CN117117858B - Wind turbine generator power prediction method, device and storage medium - Google Patents

Abstract

The invention relates to the technical field of fan unit power prediction, in particular to a method, a device and equipment for predicting the power of a wind turbine unit and a computer storage medium. According to the wind turbine generator power prediction method, interval division in a parameter method is combined with a gradient lifting regression tree in a non-parameter method, and data in each wind speed area is input as a training set of the gradient lifting regression tree to obtain reference power of each wind speed area and used for fitting a power curve; the invention also improves the gradient lifting regression tree by k neighbor weighted average so as to improve the contribution value of the sample which is closer to the predicted sample, thereby ensuring more accurate prediction effect.

Description

A wind turbine power prediction method, device and storage medium

Technical field

The invention relates to the technical field of wind turbine power prediction, and in particular, to a wind turbine power prediction method, device, equipment and computer storage medium.

Background technique

In modern society where traditional fossil energy resources are scarce and pollution is serious, wind energy is widely favored by the public as a pollution-free, renewable new energy source. The wind power industry has thus become one of the new renewable energy industries being vigorously developed at home and abroad. . In our country, the construction of wind farms and related research work have significantly improved in terms of quantity and quality in the past ten years. However, while vigorously developing the wind power industry, it has also been accompanied by the continuous degradation of the wind turbines themselves. a series of negative factors. During the use of today's wind turbines, due to the intermittent and highly uncertain characteristics of wind speed, it has a great impact on the performance evaluation of the wind turbine itself. It is necessary to correctly evaluate the performance and health of the wind turbine. Assessment and diagnosis are important for reasonable planning in wind power operation and maintenance.

The actual operation of wind farms is affected by the environment, unit status, operation methods, power grid dispatching, etc., and cannot maintain a high efficiency level at all times, resulting in loss of power generation. Wind turbine power curve modeling is used to predict wind turbine power. It is a necessary link to evaluate wind farm power generation efficiency and find ways to improve efficiency. Wind turbine power curve modeling methods are divided into parametric methods and non-parametric methods. Parametric methods mainly include the piecewise averaging method (IEC), piecewise linear model method, polynomial fitting, multi-parameter logistic function regression, etc.; non-parametric methods mainly include fuzzy logistic regression, neural network, K nearest neighbor method, etc.; however, conventional Parametric methods are not accurate enough for power estimation and are greatly affected by outliers, while non-parametric methods require a large number of iterative calculations. Therefore, modeling requires a lot of time under large-scale data.

Contents of the invention

To this end, the technical problem to be solved by the present invention is to overcome the problem of low power prediction accuracy in the prior art.

In order to solve the above technical problems, the present invention provides a wind turbine power prediction method, which includes:

Obtain a wind farm fan status data set, and divide the wind farm fan status data set according to wind speed zones;

Use the data in each wind speed zone as the training input of the gradient boosting regression tree to obtain the reference power of each wind speed zone;

Use the least squares method to fit the reference power of multiple wind speed zones to obtain the wind turbine power curve;

The wind turbine power is predicted according to the wind turbine power curve.

Preferably, obtaining the wind farm fan status data set and dividing the wind farm fan status data set according to wind speed zones includes:

Get wind farm turbine status data collection , among which, the fan status at the i-th moment/> ,/> Indicates wind speed,/> Represents active power,/> Indicates ambient air pressure,/> Represents the ambient temperature;

Initialize the wind speed zone collection , among which, the number of wind speed zones ,/> and/> are the cut-in wind speed and cut-out wind speed of the wind farm turbine respectively, and the kth wind speed zone ,/> Represents the fan status data set contained in the k-th wind speed zone,/> Represents the central wind speed of the kth wind speed zone,/> Represents the reference power of the kth wind speed zone;

Traverse the wind farm turbine status data collection , according to the fan status at the i-th moment/> of wind speed Calculate wind speed interval number/> , and change the fan status at the i-th moment/> Join the kth wind speed zone/> fan status data collection/> middle.

Preferably, obtaining the wind farm turbine status data set further includes:

The active power Corrected to power value under standard atmospheric pressure and standard ambient temperature/> :

in, is the air density,/> is the atmospheric pressure,/> is the ambient temperature,/> Represents standard atmospheric pressure,/> Indicates standard ambient temperature.

Preferably, the gradient boosting regression tree is a gradient boosting regression tree improved by using k nearest neighbor weighted average.

Preferably, using the data in each wind speed zone as the training input of the gradient boosting regression tree to obtain the reference power of each wind speed zone includes:

Step a: Calculate the wind speed of the tth wind speed zone As training set related variables/> , active power/> As the training set target variable/> , obtain the training set and initialize the weak learner;

Step b: Calculate the current regression tree model residuals ;

Step c: Use the residual as the target variable of the training set to obtain a new training set, and use the cart algorithm to fit the mth regression tree;

Step d: Calculate the leaf node area of the m-th regression tree The distance between all training samples and the sample mean;

Step e: Screen out the K training samples with the smallest distance, and calculate the respective weights of the K training samples;

Step f: Calculate the predicted value of the mth regression tree leaf node and update the strong learner;

Step g: When the current number of training times is not less than the maximum number of training times, the final regression tree is obtained;

Step h: Calculate the current wind speed zone The central wind speed/> , and input the final regression tree to obtain the reference power of the current wind speed zone.

Preferably, the calculated current wind speed zone The central wind speed/> , and input the final regression tree. After obtaining the reference power of the current wind speed area, it also includes:

Update wind speed-power counter t=t+1;

when , jump to step a.

Preferably, the method of fitting the reference power of multiple wind speed zones using the least squares method to obtain the wind turbine power curve includes:

Collect the wind speed zones The kth wind speed zone in the middle/> The central wind speed/> As the abscissa, the reference power/> As the ordinate, the least squares method is used to perform power curve fitting to obtain the wind turbine power curve.

The invention also provides a wind turbine power prediction device, which includes:

A data set dividing module is used to obtain a wind farm fan status data set and divide the wind farm fan status data set according to wind speed zones;

The reference power calculation module is used to use the data in each wind speed zone as the training input of the gradient boosting regression tree to obtain the reference power of each wind speed zone;

The power curve fitting module is used to fit the reference power of multiple wind speed zones using the least squares method to obtain the wind turbine power curve;

A power prediction module is used to predict wind turbine power according to the wind turbine power curve.

The invention also provides a wind turbine power prediction device, including:

Memory, used to store computer programs;

A processor, configured to implement the steps of the above-mentioned wind turbine power prediction method when executing the computer program.

The present invention also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned wind turbine power prediction method are implemented.

The above technical solution of the present invention has the following advantages compared with the existing technology:

The wind turbine power prediction method of the present invention combines the interval division in the parametric method with the gradient boosting regression tree in the non-parametric method, and inputs the data in each wind speed zone as the training set of the gradient boosting regression tree to obtain The reference power of each wind speed zone is used to fit the power curve; the present invention also uses k nearest neighbor weighted average to improve the gradient boosting regression tree to increase the contribution value of samples closer to the predicted sample, making the prediction effect more accurate .

Description of the drawings

In order to make the content of the present invention easier to understand clearly, the present invention will be further described in detail below based on specific embodiments of the present invention and in conjunction with the accompanying drawings, wherein:

Figure 1 is an implementation flow chart of a wind turbine power prediction method provided by the present invention;

Figure 2 is a flow chart for realizing wind turbine power prediction provided by an embodiment of the present invention.

Detailed ways

The core of the present invention is to provide a wind turbine power prediction method, device, equipment and computer storage medium, which effectively improves the accuracy of power prediction.

In order to enable those skilled in the art to better understand the solution of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

The main purpose of the present invention is to improve the accuracy of wind turbine unit power prediction. The method based on interval division divides the wind speed into different wind speed intervals, and obtains a power estimate by certain means in different wind speed intervals. Among the conventional power curve modeling methods, the piecewise mean method (IEC), the maximum probability point method, and the piecewise linear model all belong to this category. However, these methods are not accurate enough for power estimation and are affected by outliers. The impact is relatively large. The method proposed by the present invention is used for power prediction, and the interval division is combined with the machine learning method to improve the accuracy of power prediction.

The present invention proposes a wind turbine power prediction method based on an improved gradient boosting regression tree. The main idea of the method is: after obtaining the wind farm wind turbine data set, the power is corrected to the power under standard atmospheric pressure and standard ambient temperature, and according to The wind speed zone partitions the data set. For the wind power data in each wind speed zone, an improved gradient boosting regression tree is used to obtain a regression tree function, and the central wind speed of the wind speed zone is input to obtain a reference power. After obtaining the values of several wind speed zones After referencing the power, these points are fitted using the least squares method to obtain a power curve for power prediction of the wind turbine:

Please refer to Figure 1, which is an implementation flow chart of a wind turbine power prediction method provided by the present invention; the specific operation steps are as follows:

S101: Obtain the wind farm fan status data set, and divide the wind farm fan status data set according to wind speed zones;

S102: Use the data in each wind speed zone as the training input of the gradient boosting regression tree to obtain the reference power of each wind speed zone;

S103: Use the least squares method to fit the reference power of multiple wind speed zones to obtain the wind turbine power curve;

S104: Predict the wind turbine power according to the wind turbine power curve.

Based on the above embodiment, this embodiment describes step S101 in detail:

Get wind farm turbine status data collection , among which, the fan status at the i-th moment/> ,/> Indicates wind speed,/> Represents active power,/> Indicates ambient air pressure,/> Represents the ambient temperature;

Initialize the wind speed zone collection , among which, the number of wind speed zones ,/> and/> are the cut-in wind speed and cut-out wind speed of the wind farm turbine respectively, and the kth wind speed zone ,/> Represents the fan status data set contained in the k-th wind speed zone, initialized to empty, /> Represents the central wind speed of the kth wind speed zone,/> Represents the reference power of the kth wind speed zone (initially 0), and the initial wind speed power counter t=1;

Traverse the wind farm turbine status data collection , according to the fan status at the i-th moment/> of wind speed Calculate wind speed interval number/> , and change the fan status at the i-th moment/> Join the kth wind speed zone/> fan status data collection/> middle.

Based on the above embodiment, obtaining the wind farm turbine status data set also includes:

The active power Corrected to power value under standard atmospheric pressure and standard ambient temperature/> :

in, is the air density,/> is the atmospheric pressure,/> is the ambient temperature,/> Represents the standard atmospheric pressure, which is 101.325kPa,/> Indicates the standard ambient temperature, which is 20℃.

Based on the above embodiment, this embodiment describes step S102 in detail:

The gradient boosting regression tree is a gradient boosting regression tree improved by using k nearest neighbor weighted average.

Using the data in each wind speed zone as the training input of the gradient boosting regression tree, obtaining the reference power of each wind speed zone includes:

Step a: Calculate the wind speed of the tth wind speed zone As training set related variables/> , active power/> As the training set target variable/> , obtain the training set and initialize the weak learner;

Gather from the wind speed zone Select the tth wind speed zone/> Wind power data set/> The wind speed/> As training set related variables/> , the corrected power value is/> As the training set target variable/> , get the input training set/> , the number of samples in the training data set is N, the maximum number of training times is M, and the loss function is L. And calculate the initialized weak learner according to the following formula/> .

Step b: Calculate the current regression tree Model residual/> , where m=1,2,…,M, i=1,2,…N.

Step c: Use the residual as the target variable of the training set to obtain a new training set , and use the cart algorithm to fit to obtain the mth regression tree/> ;

Step d: Calculate the leaf node area of the m-th regression tree All training samples on and sample mean/> the distance between:

Among them, j=1,2,..., J, J is the number of leaf nodes of the regression tree m, and S is number of samples on;

Step e: Screen out the K training samples with the smallest distance, and calculate the respective weights of the K training samples:

find distance The smallest K training samples, record the output variable value of these K training samples as , they are related to the sample mean/> The distance is/> ,The weight of each training sample is/> ,weight/> It is calculated according to the following formula:

Step f: Calculate the predicted value of the leaf node of the mth regression tree and update the strong learner:

Step g: When the current number of training times is not less than the maximum number of training times, the final regression tree is obtained:

Determine whether the current training times m is less than the maximum training times M. If it is less, then m=m+1 and jump to step a. Otherwise, continue execution.

Step h: Calculate the current wind speed zone The central wind speed/> , and input the final regression tree/> , get the reference power of the current wind speed area/> =/> :

Based on the above embodiment, the calculation of the current wind speed zone The central wind speed/> , and input the final regression tree. After obtaining the reference power of the current wind speed area, it also includes:

Update wind speed-power counter t=t+1;

when , jump to step a.

Based on the above embodiment, this embodiment describes step S103 in detail:

Collect wind speed areas in/> The central wind speed/> As the abscissa, the reference power/> As the ordinate, the least squares method is used to perform power curve fitting, and finally the wind turbine power curve is obtained, which is used to predict wind turbine power prediction.

Step a uses the method of combining wind speed partitioning and gradient boosting regression tree. The data in each wind speed zone is input as the training set of the gradient boosting regression tree to obtain the reference power of each wind speed zone for fitting the power curve. , the document is original among similar literature on related work.

Steps d, e, f, h improve the gradient boosting regression tree algorithm and introduce the k nearest neighbor weighted average method to replace the simple average, calculate the distance between the training sample and the prediction sample on the leaf node, and select the smallest k prediction sample , and calculate the weight of the prediction sample, and improve the contribution of the training samples that are close to the prediction sample to the reference power of the predicted wind speed area. It is original in the relevant literature of the same type.

The present invention proposes a prediction method that has been based on improved gradient boosting regression trees. Currently, the mainstream wind turbine power curve modeling methods are divided into parametric methods such as interval division and polynomial fitting, and non-parametric methods such as machine learning. The main advantage of this strategy is that it combines the interval division in the parametric method with the gradient boosting regression tree in the non-parametric method, and uses k-nearest neighbor weighted average to improve the gradient boosting regression tree to increase the contribution of samples closer to the predicted sample. value, making the prediction effect closer.

For most wind power prediction methods, the main points to consider are: screening and optimizing data sets and improving the efficiency and accuracy of modeling. The present invention takes these two points into consideration and is essentially different from other prediction methods. The main differences are as follows:

1. Combine the Bien method in the parametric method with the gradient boosting regression tree in the non-parametric method. The conventional parametric method is not accurate enough for power estimation and is greatly affected by outliers. The non-parametric method needs to be A large number of iterative calculations, therefore, modeling requires a lot of time with large-scale data. The present invention combines parametric methods with non-parametric methods to take into account both the efficiency and modeling accuracy of wind power power curve modeling;

2. Use k-nearest neighbor weighted average to improve the gradient boosting regression tree to increase the contribution value of samples that are closer to the predicted sample, making the prediction effect closer, and further reducing the impact of outliers to improve the accuracy of prediction.

Please refer to Figure 2, which is a flow chart for realizing wind turbine power prediction provided by an embodiment of the present invention.

The embodiment of the present invention also provides a wind turbine power prediction device; the specific device may include:

The data set dividing module 100 is used to obtain a wind farm fan status data set and divide the wind farm fan status data set according to wind speed zones;

The reference power calculation module 200 is used to use the data in each wind speed zone as the training input of the gradient boosting regression tree to obtain the reference power of each wind speed zone;

The power curve fitting module 300 is used to fit the reference power of multiple wind speed zones using the least square method to obtain the wind turbine power curve;

The power prediction module 400 is used to predict the wind turbine power according to the wind turbine power curve.

The wind turbine power prediction device of this embodiment is used to implement the aforementioned wind turbine power prediction method. Therefore, the specific implementation of the wind turbine power prediction device can be found in the embodiments of the wind turbine power prediction method described above. For example, the data set partition module 100 , the reference power calculation module 200, the power curve fitting module 300, and the power prediction module 400 are respectively used to implement steps S101, S102, S103, and S104 in the above wind turbine power prediction method. Therefore, the specific implementation can refer to the corresponding parts. The description of the embodiment will not be repeated here.

Specific embodiments of the present invention also provide a wind turbine power prediction device, including: a memory for storing a computer program; and a processor for implementing the steps of the above wind turbine power prediction method when executing the computer program.

Specific embodiments of the present invention also provide a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps of the above wind turbine power prediction method are implemented.

Those skilled in the art will understand that embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment that combines software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each process and/or block in the flowchart illustrations and/or block diagrams, and combinations of processes and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce a use A device for realizing the functions specified in a process or processes in a flowchart and/or a block or blocks in a block diagram.

These computer program instructions may also be stored in a computer-readable memory that causes a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including the instruction means, the instructions The device implements the functions specified in a process or processes in the flowchart and/or in a block or blocks in the block diagram.

These computer program instructions may also be loaded onto a computer or other programmable data processing device, causing a series of operating steps to be performed on the computer or other programmable device to produce computer-implemented processing, thereby executing on the computer or other programmable device. Instructions provide steps for implementing the functions specified in a process or processes of a flowchart diagram and/or a block or blocks of a block diagram.

Obviously, the above-mentioned embodiments are only examples for clear explanation and are not intended to limit the implementation. For those of ordinary skill in the art, other changes or modifications may be made based on the above description. An exhaustive list of all implementations is not necessary or possible. The obvious changes or modifications derived therefrom are still within the protection scope of the present invention.

Claims (9)

1. A wind turbine power prediction method, characterized by including: Obtain a wind farm fan status data set, and divide the wind farm fan status data set according to wind speed zones; Use the data in each wind speed zone as the training input of the gradient boosting regression tree to obtain the reference power of each wind speed zone; Use the least squares method to fit the reference power of multiple wind speed zones to obtain the wind turbine power curve; Predict the wind turbine power according to the wind turbine power curve; Wherein, dividing the wind farm turbine status data set according to wind speed zones includes: Obtain wind farm turbine status data collection; Initialize the wind speed zone collection; Traverse the wind farm fan status data set, calculate the wind speed interval number based on the wind speed of the fan status at the i-th moment, and add the i-th fan status to the fan status data set of the k-th wind speed zone; The data in each wind speed zone is used as the training input of the gradient boosting regression tree to obtain the reference power of each wind speed zone, including: The wind speed of the tth wind speed zone is As training set related variables/> , active power/> As the training set target variable/> , obtain the training set and initialize the weak learner; Calculate the current regression tree model residuals ; Use the residual as the target variable of the training set to obtain a new training set, and use the cart algorithm to fit the mth regression tree; Calculate the leaf node area of the mth regression tree The distance between all training samples and the sample mean; Screen out the K training samples with the smallest distance, and calculate the respective weights of the K training samples; Calculate the predicted value of the mth regression tree leaf node and update the strong learner; When the current number of training times is not less than the maximum number of training times, the final regression tree is obtained; Calculate current wind speed zone The central wind speed/> , and input the final regression tree to obtain the reference power of the current wind speed area. 2. The wind turbine power prediction method according to claim 1, characterized in that said obtaining a wind farm fan status data set and dividing the wind farm fan status data set according to wind speed zones includes: The wind farm turbine status data set is , where the fan status at the i-th moment is/> ,/> Indicates wind speed,/> Represents active power,/> Indicates ambient air pressure,/> Represents the ambient temperature; The initialized wind speed zone set is , among which, the number of wind speed zones ,/> and/> are the cut-in wind speed and cut-out wind speed of the wind farm turbine respectively, and the kth wind speed zone/> ,/> Represents the fan status data set contained in the kth wind speed zone, Represents the central wind speed of the kth wind speed zone, /> Represents the reference power of the kth wind speed zone; The fan status at the i-th moment The wind speed is/> , the wind speed interval number is/> . 3. The wind turbine power prediction method according to claim 2, characterized in that, after obtaining the wind farm fan status data set, it also includes: The active power Corrected to power value under standard atmospheric pressure and standard ambient temperature/> : in, is the air density,/> is the atmospheric pressure,/> is the ambient temperature,/> Represents standard atmospheric pressure,/> Indicates standard ambient temperature. 4. The wind turbine power prediction method according to claim 2, characterized in that the gradient boosting regression tree is a gradient boosting regression tree improved by using k nearest neighbor weighted average. 5. The wind turbine power prediction method according to claim 1, characterized in that the calculation of the current wind speed zone The central wind speed/> , and input the final regression tree. After obtaining the reference power of the current wind speed area, it also includes: Update wind speed-power counter ; when , jump to step a. 6. The wind turbine power prediction method according to claim 2, characterized in that the use of the least square method to fit the reference power of multiple wind speed zones to obtain the wind turbine power curve includes: Collect the wind speed zones The kth wind speed zone in the middle/> The central wind speed/> As the abscissa, the reference power/> As the ordinate, the least squares method is used to perform power curve fitting to obtain the wind turbine power curve. 7. A wind turbine power prediction device, characterized in that it includes: A data set dividing module is used to obtain a wind farm fan status data set and divide the wind farm fan status data set according to wind speed zones; The reference power calculation module is used to use the data in each wind speed zone as the training input of the gradient boosting regression tree to obtain the reference power of each wind speed zone; The power curve fitting module is used to fit the reference power of multiple wind speed zones using the least squares method to obtain the wind turbine power curve; A power prediction module, used to predict the power of the wind turbine according to the wind turbine power curve; Wherein, the data set division module is also used to: obtain the wind farm fan status data set, initialize the wind speed zone set, traverse the wind farm fan status data set, calculate the wind speed interval number according to the wind speed of the wind turbine status at the i-th moment, and The fan status at the i-th moment is added to the fan status data set of the k-th wind speed zone; The reference power calculation module is also used to: convert the wind speed of the tth wind speed zone As training set related variables/> , active power/> As the training set target variable/> , obtain the training set, initialize the weak learner, and calculate the current regression tree model residual , use the residual as the target variable of the training set to obtain a new training set, and use the cart algorithm to fit to obtain the mth regression tree, and calculate the leaf node area of the mth regression tree/> based on the distance between all training samples and the sample mean, select the K training samples with the smallest distance, calculate the respective weights of the K training samples, calculate the predicted value of the mth regression tree leaf node, and update Strong learner, when the current number of training times is not less than the maximum number of training times, the final regression tree is obtained and the current wind speed zone is calculated/> The central wind speed/> , and input the final regression tree to obtain the reference power of the current wind speed zone. 8. A wind turbine power prediction device, characterized by including: Memory, used to store computer programs; A processor, configured to implement the steps of a wind turbine power prediction method according to any one of claims 1 to 6 when executing the computer program. 9. 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 a processor, the method of any one of claims 1 to 6 is implemented. Steps of wind turbine power prediction method.