of the first-class wind speed sample transferring to the K-class wind speed sample in spring;

[0053] p 1(2,1) represents the probability of the second-class wind speed sample transferring to the first-class wind speed sample in spring, p 1(2,2) It represents the probability of the second type of wind speed sample in spring transferring to the second type of wind speed sample, p 1(2,K) It represents the probability of the second-class wind speed sample transferring to the K-class wind speed sample in spring;

[0054] p 1(K,1) represents the probability of the K-th wind speed sample transferring to the 1st wind speed sample in spring, p 1(K,2) represents the probability of the K-th wind speed sample transferring to the second-class wind speed sample in spring, p 1(K,K) It represents the probability of the Kth wind speed sample transferring to the Kth wind speed sample in spring.

[0055] In step 3, wind speed simulation is performed on each cluster in step S2 to generate simulated wind speed samples that take into account seasonality, randomness and autocorrelation; specifically, as follows:

[0056] The wind speed samples of the same cluster are decomposed into multiple components using EMD decomposition method, the low-frequency component is taken as the long-term trend component of the cluster, and the sum of the other components is the random fluctuation component of the cluster;

[0057] The random fluctuation component is modeled using the Ito random process, and the simulated wind speed of the cluster is obtained by adding it to the long-term trend component. If the wind speed at a certain time point is less than zero, the wind speed at that time point is re-simulated until the wind speed is greater than zero.

[0058] The Markov chain is used to calculate the state transfer matrix of each cluster wind speed, and the cluster of the next day is simulated according to the state transfer matrix, and the simulated wind speed of the next day is sequentially extracted from the simulated wind speed of the cluster.

[0059] The following steps are involved:

[0060] S3.1: Connect the wind speeds of the qth clusters in order, forming a series with T q Hourly wind speed samples v1, v2, …v Tq , where v t is the wind speed sample at time t, t=1,2,…,T q They represent the hourly time of the wind speed of the qth cluster day,

[0061] S3.2: Decompose the qth cluster wind speed samples v1, v2, …, v Tq Decomposed into seasonal trend components s1, s2, …, s Tq and random fluctuation components r1, r2, …, r Tq ,,s1,s2,…,s Tq Respectively represent the 1st to T q seasonal trend components at time, r1, r2, …, r Tq Respectively represent the 1st to T q The random fluctuation component at each moment;

[0062] Taking the first cluster of wind speed samples as an example, the specific steps are as follows:

[0063] (1) Envelope fitting: wind speed samples v1, v2,…, v Tq For the original input signal v(t), find all its maximum points, and connect the maximum points into the maximum envelope y through the cubic spline function max (t); Similarly, find all the minimum points of v(t) and connect the minimum values ​​into the minimum envelope y through the cubic spline function. min (t);

[0064] (2) Calculate the intermediate wind speed: Use the following formula to calculate the intermediate wind speed v1(t):

[0065] v1(t)=v(t)-m1(t)

[0066]

[0067] Where: m1(t) represents the median wind speed of the first cluster of wind speed samples;

[0068] (3) Calculate the intrinsic mode function: If the difference between the number of extreme points and the number of zero-crossing points in the entire data segment of v1(t) is not greater than 1, and the mean of the upper envelope and the lower envelope is zero, then v1(t) is an intrinsic mode function component of the wind speed sample v(t); otherwise, v1(t) is the original input signal, and steps (1) to (3) are repeated;

[0069] (4) Loop: e k (t) is the new original signal, where e k (t) = v(t) - v1(t), v(t) represents the input wind speed time series, v1(t) represents the intrinsic mode function component of the input wind speed time series, k is the number of cycles, if e k (t) is a monotonic function or a constant, then the process ends. Otherwise, repeat steps (1) to (4).

[0070] (5) Determine the weight: e k (t) is the seasonal trend component s(t) of d(t), that is, s(t) = e k (t), and the sum of k-1 intrinsic mode function components is taken as the random fluctuation component r(t), that is, e i (t) represents the i-th natural mode component;

[0071] S3.3: Based on the decomposition results of step S3.2, use the Ito random process to model the random fluctuation component and calculate the simulated random fluctuation component at time t

[0072]

[0073] Where: represents the simulated random fluctuation component at time t, t=1,2,…,L i The hour and time of the simulated wind speed respectively;

[0074] μ() is the drift term, which reflects the tendency of the random variable to move toward the expected value. Its calculation formula is: For r t The mean of θ is the time autocorrelation function, and its calculation formula is R(τ)=e -θτ ; σ() is the diffusion term, reflecting W t Randomness affects simulation value; W t represents the Wiener process.

[0075] It can be solved by choosing a suitable random sequence probability distribution.

[0076] S3.4: The simulated wind speed seasonal trend component s in step S3.2t As the seasonal trend component of simulated wind speed and the random fluctuation component of wind speed simulated in step S3.3 Add together to get the simulated wind speed:

[0077]

[0078] verify Is it in the interval [v min ,v max ], if yes, accept it, if no, reject it and repeat step S3.3 until the sample is accepted.

[0079] S3.5: Repeat steps S3.3 and S3.4 until T is generated i The wind speed samples of all clusters are simulated, and steps S31 to S35 are repeated until the simulation of wind speed samples of all clusters is completed.

[0080] S3.6: According to the state transfer matrices P1, P2, P3 and P4 of the four seasons of spring, summer, autumn and winter, one day's simulated wind speed samples are sequentially extracted from the k clusters of simulated wind speed samples until N groups of simulated wind speeds are generated, where N = T1 + T2 + T3 + T4, T1, T2, T3, T4 represent the number of simulated wind speed samples in spring, the number of simulated wind speed samples in summer, the number of simulated wind speed samples in autumn and the number of simulated wind speed samples in winter respectively.

[0081] In the step 4, the quarterly wind energy evaluation indicators such as the quarterly wind speed mean, the quarterly wind speed standard deviation, the quarterly wind power density, and the average effective wind power are calculated according to the simulated wind speed, and the wind energy resources in the region are evaluated through the calculation results; including the following:

[0082] 4.1: Use the following formula to calculate the quarterly mean wind speed v of the simulated wind speed sample: m.p and seasonal wind speed standard deviation v s.p :

[0083]

[0084] Where: the superscript “^” indicates the wind speed simulation value, and The seasonal wind speed mean and seasonal wind speed standard deviation of the simulated wind speed are respectively represented, where p = (1, 2, 3, 4) represents spring, summer, autumn and winter respectively; taking spring as an example, N1 is the number of spring wind speed samples in the simulated wind speed samples; is the simulated wind speed sample at the i-th moment in spring;

[0085] 4.2: Use the following formula to calculate the seasonal average wind power density of the simulated wind speed:

[0086]

[0087] Where: ρ is the air density, usually 1.29kg / m 3 ;

[0088] 4.3: Use the following formula to calculate the seasonal average effective wind power density of the simulated wind speed:

[0089]

[0090] Where: v ci and v co are the cut-in wind speed and cut-out wind speed of the wind turbine respectively; I pi is the indicator factor, dimensionless;

[0091] 4.4: Use the following formula to calculate the quarterly effective hours T of the simulated wind speed: p and quarterly utilization efficiency η p :

[0092]

[0093] Where: N bp is the number of hours in the four quarters of the year, N b1 、N b2 、N b3 and N b4 The values ​​are 2208, 2208, 2184 and 2160 respectively;

[0094] 4.5: Use the following formula to calculate the quarterly power generation E of the simulated wind speed p :

[0095]

[0096] Where: P p () is the wind power conversion function of the wind turbine, which is shown in the following formula:

[0097]

[0098] Where: v is the instantaneous wind speed; v τ and P τ are the rated wind speed and rated capacity of the wind turbine respectively; the coefficients A, B and C can be obtained from v ci and v τ The specific calculation formula is as follows:

[0099]

[0100] 4.6: Use the following formula to calculate the seasonal capacity factor C of the simulated wind speed: p :

[0101]

[0102] The present invention provides a seasonal wind energy assessment method taking into account seasonal wind speed simulation, and the technical effects are as follows:

[0103] 1) Step 1 of the present invention first optimizes two parameters of Canopy through the PSO optimization algorithm to avoid the experience of artificial setting and make the parameter selection more accurate.

[0104] 2) The improved K-means clustering in step 2 of the present invention avoids the subjectivity in the selection of the initial number of clusters and the randomness in the selection of cluster centers, making the final clustering result of daily wind speed samples more objective and reliable, thereby better taking into account the daily wind speed variation characteristics; in addition, the state transition Markov chain between seasonal clusters shows the statistical characteristics of different seasons, thereby better taking into account the seasonal characteristics of daily wind speed samples.

[0105] 3) Step 3 of the present invention divides the quarterly wind speed into a long-term trend component and a random fluctuation component through EMD decomposition. The long-term trend component further takes into account the seasonal variation characteristics of the wind speed. The Ito random process simulates the random fluctuation component of the wind speed. This method simultaneously considers the randomness and autocorrelation of the wind speed.

[0106] 4) Step 4 of the present invention can accurately evaluate the wind resources in a certain area by calculating the wind energy evaluation index of the simulated wind speed, and provide a reference basis for the planning of the wind farm in the area.

[0107] 5) The simulated wind speed generated by the wind speed simulation model of the present invention has consistent seasonal statistical characteristics, probability distribution characteristics, and autocorrelation with the measured wind speed. Further calculation of wind energy evaluation indicators can accurately evaluate the wind energy resources in a certain area and further guide wind farm planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0109] Figure 1 The present invention is a flow chart of a wind speed simulation model taking seasonality into account.

[0110] Figure 2 Optimize the Canopy clustering algorithm flow chart for the PSO optimization algorithm.

[0111] Figure 3 This is the flow chart of wind speed based on optimized K-means clustering.

[0112] Figure 4 This is the flow chart of wind speed simulation based on EMD decomposition and Ito random process.

[0113] Figure 5(a) shows the clustering results of the wind speed samples measured at the Linton Observatory. Figure 1 ;

[0114] Figure 5(b) shows the clustering results of the wind speed samples measured at the Linton Observatory. Figure 2 ;

[0115] Figure 5(c) shows the clustering results of the wind speed samples measured at the Linton Observatory. Figure 3 ;

[0116] Figure 5(d) shows the clustering results of the wind speed samples measured at the Linton Observatory. Figure 4 ;

[0117] Figure 5(e) is a schematic diagram of the clustering results of the measured wind speed samples at the Linton Observation Station.

[0118] Figure 6 Schematic diagram of the quarterly trend component and random fluctuation component of the first cluster of wind speed decomposition.

[0119] Figure 7 This is the frequency distribution histogram of the wind speed samples measured at the Linton Observatory in the four quarters.

[0120] Figure 8 Frequency distribution histogram of simulated wind speed samples for the Linton Observatory in four quarters.

[0121] Fig. 9 This is the quarterly autocorrelation coefficient curve of the measured wind speed samples and simulated wind speed samples at the Linton Observatory. DETAILED DESCRIPTION

[0122] The specific implementation manner and working principle of the present invention are further described in detail below with reference to the accompanying drawings.

[0123] A method for seasonal wind speed simulation and wind energy assessment is proposed. The method simulates long-term wind speed and conducts wind energy assessment. First, the daily wind speed samples are roughly clustered by the optimized Canopy clustering method to obtain the optimal number of clusters and the initial cluster center. Then, the K-means clustering is used to divide the daily wind speed samples into clusters with different wind speed characteristics based on the initial cluster center. Affected by seasonal characteristics, most of the daily wind speeds in the same cluster belong to the same season, which can better reflect the seasonal characteristics of wind speed. Finally, the empirical mode decomposition (EMD) is applied to decompose the wind speed samples in the same cluster into seasonal trend components and random fluctuation components. Finally, the random fluctuation component of the cluster wind speed is simulated by the Ito random process to obtain the simulated random fluctuation component taking into account the autocorrelation, and then the final simulated wind speed is obtained by adding it to the seasonal trend component. This method accurately evaluates the wind energy in the region by generating simulated wind speed samples with the same seasonal characteristics as the measured wind speed samples and combining them with wind energy assessment indicators.

[0124] from Figure 1 It can be seen that a method for seasonal wind speed simulation and quarterly wind energy assessment includes the following steps:

[0125] S1: Establish the PSO-optimized Canopy algorithm, obtain the two optimal parameters of the Canopy clustering algorithm, and then use the optimal Canopy clustering algorithm to perform coarse clustering on the daily wind speed samples to obtain the optimal number of clusters and initial cluster centers;

[0126] S11: Collect wind speed samples measured at meteorological observation stations over a long period of time, record v1, v2, …v N is N groups of measured wind speed samples, which are separated by one hour. The wind speed samples are divided into units of days to obtain the daily wind speed matrix V = [V1, V2, …, V M ], where V m =[v 24(m-1)+1 ,v 24(m-1)+2 ,…,v 24(m-1)+24 ] is the 24-hour measured wind speed sample on the mth day (m=1,2,…,M);

[0127] S12: x is the position of a single particle on the one-dimensional coordinate axis, and its range of variation is limited to [X min ,X max ], v is the velocity of a single particle, and its range of variation is limited to [V min ,V max ], where X min and X max The following steps are performed:

[0128] (1) Calculate the average value: Calculate the daily wind speed average P = [p1, p2, ..., p24 ];

[0129] P=(V1+V2+...+V M ) / M

[0130] (2) Calculate the Euclidean distance: Calculate the wind speed sample V on the mth day according to the following formula m The Euclidean distance L with P constitutes the distance matrix L = [L1, L2, .., L M ];

[0131]

[0132] (3) Calculate X min and X max :The specific calculation formula is as follows;

[0133] X max =(L max -L min )×0.6

[0134] X min =L min ×0.35

[0135] Where L max is the maximum value in L, L min is the minimum value in L;

[0136] S13: This model sets two groups of particle populations, each group of populations consists of K particles. Taking the kth (k = 1, 2, ..., K) particle in the two groups of populations as an example, the position coordinate vector is X k =[x 1k ,x 2k ], where x 2k <x 1k , the fitness value is f k , f k The calculation process is as follows:

[0137] (1) The number of cluster centers Q = 1. A day is randomly selected from the daily wind speed matrix V as the first cluster center C. q , C=[C1,C2,..,C Q ], mark the wind speed of that day as Q and delete the data of that day from V;

[0138] (2) Traverse V and calculate the distance between the daily wind speed sample in V and the Q cluster centers in C. If the wind speed sample on that day is close to the cluster center C q The distance of (q=1,2,..,Q) is less than x 1k , mark the wind speed of the day as q and delete the sample data of the day from V, and then update the cluster center C with the mean of the wind speed samples marked as q q; If the distance to each cluster center is greater than x 2k , then Q = Q + 1, and the wind speed sample of that day is used as the new cluster center C Q , mark the wind speed of that day as Q, and then delete the data of that day from V;

[0139] (3) If V is not empty, repeat step (2); otherwise, end the algorithm;

[0140] (4) Calculate the Davidson-Boulding Index (DBI) as the fitness value f k , the calculation formula is as follows:

[0141]

[0142] in: is the average Euclidean distance from the i-th (i=1,2,…Q) class to its class center; ||w i -w j ||2 is the Euclidean distance between the class centers of the i-th class and the j-th (j=1,2,…Q) class;

[0143] The overall flow chart is as follows Figure 2 As shown:

[0144] (1) Initialization: The maximum number of iterations is N; in the particle position range [X min ,X max ] Randomly generate initial particle positions In the particle velocity range [V min ,V max ] Randomly generate initial particle speed Particle self-learning factor s1, group learning factor s2, maximum inertia weight W max , minimum inertia weight W min ;

[0145] (2) Calculate the fitness value f k :If the initial particle position According to the fitness value calculation process, the fitness value fk of the kth particle in the two groups of particle populations is calculated in turn, where k = (1, 2, ..., K), otherwise the fitness value f k is a larger number, which is set to e in the present invention 10 , two groups of particle populations K constitute the vector F = [f1,f2,…,f K ]:

[0146] (3) Optimal individual selection: Select the minimum fitness value f from the calculated result F bn , and record the minimum fitness value f during the nth iteration bn , forming a matrix F2 = [f b1 ,f b2 ,…fbn ]; corresponding to X bn Construct vector X B =[X b1 ,X b2 ,…X bn ],;

[0147] (4) Position update: Taking population 1 as an example, the following formula is used to update the particle position:

[0148]

[0149] Where r1, r2 are random numbers between [0,1], f m is the minimum value in matrix F2;

[0150] (5) Loop: Repeat steps (2) to (5) until the maximum number of iterations is reached, then stop the loop and the minimum value f in F2 is bmin Corresponding X bmin That is, the optimal threshold of the Canopy algorithm.

[0151] S14: according to the optimal threshold of the Canopy algorithm obtained in step S13, the daily wind speed samples are roughly clustered using the process described in step S13 to obtain the optimal number of clusters Q and the initial cluster center C;

[0152] from Figure 3 It can be seen that the daily wind speed is divided into clusters with different wind speed characteristics by using the number of clusters and the initial cluster center in step S1 through the K-means clustering model, including the following steps:

[0153] S21: taking the optimal number of clusters Q and the initial cluster center C obtained in step S1 as initial clustering input parameters of K-means clustering;

[0154] S22: Calculate the daily wind speed D on the mth day in the daily wind speed sample m The Euclidean distance to each cluster center Cq constitutes the matrix L m =[L1,L2,…,L Q ], find L m Minimum value L min The corresponding cluster number is q, and the wind speed on that day is marked as q. Apply formula C q =(C q +V m ) / 2 Update the cluster center C q ;

[0155] S23: repeat step S22 until all wind speed samples are divided;

[0156] S24: Use Markov chain to describe the random transfer relationship of daily wind speed in spring, summer, autumn and winter, P1, P2, P3 and P4, and then take into account the seasonality of wind speed. Taking spring as an example, the transfer probability of the daily wind speed sample of cluster i (i = 1, 2, ..., Q) to the daily wind speed sample of cluster j (j = 1, 2, ..., Q) in spring can be calculated by the following formula:

[0157]

[0158] Where N 1(i,j) is the number of times the measured spring daily wind speed samples are transferred from the i-th cluster wind speed to the j-th cluster wind speed, so the spring state transition probability matrix P1 is:

[0159]

[0160] from Figure 4 It can be seen that performing wind speed simulation on each cluster wind speed in step S3 to generate simulated wind speed samples taking into account seasonality, randomness and autocorrelation includes the following steps:

[0161] S31: Connect the wind speeds of the qth clusters in order, forming a series with T q Hourly wind speed samples v1, v2, …v Tq , where v t (t=1,2,…,T q ) is the wind speed sample at the tth moment;

[0162] S32: Decompose the qth cluster wind speed samples v1, v2, …, v Tq Decomposed into seasonal trend components s1, s2, …, s Tq and random fluctuation components r1, r2, …, r Tq , Taking the first cluster of wind speed samples as an example, the specific steps are as follows:

[0163] (1) Envelope fitting: wind speed samples v1, v2,…, v Tq For the original input signal v(t), find all its maximum points, and connect the maximum points into the maximum envelope y through the cubic spline function max (t); Similarly, find all the minimum points of v(t) and connect the minimum values ​​into the minimum envelope y through the cubic spline function. min (t);

[0164] (2) Calculate the intermediate wind speed: Use the following formula to calculate the intermediate wind speed v1(t):

[0165] v1(t)=v(t)-m1(t)

[0166]

[0167] (3) Calculate the intrinsic mode function: If the difference between the number of extreme points and the number of zero-crossing points in the entire data segment of v1(t) is not greater than 1, and the mean of the upper envelope and the lower envelope is zero, then v1(t) is an intrinsic mode function component of the wind speed sample v(t); otherwise, v1(t) is the original input signal, and steps (1) to (3) are repeated;

[0168] (4) Loop: e k (t) is the new original signal, where e k (t) = d(t) - d1(t), k is the number of cycles, if e k (t) is a monotonic function or a constant, then the process ends. Otherwise, repeat steps (1) to (4).

[0169] (5) Determine the weight: e k (t) is the seasonal trend component s(t) of d(t), that is, s(t) = e k (t), and the sum of k-1 intrinsic mode function components is taken as the random fluctuation component r(t), that is,

[0170] S33: Based on the decomposition result of step S32, use the Ito random process to model the random fluctuation component and calculate the simulated random fluctuation component at time t

[0171]

[0172] Where: Indicates t(t=1,2,…,L i ) is the simulated random fluctuation component at the moment; μ() is the drift term, which reflects the trend of the random variable moving toward the expected value. Its calculation formula is: For r t The mean of θ is the time autocorrelation function, and its calculation formula is R(τ)=e -θτ ; σ() is the diffusion term, reflecting W t Randomness affects simulation value; W t represents the Wiener process. It can be solved by choosing a suitable random sequence probability distribution.

[0173] S34: The simulated wind speed seasonal trend component s in step S32 is t As the seasonal trend component of simulated wind speed and the random fluctuation component of wind speed simulated in step S33 Add together to get the simulated wind speed:

[0174]

[0175] verify Is it in the interval [v min ,v max ], if yes, accept, if no, reject and repeat step S33 until the sample is accepted.

[0176] S35: Repeat steps S33 and S34 until T is generated i The wind speed samples of all clusters are simulated, and steps S31 to S35 are repeated until the simulation of wind speed samples of all clusters is completed.

[0177] S36: According to the state transfer matrices P1, P2, P3 and P4 of the four seasons of spring, summer, autumn and winter, a simulated wind speed sample of one day is sequentially extracted from the k clusters of simulated wind speed samples until N groups of simulated wind speeds are generated, where N = T1 + T2 +, ... + T k .

[0178] In step S4, the simulated wind speed sample in step S3 is used to calculate the wind energy evaluation index and evaluate the wind energy resources in the region. The specific process is as follows:

[0179] S41: Use the following formula to calculate the quarterly mean wind speed v of the simulated wind speed sample m.p and seasonal wind speed standard deviation v s.p :

[0180]

[0181] The superscript “^” in the formula represents the simulated wind speed value. and The seasonal wind speed mean and seasonal wind speed standard deviation of the simulated wind speed are respectively represented, where p = (1, 2, 3, 4) represents spring, summer, autumn and winter respectively; taking spring as an example, N1 is the number of spring wind speed samples in the simulated wind speed samples; is the simulated wind speed sample at the i-th moment in spring;

[0182] S42: Calculate the quarterly average wind power density of the simulated wind speed using the following formula

[0183]

[0184] Where ρ is the air density, usually 1.29 kg / m 3 ;

[0185] S43: Calculate the quarterly average effective wind power density of the simulated wind speed using the following formula

[0186]

[0187] Where v ci and v coare the cut-in wind speed and cut-out wind speed of the wind turbine, respectively, pi is the indicator factor, dimensionless;

[0188] S44: Use the following formula to calculate the quarterly effective hours T of the simulated wind speed p and quarterly utilization efficiency η p :

[0189]

[0190] Where N bp is the number of hours in the four quarters of the year, N b1 、N b2 、N b3 and N b4 The values ​​are 2208, 2208, 2184 and 2160 respectively;

[0191] S45: Calculate the quarterly power generation E of the simulated wind speed using the following formula p :

[0192]

[0193] Where P p () is the wind power conversion function of the wind turbine, which is shown in the following formula:

[0194]

[0195] Where v is the instantaneous wind speed; v τ and P τ are the rated wind speed and rated capacity of the wind turbine respectively. The coefficients A, B and C can be obtained from v ci and v τ The specific calculation formula is as follows:

[0196]

[0197] S46: Calculate the seasonal capacity factor C of the simulated wind speed using the following formula p :

[0198]

[0199] Example:

[0200] The present invention collects the open meteorological database North Dakota agricultural weather network (NDAWN) and obtains the measured wind speed samples with a time step of 1 hour from March 2011 to February 2020 at the Linton meteorological observation station. A wind speed simulation model taking into account seasonality in the present invention is used to simulate ten years of wind speed samples. By comparing the seasonal wind speed mean and standard deviation, seasonal probability distribution characteristics and seasonal autocorrelation characteristics of the measured wind speed samples of the observation station and the simulated wind speed of the present invention, it is verified that the present invention can accurately take into account the seasonality of wind speed.

[0201] In step S1, the particle range of the PSO optimization algorithm is [203.98, 335.13]; its movement speed range is [-1, 1], the particle self-learning factor s1 and the group learning factor s2 are both 2, and the maximum inertia weight W max =0.8, minimum inertia weight W min =0.3.

[0202] Figure 5(a) to Figure 5(e) The figure shows the clustering results of the measured wind speed samples at the Linton Observatory. It can be seen that the improved K-means clustering model clusters the daily wind speeds at the Linton Meteorological Observatory from March 2011 to February 2020 into five wind speed clusters with different characteristics. Among them: the wind speed in the cluster shown in Figure 5(a) is relatively low, and there is no significant change within a day; the wind speed in the cluster shown in Figure 5(b) has obvious day and night characteristics, and the wind speed gradually increases from the 10th hour; the day and night characteristics of the wind speed in the cluster shown in Figure 5(c) are not obvious, but the overall wind speed is higher than that in Figure 5(a); the wind speed in the cluster shown in Figure 5(b) is relatively high, and has obvious periodic changes, and the wind speed gradually decreases from noon; the wind speed in the cluster shown in Figure 5(e) is generally high. The proportion of the five types of wind speeds in the fourth quarter is shown in Table 1. Among them, the five types of wind speeds account for a relatively even proportion in spring, the first type of wind speed accounts for the largest proportion in summer, and the fifth type of wind speed accounts for the smallest proportion. The proportions of the five types of wind speeds in autumn and winter are relatively similar. The different proportions of the five types of wind speeds in the fourth quarter indicate that wind speed has obvious quarterly characteristics.

[0203] Table 1 Proportion of five types of wind speed in the fourth quarter (unit: %)

[0204]

[0205] Figure 6 The measured wind speed samples of the first cluster of wind speed after clustering and the random fluctuation component and seasonal trend component obtained after decomposition. It can be seen that the random fluctuation component has similar fluctuations to the measured wind speed samples. The seasonal trend component shows the changes of the samples at the measured sample scale, and the overall trend is slowly decreasing.

[0206] The seasonal wind speed simulation and quarterly wind energy assessment method of the present invention are applied to simulate wind speeds for ten years, and the statistical index results are shown in Table 2. The annual and quarterly means and standard deviations of the simulated wind speed samples and the measured wind speed samples of the present invention are very close, indicating that the present invention can better retain the annual and quarterly statistical indicators and quarterly trends close to the measured wind speed samples.

[0207] Table 2 Annual and quarterly means and standard deviations of measured and simulated wind speeds

[0208]

[0209] Figure 7 and Figure 8 They are the quarterly frequency distribution histograms of the measured wind speed samples and the quarterly frequency distribution histograms of the simulated wind speed samples, respectively. It can be seen that the wind speed at the Linton Observatory has obvious quarterly characteristics, with a higher probability of low wind speed in summer and a higher probability of high wind speed in spring, which is consistent with the fact that the mean wind speed in spring is greater than the mean wind speed in summer in Table 1. The annual and quarterly probability distributions of the simulated wind speed samples of the present invention are highly similar to those of the measured wind speed samples, indicating that the simulated wind speed of the present invention can better retain the statistical characteristics of the measured wind speed.

[0210] Fig. 9 It is the autocorrelation coefficient curve of the measured wind speed sample and the simulated wind speed of the present invention in four seasons. It can be seen that the autocorrelation coefficient curves of the measured wind speed sample and the simulated wind speed of the present invention in spring, autumn and winter have a high degree of fit within a lag of 5 hours, and the autocorrelation coefficient curve of the measured wind speed sample and the simulated wind speed of the present invention in summer has a relatively low degree of fit within a lag of 5 hours, but overall, this simulation model can better account for the autocorrelation of wind speed in four seasons.

[0211] Take a single 1.5MW wind turbine generator as an example to calculate the wind energy density in the area. The cut-in, rated, and cut-out wind speeds of the wind turbine are 3m / s, 10.5m / s, and 25m / s, respectively. Its rated power and hub height are 1.5MW and 100m, respectively. The wind speed measurement height of the Linton Observatory is 3m. According to the formula Extrapolate the simulated wind speed vertically to the hub height, where v hub is the hub height of the wind turbine, v0 is the measurement height of the observation station, is a power law exponent that depends on the surface roughness, and in this embodiment, the value is 0.15. The wind power density and the average effective wind power density in the four seasons at the hub height are shown in Table 3. The wind power density of the Linton Observatory in the four quarters remains at a high level. According to the wind power density classification table, the Linton area has good wind energy resources. The average effective wind power density in the four quarters is lower than the wind power density, but still has the same trend of change. The wind energy resources in spring are the most abundant, the wind energy resources in summer are relatively low, but there are still objective wind energy resources, and the wind energy resources in autumn and winter are relatively close.

[0212] Table 3 Simulated wind speed seasonal monsoon power density and average effective wind power density in four seasons

[0213]

[0214] The effective hours and effective utilization rate of the simulated wind speed in the four quarters at the hub height are shown in Table 4. The effective hours of wind speed in spring are the highest, and the effective hours of wind speed in autumn are the lowest, but the effective utilization rate in the four quarters is not much different, all maintained at around 90%, with a high effective utilization rate, further indicating that the Linton area has good wind energy resources. The quarterly power generation and quarterly capacity factor of a single 1.5MW wind turbine generator under simulated wind speed are shown in Table 5. The power generation in the fourth quarter is consistent with the average wind speed change trend in Table 1. The average wind speed in spring is the highest, and this quarter also has the highest power generation point. The power generation in summer is the lowest, and the power generation in autumn and winter is not much different. The capacity factor combines wind speed and wind turbine characteristics to better measure the power generation utilization rate from an energy perspective. As shown in Table 5, the fourth quarter is maintained at a good level, indicating that the Linton area has abundant wind energy resources.

[0215] Table 4 Effective hours and effective utilization rate of simulated wind speed in four quarters

[0216]

[0217] Table 5 Quarterly power generation and capacity factor of wind turbines in the four simulated wind speed quarters

[0218]

[0219] From the above results, it can be seen that the wind speed simulation method of the present invention can better take into account the seasonality of the measured wind speed, and the generated simulated wind speed samples have consistent seasonal statistical indicators, probability distribution characteristics and autocorrelation characteristics with the measured wind speed samples. The wind energy evaluation index of the Linton area is calculated using the simulated wind speed of the present invention. The results show that the area has rich wind energy resources, providing a reference for wind farm investment in the area.

Claims

1. A seasonal wind energy assessment method taking into account seasonal wind speed simulation, characterized in that The following steps are involved: Step 1: Optimize the Canopy clustering algorithm based on the particle swarm optimization algorithm. After obtaining the two optimal parameters of the Canopy clustering algorithm, use the optimal Canopy clustering algorithm to perform coarse clustering on the daily wind speed samples to obtain the optimal number of clusters and initial cluster centers; Step 2: Using the number of clusters and the initial cluster centers in step 1, the K-means clustering model is combined to divide the daily wind speed into clusters with different wind speed characteristics, and the state transition Markov chain between clusters in each season is combined to calculate the seasonality of wind speed; Step 3: Perform wind speed simulation on each cluster in step 2 to generate simulated wind speed samples that take seasonality into account; Step 4: Using the simulated wind speed samples in step 3, calculate the wind energy assessment index and evaluate the wind energy resources in the area.

2. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 1, characterized in that: The step 1 comprises the following steps: S1.1: Collect the measured wind speed samples of a certain observation station, divide the measured wind speed samples into 24 hours a day, and obtain the daily wind speed samples; S1.2: Apply the PSO optimization algorithm to optimize the two distance thresholds in the Canopy clustering algorithm to obtain the optimal Canopy clustering algorithm; S1.3: Use the optimal Canopy algorithm of S1.2 to roughly cluster the daily wind speed samples and obtain the optimal number of clusters and initial cluster centers.

3. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 2, characterized in that: In S1.1, the wind speed samples measured over a long period of time at the meteorological observation station are collected and recorded as v1, v2, …v N is N groups of measured wind speed samples, which are separated by one hour. The measured wind speed samples are divided into units of days to obtain the daily wind speed matrix V = [V1, V2, …, V M ],V1,V2,…,V M Respectively represent the matrix composed of 24-hour wind speed samples on the mth day, where V m =[v 24(m-1)+1 ,v 24(m-1)+2 ,…,v 24(m-1)+24 ] is the 24-hour measured wind speed sample on the mth day, v 24(m-1)+1 ,v 24(m-1)+2 ,…,v 24(m-1)+24 They represent the measured wind speed samples from the 1st hour to the 24th hour of the mth day, respectively, m=1, 2, …, M, and M represents the number of days in a year.

4. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 2, characterized in that: In S1.2, x is the position of a single particle on a one-dimensional coordinate axis, and its range of variation is limited to [X min ,X max ], X min ,X max They represent the lower and upper limits of the position of a single particle, respectively. v is the velocity of a single particle, and its range of variation is limited to [V min ,V max ],V min ,V max They represent the lower and upper limits of the speed of a single particle, respectively, where X min and X max The following steps are performed: (1) Find the mean: The daily wind speed average P is calculated according to the following formula: [p1, p2, ..., p 24 ]; p1, p2, …, p 24 Respectively represent the mean wind speed of all daily wind speed samples from the 1st to the 24th hour: P=(V1+V2+...+V M ) / M (2) Calculate the Euclidean distance: Calculate the wind speed sample V on the mth day according to the following formula m The Euclidean distance L with P constitutes the distance matrix L = [L1,L2,..,L M ],L1,L2,..,L M Respectively represent the wind speed sample V on the mth day m Euclidean distance from P: (3) Calculate X min and X max , the specific calculation formula is as follows: X max =(L max -L min )×0.6 X min =L min ×0.35 Where: L max is the maximum value in L, L min is the minimum value in L.

5. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 4, characterized in that: In S1.2, the PSO optimization algorithm sets two groups of particle populations, each group of populations consists of K particles. Taking the kth particle in the two groups of populations as an example, k = 1, 2, ..., K, K represents the number of particles in the two groups of populations, and the position coordinate vector is X k =[x 1k ,x 2k ], x 1k , x2 represent the position of the kth particle in particle population 1 and the position of the kth particle in population 2, respectively, where x 2k <x 1k , the fitness value is f k , f k The calculation process is as follows: 1) The number of cluster centers Q = 1, and a day is randomly selected from the daily wind speed matrix V as the first cluster center C q , C=[C1,C2,..,C Q ], mark the wind speed of that day as Q and delete the data of that day from V; 2) Traverse V and calculate the distances from the daily wind speed samples in V to the Q cluster centers in C; If the wind speed sample of the day reaches the cluster center C q The distance is less than x 1k , q=1,2,..,Q, Q represents the total number of cluster centers; The wind speed of the day is marked as q and the sample data of the day is deleted from V. Then the cluster center C is updated with the mean of the wind speed samples marked as q. q ; If the distance to each cluster center is greater than x 2k , then Q = Q + 1, and the wind speed sample of that day is used as the new cluster center C Q , mark the wind speed of that day as Q, and then delete the data of that day from V; 3) If V is not empty, repeat step 2); otherwise, end the algorithm; 4) Calculate the Davidson-Boulding Index (DBI) as the fitness value f k , the calculation formula is as follows: in: is the average Euclidean distance from the i-th class to its class center, i = 1, 2, ... Q, Q represents the total number of cluster centers; S j represents the average Euclidean distance from the jth class to its class center, w i represents the Euclidean distance of the center of the i-th class, w j represents the Euclidean distance of the center of the i-th class.

6. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 2, characterized in that: The S1.2 comprises the following steps: Step 1. Initialization: The maximum number of iterations is N; in the particle position range [X min ,X max ] Randomly generate initial particle positions represents the initial position of the kth particle in particle population 1, represents the initial position of the kth particle in particle population 2, in the particle velocity range [V min ,V max ] Randomly generate initial particle speed represents the initial motion speed of the kth particle in particle population 1, represents the initial motion speed of the kth particle in particle population 2, particle self-learning factor s1, group learning factor s2, and maximum inertia weight W max , minimum inertia weight W min ; Step 2. Find the fitness value f k :If the initial particle position According to the fitness value calculation process, the fitness value fk of the kth particle in the two groups of particle populations is calculated in turn, where k = (1, 2, ..., K), otherwise the fitness value f k is a larger number, two groups of particle populations K constitute the vector F = [f1,f2,…,f K ],f1,f2,…,f K They represent the fitness values ​​calculated for the kth particle in the two populations respectively; Step 3. Optimize individuals: Select the minimum fitness value f from the calculated result F bn , and record the minimum fitness value f during the nth iteration bn , forming a matrix F2 = [f b1 ,f b2 ,…f bn ],f b1 ,f b2 ,…f bn They represent the minimum fitness value in the vector F formed by the nth iteration, and the corresponding X bn Construct vector X B =[X b1 ,X b2 ,…X bn ],X b1 ,X b2 ,…X bn They respectively represent the position coordinates corresponding to the minimum fitness value in the vector F formed by the nth iteration; Step 4. Position update: Taking population 1 as an example, apply the following formula to update the particle position: Where: r1, r2 are random numbers between [0,1]; f m is the minimum value in the matrix F2; s1 represents the self-learning factor of the particle; s2 represents the group learning factor of the particle; Step 5. Loop: Repeat the process from Step 2 to Step 4 until the maximum number of iterations is reached, then stop the loop and the minimum value f in F2 is bmin Corresponding X bmin That is, the optimal threshold of the Canopy clustering algorithm.

7. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 2, characterized in that: In S1.3, according to the optimal threshold of the Canopy clustering algorithm obtained in step S1.3, the daily wind speed samples are roughly clustered using the process described in S1.3 to obtain the optimal number of clusters Q and the initial cluster center C.

8. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 1, characterized in that: In step 2, the initial cluster center in step S1 is used as the cluster center randomly selected initially by K-means clustering, and K-means clustering is performed on the daily wind speed samples to divide the daily wind speed samples into wind speed clusters with different characteristics; based on the clustering results, the daily wind speed state transition Markov chain of the fourth quarter is calculated; The following steps are involved: S2.1: Use the optimal number of clusters Q and the initial cluster center C obtained in step S1 as the initial clustering input parameters of K-means clustering; S2.2: Calculate the daily wind speed D on the mth day in the daily wind speed sample m The Euclidean distance to each cluster center Cq constitutes the matrix L m =[L1,L2,…,L Q ],L1,L2,…,L Q represents the daily wind speed D on the mth day m To the Euclidean distance from the 1st cluster center to the Qth cluster center, find L m Minimum value L min The corresponding cluster number is q, and the wind speed on that day is marked as q. Apply formula C q =(C q +V m ) / 2 Update the cluster center C q ; C q represents the qth cluster center, V m represents the daily wind speed on the mth day; S2.3: repeat step S2.2 until all wind speed samples are divided; S2.4: Use Markov chain to describe the random transfer relationship of daily wind speed in spring, summer, autumn and winter, P1, P2, P3 and P4, and then take into account the seasonality of wind speed; the transfer probability of the daily wind speed sample of the i-th cluster to the daily wind speed sample of the j-th cluster in spring can be calculated by the following formula: Where: p 1(i,j) represents the probability of the i-th wind speed sample transferring to the j-th wind speed sample in spring, K represents the number of daily wind speed sample types, N 1(i,j) is the number of times the measured spring daily wind speed samples are transferred from the i-th cluster wind speed to the j-th cluster wind speed, so the spring state transition probability matrix P1 is: Where: p 1(1,1) It represents the probability of the first type of wind speed sample in spring transferring to the second type of wind speed sample, p 1(1,2) It represents the probability of the second type of wind speed sample in spring transferring to the second type of wind speed sample, p 1(1,K) It represents the probability of the first-class wind speed sample transferring to the K-class wind speed sample in spring; p 1(2,1) represents the probability of the second-class wind speed sample transferring to the first-class wind speed sample in spring, p 1(2,2) It represents the probability of the second type of wind speed sample in spring transferring to the second type of wind speed sample, p 1(2,K) It represents the probability of the second-class wind speed sample transferring to the K-class wind speed sample in spring; p 1(K,1) represents the probability of the K-th wind speed sample transferring to the 1st wind speed sample in spring, p 1(K,2) represents the probability of the K-th wind speed sample transferring to the second-class wind speed sample in spring, p 1(K,K) It represents the probability of the Kth wind speed sample transferring to the Kth wind speed sample in spring.

9. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 1, characterized in that: In the step 3, wind speed simulation is performed on each cluster in step S2 to generate simulated wind speed samples taking into account seasonality, randomness and autocorrelation; the steps include: S3.1: Connect the wind speeds of the qth clusters in order, forming a series with T q Hourly wind speed samples v1, v2, …v Tq , where v t is the wind speed sample at time t, t=1,2,…,T q They represent the hourly time of the wind speed of the qth cluster day, S3.2: Decompose the qth cluster wind speed samples v1, v2, …, v Tq Decomposed into seasonal trend components s1, s2, …, s Tq and random fluctuation components r1, r2, …, r Tq ,,s1,s2,…,s Tq Respectively represent the 1st to T q seasonal trend components at time, r1, r2, …, r Tq Respectively represent the 1st to T q The random fluctuation component at each moment; The first cluster of wind speed samples, the specific steps are as follows: (1) Envelope fitting: wind speed samples v1, v2,…, v Tq For the original input signal v(t), find all its maximum points, and connect the maximum points into the maximum envelope y through the cubic spline function max (t); Similarly, find all the minimum points of v(t) and connect the minimum values ​​into the minimum envelope y through the cubic spline function. min (t); (2) Calculate the intermediate wind speed: Use the following formula to calculate the intermediate wind speed v1(t): v1(t)=v(t)-m1(t) Where: m1(t) represents the median wind speed of the first cluster of wind speed samples; (3) Calculate the intrinsic mode function: If the difference between the number of extreme points and the number of zero-crossing points in the entire data segment of v1(t) is not greater than 1, and the mean of the upper envelope and the lower envelope is zero, then v1(t) is an intrinsic mode function component of the wind speed sample v(t); otherwise, v1(t) is the original input signal, and steps (1) to (3) are repeated; (4) Loop: e k (t) is the new original signal, where e k (t) = v(t) - v1(t), v(t) represents the input wind speed time series, v1(t) represents the intrinsic mode function component of the input wind speed time series, k is the number of cycles, if e k (t) is a monotonic function or a constant, then the process ends. Otherwise, repeat steps (1) to (4). (5) Determine the weight: e k (t) is the seasonal trend component s(t) of d(t), that is, s(t) = e k (t), and the sum of k-1 intrinsic mode function components is taken as the random fluctuation component r(t), that is, e i (t) represents the i-th natural mode component; S3.3: Based on the decomposition results of step S3.2, use the Ito random process to model the random fluctuation component and calculate the simulated random fluctuation component at time t Where: represents the simulated random fluctuation component at time t, t=1,2,…,L i The hour and time of the simulated wind speed respectively; μ() is the drift term, which reflects the tendency of the random variable to move toward the expected value. Its calculation formula is: For r t The mean of θ is the time autocorrelation function, and its calculation formula is R(τ)=e -θτ ; σ() is the diffusion term, reflecting W t Randomness affects simulation value; W t represents the Wiener process; By choosing a suitable random sequence probability distribution, it can be solved; S3.4: The simulated wind speed seasonal trend component s in step S3.2 t As the seasonal trend component of simulated wind speed and the random fluctuation component of wind speed simulated in step S3.3 Add together to get the simulated wind speed: verify Is it in the interval [v min ,v max ], if yes, accept, if no, reject and repeat step S3.3 until the sample is accepted; S3.5: Repeat steps S3.3 and S3.4 until T is generated i The wind speed samples of all clusters are simulated, and steps S31 to S35 are repeated until the simulation of wind speed samples of all clusters is completed; S3.6: According to the state transfer matrices P1, P2, P3 and P4 of the four seasons of spring, summer, autumn and winter, one day's simulated wind speed samples are sequentially extracted from the k clusters of simulated wind speed samples until N groups of simulated wind speeds are generated, where N = T1 + T2 + T3 + T4, T1, T2, T3, T4 represent the number of simulated wind speed samples in spring, the number of simulated wind speed samples in summer, the number of simulated wind speed samples in autumn and the number of simulated wind speed samples in winter respectively.

10. A seasonal wind energy assessment method taking into account seasonal wind speed simulation according to claim 1, characterized in that: In step 4, quarterly wind energy assessment indicators such as quarterly wind speed mean, quarterly wind speed standard deviation, quarterly wind power density, and average effective wind power are calculated according to the simulated wind speed, and the wind energy resources in the region are evaluated through the calculation results; Includes the following: 4.1: Use the following formula to calculate the quarterly mean wind speed v of the simulated wind speed sample: m.p and seasonal wind speed standard deviation v s.p : Where: the superscript "^" represents the wind speed simulation value, and The seasonal wind speed mean and seasonal wind speed standard deviation of the simulated wind speed are respectively represented, where p = (1, 2, 3, 4) represents spring, summer, autumn and winter respectively; taking spring as an example, N1 is the number of spring wind speed samples in the simulated wind speed samples; is the simulated wind speed sample at the i-th moment in spring; 4.2: Use the following formula to calculate the seasonal average wind power density of the simulated wind speed: Where: ρ is the air density; 4.3: Use the following formula to calculate the seasonal average effective wind power density of the simulated wind speed: Where: v ci and v co are the cut-in wind speed and cut-out wind speed of the wind turbine respectively; I pi is the indicator factor; 4.4: Use the following formula to calculate the quarterly effective hours T of the simulated wind speed: p and quarterly utilization efficiency η p : Where: N bp The number of hours for the four quarters of the year; 4.5: Use the following formula to calculate the quarterly power generation E of the simulated wind speed p : Where: P p () is the wind power conversion function of the wind turbine, which is shown in the following formula: Where: v is the instantaneous wind speed; v τ and P τ are the rated wind speed and rated capacity of the wind turbine respectively; coefficients A, B and C are given by v ci and v τ The specific calculation formula is as follows: 4.6: Use the following formula to calculate the seasonal capacity factor C of the simulated wind speed: p :

Images