Photovoltaic output schedulable space prediction method based on confidence coefficient

Abstract

The invention belongs to the technical field of photovoltaic power generation, and particularly relates to a photovoltaic output schedulable space prediction method based on confidence. The method comprises the following steps: acquiring historical illumination intensity data of n groups of similar days, constructing a likelihood function based on the historical illumination intensity data for normal distribution, logarithmic normal distribution and Weibull distribution probability distribution models of illumination intensity, and calculating illumination intensity probability distribution model parameters by using a maximum likelihood estimation method. By setting the total area, efficiency, confidence interval quantile and other parameters of a photovoltaic panel, the output upper limit value and the output lower limit value of a photovoltaic output schedulable space are obtained by using a photovoltaic output calculation formula. According to the method, on the basis of historical illumination intensity data of similar days, a photovoltaic output schedulable space can be predicted under different confidence levels by setting a confidence interval quantile.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic power generation, and particularly relates to a method for predicting the schedulable space of photovoltaic output based on confidence. Background Art

[0002] In recent years, solar energy has gradually become an indispensable part of the global energy structure. Especially in the power system, the application of solar energy is becoming increasingly widespread. As one of the main ways of solar energy utilization, photovoltaic power generation has become an effective way to solve the global energy crisis and environmental problems due to its advantages of environmental protection and low carbon emissions. At the same time, the low operation and maintenance costs of the photovoltaic system also bring considerable economic benefits to users. However, although significant progress has been made in converting solar energy into electrical energy in photovoltaic technology, it still faces a series of challenges such as large power fluctuations and significant impacts from the environment and climate change. After photovoltaic power generation is connected to the power grid, it may cause power grid fluctuations, increasing the operation difficulty and uncertainty of the power system. Therefore, when scheduling a power generation system containing photovoltaic power generation, it is necessary to reasonably consider the uncertainty of photovoltaic output. When existing technologies schedule a photovoltaic system, most of them set the predicted value of photovoltaic output as a point prediction value and a fixed interval, and fail to fully consider the schedulable space of the output of the photovoltaic system.

[0003] Literature 1, "Research on a combined prediction model of photovoltaic output based on similar day selection and PCA-LSTM" (Acta Energiae Solaris Sinica, Vol. 45, No. 7, pp. 454-456) uses an improved K-means clustering method and the DTW algorithm to generate a historical day sample set with high internal correlation and similar weather characteristics to the day to be predicted, and then combines an LSTM neural network to propose a photovoltaic power prediction model based on similar day selection.

[0004] Literature 2, "Probability prediction of photovoltaic output based on quantile interpolation and deep autoregressive network" (Automation of Electric Power Systems, Vol. 47, No. 9, pp. 80-82) proposes a probability model of photovoltaic output based on quantile linear interpolation, then proposes using the continuous ranked probability score as the loss function for training the prediction model, and finally uses a deep autoregressive recurrent neural network to model the time series of photovoltaic output, combined with the proposed probability model of photovoltaic output, to form a new method for predicting the probability of photovoltaic output.

[0005] Literature 3, "Research on a method for predicting photovoltaic power generation based on improved LSTM" (Acta Energiae Solaris Sinica, Vol. 45, No. 11, pp. 297-299) analyzes the meteorological characteristics with strong correlation with photovoltaic output, uses the t-distribution neighborhood embedding dimensionality reduction technology to reduce the selected feature data to two dimensions, then clusters the reduced data through density peak clustering, and trains a long short-term memory neural network prediction model to achieve the prediction of photovoltaic output.

[0006] Existing literature has proposed methods for predicting photovoltaic power output, but the schedulable space of photovoltaic power output at different confidence levels has not been given, which restricts the application efficiency of the prediction results in power grid scheduling. Summary of the Invention

[0007] The object of the present invention is to overcome the deficiencies of the prior art. The present invention provides a method for predicting the schedulable space of photovoltaic power output based on confidence. For the normal distribution, lognormal distribution, and Weibull distribution functions that fit the probability distribution characteristics of light intensity, the present invention constructs a likelihood function based on historical light intensity data, calculates the probability distribution parameters of each distribution function using the maximum likelihood estimation method, and constructs a weighted probability distribution model of light intensity based on the above three distributions. The weight coefficient values are determined based on the distribution characteristics of historical light intensity data. On this basis, the confidence interval quantile Z(C) is determined based on the confidence setting value C, and the upper and lower limit values of light intensity corresponding to the weighted probability distribution model of light intensity are determined. Finally, based on parameters such as the total area of photovoltaic panels, efficiency, and confidence interval quantiles, the upper and lower limit values of the output of the schedulable space of photovoltaic power output are obtained using the photovoltaic power output calculation formula. Dispatchers can set different confidence levels according to the grid conditions and obtain the corresponding schedulable space of photovoltaic power output, optimize the dispatching plan within a larger range of photovoltaic power output, and improve the operating efficiency of the power grid.

[0008] To achieve the above object of the present invention, the following technical solutions are adopted:

[0009] A method for predicting the schedulable space of photovoltaic power output based on confidence, comprising the following steps:

[0010] Obtain historical light intensity data of n sets of similar days; respectively for the normal distribution, lognormal distribution, and Weibull distribution probability distribution models of light intensity, construct a likelihood function based on historical light intensity data, and calculate the light intensity probability distribution model parameters using the maximum likelihood estimation method; construct a weighted probability distribution model of light intensity that combines the normal distribution, lognormal distribution, and Weibull distribution, and determine the weight coefficient values based on the distribution characteristics of historical light intensity data; determine the confidence interval quantile Z(C) based on the confidence setting value C, and determine the upper limit value G max and lower limit value G min of light intensity corresponding to the weighted probability distribution model of light intensity; based on the total area A of photovoltaic panels, the efficiency η of photovoltaic panels, the upper limit value G max and lower limit value G min of light intensity corresponding to the weighted probability distribution model of light intensity, obtain the upper and lower limit values of the output of the schedulable space of photovoltaic power output using the photovoltaic power output calculation formula.

[0011] As a further preferred technical solution of the present invention, the construction of the likelihood function and the calculation of the parameters of the light intensity probability distribution model using the maximum likelihood estimation method include:

[0012] For the normal distribution probability distribution model of light intensity, taking the historical light intensity data of n groups of similar days as input, a likelihood function L of the normal distribution at time j is established 1 (μ 1 ,σ 1 |G 1,j ,G 2,j ,...,G n,j ), as shown in Equation (1):

[0013]

[0014] In the formula, G i,j is the light intensity at the j-th moment of the i-th group of similar days, μ 1 and σ 1 are the mean and standard deviation of the normal distribution to be solved, respectively;

[0015] Take the logarithm of the likelihood function L 1 (μ 1 ,σ 1 |G 1,j ,G 2,j ,...,G n,j ), as shown in Equation (2):

[0016]

[0017] Find the partial derivatives of the likelihood function L 1 (μ 1 ,σ 1 |G 1,j ,G 2,j ,...,G n,j ) with respect to the square of the mean and standard deviation of the normal distribution. The value of the partial derivative solution being zero indicates that the log-likelihood function reaches its maximum value, as shown in Equations (3) and (4):

[0018]

[0019] Based on Equations (3) and (4), solve for the mean μ 1 and standard deviation σ 1 of the normal distribution, as shown in Equations (5) and (6) respectively:

[0020]

[0021] Thus, construct the probability density function f 1 (G i,j |μ1 , σ 1 ):

[0022]

[0023] For the lognormal distribution probability distribution model of light intensity, taking the historical light intensity data of n groups of similar days as input, establish the likelihood function L of the lognormal distribution at time j 2 (μ 2 , σ 2 | G 1,j , G 2,j ,..., G n,j ), as shown in Equation (8):

[0024]

[0025] Where μ 2 and σ 2 are the mean and standard deviation of the lognormal distribution to be solved, respectively;

[0026] Take the logarithm of the likelihood function L 2 (μ 2 , σ 2 | G 1,j , G 2,j ,..., G n,j ), as shown in Equation (9):

[0027]

[0028] Find the partial derivatives of the likelihood function L 2 (μ 2 , σ 2 | G 1,j , G 2,j ,..., G n,j ) with respect to the squares of the mean and standard deviation of the lognormal distribution. The fact that the partial derivative solution value is zero indicates that the log-likelihood function reaches its maximum value, as shown in Equations (10) and (11):

[0029]

[0030] Based on Equations (10) and (11), solve for the mean μ 2 and standard deviation σ 2 of the lognormal distribution, as shown in Equations (12) and (13) respectively:

[0031]

[0032] Thus, construct the probability density function f 2 (G i,j | μ2 , σ 2 ):

[0033]

[0034] For the Weibull distribution probability distribution model of light intensity, taking the historical light intensity data of n groups of similar days as input, establish the likelihood function L of the Weibull distribution at time j 3 (λ, k|G 1,j , G 2,j ,..., G n,j ), as shown in Equation (15):

[0035]

[0036] Where λ and k are the shape parameter and scale parameter of the Weibull distribution to be solved respectively;

[0037] Take the logarithm of the likelihood function L 3 (λ, k|G 1,j , G 2,j ,..., G n,j ), as shown in Equation (16):

[0038]

[0039] Find the partial derivatives of the likelihood function L 3 (λ, k|G 1,j , G 2,j ,..., G n,j ) with respect to the shape parameter and scale parameter of the Weibull distribution. The solution value of the partial derivative being zero indicates that the log-likelihood function reaches its maximum value, as shown in Equations (17) and (18):

[0040]

[0041]

[0042] By using numerical solution software such as Python to solve Equations (17) and (18), obtain the shape parameter λ and scale parameter k of the Weibull distribution, thereby constructing the probability density function f 3 (G i,j |λ, k):

[0043]

[0044] Based on the shape parameter λ and scale parameter k of the Weibull distribution, calculate the mean E of the Weibull distribution and the standard deviation σ of the Weibull distribution respectively based on Equations (20) and (21) 3 :

[0045]

[0046] where Γ is the gamma function.

[0047] Further, as a preferred technical solution of the present invention, the construction of the weighted probability distribution model of light intensity combining normal distribution, lognormal distribution and Weibull distribution specifically includes:

[0048] Based on historical light intensity data, the model parameters of the normal distribution, lognormal distribution and Weibull distribution of light intensity are solved respectively. Weight coefficients a, b, and c are introduced to examine the importance of each distribution model in the weighted probability distribution model of light intensity, where the sum of a, b, and c is 1; construct the weighted probability distribution model f 4 (G i,j |μ 1 ,σ 1 ,μ 2 ,σ 2 ,λ,k), as shown in Equation (22):

[0049]

[0050] Further, as a preferred technical solution of the present invention, based on the distribution characteristics of historical light intensity data, the values of the weight coefficients are determined, specifically including:

[0051] If among the historical light intensity data of n groups of similar days, there are n 1 groups of data whose distributions are relatively close to the normal distribution, that is, satisfying the criterion shown in Equation (23):

[0052]

[0053] where p 1 is the standard score based on the normal distribution;

[0054] There are n 2 groups of data whose distributions are relatively close to the lognormal distribution, that is, satisfying the criterion shown in Equation (24):

[0055]

[0056] where p 2 is the standard score based on the lognormal distribution;

[0057] There are n 3 groups of data whose distributions are relatively close to the Weibull distribution, that is, satisfying the criterion shown in Equation (25):

[0058]

[0059] Based on n 1 、n2 and n 3 , determine the weight coefficients a, b, and c, as shown in Equation (26):

[0060]

[0061] Furthermore, as a preferred technical solution of the present invention, the confidence interval quantile Z(C) is determined based on the confidence level setting value C, and the upper limit value G of the light intensity corresponding to the light intensity weighted probability distribution model is calculated max and the lower limit value G min , as shown in Equations (27) and (28):

[0062]

[0063] Furthermore, as a preferred technical solution of the present invention, based on the upper limit value G of the light intensity and the lower limit value G of the light intensity corresponding to the light intensity weighted probability distribution model max and the lower limit value G min , the upper limit value P of the output power of the photovoltaic output dispatchable space is calculated using Equations (29) and (30) max and the lower limit value P min :

[0064] P max = G max ·A·η (29);

[0065] P min = G min ·A·η (30);

[0066] where A is the total area of the photovoltaic panels and η is the efficiency of the photovoltaic panels.

[0067] A method for predicting the dispatchable space of photovoltaic output based on confidence level of the present invention, compared with the prior art using the above technical solutions, has the following technical effects

[0068] (1) For the normal distribution, lognormal distribution, and Weibull distribution probability distribution models of light intensity respectively, the present invention constructs a likelihood function based on historical light intensity data, calculates the parameters of the light intensity probability distribution model using the maximum likelihood estimation method, constructs a light intensity weighted probability distribution model that combines the normal distribution, lognormal distribution, and Weibull distribution, and determines the value of the weight coefficient based on the distribution characteristics of historical light intensity data. Using three typical probability distribution models to fit the actual probability distribution model can improve its adaptability to historical light intensity data with different characteristics in different regions, making the light intensity weighted probability distribution model constructed based on the method of the present invention have high reliability and accuracy when predicting the photovoltaic output interval.

[0069] (2) By determining the dispatchable space of photovoltaic power output through setting the confidence level, the day-ahead dispatch plan can be made closer to the actual operating conditions of the photovoltaic power within the day, and a better dispatch plan can be sought within a larger dispatch range of photovoltaic power output. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a flowchart of the method according to an embodiment of the present invention;

[0071] Figure 2 It is a comparison graph of the normal distribution fitting probability density function curve and the lognormal distribution fitting probability density function curve of the light intensity data constructed by using the maximum likelihood estimation method according to an embodiment of the present invention;

[0072] Figure 3 It is a Weibull distribution fitting probability density function curve of the light intensity data constructed by using the maximum likelihood estimation method according to an embodiment of the present invention;

[0073] Figure 4 It is a probability density function curve of the weighted probability distribution model of the light intensity constructed by using the maximum likelihood estimation method according to an embodiment of the present invention;

[0074] Figure 5 It is the corresponding relationship between the confidence level setting value C and the confidence interval quantile Z(C);

[0075] Figure 6 It is the dispatchable space of the photovoltaic power output according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0076] The following will describe the present invention in detail with reference to the accompanying drawings for further explanation, so that those skilled in the art can understand the present invention more deeply and be able to implement it. However, the following is only for explaining the present invention by referring to examples and does not limit the present invention.

[0077] As Figure 1 shown, a method for predicting the dispatchable space of photovoltaic power output based on the confidence level includes the following steps:

[0078] Obtain historical light intensity data of 50 similar days;

[0079] For the normal distribution probability distribution model of the light intensity, take the historical light intensity data of n similar days as input, and establish the likelihood function L of the normal distribution at time j 1 (μ 1 ,σ 1 |G 1,j ,G 2,j ,...,G n,j ), as shown in Equation (1):

[0080]

[0081] Take the logarithm of the likelihood function L 1 (μ 1 , σ 1 |G 1,j , G 2,j ,..., G n,j ) as shown in Equation (2):

[0082]

[0083] Find the likelihood function L 1 (μ 1 , σ 1 |G 1,j , G 2,j ,..., G n,j ) with respect to the partial derivatives of the square of the mean and standard deviation of the normal distribution. The value of the partial derivative solution being zero indicates that the log-likelihood function reaches its maximum value (as shown in Equations (3) and (4)). The formulas are as follows:

[0084]

[0085] Based on Equations (3) and (4), obtain the mean μ 1 and standard deviation σ 1 (as shown in Equations (5) and (6) respectively):

[0086]

[0087] Thus, construct the probability density function f 1 (G i,j |μ 1 , σ 1 ) of the normal distribution of the light intensity at the j-th moment of the scheduling day:

[0088]

[0089] For the log-normal distribution probability distribution model of the light intensity, use the historical light intensity data of n groups of similar days as input to establish the likelihood function L 2 (μ 2 , σ 2 |G 1,j , G 2,j ,..., G n,j ) at the j-th moment, as shown in Equation (8):

[0090]

[0091] Take the logarithm of the likelihood function L 2 (μ 2 , σ 2 |G 1,j , G 2,j,...,G n,j ) Take the logarithm as shown in Equation (9):

[0092]

[0093] Find the likelihood function L 2 (μ 2 , σ 2 |G 1,j , G 2,j ,..., G n,j ) The partial derivatives of the mean and the square of the standard deviation of the log-normal distribution are taken. The fact that the solution of the partial derivative is zero indicates that the log-likelihood function reaches its maximum value (as shown in Equations (10) and (11)). The formulas are as follows:

[0094]

[0095] Based on Equations (10) and (11), solve for the mean μ of the log-normal distribution 2 and the standard deviation σ 2 (as shown in Equations (12) and (13) respectively):

[0096]

[0097] Thus, construct the probability density function f of the log-normal distribution of the light intensity at the j-th moment of the scheduling day 2 (G i,j |μ 2 , σ 2 ):

[0098]

[0099] Specifically in the embodiment of the present invention, 50 groups of historical light intensity data at 12:00 are selected for the probability density function curve fitting of the normal distribution and the log-normal distribution respectively. The comparison of the two curves is as Figure 2 shown.

[0100] For the Weibull distribution probability distribution model of the light intensity, take the historical light intensity data of n groups of similar days as the input, and establish the likelihood function L of the Weibull distribution at the j-th moment 3 (λ, k|G 1,j , G 2,j ,..., G n,j ), as shown in Equation (15):

[0101]

[0102] Take the partial derivative of the likelihood function L 3 (λ, k|G 1,j , G 2,j ,..., G n,j)Take the logarithm, as shown in Equation (16):

[0103]

[0104]

[0105] Obtain the likelihood function L 3 (λ, k|G 1,j , G 2,j ,..., G n,j )The partial derivatives of the Weibull distribution shape parameter and scale parameter, and the fact that the solution value of the partial derivative is zero indicates that the log-likelihood function reaches its maximum value (as shown in Equations (17) and (18)). The formula is as follows:

[0106]

[0107] By using numerical solution software such as Python to solve Equations (17) and (18), obtain the shape parameter λ and scale parameter k of the Weibull distribution, and thus construct the probability density function f 3 (G i,j |λ, k):

[0108]

[0109] Based on the shape parameter λ and scale parameter k of the Weibull distribution, calculate the mean E of the Weibull distribution and the standard deviation of the Weibull distribution respectively based on Equations (20) and (21):

[0110]

[0111] Specifically in the embodiment of the present invention, select 50 groups of historical light intensity data at 12:00 for curve fitting of the Weibull distribution probability density function, and the curve is as Figure 3 shown.

[0112] Construct a weighted probability distribution model of light intensity integrating the normal distribution, log-normal distribution and Weibull distribution, specifically including:

[0113] Based on the historical light intensity data, solve the model parameters of the normal distribution, log-normal distribution and Weibull distribution of the light intensity respectively, introduce weight coefficients a, b, c (the sum of a, b, c is 1) to examine the importance of each distribution model in the weighted probability distribution model of light intensity, and construct the weighted probability distribution model f 4 (G i,j |μ 1 , σ 1 , μ 2 , σ 2 , λ, k), as shown in Equation (22):

[0114]

[0115] Determine the value of the weight coefficient based on the distribution characteristics of historical light intensity data, specifically including:

[0116] If among the historical light intensity data of n groups of similar days, n 1 groups of data are more closely distributed to a normal distribution, that is, satisfy the criterion shown in Equation (23):

[0117]

[0118] where p 1 is the standard score based on the normal distribution.

[0119] There are n 2 groups of data that are more closely distributed to a lognormal distribution, that is, satisfy the criterion shown in Equation (24):

[0120]

[0121] where p 2 is the standard score based on the lognormal distribution.

[0122] There are n 3 groups of data that are more closely distributed to a Weibull distribution, that is, satisfy the criterion shown in Equation (25):

[0123]

[0124] Based on n 1 、n 2 、n 3 ,determine the weight coefficients a, b, c, as shown in Equation (26):

[0125]

[0126] Based on the historical light intensity data of 50 groups of similar days, calculate a as 0.342, b as 0.393, c as 0.265 through Equation (26) and obtain the weighted probability distribution model of light intensity:

[0127]

[0128] By querying Figure 5 the corresponding relationship between the confidence level setting value C and the confidence interval quantile Z(C) shown, determine Z(C). For example, when the confidence level setting value C takes 95%, the result of Figure 2For the cell with a value of 0.975, the corresponding values in its row and column are 1.9 and 0.06 respectively. The sum of these two values gives the confidence interval quantile Z(C) as 1.96. Calculate the upper limit value G of the light intensity corresponding to the light intensity weighted probability distribution model max and the lower limit value G min , as shown in Equations (28) and (29):

[0129]

[0130] Specifically, in the embodiment of the present invention, the probability density function curve of the light intensity weighted probability distribution model constructed by the maximum likelihood estimation method is as Figure 5 shown.

[0131] Based on the upper limit value G max and the lower limit value G min of the light intensity corresponding to the light intensity weighted probability distribution model, use Equations (30) and (31) to calculate the upper limit value P max and the lower limit value P min of the dispatchable space of the photovoltaic output:

[0132] P max = G max ·A·η (30);

[0133] P min = G min ·A·η (31);

[0134] Through a solution software such as Python, calculate the dispatchable space of the photovoltaic output under the condition of a 95% confidence level (A takes 5000 square meters and η takes 20%), as Figure 6 shown. The abscissa is time (hours), and the ordinate is the photovoltaic output (unit: kilowatt). The broken line represents the day-ahead prediction value of the photovoltaic output. The short dashed line and the long dashed line respectively represent the upper limit value curve and the lower limit value curve of the dispatchable space of the photovoltaic output. In the dispatchable space of the photovoltaic output, the photovoltaic output can be set to any value within the range between the short dashed line and the long dashed line, rather than being limited to the day-ahead prediction value of the photovoltaic output corresponding to the broken line.

[0135] The above specific implementation solutions have further detailed the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above is only the specific implementation solutions of the present invention, and is not intended to limit the scope of the present invention. Any equivalent changes and modifications made by those skilled in the art without departing from the concept and principles of the present invention shall fall within the scope of protection of the present invention.

Claims

1. A confidence-based method for predicting the dispatchable space of photovoltaic output, characterized in that: The following steps are involved: Get historical light intensity data of n groups of similar days; For the normal distribution, lognormal distribution and Weibull distribution probability distribution models of light intensity, likelihood functions were constructed based on historical light intensity data, and the parameters of the light intensity probability distribution model were calculated using the maximum likelihood estimation method. A light intensity weighted probability distribution model combining normal distribution, lognormal distribution and Weibull distribution was constructed, and the weight coefficient value was determined based on the distribution characteristics of historical light intensity data. Determine the confidence interval quantile Z(C) based on the confidence setting value C, and determine the upper limit value G of the light intensity corresponding to the light intensity weighted probability distribution model max With the lower limit G min ; Based on the total area A of photovoltaic panels, the efficiency η of photovoltaic panels, and the upper limit value G of light intensity corresponding to the weighted probability distribution model of light intensity max With the lower limit G min , use the photovoltaic output calculation formula to obtain the upper and lower limits of the photovoltaic output dispatchable space.

2. The method for predicting the schedulable space of photovoltaic output based on confidence according to claim 1, characterized in that: The construction of likelihood function and the calculation of light intensity probability distribution model parameters using maximum likelihood estimation method include: For the normal distribution probability distribution model of light intensity, the historical light intensity data of n groups of similar days are taken as input to establish the likelihood function L1(μ1,σ1|G 1,j ,G 2,j ,...,G n,j ), as shown in formula (1): Where G i,j is the light intensity at the jth moment of the i-th group of similar days, μ1 and σ1 are the mean and standard deviation of the normal distribution to be solved respectively; For the likelihood function L1(μ1,σ1|G 1,j ,G 2,j ,...,G n,j ) takes the logarithm, as shown in formula (2): Obtain the likelihood function L1(μ1,σ1|G 1,j ,G 2,j ,...,G n,j ) is the partial derivative of the mean and the square of the standard deviation of the normal distribution. The partial derivative solution is zero, indicating that the log-likelihood function has reached a maximum value, as shown in equations (3) and (4): Based on equations (3) and (4), the mean μ1 and standard deviation σ1 of the normal distribution are solved as shown in equations (5) and (6), respectively: Thus, the probability density function f1(G i,j |μ1,σ1): For the log-normal distribution probability distribution model of light intensity, the historical light intensity data of n groups of similar days are taken as input, and the likelihood function L2(μ2,σ2|G 1,j ,G 2,j ,...,G n,j ), as shown in formula (8): Where μ2 and σ2 are the mean and standard deviation of the lognormal distribution to be solved respectively; For the likelihood function L2(μ2,σ2|G 1,j ,G 2,j ,...,G n,j ) takes the logarithm, as shown in formula (9): Obtain the likelihood function L2(μ2,σ2|G 1,j ,G 2,j ,...,G n,j ) is the partial derivative of the mean and the square of the standard deviation of the lognormal distribution. The partial derivative solution is zero, indicating that the log-likelihood function has reached a maximum value, as shown in equations (10) and (11): Based on equations (10) and (11), the mean μ2 and standard deviation σ2 of the lognormal distribution are solved as shown in equations (12) and (13), respectively: Thus, the probability density function f2(G i,j |μ2,σ2): For the Weibull distribution probability distribution model of light intensity, the historical light intensity data of n groups of similar days are taken as input, and the likelihood function L3(λ,k|G 1,j ,G 2,j ,...,G n,j ), as shown in formula (15): Where λ and k are the shape parameter and scale parameter of the Weibull distribution to be solved respectively; For the likelihood function L3(λ,k|G 1,j ,G 2,j ,...,G n,j ) takes the logarithm, as shown in formula (16): Obtain the likelihood function L3(λ,k|G 1,j ,G 2,j ,...,G n,j ) is the partial derivative of the shape parameter and scale parameter of the Weibull distribution. The partial derivative solution is zero, indicating that the log-likelihood function has reached a maximum value, as shown in equations (17) and (18): By using numerical solution software such as Python to solve equations (17) and (18), we can obtain the shape parameter λ and scale parameter k of the Weibull distribution, thereby constructing the probability density function f3(G i,j |λ,k): Based on the shape parameter λ and scale parameter k of the Weibull distribution, the mean E of the Weibull distribution and the standard deviation σ3 of the Weibull distribution are calculated based on equations (20) and (21) respectively: Where Γ is the gamma function.

3. The confidence-based photovoltaic output dispatchable spatial prediction method according to claim 2, characterized in that: The method of constructing a light intensity weighted probability distribution model integrating normal distribution, lognormal distribution and Weibull distribution specifically includes: solving the model parameters of normal distribution, lognormal distribution and Weibull distribution of light intensity respectively based on historical light intensity data, introducing weight coefficients a, b and c to examine the importance of each distribution model in the light intensity weighted probability distribution model, wherein the sum of a, b and c is 1; constructing a light intensity weighted probability distribution model f4(G i,j |μ1,σ1,μ2,σ2,λ,k), as shown in equation (22):

4. The confidence-based photovoltaic output dispatchable spatial prediction method according to claim 3 is characterized in that: Based on the distribution characteristics of historical light intensity data, the weight coefficient value is determined, including: If among n groups of historical light intensity data on similar days, n1 groups of data have a distribution close to a normal distribution, that is, they satisfy the criterion shown in formula (23): Where p1 is the standard score based on normal distribution; There are n2 groups of data whose distribution is close to the log-normal distribution, that is, they satisfy the criterion shown in formula (24): Where p2 is the standard score based on lognormal distribution; There are n3 groups of data whose distribution is close to the Weibull distribution, that is, they satisfy the criterion shown in formula (25): Based on n1, n2, and n3, the weight coefficients a, b, and c are determined as shown in formula (26):

5. The confidence-based photovoltaic output dispatchable spatial prediction method according to claim 4 is characterized in that: The confidence interval quantile Z(C) is determined based on the confidence setting value C, and the upper limit value G of the light intensity corresponding to the light intensity weighted probability distribution model is calculated. max With the lower limit G min , as shown in formula (27) and formula (28):

6. The confidence-based photovoltaic output dispatchable spatial prediction method according to claim 5, characterized in that: The upper limit value of light intensity G corresponding to the light intensity weighted probability distribution model max With the lower limit G min , using equations (29) and (30) to calculate the output upper limit P of the PV output dispatchable space max With the lower limit value P min : P max =G max ·A·h (29); P min =G min ·A·h (30); Where A is the total area of ​​the photovoltaic panel, and η is the efficiency of the photovoltaic panel.

Images