New energy output power random generation method based on hybrid distribution function and high-dimensional vine Copula function

Abstract

The invention discloses a new energy output power random generation method based on a hybrid distribution function and a high-dimensional vine Copula function, and the method comprises the steps: collecting wind and light natural resource data and engineering data, calculating the hour-scale wind and light output power, and building a wind and light output feature descriptive index system; defining a discrete-continuous hybrid marginal distribution function, and describing intermittent characteristics of hour-scale wind and light output power; constructing a high-dimensional mixed vine Copula function of a self-defined structure, and describing correlation characteristics between wind and light output power in each hour in a day; and generating a wind and light intra-day output power scene set through random simulation based on a discrete-continuous hybrid marginal distribution function and a high-dimensional vine Copula joint distribution function of a self-defined structure. The method can accurately simulate the hourly-scale wind and light output power value and maintain the intermittent and correlation characteristics, and has important significance for planning the operation of a novel power system taking new energy as a main body.

Description

Technical Field

[0001] The present invention belongs to the technical field of multi-energy complementary power generation scheduling, and in particular to a method for randomly generating new energy output power based on a hybrid distribution function and a high-dimensional vine Copula function. Background Art

[0002] At present, the installed capacity and grid-connected power of renewable energy continue to increase. Due to the inherent intermittent and random volatility characteristics of these energy sources, large-scale renewable energy access poses severe challenges to the security, stability and power quality of the power grid. Research on the characteristics of renewable energy generation and simulation modeling is of great value to improving the operational stability of new power systems.

[0003] Currently, methods for simulating renewable energy output typically include physical modeling, machine learning, and mathematical statistics. Physical modeling requires high-resolution spatiotemporal natural resource datasets, resulting in high computational complexity. Machine learning methods require a large number of data samples and have low generalization capabilities. Mathematical statistics methods offer simple structure and high reliability, enabling rapid modeling of renewable energy output. However, traditional mathematical statistics modeling methods, such as Monte Carlo and Copula functions, struggle to effectively characterize the intermittent nature of renewable energy output and its high-order intraday time-course correlations, ultimately reducing the accuracy of renewable energy output simulations. Summary of the Invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a method for randomly generating renewable energy output power based on a mixed distribution function and a high-dimensional vine copula function, which can accurately simulate the hourly wind and solar output power values ​​and retain their intermittent and correlation characteristics, and is of great significance for planning the operation of new power systems based on renewable energy.

[0005] Technical solution: The method of the present invention comprises the following steps:

[0006] S1. Based on hourly wind and solar natural resource data and engineering data of new energy bases, calculate hourly wind and solar output power samples of bases as observation samples and establish a descriptive index system for wind and solar output power characteristics;

[0007] S2. Construct a hybrid marginal distribution function based on the Dirac function and the Kernal function, divide the wind and solar power output observation samples of each hour of the day into discrete parts and continuous parts, and fit the output power distribution of the discrete part and the continuous part based on the Dirac function and the Kernal function in the hybrid marginal distribution function;

[0008] S3. Use a high-dimensional mixed vine Copula to connect the mixed marginal distribution function, select the corresponding vine structure, Copula function type and parameters, and use the goodness-of-fit evaluation index to evaluate the fitting effect of the distribution function, so as to construct a custom structure high-dimensional mixed vine Copula joint distribution function that describes the correlation characteristics between the output power of each hour within a day;

[0009] S4. Based on the constructed custom high-dimensional mixed vine copula joint distribution function, the wind and solar power output for each hour of the day is derived through inverse sampling, thereby randomly generating a set of wind and solar power daily output scenarios that retain intermittent and correlation characteristics as simulation samples;

[0010] S5. Based on the descriptive index system of wind and solar power output characteristics, the evaluation index values ​​of the observation samples and the simulation samples are calculated, and the evaluation index values ​​of the observation samples and the evaluation index values ​​of the simulation samples are compared to evaluate the description effect of the method on the randomness, intermittency and correlation characteristics of wind and solar power output on the hourly scale.

[0011] Furthermore, step S1 includes the following steps:

[0012] S11. Collect hourly wind and solar natural resource data and engineering data for the new energy base. The wind and solar natural resource data includes hourly wind speed, air density, altitude, solar radiation, and temperature. The engineering data includes the number of wind turbines, installed capacity, output coefficient, cut-in wind speed, rated wind speed, cut-out wind speed, hub area, hub height, wind speed conversion coefficient, number of photovoltaic panels, installed capacity, area, and heat loss efficiency coefficient.

[0013] S12. Calculate hourly wind and solar power output at the new energy base based on a single wind turbine and photovoltaic power output calculation model, wind and solar natural resource data, and engineering data;

[0014] S13. Establish a descriptive indicator system for the randomness, intermittency, and correlation characteristics of wind and solar power output at the hourly scale, including:

[0015] (1) The random characteristics of wind and solar power output at the hourly scale are measured using sample mean, standard deviation, coefficient of variation, kurtosis coefficient, and output volatility;

[0016] (2) Using output activity to measure the intermittent characteristics of wind and solar power output on an hourly scale;

[0017] (3) The Kendall correlation coefficient is used to measure the temporal correlation characteristics between wind and solar power output at the hourly scale.

[0018] Furthermore, the step S2 specifically includes:

[0019] S21. Arrange the hourly output of wind power and photovoltaic power into a set of samples with a length of N, a dimension of T=24, and a sample space of 24-dimensional random variable

[0020] S22. Considering the double boundary of wind and solar power output at the hourly scale, the Dirac function and Kernal function are introduced to represent the upper and lower boundary values ​​and continuous values ​​of wind and solar power output respectively, and a mixed marginal distribution function is constructed to describe the double boundary distribution characteristics of wind and solar power output variables at the hourly scale;

[0021] S23, using root mean square error RMSE and goodness of fit R 2 The index evaluates the fitting effect of the mixed marginal distribution function constructed in step S22 on the hourly double-bounded wind and solar power output power variable distribution containing a large number of zero values ​​and rated output values.

[0022] Furthermore, the step S22 specifically includes:

[0023] (1) Construct a mixed probability density function of the double-boundary wind and solar power output variable at the hourly scale, expressed as:

[0024]

[0025] in, is the probability density function of wind power output; is the probability density function of the photovoltaic output power; δ(x) is the Dirac function; h W (x),h PV (x) represents the Kernal function obtained by fitting the wind power or photovoltaic power output variable; x is the value of the random variable X, and the random variable X is the wind power or photovoltaic power output in a certain period t; is the intermittent coefficient;

[0026] (2) The intermittency coefficient is determined by the ratio of the number of days with zero wind power or photovoltaic power output and the number of days with rated power to the total number of days in a certain period; the expression is:

[0027]

[0028] (3) The final cumulative distribution function is derived from the constructed mixed probability density function, that is, the mixed marginal distribution function is:

[0029]

[0030] Among them, P{X≤x} is the definition of the cumulative distribution function, which is the probability that the random variable X is less than or equal to x; P{0<X≤x|x> 0} represents the probability that the random variable X is less than or equal to x when x>0.

[0031] Furthermore, the Dirac function δ(x) is used to represent the probability density of the discrete part, and the expression is:

[0032]

[0033]

[0034] The kernel function h(x) is used to represent the probability density of the continuous part, and its expression is:

[0035]

[0036] Where N is the number of wind / photovoltaic output power samples in a certain period t; x n represents the nth wind power / photovoltaic output power value; B is the kernel function parameter, B best is the optimal parameter; σ is the standard deviation of the random variable X.

[0037] Furthermore, the step S3 specifically includes:

[0038] (1) Considering the wind and solar power output variable of each hour in a day as a node, the Kendall correlation coefficient γ between all nodes in the first-level tree structure and the remaining nodes is calculated in sequence;

[0039] (2) Estimate the type and parameter value of the Copula function connecting the two nodes in the first-level tree structure, and use the AIC criterion to determine the optimal Copula function type from the two aspects of goodness of fit and complexity; AIC = 2k-2ln(L), where k and L represent the number of parameters of the currently selected Copula function and the value of the maximum likelihood function of the Copula function, respectively;

[0040] (3) Determine the parameters of the Copula function using the maximum likelihood estimation method;

[0041] (4) After determining the first-layer Copula function, sample the Copula functions of each edge in turn to obtain the corresponding node values ​​of the next layer of tree, and repeat the above steps (1) to (3) to determine the tree structure and Copula function type of the next layer until the tree structure and Copula function type of all layers are determined;

[0042] (5) Based on the difference between the Cramer-von Mises statistical empirical Copula function and the constructed Copula function, the vine Copula model with the best fitting effect is selected as: D n =∫|Cn (u)-C e (u)| 2 dC e (u), where D n C is an indicator to test the degree of fit between the empirical Copula function and the constructed Copula function. n (u) and C e (u) represent the empirical Copula distribution function and the constructed Copula distribution function respectively.

[0043] Furthermore, the step S4 specifically includes:

[0044] (1) Derivation of hourly wind and solar power output samples X1, X2, X3, ... X t The corresponding marginal distribution function, joint distribution function and joint conditional distribution function;

[0045] (2) Generate a t-dimensional random vector that obeys a uniform distribution based on the Monte Carlo method Iteratively solve the distribution function of wind and solar power output sequence in each period;

[0046]

[0047] Among them, x1,x2,x3,…x t Corresponding representations are X1, X2, X3, ... X t The value in F1 -1 (ω1) is the inverse distribution function of the generated random variable ω1 that obeys the uniform distribution, They are all conditional distribution functions in the high-dimensional mixed vine copula joint distribution function of the custom structure constructed;

[0048] (3) According to the formula in step (2), the inverse function is used to derive the output power distribution function of each period; then the output power distribution function F in the tth period is t (x t )for;

[0049]

[0050] in, for x1,x2,x3,...,x t The joint conditional distribution function F t|t-1,t-2,…,2,1 (x t |x t-1 ,x t-2 ,…,x2,x1).

[0051] (4) Repeat the sampling process of steps (2) to (3) in each layer of the subsequent vine copula structure, derive the distribution function of the nodes associated with each time period layer by layer according to the optimal vine copula structure, function type and parameters, and finally obtain the wind and solar power output distribution function of each hour containing the complete correlation relationship;

[0052] (5) The hourly wind and solar power output power distribution function obtained by random generation is substituted into the inverse function of the fitted hourly wind and solar power output cumulative distribution function in turn to obtain the hourly wind and solar power output power sample generated by random simulation, namely the simulation sample.

[0053] The system corresponding to the method includes:

[0054] The data and indicator system unit is used to calculate hourly wind and solar power output samples based on hourly wind and solar power natural resource data and engineering data of new energy bases as observation samples, and to establish a descriptive indicator system for wind and solar power output characteristics;

[0055] A mixed marginal distribution function construction unit is used to construct a mixed marginal distribution function based on the Dirac function and the Kernal function, divide the wind and solar power output observation samples of each hour of the day into discrete and continuous parts, and fit the output power distribution of the discrete and continuous parts based on the Dirac function and the Kernal function in the mixed marginal distribution function;

[0056] The joint distribution function construction unit is used to connect the mixed marginal distribution functions using a high-dimensional mixed vine copula, optimize the corresponding vine structure, copula function type and parameters, and use the goodness-of-fit evaluation index to evaluate the fitting effect of the distribution function, thereby constructing a custom high-dimensional mixed vine copula joint distribution function that describes the correlation characteristics between the output power of each hour within a day;

[0057] The wind and solar power output simulation unit is used to derive the wind and solar power output for each hour of the day through inverse sampling based on a custom-structured high-dimensional mixed vine copula joint distribution function, thereby randomly generating a set of wind and solar power daily output scenarios that retain intermittent and correlation characteristics as simulation samples;

[0058] An evaluation unit is used to calculate the evaluation index values ​​of the observation samples and the simulation samples based on the descriptive index system of wind and solar power output characteristics, and compare the evaluation index values ​​of the observation samples with the evaluation index values ​​of the simulation samples to evaluate the description effect of the method on the randomness, intermittency and correlation characteristics of wind and solar power output on the hourly scale.

[0059] The electronic device described in the present invention includes a memory, a processor, and a computer program / instruction stored in the memory and capable of running on the processor. When the computer program / instruction is executed by the processor, the steps of the method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional Copula function are implemented.

[0060] The computer-readable storage medium of the present invention stores computer instructions, which, when called, are used to execute the steps of the method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional Copula function.

[0061] Beneficial effects: Compared with the existing technology, the significant technical effects of the present invention are: considering the inherent intermittent and high-order time-correlation characteristics of wind and solar energy, introducing a "discrete-continuous" mixed marginal distribution function to solve the double-boundary distribution and intermittent problems of wind and solar output power; constructing a high-dimensional vine Copula model, and combining it with the "discrete-continuous" mixed marginal distribution function to characterize the complex correlation structure and high-order time-correlation of wind power and photovoltaic power output, accurately simulating the hourly output power value of wind and solar while retaining its output characteristics, thereby improving the operating efficiency of the new power system with new energy as the main body. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a flow chart of the method of the present invention;

[0063] Figure 2 This is a comparison diagram of the improvement effect of the hybrid marginal distribution function on the fitting of the hourly wind and solar power output power distribution characteristics, where (a) is a comparison diagram of the improvement effect of the wind power output power distribution characteristics fitting, and (b) is a comparison diagram of the improvement effect of the photovoltaic power output power distribution characteristics fitting;

[0064] Figure 3 This is a schematic diagram of the optimal 24-dimensional vine Copula structure;

[0065] Figure 4 It is a comparison diagram of randomly generated wind and solar power output simulation samples and observation samples, among which (a) is the comparison diagram of wind power output simulation samples and observation samples, and (b) is the comparison diagram of photovoltaic power output simulation samples and observation samples. DETAILED DESCRIPTION

[0066] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. However, it will be apparent to those skilled in the art that the present invention can be practiced without one or more of these details. In other examples, some technical features known in the art are not described to avoid confusion with the present invention. The order of the relevant steps in the present invention is not limiting, that is, those skilled in the art can adjust it. The order in the present invention is a case-based writing method, not a limiting description.

[0067] The present invention discloses a random simulation method for renewable energy output using a hybrid distribution function and a high-dimensional vine copula function. This method constructs a model that couples a "discrete-continuous" hybrid marginal distribution function with a high-dimensional vine copula joint distribution function. This method can accurately simulate hourly wind and solar power output values ​​while retaining their randomness, intermittency, and correlation characteristics, thereby improving the operating efficiency of new power systems based on renewable energy. Compared with traditional single marginal distribution functions and low-dimensional copula functions, this method can improve the problems of inaccurate characterization of intermittent distribution characteristics and correlation structures, and accurately simulate hourly wind and solar power output values ​​while retaining their intermittency and correlation characteristics. This method is of great significance for planning the operation of new power systems based on renewable energy.

[0068] like Figure 1 As shown, the method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional vine copula function of the present invention includes the following steps:

[0069] S1. Collect hourly wind and solar natural resource data and engineering data of new energy bases, calculate hourly wind and solar output power samples (i.e., observation samples) of the bases, and establish a descriptive indicator system of wind and solar output power characteristics to characterize the randomness, intermittency, and correlation characteristics of wind and solar output.

[0070] According to one aspect of the present application, step S1 further comprises:

[0071] S11. Collect hourly wind and solar natural resource data and engineering data of the new energy base. The wind and solar natural resource data include hourly wind speed, air density, altitude, solar radiation and temperature. The engineering data include the number of wind turbines, installed capacity, output coefficient, cut-in wind speed, rated wind speed, cut-out wind speed, hub area, hub height, wind speed conversion coefficient, number of photovoltaic panels, installed capacity, area and heat loss efficiency coefficient.

[0072] In this embodiment, the above-mentioned wind and solar natural resource data can be obtained from public data, with the aim of gaining an in-depth understanding of the basic conditions of the study area and the wind and solar output power characteristics, and providing basic information and data support for subsequent model design and construction.

[0073] S12. Calculate the hourly wind and solar power output of the new energy base based on the single wind turbine and photovoltaic output power calculation model, wind and solar natural resource data, and engineering data.

[0074] According to one aspect of the present application, step S12 is further as follows:

[0075] S12a. Calculate the hourly wind power output value based on the output power calculation model of a single wind turbine, wind resource data, and engineering data of the wind turbine:

[0076]

[0077] Among them, P t w is the output of a single wind turbine during period t, kW; is the rated output of the wind turbine, kW; C p is the wind turbine output coefficient; S A is the area of ​​the wind turbine hub, m 3 ρ air is the air density, kg / m 3 ; V cutin 、V rated and V cutout are the cut-in, rated and cut-out wind speeds of the wind turbine, m / s, corresponding to the critical wind speeds at which the wind turbine starts generating electricity, operates at rated state and stops generating electricity; u t is the wind speed at the hub of the wind turbine, m / s;

[0078] Considering that the height of wind speed data does not match the height of wind turbine generator, we can build a wind speed conversion relationship to convert the data height wind speed u 1,t Approximately converted to the hub elevation u 2,t Wind speed:

[0079]

[0080] Among them, h1 and h2 are the hub heights of different wind turbines, m; u 1,t 、u 2,t h1 and h2 are the corresponding wind speeds, m / s; α is the wind speed conversion coefficient, which changes with the height of h1 and h2.

[0081] S12b. Calculate the hourly photovoltaic output power value based on the single photovoltaic output power calculation model, photovoltaic resource data, and photovoltaic panel engineering data:

[0082]

[0083] Among them, Pt s is the output of the photovoltaic power station during period t, kW; P stc Under standard conditions (corresponding to the solar radiation intensity G stc =1000W / m 2 , temperature T ref =25℃) Photovoltaic panel output, kW; G tot,t is the actual solar radiation intensity of the photovoltaic power station during period t, W / m 2 β is the heat loss efficiency coefficient, % / ℃ (for single crystal silicon photovoltaic cells, it is generally 0.45% / ℃); T t is the temperature during time period t, ℃; A PV is the area of ​​photovoltaic panels in the photovoltaic power station, m 2 .

[0084] S12c. Based on the output power calculation model of a single wind power station and photovoltaic power station in steps S12a and S12b, multiply it by the number of wind power and photovoltaic power stations in the new energy base to obtain the hourly wind and solar power output power sample (i.e., observation sample) of the base.

[0085] In this embodiment, since wind energy and solar energy are significantly affected by external environmental conditions such as meteorological and topographical factors and react quickly, have a high degree of uncertainty, significant time variations, and intermittent output, it is necessary to establish a comprehensive feature description index system to accurately describe the fluctuations and related characteristics of the output power of wind and solar power stations in adjacent time periods.

[0086] S13. Establish a descriptive indicator system for the randomness, intermittency and correlation characteristics of wind and solar power output at the hourly scale.

[0087] According to one aspect of the present application, step S13 is further as follows:

[0088] S13a. The random characteristics of wind and solar power output at the hourly scale are measured using the sample mean, standard deviation, coefficient of variation, kurtosis coefficient, and output fluctuation rate. The output fluctuation rate is calculated as follows:

[0089]

[0090] Among them, ω t is the wind power or photovoltaic output fluctuation rate in the current period t; P t is the average output power of wind power or photovoltaic power in the current period t, kW; P t+1 is the average output power of wind power or photovoltaic power in the t+1 period, kW; P max is the installed capacity of wind power or photovoltaic power, kW.

[0091] S13b. Use output activity to measure the intermittent characteristics of wind and solar power output on an hourly basis:

[0092]

[0093] Among them, A is the output activity; D n Indicates the number of days in a period when the output is not 0; TD indicates the total number of days in the period; x n Represents the nth wind power / photovoltaic output power value, n = 1, 2, 3, ..., N; N is the total number of hourly wind power / photovoltaic output power samples.

[0094] S13c, using the Kendall correlation coefficient to measure the temporal correlation characteristics between wind and solar power output at the hourly scale:

[0095]

[0096] Where γ is the Kendall correlation coefficient; N c 、N d They represent the number of pairs with the same and different serial numbers after the two groups of samples are sorted by numerical size.

[0097] In this embodiment, the hourly scale wind and solar power output presents a significant double-boundary intermittent characteristic. In a wind power generation system, when the wind speed is lower than the wind turbine's cut-in wind speed threshold, the wind power output is zero; when the wind speed exceeds the cut-out wind speed threshold, the wind power output reaches the rated power. In a photovoltaic power generation system, due to the influence of the diurnal cycle, power output is generated during the day, while the output is zero at night. In addition, the intermittent nature of photovoltaic power output varies with the seasons, and the power output sequence is discontinuous. By constructing a "discrete-continuous" mixed marginal distribution function to describe the double-boundary distribution characteristics of hourly scale wind and solar power output samples containing a large number of zero values ​​and rated output values, compared to the unified modeling marginal distribution function, the present invention represents the discrete part of the distribution in the wind and solar power output with a Dirac function and the continuous part with a Kernal function, thereby solving the problem of inaccurate distribution characteristics characterization of a single marginal distribution function.

[0098] S2. Based on the Dirac function and Kernal function, a "discrete-continuous" mixed marginal distribution function is constructed. The wind and solar power output observation samples of each hour in the day are divided into discrete part and continuous part. The output power distribution of the discrete part and the continuous part is fitted based on the Dirac function and Kernal function in the mixed marginal distribution function, thereby realizing the description of the intermittent characteristics of wind and solar power output on the hourly scale.

[0099] According to one aspect of the present application, step S2 further comprises:

[0100] S21. Arrange the output power of wind power and photovoltaic power in each hour of the day into a set of samples with a length of N (i.e., the total number of days), a dimension of T = 24 (i.e., 24 hours in a day), and a sample space of 24-dimensional random variable

[0101]

[0102] in, is the installed capacity of wind power (i.e. the maximum output of wind power), kW; is the photovoltaic installed capacity (i.e. the maximum photovoltaic output), kW; They represent the observation samples of wind power and photovoltaic power output at each hour of the day, They represent the wind power and photovoltaic power output at time t on the nth day, kW, n=1,2,…,N, t=1,2,…,T.

[0103] S22. Considering the double boundary of wind and solar power output at hourly scale, Dirac function and Kernal function are introduced to represent the upper and lower boundary values ​​and continuous values ​​of wind and solar power output respectively, and a "discrete-continuous" mixed marginal distribution function is constructed to describe the double boundary distribution characteristics of wind and solar power output variables at hourly scale. Figure 2 (a) and (b) show the improvement effect of the “discrete-continuous” mixed marginal distribution function constructed by this step on the fitting of the hourly wind and solar power output power distribution characteristics.

[0104] According to one aspect of the present application, step S22 is further as follows:

[0105] S22a. Construct a mixed probability density function that can characterize the hourly double-bounded wind / solar output power variable containing a large number of zero values ​​and rated output values. Let the wind / photovoltaic output power of a certain period t (t = 1, 2, ..., 24) be recorded as a random variable X, and its probability density function can be written as:

[0106]

[0107] in, is the probability density function of wind power output; is the probability density function of the photovoltaic output power; δ(x) is the Dirac function, see step S22b; h W (x),h PV (x) represents the Kernal function obtained by fitting the wind power / photovoltaic power output variable, and the calculation formula is shown in step S22c; x is the value of the random variable X (i.e., the wind power / photovoltaic power output in a certain period t); is the intermittent coefficient, see step S22d.

[0108] S22b, step S22a The Dirac function δ(x) in equations (8) and (9) is used to represent the probability density of the discrete part:

[0109]

[0110] S22c, step S22a Kernal function h(x) (i.e. h W (x),h PV (x)) is used to represent the probability density of the continuous part:

[0111]

[0112] Where N is the number of wind / photovoltaic output power samples in a certain period t (i.e. the total number of days); x n represents the nth wind power / photovoltaic output power value; B is the kernel function parameter, bandwidth, and optimal parameter B best Calculated by empirical formula; σ is the standard deviation of the random variable X.

[0113] S22d, intermittent coefficient The ratio of the number of days with zero wind / photovoltaic output power and the number of days with rated power to the total number of days in a certain period t is used to determine:

[0114]

[0115] in, They represent the proportion of days with zero wind power output and days with rated power output in the total number of days in period t respectively; It indicates the ratio of the number of days with zero photovoltaic power output and the number of days with rated power output to the total number of days in period t.

[0116] S22e, the mixed probability density function of the random variable X constructed in step S22a The cumulative distribution function of wind and solar power output at the hourly scale (i.e., the “discrete-continuous” mixed marginal distribution function) can be derived separately. The corresponding variable Distribution function:

[0117]

[0118] Among them, P W {X≤x},P PV {X≤x} are the cumulative distribution functions of wind power and photovoltaic power output respectively. The definition of derives the expression for the probability that the random variable X is less than or equal to x;

[0119] P W{0<X≤x|x> 0},P PV {0<X≤x|x> 0} represent the definition and derivation expressions of the cumulative distribution functions when the wind power and photovoltaic power output values ​​are greater than 0, respectively, and are the probabilities that the random variable X is less than or equal to x when x>0.

[0120] They are the mixed marginal distribution functions of wind power output and photovoltaic power output respectively.

[0121] S23, using root mean square error RMSE and goodness of fit R 2 The index evaluates the fitting effect of the "discrete-continuous" mixed marginal distribution function constructed in step S22 on the hourly double-boundary wind and solar power output power variable distribution containing a large number of zero values ​​and rated output values.

[0122] In this embodiment, a high-dimensional mixed vine Copula joint distribution function with a custom structure is used to describe the correlation between the output powers of each hour in a day. Compared with the low-dimensional unstructured Copula group, the high-dimensional mixed vine Copula joint distribution function with a custom structure used in the present invention can efficiently describe the complex correlation and structure between high-dimensional variables.

[0123] S3. Use a high-dimensional mixed vine Copula with a customizable structure to connect the "discrete-continuous" mixed marginal distribution function (i.e., the cumulative distribution function of wind and solar power output at each hour of the day) constructed in step S2, optimize the corresponding vine structure, Copula function type and parameters, and use the goodness of fit evaluation index to evaluate the fitting effect of the distribution function, thereby constructing a custom structure high-dimensional mixed vine Copula joint distribution function that accurately describes the correlation characteristics between the output power at each hour of the day.

[0124] According to one aspect of the present application, step S3 is further:

[0125] S31. Derivation of the cumulative distribution function of the wind and solar power output for 24 hours in a day (i.e., a "discrete-continuous" mixed marginal distribution function) according to step S22e.

[0126] S32. Use a high-dimensional mixed vine Copula joint distribution function with a customizable structure to connect the cumulative distribution function of the wind and solar power output within 24 hours of the day, and select the corresponding optimal vine structure, Copula function type and parameters.

[0127] According to one aspect of the present application, step S32 is further as follows:

[0128] S32a, treating the wind and solar power output variable of each hour in the day as a node, and sequentially calculating the Kendall correlation coefficient γ between all nodes in the first-level tree structure and the remaining nodes (the calculation method is the same as step S13c);

[0129] S32b. Estimate the Copula function type and parameter value connecting two nodes in the first-level tree structure, and use the AIC criterion to determine the optimal Copula function type from the two aspects of goodness of fit and complexity:

[0130] AIC=2k-2ln(L) (17)

[0131] Where k and L represent the number of parameters of the currently selected Copula function and the value of the maximum likelihood function of the Copula function, respectively;

[0132] S32c. Use the maximum likelihood estimation method to determine the parameters of the Copula function:

[0133]

[0134] Where L represents the likelihood function (i.e. the currently selected Copula function type); p(x i |θ) represents the probability of each variable under a given parameter θ;

[0135] S32d, after determining the first layer Copula function, sample the Copula function of each edge in turn to obtain the corresponding node value of the next layer of tree, repeat the above steps S32a to S32c to determine the tree structure and Copula function type of the next layer, until the tree structure and Copula function type of all layers are determined. The final preferred 24-dimensional vine Copula structure is as follows Figure 3 As shown;

[0136] S32e. Based on the difference between the Cramer-von Mises statistical empirical Copula function and the constructed Copula function, the vine Copula model with the best fitting effect is selected:

[0137] D n =∫|C n (u)-C e (u)| 2 dC e (u) (19)

[0138] Among them, D n C is an indicator to test the degree of fit between the empirical Copula function and the constructed Copula function. n (u) and C e (u) represent the empirical Copula distribution function and the constructed Copula distribution function respectively.

[0139] In this embodiment, based on the high-dimensional mixed vine Copula joint distribution function of the custom structure constructed in step S3, the wind and solar power output of each hour is continuously derived through inverse sampling, thereby randomly generating wind and solar power output scenarios within the day. The coupling method adopted in this application can accurately simulate the intermittent characteristics of wind and solar power output while retaining the complex correlation and structure between high-dimensional variables.

[0140] S4. Based on the custom high-dimensional mixed vine Copula joint distribution function constructed in step S3, the wind and solar power output for each hour of the day is derived through inverse sampling, thereby randomly generating a set of wind and solar power daily output scenarios that retain intermittent and correlation characteristics as simulation samples.

[0141] According to one aspect of the present application, step S4 is further:

[0142] S41. Assume that there are t-dimensional (t=1,2,…,T) random variables X1,X2,X3,…X t (i.e. corresponding to 24 hours of wind and solar power output samples), x1, x2, x3, ... x t Then the corresponding random variables X1, X2, X3, ... X t The value in F t (x t ) is x t The marginal distribution function, F 2|1 (x2|x1) is the joint conditional distribution function of x1 and x2, and the joint distribution function F of x1 and x2 is 12 (x1,x2) is obtained by taking the partial derivative of x1, which can also be simply expressed as H 1；1,2 (F1(x1),F2(x2)):

[0143]

[0144] Similarly, when it comes to 3D random variables, F 3|1,2 (x3|x1,x2) is the joint conditional distribution function of x1, x2, and x3, which is obtained by taking the partial derivative of the corresponding joint distribution function with respect to the known part, and can be simply expressed as H 1；1,3|2 (G 1|2 ,G 3|2 ):

[0145]

[0146] In formula (21), G 1|2 ,G 3|2 They are brief expressions of the following joint conditional distribution functions:

[0147] G 1|2 =F1|2 (x1|x2), G 3|2 =F 3|2 (x3|x2) (22)

[0148] Among them, F 1|2 (x1|x2) is the joint conditional distribution function of x1 and x2, F 3|2 (x3|x2) is the joint conditional distribution function of x2 and x3.

[0149] By analogy to the t-dimensional random variable, F t|t-1,t-2,...,2,1 (x t |x t-1 ,x t-2 ,…,x2,x1) is x1,x2,x3,...,x t The joint conditional distribution function can be expressed as:

[0150]

[0151] S42. Generate t-dimensional random vectors that obey uniform distribution based on Monte Carlo method Iteratively solve the distribution function of the wind and solar power output sequence in each period, let:

[0152]

[0153] in, is the inverse distribution function of the generated random variable ω1 that obeys the uniform distribution, These are all conditional distribution functions in the high-dimensional mixed vine copula joint distribution function of the custom structure constructed in step S32.

[0154] S43, according to formula (24) The output power distribution function F1(x1) of the first period is derived by inverting the function:

[0155] F1(x1)=ω1 (25)

[0156] S44. From the first-level structure of the vine copula determined in step S32, it can be seen that the output power distribution functions between two adjacent time periods are connected by a certain copula function, that is, the output power distribution function of the second time period can be derived from the output power distribution function of the first time period. From formula (24), Find the inverse function:

[0157] F2(x2)=F 2|1 (ω2|x1) (26)

[0158] The output power distribution function of the second period can be derived from the derivation principle of step S41:

[0159] F 2|1 (x2|x1)=H 1；1,2 (F1(x1),F2(x2))=ω2|(F1(x1)=ω1) (27)

[0160]

[0161] According to the definition in step S41, is the joint conditional distribution function F of x1 and x2 2|1 The inverse function of the concise expression (x2|x1).

[0162] S45. Similarly, the output power distribution function of the third period can be derived:

[0163]

[0164] According to the definition in step S41, is the joint conditional distribution function F of x1, x2, x3 3|2,1 The inverse function of the concise expression (x3|x2,x1).

[0165] S46. Similarly, the output power distribution function of the t-th period is obtained:

[0166]

[0167] According to the definition in step S41, for x1,x2,x3,...,x t The joint conditional distribution function F t|t-1,t-2,…,2,1 (x t |x t-1 ,x t-2 ,…,x2,x1).

[0168] S47. A new set of random numbers is obtained through the reverse sampling process of steps S42 to S46. These random numbers contain the dependency information between some variables. In each subsequent layer of the vine copula, the reverse sampling process of steps S42 to S46 is repeated, and the distribution function of the nodes associated with each time period is derived layer by layer based on the optimal vine copula structure, function type and parameters determined in step S32. Based on the variable correlation relationship established in the previous layer, the correlation relationship between the variables is further described, and finally the wind and solar power output distribution function for each hour containing the complete correlation relationship is obtained. and in, is the wind power output distribution function at hour t, is the wind power output variable at hour t, is the photovoltaic output power distribution function at hour t, is the photovoltaic output power variable at hour t.

[0169] S48: Substitute the hourly wind / solar output power distribution function randomly generated in step S47 into the inverse function of the hourly cumulative distribution function fitted in step S22e, and obtain the hourly wind / solar output power sample (i.e., simulated sample) generated by random simulation. and The final statistical distribution of simulation samples and observation samples is as follows Figure 4 As shown in (a) and (b).

[0170] S5. Based on the descriptive index system of wind and solar power output characteristics, the evaluation index values ​​of the observation samples and the simulation samples are calculated, and the evaluation index values ​​of the observation samples and the evaluation index values ​​of the simulation samples are compared to evaluate the description effect of the method on the randomness, intermittency and correlation characteristics of wind and solar power output on the hourly scale.

[0171] The characteristic description index system established in step S1 is used as the evaluation index to calculate the evaluation index values ​​of the hourly-scale wind and solar power output samples (i.e., observation samples) of the base. These values ​​are compared with the evaluation index values ​​of the hourly-scale wind and solar power output samples (i.e., simulation samples) obtained by random simulation of the model constructed by this method (i.e., the new energy output random simulation model based on the "discrete-continuous" mixed marginal distribution function and the high-dimensional vine Copula joint distribution function constructed in steps S2, S3, and S4) to evaluate the simulation effect of this method on the randomness, intermittency, and correlation characteristics of hourly-scale wind and solar power output.

[0172] In the study area, by applying the random generation method of renewable energy output power based on mixed distribution function and high-dimensional vine Copula function proposed in this invention, a set of wind and solar power output scenarios at the hourly scale can be quickly simulated and generated. While accurately simulating the hourly wind and solar power output values, their intermittent and correlation characteristics are retained, thereby improving the operating efficiency of the new power system with renewable energy as the main body.

[0173] In this embodiment, after all the above steps, hourly wind and solar power output can be obtained, providing effective and reliable reference information for the operation of a new power system based on renewable energy, which has high technical innovation and practicality. The above describes the preferred embodiments of the present invention in detail, but the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solution of the present invention, and these equivalent transformations all fall within the scope of protection of the present invention.

Claims

1. A method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional vine copula function, characterized in that: The following steps are involved: S1. Based on hourly wind and solar natural resource data and engineering data of new energy bases, calculate hourly wind and solar output power samples of bases as observation samples and establish a descriptive index system for wind and solar output power characteristics; S2. Construct a hybrid marginal distribution function based on the Dirac function and the Kernal function, divide the wind and solar power output observation samples of each hour of the day into discrete parts and continuous parts, and fit the output power distribution of the discrete part and the continuous part based on the Dirac function and the Kernal function in the hybrid marginal distribution function; S3. Use a high-dimensional mixed vine Copula to connect the mixed marginal distribution function, select the corresponding vine structure, Copula function type and parameters, and use the goodness-of-fit evaluation index to evaluate the fitting effect of the distribution function, so as to construct a custom structure high-dimensional mixed vine Copula joint distribution function that describes the correlation characteristics between the output power of each hour within a day; S4. Based on the constructed custom high-dimensional mixed vine copula joint distribution function, the wind and solar power output for each hour of the day is derived through inverse sampling, thereby randomly generating a set of wind and solar power daily output scenarios that retain intermittent and correlation characteristics as simulation samples; S5. Based on the descriptive index system of wind and solar power output characteristics, the evaluation index values ​​of the observation samples and the simulation samples are calculated, and the evaluation index values ​​of the observation samples and the evaluation index values ​​of the simulation samples are compared to evaluate the description effect of the method on the randomness, intermittency and correlation characteristics of wind and solar power output on the hourly scale.

2. The method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional Copula function according to claim 1 is characterized in that: Step S1 includes the following steps: S11. Collect hourly wind and solar natural resource data and engineering data for the new energy base. The wind and solar natural resource data includes hourly wind speed, air density, altitude, solar radiation, and temperature. The engineering data includes the number of wind turbines, installed capacity, output coefficient, cut-in wind speed, rated wind speed, cut-out wind speed, hub area, hub height, wind speed conversion coefficient, number of photovoltaic panels, installed capacity, area, and heat loss efficiency coefficient. S12. Calculate hourly wind and solar power output at the new energy base based on a single wind turbine and photovoltaic power output calculation model, wind and solar natural resource data, and engineering data; S13. Establish a descriptive indicator system for the randomness, intermittency, and correlation characteristics of wind and solar power output at the hourly scale, including: (1) The random characteristics of wind and solar power output at the hourly scale are measured using sample mean, standard deviation, coefficient of variation, kurtosis coefficient, and output volatility; (2) Using output activity to measure the intermittent characteristics of wind and solar power output on an hourly scale; (3) The Kendall correlation coefficient is used to measure the temporal correlation characteristics between wind and solar power output at the hourly scale.

3. The method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional Copula function according to claim 1 is characterized in that: The step S2 specifically includes: S21. Arrange the output power of wind power and photovoltaic power in each hour of the day into a set of samples with length N, dimension T, and sample space T-dimensional random variable is the installed capacity of wind power, i.e. the maximum output value of wind power; is the photovoltaic installed capacity, that is, the maximum photovoltaic output value; They represent the observation samples of wind power and photovoltaic power output at each hour of the day; S22. Considering the double boundary of wind and solar power output at the hourly scale, the Dirac function and Kernal function are introduced to represent the upper and lower boundary values ​​and continuous values ​​of wind and solar power output respectively, and a mixed marginal distribution function is constructed to describe the double boundary distribution characteristics of wind and solar power output variables at the hourly scale; S23, using root mean square error RMSE and goodness of fit R 2 The index evaluates the fitting effect of the mixed marginal distribution function constructed in step S22 on the hourly double-bounded wind and solar power output power variable distribution containing a large number of zero values ​​and rated output values.

4. The method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional vine copula function according to claim 3 is characterized in that: The step S22 specifically includes: (1) Construct a mixed probability density function of the double-boundary wind and solar power output variable at the hourly scale, expressed as: in, is the probability density function of wind power output; is the probability density function of the photovoltaic output power; δ(x) is the Dirac function; h W (x), h PV (x) represents the Kernal function obtained by fitting the wind power and photovoltaic power output variables respectively; x is the value of the random variable X, and the random variable X is the wind power or photovoltaic power output in a certain period t; is the intermittent coefficient; (2) The intermittency coefficient is determined by the ratio of the number of days with zero wind power or photovoltaic power output and the number of days with rated power to the total number of days in a certain period; the expression is: (3) The hybrid probability density function of wind power and photovoltaic power output is constructed The corresponding cumulative distribution function is deduced respectively, that is, the mixed marginal distribution function is Among them, P W {X≤x},P PV {X≤x} are the cumulative distribution functions of wind power and photovoltaic power output respectively. The definition of derives the expression for the probability that the random variable X is less than or equal to x; P W {0<X≤x|x> 0},P PV {0<X≤x|x> 0} represent the definition and derivation expressions of the cumulative distribution functions when the wind power and photovoltaic power output values ​​are greater than 0, respectively, and are the probabilities that the random variable X is less than or equal to x when x>0.

5. The method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional Copula function according to claim 4 is characterized in that: The Dirac function δ(x) is used to represent the probability density of the discrete part, and the expression is: The kernel function h(x) is used to represent the probability density of the continuous part, and its expression is: Where N is the number of wind power or photovoltaic power output samples in a certain period t; x n represents the nth wind power / photovoltaic output power value; B is the kernel function parameter, B best is the optimal parameter; σ is the standard deviation of the random variable X.

6. The method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional Copula function according to claim 1, characterized in that: The step S3 specifically includes: (1) Considering the wind and solar power output variable of each hour in a day as a node, the Kendall correlation coefficient γ between all nodes in the first-level tree structure and the remaining nodes is calculated in sequence; (2) Estimate the type and parameter value of the Copula function connecting the two nodes in the first-level tree structure, and use the AIC criterion to determine the optimal Copula function type from the two aspects of goodness of fit and complexity; AIC = 2k-2ln(L), where k and L represent the number of parameters of the currently selected Copula function and the value of the maximum likelihood function of the Copula function, respectively; (3) Determine the parameters of the Copula function using the maximum likelihood estimation method; (4) After determining the first-layer Copula function, sample the Copula functions of each edge in turn to obtain the corresponding node values ​​of the next layer of tree, and repeat the above steps (1) to (3) to determine the tree structure and Copula function type of the next layer until the tree structure and Copula function type of all layers are determined; (5) Based on the difference between the Cramer-von Mises statistical empirical Copula function and the constructed Copula function, the vine Copula model with the best fitting effect is selected as: D n =∫|C n (u)-C e (u)| 2 dC e (u), where D n C is an indicator to test the degree of fit between the empirical Copula function and the constructed Copula function. n (u) and C e (u) represent the empirical Copula distribution function and the constructed Copula distribution function respectively.

7. The method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional Copula function according to claim 1, characterized in that: The step S4 specifically includes: (1) Derivation of hourly wind and solar power output samples X1, X2, X3, ... X t The corresponding marginal distribution function, joint distribution function and joint conditional distribution function; (2) Generate a t-dimensional random vector that obeys a uniform distribution based on the Monte Carlo method Iteratively solve the distribution function of wind and solar power output sequence in each period; Among them, x1,x2,x3,…x t Corresponding representations are X1, X2, X3, ... X t The value in ; is the inverse distribution function of the generated random variable ω1 that obeys the uniform distribution, They are all conditional distribution functions in the high-dimensional mixed vine copula joint distribution function of the custom structure constructed; (3) According to the formula in step (2), the inverse function is used to derive the output power distribution function of each period; then the output power distribution function F in the tth period is t (x t )for; in, for x1,x2,x3,...,x t The joint conditional distribution function of The inverse function of the concise expression of . (4) Repeat the sampling process of steps (2) to (3) in each layer of the subsequent vine copula structure, derive the distribution function of the nodes associated with each time period layer by layer according to the optimal vine copula structure, function type and parameters, and finally obtain the wind and solar power output distribution function of each hour containing the complete correlation relationship; (5) The hourly wind and solar power output power distribution function obtained by random generation is substituted into the inverse function of the fitted hourly wind and solar power output cumulative distribution function in turn to obtain the hourly wind and solar power output power sample generated by random simulation, namely the simulation sample.

8. A new energy output power random generation system based on a mixed distribution function and a high-dimensional vine copula function, characterized in that: include: The data and indicator system unit is used to calculate hourly wind and solar power output samples based on hourly wind and solar power natural resource data and engineering data of new energy bases as observation samples, and to establish a descriptive indicator system for wind and solar power output characteristics; A mixed marginal distribution function construction unit is used to construct a mixed marginal distribution function based on the Dirac function and the Kernal function, divide the wind and solar power output observation samples of each hour of the day into discrete and continuous parts, and fit the output power distribution of the discrete and continuous parts based on the Dirac function and the Kernal function in the mixed marginal distribution function; The joint distribution function construction unit is used to connect the mixed marginal distribution functions using a high-dimensional mixed vine copula, optimize the corresponding vine structure, copula function type and parameters, and use the goodness-of-fit evaluation index to evaluate the fitting effect of the distribution function, thereby constructing a custom high-dimensional mixed vine copula joint distribution function that describes the correlation characteristics between the output power of each hour within a day; The wind and solar power output simulation unit is used to derive the wind and solar power output for each hour of the day through inverse sampling based on a custom-structured high-dimensional mixed vine copula joint distribution function, thereby randomly generating a set of wind and solar power daily output scenarios that retain intermittent and correlation characteristics as simulation samples; An evaluation unit is used to calculate the evaluation index values ​​of the observation samples and the simulation samples based on the descriptive index system of wind and solar power output characteristics, and compare the evaluation index values ​​of the observation samples with the evaluation index values ​​of the simulation samples to evaluate the description effect of the method on the randomness, intermittency and correlation characteristics of wind and solar power output on the hourly scale.

9. An electronic device, characterized in that: It includes a memory, a processor, and a computer program / instruction stored in the memory and executable on the processor. When the computer program / instruction is executed by the processor, the steps of the method for randomly generating the output power of a new energy source based on a mixed distribution function and a high-dimensional Copula function are implemented according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, which, when called, are used to execute the steps of the method for randomly generating new energy output power based on a mixed distribution function and a high-dimensional Copula function as described in any one of claims 1 to 7.

Images