Solar irradiance probability modeling method and device based on non-parametric solution set

Abstract

The present invention discloses a solar irradiance probability modeling method and device based on a non-parametric solution set. The method comprises the following steps: Step 1: defining adjacent daily irradiance and hourly irradiance, obtaining daily irradiance sample data and hourly irradiance sample data, and respectively calculating the sample covariance and bandwidth coefficient of the adjacent daily irradiance and hourly irradiance; Step 2: selecting a Gaussian function as a kernel function, constructing a conditional kernel density estimation model for daily irradiance, and generating daily irradiance random sample data by randomly generating Gaussian functions; Step 3: selecting a Gaussian function as a kernel function, expressing daily irradiance as the sum of hourly irradiances, and establishing an hourly irradiance conditional kernel density estimation model based on daily irradiance and the previous day's hourly irradiance, and generating hourly irradiance random sample data by randomly selecting a multidimensional Gaussian function. The present invention has the advantages of simple implementation method, high modeling accuracy, and wide application range.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation reliability assessment, and in particular to a solar irradiance probability modeling method and device based on a non-parametric solution set. Background Art

[0002] Solar irradiance is the primary factor affecting photovoltaic output. Establishing a reasonable and accurate irradiance probabilistic model is key to assessing the reliability of power systems containing photovoltaic power plants. Due to environmental factors, solar irradiance is random, intermittent, and fluctuating. Furthermore, the output of different photovoltaic power plants in adjacent regions exhibits a certain degree of correlation. The correlation structure and magnitude of irradiance can affect the output of adjacent distributed photovoltaic power plants, and thus, the reliability assessment of the power system.

[0003] In the analysis of the randomness of irradiance, the existing technology generally adopts methods based on the Normal distribution, Lognormal distribution, Weibull distribution, Extreme Value distribution, Beta distribution, etc. to establish the probability model of irradiance. Since the correlation problem must be considered when modeling the irradiance of multiple adjacent photovoltaic power stations, the existing technology generally adopts methods such as ARMA model and the formation of solar irradiance time series dynamic probability distribution model based on the superposition of time functions to establish the photovoltaic output time series probability model, and then establish various irradiance probability models that take into account the correlation. However, solar irradiance is also random, and the above models cannot effectively represent the randomness of solar irradiance. Although the method of using the half-sine model to distribute the total daily radiation hourly according to a preset time function can represent the randomness to a certain extent, this mechanical method of obtaining hourly irradiance according to a deterministic law is difficult to accurately reflect the random fluctuation of hourly irradiance.

[0004] In addition to its randomness and correlation, solar irradiance also exhibits an additive characteristic between daily and hourly irradiance, and there is a strong correlation between the hourly irradiance of the previous day and the hourly irradiance of the current simulated day. The solar irradiance probabilities established in existing technologies cannot simultaneously account for the randomness, correlation, and additive characteristics of irradiance, nor can they reflect the correlation between the hourly irradiance of the previous day and the hourly irradiance of the current simulated day. Some practitioners use solution set theory to establish a probability solution set parameter model for converting total daily irradiance to hourly irradiance. However, the probability distribution type of solar irradiance in this model must be artificially assumed in advance, and the consistency between the subjectively assumed probability distribution type and the actual distribution is often difficult to ensure, resulting in low actual model accuracy. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: in response to the technical problems existing in the prior art, the present invention provides a solar irradiance probability modeling method and device based on a non-parametric solution set, which has a simple implementation method, high modeling accuracy and a wide range of applications. The method and device can comprehensively consider the correlation between the hourly irradiance of the previous day and the hourly irradiance of the current simulated day, accurately simulate the additive characteristics of the irradiance at each moment and the total daily irradiance, and then realize the accurate simulation of the probability distribution of the irradiance of the photovoltaic power station at each moment, thereby improving the accuracy of the reliability assessment of the new power system containing the photovoltaic power station.

[0006] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0007] A solar irradiance probability modeling method based on a non-parametric solution set comprises the following steps:

[0008] Step 1: Initialization of sample parameters: define the adjacent day irradiance I t =[I t ,I t-1 ] and r t =[r 1,t ,r 2,t ,…,r d,t ] T , where I t ,I t-1 The daily irradiance data of the tth day and the t-1th day are obtained respectively. The daily irradiance data of multiple days are formed into daily irradiance sample data and hourly irradiance sample data. t represents the random vector composed of hourly irradiance on day t, d is the number of hours in a day, and the irradiance of adjacent days I is calculated based on the obtained daily irradiance and hourly irradiance sample data. t and hourly irradiance r t The sample covariance and bandwidth coefficient of ;

[0009] Step 2: Randomly generate daily irradiance samples: select Gaussian function as kernel function, based on the adjacent daily irradiance I t The sample covariance and bandwidth coefficient of the daily irradiance are used to construct a conditional kernel density estimation model; by randomly generating a Gaussian function, the known I is calculated according to the conditional kernel density estimation model of the daily irradiance. t-1 Daily irradiance under conditions I t , generate random sample data of daily irradiance;

[0010] Step 3: Randomly generate hourly irradiance samples: select Gaussian function as kernel function and transform daily irradiance I t Expressed as hourly irradiance r t The sum of the time and based on the daily irradiance I t and the hourly irradiance r of the previous day t-1A non-parametric solution set model of hourly irradiance is established, and the hourly irradiance r of the previous day is calculated based on the non-parametric solution set model of hourly irradiance by randomly selecting a multi-dimensional Gaussian function. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t , generate random sample data of hourly irradiance.

[0011] Furthermore, the hourly irradiance non-parametric solution model established in step 3 is:

[0012]

[0013] in:

[0014]

[0015]

[0016]

[0017]

[0018]

[0019] Among them, U t =(y 1,t ,y 2,t ,…,y d-1,t ) T , Y t =(y 1,t ,y 2,t ,…,y d,t ) T , Y t =Rr t , R is the unit orthogonal matrix, d is the number of hours per day, V t ′=(U t-1 ,I t ′) T , λ V′ Indicates V t ′’s bandwidth coefficient, λ UV′ Indicates (U t ,V t Bandwidth coefficient of ′), S UV′ For U t and V t ′ between the (d-1) × d order sample covariance matrix, S V′ V t ′’s d×d-order sample covariance matrix, i represents the serial number of the day, and n is the total number of days of daily irradiance sample data.

[0020] Furthermore, in step 3, the hourly irradiance r of the previous day is calculated by randomly selecting a multidimensional Gaussian function according to the hourly irradiance non-parametric solution model. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t , the steps to generate random sample data of hourly irradiance include:

[0021] Step 3.1: Set the initial value of hourly irradiance to r t-1 ;

[0022] Step 3.2: According to the current daily irradiance I t , hourly irradiance r t-1 The values of sample data V are calculated respectively. i ′’s weight value ω i ,i=1,2,…,n,to divide the interval [0,1] into n intervals of length ω i subintervals of ;

[0023] Step 3.3: Generate a random number x that follows a uniform distribution between [0,1]. If x falls in the kth subinterval of the interval [0,1], that is, Then the kth Gaussian function is selected in the current sampling, where the mean vector is B k , the covariance matrix is

[0024] Step 3.4: Generate a (d-1)-dimensional Gaussian column vector Q and obtain a random vector U t =B k +λ UV′ LQ, and then get Y t =(U t T ,I t ′) T ;

[0025] Step 3.5: By linear transformation r t =R T Y t , we get the solution set of daily irradiance to hourly irradiance, that is, we get the random sample of hourly irradiance;

[0026] Step 3.6: Determine whether t reaches the preset threshold. If not, set t=t+1 and go to step 3.3). Otherwise, the calculation is terminated and the final hourly irradiance random sample output is generated.

[0027] Furthermore, the conditional probability density estimation model of daily irradiance established in step 2 is:

[0028]

[0029] in:

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

[0036] Among them, λ t ,λ t-1 The daily irradiance I t , I t-1 The sample variance of λ t-1 =[(n-1)(4π) 1 / 2 R(f t-1 )] -1 / 5 , R(f t )=∫tr 2 {S t ·f″(I t )}dI t , R(f t-1 )=∫tr 2 {S q ·f″(I t-1 )}dI t-1 , S t is the irradiance of the adjacent day I t The sample covariance matrix, S h is the daily irradiance I t The sample variance of S hq is the daily irradiance I t and I t-1 The sample covariance of S q is the daily irradiance I t-1 The sample variance of , n is the total number of days of daily irradiance sample data.

[0037] Furthermore, in step 2, the conditional kernel density estimation model of the daily irradiance is used to calculate the known I t-1 Daily irradiance under conditions I t , the steps of generating random sample data of daily irradiance include:

[0038] Step 2.1: Set t = 2 and set the initial value of daily irradiance I t-1 ;

[0039] Step 2.2: Calculate I for each sample ti-1 In the conditional quantity I t-1 The contribution weight ω under ti (i=1,2,…,n-1), which divides the [0,1] interval into n-1 intervals of length ω ti subintervals of ;

[0040] Step 2.3: Generate a random number x that follows a uniform distribution between [0,1]. If x falls in the kth subinterval of the interval [0,1], that is, Then the kth Gaussian function is selected for this sampling, and its mean is B tk , the variance is

[0041] Step 2.4: Randomly generate a Gaussian function V k , thus obtaining the known condition quantity I t-1 The daily irradiance I on the tth day t =B tk +λ t L t V k ;

[0042] Step 2.5: Determine whether t is less than the preset threshold. If it is, set I t Assign I t-1 As the new conditional quantity, let t=t+1 and go to step 2.3), otherwise go to step 3.

[0043] Furthermore, in step 1, the adjacent day irradiance I is calculated based on the obtained daily irradiance and hourly irradiance sample data. t and hourly irradiance r t Steps to calculate the sample covariance matrix:

[0044] Hourly irradiance r t Perform the following coordinate transformation:

[0045] Y t =Rr t

[0046] Where Y t =(y 1,t ,y 2,t ,…,y d,t ) T ; R is the unit orthogonal matrix, that is, R T =R -1 ;definition U t =(y 1,t ,y 2,t ,…,y d-1,t )T , then Y t =(y 1,t ,y 2,t ,…,y d,t ) T =(U t T ,I t ′) T , and let V t ′=(U t-1 ,I t ′) T ;

[0047] Calculate daily irradiance I t The sample covariance matrix S t :

[0048]

[0049] Among them, S h For I t The sample variance of S hq For I t and I t-1 The sample covariance of S q For I t-1 The sample variance of .

[0050] Calculation (U t ,V t The sample covariance matrix S between ′) is:

[0051]

[0052] Where S U For U t The (d-1)×(d-1)-order symmetric sample covariance matrix, S UV′ For U t and V t ′’s (d-1)×d-order sample covariance matrix, S V′ V t ′ is the d×d order sample covariance matrix.

[0053] Furthermore, step 1 also includes:

[0054] Pair S t Perform the following elementary transformations

[0055]

[0056] but

[0057] And make the following elementary transformations on S:

[0058]

[0059] Among them, E d-1 is a d-1 order unit matrix, E d Then it is a unit matrix of order d;

[0060] make To A t and A to perform Cholesky decomposition to obtain L t and L, where A t =L t L t T 、A=LL T .

[0061] A solar irradiance probability modeling device based on a non-parametric solution set, comprising:

[0062] Sample parameter initialization module, used to define I t =[I t ,I t-1 ] and r t =[r 1,t ,r 2,t ,…,r d,t ] T , where I t ,I t-1 are the daily irradiance on day t and day t-1, r t represents a random vector composed of hourly irradiance on the tth day, obtains daily irradiance and hourly irradiance data of multiple days to form daily irradiance and hourly irradiance sample data, and calculates the adjacent day irradiance I according to the obtained daily irradiance and hourly irradiance sample data t and hourly irradiance r t The sample covariance and bandwidth of

[0063] The daily irradiance random sample generation module is used to select the Gaussian function as the kernel function, based on the adjacent daily irradiance I t The sample covariance and bandwidth coefficient of the daily irradiance are used to construct a conditional kernel density estimation model of the daily irradiance; by randomly generating a Gaussian function, the known I is calculated according to the conditional kernel density estimation model of the daily irradiance. t-1 Daily irradiance under conditions I t , generate random sample data of daily irradiance;

[0064] The hourly irradiance random sample generation module is used to select the Gaussian function as the kernel function and convert the daily irradiance I t Expressed as hourly irradiance r t The sum of the time and based on the daily irradiance I t and the hourly irradiance r of the previous day t-1A non-parametric solution set model of hourly irradiance is established, and the hourly irradiance r of the previous day is calculated based on the non-parametric solution set model of hourly irradiance by randomly selecting a multi-dimensional Gaussian function. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t , generate random sample data of hourly irradiance.

[0065] A computer device includes a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to perform the above method.

[0066] A computer-readable storage medium storing a computer program, wherein the computer program implements the above method when executed by a processor.

[0067] Compared with the prior art, the advantages of the present invention are: the present invention comprehensively considers the correlation between the hourly irradiance of the previous day and the hourly irradiance of the current simulated day, and can fully reflect the correlation structure of the hourly irradiance of adjacent days, so that the randomness of the hourly irradiance, the correlation between adjacent or non-adjacent hourly irradiances, and the additive characteristics of the total daily irradiance to the hourly irradiance solution set can be accurately taken into account, thereby realizing accurate simulation of the probability distribution of daily irradiance and hourly irradiance; and based on the measured data of irradiance, the time series probability modeling of hourly irradiance is performed based on non-parametric conditional kernel density estimation and solution set theory, without any assumptions about the probability distribution form and parameters. Not only is the modeling accuracy high and the versatility strong, but also good simulation effects can be achieved for photovoltaic power stations in different regions and with different climatic characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a schematic diagram of the implementation flow of the solar irradiance probability modeling method based on the non-parametric solution set in this embodiment.

[0069] Figure 2 It is a box plot of various statistical parameters of hourly irradiance and daily irradiance in a certain area obtained in a specific application embodiment of the present invention.

[0070] Figure 3 It is the probability density distribution curve of hourly irradiance and daily irradiance in a certain area obtained in a specific application embodiment of the present invention. DETAILED DESCRIPTION

[0071] The present invention will be further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the scope of protection of the present invention is not limited thereby.

[0072] The present invention utilizes the relationship between conditional kernel density estimation and multidimensional normal distribution to realize solar irradiance probability modeling in three stages: in the first stage, sample parameters are initialized, and daily irradiance data and hourly irradiance data of multiple days are obtained to form daily irradiance sample data and hourly irradiance sample data, and the irradiance of adjacent days I is calculated respectively. t and hourly irradiance r t The sample covariance and bandwidth coefficient of the daily irradiance are obtained. In the second stage, daily irradiance random samples are generated by constructing a conditional kernel density estimation model of daily irradiance. In the third stage, hourly irradiance time series probability model is constructed based on the non-parametric solution set to generate hourly irradiance random samples. The non-parametric conditional kernel density estimation method can be fully applied to comprehensively consider the correlation between the hourly irradiance of the previous day and the hourly irradiance of the current simulated day, and accurately simulate the additive characteristics of the irradiance at each moment and the total daily irradiance. The constructed hourly irradiance time series probability model can not only take into account the randomness, correlation and additive characteristics of the irradiance, but also reflect the correlation between the hourly irradiance of the previous day and the hourly irradiance of the current simulated day, thereby realizing the accurate simulation of the probability distribution of the irradiance of the photovoltaic power station at each moment, thereby improving the accuracy of the reliability assessment of the new power system containing photovoltaic power stations.

[0073] like Figure 1 As shown, the steps of the solar irradiance probability modeling method based on the non-parametric solution set in this embodiment include:

[0074] Step 1: Initialization of sample parameters: define the adjacent day irradiance I t =[I t ,I t-1 ] and define r t =[r 1,t ,r 2,t ,…,r d,t ] T , where I t ,I t-1 are the daily irradiance on day t and day t-1, r t represents the random vector composed of hourly irradiance on day t, where d is the number of hours per day; daily irradiance data of multiple days are obtained to form daily irradiance sample data and hourly irradiance sample data, and the irradiance of adjacent days I is calculated based on the obtained daily irradiance and hourly irradiance sample data. t and hourly irradiance r t The sample covariance and bandwidth coefficient of ;

[0075] Step 2: Daily irradiance sample generation: Select Gaussian function as kernel function, based on adjacent daily irradiance I tThe sample covariance and bandwidth coefficient of the daily irradiance are used to construct a conditional kernel density estimation model; by randomly generating a Gaussian function, the known I is calculated according to the conditional kernel density estimation model of the daily irradiance. t-1 Daily irradiance under conditions I t , generate random sample data of daily irradiance;

[0076] Step 3: Hourly irradiance sample generation: Select Gaussian function as kernel function and transform daily irradiance I t Expressed as hourly irradiance r t The sum of the time and based on the daily irradiance I t and the hourly irradiance r of the previous day t-1 A non-parametric solution set model of hourly irradiance is established, and the hourly irradiance r of the previous day is calculated based on the non-parametric solution set model of hourly irradiance by randomly selecting a multi-dimensional Gaussian function. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t , generate random sample data of hourly irradiance.

[0077] In this embodiment, first, I t represents the daily irradiance on day t, r t =[r 1,t ,r 2,t ,…,r d,t ] T represents the random vector of hourly irradiance of photovoltaic power source on day t, where d is the number of hours per day, and the unit is W / m 2 ; Input the measured data of solar irradiance for m years and n days as I ti (i=1,2,…,n), the measured data for each hour on the i-th day is r i =[r i1 ,r i2 ,…,r id ] T , obviously the subscript i represents the variable I t and r t The i-th sample data. Let I t =[I t ,I t-1 ] T , then I ti =[I ti ,I ti-1 ] T The sample data of adjacent day irradiance is calculated based on the obtained daily irradiance and hourly irradiance sample data in step 1. t and hourly irradiance r t The specific steps of calculating the sample covariance include:

[0078] Hourly irradiance r t Perform the following coordinate transformation:

[0079] Y t =Rr t (1)

[0080] Where Y t =(y 1,t ,y 2,t ,…,y d,t ) T ; R is the unit orthogonal matrix, that is, R T =R -1 ;definition U t =(y 1,t ,y 2,t ,…,y d-1,t ) T , then Y t =(y 1,t ,y 2,t ,…,y d,t ) T =(U t T ,I t ′) T , and let V t ′=(U t-1 ,I t ′) T ; Then the sample data U i and V i ′=(U i-1 ,I ti ′) T It can be obtained by the above steps.

[0081] Computation I t The sample covariance matrix S t :

[0082]

[0083] Among them, S h For I t The sample variance of S hq For I t and I t-1 The sample covariance of S q For I t-1 The sample variance of .

[0084] Calculation (U t ,V t The sample covariance matrix S between ′) is:

[0085]

[0086] Where S U For U t The (d-1)×(d-1)-order symmetric sample covariance matrix, S UV′ For U t and V t ′’s (d-1)×d-order sample covariance matrix, S V′ V t ′ is the d×d order sample covariance matrix.

[0087] Random variable I t , I t-1 、(U t ,V t ′) and V t ′’s bandwidth coefficient λ t ,λ t-1 ,λ UV′ and λ V′ The bandwidth coefficient can be calculated using an analytical method, such as the one described in “Multivariate online kernel density estimation with Gaussian kernels” (Pattern Recognition, Vol. 44, No. 10, 2011).

[0088] Because I t-1 The sample variance of is generally not zero, so we can use S in formula (2) t Perform the following elementary transformations:

[0089]

[0090] get:

[0091]

[0092] Likewise, since V t The sample covariance matrix of ′ is generally not zero, and the following elementary transformation can be performed on S in formula (3):

[0093]

[0094] Among them, E d-1 is a d-1 order unit matrix, E d It is a unit matrix of order d.

[0095] make To A t and A to perform Cholesky decomposition to obtain L t and L, where A t =L t L t T、A=LL T .

[0096] The process of constructing the conditional probability density estimation model of daily irradiance in this embodiment is as follows:

[0097] Definition based on the univariate kernel density estimation theory to estimate f(I t-1 ), the two-variable kernel density estimation theory estimates f(I t ,I t-1 ), select Gaussian function as kernel function, then f(I t-1 ) and f(I t ,I t-1 ) is expressed as:

[0098]

[0099]

[0100] According to the conditional probability theory and matrix operation rules, the conditional probability density estimation model of daily irradiance can be obtained as follows:

[0101]

[0102]

[0103]

[0104]

[0105]

[0106] Among them, λ t ,λ t-1 The daily irradiance I t , I t-1 The sample variance of , and satisfies:

[0107] λ t-1 =[(n-1)(4π) 1 / 2 R(f t-1 )] -1 / 5 ;

[0108] R(f t )=∫tr 2 {S t ·f″(I t )}dI t , R(f t-1 )=∫tr 2 {S q ·f″(I t-1 )}dI t-1 , S t For I t The sample covariance matrix, S h is the daily irradiance I t The sample variance of S hq is the daily irradiance I t and I t-1 The sample covariance of S q is the daily irradiance I t-1 The sample variance of , n is the total number of days of daily irradiance sample data.

[0109] Based on the above, the conditional probability density estimation model of the daily irradiance required in this embodiment is established.

[0110] From equations (8)-(14), we can see that the conditional density estimation is n-1 1-dimensional Gaussian functions (with mean B ti , the variance is ) weighted (weight is ω ti ) and weight ω ti Depends on I t-1 To historical sample value I ti-1 The distance between (sample data of daily irradiance on day t-1).

[0111] In step 2 of this embodiment, the known I is calculated based on the conditional kernel density estimation model of the daily irradiance. t-1 Daily irradiance under conditions I t , the steps of generating random sample data of daily irradiance include:

[0112] Step 2.1: Set t = 2 and set the initial value of daily irradiance I t-1 ;

[0113] Step 2.2: Calculate I for each sample ti-1 In the conditional quantity I t-1 The contribution weight ω under ti (i=1,2,…,n-1), which divides the [0,1] interval into n-1 intervals of length ω ti subintervals of ;

[0114] Step 2.3: Generate a random number x that follows a uniform distribution between [0,1]. If x falls in the kth subinterval of the interval [0,1], that is, Then the kth Gaussian function is selected for this sampling, and its mean is B tk , the variance is

[0115] Step 2.4: Randomly generate a Gaussian function V k , thus obtaining the known condition quantity I t-1The daily irradiance I on the tth day t =B tk +λ t L t V k ;

[0116] Step 2.5: Determine whether t is less than the preset threshold. If it is, set I t Assign I t-1 As the new conditional quantity, let t=t+1 and go to step 2.3), otherwise go to step 3.

[0117] For example, if the threshold is set to 365, if t≤365 days, I t Assign I t-1 As the new conditional quantity, let t = t + 1 and go to step 2.3, otherwise go to step 3 to execute stage three. In stage three, under the simulated daily irradiance conditions generated in stage two, a non-parametric solution set model of hourly irradiance is constructed to generate random samples of hourly irradiance.

[0118] The process of constructing the hourly irradiance non-parametric solution set model in this embodiment is as follows:

[0119] Considering the daily irradiance I t The summation characteristic between irradiance and hourly irradiance can be expressed as:

[0120] I t =r 1,t +r 2,t +…+r d,t (15)

[0121] Considering the hourly irradiance r t Not only with the solar irradiance I t There is an additive constraint, and it is also related to the hourly irradiance r of the previous day t-1 Related, in r t-1 and I generated by stage 2 sampling t Under known conditions, r t The conditional probability density function is:

[0122]

[0123] Based on formula (1), t After coordinate transformation, Equation (16) can be transformed into:

[0124]

[0125] Where: V t ′=(U t-1 ,I t ′) T ;f(U t ,Vt ′) is the 2d-1 dimensional joint probability density function; f(V t ′) is the d-dimensional marginal probability density function.

[0126] The multivariate kernel density estimation theory and bandwidth coefficient analytical method are used to estimate f(U t ,V t ′) and f(V t ′), we can get:

[0127]

[0128]

[0129] According to the block matrix inversion formula and matrix operation rules, we can deduce f(U t |V t The conditional kernel density estimation expression of ′) is:

[0130]

[0131] in:

[0132]

[0133]

[0134]

[0135]

[0136] Among them, U t =(y 1,t ,y 2,t ,…,y d-1,t ) T , Y t =(y 1,t ,y 2,t ,…,y d,t ) T , Y t =Rr t , R is the unit orthogonal matrix, d is the number of hours per day, V t ′=(U t-1 ,I t ′) T , λ V′ Indicates V t ′’s bandwidth coefficient, λ UV′ Indicates (U t ,V t Bandwidth coefficient of ′), S UV′ For U t and Vt ′ between the (d-1) × d order sample covariance matrix, S V′ V t ′’s d×d-order sample covariance matrix, i represents the serial number of the day, and n is the total number of days of daily irradiance sample data.

[0137] From formula (20), we can see that the conditional density estimation is n d-1 dimensional Gaussian functions (mean vector is B i , the covariance matrix is ) weighted (weight is ω i ) and the Gaussian function The contribution weight ω i Depends on V t ' to sample value V i ′The distance between them.

[0138] In step 3 of this embodiment, a multidimensional Gaussian function is randomly selected to calculate the hourly irradiance r of the previous day according to the hourly irradiance non-parametric solution model. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t ,The specific steps to generate hourly irradiance random sample data include:

[0139] Step 3.1: Let t = 2. Since the initial value of the solar irradiance is I t-1 It is known that the initial value of hourly irradiance is set to r t-1 ;

[0140] Step 3.2: In r t-1 , I t Under known conditions (t=1,2,…,365), then V t ′ is also a known quantity, according to the current daily irradiance I t , hourly irradiance r t-1 The values of sample data V are calculated respectively. i ′’s weight value ω i ,i=1,2,…,n,to divide the interval [0,1] into n intervals of length ω i subintervals of ;

[0141] Step 3.3: Generate a random number x that follows a uniform distribution between [0,1]. If x falls in the kth subinterval of the interval [0,1], that is, Then the kth Gaussian function is selected in the current sampling, where the mean vector is B k , the covariance matrix is

[0142] Step 3.4: Generate a (d-1)-dimensional Gaussian column vector Q and obtain a random vector U t =B k +λ UV′ LQ, and then get Y t =(U t T ,I t ′) T ;

[0143] Step 3.5: By linear transformation r t =R T Y t , we get the solution set of daily irradiance to hourly irradiance, that is, we get the random sample of hourly irradiance;

[0144] Step 3.6: Determine whether t reaches the preset threshold. If not, set t = t + 1 and go to step 3.3. Otherwise, the calculation is terminated and the final hourly irradiance random sample output is generated.

[0145] The present invention constructs a non-parametric solution set model of hourly irradiance by comprehensively considering the correlation between the hourly irradiance of the previous day and the hourly irradiance of the current simulated day. This model can fully reflect the correlation structure of the hourly irradiance of adjacent days, accurately take into account the randomness of hourly irradiance, the correlation between adjacent or non-adjacent hourly irradiances, and the additive characteristics of the total daily irradiance to the hourly irradiance solution set, thereby achieving accurate simulation of the probability distribution of daily irradiance and hourly irradiance. In addition, the time series probability modeling of hourly irradiance is carried out based on the non-parametric conditional kernel density estimation and solution set theory according to the measured data of irradiance. It does not require any assumptions about the probability distribution form and parameters. Not only is the modeling accuracy high and the versatility strong, but it can also achieve good simulation effects for photovoltaic power stations in different regions and with different climate characteristics.

[0146] In order to verify the effect of the present invention, the present invention is used to perform probabilistic modeling of the solar irradiance of a photovoltaic power station in a certain area. The detailed steps are as follows:

[0147] Phase 1: Initialization of sample parameters

[0148] 1.1) Let I t represents the daily irradiance on day t, r t =[r 1,t ,r 2,t ,…,r d,t ] T Represents the random vector of hourly irradiance of photovoltaic power source on day t, in W / m 2; d is the number of hours per day. Since the light is zero at night, according to the measured data of a certain area, only the situation from 9:00 to 16:00 when the irradiance is greater is studied, d=8; input the measured data of daily irradiance and hourly irradiance of a certain area for 3*365 days from 2020 to 2022. The measured data of daily irradiance is I ti (i=1,2,…,n), the measured data for each hour on the i-th day is r i =[r i1 ,r i2 ,…,r id ] T .

[0149] 1.2) Let I t =[I t ,I t-1 ] T , then I ti =[I ti ,I ti-1 ] T The sample data consists of irradiance of adjacent days.

[0150] 1.3) For r t Perform coordinate transformation: Y t =Rr t , Y t =(y 1,t ,y 2,t ,…,y d,t ) T , R is the unit orthogonal matrix, U t =(y 1,t ,y 2,t ,…,y d-1,t ) T , then Y t =(y 1,t ,y 2,t ,…,y d,t ) T =(U t T ,I t ′) T And let V t ′=(U t-1 ,I t ′) T Calculate the sample data U of the current region in the same way as steps 1.1) and 1.2) i and V i ′=(U i-1 ,I ti ′) T .

[0151] 1.4) Calculation I tThe sample covariance matrix of get Among them S h For I t The sample variance of S hq For I t and I t-1 The sample covariance of S q For I t-1 The sample variance of .

[0152] 1.5) Calculation (U t ,V t The sample covariance matrix between The calculation results are shown in Table 1.

[0153] Table 1: Sample covariance matrix calculation results

[0154]

[0155] 1.6) Calculate the random variable I t , I t-1 、(U t ,V t ′) and V t ′’s bandwidth coefficient λ t ,λ t-1 ,λ UV′ and λ V′ is: t =0.0932,λ t-1 =0.1201,λ UV′ = 0.2539 and λ V′ =0.1647.

[0156] 1.7) To A t =L t L t T and A=LL T Perform Cholesky decomposition to obtain L t and L, we get L t =1061.36, the results of matrix L are shown in Table 2.

[0157] Table 2: Calculation results of matrix L

[0158]

[0159]

[0160] Phase 2: Construct a conditional kernel density estimation model for daily irradiance and generate random samples of daily irradiance

[0161] 2.1) Obtain the conditional probability density estimation model of daily irradiance, namely:

[0162]

[0163] 2.2) Let t = 2, and set the initial value of daily irradiance to I t-1 =1790W / m 2 .

[0164] 2.3) Calculate each sample I using formula (12) ti-1 In the conditional quantity I t-1 The contribution weight ω under ti (i=1,2,…,n-1), which divides the [0,1] interval into n-1 intervals of length ω ti subinterval of .

[0165] 2.4) Generate a random number x=0.1576 from a uniform distribution between [0,1]. x falls in the 75th subinterval of the interval [0,1], i.e., 0.1531≤x≤0.1641. Then select the 75th Gaussian function for this sampling, and its mean is B tk =2675W / m 2 , the variance is

[0166] 2.5) Randomly generate a Gaussian function V k =3.0349, thus obtaining the known condition quantity I t-1 The solar irradiance on the second day is I t =B tk +λ t L t V k =2975. If t≤365 days, then I t Assign I t-1 As the new conditional quantity, let t = t + 1 and go to step (2.3), otherwise go to stage three.

[0167] Phase 3: Under the simulated daily irradiance conditions generated in Phase 2, a non-parametric solution set model of hourly irradiance is constructed to form a kernel density estimate of hourly irradiance conditions, and then a random sample of hourly irradiance is generated.

[0168] 3.1) Obtain the kernel density estimate of hourly irradiance conditions as:

[0169]

[0170] 3.2) Let t = 2. Since the initial value of the solar irradiance is I t-1 =1790W / m 2 Known, the initial value of hourly irradiance is r t-1=[135,204,263,235,287,267,289,106] T .

[0171] 3.3) In r t-1 , I t Under known conditions (t=1,2,…,365), then V t ′=[20,68,98,51,89,57,70,1052] T It is also a known quantity. Use formula (22) to calculate the sample V i ′’s contribution weight ω i (i=1,2,…,n), which divides the [0,1] interval into n intervals of length ω i subinterval of .

[0172] 3.4) Generate a random number x=0.9572 from a uniform distribution between [0,1]. x falls in the 69th subinterval of the interval [0,1], i.e., 0.9427≤x≤0.9699. Then select the 69th Gaussian function for this sampling, and its mean vector is B k =[-18,30,111,186,330,198,-103] T , covariance matrix As shown in Table 3.

[0173] Table 3: Calculation results of the covariance matrix

[0174]

[0175]

[0176] 3.5) Generate a (d-1)-dimensional Gaussian column vector Q = [-0.0631, 0.7147, -0.2050, -0.1242, 1.4897, 1.4090, 1.4172] T , thus we can get the random vector:

[0177] U t =B k +λ UV′ LQ=[-19.3251,39.4346,113.3665,184.6841,355.8726,232.8991,-74.8856] T ,but

[0178] Y t =(U t T ,I t ′) T=[-19.3251,39.4346,113.3665,184.6841,355.8726,232.8991,-74.8856,1052] T .

[0179] 3.6) Through linear transformation r t =R T Y t =[177,240,337,446,670,597,302,204] T , which means that the solution set of converting daily irradiance to hourly irradiance is realized. If t≤365 days, set t=t+1 and go to step (3.3); otherwise, the calculation terminates.

[0180] In this embodiment, the measured data of daily irradiance and hourly irradiance for a total of 3*365 days from 2020 to 2022 are input for a photovoltaic power station in a certain area. Since the light is zero at night, only the situation from 9:00 to 16:00, when the irradiance is relatively large, is studied based on the measured data of the region. The method of the present invention is used to establish a solar irradiance probability model and random sampling is performed; then, based on the measured data and the sampled data, a box plot is used to respectively display the test results of statistical parameters such as the mean, mean square error, coefficient of variation and first-order autocorrelation coefficient of the daily irradiance and hourly irradiance in the region, as shown in FIG. Figure 2 As shown; the probability density distribution curves of hourly irradiance and daily irradiance are drawn based on the measured data and sample data, as shown Figure 3 As shown. From the experimental results, we can get:

[0181] 1) The method of the present invention can accurately take into account the randomness of hourly irradiance, the correlation between adjacent or non-adjacent hourly irradiances, and the additive characteristics of the total daily irradiance to the hourly irradiance solution set, and can achieve accurate simulation of the probability distribution of daily irradiance and hourly irradiance;

[0182] 2) The method of the present invention can comprehensively consider the correlation between the hourly irradiance of the previous day and the hourly irradiance of the current simulated day, and can fully reflect the correlation structure of the hourly irradiance of adjacent days. The statistical parameters of the measured data all fall within the box range, with good fitting effect and high modeling accuracy.

[0183] 3) The method of the present invention is based on the measured data of irradiance, and uses non-parametric conditional kernel density estimation and solution set theory to perform time-series probability modeling of hourly irradiance. It does not require any assumptions about the probability distribution form and parameters. Therefore, the method is highly accurate and versatile, and can achieve good simulation effects for photovoltaic power stations located in different regions and with different climatic characteristics.

[0184] The solar irradiance probability modeling device based on the non-parametric solution set of this embodiment includes:

[0185] Sample parameter initialization module, used to define I t =[I t ,I t-1 ] and r t =[r 1,t ,r 2,t ,…,r d,t ] T , where I t ,I t-1 are the daily irradiance on day t and day t-1, r t represents a random vector composed of hourly irradiance on the tth day, obtains daily irradiance and hourly irradiance data of multiple days to form daily irradiance and hourly irradiance sample data, and calculates the adjacent day irradiance I according to the obtained daily irradiance and hourly irradiance sample data t and hourly irradiance r t The sample covariance and bandwidth coefficient of ;

[0186] The daily irradiance random sample generation module is used to select the Gaussian function as the kernel function, based on the adjacent daily irradiance I t The sample covariance and bandwidth coefficient of the daily irradiance are used to construct a conditional kernel density estimation model of the daily irradiance; by randomly generating a Gaussian function, the known I is calculated according to the conditional kernel density estimation model of the daily irradiance. t-1 Daily irradiance under conditions I t , generate random sample data of daily irradiance;

[0187] The hourly irradiance random sample generation module is used to select the Gaussian function as the kernel function and convert the daily irradiance I t Expressed as hourly irradiance r t The sum of the time and based on the daily irradiance I t and the hourly irradiance r of the previous day t-1 A non-parametric solution set model of hourly irradiance is established, and the hourly irradiance r of the previous day is calculated based on the non-parametric solution set model of hourly irradiance by randomly selecting a multi-dimensional Gaussian function. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t , generate random sample data of hourly irradiance.

[0188] The solar irradiance probability modeling device based on the non-parametric solution set in this embodiment corresponds one-to-one to the above-mentioned solar irradiance probability modeling device based on the non-parametric solution set, and will not be described in detail here.

[0189] This embodiment further provides a computer device, including a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to perform the above method.

[0190] It is understandable that the above method of this embodiment can be executed by a single device, such as a computer or server, etc., and can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of a distributed scenario, one of the multiple devices can only execute one or more steps in the above method of this embodiment, and multiple devices interact to complete the above method. The processor can be implemented in the form of a general-purpose CPU, a microprocessor, an application-specific integrated circuit, or one or more integrated circuits, etc., for executing relevant programs to implement the above method of this embodiment. The memory can be implemented in the form of a read-only memory ROM, a random access memory RAM, a static storage device, and a dynamic storage device. The memory can store an operating system and other application programs. When the above method of this embodiment is implemented by software or firmware, the relevant program code is stored in the memory and called and executed by the processor.

[0191] This embodiment further provides a computer-readable storage medium storing a computer program, which implements the above method when executed by a processor.

[0192] Those skilled in the art will appreciate that the above-mentioned embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including the instruction device, which implements the function specified in the process. Figure 1 a process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0193] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed above with reference to the preferred embodiment, it is not intended to limit the present invention. Therefore, any simple modifications, equivalent variations, and modifications to the above embodiment that do not depart from the technical solution of the present invention and are based on the technical essence of the present invention shall fall within the scope of protection of the technical solution of the present invention.

Claims

1. A solar irradiance probability modeling method based on non-parametric solution set, characterized in that the steps include: Step 1: Initialization of sample parameters: define the adjacent day irradiance I t =[I t ,I t-1 ] and r t =[r 1,t ,r 2,t ,…,r d,t ] T , where I t ,I t-1 The daily irradiance data of the tth day and the t-1th day are obtained respectively. The daily irradiance data of multiple days are formed into daily irradiance sample data and hourly irradiance sample data. t represents the random vector composed of hourly irradiance on day t, d is the number of hours in a day, and the irradiance of adjacent days I is calculated based on the obtained daily irradiance and hourly irradiance sample data. t and hourly irradiance r t The sample covariance and bandwidth coefficient of ; Step 2: Randomly generate daily irradiance samples: select Gaussian function as kernel function, based on the adjacent daily irradiance I t The sample covariance and bandwidth coefficient of the daily irradiance are used to construct a conditional kernel density estimation model; by randomly generating a Gaussian function, the known I is calculated according to the conditional kernel density estimation model of the daily irradiance. t-1 Daily irradiance under conditions I t , generate random sample data of daily irradiance; Step 3: Randomly generate hourly irradiance samples: select Gaussian function as kernel function and transform daily irradiance I t Expressed as hourly irradiance r t The sum of the time and based on the daily irradiance I t and the hourly irradiance r of the previous day t-1 A non-parametric solution set model of hourly irradiance is established, and the hourly irradiance r of the previous day is calculated based on the non-parametric solution set model of hourly irradiance by randomly selecting a multi-dimensional Gaussian function. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t , generate hourly irradiance random sample data; The hourly irradiance non-parametric solution model established in step 3 is: in: Among them, U t =(y 1,t ,y 2,t ,…,y d-1,t ) T , Y t =(y 1,t ,y 2,t ,…,y d,t ) T , Y t =Rr t ,R is the unit orthogonal matrix, ,d is the number of hours per day, V t ′=(U t-1 ,I t ′) T , λ V′ Indicates V t ′’s bandwidth coefficient, λ UV′ Indicates (U t ,V t Bandwidth coefficient of ′), S UV′ For U t and V t ′ between the (d-1) × d order sample covariance matrix, S V′ V t ′’s d×d-order sample covariance matrix, i represents the serial number of the day, and n is the total number of days of daily irradiance sample data; In step 2, by calculating each sample I ti-1 In the conditional quantity I t-1 The contribution weight ω under ti (i=1,2,…,n-1), which divides the [0,1] interval into n-1 intervals of length ω ti Generate a random number x between [0,1] that obeys uniform distribution. If x falls in the kth subinterval of the interval [0,1], that is, Then the kth Gaussian function is selected for this sampling, and its mean is B tk , the variance is Randomly generate a Gaussian function V k , thus obtaining the known condition quantity I t-1 The daily irradiance I on the tth day t =B tk +λ t L t V k ; In step 3, according to the current daily irradiance I t , hourly irradiance r t-1 The values of sample data V are calculated respectively. i ′’s weight value ω i ,i=1,2,…,n,to divide the interval [0,1] into n intervals of length ω i A subinterval of [0,1] generates a random number x that obeys a uniform distribution. If x falls in the kth subinterval of the [0,1] interval, that is, Then the kth Gaussian function is selected in the current sampling, where the mean vector is B k , the covariance matrix is Generate a (d-1)-dimensional Gaussian column vector Q and get a random vector U t =B k +λ UV′ LQ, and then get Y t =(U t T ,I t ′) T ; Through the linear transformation r t =R T Y t , and obtain the solution set of daily irradiance to hourly irradiance, that is, obtain the random sample of hourly irradiance.

2. The solar irradiance probability modeling method based on non-parametric solution set according to claim 1 is characterized in that: In step 3, multiple Gaussian functions are randomly selected to calculate the hourly irradiance r of the previous day according to the hourly irradiance non-parametric solution model. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t , the steps to generate random sample data of hourly irradiance include: Step 3.1: Set the initial value of hourly irradiance to r t-1 ; Step 3.2: According to the current daily irradiance I t , hourly irradiance r t-1 The values of sample data V are calculated respectively. i ′’s weight value ω i ,i=1,2,…,n,to divide the interval [0,1] into n intervals of length ω i subintervals of ; Step 3.3: Generate a random number x that follows a uniform distribution between [0,1]. If x falls in the kth subinterval of the interval [0,1], that is, Then the kth Gaussian function is selected in the current sampling, where the mean vector is B k , the covariance matrix is Step 3.4: Generate a (d-1)-dimensional Gaussian column vector Q and obtain a random vector U t =B k +λ UV′ LQ, and then get Y t =(U t T ,I t ′) T ; Step 3.5: By linear transformation r t =R T Y t , we get the solution set of daily irradiance to hourly irradiance, that is, we get the random sample of hourly irradiance; Step 3.6: Determine whether t reaches the preset threshold. If not, set t=t+1 and go to step 3.3). Otherwise, the calculation is terminated and the final hourly irradiance random sample output is generated.

3. The solar irradiance probability modeling method based on non-parametric solution set according to claim 1 is characterized in that: The conditional probability density estimation model of daily irradiance established in step 2 is: in: Among them, λ t ,λ t-1 The daily irradiance I t , I t-1 The sample variance of λ t-1 =[(n-1)(4π) 1 / 2 R(f t-1 )] -1 / 5 , R(f t )=∫tr 2 {S t ·f″(I t )}dI t , R(f t-1 )=∫tr 2 {S q ·f″(I t-1 )}dI t-1 , S t is the irradiance of the adjacent day I t The sample covariance matrix, S h is the daily irradiance I t The sample variance of S hq is the daily irradiance I t and I t-1 The sample covariance of S q is the daily irradiance I t-1 The sample variance of , n is the total number of days of daily irradiance sample data.

4. The solar irradiance probability modeling method based on non-parametric solution set according to claim 3 is characterized in that: In step 2, the known I is calculated based on the conditional kernel density estimation model of the daily irradiance. t-1 Daily irradiance under conditions I t , the steps of generating random sample data of daily irradiance include: Step 2.1: Set t = 2 and set the initial value of daily irradiance I t-1 ; Step 2.2: Calculate I for each sample ti-1 In the conditional quantity I t-1 The contribution weight ω under ti (i=1,2,…,n-1), which divides the [0,1] interval into n-1 intervals of length ω ti subintervals of ; Step 2.3: Generate a random number x that follows a uniform distribution between [0,1]. If x falls in the kth subinterval of the interval [0,1], that is, Then the kth Gaussian function is selected for this sampling, and its mean is B tk , the variance is Step 2.4: Randomly generate a Gaussian function V k , thus obtaining the known condition quantity I t-1 The daily irradiance I on the tth day t =B tk +λ t L t V k ; Step 2.5: Determine whether t is less than the preset threshold. If it is, set I t Assign I t-1 As the new conditional quantity, let t=t+1 and go to step 2.3), otherwise go to step 3.

5. The solar irradiance probability modeling method based on non-parametric solution set according to any one of claims 1 to 4, characterized in that: In step 1, the adjacent day irradiance I is calculated based on the obtained daily irradiance and hourly irradiance sample data. t and hourly irradiance r t Steps to calculate the sample covariance matrix: Hourly irradiance r t Perform the following coordinate transformation: Y t =Rr t Where Y t =(y 1,t ,y 2,t ,…,y d,t ) T ; R is the unit orthogonal matrix, that is, R T =R -1 ;definition U t =(y 1,t ,y 2,t ,…,y d-1,t ) T , then Y t =(y 1,t ,y 2,t ,…,y d,t ) T =(U t T ,I t ′) T , and let V t ′=(U t-1 ,I t ′) T ; Calculate daily irradiance I t The sample covariance matrix S t : Among them, S h For I t The sample variance of S hq For I t and I t-1 The sample covariance of S q For I t-1 The sample variance of Calculation (U t ,V t The sample covariance matrix S between ′) is: Where S U For U t The (d-1)×(d-1)-order symmetric sample covariance matrix, S UV′ For U t and V t ′’s (d-1)×d-order sample covariance matrix, S V′ V t ′ is the d×d order sample covariance matrix.

6. The solar irradiance probability modeling method based on non-parametric solution set according to claim 5 is characterized in that: Step 1 also includes: Pair S t Perform the following elementary transformations but And make the following elementary transformations on S: Among them, E d-1 is a d-1 order unit matrix, E d Then it is a unit matrix of order d; make To A t and A to perform Cholesky decomposition to obtain L t and L, where A t =L t L t T 、A=LL T .

7. A solar irradiance probability modeling device based on non-parametric solution set, characterized in that: include: Sample parameter initialization module, used to define I t =[I t ,I t-1 ] and r t =[r 1,t ,r 2,t ,…,r d,t ] T , where I t ,I t-1 are the daily irradiance on day t and day t-1, r t represents a random vector composed of hourly irradiance on the tth day, obtains daily irradiance and hourly irradiance data of multiple days to form daily irradiance and hourly irradiance sample data, and calculates the adjacent day irradiance I according to the obtained daily irradiance and hourly irradiance sample data t and hourly irradiance r t The sample covariance and bandwidth coefficient of ; The daily irradiance random sample generation module is used to select the Gaussian function as the kernel function, based on the adjacent daily irradiance I t The sample covariance and bandwidth coefficient of the daily irradiance are used to construct a conditional kernel density estimation model of the daily irradiance; by randomly generating a Gaussian function, the known I is calculated according to the conditional kernel density estimation model of the daily irradiance. t-1 Daily irradiance under conditions I t , generate random sample data of daily irradiance; The hourly irradiance random sample generation module is used to select the Gaussian function as the kernel function and convert the daily irradiance I t Expressed as hourly irradiance r t The sum of the time and based on the daily irradiance I t and the hourly irradiance r of the previous day t-1 A non-parametric solution set model of hourly irradiance is established, and the hourly irradiance r of the previous day is calculated based on the non-parametric solution set model of hourly irradiance by randomly selecting a multi-dimensional Gaussian function. t-1 , daily irradiance I t Hourly irradiance r under the conditions of t , generate hourly irradiance random sample data; The established hourly irradiance non-parametric solution set model is: in: Among them, U t =(y 1,t ,y 2,t ,…,y d-1,t ) T , Y t =(y 1,t ,y 2,t ,…,y d,t ) T , Y t =Rr t ,R is the unit orthogonal matrix, ,d is the number of hours per day, V t ′=(U t-1 ,I t ′) T , λ V′ Indicates V t ′’s bandwidth coefficient, λ UV′ Indicates (U t ,V t Bandwidth coefficient of ′), S UV′ For U t and V t ′ between the (d-1) × d order sample covariance matrix, S V′ V t ′’s d×d-order sample covariance matrix, i represents the serial number of the day, and n is the total number of days of daily irradiance sample data; In the random sample generation module of daily irradiance, by calculating each sample I ti-1 In the conditional quantity I t-1 The contribution weight ω under ti (i=1,2,…,n-1), which divides the [0,1] interval into n-1 intervals of length ω ti Generate a random number x between [0,1] that obeys uniform distribution. If x falls in the kth subinterval of the interval [0,1], that is, Then the kth Gaussian function is selected for this sampling, and its mean is B tk , the variance is Randomly generate a Gaussian function V k , thus obtaining the known condition quantity I t-1 The daily irradiance I on the tth day t =B tk +λ t L t V k ; In the hourly irradiance random sample generation module, the current daily irradiance I t , hourly irradiance r t-1 The values of sample data V are calculated respectively. i ′’s weight value ω i ,i=1,2,…,n,to divide the interval [0,1] into n intervals of length ω i A subinterval of [0,1] generates a random number x that obeys a uniform distribution. If x falls in the kth subinterval of the [0,1] interval, that is, Then the kth Gaussian function is selected in the current sampling, where the mean vector is B k , the covariance matrix is Generate a (d-1)-dimensional Gaussian column vector Q and get a random vector U t =B k +λ UV′ LQ, and then get Y t =(U t T ,I t ′) T ; Through the linear transformation r t =R T Y t , and obtain the solution set of daily irradiance to hourly irradiance, that is, obtain the random sample of hourly irradiance.

8. A computer device comprising a processor and a memory, wherein the memory is used to store a computer program, wherein: The processor is configured to execute the computer program to perform the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images