Data dimensionality reduction method, device, computer equipment and storage medium

Abstract

The present application relates to a data dimension reduction method, device, computer equipment, storage medium and computer program product. The method includes: obtaining multiple influencing factors and influencing factor data of wind power; performing multicollinearity diagnosis on the multiple influencing factors according to the influencing factor data of the target season, and if there is multicollinearity between the multiple influencing factors, the ridge regression coefficient of each influencing factor is calculated according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter; screening each influencing factor according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, eliminating the influencing factors whose ridge regression coefficient does not meet the preset ridge regression coefficient change conditions, and obtaining multiple target influencing factors. By adopting this method, multiple influencing factors can be screened by the ridge regression coefficients of each influencing factor, and data redundancy can be reduced while ensuring the application value of the data.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a data dimensionality reduction method, apparatus, computer equipment, storage medium and computer program product. Background Art

[0002] As people put forward higher requirements for low-carbon electric energy, renewable energy, mainly wind power and photovoltaic power, has gradually replaced fossil energy and occupied an increasingly large proportion in the field of power generation. As the most widely used new energy, wind power generation has typical randomness and unpredictability, which increases the difficulty of formulating reasonable power planning schemes. How to calculate the power generation capacity of new energy as accurately as possible under such unpredictable natural conditions is an urgent problem to be solved.

[0003] There is a huge amount of data in the current power grid that can be used to analyze and mine the characteristics of wind power generation. Since there is often a linear or nonlinear relationship between the different types of influencing factors collected by the wind tower, there are redundant influencing factors. These redundant data not only increase the complexity of modeling, but also lead to inaccurate related models. In order to better mine the data value of numerical weather forecasts, it is necessary to study the problem of data redundancy, use effective methods to filter and reduce the dimension of data, reduce data redundancy, and improve the application value of data.

[0004] The dimensionality reduction methods in traditional technologies can be divided into linear dimensionality reduction methods and nonlinear dimensionality reduction methods. Representative algorithms for linear dimensionality reduction include principal component analysis (PCA) and linear discriminant analysis (LDA). Linear methods embed the original data set into a linear structure, and the model calculation is simple, which works well for linear and Gaussian distributed data. However, when the actual data has an obvious nonlinear structure, the dimensionality reduction effect is obviously poor. Among the commonly used nonlinear dimensionality reduction methods, the most critical kernel selection in the kernel principal component analysis (KPCA) is difficult, and currently can only be judged based on experience. The t-distributed neighborhood embedding (t-SNE) algorithm has high computational complexity and can only reduce dimensions to 2 or 3 dimensions. For wind power prediction, the data points may be mixed together and cannot be distinguished, and the dimensionality reduction effect is not good. Summary of the invention

[0005] Based on this, it is necessary to provide a data dimensionality reduction method, device, computer equipment, computer-readable storage medium and computer program product that can improve the application value of data in response to the above-mentioned technical problems.

[0006] In a first aspect, the present application provides a data dimensionality reduction method. The method comprises:

[0007] Acquire multiple influencing factors of wind power and influencing factor data, wherein the influencing factor data includes parameter values ​​of the multiple influencing factors at multiple times in a target season;

[0008] According to the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor;

[0009] If it is determined that there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnostic parameters of each of the influencing factors, the ridge regression coefficient of each of the influencing factors is calculated according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter;

[0010] The influencing factors are screened according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, and the influencing factors whose ridge regression coefficients do not meet the preset ridge regression coefficient change conditions are eliminated to obtain multiple target influencing factors.

[0011] In one embodiment, after the step of performing multicollinearity diagnosis on the multiple influencing factors according to the influencing factor data of the target season to obtain multicollinearity diagnosis parameters of each influencing factor, the method further includes:

[0012] According to the multicollinearity diagnostic parameter of each influencing factor, it is determined whether there is multicollinearity between the multiple influencing factors.

[0013] In one of the embodiments, the multicollinearity diagnostic parameters of the influencing factors include tolerance;

[0014] According to the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor, including:

[0015] For each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor;

[0016] Calculating the tolerance according to a preset target value and the square of the complex correlation coefficient;

[0017] The step of judging whether there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor includes:

[0018] If the tolerance of each of the influencing factors is less than the first preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0019] In one embodiment, the multicollinearity diagnostic parameters of the influencing factors include tolerance and variance inflation factor;

[0020] According to the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor, including:

[0021] For each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor;

[0022] Calculating the tolerance according to a preset target value and the square of the complex correlation coefficient;

[0023] The reciprocal of the tolerance is used as the variance inflation factor of the influencing factor;

[0024] The step of judging whether there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor includes:

[0025] According to the variance inflation factor of each influencing factor, determine whether there is multicollinearity among the multiple influencing factors;

[0026] If the variance inflation factor of each influencing factor is greater than a second preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0027] In one embodiment, the multicollinearity diagnostic parameter of the influencing factors includes a condition index;

[0028] According to the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor, including:

[0029] Determining a first data matrix according to the influencing factor data of the target season;

[0030] Determine a first measurement matrix according to the first data matrix and a transposed matrix of the first data matrix;

[0031] Calculating at least one eigenvalue of the first measurement matrix;

[0032] The ratio of the largest characteristic root to the smallest characteristic root is taken as the maximum condition index;

[0033] The step of judging whether there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor includes:

[0034] Judging whether there is multicollinearity among the multiple influencing factors according to the maximum condition index;

[0035] If the maximum condition index is greater than or equal to a third preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0036] In one embodiment, the calculating of the ridge regression coefficient of each of the influencing factors according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter includes:

[0037] Determining a first data matrix according to the influencing factor data of the target season;

[0038] Determine a first measurement matrix according to the first data matrix and a transposed matrix of the first data matrix;

[0039] Determining a target matrix according to the first measurement matrix, the ridge parameter and a preset standard value;

[0040] Determine a second observation matrix according to the transposed matrix of the first data matrix and the preset influencing factor observation data;

[0041] According to the target matrix and the second observation matrix, the ridge regression coefficient of each of the influencing factors is obtained.

[0042] In one embodiment, the influencing factors are screened according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, and the influencing factors whose ridge regression coefficients do not meet the preset ridge regression coefficient change conditions are eliminated to obtain multiple target influencing factors, including:

[0043] When the ridge parameter is within a preset range, determining the influencing factor whose ridge regression coefficient meets the preset stability condition and whose absolute value is less than the preset stability threshold as the first influencing factor;

[0044] When the ridge parameter is within a preset range, determining the influencing factor whose ridge regression coefficient does not meet the preset stability condition and meets the preset ridge regression coefficient change trend condition as the second influencing factor;

[0045] Each influencing factor is screened, the first influencing factor and the second influencing factor are eliminated, and a plurality of target influencing factors are obtained.

[0046] In one embodiment, the method further comprises:

[0047] The influencing factor data of the target season corresponding to the multiple target influencing factors are input into a pre-trained convolutional neural network to obtain the predicted wind power generation power of the target season.

[0048] In a second aspect, the present application also provides a data dimension reduction device. The device comprises:

[0049] An acquisition module, used to acquire multiple influencing factors of wind power and influencing factor data of a target season, wherein the influencing factor data includes parameter values ​​of the multiple influencing factors at multiple times;

[0050] A diagnosis module, used for performing multicollinearity diagnosis on the multiple influencing factors according to the influencing factor data of the target season, and obtaining multicollinearity diagnosis parameters of each influencing factor;

[0051] A calculation module, for calculating the ridge regression coefficient of each of the influencing factors according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter, if it is determined that there is multicollinearity between the multiple influencing factors according to the multicollinearity diagnostic parameters of the influencing factors;

[0052] The screening module is used to screen various influencing factors according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, eliminate the influencing factors whose ridge regression coefficient does not meet the preset ridge regression coefficient change condition, and obtain multiple target influencing factors.

[0053] In one embodiment, the device further comprises:

[0054] The judgment module is used to judge whether there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnosis parameter of each influencing factor.

[0055] In one of the embodiments, the multicollinearity diagnostic parameters of the influencing factors include tolerance;

[0056] The diagnostic module is specifically used for:

[0057] For each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor;

[0058] Calculating the tolerance according to a preset target value and the square of the complex correlation coefficient;

[0059] The judgment module is specifically used for:

[0060] If the tolerance of each of the influencing factors is less than the first preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0061] In one embodiment, the multicollinearity diagnostic parameters of the influencing factors include tolerance and variance inflation factor;

[0062] The diagnostic module is specifically used for:

[0063] For each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor;

[0064] Calculating the tolerance according to a preset target value and the square of the complex correlation coefficient;

[0065] The reciprocal of the tolerance is used as the variance inflation factor of the influencing factor;

[0066] The judgment module is specifically used for:

[0067] According to the variance inflation factor of each influencing factor, determine whether there is multicollinearity among the multiple influencing factors;

[0068] If the variance inflation factor of each influencing factor is greater than a second preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0069] In one embodiment, the multicollinearity diagnostic parameter of the influencing factors includes a condition index;

[0070] The diagnostic module is specifically used for:

[0071] Determining a first data matrix according to the influencing factor data of the target season;

[0072] Determine a first measurement matrix according to the first data matrix and a transposed matrix of the first data matrix;

[0073] Calculating at least one eigenvalue of the first measurement matrix;

[0074] The ratio of the largest characteristic root to the smallest characteristic root is taken as the maximum condition index;

[0075] The judgment module is specifically used for:

[0076] Judging whether there is multicollinearity among the multiple influencing factors according to the maximum condition index;

[0077] If the maximum condition index is greater than or equal to a third preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0078] In one embodiment, the computing module is specifically configured to:

[0079] Determining a first data matrix according to the influencing factor data of the target season;

[0080] Determine a first measurement matrix according to the first data matrix and a transposed matrix of the first data matrix;

[0081] Determining a target matrix according to the first measurement matrix, the ridge parameter and a preset standard value;

[0082] Determine a second observation matrix according to the transposed matrix of the first data matrix and the preset influencing factor observation data;

[0083] According to the target matrix and the second observation matrix, the ridge regression coefficient of each of the influencing factors is obtained.

[0084] In one embodiment, the screening module is specifically used for:

[0085] When the ridge parameter is within a preset range, determining the influencing factor whose ridge regression coefficient meets the preset stability condition and whose absolute value is less than the preset stability threshold as the first influencing factor;

[0086] When the ridge parameter is within a preset range, determining the influencing factor whose ridge regression coefficient does not meet the preset stability condition and meets the preset ridge regression coefficient change trend condition as the second influencing factor;

[0087] Each influencing factor is screened, the first influencing factor and the second influencing factor are eliminated, and a plurality of target influencing factors are obtained.

[0088] In one embodiment, the device further comprises:

[0089] The prediction module is used to input the influencing factor data of the target season corresponding to multiple target influencing factors into a pre-trained convolutional neural network to obtain the predicted wind power generation power of the target season.

[0090] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the above steps when executing the computer program.

[0091] In a fourth aspect, the present application further provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and the computer program implements the above steps when executed by a processor.

[0092] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and the above steps are implemented when the computer program is executed by a processor.

[0093] The above-mentioned data dimension reduction method, device, computer equipment, storage medium and computer program product, the method includes, obtaining multiple influencing factors and influencing factor data of wind power, the influencing factor data including parameter values ​​of multiple influencing factors at multiple times in the target season; according to the influencing factor data of the target season, performing multicollinearity diagnosis on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor; if it is determined that there is multicollinearity between the multiple influencing factors according to the multicollinearity diagnosis parameters of each influencing factor, then according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter, the ridge regression coefficient of each influencing factor is calculated; according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, each influencing factor is screened, and the influencing factors whose ridge regression coefficient does not meet the preset ridge regression coefficient change conditions are eliminated to obtain multiple target influencing factors. By adopting this method, multiple influencing factors can be screened by the ridge regression coefficient of each influencing factor, and data redundancy can be reduced under the condition of ensuring the application value of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 Schematic diagram of a flow chart of a data dimension reduction method in one embodiment;

[0095] Figure 2 A schematic diagram of a process for calculating tolerance in one embodiment;

[0096] Figure 3 It is a flowchart of the steps of making judgments according to tolerance in one embodiment;

[0097] Figure 4 A schematic diagram of a flow chart of a step of calculating a variance inflation factor in one embodiment;

[0098] Figure 5 It is a flowchart of the determination step according to the variance inflation factor in one embodiment;

[0099] Figure 6 A schematic diagram of a flow chart of a step of calculating a maximum condition index in one embodiment;

[0100] Figure 7 A schematic diagram of a flow chart of a determination step according to a maximum condition index in an embodiment;

[0101] Figure 8 It is a schematic flow chart of the ridge regression coefficient step of each influencing factor in one embodiment;

[0102] Fig. 9 It is a schematic diagram of a curve showing a change of the ridge regression coefficient of each influencing factor along with the ridge parameter in one embodiment;

[0103] Fig.10 It is a schematic diagram of a curve showing a change in the ridge regression coefficients of the influencing factors No. 1, 5, 12, 17, 22, and 23 along with the ridge parameters in one embodiment;

[0104] Fig.11 It is a schematic diagram of a curve showing a change in the ridge regression coefficients of the influencing factors No. 2, 4, 5, 6, and 7 along with the ridge parameters in one embodiment;

[0105] Fig. 12A It is a schematic diagram of the error frequency distribution of the spring data of the influencing factors before and after dimensionality reduction in one embodiment;

[0106] Fig. 12B A schematic diagram of error frequency distribution of summer data of influencing factors before and after dimensionality reduction in one embodiment;

[0107] Fig. 12C It is a schematic diagram of the error frequency distribution of autumn data of influencing factors before and after dimensionality reduction in one embodiment;

[0108] Fig.12D A schematic diagram of the error frequency distribution of winter data of influencing factors before and after dimensionality reduction in one embodiment;

[0109] Fig.13 It is a schematic diagram of comparison between predicted values ​​and actual values ​​of sample points of a part of the test set of influencing factors in autumn in one embodiment before and after dimensionality reduction;

[0110] Fig.14 is a structural block diagram of a data dimension reduction device in an embodiment;

[0111] Fig.15 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0112] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0113] In one embodiment, a data dimension reduction method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. The above-mentioned terminals can be, but are not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server can be implemented with an independent server or a server cluster composed of multiple servers. The main purpose of this embodiment is to overcome the shortcomings of existing dimensionality reduction methods. The ridge regression algorithm is used to reduce the dimensionality of the data to screen out multiple influencing factors that have a greater impact on wind power prediction. Figure 1 As shown, the data dimension reduction method includes the following steps:

[0114] Step 102: Acquire multiple influencing factors of wind power and influencing factor data.

[0115] The influencing factor data include parameter values ​​of various influencing factors at multiple times in the target season.

[0116] Specifically, the various influencing factors of wind power are influencing factors that affect the wind power prediction of the wind farm, for example, 35m temperature (temperature at a height of 35m at the wind farm location), momentum flux, 30m wind direction, 170m wind speed, and 100m wind speed, etc. The target season can be any one of the first season, the second season, the third season, and the fourth season. For example, March to May of each year is spring (the first season), June to August is summer (the second season), September to November is autumn (the third season), and December to February of the following year is winter (the fourth season). When the terminal performs data dimensionality reduction, it is necessary to obtain a variety of influencing factors that are related to (influenced by) the prediction of the wind power of the wind farm, as well as a sequence of parameter values ​​of the above-mentioned various influencing factors at multiple moments in the target season.

[0117] Optionally, after determining the wind farm, the terminal may determine 24 influencing factors related to the prediction of wind power according to the specific conditions of the wind farm. The specific types of influencing factors may be shown in the following Table 1:

[0118] Table 1

[0119]

[0120]

[0121] Step 104 , based on the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor.

[0122] Among them, multicollinearity diagnosis is to calculate the multicollinearity diagnostic parameters of various influencing factors. The multicollinearity diagnostic parameters are used to characterize the degree of linear correlation between different influencing factors. Multicollinearity refers to the linear correlation between different influencing factors.

[0123] Specifically, the terminal calculates the multicollinearity diagnostic parameters (values ​​of multiple indicators) of each influencing factor based on the influencing factor data of each influencing factor of the target season. The multicollinearity diagnostic parameters of the influencing factor indicate whether there is a multicollinearity relationship between the influencing factor and multiple other influencing factors.

[0124] Step 106, if it is determined that multicollinearity exists between the multiple influencing factors according to the multicollinearity diagnostic parameters of each influencing factor, the ridge regression coefficient of each influencing factor is calculated according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter.

[0125] Specifically, when the terminal determines that multicollinearity exists among multiple influencing factors based on the multicollinearity diagnostic parameters of each influencing factor, the terminal can calculate the ridge regression coefficient of each influencing factor based on the influencing factor data of each influencing factor in the target season, the preset influencing factor observation data and the ridge parameters.

[0126] Optionally, the ridge parameter k may be any value within a target range, and the target range may be, for example, 0≤k≤1. In this way, for each influencing factor, the terminal may calculate the ridge regression coefficient corresponding to the ridge parameter k according to different ridge parameters k, that is, obtain the corresponding relationship between the ridge regression coefficient of each influencing factor and the ridge parameter of the target range.

[0127] Step 108 , screening various influencing factors according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, eliminating the influencing factors whose ridge regression coefficients satisfy the preset ridge regression coefficient change conditions, and obtaining a plurality of target influencing factors.

[0128] Specifically, the preset ridge regression coefficient change condition may be that the ridge regression coefficient is stable and the absolute value of the ridge regression coefficient is less than a preset threshold; the preset ridge regression coefficient change condition may also be that the ridge regression coefficient is unstable and as the ridge parameter k increases, the ridge regression coefficient tends to 0. The terminal may screen among multiple influencing factors, remove influencing factors whose ridge regression coefficients meet the preset ridge regression coefficient change condition, retain influencing factors whose ridge regression coefficients do not meet the preset ridge regression coefficient change condition, and obtain multiple target influencing factors.

[0129] Optionally, the terminal can also screen among multiple target influencing factors. For example, the terminal can determine the target parameter value of the ridge parameter, and calculate the absolute value of the ridge regression coefficient of each influencing factor when the ridge parameter is the target parameter value. The terminal can extract the target influencing factors according to the absolute value of the ridge regression coefficient of each influencing factor, and take the target influencing factors with the largest absolute value as the updated target influencing factors. The target number can be five.

[0130] In one example, the terminal determines a first range of ridge parameters k (for example, k>0.4) that makes the ridge regression coefficients of each influencing factor tend to be stable based on the correspondence between the ridge regression coefficients of each influencing factor and the ridge parameters. The terminal can randomly determine an arbitrary value within the first range as the target parameter value of the ridge parameter (for example, it can be 0.8).

[0131] In the above data dimension reduction method, multiple influencing factors of wind power and influencing factor data of the target season are obtained. According to the influencing factor data of the target season, multicollinearity diagnosis is performed on multiple influencing factors to obtain multicollinearity diagnostic parameters of each influencing factor. If it is determined that there is multicollinearity between multiple influencing factors based on the multicollinearity diagnostic parameters of each influencing factor, the ridge regression coefficient of each influencing factor is calculated based on the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter. According to the ridge regression coefficient, each influencing factor is screened, and the influencing factors whose ridge regression coefficients meet the preset ridge regression coefficient change conditions are eliminated to obtain multiple target influencing factors. By adopting this method, multiple influencing factors can be screened by the ridge regression coefficients of each influencing factor, and data redundancy can be reduced while ensuring the application value of the data, so as to screen out the influencing factors that have a greater impact on the wind power prediction of the wind farm.

[0132] In one embodiment, after performing multicollinearity diagnosis on multiple influencing factors according to the influencing factor data of the target season to obtain the multicollinearity diagnosis parameter of each influencing factor, the data dimensionality reduction method further includes:

[0133] According to the multicollinearity diagnostic parameters of each influencing factor, it is determined whether there is multicollinearity among multiple influencing factors.

[0134] Specifically, the multicollinearity diagnostic parameters of each influencing factor may include one or more of the tolerance of the influencing factor, the variance inflation factor, and the conditional index. The terminal can judge whether multicollinearity exists or does not exist between multiple factors based on the parameter values ​​of each indicator included in the multicollinearity diagnostic parameters of each influencing factor.

[0135] In one embodiment, the multicollinearity diagnostic parameter of the influencing factor includes a parameter value of tolerance.

[0136] Accordingly, if Figure 2 As shown, the specific processing process of step 104 "performing multicollinearity diagnosis on multiple influencing factors according to the influencing factor data of the target season to obtain multicollinearity diagnosis parameters of each influencing factor" includes:

[0137] Step 202: for each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor.

[0138] Specifically, for each influencing factor, the terminal performs multivariate linear regression processing by a preset least squares method based on the first influencing factor data of the influencing factor in the target season and the second influencing factor data of multiple other influencing factors in the target season, and calculates the complex correlation coefficient R corresponding to the influencing factor. j .

[0139] Optionally, there may be n factors influencing the prediction of the wind power of the wind farm. Thus, R j is the jth influencing factor x j The multiple correlation coefficient with other n-1 influencing factors. The larger the multiple correlation coefficient, the closer the linear correlation between the above-mentioned influencing factors.

[0140] Step 204, calculating the tolerance according to the preset target value and the square of the complex correlation coefficient.

[0141] Specifically, for the jth influencing factor x j , the terminal can calculate the square of the complex correlation coefficient by the preset target value (eg, 1). The difference between j .

[0142] Optionally, the terminal can calculate the jth influencing factor x by the following formula: j Tolerance TOL j :

[0143]

[0144] Wherein, n represents the number of multiple influencing factors.

[0145] Accordingly, if Figure 3 As shown in FIG. 1 , the specific implementation process of the step “determining whether there is multicollinearity between multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor” includes:

[0146] Step 302: If the tolerance of each influencing factor is less than the first preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0147] Specifically, the first preset threshold value may be determined by the terminal according to an actual application scenario. The terminal compares the tolerance of the influencing factor with the first preset threshold value. If the tolerance of all influencing factors is less than the first preset threshold value, it can be determined that there is multicollinearity between the multiple influencing factors. If the tolerance of all influencing factors is greater than or equal to the first preset threshold value, it can be determined that there is no multicollinearity between the multiple influencing factors.

[0148] In one example, the larger the multiple correlation coefficient, the closer the linear correlation between the influencing factors is. The larger the TOL j The smaller. If the jth influencing factor x j If there is multicollinearity (serious collinearity) with all other influencing factors, then TOL j ≈0, otherwise, TOL j ≈1. TOL j The closer it is to 0, the more it indicates the jth influencing factor x j The closer the relationship with all other influencing factors, the more likely multicollinearity will occur.

[0149] In this embodiment, by calculating the tolerance of the influencing factors and comparing the tolerance with the first preset threshold, it is possible to accurately determine whether multicollinearity exists between the multiple influencing factors.

[0150] In one embodiment, the multicollinearity diagnostic parameters of the influencing factors include tolerance and variance inflation factor. Figure 4 As shown in FIG. 1 , the specific processing process of the step “performing multicollinearity diagnosis on multiple influencing factors according to the influencing factor data of the target season to obtain the multicollinearity diagnosis parameters of each influencing factor” includes:

[0151] Step 402, for each influencing factor, performing least squares calculation on the influencing factor data of the target season corresponding to the influencing factor to determine the complex correlation coefficient of the influencing factor;

[0152] Specifically, the specific execution process of this step is similar to the execution process of step 202, which will not be repeated here. Please refer to the description of the above step 202.

[0153] Step 404, calculating the tolerance according to the preset target value and the square of the complex correlation coefficient;

[0154] Specifically, the specific execution process of this step is similar to the execution process of step 204, which will not be repeated here. Please refer to the description of the above step 204.

[0155] Step 406, using the inverse of the tolerance as the variance inflation factor of the influencing factor;

[0156] Specifically, the terminal can calculate the jth influencing factor x by the following formula: j Variance inflation factor VIF j :

[0157] VIF j =1 / TOL j .

[0158] Accordingly, if Figure 5 As shown in FIG. 1 , the specific processing process of the step “determining whether there is multicollinearity among multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor” includes:

[0159] Step 502: If the variance inflation factor of each influencing factor is greater than a second preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0160] Specifically, the second preset threshold value can be determined by the terminal according to the actual application scenario. The terminal compares the variance inflation factor of the influencing factor with the second preset threshold value. If the variance inflation factors of all influencing factors are greater than or equal to the second preset threshold value, it can be determined that there is multicollinearity between the multiple influencing factors. If the variance inflation factors of all influencing factors are less than the second preset threshold value, it can be determined that there is no multicollinearity between the multiple influencing factors.

[0161] In one example, if the jth influencing factor x j There is no linear correlation between all other influencing factors (e.g., n-1 factors), then VIF j =1; if the jth influencing factor x j There is a linear correlation with all other influencing factors (e.g., n-1 factors), then TOL j <1, VIF j >1, and VIF j The larger the value, the more serious the collinearity between the influencing factors. i ≥10, it can be considered that there is serious collinearity among various influencing factors, and the second preset threshold can be 10.

[0162] In this embodiment, by calculating the tolerance of the influencing factors and the variance inflation factor, and comparing the tolerance with the first preset threshold, and comparing the variance inflation factor with the second preset threshold, it is possible to accurately determine whether multicollinearity exists between multiple influencing factors.

[0163] In one embodiment, the multicollinearity diagnostic parameter of the influencing factors includes a condition index.

[0164] Accordingly, if Figure 6 As shown in FIG. 1 , the specific processing process of the step “performing multicollinearity diagnosis on multiple influencing factors according to the influencing factor data of the target season to obtain the multicollinearity diagnosis parameters of each influencing factor” includes:

[0165] Step 602: Determine a first data matrix based on the influencing factor data of the target season.

[0166] The influencing factor data of the target season may include parameter values ​​of multiple influencing factors at multiple moments of the target season. The multiple influencing factors may be n influencing factors, for example, including: x1, x2, ..., x n . x1 can represent the parameter value of the first influencing factor at multiple moments in the target season.

[0167] Specifically, the terminal can determine the first data matrix X according to the parameter values ​​of various influencing factors at multiple moments in the target season, that is, X=[x1, x2, ..., x n ].

[0168] Step 604: determine a first measurement matrix according to the first data matrix and the transposed matrix of the first data matrix.

[0169] Specifically, the transposed matrix of the first data matrix can be X T , the terminal can convert the first data matrix X and the transposed matrix X of the first data matrix T The product of is the first observation matrix XX T .

[0170] Step 606, calculating at least one eigenvalue of the first measurement matrix.

[0171] Specifically, the terminal can calculate all the characteristic roots of the first measurement matrix according to a preset characteristic root calculation method, and the arrangement according to the size of each characteristic root can be λ1≥λ2≥…≥λ n .

[0172] Step 608, taking the ratio of the largest characteristic root to the smallest characteristic root as the largest condition index;

[0173] Specifically, the terminal can calculate the kth conditional index η of the first observation matrix by the following formula: k :

[0174] η k =λ1 / λ k , k=2,3,…,n.

[0175] Among them, λ1 can be the largest characteristic root, λ k It can be the kth largest characteristic root. In this way, the terminal can determine that the maximum condition index is the ratio between the largest characteristic root and the smallest characteristic root.

[0176] Accordingly, if Figure 7 As shown in FIG. 1 , the specific implementation process of the step “determining whether there is multicollinearity between multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor” includes:

[0177] Step 702: If the maximum condition index is greater than or equal to a third preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0178] Specifically, the third preset threshold value can be determined by the terminal according to the actual application scenario. The terminal compares the maximum condition index determined by the influencing factor data of the target season with the third preset threshold value. If the maximum condition index is less than the third preset threshold value, it can be determined that there is multicollinearity between the multiple influencing factors. If the maximum condition index is greater than or equal to the third preset threshold value, it can be determined that there is no multicollinearity between the multiple influencing factors.

[0179] In one example, the terminal may determine a fourth preset threshold and a fifth preset threshold. In this way, the terminal determines that if the largest condition index η n is 0≤η n ≤ the fourth preset threshold, then n influencing factors x1, x2, ..., x n There is no multicollinearity between them; if the maximum conditional index η n is the fourth preset threshold <η n ≤ the fifth preset threshold, then n influencing factors x1, x2, ..., x n There is a moderate degree of multicollinearity; if the largest conditional index η n is η n > The fifth preset threshold value, then n influencing factors x1, x2, ..., x n There is serious multicollinearity.

[0180] Optionally, the fourth preset threshold may be 100, and the fifth preset threshold may be 100. In this way, the terminal determines that if the maximum condition index ηn is 0≤η n ≤100, then there are n influencing factors x1, x2, ..., x n There is no multicollinearity between them; if the maximum conditional index η n is 100<η n ≤1000, then there are n influencing factors x1, x2, ..., x n There is a moderate degree of multicollinearity; if the largest conditional index η n is η n > 1000, then determine n influencing factors x1, x2, ..., x n There is serious multicollinearity.

[0181] In this embodiment, by calculating the maximum condition index of the influencing factors and comparing the maximum condition index with the third preset threshold, it is possible to accurately determine whether multicollinearity exists between the multiple influencing factors.

[0182] In one embodiment, Figure 8 As shown, the specific processing process of step 106 "calculating the ridge regression coefficient of each influencing factor according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter" includes:

[0183] Step 802: Determine a first data matrix according to the influencing factor data of the target season.

[0184] The influencing factor data of the target season may include parameter values ​​of multiple influencing factors at multiple moments of the target season. The multiple influencing factors may be n influencing factors, for example, including: x1, x2, ..., x n . x1 can represent the parameter value of the first influencing factor at multiple moments in the target season.

[0185] Specifically, the terminal can determine the first data matrix X according to the parameter values ​​of various influencing factors at multiple moments in the target season, that is, X=[x1, x2, ..., x n ].

[0186] Step 804: determine a first measurement matrix according to the first data matrix and the transposed matrix of the first data matrix.

[0187] Specifically, the transposed matrix of the first data matrix can be X T , the terminal can convert the first data matrix X and the transposed matrix X of the first data matrix T The product of is the first observation matrix XX T .

[0188] Step 806, determining a target matrix according to the first measurement matrix, the ridge parameter and the preset standard value.

[0189] Specifically, the ridge parameter may be k, and the preset standard value may be a value that can make the characteristic root of the matrix far away from zero and have a better prediction effect, for example, it may be the unit matrix I. In this way, the terminal may obtain the first product kI of the ridge parameter k and the preset standard value I, and obtain the first product kI and the first observation matrix XX. T The sum of is taken as the target matrix.

[0190] Step 808: Determine a second observation matrix according to the transposed matrix of the first data matrix and the preset influencing factor observation data.

[0191] Specifically, the preset influencing factor observation data may be Y, and the terminal may convert the transposed matrix X of the first data matrix into T The product of the preset influencing factor observation data Y is used as the second observation matrix X T Y.

[0192] In one example, Y is a set of values ​​of an observed variable at different times, and the observed variable can be calculated by the following formula:

[0193] Y=Xβ+ε,

[0194] Among them, X is the influencing factors, which is the independent variable; Y is a set of observations consisting of the power generation of the wind farm at multiple times, which is the dependent variable; β is a set of regression coefficients corresponding to the independent variables in the multivariate linear regression model, and ε is the random error term.

[0195] Step 810, obtaining the ridge regression coefficient of each influencing factor according to the target matrix and the second measurement matrix.

[0196] Specifically, the terminal obtains the ridge regression model matrix through the target matrix and the second observation matrix. In this way, the terminal can calculate the ridge regression coefficient of each influencing factor through the following formula (ridge regression model matrix):

[0197]

[0198] Among them, the data matrix X = [x1, x2, ..., x n ], the x that constitutes a column of the matrix X i (i=1, 2, ..., n) represents a set of values ​​of a certain influencing factor at different times, the ridge parameter k is a parameter varying between 0 and 1, and Y is a set of values ​​of the observed variable at different times. It is a set of ridge regression coefficients for n influencing factors under a certain parameter k.

[0199] In an example, for a multivariate linear regression model Y=Xβ+ε, Y is an observation vector, and the least squares method is used to obtain an estimated value of the vector β, which can be specifically calculated by the following formula:

[0200]

[0201]

[0202] Thus, if there is multicollinearity among the above-mentioned influencing factors, some characteristic roots of the matrix are close to zero, and X T X is singular. That is, when multiple influencing factors are linearly related, X T X is singular, and some of its characteristic roots are close to zero. The value of will be very large, resulting in a large deviation between the estimated value and the actual value of the ridge regression coefficient, causing the traditional least squares method to lose stability and reliability. In this case, the present invention re-determines the ridge regression model matrix through the method in the above embodiment, so that the characteristic root of the matrix is ​​far away from zero. The value of is reduced, and the estimated value of the vector β is obtained.

[0203] In this embodiment, by using X T X+kI replaces X T X, which makes the characteristic roots of the ridge regression model matrix far away from zero, and because X T X+kI is non-singular, so the calculated The error can be reduced to obtain more accurate ridge regression coefficients.

[0204] In one embodiment, the specific processing process of the step of "screening various influencing factors according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, eliminating the influencing factors whose ridge regression coefficients meet the preset ridge regression coefficient change conditions, and obtaining multiple target influencing factors" includes:

[0205] When the ridge parameter is within the preset range, the influencing factor whose ridge regression coefficient meets the preset stability condition and whose absolute value is less than the preset stability threshold is determined as the first influencing factor; when the ridge parameter is within the preset range, the influencing factor whose ridge regression coefficient does not meet the preset stability condition and meets the preset ridge regression coefficient change trend condition is determined as the second influencing factor; each influencing factor is screened, the first influencing factor and the second influencing factor are eliminated, and multiple target influencing factors are obtained.

[0206] Specifically, the terminal can calculate the corresponding relationship between the ridge regression coefficient and the ridge parameter of each influencing factor according to the ridge regression model matrix. The terminal can screen each influencing factor according to the corresponding relationship between the ridge regression coefficient and the ridge parameter of each influencing factor to obtain multiple target influencing factors. The preset stability condition can be that the rate of change of the ridge regression coefficient is less than the sixth preset threshold, and the sixth preset threshold can be ten percent of the maximum rate of change of the ridge regression coefficient. The preset ridge regression coefficient change trend condition can be that the ridge regression coefficient of the influencing factor tends to zero as the ridge parameter increases. The determination of the preset stability threshold can be determined based on the actual value of the preset standard value.

[0207] Optionally, the terminal can select each influencing factor based on the corresponding relationship between the ridge regression coefficient and the ridge parameter of each influencing factor, and convert the ridge regression coefficient Stable factors with relatively small absolute values ​​are eliminated; the ridge regression coefficient The unstable influencing factors that quickly tend to zero with the increase of the ridge parameter k are eliminated.

[0208] In one example, as the ridge parameter k increases, the final influencing factor x i The ridge regression coefficient Stable value b i , the stable value b i The influencing factor x can be expressed as i The influence coefficient on the observed variable Y. When the stable value b i is a positive number, indicating that the influencing factor x i It has a positive impact on the observed variables. i is a negative number, indicating that the influencing factor x i It has a reverse effect on the observed variable. i When the absolute value of is small, it indicates that the influencing factor x i It has little impact on Y and can be screened out, thus achieving dimensionality reduction.

[0209] In one example, the terminal may determine the growth step size of the ridge parameter (eg, 0.01), and the growth range of the ridge parameter may be 0.01 to 0.8. The terminal may calculate the ridge regression coefficients of different influencing factors. As the ridge parameter k changes, the terminal can calculate the change value of the ridge regression coefficient when k ranges from 0.01 to 0.8 based on the influencing factor data of each influencing factor in the target season. And the change values ​​of different influencing factors The absolute maximum value of For the change in ridge regression coefficient The absolute value of is less than that of the corresponding season 10% of the influencing factors, that is, the influencing factor xi Meet the preset stability conditions.

[0210] In one embodiment, the data dimension reduction method further includes:

[0211] The influencing factor data of the target season corresponding to the multiple target influencing factors are input into a pre-trained convolutional neural network to obtain the predicted wind power generation power of the target season.

[0212] Specifically, the terminal may divide the influencing factor data of the target season of the multiple target influencing factors into training data and test data according to time, and the training data also includes the actual power generation power of the wind farm corresponding to the target season.

[0213] In this embodiment, by predicting the power generation of the wind farm through the data of multiple target influencing factors with reduced dimensions, better training and prediction effects can be achieved, and more accurate power generation capacity evaluation can be realized.

[0214] The following describes the specific implementation process of the data dimensionality reduction method in conjunction with a detailed embodiment:

[0215] For example, the data dimension reduction method can be applied to a target wind farm in a target area, the wind farm capacity of the wind farm can be 75MW, the location can be xx°E, xx°N, and the collection interval of multiple influencing factors can be to collect power data of one point every 15 minutes. At the same longitude and latitude, the influencing factor data can be provided by the target wind tower.

[0216] For example, the influencing factor data of multiple influencing factors can be the data between 20:00 on January 1, xxx9 and 20:00 on December 30, xx10. Each wind farm has a corresponding nearest wind tower to provide numerical weather forecasts. There are 24 possible influencing factors of wind power in the numerical weather forecast. Take the first 20 days of data of each month as the training set, and the last 10 days of data as the test set. March to May of each year is spring, June to August is summer, September to November is autumn, and December to February of the following year is winter. Training and testing are carried out in different seasons, and the target season can be any of the four seasons of spring, summer, autumn and winter.

[0217] Step 1: The terminal can perform multivariate linear regression processing on the 24-dimensional possible influencing factors and wind power data (influencing factor data of multiple influencing factors) in the numerical weather forecast NWP data of the target wind farm by the least squares method, and calculate the R-square statistic. The result can be 0.5762, which can preliminarily determine that the regression equation has multicollinearity.

[0218] Step 2: Perform multicollinearity diagnosis on the 24 influencing factors (X1-X24) corresponding to the 24 dimensions. The tolerance and variance inflation factor of each influencing factor are shown in Table 2. The tolerance of a large number of influencing factors is less than or equal to 0.1, and the variance inflation factor is greater than or equal to 10, indicating that there is serious multicollinearity between these influencing factors and the remaining influencing factors. The maximum variance inflation factor is 2467.2, and the minimum tolerance is 0.0004053, which is much less than 0.1. There is a high degree of multicollinearity between the independent variables. The matrix X composed of the observation data T The maximum conditional index of X is 7.6806*1017, which is much larger than 100, indicating serious multicollinearity.

[0219] Table 2

[0220]

[0221]

[0222] Step 3, solve the multicollinearity problem between various influencing factors through ridge regression method, use ridge trace method to determine ridge parameter k, and finally obtain the independent variable after dimensionality reduction. The preset screening condition 1 can be: eliminate the ridge regression coefficient Relatively stable independent variable with relatively small absolute value; the preset screening condition 2 can be: ridge regression coefficient The independent variable is unstable but tends to zero rapidly with the increase of ridge parameter k. The curve of ridge regression coefficient of all kinds of influencing factors with the change of ridge parameter can be as follows Fig. 9 As shown in the curve graph, when the ridge parameter k is greater than 0.4, the ridge regression coefficients of various influencing factors tend to be basically stable. When the ridge parameter k is 0.8, the value of the ridge regression coefficient can represent the influence of different influencing factors on wind power prediction, and perform subsequent data dimensionality reduction.

[0223] According to the above preset screening conditions 1 and 2, the terminal can eliminate the influencing factors with relatively stable ridge regression coefficients and relatively small absolute values, and the influencing factors with unstable ridge regression coefficients but rapidly tending to zero with the increase of ridge parameters. The ridge parameter k increases from 0.01 to 0.8 with a step size of 0.01, and the ridge regression coefficients considering different factors are As k changes, the ridge regression coefficient of each season training set data is calculated when k changes from 0.01 to 0.8. And the change values ​​of different influencing factors The absolute maximum value of For the change in ridge regression coefficient The absolute value of is less than that of the corresponding season 10% of the influencing factors, that is, the influencing factor x iRelatively stable.

[0224] Specifically, the ridge regression coefficient Near zero, the size is almost constant, such as Fig.10 The curves of the influencing factor No. 12 and No. 22 are shown. Fig.10 Also shown are the curves of influencing factor No. 1, No. 5, No. 17, and No. 23.

[0225] Specifically, when k is 0.8, if the influencing factor x i The ridge regression coefficient The absolute value does not exceed 10% It is considered that the influencing factor meets the preset screening condition 1, and the spring training data of the corresponding station should exclude the influencing factors X3 (30m wind direction), X10 (100m wind direction), X11 (10m wind direction), X12 (10m sea level wind direction), X13 (sea level pressure), X14 (cloud cover), and X19 (ground pressure). Then the influencing factor x is filtered according to the preset screening condition 2. i Ridge regression coefficient at k = 0.8 The larger the absolute value of is, the greater its influence on the observed value of wind power is. In this patent, 5 influencing factors are retained in each season, so among the remaining influencing factors, select For the data of spring, autumn and winter, the five influencing factors X2 (momentum flux), X4 (170m wind speed), X5 (100m wind speed), X6 (30m wind speed) and X7 (10m wind speed) are retained for subsequent analysis. The curves of the above five influencing factors can be as follows: Fig.11 As shown; note that X7 (10m wind speed) and X8 (10m sea level wind speed) in the data are exactly the same, here we screen out the repeated influencing factor X8 and retain X7. For summer, the five influencing factors with the largest absolute values ​​of the ridge regression coefficients are X4 (170m wind speed), X5 (100m wind speed), X6 (30m wind speed), X16 (sensible heat flux), and X24 (2m relative humidity), which may be related to the impact of high temperatures in summer on the operation of wind turbines and the characteristics of the marine climate in the target area. The five influencing factors retained after dimensionality reduction of the target wind farm in the target area can be as shown in Table 3:

[0226] Table 3

[0227]

[0228]

[0229] , Step 4, select all normalized independent variables and independent variables selected by ridge regression, and perform CNN convolutional neural network prediction respectively. Among them, the first 20 days of data in each month in different seasons are taken as training sets, and the last 10 days of data are taken as test sets. Since the results of CNN have certain random fluctuations, each case is predicted 10 times. The root mean square RMSE between the predicted results and the actual values ​​is used as the error index to perform comparative analysis of the errors. The comparative analysis can be as shown in Table 4 and Table 5. It can be determined that the accuracy of wind power prediction by using the data after dimensionality reduction is greatly improved. Among them, the "Spring-Before Dimensionality Reduction" column represents the error of the wind farm selecting all 24 independent variables for prediction, and the "Spring-After Dimensionality Reduction" column represents the error of the wind farm selecting 5 independent variables after dimensionality reduction for prediction. The last column is the prediction error obtained by training with the data obtained after dimensionality reduction regardless of season using the training set of the whole year. The prediction error is greater than the result when it is divided by season, indicating that the data must be processed by season.

[0230] Table 4

[0231]

[0232]

[0233] Table 5

[0234]

[0235] The error frequency distribution diagrams of data before and after dimensionality reduction in different seasons are as follows: Fig. 12A , Fig. 12B , Fig. 12C , as shown in Figure 12D. After dimensionality reduction, the prediction effects of summer, autumn and winter data have been improved, among which the prediction effect of autumn has been improved most significantly. The predicted values ​​of sample points of some test sets in autumn before and after dimensionality reduction are taken out and compared with the actual values, as shown in Figure 12D. Fig.13 The predicted value after dimensionality reduction is represented by a dotted line. Compared with the predicted value before dimensionality reduction, it is closer to the actual value at most sampling points. It can be seen that the prediction after dimensionality reduction can indeed improve the prediction effect.

[0236] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0237] Based on the same inventive concept, the embodiment of the present application also provides a data dimensionality reduction device for implementing the data dimensionality reduction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more data dimensionality reduction device embodiments provided below can refer to the limitations of the data dimensionality reduction method above, and will not be repeated here.

[0238] In one embodiment, Fig.14 As shown, a data dimension reduction device 900 is provided, comprising:

[0239] The acquisition module 901 is used to acquire multiple influencing factors of wind power and influencing factor data, where the influencing factor data includes parameter values ​​of the multiple influencing factors at multiple moments in a target season.

[0240] The diagnosis module 902 is used to perform multicollinearity diagnosis on multiple influencing factors according to the influencing factor data of the target season, and obtain multicollinearity diagnosis parameters of each influencing factor.

[0241] The calculation module 903 is used to calculate the ridge regression coefficient of each influencing factor according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter if it is determined that there is multicollinearity between multiple influencing factors according to the multicollinearity diagnostic parameters of each influencing factor.

[0242] The screening module 904 is used to screen various influencing factors according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, eliminate the influencing factors whose ridge regression coefficients meet the preset ridge regression coefficient change conditions, and obtain multiple target influencing factors.

[0243] Each module in the above-mentioned data dimension reduction device 900 can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0244] In one embodiment, the device further comprises:

[0245] The judgment module is used to judge whether there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnosis parameter of each influencing factor.

[0246] In one of the embodiments, the multicollinearity diagnostic parameters of the influencing factors include tolerance;

[0247] The diagnostic module is specifically used for:

[0248] For each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor;

[0249] Calculating the tolerance according to a preset target value and the square of the complex correlation coefficient;

[0250] The judgment module is specifically used for:

[0251] If the tolerance of each of the influencing factors is less than the first preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0252] In one embodiment, the multicollinearity diagnostic parameters of the influencing factors include tolerance and variance inflation factor;

[0253] The diagnostic module is specifically used for:

[0254] For each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor;

[0255] Calculating the tolerance according to a preset target value and the square of the complex correlation coefficient;

[0256] The reciprocal of the tolerance is used as the variance inflation factor of the influencing factor;

[0257] The judgment module is specifically used for:

[0258] According to the variance inflation factor of each influencing factor, determine whether there is multicollinearity among the multiple influencing factors;

[0259] If the variance inflation factor of each influencing factor is greater than a second preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0260] In one embodiment, the multicollinearity diagnostic parameter of the influencing factors includes a condition index;

[0261] The diagnostic module is specifically used for:

[0262] Determining a first data matrix according to the influencing factor data of the target season;

[0263] Determine a first measurement matrix according to the first data matrix and a transposed matrix of the first data matrix;

[0264] Calculating at least one eigenvalue of the first measurement matrix;

[0265] The ratio of the largest characteristic root to the smallest characteristic root is taken as the maximum condition index;

[0266] The judgment module is specifically used for:

[0267] Judging whether there is multicollinearity among the multiple influencing factors according to the maximum condition index;

[0268] If the maximum condition index is greater than or equal to a third preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

[0269] In one embodiment, the computing module is specifically configured to:

[0270] Determining a first data matrix according to the influencing factor data of the target season;

[0271] Determine a first measurement matrix according to the first data matrix and a transposed matrix of the first data matrix;

[0272] Determining a target matrix according to the first measurement matrix, the ridge parameter and a preset standard value;

[0273] Determine a second observation matrix according to the transposed matrix of the first data matrix and the preset influencing factor observation data;

[0274] According to the target matrix and the second observation matrix, the ridge regression coefficient of each of the influencing factors is obtained.

[0275] In one embodiment, the screening module is specifically used for:

[0276] When the ridge parameter is within a preset range, determining the influencing factor whose ridge regression coefficient meets the preset stability condition and whose absolute value is less than the preset stability threshold as the first influencing factor;

[0277] When the ridge parameter is within a preset range, determining the influencing factor whose ridge regression coefficient does not meet the preset stability condition and meets the preset ridge regression coefficient change trend condition as the second influencing factor;

[0278] Each influencing factor is screened, the first influencing factor and the second influencing factor are eliminated, and a plurality of target influencing factors are obtained.

[0279] In one embodiment, the device further comprises:

[0280] The prediction module is used to input the influencing factor data of the target season corresponding to multiple target influencing factors into a pre-trained convolutional neural network to obtain the predicted wind power generation power of the target season.

[0281] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.15 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store relevant data of influencing factors. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a data dimensionality reduction method is implemented.

[0282] Those skilled in the art will understand that Fig.15 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0283] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0284] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0285] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0286] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0287] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited thereto.

[0288] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0289] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A data dimensionality reduction method, characterized in that: The method comprises: Acquire multiple influencing factors of wind power and influencing factor data, wherein the influencing factor data includes parameter values ​​of the multiple influencing factors at multiple times in a target season; According to the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor; If it is determined that there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnostic parameters of each of the influencing factors, the ridge regression coefficient of each of the influencing factors is calculated according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter; Screening each influencing factor according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, eliminating the influencing factors whose ridge regression coefficients do not meet the preset ridge regression coefficient change conditions, and obtaining multiple target influencing factors; The step of calculating the ridge regression coefficient of each influencing factor according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter comprises: Determine a first data matrix based on the influencing factor data of the target season; determine a first observation matrix based on the first data matrix and the transposed matrix of the first data matrix; determine a target matrix based on the first observation matrix, the ridge parameters and preset standard values; determine a second observation matrix based on the transposed matrix of the first data matrix and the preset influencing factor observation data; obtain the ridge regression coefficients of each of the influencing factors based on the target matrix and the second observation matrix.

2. The method according to claim 1, characterized in that After the step of performing multicollinearity diagnosis on the multiple influencing factors according to the influencing factor data of the target season to obtain multicollinearity diagnosis parameters of each influencing factor, the method further includes: According to the multicollinearity diagnostic parameter of each influencing factor, it is determined whether there is multicollinearity between the multiple influencing factors.

3. The method according to claim 2, characterized in that The multicollinearity diagnostic parameters of the influencing factors include tolerance; According to the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor, including: For each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor; Calculating the tolerance according to a preset target value and the square of the complex correlation coefficient; The step of judging whether there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor includes: If the tolerance of each of the influencing factors is less than the first preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

4. The method according to claim 2, characterized in that: The multicollinearity diagnostic parameters of the influencing factors include tolerance and variance inflation factor; According to the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor, including: For each influencing factor, the influencing factor data of the target season corresponding to the influencing factor are calculated by the least square method to determine the complex correlation coefficient of the influencing factor; Calculating the tolerance according to a preset target value and the square of the complex correlation coefficient; The reciprocal of the tolerance is used as the variance inflation factor of the influencing factor; The step of judging whether there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor includes: If the variance inflation factor of each influencing factor is greater than a second preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

5. The method according to claim 2, characterized in that: The multicollinearity diagnostic parameters of the influencing factors include condition index; According to the influencing factor data of the target season, multicollinearity diagnosis is performed on the multiple influencing factors to obtain multicollinearity diagnosis parameters of each influencing factor, including: Determining a first data matrix according to the influencing factor data of the target season; Determine a first measurement matrix according to the first data matrix and a transposed matrix of the first data matrix; Calculating at least one eigenvalue of the first measurement matrix; The ratio of the largest characteristic root to the smallest characteristic root is taken as the maximum condition index; The step of judging whether there is multicollinearity among the multiple influencing factors according to the multicollinearity diagnostic parameter of each influencing factor includes: If the maximum condition index is greater than or equal to a third preset threshold, it is determined that multicollinearity exists between the multiple influencing factors.

6. The method according to claim 1, characterized in that The influencing factors are screened according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, and the influencing factors whose ridge regression coefficients do not meet the preset ridge regression coefficient change conditions are eliminated to obtain multiple target influencing factors, including: When the ridge parameter is within a preset range, determining the influencing factor whose ridge regression coefficient meets the preset stability condition and whose absolute value is less than the preset stability threshold as the first influencing factor; When the ridge parameter is within a preset range, determining the influencing factor whose ridge regression coefficient does not meet the preset stability condition and meets the preset ridge regression coefficient change trend condition as the second influencing factor; Each influencing factor is screened, the first influencing factor and the second influencing factor are eliminated, and a plurality of target influencing factors are obtained.

7. The method according to claim 1, characterized in that The method further comprises: The influencing factor data of the target season corresponding to the multiple target influencing factors are input into a pre-trained convolutional neural network to obtain the predicted wind power generation power of the target season.

8. A data dimension reduction device, characterized in that: The device comprises: An acquisition module, used to acquire multiple influencing factors of wind power and influencing factor data, wherein the influencing factor data includes parameter values ​​of the multiple influencing factors at multiple moments in a target season; A diagnosis module, used for performing multicollinearity diagnosis on the multiple influencing factors according to the influencing factor data of the target season, and obtaining multicollinearity diagnosis parameters of each influencing factor; A calculation module, for calculating the ridge regression coefficient of each of the influencing factors according to the influencing factor data of the target season, the preset influencing factor observation data and the ridge parameter, if it is determined that there is multicollinearity between the multiple influencing factors according to the multicollinearity diagnostic parameters of the influencing factors; A screening module, used to screen various influencing factors according to the corresponding relationship between the ridge regression coefficient and the ridge parameter, eliminate the influencing factors whose ridge regression coefficient does not meet the preset ridge regression coefficient change condition, and obtain multiple target influencing factors; The calculation module is specifically used to determine a first data matrix based on the influencing factor data of the target season; determine a first observation matrix based on the first data matrix and the transposed matrix of the first data matrix; determine a target matrix based on the first observation matrix, the ridge parameters and preset standard values; determine a second observation matrix based on the transposed matrix of the first data matrix and the preset influencing factor observation data; and obtain the ridge regression coefficient of each of the influencing factors based on the target matrix and the second observation matrix.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images