Wind power plant power prediction method based on regularized polynomial regression
By adopting a regularized polynomial regression model in wind farm power prediction, combining K-fold cross-validation and grid search to tune hyperparameters, the problem of easy overfitting of existing wind power prediction methods is solved, and the prediction accuracy and model generalization ability are significantly improved.
Patent Information
- Application Number
- CN202510249322.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-04
- Publication Date
- 2025-06-24
AI Technical Summary
The existing wind power prediction methods are prone to overfitting, have high degrees of freedom, and are prone to overcapture noise while flexibly fitting the model, resulting in low accuracy in wind power prediction in wind farms.
The wind farm power prediction method based on regularization polynomial regression is adopted. By combining regularization with polynomial regression model, the model parameter size is limited, the model generalization ability is improved, and the hyperparameters are tuned through K-fold cross-validation combined with grid search to determine the best regularization parameter combination.
It significantly improves the accuracy of wind farm power prediction, avoids model overfitting, reduces model complexity, and improves the generalization ability of the model.
Smart Images
Figure CN120200217A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for predicting the power of a wind farm, and particularly to a method for predicting the power of a wind farm based on regularized polynomial regression. Background Art
[0002] Wind power, as a renewable and pollution-free energy source, works by converting energy. First, the kinetic energy of the wind drives the rotation of the fan blades to convert wind energy into mechanical energy, and then the mechanical energy of the wind wheel is transmitted to the generator through the transmission system to be converted into electrical energy.
[0003] For wind farm enterprises, during their operation, whether it is formulating a power generation plan, arranging maintenance and repair, or participating in the electricity market transaction, there are clear and strong demands for wind power prediction. Accurate wind power prediction can not only help enterprises understand the power generation capacity at different times in advance, scientifically formulate power generation plans, and ensure the balanced supply of electricity and the stability of the power grid; but also can judge the peak and trough periods of the wind farm's power generation load, reasonably arrange maintenance and repair work, improve the efficiency and quality of maintenance, and ensure the safe, reliable and efficient operation of the unit equipment; at the same time, in electricity trading, it can estimate the trend of wind power and price, scientifically and reasonably quote prices and quantities, improve the accuracy of trading decisions, avoid trading default risks, and enhance the competitiveness of enterprises.
[0004] However, the existing wind power prediction methods are prone to overfitting, have a high degree of freedom, and are likely to overly capture noise while flexibly fitting the model, resulting in low accuracy of wind power prediction for wind farms. Summary of the Invention
[0005] In order to solve the deficiencies of the above technologies, the present invention provides a method for predicting the power of a wind farm based on regularized polynomial regression.
[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for predicting the power of a wind farm based on regularized polynomial regression, which specifically includes the following steps:
[0007] Step S1: Collect the historical measured data of the wind farm to form a historical record data set;
[0008] Step S2: Preprocess the historical measured data of the wind farm in Step S1; the preprocessing includes filling in missing values, detecting and processing outliers, and normalizing the data;
[0009] Step S3: According to the specified random seed, divide the historical record data set into two parts, namely a training set and a test set, according to a certain ratio;
[0010] Step S4: Construct a multi-type polynomial regression model based on the training set; the multi-type polynomial regression model includes a polynomial regression model based on L1 regularization, a polynomial regression model based on L2 regularization, and a polynomial regression model based on a combined L1 and L2 regularization;
[0011] Step S5: Use K-fold cross-validation combined with grid search to tune the hyperparameters in the multi-type polynomial regression model, determine the best parameter combination in the multi-type polynomial regression model, that is, determine the optimal polynomial regression model based on L1 regularization, the optimal polynomial regression model based on L2 regularization, and the optimal polynomial regression model based on a combined L1 and L2 regularization;
[0012] Step S6: Use the test set as the input of the multi-type polynomial regression model, and respectively enter the optimal polynomial regression model based on L1 regularization, the optimal polynomial regression model based on L2 regularization, and the optimal polynomial regression model based on a combined L1 and L2 regularization for wind power prediction to obtain the predicted power values of the three models;
[0013] Step S7: Compare the predicted power values of the three models with the actual observed power values respectively; select the model with the smallest square root of the mean square difference between the predicted power value of the model and the actual observed power value, that is, the model with the highest prediction accuracy, as the wind farm power prediction model;
[0014] Step S8: Input the new feature data into the wind farm power prediction model in Step S7 to obtain the power value of the wind farm in the future period.
[0015] Further, in Step S2, the preprocessing of the historical measured data of the wind farm specifically includes the following steps:
[0016] Step S21: Delete the duplicate data records in the historical measured data of the wind farm;
[0017] Step S22: Fill in the missing data by means of forward filling, backward filling, and interpolation filling;
[0018] Step S23: Process the outliers in the wind power historical record data by means of segmented detection.
[0019] Further, the usage rules for filling in the missing data in Step S21 include:
[0020] First, if the wind speed data at a certain isolated time point is missing, then select the forward filling method for filling;
[0021] II. When the wind speed data for two consecutive time points are missing, the wind speed data at the previous time point is filled by forward filling, and the wind speed data at the subsequent time point is filled by backward filling;
[0022] III. When the wind speed data for multiple consecutive time points are missing, i.e., more than two time points of wind speed data are missing, the interpolation filling method is selected to fill the wind speed data.
[0023] Furthermore, the processing of outliers in step S23 specifically includes the following steps:
[0024] Step S23-1: Set the interval and segment the wind speed according to the equal interval;
[0025] Step S23-2: Adopt the mean standard deviation detection method to calculate the average value and standard deviation of the power data of the wind power historical record data within each segment;
[0026] The average value of the power data of the wind power historical record data, the calculation formula is:
[0027]
[0028] where n is the number of power data within this segment; x i is the i-th data value within this segment;
[0029] The standard deviation of the power data of the wind power historical record data, the calculation formula is:
[0030]
[0031] where n is the number of power data within this segment; x i is the i-th data value within this segment; is the average value of the power data within this segment;
[0032] Step S23-3: Set the threshold multiple. If the multiple range between the average value and the standard deviation is less than the threshold multiple, the power data within this segment is a normal value; if the multiple range between the average value and the standard deviation is not less than the threshold multiple, the power data within this segment is an outlier;
[0033] Step S23-4: Replace the outliers within this segment with the average value of the power data within this segment.
[0034] Furthermore, the formula of the polynomial regression model in step S4 is:
[0035] Y = β0 + β1x + β2x 2 +... + β p x p + ∈;
[0036] Among them, β0, β1, β2,..., β p are the parameters of the polynomial regression model; p is the degree of the polynomial; ∈ is the error term.
[0037] Furthermore, the parameters of the polynomial regression model are calculated based on the least squares method; the formula for calculating the parameters of the polynomial regression model is:
[0038]
[0039] where y i is the actual value of the i-th observed data; is the predicted value corresponding to the i-th observed data; n is the number of observed data.
[0040] Furthermore, the loss function based on L1 regularization in step S4 is:
[0041]
[0042] where λ1 is the parameter of L1 regularization, used to control the regularization strength; y i is the actual value of the i-th observed data; is the predicted value corresponding to the i-th observed data; n is the number of observed data; p is the degree of the polynomial; |β j | is the absolute value of the parameters of the polynomial regression model.
[0043] Furthermore, the loss function based on L2 regularization in step S4 is:
[0044]
[0045] where λ2 is the parameter of L2 regularization, used to control the regularization strength; y i is the actual value of the i-th observed data; is the predicted value corresponding to the i-th observed data; n is the number of observed data; p is the degree of the polynomial; β j 2 is the square of the parameters of the polynomial regression model.
[0046] Furthermore, the loss function based on the combined L1 and L2 regularization in step S4 is:
[0047]
[0048] where λ1 is the parameter of L1 regularization; λ2 is the parameter of L2 regularization; y i is the actual value of the i-th observed data; is the predicted value corresponding to the i-th observed data; n is the number of observed data; p is the degree of the polynomial; |βj | is the absolute value of the polynomial regression model parameter; β j 2 is the square of the polynomial regression model parameter.
[0049] Further, the formula for calculating the prediction accuracy in step S7 is:
[0050]
[0051] where N is the number of samples in the test set; y i is the actual observed power value; is the power value predicted by the model.
[0052] The present invention discloses a wind farm power prediction method based on regularized polynomial regression. By combining regularization with the polynomial regression model, it can not only limit the size of the model parameters and improve the generalization ability of the model, but also achieve a balance between the model complexity and performance, reduce the model complexity, avoid model overfitting, and significantly improve the accuracy of wind farm power prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] Figure 1 is the flowchart of Embodiment 1 of the present invention.
[0054] Figure 2 is the comparison chart of prediction results based on the polynomial regression model in Embodiment 2.
[0055] Figure 3 is the comparison chart of prediction results based on the polynomial regression model with L1 regularization in Embodiment 2.
[0056] Figure 4 is the comparison chart of prediction results based on the polynomial regression model with L2 regularization in Embodiment 2.
[0057] Figure 5 is the comparison chart of prediction results based on the polynomial regression model with combined L1 and L2 regularization in Embodiment 2. DETAILED DESCRIPTION OF THE INVENTION
[0058] The present invention will be further described in detail below with reference to the drawings and specific embodiments.
[0059] Embodiment 1:
[0060] As Figure 1 shown, the wind farm power prediction method based on regularized polynomial regression specifically includes the following steps:
[0061] Step S1: Collect the historical measured data of the wind farm, namely the historical wind power data, to form a historical record data set;
[0062] Step S2: Preprocess the historical measured data of the wind farm in Step S1, that is, clean and process the historical data, namely, data cleaning and feature engineering operations; The preprocessing includes filling missing values, detecting and processing outliers, and normalizing the data; The preprocessing specifically includes the following steps:
[0063] Step S21: Delete the duplicate data records in the historical measured data of the wind farm;
[0064] Step S22: Fill the missing data by means of forward filling, backward filling, and interpolation filling;
[0065] In this embodiment, the historical record data of the wind farm is time-series data with a time interval of 15 minutes; The usage rules for filling the missing data include that if the wind speed data of a certain isolated time point is missing, the forward filling method is selected for filling; If the wind speed data of two consecutive time points is missing, the wind speed data of the previous time point is filled by the forward filling method, and the wind speed data of the latter time point is filled by the backward filling method; If the wind speed data of multiple consecutive time points is missing, that is, the wind speed data of more than two time points is missing, the interpolation filling method is selected for filling the wind speed data;
[0066] Step S23: Process the outliers in the wind power historical record data by means of segmented detection; Specifically, it includes the following steps:
[0067] Step S23-1: In this embodiment, the wind speed is segmented at an equidistant interval of 0.5 m / s;
[0068] Step S23-2: Adopt the mean standard deviation detection method to calculate the mean and standard deviation of the power data of the wind power historical record data in each segment;
[0069] The formula for calculating the mean of the power data of the wind power historical record data is:
[0070]
[0071] where n is the number of power data in this segment; x i is the i-th data value in this segment;
[0072] The formula for calculating the standard deviation of the power data of the wind power historical record data is:
[0073]
[0074] where n is the number of power data in this segment; x i is the i-th data value in this segment; is the mean of the power data in this segment;
[0075] Step S23-3: Set the threshold multiple. If the multiple range between the average value and the standard deviation is less than the threshold multiple, the power data within this segment is normal; if the multiple range between the average value and the standard deviation is not less than the threshold multiple, the power data within this segment is an outlier.
[0076] Step S23-4: Replace the outliers within this segment with the average value of the power data within this segment.
[0077] Step S3: According to the specified random seed, divide the historical record data set into two parts, namely the training set and the test set, according to a certain proportion; the training set is the modeling sample, and the test set is the prediction sample.
[0078] The random seed is usually determined by the current computer state and is the initial value used when generating a pseudo-random number sequence. The random seed can ensure that the same random number sequence is obtained starting from the same seed.
[0079] Step S4: Based on the training set, utilize the regression analysis theory and combine with the regularization method to construct multi-type polynomial regression models; the multi-type polynomial regression models include the polynomial regression model based on L1 regularization, the polynomial regression model based on L2 regularization, and the polynomial regression model based on the combined regularization of L1 and L2.
[0080] The polynomial regression model is an extension of the linear regression model. By adding high-order terms of the original features, a non-linear model is constructed to fit data relationships of different complexities; the formula of the polynomial regression model in Step S4 is:
[0081] Y = β0 + β1x + β2x 2 +... + β p x p + ∈;
[0082] Among them, β0, β1, β2,..., β p are the parameters of the polynomial regression model; p is the degree of the polynomial; ∈ is the error term.
[0083] Among them, the parameters of the polynomial regression model are calculated based on the least squares method, that is, find a set of parameter values to minimize the sum of the squares of the errors between the observed values and the predicted values, that is, minimize the loss function; the formula for calculating the parameters of the polynomial regression model is:
[0084]
[0085] Among them, y i is the actual value of the i-th observed data; is the predicted value corresponding to the i-th observed data; n is the number of observed data.
[0086] Regularization refers to a technique in machine learning and statistical modeling. It is an important technical means in machine learning and statistics to prevent model overfitting and improve the generalization ability of the model.
[0087] L1 regularization, also known as lasso regression, adds the sum of the absolute values of the model parameters as a regularization term to the original loss function; L1 regularization tends to produce a sparse weight matrix, that is, pushing some weights to zero, thus achieving the effect of feature selection; The loss function based on L1 regularization is:
[0088]
[0089] where λ1 is the parameter of L1 regularization, used to control the regularization strength; y i is the actual value of the i-th observed data; is the predicted value corresponding to the i-th observed data; n is the number of observed data; p is the degree of the polynomial; |β j | is the absolute value of the polynomial regression model parameters.
[0090] L2 regularization, also known as ridge regression, adds the sum of the squares of the model parameters as a regularization term to the original loss function; L2 regularization makes the weight values smaller, does not directly lead to weight sparsity, and can effectively control the model complexity; The loss function based on L2 regularization is:
[0091]
[0092] where λ2 is the parameter of L2 regularization, used to control the regularization strength; y i is the actual value of the i-th observed data; is the predicted value corresponding to the i-th observed data; n is the number of observed data; p is the degree of the polynomial; β j 2 is the square of the polynomial regression model parameters.
[0093] L1 and L2 combined regularization, also known as elastic net regression, adds L1 regularization and L2 regularization to the original loss function at the same time; The loss function based on L1 and L2 combined regularization is:
[0094]
[0095] where λ1 is the parameter of L1 regularization; λ2 is the parameter of L2 regularization; y i is the actual value of the i-th observed data; is the predicted value corresponding to the i-th observed data; n is the number of observed data; p is the degree of the polynomial; |β j | is the absolute value of the polynomial regression model parameters; β j 2It is the square of the parameters of the polynomial regression model.
[0096] Step S5: Use K-fold cross-validation combined with grid search to tune the hyperparameters in the multi-type polynomial regression model, that is, evaluate and tune the multi-type polynomial regression model, and determine the best parameter combination in the multi-type polynomial regression model, that is, determine the optimal polynomial regression model based on L1 regularization, the optimal polynomial regression model based on L2 regularization, and the optimal polynomial regression model based on the combined regularization of L1 and L2; In this embodiment, the five-fold cross-validation method is used in combination with grid search to tune the hyperparameters in the polynomial regression model;
[0097] Using K-fold cross-validation specifically includes the following steps:
[0098] Step S51: Randomly divide the historical record data set into K equal proportions;
[0099] Step S52: Take one of them as the validation data, and the remaining K - 1 parts as the training data;
[0100] Step S53: Obtain the training model and obtain the evaluation score;
[0101] Step S54: Repeat steps S51 - S53 K times to obtain K training models and the evaluation scores corresponding to each training model;
[0102] Step S55: Evaluate the performance of the multi-type polynomial regression model according to the average value of the K evaluation scores.
[0103] Grid search is to traverse the pre-given parameter combinations, train and evaluate the models corresponding to each group of parameters, and select the parameter combination with the best performance from them.
[0104] Step S6: Use the test set as the input of the multi-type polynomial regression model, and respectively enter the optimal polynomial regression model based on L1 regularization, the optimal polynomial regression model based on L2 regularization, and the optimal polynomial regression model based on the combined regularization of L1 and L2 for wind power prediction to obtain three model prediction power values;
[0105] Step S7: Compare the three model prediction power values with the actual observed power values respectively; Select the model with the smallest square root of the mean square difference between the model prediction power value and the actual observed power value, that is, the model with the highest prediction accuracy, as the wind farm power prediction model; The formula for calculating the prediction accuracy is:
[0106]
[0107] Among them, N is the number of samples in the test set; y i is the actual observed power value; is the predicted power value of the model;
[0108] Step S8: Use the wind farm power prediction model for prediction. Input the new feature data into the wind farm power prediction model in Step S7 to obtain the power value for the future period of the wind farm.
[0109] Embodiment 2:
[0110] This embodiment provides the wind power prediction process of a wind farm with an installed capacity of 100 MW in North China, which specifically includes the following steps:
[0111] Step 1: Collect the historical data of the wind farm from April 2024 to June 2024; the historical data includes information such as the basic information of the wind farm, measured meteorological parameters, and actual output power, etc.;
[0112] Step 2: Set the interval duration of the historical data to 15 minutes and preprocess the historical data; among them, a total of 8376 historical data records are used as sample data to establish the power prediction model of the wind farm;
[0113] Step 3: Based on the sample data, use the polynomial regression model, the polynomial regression model based on L1 regularization, the polynomial regression model based on L2 regularization, and the polynomial regression model based on the combined regularization of L1 and L2 respectively for data fitting to obtain the fitting result data; compare the fitting result data and select the best model to predict the output power of the wind farm; the fitting result data is shown in Table 1 as follows: Table 1 is as follows:
[0114] Table 1
[0115] Model type Mean absolute error Root mean square error Coefficient of determination Polynomial regression model 2.1078 2.8402 0.9755 Polynomial regression model based on L1 regularization 1.9752 2.7880 0.9805 Polynomial regression model based on L2 regularization 1.9467 2.7514 0.9817 Polynomial regression model based on combined L1 and L2 regularization 1.9133 2.6889 0.9833
[0116] As shown in Table 1, using the regularized polynomial regression model, the root mean square error is smaller and the accuracy is higher than that of the polynomial regression model without regularization; among them, the polynomial regression model based on the combined regularization of L1 and L2 has the highest accuracy, the root mean square error is reduced from 2.8402 to 2.6889, a decrease of about 5%, and the coefficient of determination is increased from 0.9755 to 0.9833;
[0117] As Figures 2 - 5 shown, the test sample numbers range from 0 to 350, and their true values are arranged in ascending order, and the power of the wind power is from 0 to 80 MW; Figure 2It is a comparison graph between the prediction of the polynomial regression model and the true value. The solid line represents the true value, and the dashed line represents the predicted value. The value predicted based on the polynomial regression model differs significantly from the true value for a single test sample. Among the samples numbered from 0 to 50, most of the predicted values are higher than the true value. Among the samples numbered from 90 to 100, most of the predicted values are higher than the true value. Among the samples numbered from 260 to 290, most of the predicted values are lower than the true value; Figure 3 It is a comparison graph between the prediction of the polynomial regression model based on L1 regularization and the true value. The solid line represents the true value, and the dashed line represents the predicted value. Its predicted value converges to the true value compared with the predicted value in Figure 2 Among the samples numbered from 130 to 150, the predicted value is higher than the true value. Among the samples numbered from 190 to 200, most of the predicted values are higher than the true value, and the difference between the predicted value and the true value is relatively large; Figure 4 It is a comparison graph between the prediction of the polynomial regression model based on L2 regularization and the true value. The solid line represents the true value, and the dashed line represents the predicted value. Its predicted value converges further to the true value compared with the predicted value in Figure 3 Among the samples numbered from 190 to 200, most of the predicted values are higher than the true value, and there is still a relatively large difference between the predicted value and the true value; Figure 5 It is a comparison graph between the prediction of the polynomial regression model based on the combined L1 and L2 regularization and the true value. The solid line represents the true value, and the dashed line represents the predicted value. Its predicted value compared with Figure 2 、 Figure 3 and Figure 4 has the best convergence effect and is close to the true value in the test values of each test sample;
[0118] Step 4: In this embodiment, the polynomial regression model based on the combined L1 and L2 regularization is used to predict the output power of this wind farm, which significantly improves the accuracy of the wind farm power prediction.
[0119] By combining regularization with the polynomial regression model, the size of the model parameters is restricted, and the generalization ability of the model is improved. At the same time, it can achieve a balance between the model complexity and performance, reduce the model complexity, avoid model overfitting, and significantly improve the accuracy of the wind farm power prediction.
[0120] The above embodiments are not limitations to the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions, or substitutions made by those skilled in the art within the scope of the technical solution of the present invention also belong to the protection scope of the present invention.
Claims
1. A wind farm power prediction method based on regularized polynomial regression, characterized in that: The specific steps include: Step S1: Collect historical measured data of the wind farm to form a historical record data set; Step S2: preprocessing the historical measured data of the wind farm in step S1; the preprocessing includes filling missing values, detecting and processing abnormal values, and normalizing the data; Step S3: according to the specified random seed, the historical record data set is divided into two parts, namely, a training set and a test set, in a certain ratio; Step S4: constructing a multi-type polynomial regression model based on the training set; the multi-type polynomial regression model includes a polynomial regression model based on L1 regularization, a polynomial regression model based on L2 regularization, and a polynomial regression model based on L1 and L2 combined regularization; Step S5: using K-fold cross validation combined with grid search to tune the hyperparameters in the multi-type polynomial regression model, and determining the best parameter combination in the multi-type polynomial regression model, that is, determining the best polynomial regression model based on L1 regularization, the best polynomial regression model based on L2 regularization, and the best polynomial regression model based on L1 and L2 combined regularization; Step S6: using the test set as the input of the multi-type polynomial regression model, respectively entering the optimal polynomial regression model based on L1 regularization, the optimal polynomial regression model based on L2 regularization, and the optimal polynomial regression model based on L1 and L2 combined regularization to perform wind power prediction, and obtaining power values predicted by three models; Step S7: Compare the power values predicted by the three models with the actual observed power values respectively; select the model with the smallest square root of the square mean of the difference between the model predicted power value and the actual observed power value, i.e., the model with the highest prediction accuracy, as the wind farm power prediction model; Step S8: input the new characteristic data into the wind farm power prediction model of step S7 to calculate the power value of the wind farm in the future period.
2. The wind farm power prediction method based on regularized polynomial regression according to claim 1, characterized in that: The step S2 pre-processes the historical measured data of the wind farm, specifically including the following steps: Step S21: deleting duplicate data records in the historical measured data of the wind farm; Step S22: fill the missing data by forward filling, backward filling and interpolation filling; Step S23: using a segmented detection method to process abnormal values in the wind power historical record data.
3. The wind farm power prediction method based on regularized polynomial regression according to claim 2, characterized in that: The usage rules for filling the missing data in step S21 include:
1. If the wind speed data for a certain isolated time point is missing, fill it in by forward filling; 2. If the wind speed data for two consecutive time points are missing, the wind speed data at the previous time point is filled in by forward filling, and the wind speed data at the next time point is filled in by backward filling; 3. If the wind speed data for multiple consecutive time points are missing, that is, the wind speed data for more than two time points are missing, the interpolation filling method is selected to fill the wind speed data.
4. The wind farm power prediction method based on regularized polynomial regression according to claim 2, characterized in that: The abnormal value is processed in step S23, which specifically includes the following steps: Step S23-1: setting intervals, dividing the wind speed into segments according to equidistant intervals; Step S23-2: using a mean standard deviation detection method to calculate the mean value and standard deviation of the power data of the wind power historical record data in each section; The average value of the power data of the wind power historical record data is calculated as follows: Where n is the number of power data in the segment; x i is the i-th data value in the segment; The standard deviation of the power data of wind power historical record data is calculated as follows: Where n is the number of power data in the segment; x i is the i-th data value in the segment; is the average value of the power data in this segment; Step S23-3: setting a threshold multiple. If the multiple range between the mean value and the standard deviation is less than the threshold multiple, the power data in the segment is a normal value; if the multiple range between the mean value and the standard deviation is not less than the threshold multiple, the power data in the segment is an abnormal value. Step S23 - 4 : Replace the abnormal value in the segment with the average value of the power data in the segment.
5. The wind farm power prediction method based on regularized polynomial regression according to claim 1, characterized in that: The formula of the polynomial regression model in step S4 is: Y=β0+β1x+β2x 2 +...+b p x p +∈; Among them, β0,β1,β2,...,β p is the parameter of the polynomial regression model; p is the degree of the polynomial; ∈ is the error term.
6. The wind farm power prediction method based on regularized polynomial regression according to claim 5, characterized in that: The parameters of the polynomial regression model are calculated based on the least squares method; the formula for calculating the parameters of the polynomial regression model is: Among them, y i is the actual value of the i-th observation; is the predicted value corresponding to the i-th observation data; n is the number of observation data.
7. The wind farm power prediction method based on regularized polynomial regression according to claim 6, characterized in that: The loss function based on L1 regularization in step S4 is: Among them, λ1 is the parameter of L1 regularization, which is used to control the regularization strength; y i is the actual value of the i-th observation; is the predicted value corresponding to the i-th observation data; n is the number of observation data; p is the degree of the polynomial; |β j | is the absolute value of the polynomial regression model parameter.
8. The wind farm power prediction method based on regularized polynomial regression according to claim 7, characterized in that: The loss function based on L2 regularization in step S4 is: Among them, λ2 is the parameter of L2 regularization, which is used to control the regularization strength; y i is the actual value of the i-th observation; is the predicted value corresponding to the i-th observation data; n is the number of observation data; p is the degree of the polynomial; β j 2 is the square of the polynomial regression model parameter.
9. The wind farm power prediction method based on regularized polynomial regression according to claim 8, characterized in that: The loss function based on L1 and L2 combined regularization in step S4 is: Among them, λ1 is the parameter of L1 regularization; λ2 is the parameter of L2 regularization; y i is the actual value of the i-th observation; is the predicted value corresponding to the i-th observation data; n is the number of observation data; p is the degree of the polynomial; |β j | is the absolute value of the polynomial regression model parameter; β j 2 is the square of the polynomial regression model parameter.
10. The wind farm power prediction method based on regularized polynomial regression according to claim 1, characterized in that: The formula for calculating the prediction accuracy in step S7 is: Where N is the number of samples in the test set; y i is the actual observed power value; Predict power values for the model.