Wind power generation power prediction method and system based on RF and LR twin models
By using RF and LR Gemini models to identify and fuse linear and nonlinear features in the wind power prediction model, the shortcomings of existing models in data dependence, generalization ability, real-time data sensitivity, extreme weather response, complexity and interpretability are solved, and more efficient and accurate wind power prediction is achieved.
Patent Information
- Application Number
- CN202510017470.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-06
- Publication Date
- 2025-05-30
AI Technical Summary
The existing wind power prediction models have problems such as high data dependence, poor model generalization ability, low sensitivity to real-time data, difficulty in dealing with extreme weather conditions, high model complexity and poor model interpretability.
The wind power generation power prediction method based on the RF and LR Gemini model is adopted, by identifying the linear and nonlinear characteristics in the real-time operation data, the first wind power generation predicted power and the second wind power generation predicted power are respectively output, and the fusion calculation is performed to generate the final wind power generation power.
The sensitivity and prediction accuracy of the wind power prediction model to wind power data is improved, the generalization ability of the model and the ability to cope with extreme weather is enhanced, while reducing the complexity of the model and improving interpretability.
Smart Images
Figure CN120069158A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and particularly to a wind power prediction method based on an RF and LR twin model, a wind power prediction system based on the RF and LR twin model, an electronic device, and a computer-readable storage medium. Background Art
[0002] With the rapid development of renewable energy, wind power generation has become an important part of the global energy structure. Accurate wind power prediction is of great significance for the stable operation of the power grid, the economic dispatching of wind farms, and the reduction of wind curtailment.
[0003] The training data of traditional wind power prediction models mainly consists of historical meteorological data such as wind speed, wind direction, temperature, humidity, etc., and historical power generation data of wind turbines. When using these training data for model training, generally, based on a single model such as a CNN or a statistical model, etc., feature data extraction and learning training are performed on the training data. Although a wind power prediction model can be obtained based on the training of the original model, in practical applications of the wind power prediction model, it is found that such an independently trained model has the following application drawbacks, including:
[0004] High data dependence: The model has high requirements for the accuracy of historical data. Once the data is deviated, the prediction result will also be affected.
[0005] Poor model generalization ability: Traditional models are often trained for data in specific regions or time periods, resulting in poor generalization ability in other regions or time periods. In addition, there are various structured and unstructured data in historical data. Using a single model for training and learning of historical data cannot enable the model to identify the linear characteristics of various data.
[0006] Low sensitivity to real-time data: Since the model mainly relies on historical data, its sensitivity to real-time data is low, and it is difficult to capture the real-time changes in wind power generation.
[0007] Difficulty in coping with extreme weather conditions: Under extreme weather conditions, traditional models may not be able to accurately predict wind power generation, resulting in a large deviation between the prediction result and the actual value.
[0008] High model complexity: Traditional models usually require a large amount of computing resources and time for training, and the model structure is complex, making it difficult to adjust in real time.
[0009] Poor model interpretability: Traditional models often lack interpretability and it is difficult to understand the reasons behind the prediction results.
[0010] Therefore, it is necessary to perform feature enhancement processing on the training data of the wind power prediction model, so as to improve the performance of the wind power prediction model, such as its sensitivity to wind power data, and thus improve the prediction accuracy of the wind power prediction model. Summary of the Invention
[0011] In order to solve the technical problems existing in the prior art, the present invention provides the following technical solutions:
[0012] On the one hand, a wind power prediction method based on an RF and LR twin model is provided. This method is implemented by an electronic device and includes:
[0013] S1. Collect the real-time operation data of the wind turbine generator set;
[0014] S2. Input the real-time operation data into a pre-deployed wind power prediction model, where the wind power prediction model is constructed based on a twin model: an RF model and an LR model in a model fusion training manner;
[0015] S3. Identify the linear operation data features and non-linear operation data features related to wind power in the real-time operation data through the wind power prediction model, and respectively output a first wind power prediction power matching the linear operation data features and a second wind power prediction power matching the non-linear operation data features;
[0016] S4. Perform fusion calculation on the first wind power prediction power and the second wind power prediction power to generate and output the final wind power.
[0017] Preferably, in step S2, the construction method of the wind power prediction model includes:
[0018] (1) Data collection
[0019] Collect a linear historical operation data set and a non-linear historical operation data set related to wind power;
[0020] (2) Data preprocessing
[0021] Perform data preprocessing on the linear historical operation data set and the non-linear historical operation data set in the linear and non-linear directions respectively;
[0022] (3) Feature engineering
[0023] Perform feature engineering, and respectively extract the linear data features and non-linear data features related to wind power in the linear historical operation data set and the non-linear historical operation data set;
[0024] (4) Model pre-training
[0025] Input the linear data features into the preset LR model, and let the LR model learn the linear data features related to wind power generation to obtain the first wind power generation prediction model;
[0026] Input the non-linear data features into the preset RF model, and let the RF model learn the non-linear data features related to wind power generation to obtain the second wind power generation prediction model;
[0027] (5)Model fusion training
[0028] Randomly collect several groups of real-time operation training data sets of wind turbine generators, input them into the first wind power generation prediction model and the second wind power generation prediction model respectively, and collect the wind power generation prediction results output by the two models respectively;
[0029] Collect the wind power generation prediction results and form a new training set with the real-time operation training data set;
[0030] Based on feature engineering, extract the training features of the new training set and divide them into a new feature training set and a new feature validation set according to a preset ratio;
[0031] Input the new feature training set into a preset support vector machine (SVM), train and learn the support vector machine (SVM) to generate the wind power generation prediction model;
[0032] (6)Model verification
[0033] Use the new feature validation set to verify the prediction performance of the wind power generation prediction model:
[0034] If the verification passes, deploy and apply the wind power generation prediction model;
[0035] Otherwise, repeat the above steps.
[0036] Preferably, in S3, identify the linear operation data features and non-linear operation data features related to wind power generation in the real-time operation data through the wind power generation prediction model, and respectively output the first wind power generation prediction power matching the linear operation data features and the second wind power generation prediction power matching the non-linear operation data features, including:
[0037] Organize retrieval prompt words related to linear data and configure them into a preset large language model (LLM);
[0038] Through the LLM large language model, retrieve data from the real-time operation data, screen out the real-time operation linear data related to linear data, and mark the remaining real-time operation data as real-time operation non-linear data;
[0039] Through the LLM large language model, input the real-time operation linear data into the wind power prediction model, and the wind power prediction model identifies the linear operation data characteristics in the real-time operation linear data and outputs the first wind power prediction power matching the linear operation data characteristics;
[0040] Through the LLM large language model, input the real-time operation non-linear data into the wind power prediction model, and the wind power prediction model identifies the non-linear operation data characteristics in the real-time operation non-linear data and outputs the second wind power prediction power matching the non-linear operation data characteristics;
[0041] According to the above steps, collect multiple groups of the first wind power prediction power and the second wind power prediction power.
[0042] Preferably, in S4, perform fusion calculation on the first wind power prediction power and the second wind power prediction power, generate and output the final wind power, including:
[0043] Calculate the mean value of the first wind power prediction power and the second wind power prediction power according to a preset weight ratio;
[0044] Collect the mean values of multiple groups of the first wind power prediction power and the second wind power prediction power;
[0045] Statistically calculate the final weighted average value of multiple groups of mean values as the final wind power and output it.
[0046] On the other hand, a wind power prediction system based on the RF and LR twin models is provided. The wind power prediction system based on the RF and LR twin models is used to implement the wind power prediction method based on the RF and LR twin models. The system includes:
[0047] An industrial control computer for collecting the real-time operation data of the wind turbine generator set;
[0048] A wind power prediction system is used to input the real-time operation data into a pre-deployed wind power prediction model. Through the wind power prediction model, the linear operation data features and non-linear operation data features related to wind power in the real-time operation data are identified, and the first wind power prediction power matching the linear operation data features and the second wind power prediction power matching the non-linear operation data features are respectively output;
[0049] A power generation power calculation unit is used to perform fusion calculation on the first wind power prediction power and the second wind power prediction power, generate the final wind power generation power and output it;
[0050] The industrial control computer is communicatively connected with the wind power prediction system;
[0051] The wind power prediction system is communicatively connected with the power generation power calculation unit.
[0052] On the other hand, an electronic device is provided. The electronic device includes: a processor; a memory, and computer-readable instructions are stored on the memory. When the computer-readable instructions are executed by the processor, any one of the methods in the above-mentioned wind power prediction method based on the RF and LR twin models is implemented.
[0053] On the other hand, a computer-readable storage medium is provided. At least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement any one of the methods in the above-mentioned wind power prediction method based on the RF and LR twin models.
[0054] The beneficial effects brought by the technical solutions provided in the embodiments of the present invention at least include:
[0055] The wind power prediction model of the present invention is constructed by fusing and training two models: the RF model and the LR model. The linear operation data features and non-linear operation data features related to wind power in the real-time operation data are identified, and the first wind power prediction power matching the linear operation data features and the second wind power prediction power matching the non-linear operation data features are respectively output; the first wind power prediction power and the second wind power prediction power are subjected to fusion calculation to generate the final wind power generation power and output it. It can use the fusion model to fuse and identify the linear or non-linear features affecting the power generation power, thereby improving the performance of the wind power prediction model such as the sensitivity to wind power generation data, and thus improving the prediction accuracy of the wind power prediction model. Description of the Drawings
[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.
[0057] Figure 1 is a flowchart of a wind power prediction method based on an RF and LR twin model provided by an embodiment of the present invention;
[0058] Figure 2 is a schematic diagram of model fusion training provided by an embodiment of the present invention;
[0059] Figure 3 is a block diagram of a wind power prediction system based on an RF and LR twin model provided by an embodiment of the present invention;
[0060] Figure 4 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Detailed implementation manners
[0061] The following will describe the technical solutions in the present invention with reference to the accompanying drawings.
[0062] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to give examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly, the use of the word "example" is intended to present concepts in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.
[0063] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.
[0064] In the embodiments of the present invention, sometimes subscripts such as W 1 may be miswritten as non-subscript forms such as W1. When the difference is not emphasized, the meanings they express are the same.
[0065] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.
[0066] An embodiment of the present invention provides a wind power prediction method based on an RF and LR twin model. This method can be implemented by an electronic device, which can be a terminal or a server. As Figure 1 shown in the flowchart of the wind power prediction method based on the RF and LR twin model, the processing flow of this method can include the following steps:
[0067] S1. Collect the real-time operation data of the wind turbine generator set;
[0068] S2. Input the real-time operation data into a pre-deployed wind power prediction model. Among them, the wind power prediction model is constructed by fusing and training two sub-models: the RF model and the LR model;
[0069] S3. Identify the linear operation data features and non-linear operation data features related to wind power in the real-time operation data through the wind power prediction model, and respectively output the first wind power prediction power matching the linear operation data features and the second wind power prediction power matching the non-linear operation data features;
[0070] S4. Perform a fusion calculation on the first wind power prediction power and the second wind power prediction power to generate and output the final wind power.
[0071] The present invention constructs a wind power prediction model based on a twin model (the RF model and the LR model are constructed by fusing and training), identifies the linear operation data features and non-linear operation data features related to wind power in the real-time operation data, and respectively outputs the first wind power prediction power matching the linear operation data features and the second wind power prediction power matching the non-linear operation data features; performs a fusion calculation on the first wind power prediction power and the second wind power prediction power to generate and output the final wind power.
[0072] The real-time operation data of the wind turbine generator set includes: collecting historical meteorological data such as wind speed, wind direction, temperature, humidity, and air pressure, as well as historical operation data of the wind turbine generator set (such as wind power output, peak power generation voltage, peaks and troughs of output power, times of peaks and troughs, and corresponding fault records, etc.). And the historical operation data with a linear relationship between power generation and the following factors: load current, load voltage, generator speed, generator temperature, water turbine guide vane opening (for hydroelectric power), fuel consumption rate (for thermal power), wind speed (for wind power), radiation intensity (for solar power).
[0073] Preferably, in step S2, the construction method of the wind power prediction model includes:
[0074] (1)Data collection
[0075] Collect linear historical operation datasets and non-linear historical operation datasets related to wind power generation
[0076] Wind power generation is affected by a variety of linear and non-linear factors. To accurately predict and evaluate wind power generation, these factors need to be comprehensively considered, and appropriate models and methods are adopted for analysis and calculation
[0077] The main difference between linear data and non-linear data lies in the growth or change pattern of the data. Linear data usually follows a fixed ratio or proportional relationship, that is, the change between data points is linear. Non-linear data does not follow this fixed proportional relationship, and its growth or change may be curvilinear, periodic or other complex forms
[0078] Example of linear data: If for every one unit increase, the data volume also increases by one unit correspondingly, then this is linear data
[0079] Example of non-linear data: If the data volume increases in a square or exponential form as a certain factor increases, then this is non-linear data
[0080] In data analysis, the methods and techniques for dealing with linear data and non-linear data are also different. Linear data can usually be analyzed through simple linear regression models, while non-linear data may require more complex statistical or machine learning models to capture its complex change patterns
[0081] Specifically, the data is as follows
[0082] Wind power generation is affected by a variety of factors, which can be divided into linear operation data and non-linear operation data. The following are some examples of these two types of data
[0083] Linear operation data
[0084] Linear operation data usually refers to factors that have a linear or approximately linear relationship with wind power generation, and the changes of these factors will directly affect the power generation in proportion
[0085] Wind speed: Wind speed is the most direct factor affecting wind power generation. Generally, the higher the wind speed, the greater the output power of the wind turbine. This relationship can be regarded as linear within a certain range
[0086] Air density: Air density also affects wind power generation. The greater the air density, the greater the wind force on the wind turbine blades, and thus the greater the power generated. Air density is related to factors such as temperature and altitude
[0087] Wind direction: Although the impact of wind direction on wind power generation is not as direct as that of wind speed, changes in wind direction can also affect the output power of wind turbines. Especially when the wind direction is inconsistent with the arrangement direction of the wind turbine blades, it will lead to a power drop.
[0088] Non-linear operation data
[0089] Non-linear operation data refers to factors that have a non-linear relationship with wind power generation. Changes in these factors do not directly affect the power generation proportionally, but may affect it in a more complex way.
[0090] Turbulence intensity: Turbulence is the rapid change of wind speed and direction, which will affect the stability and output power of wind turbines. Turbulence intensity is related to factors such as terrain, obstacles, and meteorological conditions, and its impact on power generation is non-linear.
[0091] Wind turbine characteristics: Different wind turbines have different power curves and efficiency characteristics. These characteristics will affect the output power of wind turbines at different wind speeds, and this relationship is usually non-linear.
[0092] Temperature: Temperature will affect air density and the operating efficiency of wind turbines, thus indirectly affecting power generation. However, the relationship between temperature and power generation is usually not a simple linear relationship.
[0093] Humidity: Humidity may also have a certain impact on wind power generation, especially under extreme humidity conditions. Humidity may affect the physical properties of air, such as density and viscosity, and thus affect the performance of wind turbines. However, this impact is usually non-linear and may interact with other factors.
[0094] Maintenance status: The maintenance status of wind turbines will also affect their output power. For example, blade wear, gearbox failure, or motor aging may all lead to a power drop. This impact is usually non-linear and difficult to describe with a simple mathematical model.
[0095] Data collection can be retrieved from the system background database, and it is specifically collected by the administrator.
[0096] (2)Data preprocessing
[0097] Perform data preprocessing on the linear historical operation dataset and the non-linear historical operation dataset in the linear and non-linear directions respectively;
[0098] Clean the data, including removing outliers, filling in missing values, normalization processing, etc., to ensure data quality.
[0099] (3)Feature engineering
[0100] Perform feature engineering to extract the linear data features and non-linear data features related to wind power generation from the linear historical operation dataset and the non-linear historical operation dataset respectively;
[0101] Feature extraction methods such as data analysis and mining techniques can be used to extract corresponding data features from the linear operation data and non-linear operation data that affect wind power generation. The main methods are as follows:
[0102] Extracting data features from the linear operation data and non-linear operation data that affect wind power generation is a key step in wind power generation prediction and modeling. The following are some common methods:
[0103] 1) Feature extraction methods for linear operation data
[0104] Statistical features:
[0105] Mean value: Calculate the average values of linear factors such as wind speed and air density to reflect their overall levels.
[0106] Standard deviation: Calculate the degree of dispersion of the data to understand the fluctuation of the data.
[0107] Maximum value / minimum value: Find the extreme values in the data and analyze their potential impacts on wind power generation.
[0108] Time series analysis:
[0109] Autocorrelation function: Analyze the correlation of time series data such as wind speed at different time lags to identify periodic or trend changes.
[0110] Moving average: Eliminate noise by smoothing the data to highlight the main trend.
[0111] Linear regression analysis:
[0112] Use a linear regression model to analyze the linear relationship between wind speed and wind power generation, and determine the coefficient and intercept.
[0113] 2) Feature extraction methods for non-linear operation data
[0114] Non-linear transformation:
[0115] Perform non-linear transformation on the original data, such as logarithmic transformation, square root transformation, etc., to reveal hidden non-linear relationships.
[0116] Spectrum analysis:
[0117] Perform spectral analysis on time series data such as wind speed to identify the frequency components therein and understand the periodic changes of non-linear factors such as turbulence.
[0118] Wavelet analysis:
[0119] Use wavelet transform to analyze the time-frequency characteristics of data and capture local changes and non-linear features in signals such as wind speed.
[0120] In addition to the above methods, machine learning algorithms can also be used:
[0121] Use machine learning algorithms such as decision trees, random forests, and support vector machines to automatically learn and extract non-linear features from data.
[0122] Deep learning algorithms, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), can handle more complex data structures and extract deep non-linear features.
[0123] Chaos theory:
[0124] For data showing chaotic characteristics (such as some turbulence data), methods of chaos theory can be used for analysis to extract their non-linear dynamic features.
[0125] Feature engineering:
[0126] Combining domain knowledge and data characteristics, manually construct new features, such as the square and cube of wind speed, etc., to capture non-linear relationships.
[0127] Use feature selection algorithms to select the most useful features for wind power prediction from a large number of candidate features.
[0128] In practical applications, it is often necessary to combine multiple methods to extract data features. For example, statistical analysis of linear factors can be performed first, then spectral analysis or wavelet analysis of non-linear factors, and finally machine learning algorithms are used to integrate all features for modeling and prediction. The specific method is selected by the administrator for feature extraction.
[0129] In addition, data preprocessing is also a very important step, including data cleaning, missing value handling, outlier detection, etc., to ensure that the extracted features can accurately reflect the true situation of the data.
[0130] Next, linear and non-linear model training will be carried out respectively. Combined with the attached Figure 3 as shown:
[0131] (4) Model pre-training
[0132] Input the linear data features into the preset LR model, and let the LR model learn the linear data features related to wind power generation to obtain the first wind power generation prediction model;
[0133] Input the non-linear data features into the preset RF model, and let the RF model learn the non-linear data features related to wind power generation to obtain the second wind power generation prediction model;
[0134] For linear and non-linear training models, it is necessary to select the corresponding model for learning and training of the corresponding type of data features.
[0135] In this embodiment:
[0136] RF model: Using the random forest algorithm, construct multiple decision trees as base learners, and improve the prediction accuracy through the way of ensemble learning. The RF model is good at dealing with non-linear relationships and high-dimensional data, and can capture the complex relationship between wind power generation and meteorological factors.
[0137] LR model: Although logistic regression is usually used for classification problems, it can be used as a linear model here to capture the linear relationship between wind power generation and main power factors. Through regularization techniques, the LR model can effectively prevent overfitting and improve the generalization ability of the model.
[0138] Therefore, use the RF model and the LR model to perform feature training and learning on non-linear data features and linear data features respectively, so that the model has good feature recognition and learning abilities, and makes the model output results more accurate.
[0139] Subsequent model fusion
[0140] (5)Model fusion training
[0141] Randomly collect several groups of real-time operation training data sets of wind turbine generators, input them into the first wind power generation prediction model and the second wind power generation prediction model respectively, and collect the wind power generation prediction results output by the two models respectively;
[0142] Collect the wind power generation prediction results and form a new training set with the real-time operation training data set;
[0143] Based on feature engineering, extract the training features of the new training set and divide them into a new feature training set and a new feature validation set according to a preset ratio;
[0144] Input the new feature training set into a preset support vector machine (SVM), train and learn the support vector machine (SVM) to generate the wind power generation prediction model;
[0145] After training and learning the data features of the corresponding type using the twin models, the two models can respectively identify the feature data of the corresponding type. Then, the new training data is input into the two prediction models respectively, and the two constructed prediction models are used to predict the new training data. The model output results of the two models are collected, and then the model output results are integrated to form a new training set. Feature engineering is performed again, and a support vector machine is used for model training to train and generate a wind power prediction model.
[0146] Adopt the Stacking strategy, and use the prediction results of the RF model and the LR model as new features to input into a meta-learner (such as linear regression or support vector machine) for the final wind power prediction. This fusion method can make full use of the advantages of the two models and improve the prediction accuracy. The main steps are as follows:
[0147] Extract new features
[0148] Use the prediction results of the RF model and the LR model as new features.
[0149] These new features can be used alone or in combination with other original features.
[0150] Construct an SVM model
[0151] Use the support vector machine algorithm to construct a new prediction model.
[0152] Use the extracted new features (prediction results of RF and LR) as the input features of the SVM model.
[0153] If other original features are combined, these features also need to be input into the SVM model together.
[0154] Train the SVM model
[0155] Use the training data set to train the SVM model.
[0156] Optimize the model performance by adjusting the hyperparameters of the SVM (such as C, gamma, etc.).
[0157] Model evaluation and verification
[0158] Evaluate the performance of the SVM model on the validation data set, and metrics such as mean squared error (MSE), root mean squared error (RMSE), and accuracy can be used. (Model verification can be specifically implemented by the administrator)
[0159] If the model performance is not good, you can return and adjust the feature selection or the hyperparameters of the SVM.
[0160] Model deployment and prediction
[0161] Deploy the trained SVM model to actual applications.
[0162] Input new data into the model to obtain the prediction results of wind power generation.
[0163] (6) Model verification
[0164] Use the new feature validation set to verify the prediction performance of the wind power generation prediction model:
[0165] If the verification passes, deploy and apply the wind power generation prediction model;
[0166] Conversely, repeat the above steps.
[0167] In model training, verifying the usage performance of the training model is a key step to ensure the accuracy and reliability of the model, mainly including the following methods:
[0168] Hold-out method : Divide the dataset into a training set and a test set. The training set is used for model training, and the test set is used to evaluate the model performance, ensuring that the data distributions of the test set and the training set are consistent.
[0169] Cross-validation method : Such as k-fold cross-validation, evenly divide the dataset into k subsets, take turns as the test set, and the rest as the training set. The model is trained k times, and the average value of the performance evaluation metrics is taken.
[0170] Evaluation metrics : Use metrics such as accuracy, recall rate, F1 score, mean squared error, ROC curve, and AUC value to measure the performance of the model.
[0171] Through these methods, the prediction ability and generalization performance of the model can be comprehensively evaluated to ensure the stability and reliability of the model in actual applications.
[0172] Preferably, in S3, identify the linear operation data features and non-linear operation data features related to wind power generation in the real-time operation data through the wind power generation prediction model, and respectively output the first wind power generation prediction power matching the linear operation data features and the second wind power generation prediction power matching the non-linear operation data features, including:
[0173] Organize retrieval prompt words related to linear data and configure them to a preset large language model (LLM);
[0174] Through the LLM, perform data retrieval on the real-time operation data, screen out the real-time operation linear data related to linear data, and mark the remaining real-time operation data as real-time operation non-linear data;
[0175] Through the LLM (Large Language Model), input the real-time running linear data into the wind power prediction model. The wind power prediction model identifies the linear operation data features in the real-time running linear data and outputs the first wind power prediction power matching the linear operation data features.
[0176] Through the LLM (Large Language Model), input the real-time running non-linear data into the wind power prediction model. The wind power prediction model identifies the non-linear operation data features in the real-time running non-linear data and outputs the second wind power prediction power matching the non-linear operation data features.
[0177] According to the above steps, collect multiple groups of the first wind power prediction power and the second wind power prediction power.
[0178] The LLM (Large Language Model) can be self-developed by the user or call a third-party LLM (Large Language Model), such as Wenxin Yiyan of Baidu.
[0179] Here, the LLM is used for data screening to quickly classify data into linear and non-linear types. The LLM can retrieve corresponding types of data based on prompt words. The prompt words contain the retrieval keywords and requirements for the current screening target. For example, the retrieval prompt words contain the keyword of the data type defined linearly (keywords related to linear data, such as wind speed) and the retrieval logic of the linear keyword (the value of the corresponding linear keyword needs to be extracted). Input the prompt words into the LLM, and let the LLM retrieve linear data from the library based on the retrieval prompt words. The retrieved data that meets the prompt words is used as linear data, and the remaining data is used as non-linear data. Therefore, the data classification and application efficiency can be greatly improved.
[0180] To retrieve real-time running data through the LLM (Large Language Model) and screen out linear real-time running linear data, the following steps can be taken:
[0181] Data collection and preprocessing:
[0182] First, collect data from the real-time running system, which may include sensor readings, system logs, user behavior records, etc.
[0183] Preprocess the collected data, including cleaning the data (removing invalid or incorrect data), normalizing (scaling the data to a unified range), denoising, etc.
[0184] Data feature extraction:
[0185] Extract features from the preprocessed data, and these features should be related to the linear relationship to be found.
[0186] Feature extraction may include timestamps, specific measurements, event identifiers, etc.
[0187] Construct a query statement:
[0188] Based on the characteristics of the linear data to be screened, construct an appropriate query statement or instruction (keywords) for the LLM.
[0189] The query statement should clearly specify the data type to be searched, time range, feature conditions (logic), etc.
[0190] Utilize the LLM for data retrieval:
[0191] Input the query statement into the LLM and utilize the natural language processing ability of the LLM to parse and execute the query.
[0192] The LLM will return a dataset or data fragment that matches the query conditions.
[0193] Linear data screening:
[0194] Further screen out the data with a linear relationship from the data returned by the LLM.
[0195] This requires applying mathematical or statistical methods, such as linear regression analysis, to determine which data points or datasets exhibit a linear trend.
[0196] Verification and confirmation:
[0197] Verify the screened linear data to confirm that it indeed conforms to a linear relationship.
[0198] This may include visual analysis (such as plotting scatter plots or line graphs), calculating correlation coefficients, etc.
[0199] Storage and application:
[0200] Store the verified linear data in an appropriate database or data warehouse for subsequent analysis or application.
[0201] According to specific requirements, these linear data can be used in scenarios such as prediction, monitoring, optimization, etc.
[0202] The steps of collecting multiple sets of the first wind power prediction power and the second wind power prediction power should be understood in combination with the functions of the above wind power prediction model (incorporating the capabilities of 2 prediction models), and will not be elaborated here.
[0203] Preferably, S4. Perform a fusion calculation on the first wind power prediction power and the second wind power prediction power to generate and output the final wind power, including:
[0204] Calculate the mean value of the first wind power generation prediction power and the second wind power generation prediction power according to a preset weight ratio;
[0205] Collect multiple groups of the mean values of the first wind power generation prediction power and the second wind power generation prediction power;
[0206] Statistically calculate the final weighted average value of multiple groups of mean values as the final wind power generation power and output it.
[0207] For each group of the first wind power generation prediction power and the second wind power generation prediction power respectively identified and output by the wind power generation power prediction model for linear data and non-linear data, calculate the mean value to obtain the final wind power generation power.
[0208] Here, in order to improve data accuracy, multiple groups of mean values of the two values output by the wind power generation power prediction model can be collected, and the mean value calculation is performed again, and this is used as the final wind power generation power and output, so as to improve the accuracy of the model for predicting the power generation power. The weight for calculating the mean value of the two prediction powers can be calculated according to 0.5 respectively. It is also possible to determine the weight ratio according to the influence granularity of linear or non-linear data on the power generation power. For example, if it is found according to experience that non-linear data has a greater influence on the wind power generation power, then a weight configuration of about 0.7 is performed on the second wind power generation prediction power, so that the result is biased towards the influence angle of non-linear data.
[0209] Figure 3 It is a block diagram of a wind power generation power prediction system based on an RF and LR twin model shown according to an exemplary embodiment. This system is used for a wind power generation power prediction method based on an RF and LR twin model. Refer to Figure 3 , this system includes an industrial control computer 310, a wind power generation power prediction system 320, and a power generation power calculation unit 330. Among them:
[0210] The industrial control computer is used to collect the real-time operation data of the wind turbine generator set;
[0211] The wind power generation power prediction system is used to input the real-time operation data into a pre-deployed wind power generation power prediction model, identify the linear operation data characteristics and non-linear operation data characteristics related to the wind power generation power in the real-time operation data through the wind power generation power prediction model, and respectively output a first wind power generation prediction power matching the linear operation data characteristics and a second wind power generation prediction power matching the non-linear operation data characteristics;
[0212] The power generation power calculation unit is used to perform fusion calculation on the first wind power generation prediction power and the second wind power generation prediction power, generate the final wind power generation power and output it;
[0213] The industrial control computer is communicatively connected to the wind power prediction system;
[0214] The wind power prediction system is communicatively connected to the power generation calculation unit.
[0215] Figure 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. As Figure 4 shown, the electronic device may include the above-mentioned Figure 3 wind power prediction system based on the RF and LR twin models shown. Optionally, the electronic device 410 may include a first processor 2001.
[0216] Optionally, the electronic device 410 may further include a memory 2002 and a transceiver 2003.
[0217] Among them, the first processor 2001, the memory 2002, and the transceiver 2003 may be connected through a communication bus, for example.
[0218] Next, a specific introduction to the various components of the electronic device 410 will be given in conjunction with Figure 4 :
[0219] Among them, the first processor 2001 is the control center of the electronic device 410, which may be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or may be an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention, such as: one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).
[0220] Optionally, the first processor 2001 may execute various functions of the electronic device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.
[0221] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 4 CPU0 and CPU1 shown in
[0222] In a specific implementation, as an embodiment, the electronic device 410 may also include multiple processors, such asFigure 4 The first processor 2001 and the second processor 2004 shown in the figure. Each of these processors can be a single-CPU or a multi-CPU. Here, the processor can refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).
[0223] Among them, the memory 2002 is used to store the software program for implementing the solution of the present invention and is controlled by the first processor 2001 for execution. The specific implementation manner can refer to the above method embodiments and will not be elaborated here.
[0224] Optionally, the memory 2002 can be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or can also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 can be integrated with the first processor 2001 or can exist independently and is coupled to the first processor 2001 through the interface circuit of the electronic device 410 ( Figure 4 not shown in the figure), and the embodiments of the present invention do not make specific limitations on this.
[0225] The transceiver 2003 is used to communicate with a network device or with a terminal device.
[0226] Optionally, the transceiver 2003 can include a receiver and a transmitter ( Figure 4 not shown separately in the figure). Among them, the receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.
[0227] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or can exist independently and is coupled to the first processor 2001 through the interface circuit of the electronic device 410 ( Figure 4 not shown in the figure), and the embodiments of the present invention do not make specific limitations on this.
[0228] It should be noted that Figure 4 the structure of the electronic device 410 shown in [the figure] does not limit the router. The actual knowledge structure recognition device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0229] In addition, the technical effects of the electronic device 410 can refer to the technical effects of the wind power prediction method based on the RF and LR twin models described in the above method embodiments, which will not be elaborated here.
[0230] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0231] It should also be understood that the memory in the embodiments of the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DR RAM).
[0232] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable systems. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.
[0233] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. Additionally, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be understood specifically with reference to the context before and after.
[0234] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.
[0235] It should be understood that in various embodiments of the present invention, the magnitudes of the sequence numbers of the above processes do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
[0236] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.
[0237] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, systems, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.
[0238] In several embodiments provided by the present invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of systems or units can be electrical, mechanical, or other forms.
[0239] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0240] In addition, the functional units in each embodiment of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.
[0241] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.
[0242] As described above, the above are only specific implementation manners of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.
Claims
1. A wind power generation prediction method based on RF and LR twin models, characterized in that: The method comprises: S1. Collect real-time operation data of wind turbines; S2. Inputting the real-time operation data into a pre-deployed wind power generation prediction model, wherein the wind power generation prediction model is constructed based on a twin model: an RF model and an LR model by means of model fusion training; S3, identifying the linear operation data characteristics and the nonlinear operation data characteristics of the wind power generation in the real-time operation data through the wind power generation prediction model, and outputting the first wind power generation prediction power matching the linear operation data characteristics and the second wind power generation prediction power matching the nonlinear operation data characteristics respectively; S4. Perform a fusion calculation on the first wind power generation prediction power and the second wind power generation prediction power to generate and output a final wind power generation power.
2. The wind power generation prediction method based on the RF and LR twin models according to claim 1 is characterized in that: In step S2, the method for constructing the wind power generation prediction model includes: (1) Data collection Collecting linear historical operation data sets and nonlinear historical operation data sets related to wind power generation; (2) Data preprocessing Performing data preprocessing in linear and nonlinear directions on the linear historical operation data set and the nonlinear historical operation data set respectively; (3) Feature Engineering Performing feature engineering to extract linear data features and nonlinear data features related to wind power generation in the linear historical operation data set and the nonlinear historical operation data set respectively; (4) Model pre-training Inputting the linear data features into the preset LR model, learning the linear data features related to wind power generation through the LR model, and obtaining a first wind power generation prediction model; Inputting the nonlinear data features into the preset RF model, and learning the nonlinear data features related to wind power generation through the RF model to obtain a second wind power generation prediction model; (5) Model fusion training Randomly collect a number of groups of real-time operation training data sets of wind turbine generator sets, input them into the first wind power generation prediction model and the second wind power generation prediction model respectively, and collect the wind power generation prediction results output by the two models respectively; Collecting the wind power generation prediction results and forming a new training set with the real-time running training data set; Based on feature engineering, the training features of the new training set are extracted and divided into a new feature training set and a new feature verification set according to a preset ratio; Inputting the new feature training set into a preset support vector machine (SVM), training and learning the support vector machine (SVM), and generating the wind power generation prediction model; (6) Model validation The prediction performance of the wind power generation prediction model is verified using the new feature verification set: If the verification is successful, the wind power generation prediction model is deployed and applied; Otherwise, repeat the above steps.
3. The wind power generation prediction method based on the RF and LR twin models according to claim 1 is characterized in that: S3, identifying the linear operation data features and the nonlinear operation data features of the wind power generation in the real-time operation data through the wind power generation prediction model, and outputting the first wind power generation prediction power matching the linear operation data features and the second wind power generation prediction power matching the nonlinear operation data features, respectively, including: Organize the search prompt words related to linear data and configure them to the preset LLM large language model; Through the LLM large language model, data retrieval is performed on the real-time operation data, real-time operation linear data related to linear data is screened out, and the remaining real-time operation data is marked as real-time operation non-linear data; The real-time linear operation data is input into the wind power generation prediction model through the LLM large language model, and the wind power generation prediction model identifies the linear operation data features in the real-time linear operation data and outputs the first wind power generation prediction power matching the linear operation data features; The real-time nonlinear operation data is input into the wind power generation prediction model through the LLM large language model, and the wind power generation prediction model identifies the nonlinear operation data features in the real-time nonlinear operation data and outputs a second wind power generation prediction power matching the nonlinear operation data features; According to the above steps, multiple groups of the first wind power generation predicted power and the second wind power generation predicted power are collected.
4. The wind power generation prediction method based on the RF and LR twin models according to claim 3 is characterized in that: S4, performing a fusion calculation on the first wind power generation prediction power and the second wind power generation prediction power to generate and output a final wind power generation power, including: Calculating the average of the first wind power generation prediction power and the second wind power generation prediction power according to a preset weight ratio; Collecting average values of multiple groups of the first wind power generation prediction power and the second wind power generation prediction power; The final weighted average of the multiple groups of mean values is calculated and output as the final wind power generation power.
5. A wind power generation power prediction system based on RF and LR twin models, wherein the wind power generation power prediction system based on RF and LR twin models is used to implement the wind power generation power prediction method based on RF and LR twin models as claimed in any one of claims 1 to 4, characterized in that: The system comprises: Industrial computer, used to collect real-time operation data of wind turbines; A wind power generation power prediction system, used for inputting the real-time operation data into a pre-deployed wind power generation power prediction model, identifying linear operation data features and non-linear operation data features of wind power generation in the real-time operation data through the wind power generation power prediction model, and outputting a first wind power generation prediction power matching the linear operation data features and a second wind power generation prediction power matching the non-linear operation data features respectively; A power generation calculation unit, used for performing a fusion calculation on the first wind power generation prediction power and the second wind power generation prediction power to generate and output a final wind power generation power; The industrial computer is communicatively connected with the wind power generation power prediction system; The wind power generation power prediction system is communicatively connected to the power generation calculation unit.
6. An electronic device, characterized in that: The electronic device comprises: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 4 is implemented.
7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program codes, which can be called by a processor to execute the method according to any one of claims 1 to 4.