Net Load Prediction Method and System Based on Copula Structure

Abstract

The present invention provides a net load prediction method and system based on a Copula structure. The implementation solution is as follows: The dictionary learning method is used to learn the new energy output and load of the first regional power grid at the first time, so as to obtain the predicted new energy output and predicted load of the first regional power grid, and the initial predicted net load is determined; The Gaussian Copula structure is used to model the prediction errors of the predicted new energy output and the predicted load to obtain the joint probability density function; Sampling of the prediction error is performed on the joint probability density function to obtain the new energy output error sample and load error sample of the first regional power grid at the first time; Based on the new energy output error sample and the load error sample, the net load error is determined; Based on the net load error, the predicted net load is corrected to obtain the target predicted net load of the first regional power grid at the first time. By using the present invention, the accuracy of the net load can be improved.

Description

Technical Field

[0001] The present invention relates to the field of power technology, and in particular, to a net load forecasting method and system based on Copula structure. Background Art

[0002] With the large-scale access of renewable energy, load forecasting in traditional power systems faces many challenges. The volatility and uncertainty of renewable energy generation have greatly reduced the forecasting accuracy of grid load. To ensure the safe and stable operation of the power system, it is necessary to accurately forecast the net load, that is, to eliminate the influence brought by renewable energy generation such as wind energy and solar energy, and forecast the actual net load that the power grid needs to dispatch without the support of these energies.

[0003] However, most traditional load forecasting methods rely on historical load data and simple time series models, and cannot fully consider the influence of renewable energy. Therefore, how to improve the accuracy, reliability and real-time performance of net load forecasting has become an important issue in current power system scheduling and optimization. Summary of the Invention

[0004] The present invention provides a net load forecasting method and system based on Copula structure, which can solve at least one of the above technical problems.

[0005] According to one aspect of the present invention, a net load forecasting method based on Copula structure is provided, including:

[0006] Adopting a dictionary learning method to learn the new energy output and load of the first regional power grid at the first time, and obtaining the predicted new energy output and predicted load of the first regional power grid at the first time;

[0007] Based on the predicted new energy output and the predicted load, determining the predicted net load of the first regional power grid within the first time;

[0008] Adopting a Gaussian Copula structure to model the prediction errors of the predicted new energy output and the prediction errors of the predicted load, and obtaining a joint probability density function;

[0009] Sampling the prediction errors of the joint probability density function to obtain the new energy output error sample and load error sample of the first regional power grid at the first time;

[0010] Based on the new energy output error sample and the load error sample, determining the net load error;

[0011] Based on the net load error, correcting the predicted net load to obtain the target predicted net load of the first regional power grid at the first time.

[0012] According to another aspect of the present invention, there is provided a net load prediction device based on a Copula structure, including:

[0013] An output and load prediction module, configured to use a dictionary learning method to learn the new energy output and load of the first regional power grid at a first time, so as to obtain the predicted new energy output and predicted load of the first regional power grid at the first time;

[0014] A net load determination module, configured to determine the predicted net load of the first regional power grid at the first time based on the predicted new energy output and the predicted load;

[0015] A prediction error modeling module, configured to use a Gaussian Copula structure to model the prediction error of the predicted new energy output and the prediction error of the predicted load, so as to obtain a joint probability density function;

[0016] A prediction error sampling module, configured to perform prediction error sampling on the joint probability density function to obtain a new energy output error sample and a load error sample of the first regional power grid at the first time;

[0017] A net load error determination module, configured to determine a net load error based on the new energy output error sample and the load error sample;

[0018] A net load correction module, configured to correct the predicted net load based on the net load error to obtain the target predicted net load of the first regional power grid at the first time.

[0019] According to another aspect of the present invention, there is provided a net load prediction system based on a Copula structure, including: at least one processor, and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the processor is configured to obtain the instructions from the memory and execute the instructions, so that the at least one processor can execute any one of the net load prediction methods based on the Copula structure in the embodiments of the present invention.

[0020] Adopting the technical solution of the present invention, a dictionary learning method is used to learn the new energy output and load of the first regional power grid at the first time, and the predicted new energy output and predicted load of the first regional power grid at the first time are obtained. Thus, based on the predicted new energy output and the predicted load, an initial predicted net load can be determined. Then, a Gaussian Copula structure is used to model the prediction errors of the predicted new energy output and the prediction errors of the predicted load, obtaining a joint probability density function, and sampling the prediction errors of the joint probability density function to obtain a new energy output error sample and a load error sample of the first regional power grid at the first time. Using the new energy output error sample and the load error sample, the corresponding net load error can be determined. Furthermore, based on the net load error, the initial predicted net load is corrected to accurately obtain the target predicted net load of the first regional power grid at the first time. Therefore, adopting the technical solution of the present invention can improve the accuracy of net load prediction.

[0021] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings are used to better understand the present solution and do not constitute a limitation to the present invention. Among them:

[0023] Figure 1 is a flowchart of a net load prediction method based on a Copula structure according to an embodiment of the present invention;

[0024] Figure 2 is a structural block diagram of a net load prediction device based on a Copula structure according to an embodiment of the present invention;

[0025] Figure 3 is a block diagram of an electronic device for implementing the method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following describes exemplary embodiments of the present invention with reference to the drawings. Various details of the embodiments of the present invention are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the present invention. Similarly, for clarity and conciseness, the description below omits the description of well-known functions and structures.

[0027] Figure 1 is a flowchart of a net load prediction method based on a Copula structure according to an embodiment of the present invention.

[0028] As Figure 1As shown, the net load prediction method based on the Copula structure may include:

[0029] S110, using the dictionary learning method to learn the new energy output and load of the first regional power grid at the first time, and obtaining the predicted new energy output and predicted load of the first regional power grid at the first time;

[0030] S120, based on the predicted new energy output and predicted load, determining the predicted net load of the first regional power grid at the first time;

[0031] S130, using the Gaussian Copula structure to model the prediction errors of the predicted new energy output and the predicted load, and obtaining the joint probability density function;

[0032] S140, sampling the prediction errors of the joint probability density function to obtain the new energy output error sample and load error sample of the first regional power grid at the first time;

[0033] S150, based on the new energy output error sample and load error sample, determining the net load error;

[0034] S160, based on the net load error, correcting the predicted net load to obtain the target predicted net load of the first regional power grid at the first time.

[0035] Exemplarily, the new energy output may include wind power output and photovoltaic power output. In some examples, it may also include hydropower output, thermal power output, etc.

[0036] Exemplarily, the first time may be divided into multiple moments on multiple dates.

[0037] It can be understood that the new energy output and load of the first regional power grid at the first time can be regarded as measured values.

[0038] Exemplarily, before learning the new energy output and load of the first regional power grid at each moment on each date, new energy output data and load data reflecting different seasons and different weather conditions can be selected, and the new energy output data and load data are processed for outliers to obtain the new energy output data and load data for dictionary learning.

[0039] Exemplarily, the method for outlier processing is: based on the new energy output and load output at each time point, respectively determine the mean and standard deviation of the new energy output, and determine the mean and standard deviation of the load output, and use the mean and standard deviation corresponding to the new energy output and load output respectively to calculate the standard scores of the new energy output and load at each time point.

[0040] Exemplarily, the following formula can be used to calculate the standard score :

[0041] ;

[0042] wherein, represents the new energy output or load output at a certain time point, represents the mean value corresponding to the new energy output or load output, represents the standard deviation corresponding to the new energy output or load output.

[0043] Next, if the standard score of the new energy output or load at each time point is greater than a preset value, such as 3, then the new energy output or load output at that time point is set to the corresponding mean value.

[0044] Exemplarily, subtract the predicted new energy output from the predicted load to obtain the predicted net load.

[0045] For example, if the new energy output includes wind power output and photovoltaic output, subtract the sum of the predicted wind power output and the predicted photovoltaic output from the predicted load to obtain the predicted net load, as shown in the following formula:

[0046] ;

[0047] wherein, represents the predicted net load at time represents the predicted load at time represents the predicted wind power output at time represents the predicted photovoltaic output at time.

[0048] Exemplarily, subtract the new energy output at the first time from the predicted new energy output to obtain the prediction error of the predicted new energy output. Subtract the load at the first time from the predicted load to obtain the prediction error of the predicted load. If the above first time is divided into multiple moments on multiple dates, these actual output data or predicted output data can be constructed into a matrix, and then corresponding calculations can be performed. For example, subtract the predicted new energy output matrix from the new energy output matrix to obtain the prediction error matrix of the predicted new energy output matrix, and subtract the predicted load matrix from the load matrix to obtain the prediction error matrix of the predicted load matrix.

[0049] Exemplarily, for example, if the new energy output includes wind power output and photovoltaic output, then through the foregoing steps, the prediction errors of the predicted load, predicted photovoltaic output, and predicted wind power output are calculated. Since there are common dependent factors among these prediction errors, such as climate, etc., through steps S130 to S140, the prediction errors with interdependent relationships are combined. In the subsequent step S150, the initial predicted values are corrected, which can improve the accuracy of the net load prediction.

[0050] Exemplarily, multiple new energy output error samples and multiple load error samples can be extracted. Calculate the mean of the extracted samples, and then subtract the mean of the new energy output error samples from the mean of the load error samples to obtain the net load error.

[0051] Exemplarily, the Monte Carlo sampling method can be used to extract new energy output error samples and load error samples from the joint probability density function obtained in the previous step until a sufficient sample size is obtained. Of course, other sampling methods can also be used to extract new energy output error samples and load error samples from the joint probability density function.

[0052] It can be understood that the errors involved in the embodiments of the present invention are all prediction errors.

[0053] For example, gradually increase the number of samples of the new energy output error and the load error samples. Observe that when the fluctuations of the means and variances of the new energy output error samples and the load error samples become very small and converge, stop increasing the sample quantity and determine the minimum sample size. .

[0054] Exemplarily, for samples , and , calculate , specifically as follows:

[0055] ;

[0056] Wherein, represents the th predicted net load error at time represents the th load error sample at time represents the th wind power output error sample at time represents the th photovoltaic output error sample at time.

[0057] Exemplarily, for those with negative values The mean value of the predicted net load error at a moment is calculated. The absolute value of this mean value is added to the initial predicted net load to obtain the upper boundary of the predicted net load, as follows:

[0058] ;

[0059] ;

[0060] Wherein, represents the upper boundary of the predicted net load at the moment, represents the predicted net load at the moment, represents the mean value of the predicted net load error at the moment that is negative.

[0061] In this example, if the error of a certain predicted value is a negative error, it means that the predicted value is less than the measured value, and the absolute value of the error needs to be added to correct this predicted value.

[0062] In this example, by adding the average absolute value of the negative net load error to correct the initial predicted net load, the maximum possible net load demand can be estimated.

[0063] Exemplarily, the mean value of the predicted net load error at the moment that is positive is calculated. The initial predicted net load is subtracted by the absolute value of this mean value to obtain the lower boundary of the predicted net load, as follows:

[0064] ;

[0065] ;

[0066] Wherein, represents the lower boundary of the predicted net load at the moment, represents the predicted net load at the moment, represents the mean value of the predicted net load error at the moment that is positive.

[0067] In this example, if the error of a certain predicted value is a positive error, it means that the predicted value is higher than the measured value, and the absolute value of the error needs to be subtracted to correct this predicted value.

[0068] In this example, by subtracting the average absolute value of the positive net load error to correct the initial predicted net load, the minimum possible net load demand can be estimated.

[0069] Exemplarily, for the corrected predicted net load, it can be: ​​​

[0070] ;

[0071] Among them, represents the predicted net load after moment correction, which is the mean value of the lower boundary and the upper boundary of the predicted net load at the moment.

[0072] According to the above embodiments, the dictionary learning method is adopted to learn the new energy output and load of the first regional power grid at the first time, and the predicted new energy output and predicted load of the first regional power grid at the first time are obtained. Thus, based on the predicted new energy output and the predicted load, the initial predicted net load can be determined. Then, by using the Gaussian Copula structure, the prediction errors of the predicted new energy output and the predicted load are modeled to obtain the joint probability density function, and the prediction error sampling is performed on the joint probability density function to obtain the new energy output error sample and the load error sample of the first regional power grid at the first time. By using the new energy output error sample and the load error sample, the corresponding net load error can be determined. Thus, the net load error considering the dependence relationship between the new energy output and the load can be obtained. Furthermore, based on the net load error, the initial predicted net load is corrected to accurately obtain the target predicted net load of the first regional power grid at the first time. Therefore, by adopting the technical solution of the present invention, the accuracy of the net load prediction can be improved.

[0073] In one embodiment, the dictionary learning method is adopted to learn the new energy output and load of the first regional power grid at each moment on each date, and the predicted new energy output and predicted load of the first regional power grid at the first time are obtained, including: initializing the new energy output dictionary and the new energy output sparse matrix, as well as the load dictionary and the load sparse matrix based on the new energy output and load of the first regional power grid at each moment on each date; for the new energy output and the load, respectively, based on the matrix obtained by multiplying the dictionary with the deleted columns and the sparse matrix with the corresponding deleted rows column by column and row by row, updating the columns of the dictionary and updating the rows of the sparse matrix to obtain the updated new energy output dictionary with updated columns and the updated new energy output sparse matrix with updated rows, as well as the updated load dictionary with updated columns and the updated load sparse matrix with updated rows; determining the predicted new energy output of the first regional power grid at the first time based on the product of the updated new energy output dictionary with updated columns and the updated new energy output sparse matrix with updated rows; determining the predicted load of the first regional power grid at the first time based on the product of the updated load dictionary with updated columns and the updated load sparse matrix with updated rows.

[0074] In one implementation, based on the new - energy output and load at each moment on each date within the first time period of the first regional power grid, initialize the new - energy output dictionary, the new - energy output sparse matrix, the load dictionary, and the load sparse matrix, including: Based on the new - energy output and load at each moment on each date of the first regional power grid, respectively determine the new - energy output sample matrix and the load sample matrix, where each row in the new - energy output sample matrix and the load sample matrix represents each moment, and each column represents each date; Based on the first column vectors in the left singular matrix after decomposing the new - energy output sample matrix, determine the new - energy output dictionary; Based on the standard normal distribution of the output of each column in the new - energy output sample matrix, determine the non - zero elements in the corresponding column of the new - energy output sparse matrix; Based on the first column vectors in the left singular matrix after decomposing the load sample matrix, determine the load dictionary; Based on the standard normal distribution of the output of each column in the load sample matrix, determine the non - zero elements in the corresponding column of the load sparse matrix.

[0075] Exemplarily, the new - energy output sample matrix may include rows and columns of new - energy output, where the th column includes the new - energy output from the 1st moment to the th moment on the th date, and the th row includes the new - energy output at the th moment from the 1st date to the th date. The new - energy output sample matrix may include a wind - power output sample matrix and a photovoltaic - power output sample matrix, is a positive integer, , is a positive integer, is less than or equal to .

[0076] Exemplarily, the load sample matrix may include rows and columns of load, where the th column includes the load from the 1st moment to the th moment on the th date, and the th row includes the load at the th moment from the 1st date to the th date at the

[0077] Exemplarily, the new - energy output dictionary may include rows and row column load, where .

[0078] Exemplarily, both the new energy output sparse matrix and the load sparse matrix can be row column structure. Among them, the process of initializing these two matrices can be:

[0079] For the sparse matrix , let have a sparsity of 0.1, which means that 10% of the elements in each column of the matrix are non-zero. Randomly select the positions of the non-zero elements in each column, and generate the numerical values of the non-zero elements at these positions from the standard normal distribution of the corresponding column elements in the corresponding sample matrix, so as to ensure the uniformity of the elements in the initialized sparse matrix.

[0080] Exemplarily, the above new energy output sample matrix can be divided into a photovoltaic output sample matrix and a wind power output sample matrix. Similarly, the photovoltaic output dictionary and the wind power output dictionary, as well as the photovoltaic output sparse matrix and the wind power output sparse matrix, can be calculated according to the above example.

[0081] According to the above embodiments, initializing the new energy output dictionary and the new energy output sparse matrix, as well as the load dictionary and the load sparse matrix, can ensure the uniformity of the elements during initialization.

[0082] In one embodiment, for the new energy output and the load, respectively, column by column and row by row, based on the matrix obtained by multiplying the dictionary of the deleted columns by the sparse matrix of the deleted corresponding rows, the dictionary is updated column by column and the sparse matrix is updated row by row, including: deleting the th column in the new energy output dictionary and the th row in the new energy output sparse matrix to obtain the new energy output intermediate dictionary and the new energy output intermediate sparse matrix; subtracting the product between the new energy output intermediate dictionary and the new energy output intermediate sparse matrix from the new energy output sample matrix to obtain the new energy output error matrix; updating the elements in the th column of the new energy output dictionary based on the first column vector in the left singular matrix after the decomposition of the new energy output error matrix; updating the elements in the th row of the new energy output sparse matrix based on the product of the first singular value in the diagonal matrix after the decomposition of the new energy output error matrix and the first column vector in the right singular matrix.

[0083] In one embodiment, the above method further includes: for the th column in the load dictionary and the Delete rows to obtain the load intermediate dictionary and the load intermediate sparse matrix; subtract the product of the load intermediate dictionary and the load intermediate sparse matrix from the load sample matrix to obtain the load error matrix; based on the first column vector in the left singular matrix after the decomposition of the load error matrix, update the elements in the th column of the load dictionary; based on the product of the first singular value in the diagonal matrix after the decomposition of the load error matrix and the first column vector in the right singular matrix, update the elements in the th row of the load sparse matrix, where is a positive integer, less than or equal to .

[0084] Exemplarily, remove the th column in the dictionary to obtain the dictionary , remove the th row in the sparse matrix to obtain the sparse matrix , and calculate the error matrix , where is the sample matrix.

[0085] Exemplarily, perform singular value decomposition on the error matrix , that is , where is the left singular matrix, is the diagonal matrix, is the right singular matrix. Then, take the first column in the left singular matrix to update the elements in the th column of the dictionary , which can ensure that the update direction of the dictionary atoms can be updated along the direction most relevant to the original data. And, update the result of multiplying the first singular value of the diagonal matrix and the first column of the right singular matrix as the elements in the th row of the sparse matrix , so that it can be ensured that the coefficients in the th row of the sparse matrix correspond to the updated dictionary .

[0086] Exemplarily, start executing the above steps from , and add 1 to after each execution of the above steps until all the column vectors in the dictionary and all the row vectors in the sparse matrix have been updated.

[0087] Exemplarily, the product of the updated new energy output dictionary for each column and the updated new energy output sparse matrix for each row is used as the predicted new energy output of the first regional power grid at the first time. For example, the product of the updated wind power output dictionary and the updated wind power output sparse matrix is used as the predicted wind power output. Another example is that the product of the updated photovoltaic power output dictionary and the updated photovoltaic power output sparse matrix is used as the predicted photovoltaic power output.

[0088] It can be understood that the predicted new energy output at the first time here can be a matrix, and this matrix includes the predicted new energy output at each moment on each date within the first time.

[0089] Exemplarily, the product of the updated load dictionary for each column and the updated load sparse matrix for each row is used as the predicted load of the first regional power grid at the first time.

[0090] It can be understood that the predicted load at the first time here can be a matrix, and this matrix includes the load at each moment on each date within the first time.

[0091] In one implementation, predicting new energy output includes predicting photovoltaic power output and predicting wind power output. Using the Gaussian Copula structure, the prediction errors of the predicted new energy output and the prediction errors of the predicted load are modeled to obtain a joint probability density function, including: determining the first conditional distribution function of the wind power output error, the second conditional distribution function of the photovoltaic power output error, and the joint conditional distribution function between the wind power output error and the photovoltaic power output error under a specified load error based on the first marginal distribution function of the prediction error of the predicted wind power output, the second marginal distribution function of the prediction error of the predicted photovoltaic power output, and the third marginal distribution function of the prediction error of the predicted load; determining the first dependence relationship between the load error and the wind power output error, the second dependence relationship between the load error and the photovoltaic power output error, and the conditional dependence relationship between the wind power output error and the photovoltaic power output error under a specified load error based on the inverse cumulative distribution functions of the standard normal distributions corresponding to the first marginal distribution function, the second marginal distribution function, the third marginal distribution function, the first conditional distribution function, the second conditional distribution function, and the joint conditional distribution function; determining the joint probability density function based on the first dependence relationship, the second dependence relationship, and the conditional dependence relationship.

[0092] Exemplarily, subtracting the actual load from the predicted load obtained above can yield the load prediction error , subtracting the actual wind power output from the predicted wind power output obtained above can yield the wind power output prediction error , subtracting the actual photovoltaic power output from the predicted photovoltaic power output obtained above can yield the photovoltaic power output prediction error . That is, . , are the differences between the predicted values , , and the corresponding actual values respectively.

[0093] Exemplarily, in order to construct the joint probability density function of these three prediction errors, in this example, let be the load prediction error , the wind power output prediction error , and the photovoltaic power output prediction error . The three-dimensional joint distribution function is as follows:

[0094] ;

[0095] where , , are the marginal distribution functions of the load prediction error, the wind power output prediction error, and the photovoltaic power output prediction error respectively, is the Copula function, which is used to describe the dependence relationship among the three.

[0096] In this example, Gaussian Copula is used for modeling. By using the covariance between the sample data of these prediction errors, the correlation coefficients between the corresponding two variables are estimated. Furthermore, by using these correlation coefficients, the linear dependence relationship among the load, wind power, and photovoltaic can be captured, and thus the joint probability density function among these three prediction error variables can be obtained.

[0097] Exemplarily, according to the marginal distributions of the three random variables (load prediction error, wind power output prediction error, and photovoltaic power output prediction error), the respective cumulative distribution functions are calculated, and then they are converted into standard normal variables through the inverse cumulative distribution function of the standard normal distribution. In this way, the scales among different variables can be unified, making the contributions of different variables in the model more balanced.

[0098] Exemplarily, the above functions may include the following: , , , , and .

[0099] where is a random variable that satisfies the standard normal distribution . is the inverse cumulative distribution function of the standard normal distribution. is the first marginal distribution function of the prediction error of the predicted wind power output, For predicting the prediction error of photovoltaic output is the second marginal distribution function of For predicting the prediction error of wind load is the third marginal distribution function of

[0100] Among them, and respectively represent the conditional distribution function of wind power output error when the load error is and the conditional distribution function of photovoltaic output error of represents the joint conditional distribution function of the joint of wind power output error and photovoltaic output error when the load error is of

[0101] Exemplarily, taking the triple partial derivative of the above three-dimensional joint distribution function, the joint probability density function is obtained as follows:

[0102] ;

[0103] Among them, represents the Copula density function, , , are respectively the probability density functions of the prediction errors , , of

[0104] Exemplarily, disassembling the dependence relationship among the three variables into the above three dependence relationships, a new joint probability density function is obtained as follows:

[0105] ;

[0106] ;

[0107] ;

[0108] ;

[0109] In one embodiment, the first dependence relationship is:

[0110] ;

[0111] Among them, represents the first dependence relationship, represents the correlation coefficient matrix between the load error and the wind power output error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the third marginal distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the first marginal distribution function.

[0112] In one embodiment, the second dependency is:

[0113] ;

[0114] wherein, represents the second dependency, represents the correlation coefficient matrix between the load error and the photovoltaic output error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the third marginal distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the second marginal distribution function.

[0115] Exemplarily, the correlation coefficient matrix between the load error and the wind power output error is:

[0116] .

[0117] Exemplarily, the correlation coefficient matrix between the load error and the photovoltaic output error is:

[0118] .

[0119] wherein, is the correlation coefficient between the load error and the wind power output error, as follows:

[0120] ;

[0121] wherein, is the correlation coefficient between the load error and the photovoltaic output error, as follows:

[0122] ;

[0123] wherein, the correlation coefficient between the wind power output error and the photovoltaic output error, as follows:

[0124] ;

[0125] wherein, represents the covariance between the load prediction error and the wind power output prediction error, represents the covariance between the load prediction error and the photovoltaic output prediction error, represents the covariance between the wind power output prediction error and the photovoltaic output prediction error, 、 、 respectively represent the standard deviations of the load prediction error, the wind power output prediction error, and the photovoltaic power output prediction error.

[0126] It can be understood that from the above analysis, the prediction errors of the above-mentioned various power outputs and loads can be matrices, including the prediction errors at each moment on each date within the first time period. By calculating the load prediction error matrix and the wind power output prediction error matrix, the covariance between them can be obtained. Calculating each prediction error matrix can obtain the standard deviation of the corresponding variable.

[0127] In one implementation manner, the above conditional dependence relationship is:

[0128]

[0129] ;

[0130] wherein, represents the conditional dependence relationship, represents the correlation coefficient matrix between the wind power output error and the photovoltaic power output error under the given load error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the first conditional distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the second conditional distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the joint conditional distribution function.

[0131] Exemplarily, the correlation coefficient matrix between the wind power output error and the photovoltaic power output error under the given load error is:

[0132] ;

[0133] wherein, is the correlation coefficient between the wind power output error and the load error under the given load error condition, is the correlation coefficient between the photovoltaic power output error and the load error under the given load error condition, is the correlation coefficient between the wind power output error and the photovoltaic power output error under the given load error condition.

[0134] Exemplarily, the correlation coefficient between the wind power output error and the load error under the given load error condition is as follows:

[0135] ;

[0136] Exemplarily, the correlation coefficient between the photovoltaic power output error and the load error under the given load error condition is as follows:

[0137] ;

[0138] Exemplarily, the correlation coefficient between the wind power output error and the photovoltaic power output error under a given load error condition is as follows:

[0139] .

[0140] According to the above embodiments, the Gaussian Copula structure can accurately describe the dependence relationships between the respective output prediction errors and between each output prediction error and the load prediction error. Thus, the joint probability density function constituted by each output prediction error and the load prediction error can be accurately described.

[0141] Figure 2 is the structural block diagram of a net load prediction device based on the Copula structure according to an embodiment of the present invention.

[0142] As Figure 2 shown, the net load prediction device based on the Copula structure may include:

[0143] An output and load prediction module 210, configured to use a dictionary learning method to learn the new energy output and load of the first regional power grid at the first time, and obtain the predicted new energy output and predicted load of the first regional power grid at the first time;

[0144] A net load determination module 220, configured to determine the predicted net load of the first regional power grid within the first time based on the predicted new energy output and the predicted load;

[0145] A prediction error modeling module 230, configured to use a Gaussian Copula structure to model the prediction errors of the predicted new energy output and the predicted load, and obtain a joint probability density function;

[0146] A prediction error sampling module 240, configured to perform prediction error sampling on the joint probability density function to obtain a new energy output error sample and a load error sample of the first regional power grid at the first time;

[0147] A net load error determination module 250, configured to determine a net load error based on the new energy output error sample and the load error sample;

[0148] A net load correction module 260, configured to correct the predicted net load based on the net load error to obtain the target predicted net load of the first regional power grid at the first time.

[0149] In one embodiment, the output and load prediction module 210 includes:

[0150] An initialization unit, configured to initialize a new energy output dictionary, a new energy output sparse matrix, a load dictionary, and a load sparse matrix based on the new energy output and load at each moment of each date within the first time period of the first regional power grid;

[0151] A matrix update unit, configured to, for the new energy output and the load, respectively update the columns of the dictionary and update the rows of the sparse matrix column by column and row by row based on the matrix obtained by multiplying the dictionary of the deleted columns by the sparse matrix of the deleted corresponding rows, so as to obtain the updated new energy output dictionary with updated columns, the updated new energy output sparse matrix with updated rows, the updated load dictionary with updated columns, and the updated load sparse matrix with updated rows;

[0152] A first prediction unit, configured to determine the predicted new energy output of the first regional power grid at the first time period based on the product of the updated new energy output dictionary with updated columns and the updated new energy output sparse matrix with updated rows;

[0153] A second prediction unit, configured to determine the predicted load of the first regional power grid at the first time period based on the product of the updated load dictionary with updated columns and the updated load sparse matrix with updated rows.

[0154] In one implementation manner, the initialization unit is specifically configured to:

[0155] Based on the new energy output and load at each moment of each date of the first regional power grid, respectively determine a new energy output sample matrix and a load sample matrix, where each row in the new energy output sample matrix and the load sample matrix represents each moment, and each column represents each date;

[0156] Based on the first column vectors in the left singular matrix obtained by decomposing the new energy output sample matrix, determine the new energy output dictionary, where is a positive integer;

[0157] Based on the standard normal distribution of the output of each column in the new energy output sample matrix, determine the non-zero elements of the corresponding column in the new energy output sparse matrix;

[0158] Based on the first column vectors in the left singular matrix obtained by decomposing the load sample matrix, determine the load dictionary;

[0159] Based on the standard normal distribution of the output of each column in the load sample matrix, determine the non-zero elements of the corresponding column in the load sparse matrix.

[0160] In one implementation manner, the matrix update unit is specifically configured to:

[0161] Delete the th column in the new energy output dictionary and the th row in the new energy output sparse matrix to obtain the intermediate new energy output dictionary and the intermediate new energy output sparse matrix;

[0162] Subtract the product of the intermediate new energy output dictionary and the intermediate new energy output sparse matrix from the new energy output sample matrix to obtain the new energy output error matrix;

[0163] Based on the first column vector in the left singular matrix after the decomposition of the new energy output error matrix, update the elements in the th column of the new energy output dictionary;

[0164] Based on the product of the first singular value in the diagonal matrix after the decomposition of the new energy output error matrix and the first column vector in the right singular matrix, update the elements in the th row of the new energy output sparse matrix;

[0165] Delete the th column in the load dictionary and the th row in the load sparse matrix to obtain the intermediate load dictionary and the intermediate load sparse matrix;

[0166] Subtract the product of the intermediate load dictionary and the intermediate load sparse matrix from the load sample matrix to obtain the load error matrix;

[0167] Based on the first column vector in the left singular matrix after the decomposition of the load error matrix, update the elements in the th column of the load dictionary;

[0168] Based on the product of the first singular value in the diagonal matrix after the decomposition of the load error matrix and the first column vector in the right singular matrix, update the elements in the th row of the load sparse matrix, where is a positive integer, is less than or equal to .

[0169] In one implementation, the predicted new energy output includes predicted photovoltaic output and predicted wind power output, and the prediction error modeling module 230 includes:

[0170] A distribution function determination unit, configured to determine a first conditional distribution function of wind power output error, a second conditional distribution function of photovoltaic power output error, and a joint conditional distribution function between wind power output error and photovoltaic power output error under a specified load error, based on a first marginal distribution function of the prediction error of the predicted wind power output, a second marginal distribution function of the prediction error of the predicted photovoltaic power output, and a third marginal distribution function of the prediction error of the predicted load;

[0171] A dependency determination unit, configured to determine a first dependency between load error and wind power output error, a second dependency between load error and photovoltaic power output error, and a conditional dependency between wind power output error and photovoltaic power output error under a specified load error, based on the inverse cumulative distribution functions of the standard normal distributions corresponding to the first marginal distribution function, the second marginal distribution function, the third marginal distribution function, the first conditional distribution function, the second conditional distribution function, and the joint conditional distribution function respectively;

[0172] A density function determination unit, configured to determine the joint probability density function based on the first dependency, the second dependency, and the conditional dependency.

[0173] In one embodiment, the first dependency is:

[0174] ;

[0175] Wherein, represents the first dependency, represents the correlation coefficient matrix between load error and wind power output error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the third marginal distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the first marginal distribution function.

[0176] In one embodiment, the second dependency is:

[0177] ;

[0178] Wherein, represents the second dependency, represents the correlation coefficient matrix between load error and photovoltaic power output error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the third marginal distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the second marginal distribution function.

[0179] In one embodiment, the conditional dependency is:

[0180]

[0181] ;

[0182] Among them, represents the conditional dependency relationship, represents the correlation coefficient matrix between the wind power output error and the photovoltaic power output error under a given load error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the first conditional distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the second conditional distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the joint conditional distribution function.

[0183] For the specific functions and examples of each module and sub-module of the system in the embodiments of the present invention, reference can be made to the relevant descriptions of the corresponding steps in the above method embodiments, which will not be elaborated here.

[0184] In the technical solution of the present invention, the acquisition, storage, and application of user personal information involved all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0185] According to the embodiments of the present invention, the present invention also provides a system and a readable storage medium.

[0186] Figure 3 FIG. shows a schematic block diagram of an exemplary electronic device 800 that can be used to implement the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described herein and / or claimed.

[0187] As Figure 3 shown, the electronic device 800 includes a computing unit 801, which can execute various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 802 or the computer program loaded from the storage unit 808 into the random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. The input / output (I / O) interface 805 is also connected to the bus 804.

[0188] Multiple components in the electronic device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disc, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0189] The computing unit 801 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 executes the various methods and processes described above, such as the net load prediction method based on the Copula structure. For example, in some embodiments, the net load prediction method based on the Copula structure can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the net load prediction method based on the Copula structure described above can be executed. Alternatively, in other embodiments, the computing unit 801 can be configured to execute the net load prediction method based on the Copula structure in any other suitable way (e.g., by means of firmware).

[0190] The various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special or general programmable processor, and can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0191] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing devices, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowchart and / or block diagram are implemented. The program codes can be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server.

[0192] In the context of the present invention, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0193] In order to provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0194] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0195] A computer system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The client-server relationship is created by computer programs that run on the respective computers and have a client-server relationship with each other. The server can be a cloud server, a server of a distributed system, or a server that incorporates blockchain.

[0196] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present invention can be achieved, and this is not limited herein.

[0197] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A net load prediction method based on Copula structure, characterized in that, Including: Using a dictionary learning method to learn the new energy output and load of the first regional power grid at the first time, and obtaining the predicted new energy output and predicted load of the first regional power grid at the first time, including: based on the new energy output and load at each moment on each date within the first time of the first regional power grid, initializing a new energy output dictionary and a new energy output sparse matrix, as well as a load dictionary and a load sparse matrix; for the new energy output and the load, respectively, based on the matrix obtained by multiplying the dictionary with the deleted columns and the sparse matrix with the corresponding deleted rows column by column and row by row, performing column update on the dictionary and row update on the sparse matrix to obtain the new energy output dictionary with updated columns and the new energy output sparse matrix with updated rows, as well as the load dictionary with updated columns and the load sparse matrix with updated rows; based on the product of the new energy output dictionary with updated columns and the new energy output sparse matrix with updated rows, determining the predicted new energy output of the first regional power grid at the first time; based on the product of the load dictionary with updated columns and the load sparse matrix with updated rows, determining the predicted load of the first regional power grid at the first time; Based on the predicted new energy output and the predicted load, determining the predicted net load within the first time of the first regional power grid; Using a Gaussian Copula structure to model the prediction errors of the predicted new energy output and the predicted load, and obtaining a joint probability density function; Sampling the prediction errors of the joint probability density function to obtain a new energy output error sample and a load error sample of the first regional power grid at the first time; Based on the new energy output error sample and the load error sample, determining a net load error; Based on the net load error, correcting the predicted net load to obtain the target predicted net load of the first regional power grid at the first time.

2. The method according to claim 1, characterized in that The initializing the new energy output dictionary and the new energy output sparse matrix, as well as the load dictionary and the load sparse matrix based on the new energy output and load at each moment on each date within the first time of the first regional power grid includes: Based on the new energy output and load at each moment on each date of the first regional power grid, respectively determining a new energy output sample matrix and a load sample matrix, where each row in the new energy output sample matrix and the load sample matrix represents each moment, and each column represents each date; Determine a dictionary of new energy output based on the first column vectors in the left singular matrix after decomposing the new energy output sample matrix, where is a positive integer; Based on the standard normal distribution of the output in each column of the new energy output sample matrix, determining the non-zero elements in the corresponding column of the new energy output sparse matrix; Based on the first column vectors in the left singular matrix after decomposing the load sample matrix, determine the load dictionary; Based on the standard normal distribution of the output in each column of the load sample matrix, determining the non-zero elements in the corresponding column of the load sparse matrix.

3. The method according to claim 2, wherein The performing column update on the dictionary and row update on the sparse matrix respectively for the new energy output and the load based on the matrix obtained by multiplying the dictionary with the deleted columns and the sparse matrix with the corresponding deleted rows column by column and row by row includes: Delete the th column in the new energy output dictionary and the th row in the new energy output sparse matrix to obtain the new energy output intermediate dictionary and the new energy output intermediate sparse matrix; Subtract the product between the new energy output intermediate dictionary and the new energy output intermediate sparse matrix from the new energy output sample matrix to obtain a new energy output error matrix; Based on the first column vector in the left singular matrix after the decomposition of the new energy output error matrix, update the column elements in the new energy output dictionary; Update the elements in the th row of the new energy output sparse matrix based on the product of the first singular value in the diagonal matrix after the decomposition of the new energy output error matrix and the first column vector in the right singular matrix; Delete the th column in the load dictionary and the th row in the load sparse matrix to obtain an intermediate load dictionary and an intermediate load sparse matrix; Subtract the product between the load intermediate dictionary and the load intermediate sparse matrix from the load sample matrix to obtain a load error matrix; Update the column elements in the load dictionary based on the first column vector in the left singular matrix after decomposing the load error matrix; Update the elements in the -th row of the load sparse matrix based on the product of the first singular value in the diagonal matrix after the load error matrix decomposition and the first column vector in the right singular matrix, where is a positive integer, less than or equal to .

4. The method according to claim 1, characterized in that, The predicted new energy output includes predicted photovoltaic output and predicted wind power output. By using the Gaussian Copula structure, the prediction errors of the predicted new energy output and the predicted load are modeled to obtain a joint probability density function, including: Based on the first marginal distribution function of the prediction error of the predicted wind power output, the second marginal distribution function of the prediction error of the predicted photovoltaic output, and the third marginal distribution function of the prediction error of the predicted load, determine the first conditional distribution function of the wind power output error, the second conditional distribution function of the photovoltaic output error, and the joint conditional distribution function between the wind power output error and the photovoltaic output error under a specified load error; Based on the inverse cumulative distribution functions of the standard normal distributions corresponding to the first marginal distribution function, the second marginal distribution function, the third marginal distribution function, the first conditional distribution function, the second conditional distribution function, and the joint conditional distribution function respectively, determine the first dependence relationship between the load error and the wind power output error, the second dependence relationship between the load error and the photovoltaic output error, and the conditional dependence relationship between the wind power output error and the photovoltaic output error under a specified load error; Based on the first dependence relationship, the second dependence relationship, and the conditional dependence relationship, determine the joint probability density function.

5. The method according to claim 4, wherein The first dependence relationship is: ； Among them, represents the first dependency relationship, represents the correlation coefficient matrix between the load error and the wind power output error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the third marginal distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the first marginal distribution function.

6. The method according to claim 4, characterized in that, The second dependence relationship is: ； Among them, represents the second dependency relationship, represents the correlation coefficient matrix between the load error and the photovoltaic output power error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the third marginal distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the second marginal distribution function.

7. The method according to claim 4, wherein The conditional dependence relationship is: ； Among them, represents the conditional dependence relationship, represents the correlation coefficient matrix between the wind power output error and the photovoltaic power output error under a given load error, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the first conditional distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the second conditional distribution function, represents the inverse cumulative distribution function of the standard normal distribution corresponding to the joint conditional distribution function.

8. A net load prediction device based on the Copula structure, characterized in that, Including: An output and load prediction module, which is used to adopt a dictionary learning method to learn the new energy output and load of the first regional power grid at the first time, and obtain the predicted new energy output and predicted load of the first regional power grid at the first time; A net load determination module, which is used to determine the predicted net load of the first regional power grid at the first time based on the predicted new energy output and the predicted load; A prediction error modeling module, which is used to adopt a Gaussian Copula structure to model the prediction errors of the predicted new energy output and the predicted load, and obtain a joint probability density function; A prediction error sampling module, which is used to sample the prediction errors of the joint probability density function to obtain the new energy output error sample and the load error sample of the first regional power grid at the first time; A net load error determination module, which is used to determine the net load error based on the new energy output error sample and the load error sample; A net load correction module, which is used to correct the predicted net load based on the net load error to obtain the target predicted net load of the first regional power grid at the first time; Among them, the output and load prediction module includes: An initialization unit, configured to initialize a new energy output dictionary, a new energy output sparse matrix, a load dictionary, and a load sparse matrix based on the new energy output and load at each moment on each date within the first time period of the first regional power grid; A matrix update unit, configured to, for the new energy output and the load, respectively update the columns of the dictionary and update the rows of the sparse matrix column by column and row by row based on the matrix obtained by multiplying the dictionary with the deleted columns and the sparse matrix with the corresponding deleted rows, so as to obtain an updated new energy output dictionary with updated columns and an updated new energy output sparse matrix with updated rows, as well as an updated load dictionary with updated columns and an updated load sparse matrix with updated rows; A first prediction unit, configured to determine the predicted new energy output of the first regional power grid at the first time period based on the product of the updated new energy output dictionary with updated columns and the updated new energy output sparse matrix with updated rows; A second prediction unit, configured to determine the predicted load of the first regional power grid at the first time period based on the product of the updated load dictionary with updated columns and the updated load sparse matrix with updated rows.

9. A net load prediction system based on Copula structure, characterized in that, Comprising: At least one processor, and a memory communicatively connected to the at least one processor; Wherein, the memory stores instructions executable by the at least one processor, and the processor is configured to obtain the instructions from the memory and execute the instructions, so that the at least one processor can execute the method according to any one of claims 1-7.

Images