New energy output probability density modeling method

Abstract

The invention discloses a new energy output probability density modeling method, which comprises the following steps of: collecting and preprocessing data to obtain a new energy power curve modeling data set; performing abnormal data cleaning on the data to obtain a new energy power curve modeling data set after partition processing; taking the new energy power curve modeling data set as input, and establishing an autocorrelation new energy power generation power condition kernel density estimation model; according to a new energy power generation power condition kernel density estimation model, determining an optimal bandwidth through a mean square integral error between a minimum estimation density and an unknown real density; and according to the optimal bandwidth, in combination with the new energy power generation power condition kernel density estimation model of the autocorrelation, calculating to obtain an output probability density model of the new energy. According to the method, the time autocorrelation of the new energy power generation data is fully considered, abnormal data can be eliminated in time, and new energy power generation characteristics under different environmental conditions can be adapted.

Description

Technical Field

[0001] This invention relates to the field of new energy power generation technology, specifically to a method for modeling the probability density of new energy output, which is particularly applicable to the analysis of power distribution characteristics and the construction of probability density curves of renewable energy power generation systems such as wind power and photovoltaic power generation. Background Technology

[0002] With the deepening of the global energy transition, renewable energy sources such as wind and solar power are increasingly accounting for a larger share of the energy mix. New energy power generation exhibits significant randomness and volatility, with its output power influenced by various environmental factors, resulting in a complex probability distribution. Accurately describing the probability density distribution of new energy power output is crucial for understanding its statistical characteristics, assessing system reliability, and developing reasonable operational strategies.

[0003] In probability density modeling of renewable energy output, kernel density estimation is widely used due to its flexibility and adaptability. This method considers the impact of environmental factors such as wind speed and solar radiation on power generation, constructing a conditional probability density function that can more accurately describe the power distribution characteristics under different environments. However, traditional methods assume that data samples are independent, ignoring the temporal autocorrelation of renewable energy output data. In reality, due to the influence of meteorological conditions and the dynamic characteristics of power generation equipment, renewable energy output data exhibits significant autocorrelation over time, such as the correlation between current wind power and power from previous moments. Ignoring autocorrelation can lead to biases in the probability density model, making it impossible to accurately characterize the power distribution. Furthermore, during data preprocessing, because renewable energy generation is affected by various environmental factors, the raw data contains outliers and noise, requiring correction and cleaning; bandwidth, as a key parameter in kernel density estimation, affects the smoothness and accuracy of the estimation and needs to be scientifically selected; and in terms of model evaluation, objective and effective evaluation indicators need to be established to comprehensively evaluate model performance.

[0004] To address the aforementioned issues, there is a need to develop a probability density modeling method for new energy output that can fully consider data autocorrelation. Therefore, this patent application is filed. Summary of the Invention

[0005] To address the above problems, the present invention aims to provide a method for modeling the probability density of new energy output, which is achieved using the following technical solution:

[0006] A method for modeling the probability density of new energy power output includes:

[0007] S1: Data collection and preprocessing to obtain a dataset for modeling new energy power curves;

[0008] S2: Divide the new energy power curve modeling dataset into N intervals according to the environmental driving factors at equal intervals, and use the quartile algorithm to clean up abnormal data in each interval to obtain the new energy power curve modeling dataset after interval processing.

[0009] S3: Using the new energy power curve modeling dataset after the interval processing as input, establish an autocorrelation-based new energy power generation conditional kernel density estimation model;

[0010] S4: Based on the kernel density estimation model for new energy power generation conditions, the optimal bandwidth is determined by minimizing the mean square integral error between the estimated density and the unknown true density.

[0011] S5: Based on the optimal bandwidth and combined with the autocorrelation-based new energy power generation condition kernel density estimation model, the output probability density model of new energy is calculated.

[0012] As a preferred design, in step S1, the collected data includes environmental driving factor data E that affects the power generation of new energy in the new energy equipment field. n And related parameters, preprocessed as data correction processing.

[0013] As a preferred design, the data correction process is performed according to a correction formula, which is:

[0014] ;

[0015] In the formula, and These are the standard air density and the actual air density, respectively. and These are the corrected power and the actual measured power, respectively. and These are the actual ambient temperature and the standard ambient temperature, respectively. and These are standard atmospheric pressure and actual atmospheric pressure, respectively.

[0016] As a preferred design, in step S2, if the amount of data in a certain interval is less than one-thousandth of the total amount of data, then the interval is considered invalid and subsequent steps are not executed.

[0017] The quartile algorithm includes:

[0018] Sort the power data within a specific range in ascending order: ;P k This represents the power generation data within the k-th interval;

[0019] Calculate the median M2, the first quartile M1, and the third quartile M3;

[0020] Based on interquartile range I QR Outliers within each interval are removed to obtain a dataset for modeling new energy power curves after interval processing.

[0021] As a preferred design, the process of establishing the autocorrelation-based kernel density estimation model for new energy power generation in S3 is as follows:

[0022] Based on the time dependence of new energy output data, the conditional mean is considered when considering time series data. for:

[0023] ;

[0024] In the formula, This represents data on factors affecting power generation. An unknown smoothing function representing the mean of the baseline conditions; An unknown smoothing function representing the fundamental conditional mean at time t-τ; An unknown smoothing function representing the conditional standard deviation; The unknown smoothing function represents the conditional standard deviation at time t-τ; n is a fixed constant. Represents the autoregressive parameters. This represents the renewable energy power generation data at input time t, using the obtained interval-processed renewable energy power curve modeling dataset as... Input data, This represents the power generation data of new energy sources at time t-τ.

[0025] For autoregressive parameters, Perform iterative updates to obtain the updated conditional mean considering the time series. for:

[0026] ;

[0027] In the formula, This represents a stationary process with a mean of 0 at time t-τ after iterative updates;

[0028] Conditional cumulative distribution function of new energy power generation data for:

[0029] ;

[0030] In the formula, G is the cumulative distribution function of d, and b is a variable. The smoothing function for the updated conditional standard deviation;

[0031] Using Gaussian curves as kernel functions :

[0032] ;

[0033] In the formula, d is an independent and identically distributed (i.e., i.i.d) random variable with zero mean and an unknown distribution. It is a residual sequence. For bandwidth;

[0034] Finally, the autocorrelation-dependent kernel density total function of new energy power generation is obtained. for:

[0035] ;

[0036] in, ;

[0037] In the formula, T represents the total time. For the total kernel function, It is a smoothing function for the updated conditional standard deviation at time T+1.

[0038] As a preferred design, ;

[0039] ;

[0040] In the formula, For a stationary process with a mean of 0, d t It is an independent and identically distributed (iid) random variable with a mean of zero and an unknown distribution, where n is a fixed constant.

[0041] As a preferred design, the autoregressive parameters, The method for iterative updates is as follows:

[0042] (1) Given The estimated values ​​are used to update the autoregressive parameters. :

[0043] ;

[0044] In the formula, This is a stationary process with a mean of 0 after iterative updates. It represents a stationary process with a mean of 0 at time tn after iterative updates;

[0045] (2) Using the updated autoregressive coefficients right The estimated values ​​have been updated:

[0046] ;

[0047] In the formula, The input data for new energy power generation at time t is updated iteratively. Let be the unknown smoothing function of the conditional standard deviation after iterative updates. These are the autoregressive parameters after iterative updates;

[0048] The two steps above are iterated until convergence.

[0049] As a preferred design, the convergence criterion during iteration is: for the residual sequence The iteration terminates when the autoregressive parameters converge and the Box-Ljung tests of the autocorrelation of all n-order lags are not rejected.

[0050] As a preferred design, in step S4, the mean square integral error value between minimizing the estimated density and the unknown true density includes the square deviation and the square deviation, i.e.: In the formula, The conditional kernel density total function of autocorrelation-dependent new energy power generation , For bandwidth, This represents the variance of the probability density function estimator. The square of the bias of the probability density function estimator. ; This refers to a higher-order term in the deviation expression. This is a higher-order variance term;

[0051] Using Gaussian curves as kernel functions To obtain the optimal bandwidth h best for:

[0052] ;

[0054] In the formula, The unknown smoothing function represents the conditional standard deviation, where n is a fixed constant.

[0055] As a preferred design, it also includes assessing the reliability and accuracy of the power output probability density model of the new energy source obtained by evaluating the average deviation and average width.

[0056] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0057] The probability density model of new energy power generation established by the method of this invention fully considers the time autocorrelation of new energy power generation data, can promptly eliminate abnormal data and adapt to the characteristics of new energy power generation under different environmental conditions, and provides more accurate technical support for power system dispatch and new energy grid-connected operation. Attached Figure Description

[0058] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0059] Figure 1 A flowchart illustrating the steps of a probability density modeling method for new energy output considering autocorrelation, provided as an embodiment of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0061] Example 1:

[0062] A probability density modeling method for new energy power output considering autocorrelation is proposed, comprising the following steps:

[0063] Step 1: Data collection and preprocessing to obtain a dataset for modeling new energy power curves.

[0064] The specific process is as follows:

[0065] Collect core environmental driving factors E that affect the power generation of new energy facilities, such as wind speed and solar radiation. n Data and actual power Ambient temperature T d Atmospheric pressure The original data was corrected using relevant data such as time t.

[0066] ;

[0067] In the formula, and These are the standard air density and the actual air density, respectively. and These are the corrected power and the actual measured power, respectively. and These are the actual ambient temperature and the standard ambient temperature, respectively. and These are standard atmospheric pressure and actual atmospheric pressure, respectively.

[0068] Through the above collection and correction preprocessing, a dataset for modeling new energy power curves is obtained.

[0069] The generation of abnormal data in the new energy system mainly stems from four aspects: equipment failures, environmental factors, operation and maintenance operations, and data processing issues. The interaction of these multi-level factors leads to various abnormalities in the power data. When affected by abnormal data, the characteristics of the power sequence (such as the change rate, mean, variance, and variance change rate) will mutate, seriously affecting the subsequent kernel density estimation and probability modeling accuracy. Therefore, effective anomaly detection and elimination must be carried out before data analysis.

[0070] Step 2: Use the quartile algorithm to clean the data for abnormal data and obtain the new energy power curve modeling data set after interval processing.

[0071] The specific steps are as follows:

[0072] According to the environmental driving factor E n Equally divide the above new energy power curve modeling data set into different N intervals at equal intervals, and select the central data of each interval as its reference data. Then the data set Can be expressed as:

[0073] ;

[0074] If the data volume in a certain interval is less than one-thousandth of the total data volume, then this interval is regarded as invalid and the subsequent steps are not executed.

[0075] Sort the power data in a specific interval (i.e., the data interval divided according to a certain requirement) in ascending order:

[0076] ;

[0077] In the formula, Represents the power generation data in the kth group interval.

[0078] Calculate three splitting points to divide the data into four parts:

[0079] The second splitting point (median) M2 is:

[0080] ;

[0081] When zn = 2i (i = 0, 1, 2,...), divide the power data samples in the interval into two parts from M2, and calculate the medians of these two parts in the above manner, denoted as M1 and M3 (M1 < M3). When zn = 2i + 1:

[0082] ;

[0083] In the formula, These are the weighting coefficients used in quartile calculation; based on linear interpolation, when zn=4i+1 Taking values ​​of 0.25 and 0.75 respectively, when zn = 4i + 3 Take values ​​of 0.75 and 0.25 respectively.

[0084] Based on interquartile range I QR Remove outliers from sample data:

[0085] ;

[0086] ;

[0087] That is, each interval arrive Data outside of these categories is identified as abnormal and deleted, ultimately resulting in a new energy power curve modeling dataset after interval processing.

[0088] Step 3: Using the new energy power curve modeling dataset obtained in Step 2 after interval processing as input, establish an autocorrelation conditional kernel density estimation method to obtain an autocorrelation conditional kernel density estimation model for new energy power generation.

[0089] The specific process is as follows:

[0090] Considering the time dependence of new energy output data, the model followed by the data is established as follows:

[0091] ;

[0092] In the formula, This indicates that the input is the renewable energy power generation data, specifically the renewable energy power curve modeling dataset obtained above after interval processing. Enter the input; This represents data on factors affecting power generation. An unknown smoothing function representing the mean of the baseline conditions; An unknown smoothing function representing the conditional standard deviation; It is a stationary process with a mean of 0, which can be represented in nth-order autoregressive form as follows:

[0093] ;

[0094] In the formula, It is an independent and identically distributed (iid) random variable with zero mean and an unknown distribution, where n is a fixed constant that depends on the autocorrelation strength of the data; Indicates the autoregressive parameters; This represents a stationary process with a mean of 0 at time t-τ.

[0095] At this point, the conditional mean related to the time series can be expressed as:

[0096] ;

[0097] In the formula, This represents data on factors affecting power generation. An unknown smoothing function representing the mean of the baseline conditions; An unknown smoothing function representing the fundamental conditional mean at time t-τ; An unknown smoothing function representing the conditional standard deviation; The unknown smoothing function represents the conditional standard deviation at time t-τ; n is a fixed constant. Represents the autoregressive parameters. This represents the power generation data of new energy sources at time t. This represents the power generation data of new energy sources at time t-τ.

[0098] To account for autocorrelation, an iterative approach is used to estimate... and autoregressive parameters The specific steps are as follows:

[0099] use The estimated values ​​are updated to the autoregressive parameters. :

[0100] ;

[0101] In the formula, This is a stationary process with a mean of 0 after iterative updates. This represents a stationary process with a mean of 0 at time tn after iterative updates. China and The symbol in "" is merely a simplified representation, representing a specific part within parentheses, and is not fixed, such as It can represent , can also mean other things.

[0102] Using the updated autoregressive coefficients right The estimated values ​​have been updated:

[0103] ;

[0104] In the formula, The input data for new energy power generation at time t is updated iteratively. Let be the unknown smoothing function of the conditional standard deviation after iterative updates. These are the autoregressive parameters after iterative updates.

[0105] The above two steps are iterated until convergence. Convergence is determined by: the residual sequence... The iteration terminates when the autoregressive parameters converge and the Box-Ljung tests of the autocorrelation of all n-order lags are not rejected.

[0106] At this point, the time-related conditional mean estimator, after iterative updates, can be expressed as:

[0107] ;

[0108] In the formula, An unknown smoothing function representing the mean of the underlying conditions.

[0109] Conditional cumulative distribution function of renewable energy power generation data at time t for:

[0110] ;

[0111] In the formula, G is the cumulative distribution function of d, and b is a variable.

[0112] Differentiating both sides of the equation for the conditional cumulative distribution function above, we obtain the conditional density function:

[0113] ;

[0114] In the formula, Let be the probability density function of d, where the symbol "•" represents the probability density function of d in the above equation. ; d is an independent and identically distributed (iid) random variable with zero mean and an unknown distribution.

[0115] Due to the assumption that the probability density functions are independent and identically distributed, we can use the residual sequence... The kernel density estimation method is used to express it as follows:

[0116] ;

[0117] In the formula, T represents the total time, and t∈T.

[0118] Considering the practicality of kernel functions in waveform synthesis calculations, a Gaussian curve is used as the kernel function:

[0119] ;

[0120] In the formula, r t It is a residual sequence. Where is the bandwidth, and n represents the number of samples.

[0121] Finally, the autocorrelation-based conditional kernel density total function of new energy power generation. for:

[0122] ;

[0123] in, .

[0124] In the formula, T represents the total time. For the total kernel function, It is a smoothing function for the updated conditional standard deviation at time T+1.

[0125] Finding the optimal bandwidth in kernel density estimation aims to achieve the best balance between bias and variance. Insufficient bandwidth leads to poor smoothing, generating excessive noise peaks and jagged density curves that fail to reflect the true probability distribution. Conversely, excessive bandwidth results in over-smoothing, losing important distribution details and local features, making the density estimate too flat. The optimal bandwidth maintains the smoothness of the density estimate while accurately capturing the true distribution characteristics of the data, thus minimizing the mean squared error and ensuring the accuracy and reliability of subsequent density-based probability predictions, risk assessments, and decision optimization.

[0126] Step 4: Find the optimal bandwidth h by minimizing the mean square integral error between the estimated density and the unknown true density. This error includes the squared deviation and the variance, i.e.:

[0127]

[0128] In the formula, The above probability density function ; .

[0129] This represents the variance of the probability density function estimator. The square of the bias of the probability density function estimator; This refers to a higher-order term in the deviation expression. This is a higher-order variance term.

[0130] When the sample size is sufficient, the asymptotic mean integral squared error It can be defined as:

[0131] ;

[0132] Given this objective function It exhibits convexity, and its minimum point corresponds to the zero of its first derivative. Therefore, the optimal bandwidth value can be obtained as:

[0133] ;

[0134] because Since the function is unknown, it can be estimated in various ways. Therefore, the Gaussian distribution used in step 3 is referenced. To approximate the data, at this point:

[0135] ;

[0136] Then the optimal bandwidth h best for:

[0137] ;

[0138] In the formula, The unknown smoothing function represents the conditional standard deviation, where n is a fixed constant.

[0139] The optimal bandwidth h obtained above best Substituting the formula into the kernel density total function of the new energy power generation condition obtained in step 3 Finally, the probability density model of new energy power output was obtained.

[0140] Step 5: Evaluate the reliability and sharpness of the new energy output probability density model obtained in Step 4 using the average deviation and average width, respectively.

[0141] ;

[0142] In the formula, M and N represent the number of quantile pairs and the sample size used in the evaluation, respectively; This is sample data for new energy power generation; These represent the upper quantile theoretical level and the lower quantile theoretical level at the j-th confidence level, respectively. These are the upper quantile theoretical level and the lower quantile theoretical level predicted for the i-th sample at the j-th confidence level, respectively. This represents the estimated upper score. R represents the estimated subdivision; R is the normalization factor, representing the range of changes in new energy power generation. This is an indicator function; its value is 1 when the condition inside the parentheses is true, and 0 otherwise.

[0143] The method of this invention should be able to effectively process abnormal data, scientifically select the optimal bandwidth, accurately capture the conditional probability distribution characteristics of new energy output, and provide a reasonable model evaluation mechanism to improve the accuracy and reliability of probability density estimation.

[0144] Those skilled in the art will understand that this application may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0145] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0146] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0147] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0148] Those skilled in the art will understand that all or part of the steps in the above facts and methods can be implemented by a program instructing related hardware. The program or the program described therein can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: at this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0149] The above specific embodiments further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A new energy output probability density modeling method, characterized in that, The method comprises the following steps: S1: data collection and preprocessing, obtaining a new energy power curve modeling data set; S2: dividing the new energy power curve modeling data set into N intervals according to the equal interval of environmental driving factors, and using the quartile algorithm to clean up the abnormal data in each interval to obtain a new energy power curve modeling data set after interval processing; S3: taking the new energy power curve modeling data set after interval processing as input, and establishing a new energy power generation power conditional kernel density estimation model with autocorrelation; S4: according to the new energy power generation power conditional kernel density estimation model, the optimal bandwidth is determined by minimizing the integral error of the mean square between the estimated density and the unknown true density; S5: according to the optimal bandwidth, combining the new energy power generation power conditional kernel density estimation model with autocorrelation, the output probability density model of new energy is calculated.

2. The method of claim 1, wherein, In the S1, the collected data includes environmental driving factor data E in the new energy equipment field that affects the new energy power generation power n and related parameters, and the preprocessing is a data correction processing.

3. The method of claim 2, wherein, The data correction processing is carried out according to the correction formula, and the correction formula is: ； wherein, and are the standard air density and the actual air density, respectively; and are the corrected power and the actual measured power, respectively; and are the actual ambient temperature and the standard ambient temperature, respectively; and are the standard atmospheric pressure and the actual atmospheric pressure, respectively.

4. The method of claim 1, wherein, In the S2, if the data quantity in a certain interval is less than one thousandth of the total data quantity, the interval is considered invalid, and the subsequent steps are not executed; The quartile algorithm comprises: The power data in the specific interval is arranged in ascending order: ; P k represents the power generation data in the kth group of intervals; Calculate the median M2 and the first quartile M1 and the third quartile M3; Based on quartile distance I QR After removing outliers in each interval, the new energy power curve modeling data set after interval processing is obtained.

5. The method of claim 1, wherein, The process of establishing a new energy power generation power conditional kernel density estimation model with autocorrelation in S3 is: Based on the dependence of new energy output data on time, consider the conditional mean of time sequence Is: ； In the formula, represents the influence factor data affecting the power generation; represents an unknown smoothing function of the base condition mean; represents an unknown smoothing function of the base condition mean at t-τ time; represents an unknown smoothing function of the condition standard deviation; represents an unknown smoothing function of the condition standard deviation at t-τ time; n is a fixed constant, represents an autoregressive parameter, represents the new energy power generation data at t time, and the obtained processed new energy power curve modeling data set in the interval is taken as input data, represents the new energy power generation data at t-τ time; For autoregressive parameters, Perform iterative updates to obtain the updated conditional mean considering the time series. for: ； wherein represents a stationary process with mean 0 updated iteratively at time t - τ; Conditional cumulative distribution function of new energy power generation power data Is: ； where G is the cumulative distribution function of d, b is a variable, is an unknown smooth function of the updated conditional standard deviation; using a gaussian curve as a kernel function : ； where d is an independent and identically distributed (i.i.d) random variable with mean zero and unknown distribution, is a residual sequence, is a bandwidth; Finally, the new energy power generation power condition kernel density total function of autocorrelation is obtained is: ； wherein ; where T is the total time, is the total kernel function, is a smoothing function of the updated conditional standard deviation at time T+1.

6. The method of claim 5, wherein, ； ； wherein is a stationary process with mean 0, d t are independent and identically distributed (i.i.d) random variables with mean zero and unknown distribution, n is a fixed constant.

7. The method of claim 5, wherein, The method for iteratively updating the autoregressive parameters, is: (1) Given an estimate of to update the autoregressive parameters : ； wherein is a stationary process with mean 0 updated iteratively, is a stationary process with mean 0 at time t - n updated iteratively, (2) Using the updated autoregressive coefficients To update the estimates: ； In the formula, is the new energy power generation data at the input t time after iterative update, is the unknown smoothing function of the conditional standard deviation after iterative update, is the autoregressive parameter after iterative update; Iterate the above two steps until convergence.

8. The method of claim 7, wherein, The convergence criterion for iteration is the Box-Ljung test on all n-lagged terms of the residual series The iteration is terminated when the autoregressive parameters converge and all n-lagged autocorrelation Box-Ljung tests are not rejected.

9. The method of claim 1, wherein, In step S4, the minimum mean squared integral error value between the estimated density and the unknown true density includes a squared bias and a squared variance, i.e.: wherein, is a new energy power generation power condition kernel density total function of autocorrelation , is a bandwidth, represents a variance of the probability density function estimator, is a square of the probability density function estimator bias, ; is a high order term in the bias expression, is a high order variance term; Gaussian curve is used as the kernel function , the optimal bandwidth h is obtained best is: ； wherein unknown smooth function representing the conditional standard deviation, n is a constant.

10. The method of claim 1, wherein, It also includes evaluating the reliability and accuracy of the output probability density model of new energy obtained by using average deviation and average width.

Images