CN114764741A - Method and system for predicting running wind power of wind driven generator - Google Patents

Abstract

The invention provides a method and a system for predicting wind power of a wind driven generator during operation, which comprises the following steps: s1: acquiring data; s2: preprocessing data; s3: processing the characteristics; s4: training a primary model; s5: selecting a neural network algorithm as a secondary model, combining the primary models, and performing secondary training by combining a plurality of model judgment results; s6: adjusting and optimizing the model; s7: adjusting model parameters; s8: power prediction is performed based on real-time weather conditions and model parameters in S7. The invention aims to improve the use efficiency and the prediction precision of the existing method, improve the method, explain the application of the method and have substantial production guidance effect and efficiency improvement effect on actual production.

Description

Method and system for predicting running wind power of wind driven generator

[ technical field ] A method for producing a semiconductor device

The invention relates to the technical field of wind power generation, in particular to a method and a system for predicting the running wind power of a wind driven generator.

[ background of the invention ]

At present, wind power generation equipment and technology are mature, but the uncertainty and the fluctuation of the power generation quality of the wind power generation equipment bring great difficulty to wind power integration and cause great waste. Accurate prediction of wind power generation is an effective solution. And a machine learning method is applied to carry out wind power prediction according to meteorological data and fan operation data, and the model is realized in a big data environment, so that the operation efficiency of the whole model is improved.

In the year 2017, the 'wind power prediction method for power system decision-oriented' explains a cost-oriented lifting regression tree algorithm to predict wind power. Su Yongxin is equal to 200 years and 1 month, and algorithms in 5 such as LSTM, SVM, RNN and the like are applied to model combination in a mountain wind turbine generator behavior prediction model based on integrated learning, so that the wind power is predicted, and a good effect is achieved.

The two documents respectively mention a method for predicting wind power by applying a single algorithm and a method for predicting wind power by performing a model combination algorithm on the basis of a single model.

In the past, the wind power is basically predicted, but the Li nation only adopts a single model to predict the wind power, and the Su-Yongxin application method is adopted, but the type reason of the used primary model is not elaborated in detail, and the selection method of the model has a decisive influence on the accuracy of the final prediction result. And the criterion input by the model index is not mentioned, which also has an influence on the prediction result. .

Accordingly, there is a need to develop a method and system for wind power prediction for wind turbine operation that addresses the deficiencies of the prior art to address or mitigate one or more of the problems set forth above.

[ summary of the invention ]

In view of the above, the invention provides a method and a system for predicting wind power of a wind driven generator, and aims to improve the use efficiency and the prediction accuracy of the existing method, improve the method, explain the application of the method, and have substantial production guidance effect and efficiency improvement effect on actual production.

In one aspect, the present invention provides a method for wind power prediction of wind power generator operation, comprising the steps of:

s1: acquiring data, namely, taking a related data source from a real-time database by a big data HIVE database, and transmitting the data source to a calculation server;

s2: data preprocessing, namely splitting, cleaning and balancing the data by a computing server to obtain a training set and a test set;

s3: performing characteristic processing, namely performing data standardization processing on the wind speed, the air temperature, the air pressure, the air humidity and the fan power which are concentrated in the training set and the test;

s4: primary model training, namely selecting multiple single algorithms as a primary model, and performing primary model training on the data after feature processing by a k-fold cross validation method;

s5: selecting a neural network algorithm as a secondary model, combining the primary models, and performing secondary training by combining a plurality of model judgment results;

s6: model optimization: applying the test set data to the secondary model to perform model accuracy test, and continuously optimizing the model;

s7: adjusting the model parameters to enable the evaluation index to reach a preset standard, outputting the model, calculating the real-time data, and storing the calculation result in a server;

s8: the power prediction is made based on the real-time weather conditions and the model parameters in S7.

As for the above-mentioned aspect and any possible implementation manner, there is further provided an implementation manner, where the S1 specifically is: and comprehensively applying an apriori algorithm and covariance analysis, selecting variables related to target variable fan power, and selecting weather forecast wind speed, air temperature, wind direction, weather, air pressure, air humidity and fan power data points as input data sources.

As to the above-mentioned aspect and any possible implementation manner, there is further provided an implementation manner, where the splitting process in S2 specifically includes: the data set is split into a training set and a test set.

As to the above-mentioned aspect and any possible implementation manner, a further implementation manner is provided, where the cleaning and balancing process in S2 specifically includes: and removing dirty data, and performing data balancing treatment to ensure that the power data are distributed in a balanced manner.

As for the above-mentioned aspect and any possible implementation manner, there is further provided an implementation manner, where the S3 specifically is: and converting the text type wind direction data into double type data between 0 and 360 degrees, and converting the meteorological data into int type data for storage.

The above-mentioned aspects and any possible implementation further provide an implementation, and the single algorithm in S4 includes LSTM, Beta-based distribution algorithm, kalman filter algorithm, and LR algorithm.

As to the above-mentioned aspect and any possible implementation manner, there is further provided an implementation manner, where the method for k-fold cross validation in S4 specifically includes:

s41: dividing the training set into 5-fold data randomly, taking four-fold data as training data and the last one-fold data as test data in sequence, and respectively training by using a primary model to obtain a prediction target variable based on the test data of the training set;

s42: predicting the test set by using the model to obtain prediction data based on the test set;

s43: repeating the steps S41-S42, and respectively applying four primary models;

s44: and regenerating new feature columns from the four model training results to obtain a predicted target variable based on cross validation and training set data and a predicted target variable based on a test set.

As to the above-mentioned aspect and any possible implementation manner, there is further provided an implementation manner, where the S5 specifically is: and taking the predicted target variable of the primary model based on the cross validation and training set data as a secondary model independent variable, taking the predicted target variable of the primary model based on the test set as a primary model target variable, and applying a neural network algorithm to train the secondary model.

The above-described aspect and any possible implementation manner further provide a system for wind power prediction of wind generator operation, the system including:

the data acquisition module is used for acquiring related data sources from the real-time database and transmitting the data sources to the calculation server;

the data preprocessing module is used for splitting, cleaning and balancing data to obtain a training set and a test set;

the characteristic processing module is used for carrying out data standardization processing on the wind speed, the air temperature, the air pressure, the air humidity and the fan power which are collected by training and testing;

the primary model training module selects multiple single algorithms as a primary model, and performs primary model training on the data after the characteristic processing by a k-fold cross validation method;

the secondary model training module selects a neural network algorithm as a secondary model, combines the primary models and performs secondary training by combining a plurality of model judgment results;

the model tuning module is used for carrying out model accuracy test on the secondary model application test set data and continuously optimizing the model;

the model parameter adjusting module is used for adjusting model parameters to enable the evaluation indexes to reach the preset standard, outputting the model, calculating the real-time data and storing the calculation result in the server;

and the real-time prediction module is used for predicting the wind power according to the real-time weather condition and the adjusted model parameters.

The above-described aspects and any possible implementation further provide a readable storage medium, which is a non-volatile storage medium or a non-transitory storage medium, having stored thereon a computer program that, when executed by a processor, performs the steps of any of the methods.

Compared with the prior art, the invention can obtain the following technical effects:

1): in the technology, aiming at the problems that the model input variable selection standard is fuzzy, the primary model selection standard is too subjective and unverified in the prior art, the method adopts a scientific algorithm to select the input variable, comprehensively selects the primary model through literature reference and actual verification, adopts a cross verification method to construct, has strong robustness, is more scientific in the model construction process, and obviously improves the model prediction accuracy.

2): in application, a big data technology is applied, the wind power prediction result is linked with a production marketing system, production activities and marketing work are organically combined, the calculation speed is increased, intermediate links are reduced, the work efficiency is improved, and the method has great significance for actual production.

Of course, it is not necessary for any product to achieve all of the above-described technical effects simultaneously in the practice of the invention.

[ description of the drawings ]

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a hardware block diagram of a system provided by one embodiment of the present invention;

fig. 2 is a flow chart of a method provided by an embodiment of the invention.

[ detailed description ] embodiments

In order to better understand the technical scheme of the invention, the following detailed description of the embodiments of the invention is made with reference to the accompanying drawings.

It should be understood that the described embodiments are only some embodiments of the invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the examples of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

The invention provides a method for predicting wind power of a wind driven generator in operation, which combines a secondary model on the basis of an output result of a primary model by applying a stacking algorithm to improve the wind power prediction accuracy. A big data mathematical analysis and mathematical modeling method is adopted, and a multilayer model is constructed on the basis of a big data platform, so that the problem of wind power prediction is solved. The method mainly adopts an application primary model algorithm, calculates results according to k-fold cross validation, takes output results as a training data set, and combines an initial target variable to construct a secondary model, so that prediction errors occurring in the primary model are emphasized and corrected in the training of the secondary model, the power prediction accuracy is improved, the prediction errors are displayed and reminded at a web end, the weather condition is monitored in real time, attention is reminded when the prediction power is large, and possible defects are eliminated in time by combining a unit health management system, and the stable operation of a unit is ensured.

The specific process of the method is shown in figure 1:

acquisition of data

And (5) a data acquisition stage. And the big data HIVE database takes relevant data sources from the real-time database and transmits the data sources to the calculation server. And comprehensively applying apriori algorithm and covariance analysis, selecting variables with large correlation with the power of the target variable fan, removing unnecessary intermediate results, avoiding overfitting of the model, and finally selecting weather forecast wind speed, air temperature, wind direction, weather, air pressure, air humidity and fan power data points as input data sources.

The method comprises the steps of collecting weather forecast wind speed, air temperature, wind direction, weather, air pressure, air humidity and fan power data, wherein the time length is 3 months, the data volume is about 1 hundred million, and the data are stored in a big data HIVE platform and are data sources of subsequent processes.

Model construction

Step 1, data preprocessing: and performing data processing on the data source acquired from the HIVE database. Since most of the original data is regular data, the data distribution is unbalanced. Training the model using unbalanced data can result in a virtual high training accuracy. Therefore, the imbalance of data is dealt with first. The method mainly comprises the steps of randomly screening power data, and hierarchically and uniformly selecting the power data in different ranges to ensure balance.

And remove the abnormal data.

Step 2, data splitting: after the data preprocessing is finished, data splitting work is carried out, the data source is split into training data and testing data, 60% -80% of the data source is used as the training data, 20% -40% of the data source is used as the testing data, and the splitting process is random.

And (5) performing model training by using the training set, and performing model tuning by using the testing set.

Step 3, feature processing: because the input original data has a large distribution range and is not uniformly distributed, the accuracy of the model is affected when the input original data is directly used, and characteristic processing is needed. The data of the weather forecast wind speed, the air temperature, the air pressure, the air humidity and the fan power are all double types, and the scales of the characteristics of the original data in different dimensions are inconsistent, so that the data of the weather forecast wind speed, the air temperature, the air pressure, the air humidity and the fan power are subjected to data standardization processing.

And converting the text type wind direction data into double type data between 0 and 360 degrees, converting the meteorological data into int type data for storage, wherein the meteorological data is text type data.

Step 4, primary model training: according to query documents and actual verification, the LSTM algorithm, the Beta distribution algorithm-based algorithm, the Kalman filtering algorithm and the LR algorithm have satisfactory effects in predicting the wind power by applying a single model, and therefore the four models are selected as primary models.

And (3) performing primary model training by using a primary model and a k-fold cross validation method. The invention trains the primary model by adopting 5-fold cross validation. And (3) dividing the training set into 5-fold data randomly, taking 4-fold data as training data and taking the last-fold data as test data.

For the LSTM model, firstly selecting 1-fold data in 80% training set after data splitting as test data, and training the rest 4-fold data as training data to obtain a target variable calculated on the 1-fold test data based on the 4-fold training data; and selecting the other 1-fold data as test data, and using the rest 4-fold data as training data to obtain a target variable calculated on the 1-fold test data based on the 4-fold training data, wherein the target variable is repeated for 5 times, and all the 5-fold data serve as over-test data. The results of 5 tests were serially concatenated. This is the training set 1 of the next secondary model.

And after 1-fold test data is predicted every time cross validation is performed, predicting 20% of test sets after data splitting by using the training model to obtain prediction data based on the test sets, and averaging output results of 5 times of cross validation models, wherein the data are the test set 1 of the subsequent secondary model.

And repeating the steps, respectively applying the four primary models, and connecting a training set 1, a training set 2, a training set 3, a training set 4 and a training set 5 into a matrix. And connecting the test set 1, the test set 2, the test set 3, the test set 4 and the test set 5 into a matrix.

Step 5, combining secondary models: and selecting a neural network algorithm as a secondary model, combining the calculation results of the primary model, and performing secondary training by combining a plurality of model judgment results to improve the model prediction accuracy.

And taking a prediction target variable matrix of the primary model based on cross validation and training set data as a secondary model independent variable, taking a prediction target variable of the primary model based on the test set as a primary model target variable, and carrying out secondary model training by applying a neural network algorithm.

Step 6, model tuning: and applying the test set data to the secondary model to perform model accuracy test, and continuously optimizing the model.

And 7, adjusting model parameters to enable the evaluation indexes to reach a certain degree, outputting the model, calculating the real-time data, and storing the calculation result in a server for subsequent calling.

③ Web exhibition

And in the centralized monitoring platform, the output result of the model is visually displayed, the prediction result is displayed in real time, and the change trend of the prediction result is known. And the marketing activities are guided by combining the report data of the marketing system.

The system hardware in the invention is shown in fig. 2, and integrates data acquisition, data storage, data analysis and mining and data display by using a method of combining software and hardware.

The method and the system for predicting the operating wind power of the wind driven generator provided by the embodiment of the application are introduced in detail above. The above description of the embodiments is only for the purpose of helping to understand the method of the present application and its core idea; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

As used in the specification and claims, certain terms are used to refer to particular components. As one skilled in the art will appreciate, manufacturers may refer to a component by different names. This specification and claims do not intend to distinguish between components that differ in name but not function. In the following description and in the claims, the terms "include" and "comprise" are used in an open-ended fashion, and thus should be interpreted to mean "include, but not limited to. "substantially" means within an acceptable error range, and a person skilled in the art can solve the technical problem within a certain error range to substantially achieve the technical effect. The description which follows is a preferred embodiment of the present application, but is made for the purpose of illustrating the general principles of the application and not for the purpose of limiting the scope of the application. The protection scope of the present application shall be subject to the definitions of the appended claims.

It is also noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a good or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such good or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a commodity or system that includes the element.

It should be understood that the term "and/or" as used herein is merely one type of association that describes an associated object, meaning that three relationships may exist, e.g., a and/or B may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.

The foregoing description shows and describes several preferred embodiments of the present application, but as aforementioned, it is to be understood that the application is not limited to the forms disclosed herein, but is not to be construed as excluding other embodiments and is capable of use in various other combinations, modifications, and environments and is capable of changes within the scope of the application as described herein, commensurate with the above teachings, or the skill or knowledge of the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the application, which is to be protected by the claims appended hereto.

Claims (9)

1. A method of wind power prediction for wind generator operation, the method comprising the steps of:

s1: acquiring data, namely acquiring a related data source from a real-time database by a big data HIVE database, and transmitting the data source to a calculation server;

s2: data preprocessing, namely splitting, cleaning and balancing the data by a computing server to obtain a training set and a test set;

s3: performing characteristic processing, namely performing data standardization processing on the wind speed, the air temperature, the air pressure, the air humidity and the fan power which are concentrated in the training set and the test;

s4: primary model training, namely selecting multiple single algorithms as a primary model, and performing primary model training on the data after feature processing by a k-fold cross validation method;

s5: selecting a neural network algorithm as a secondary model, combining the primary models, and performing secondary training by combining a plurality of model judgment results;

s6: model optimization: applying the test set data to the secondary model to perform model accuracy test, and continuously optimizing the model;

s7: adjusting the model parameters to enable the evaluation index to reach a preset standard, outputting the model, calculating the real-time data, and storing the calculation result in a server;

s8: the power prediction is made based on the real-time weather conditions and the model parameters in S7.

2. The method according to claim 1, wherein S1 is specifically: and comprehensively applying an apriori algorithm and covariance analysis, selecting variables related to the target variable fan power, and selecting weather forecast wind speed, air temperature, wind direction, weather, air pressure, air humidity and fan power data points as input data sources.

3. The method according to claim 1, wherein the cleaning and equilibrating treatment in S2 is specifically: and removing dirty data, and performing data balancing processing to enable power data to be distributed in a balanced manner.

4. The method according to claim 1, wherein S3 is specifically: and converting the text type wind direction data into double type data between 0 and 360 degrees, and converting the meteorological data into int type data for storage.

5. The method of claim 1, wherein the single algorithm in S4 includes LSTM, Beta-based distribution algorithm, kalman filter algorithm, and LR algorithm.

6. The method according to claim 5, wherein the method for k-fold cross validation in S4 specifically comprises:

s41: dividing the training set into 5-fold data randomly, taking four-fold data as training data and the last one-fold data as test data in sequence, and respectively training by using a primary model to obtain a prediction target variable based on the test data of the training set;

s42: predicting the test set by using the model to obtain prediction data based on the test set;

s43: repeating S41-S42, respectively applying four primary models;

s44: and regenerating new feature columns from the four model training results to obtain a predicted target variable based on cross validation and training set data and a predicted target variable based on a test set.

7. The method according to claim 1, wherein S5 is specifically: and (3) taking the predicted target variable of the primary model based on the cross validation and training set data as a secondary model independent variable, taking the predicted target variable of the primary model based on the test set as a primary model target variable, and carrying out secondary model training by applying a neural network algorithm.

8. A system for wind power prediction of the operation of a wind turbine, comprising the method according to any of the claims 1-7, characterized in that the system comprises:

the data acquisition module is used for acquiring related data sources from the real-time database and transmitting the data sources to the calculation server;

the data preprocessing module is used for splitting, cleaning and balancing data to obtain a training set and a test set;

the characteristic processing module is used for carrying out data standardization processing on the wind speed, the air temperature, the air pressure, the air humidity and the fan power which are collected by training and testing;

the primary model training module selects multiple single algorithms as a primary model, and performs primary model training on the data after the characteristic processing by a k-fold cross validation method;

the secondary model training module selects a neural network algorithm as a secondary model, combines the primary models and performs secondary training by combining a plurality of model judgment results;

the model tuning module is used for carrying out model accuracy test on the secondary model application test set data and continuously optimizing the model;

the model parameter adjusting module is used for adjusting model parameters to enable the evaluation indexes to reach the preset standard, outputting the model, calculating the real-time data and storing the calculation result in the server;

and the real-time prediction module is used for predicting the wind power according to the real-time weather condition and the adjusted model parameter.

9. A readable storage medium, being a non-volatile storage medium or a non-transitory storage medium, having a computer program stored thereon, wherein the computer program, when executed by a processor, performs the steps of the method according to any of the claims 1-7.

Images