Construction method and system of wind power generation power prediction model, equipment and medium

Abstract

The invention relates to the technical field of wind power generation, in particular to a construction method and system of a wind power generation power prediction model, equipment and a medium, and the method mainly comprises the steps that a training set and a test set are divided; the method comprises the steps of establishing a wind power generation power prediction model, outputting a predicted power increasing or decreasing result through the wind power generation power prediction model, establishing an evaluation model, outputting evaluation values through symbol identification and positioning data based on the evaluation model, sorting the evaluation values and outputting a sorting result. According to the method, the prediction result is further analyzed, the area which the whole wind power plant needs to pay attention to is reflected through the score, related maintenance work is carried out on the area with the high evaluation value score, the influence on the future power generation efficiency of the area is avoided, and the better power generation effect is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation, and in particular, to a method and system, device, and medium for constructing a wind power generation power prediction model. Background Art

[0002] Energy has always played an indispensable role in the development process of mankind, and its importance to human development is self-evident. Since the industrial revolution, the demand for energy required by mankind has shown a soaring growth trend. Not only coal and oil, but also natural gas are all the energy we need in daily life, and their usage has been in an increasing mode. Wind energy, because of its wide acquisition range, rich reserves, and no pollution, has been listed as one of the first-choice objects for development and utilization by countries around the world.

[0003] At present, there are many prediction models for wind power generation power, all of which can predict wind power to a certain extent. However, for the predicted results, there is no more processing and analysis, resulting in no relatively correct next operation direction being given to the staff. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system, device, and medium for constructing a wind power generation power prediction model to solve the above technical problems in the prior art.

[0005] The present invention is achieved through the following technical solutions:

[0006] In a first aspect, the present invention provides a method for constructing a wind power generation power prediction model, including:

[0007] Obtain historical data of wind power generation, preprocess the historical data, and divide the preprocessed data into a training set and a test set;

[0008] Establish a wind power generation power prediction model, and output the predicted result of power increase or decrease through the wind power generation power prediction model;

[0009] Obtain the positioning data of the wind turbines with increased and decreased power generation in the current wind farm respectively, and perform different symbol markings on the wind turbines with increased and decreased power respectively, and map them into a three-dimensional or two-dimensional data model according to the positioning data and symbol markings;

[0010] Divide the wind farm into several regions, obtain the symbol identifications and positioning data of the wind turbines in the same region, establish an evaluation model, output an evaluation value through the symbol identifications and positioning data based on the evaluation model, and output a sorting result after sorting the evaluation values.

[0011] Preferably, the preprocessing of the historical data includes:

[0012] Obtain the first historical data where the wind power output is less than 0, the second historical data where the wind speed is less than the starting wind speed power value and is not zero, the third historical data where the power value exceeds the rated power of the wind turbine, and the fourth historical data that is not continuously jumping;

[0013] And modify or replace the first historical data, the second historical data, the third historical data, and the fourth historical data respectively;

[0014] Perform feature selection on the historical data after data cleaning, perform data normalization on the historical data after feature selection, and integrate them into a data set.

[0015] Preferably, the modification or replacement includes:

[0016] Modify both the first historical data and the second historical data to 0, modify the third historical data to the rated power of the wind turbine, obtain the target historical data adjacent to the fourth historical data, and calculate the average value of the target historical data, and replace the fourth historical data with the average value of the target historical data.

[0017] Preferably, the performing feature selection on the historical data after data cleaning includes:

[0018] Calculate the independence index between the input feature and the output feature, and set an independence index threshold. When the independence index is greater than the independence index threshold, discard the feature. When the independence index is not greater than the independence index threshold, retain the feature;

[0019]

[0020] In the formula, χ is the independence index, k is the number of feature values, n is the total frequency, p i is the theoretical frequency of the i-th feature value, A i is the i-th feature value.

[0021] Preferably, the data normalization includes:

[0022]

[0023] In the formula, y is the value after normalization, x is the original data, x min is the minimum value, x max is the maximum value.

[0024] Preferably, the establishing of the wind power generation prediction model includes:

[0025] Establish an initial objective function, and perform a Taylor expansion on the objective function to obtain the final objective function:

[0026]

[0027] In the formula, Obj (s) is the objective function value, N is the number of samples, g i is the first-order partial derivative of the loss function, f s (x i ) is the base learner of the s-th, h i is the second-order partial derivative of the loss function, γ is the penalty coefficient of the regularization term for each leaf node, T is the maximum depth of the tree, λ is the regularization coefficient, ω j is the weight on the j-th leaf, L is a constant, y i is the sample label, is the training result of the (s - 1)-th round of the model;

[0028] When the said Obj (s) ≤0.5, the result of the output power reduction is output. When the said Obj (s) >0.5, the result of the output power increase is output.

[0029] Preferably, the establishment of the evaluation model includes:

[0030]

[0031] In the formula, U is the evaluation value, E is the number of wind turbines with predicted power increase in the area, G is the number of wind turbines in the area, is the objective function value of the l-th wind turbine, x r , y r and z r are the three-dimensional coordinates of the central wind turbine closest to the center of the area respectively, and x p , y p and z p are the three-dimensional coordinates of the remaining wind turbines in the area except the central wind turbine.

[0032] In the second aspect, the present invention also provides a construction system for a wind power generation power prediction model, including:

[0033] A prediction model establishment module, configured to obtain historical data of wind power generation, preprocess the historical data, divide the preprocessed data into a training set and a test set; establish a wind power generation power prediction model, and output the predicted result of power increase or decrease through the wind power generation power prediction model;

[0034] An analysis module, configured to respectively obtain the positioning data of the wind turbines with increasing and decreasing power generation in the current wind farm, respectively perform different symbol markings on the increasing and decreasing wind turbines, and map them into a three-dimensional or two-dimensional data model according to the positioning data and symbol markings; divide the wind farm into several regions, obtain the symbol identifiers and positioning data of the wind turbines in the same region, establish an evaluation model, output an evaluation value based on the evaluation model through the symbol identifiers and positioning data, and output a sorting result after sorting the evaluation values;

[0035] A main control module, connected to the prediction model establishment module and the analysis module, is used to execute the above-mentioned method for constructing a wind power generation power prediction model.

[0036] In a third aspect, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-mentioned method for constructing a wind power generation power prediction model is implemented.

[0037] In a fourth aspect, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned method for constructing a wind power generation power prediction model is implemented.

[0038] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0039] Adopting the method provided by the present invention mainly includes dividing the training set and the test set; establishing a wind power generation power prediction model, outputting the predicted result of power increase or decrease through the wind power generation power prediction model, establishing an evaluation model, outputting an evaluation value based on the evaluation model through the symbol identifier and positioning data, and outputting a sorting result after sorting the evaluation values. Through the above method, the predicted result is further analyzed, and the areas that need to be focused on in the entire wind farm are reflected by the scores. Relevant maintenance work is done on the areas with higher evaluation value scores to avoid affecting the future power generation efficiency of these areas, so as to achieve a better power generation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 It is a flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention described and illustrated in the drawings here can be arranged and designed in various different configurations.

[0043] The division of modules in this application is a logical division. In actual implementation, there may be other division methods. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0044] The independently described modules or sub-modules can be physically separated or not: they can be implemented in software or in hardware, and some of the modules or sub-modules can be implemented in software, and the functions of these modules or sub-modules are called by the processor to implement the software, and other parts of the modules or sub-modules are implemented in hardware, for example, through a hardware circuit. In addition, some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this application.

[0045] Please refer to Figure 1 , a method for constructing a wind power generation power prediction model, including:

[0046] S101: Obtain historical data of wind power generation, preprocess the historical data, and divide the preprocessed data into a training set and a test set;

[0047] Among them, the characteristic parameters of the fan data include wind speed, air pressure, wind direction, humidity, temperature, the angle of the blade strain pitch, the blade pitch angle, the generator. Rotation speed, wind turbine speed, wind turbine speed difference, nacelle temperature, nacelle direction, power.

[0048] S102: Establish a wind power generation power prediction model, and output the predicted result of power increase or decrease through the wind power generation power prediction model;

[0049] S103: Obtain the positioning data of the fans with increased and decreased power in the current wind farm respectively, mark the fans with increased and decreased power with different symbols respectively, and map them into a three-dimensional or two-dimensional data model according to the positioning data and symbol marks;

[0050] S104: Divide the wind farm into several regions, obtain the symbol identifications and positioning data of the fans in the same region, establish an evaluation model, output an evaluation value through the symbol identifications and positioning data based on the evaluation model, and output a sorting result after sorting the evaluation values.

[0051] The method provided by the present invention mainly includes dividing the training set and the test set; establishing a wind power prediction model, outputting the result of the predicted power increase or decrease through the wind power prediction model, establishing an evaluation model, outputting an evaluation value based on the evaluation model through symbol identification and positioning data, sorting the evaluation values and outputting the sorting result. Through the above method, the predicted result is further analyzed, and the area that needs to be focused on in the whole wind farm is reflected by the score. Maintenance work is carried out on the areas with higher evaluation value scores to avoid affecting the future power generation efficiency of these areas, so as to achieve a better power generation effect.

[0052] In an exemplary embodiment of the present invention, the actual operation data of wind turbines usually contains abnormal or missing data. These abnormal data mainly come from sensor acquisition errors, NWP system forecast errors, random fluctuations of wind, curtailment measures, and abnormal situations such as equipment failures and shutdowns for maintenance. When establishing a short-term output power prediction model, these outliers are mixed in the input data, and the performance of the prediction model will be greatly interfered. Therefore, it is necessary to clean the original wind power data.

[0053] The preprocessing of the historical data includes:

[0054] Obtaining target data includes the first historical data with wind power output less than 0, the second historical data with wind speed less than the starting wind speed power value and not zero, the third historical data with power value exceeding the rated power of the fan, and the fourth historical data with discontinuous jumps;

[0055] And respectively modifying or replacing the first historical data, the second historical data, the third historical data, and the fourth historical data;

[0056] The historical data after data cleaning is subjected to feature selection, the historical data after feature selection is normalized, and integrated into a data set.

[0057] Specifically, the modification or replacement includes:

[0058] Both the first historical data and the second historical data are modified to 0, the third historical data is modified to the rated power of the fan, the target historical data adjacent to the fourth historical data is obtained, and the average value of the target historical data is calculated, and the average value of the target historical data is used to replace the fourth historical data.

[0059] When the output power of wind power generation is less than 0, it means that there is a power absorption phenomenon during the operation of the wind turbine. The reason is that the wind speed has not reached the cut-in wind speed when the wind turbine is working, so the wind turbine does not generate electricity. At the same time, the control system inside the wind turbine needs to absorb energy to maintain normal operation, so the output power will be less than 0. The processing method for this type of abnormal data is to modify the power value less than 0 to 0. For the case where the output power is not 0 when the wind speed is below the starting wind speed of the wind turbine, the processing method is the same as the first type, which is also to change it to 0. For the case where the power exceeds the rated power of the wind turbine, the processing measure is to modify its value to the rated power of the wind turbine. For the case where the output power has a discontinuous jump due to a sudden change in wind speed, the processing method is to replace its value with the average value of the output powers of its adjacent points.

[0060] In an exemplary embodiment of the present invention, during the process of establishing a wind power prediction model, on the one hand, it is required that the input features of the prediction model can contain as much feature information closely related to power as possible, and the dimension of the input features is as small as possible. On the other hand, it is necessary to avoid adding redundant feature information to the input features. Therefore, feature screening is a method to solve the problem of too many dimensions of input features. An optimal feature selection method can lay a solid foundation for improving the accuracy of the prediction model later. We can also analyze the internal laws and correlations between data variables from efficient feature selection means, which plays an indispensable role in modifying and improving the prediction algorithm later.

[0061] The feature selection of the historical data after data cleaning includes:

[0062] Calculate the independence index between the input feature and the output feature, and set an independence index threshold. When the independence index is greater than the independence index threshold, the feature is discarded. When the independence index is not greater than the independence index threshold, the feature is retained;

[0063]

[0064] In the formula, χ is the independence index, k is the number of eigenvalue, n is the total frequency, p i is the theoretical frequency of the i-th eigenvalue, A i is the i-th eigenvalue.

[0065] The chi-square test can be used to examine the independence between input features and output features by us. If the internal relationship between the two is very close and the correlation is high, then the independence of the chi-square test is low, and this feature cannot be discarded. If there is not much relationship between the two and the correlation is low, then the independence of the chi-square test is high, and the impact of discarding this feature is not significant. In the actual project application, when using the chi-square to select features from data samples, there is no need to consider the degree of freedom of each sample point, nor the chi-square distribution. As long as the chi-square value of each feature is calculated and then sorted from large to small, the larger the chi-square value, the more important this feature is and it cannot be discarded. Just select several features with larger chi-square values, and the feature selection process of this chi-square test is completed.

[0066] An exemplary embodiment of the present invention, the data normalization includes:

[0067]

[0068] In the formula, y is the value after normalization, x is the original data, x min is the minimum value, x max is the maximum value.

[0069] An exemplary embodiment of the present invention, the establishment of the wind power generation power prediction model includes:

[0070] Establish an initial objective function and perform a Taylor expansion on the objective function to obtain the final objective function:

[0071]

[0072] In the formula, Obj (s) is the objective function value, N is the number of samples, g i is the first-order partial derivative of the loss function, f s (x i ) is the s-th base learner, h i is the second-order partial derivative of the loss function, γ is the penalty coefficient of the regularization term for each leaf node, T is the maximum depth of the tree, λ is the regularization coefficient, ω j is the weight on the j-th leaf, L is a constant, y j is the sample label, is the training result of the (s - 1)-th round of the model;

[0073] When the Obj (s) ≤0.5, output the result of power reduction. When the Obj (s) >0.5, output the result of power increase.

[0074] In this embodiment, in addition to supporting the learner based on the CART tree model, the above model can also be well adapted to the linear learner. At the same time, the L1 and L2 regularization terms with penalty properties are introduced, which improves the accuracy of the model and reduces the probability of overfitting.

[0075] In terms of the optimizer, the second-order Taylor expansion of the loss function is obtained, and then the first-order derivative and the second-order derivative are selected to complete the optimization, improving the optimization effect.

[0076] Construct all the subtrees of the corresponding XGBoost algorithm from top to bottom, and during the pruning process, adjust from bottom to top. This can prevent the program from falling into a local optimal solution.

[0077] Introduce the built-in poor validation algorithm into the XGBoost algorithm, which can add poor validation in each iteration, enabling it to output a better number of iterations.

[0078] Having a high adaptability is a significant feature of the XGBoost algorithm

[49] . After the corresponding iteration ends, the learning rate is assigned to the corresponding leaf nodes, and the weight coefficients of each tree itself will be adjusted. Then, the impact of each tree on the entire model will be reduced, and more improvements can be made in the learning of the new model. At the same time, users can define the optimization objective and loss function by themselves.

[0079] An exemplary implementation of the present invention, the establishment of the evaluation model includes:

[0080]

[0081] In the formula, U is the evaluation value, E is the number of wind turbines with increased predicted power in the area, G is the number of wind turbines in the area, is the objective function value of the l-th wind turbine, x r , y r and z r are the three-dimensional coordinates of the central wind turbine closest to the center of the area respectively, and x p , y p and z p are the three-dimensional coordinates of the remaining wind turbines in the area except the central wind turbine.

[0082] In the above evaluation model, the proportion of wind turbines with increased power and the objective function values of several wind turbines in a divided area are comprehensively considered, as well as the aggregation degree reflected by the distance between each wind turbine, so as to conduct a comprehensive average on the area and objectively obtain an evaluation score value. The higher the score value, the more attention should be paid by the staff, guiding the staff to do a good job in the preparation for area maintenance, and avoiding the waste of energy caused by the failure of the wind turbines in this area to generate electricity efficiently when the power increases.

[0083] The present invention also provides a system for constructing a wind power generation power prediction model, including:

[0084] A prediction model establishment module, configured to obtain historical data of wind power generation, preprocess the historical data, divide the preprocessed data into a training set and a test set; establish a wind power generation power prediction model, and output a predicted result of power increase or decrease through the wind power generation power prediction model;

[0085] An analysis module, configured to respectively obtain the positioning data of the wind turbines with increased and decreased power generation in the current wind farm, respectively perform different symbol markings on the wind turbines with increased and decreased power, and map them into a three-dimensional or two-dimensional data model according to the positioning data and symbol markings; divide the wind farm into several regions, obtain the symbol identifications and positioning data of the wind turbines in the same region, establish an evaluation model, output an evaluation value based on the symbol identifications and positioning data through the evaluation model, and output a sorting result after sorting the evaluation values;

[0086] A main control module, connected to the prediction model establishment module and the analysis module, for executing the above-mentioned method for constructing a wind power generation power prediction model.

[0087] In addition, in each embodiment of the present invention, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0088] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present invention. And the aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0089] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for constructing a wind power generation power prediction model, characterized in that Including: Obtain the historical data of wind power generation, preprocess the historical data, and divide the preprocessed data into a training set and a test set; Establish a wind power generation power prediction model, and output the result of the predicted power increase or decrease through the wind power generation power prediction model; Obtain the positioning data of the wind turbines with increased and decreased power generation in the current wind farm respectively, mark the increased and decreased wind turbines with different symbols respectively, and map them into a three-dimensional or two-dimensional data model according to the positioning data and symbol marks; Divide the wind farm into several regions, obtain the symbol identification and positioning data of the wind turbines in the same region, establish an evaluation model, output an evaluation value based on the evaluation model through the symbol identification and positioning data, and output a sorting result after sorting the evaluation values.

2. The method for constructing a wind power generation power prediction model according to claim 1, wherein The preprocessing of the historical data includes: Obtain target data including the first historical data with wind power output less than 0, the second historical data with wind speed less than the starting wind speed power value and not zero, the third historical data with power value exceeding the rated power of the wind turbine, and the fourth historical data with discontinuous jumps; Modify or replace the first historical data, the second historical data, the third historical data, and the fourth historical data respectively; Perform feature selection on the historical data after data cleaning, normalize the historical data after feature selection, and integrate them into a data set.

3. The method for constructing a wind power generation power prediction model according to claim 2, characterized in that The modification or replacement includes: Modify both the first historical data and the second historical data to 0, modify the third historical data to the rated power of the wind turbine, obtain the target historical data adjacent to the fourth historical data, calculate the average value of the target historical data, and replace the fourth historical data with the average value of the target historical data.

4. The method for constructing a wind power generation power prediction model according to claim 3, wherein, The feature selection of the historical data after data cleaning includes: Calculate the independence index between the input feature and the output feature, and set an independence index threshold. When the independence index is greater than the independence index threshold, discard the feature. When the independence index is not greater than the independence index threshold, retain the feature; Where χ is the independence index, k is the number of eigenvalues, n is the total frequency, and p i is the theoretical frequency of the i-th eigenvalue, and A i is the i-th eigenvalue.

5. The method for constructing a wind power generation power prediction model according to claim 4, wherein The data normalization includes: Where y is the normalized value, x is the original data, x min is the minimum value, and x max is the maximum value.

6. The method for constructing a wind power generation power prediction model according to claim 4, characterized in that The establishment of the wind power generation power prediction model includes: Establish an initial objective function, and perform a Taylor expansion on the objective function to obtain the final objective function: Where, Obj (s) is the objective function value, N is the number of samples, g i is the first-order partial derivative of the loss function, f s (x i ) is the base learner of the sth, h i is the second-order partial derivative of the loss function, γ is the penalty coefficient of the regularization term for each leaf node, T is the maximum depth of the tree, λ is the regularization coefficient, ω j is the weight on the jth leaf, L is a constant, y i is the sample label, is the training result of the (s - 1)th round of the model; When the Obj (s) ≤ 0.5, the result is a decrease in output power. When the Obj (s) > 0.5, the result is an increase in output power.

7. The method for constructing a wind power generation power prediction model according to claim 6, characterized in that, The establishment of the evaluation model includes: Wherein, U is the evaluation value, E is the number of wind turbines with predicted power increase in the area, G is the number of wind turbines in the area, is the objective function value of the l-th wind turbine, x r , y r and z r are the three-dimensional coordinates of the central wind turbine closest to the center of the area respectively, x p , y p and z p are the three-dimensional coordinates of the remaining wind turbines in the area except the central wind turbine.

8. A system for constructing a wind power generation power prediction model, characterized in that, Including: A prediction model establishment module, configured to obtain the historical data of wind power generation, preprocess the historical data, and divide the preprocessed data into a training set and a test set; establish a wind power generation power prediction model, and output the result of the predicted power increase or decrease through the wind power generation power prediction model; An analysis module, configured to obtain the positioning data of the wind turbines with increased and decreased power generation in the current wind farm respectively, mark the increased and decreased wind turbines with different symbols respectively, and map them into a three-dimensional or two-dimensional data model according to the positioning data and symbol marks; divide the wind farm into several regions, obtain the symbol identification and positioning data of the wind turbines in the same region, establish an evaluation model, output an evaluation value based on the evaluation model through the symbol identification and positioning data, and output a sorting result after sorting the evaluation values. The main control module, connected to the prediction model establishment module and the analysis module, is configured to execute the method for constructing the wind power generation power prediction model according to any one of claims 1-7.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for constructing the wind power generation power prediction model according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, it implements the method for constructing the wind power generation power prediction model according to any one of claims 1-7.

Images