Wind power plant large component temperature correlation analysis method

Abstract

The invention relates to the field of operation and maintenance management of a wind power plant, and discloses a wind power plant large component temperature correlation analysis method which comprises the following steps: collecting temperature data of wind power plant large components, setting parameters of temperature correlation analysis, calculating correlation indexes among the temperatures of the large components, and storing the correlation indexes in an original data set; identifying components with abnormal temperature according to the temperature data of the large components of the wind power plant; performing feature extraction on the temperature data by adopting statistical characteristics of the temperature data, such as a mean value, a variance and a range, and storing the temperature data to an original data set; constructing a correlation analysis model, and performing model training and parameter optimization by adopting data in the original data set; and the model can perform correlation analysis on the temperature change of the large part of the wind power plant so as to identify the potential temperature anomaly. According to the method, the accuracy and the efficiency of temperature anomaly identification of the large components of the wind power plant can be improved, so that potential equipment faults can be prevented, and stable operation of the wind power plant is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of operation and maintenance management of wind farms, and in particular to a temperature correlation analysis method for large components of wind farms. Background Art

[0002] In the operation and maintenance of wind farms, temperature monitoring of large components is an important means to ensure the normal operation of equipment and prevent failures. Existing technologies usually rely on a single temperature sensor to monitor the temperature of components. Although this method can provide real-time temperature data, it often lacks in-depth analysis of the correlation behind temperature changes. In addition, traditional temperature monitoring systems often only focus on temperature anomalies of a single component, while ignoring the mutual influence of temperature changes between components, which may lead to misjudgment of the cause of the failure and waste of maintenance resources.

[0003] In the temperature monitoring of large components in wind farms, the collection and processing of temperature data is the basis, but how to extract valuable information from massive data and how to use this information for effective fault prediction and health management are challenges faced by existing technologies. Existing data processing methods often rely on simple statistical analysis, such as calculating mean and variance. Although these methods can provide some basic temperature characteristics, they are difficult to capture the complex patterns and potential correlations of temperature changes.

[0004] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: when processing the temperature data of large components in a wind farm, the prior art often ignores the relationship between the temperature changes between components, resulting in the inability to fully identify and analyze the causes of temperature anomalies; at the same time, the existing monitoring system lacks effective algorithm support, making it difficult to accurately predict and respond to temperature anomalies in a timely manner, which to a certain extent limits the level of intelligence and automation in the operation and maintenance management of wind farms. Summary of the invention

[0005] The embodiment of the present invention aims to provide a temperature correlation analysis method for large components of a wind farm, so as to solve the technical problems raised in the prior art.

[0006] The embodiment of the present invention solves the technical problem by adopting the following technical solution:

[0007] A method for analyzing the temperature correlation of large components in a wind farm is provided, comprising:

[0008] Step S1, collecting temperature data of large components of the wind farm, setting parameters for temperature correlation analysis, calculating correlation indicators between the temperatures of the large components and saving them to the original data set;

[0009] Step S2: identifying components with abnormal temperatures based on the temperature data of large components of the wind farm; using the statistical characteristics of the temperature data, including mean, variance and range, to extract features from the temperature data and save them to the original data set;

[0010] Step S3: construct an association analysis model and use the data in the original data set for model training and parameter optimization;

[0011] This enables the model to correlate temperature variations in large wind farm components to identify potential temperature anomalies.

[0012] Furthermore, in the parameters of the temperature correlation analysis, the temperature data is processed by using a time series analysis method so that the time resolution of the temperature data matches the operation cycle of the wind farm.

[0013] Furthermore, the step S2 includes:

[0014] Step S201, calculating the statistical characteristics of temperature according to the temperature data of the major components of the wind farm;

[0015] Step S202: Identify components with abnormal temperatures by setting temperature thresholds;

[0016] Step S203: using a correlation analysis method to calculate the correlation between the temperatures of the large components based on the Pearson correlation coefficient;

[0017] Step S204: Save the calculated correlation index and the temperature data together in the original data set.

[0018] Furthermore, the temperature T of the ith component at time t of temperature anomaly identification is i (t) is calculated as:

[0019]

[0020] Among them, T i (t) is the temperature of the i-th component at time t; is the mean temperature of component i; Θ i Anomaly threshold set for component i; i Indicates whether the i-th component is abnormal at time t.

[0021] Further, the process of determining the temperature correlation between large components is:

[0022] 1) For n large wind farm components, calculate the temperature correlation of all components;

[0023]

[0024] Among them, r ijis the correlation coefficient between component i and component j; T i (t) and T j (t) is the temperature of component i and component j at time t; and is the mean temperature of components i and j;

[0025] 2) According to the correlation coefficient r ij , determine the temperature correlation between large components.

[0026] Furthermore, in step S3, by constructing an association analysis model, a genetic algorithm is used to select the optimal parameter combination in the model, and training is performed to establish a model in which input features include the temperature of large components, statistical characteristics and correlation indicators, and the output feature is temperature anomalies; based on the temperature correlation between large components, temperature anomalies of large components in wind farms are identified.

[0027] Furthermore, the association analysis model is a random forest model.

[0028] Furthermore, the specific process of constructing the random forest model in step S3 includes:

[0029] Step S301: collect the original data and calculate the temperature T of the i-th component at time t. i (t), statistical characteristics and correlation indicators as a sample feature, and the temperature anomaly label Y i (t) is used as the sample label; the samples in the original data set are divided into a training sample set and a test sample set;

[0030] Step S302: using the training sample set data as the input of the random forest model, and using a genetic algorithm to select the optimal parameter combination of the random forest model;

[0031] Step S303: Use the optimal parameter combination to input the test sample set data into the random forest model, and calculate the error between the model output data and the sample label until the evaluation indicators of the model on the training sample set and the test sample set meet the set requirements;

[0032] Step S304: Save the trained model and parameter combination.

[0033] Furthermore, the process of selecting the optimal parameter combination of the random forest model using the genetic algorithm includes:

[0034] 1) Define the parameter space; determine the random forest model parameters to be optimized and the corresponding value range;

[0035] 2) Define the evaluation index and fitness function: Select the accuracy A as the evaluation index to measure the performance of the random forest model under given parameters; the fitness function is the inverse of the error, that is,

[0036]

[0037] Among them, Parameters represents the model parameters; Error Rate represents the error rate of the model;

[0038] 3) Iterative optimization: The optimization process is repeated according to the specified number of iterations. In each iteration, the algorithm will adjust the parameters according to the previous evaluation results until the specified stopping condition is reached.

[0039] 4) Obtaining the best parameters: After the optimization is completed, the best parameter combination is selected from the obtained results, which corresponds to the random forest model with the best performance index.

[0040] Furthermore, the objective function of the random forest model training process includes a loss function and a regularization term, and the specific formula is:

[0041]

[0042] Among them, L(Θ) is the loss function, Θ is the model parameter, N is the number of samples, l is the loss function, y i is the true label of the sample, is the predicted label, λ is the regularization coefficient, and R(Θ) is the regularization term.

[0043] The above-mentioned embodiments of the present invention have at least the following beneficial effects: the temperature anomaly of each component in the wind farm can be effectively identified by using the temperature correlation analysis method of large components of the present invention. By collecting and analyzing the temperature data of large components, calculating correlation indicators, and combining statistical characteristics such as mean, variance and range, the method can reveal the intrinsic connection between temperature changes between components, thereby providing a scientific basis for fault prediction and health management of wind farms. In addition, the method can further improve the accuracy and efficiency of temperature anomaly identification by constructing an association analysis model, especially adopting a random forest model, and using a genetic algorithm for parameter optimization.

[0044] In addition, the embodiments of the present invention can more accurately determine whether a component is abnormal by refining the process of temperature anomaly identification, such as setting temperature thresholds and using the Pearson correlation coefficient. This method can not only improve the intelligence level of wind farm operation and maintenance, but also help reduce equipment failures and downtime caused by temperature anomalies, thereby reducing operation and maintenance costs and improving the overall operation efficiency of the wind farm. Through this method, wind farm operation and maintenance personnel can more proactively make fault prevention and maintenance decisions to ensure the stable and efficient operation of the wind farm. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] One or more embodiments are exemplarily described by the figures in the corresponding drawings, and these exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0046] Figure 1 It is a schematic flow chart of the temperature correlation analysis method for large components of a wind farm provided by the present invention. DETAILED DESCRIPTION

[0047] In order to facilitate the understanding of the present invention, the present invention is described in more detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that when an element is described as "connecting" another element, it can be directly on another element, or there can be one or more centered elements therebetween. The orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "upper end", "lower end", "top" and "bottom" used in this specification is based on the orientation or positional relationship shown in the accompanying drawings, only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0048] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0049] Combine the following Figure 1 The wind farm large component temperature correlation analysis method 100 provided in the embodiment of the present application is described in detail through a specific embodiment.

[0050] Figure 1 The following is a flow chart of a method for analyzing the temperature correlation of large components in a wind farm provided by the present invention. The method for analyzing the temperature correlation of large components in a wind farm provided by one embodiment of the present invention comprises:

[0051] Step S1, collecting temperature data of large components of the wind farm, setting parameters for temperature correlation analysis, calculating correlation indicators between the temperatures of the large components and saving them to the original data set;

[0052] Step S2: identifying components with abnormal temperatures based on the temperature data of large components of the wind farm; using the statistical characteristics of the temperature data, including mean, variance and range, to extract features from the temperature data and save them to the original data set;

[0053] Step S3: construct an association analysis model and use the data in the original data set for model training and parameter optimization;

[0054] This enables the model to correlate temperature variations in large wind farm components to identify potential temperature anomalies.

[0055] It should be noted that collecting temperature data of large components in wind farms is the basis of the entire analysis method, which involves real-time temperature monitoring of key components in wind farms to obtain necessary data for subsequent analysis. This step requires a complete temperature monitoring system that can cover all key components in wind farms to ensure the comprehensiveness and accuracy of data.

[0056] Specifically, the collection of temperature data can be achieved through sensors installed on large components of wind farms, which can measure and record the temperature of components in real time. In terms of parameter setting, the frequency and duration of data collection need to be determined according to the actual situation of the wind farm and the characteristics of the components to ensure that the collected data can reflect the actual operating status of the components. At the same time, the parameter setting of temperature correlation analysis needs to take into account the operating cycle of the wind farm to ensure the timeliness and accuracy of the analysis results.

[0057] Preferably, in order to improve the efficiency and accuracy of data collection, high-precision and high-sensitivity temperature sensors can be used in combination with advanced data acquisition systems to achieve real-time and continuous monitoring of the temperature of large components in wind farms. In addition, the collected data can be pre-processed by introducing intelligent algorithms, such as filtering and denoising, to improve the accuracy of subsequent analysis. In terms of parameter optimization, machine learning methods can be used to automatically adjust analysis parameters based on historical data and expert experience to adapt to the operating conditions and environmental changes of different wind farms.

[0058] In some embodiments, in the parameters of the temperature correlation analysis, a time series analysis method is used to process the temperature data so that the time resolution of the temperature data matches the operation cycle of the wind farm.

[0059] It should be noted that the purpose of using time series analysis to process temperature data is to match the time resolution of temperature data with the wind farm operation cycle, so as to more accurately reflect the temperature changes of large components in the wind farm. Time series analysis is a statistical method that identifies characteristics such as periodicity, trend and seasonality in data by analyzing the pattern of changes in data points over time.

[0060] Specifically, time series analysis can be implemented through a variety of algorithms, such as the autoregressive moving average model ARMA, the autoregressive integrated moving average model ARIMA, etc. In the implementation process, it is necessary to determine the appropriate time series model based on the operating data of the wind farm and the historical temperature data. Parameter settings include the order of the model, the size of the sliding window, etc. These parameters need to be adjusted according to the specific operating conditions of the wind farm and the periodicity of temperature changes. In addition, time series analysis can also include preprocessing steps such as differencing and seasonal adjustment of the data to improve the accuracy of the model.

[0061] Preferably, in order to improve the effect of time series analysis, advanced machine learning algorithms such as long short-term memory network LSTM or convolutional neural network CNN can be used, which can better capture the nonlinearity and complex patterns in temperature data. At the same time, external factors such as ambient temperature and wind speed can be introduced as input variables of the model to improve the accuracy of the analysis. In terms of parameter optimization, methods such as cross-validation can be used to determine the best model parameters to ensure that the model can accurately reflect the temperature change trend of large components in the wind farm. In addition, the model can be updated regularly to adapt to changes in the operating conditions of the wind farm.

[0062] In some embodiments, step S2 includes:

[0063] Step S201, calculating the statistical characteristics of temperature according to the temperature data of the major components of the wind farm;

[0064] Step S202: Identify components with abnormal temperatures by setting temperature thresholds;

[0065] Step S203: using a correlation analysis method to calculate the correlation between the temperatures of the large components based on the Pearson correlation coefficient;

[0066] Step S204: Save the calculated correlation index and the temperature data together in the original data set.

[0067] Specifically, step S201 involves calculating the statistical characteristics of temperature, which generally include the mean, variance and range of temperature. These statistical indicators can help us understand the distribution and variation range of temperature data. When setting parameters, the calculation cycle and threshold of statistical characteristics can be determined based on historical data and experience of the wind farm. For example, a time window can be set, such as one hour or one day, to calculate the mean and variance of temperature within the time period.

[0068] Step S202 identifies abnormal components by setting a temperature threshold. This threshold can be determined based on the historical operating data of the component and the specifications provided by the manufacturer. Once the temperature of the component exceeds this threshold, the system will mark it as abnormal.

[0069] Step S203 uses a correlation analysis method, such as the Pearson correlation coefficient, to calculate the correlation between the temperatures of the large components. The Pearson correlation coefficient is a statistical indicator that measures the degree of linear correlation between two variables, and its value is between -1 and 1. The closer the value is to 1 or -1, the stronger the correlation is.

[0070] Preferably, in order to improve the accuracy of temperature anomaly identification, more complex statistical methods or machine learning algorithms can be used to analyze temperature data. For example, cluster analysis can be used to identify components with abnormal temperature patterns, or support vector machines (SVMs) can be used to classify normal and abnormal temperature data. In addition, adaptive threshold settings can be introduced to dynamically adjust thresholds based on real-time data to adapt to factors such as environmental changes and component aging. In terms of correlation analysis, in addition to the Pearson correlation coefficient, the Spearman rank correlation coefficient or the Kendall rank correlation coefficient can also be considered, which may be more effective for identifying nonlinear relationships.

[0071] In some embodiments, the temperature T of the i-th component at time t of temperature anomaly identification is i (t) is calculated as:

[0072]

[0073] Among them, T i (t) is the temperature of the i-th component at time t; is the mean temperature of component i; Θ i Anomaly threshold set for component i; i Indicates whether the i-th component is abnormal at time t.

[0074] It should be noted that the calculation formula for temperature anomaly identification is used to determine whether the temperature of a large component of a wind farm exceeds the normal range at a specific moment. This formula works by comparing the actual temperature of the component with the set abnormal threshold to determine whether the component is abnormal.

[0075] Specifically, T in the formula i (t) represents the temperature of the i-th component at time t, is the mean temperature of component i, and Θ is the abnormal threshold set for component i. The setting of the abnormal threshold is usually based on the operating history data of the component and the manufacturer's recommended value. For example, the abnormal threshold can be set to the mean plus twice the standard deviation, so that most of the temperature data during normal operation will fall within this range, and those outside this range are considered abnormal.

[0076] Preferably, in order to improve the accuracy of temperature anomaly identification, a more complex statistical method can be used to determine the anomaly threshold. For example, a percentile method based on historical data can be introduced to set the anomaly threshold to the 95% or 99% percentile of the temperature distribution. In addition, environmental factors and the operating status of components can be considered to dynamically adjust the anomaly threshold. When calculating temperature anomalies, in addition to absolute thresholds, relative thresholds can also be introduced, such as comparing the temperature at the same time of the previous day, to identify abnormal changes in temperature.

[0077] In some embodiments, the process of determining temperature correlations between large components is:

[0078] 1) For n large wind farm components, calculate the temperature correlation of all components;

[0079]

[0080] Among them, t ij is the correlation coefficient between component i and component j; T i (t) and T j (t) is the temperature of component i and component j at time t; and is the mean temperature of components i and j;

[0081] 2) According to the correlation coefficient r ij , determine the temperature correlation between large components.

[0082] Specifically, the correlation coefficient r ij The calculation formula involves the temperature T of component i and component j at time t i (t) and T j (t), and their respective means and This formula measures the degree of linear correlation between the temperature changes of two components by calculating the sum of the products of the temperature differences and then dividing it by the square root of the product of the sum of the squares of the respective temperature differences. In terms of parameter setting, the number of data samples to calculate the correlation coefficient needs to be determined, which usually depends on the availability and time span of the data.

[0083] Preferably, in order to improve the accuracy and robustness of the temperature correlation analysis, a variety of statistical methods can be used to calculate the correlation coefficient. For example, in addition to the Pearson correlation coefficient, the Spearman rank correlation coefficient or the Kendall rank correlation coefficient can also be considered, which are more effective for identifying nonlinear relationships.

[0084] Furthermore, time-lagged correlation analysis can be introduced to explore the correlation between component temperatures at different time points, which helps to reveal the possible dynamic relationship between components. In actual operation, the stability and reliability of correlation analysis can also be improved by increasing the number of samples or using a rolling window method.

[0085] In some embodiments, in step S3, by constructing an association analysis model, a genetic algorithm is used to select the optimal parameter combination in the model, and training is performed to establish a model in which input features include large component temperature, statistical characteristics and correlation indicators, and output features are temperature anomalies; based on the temperature correlation between large components, temperature anomalies of large components in the wind farm are identified.

[0086] It should be noted that the process of building a correlation analysis model is the core of the entire wind farm large component temperature correlation analysis method. It uses the data in the original data set for model training and parameter optimization, enabling the model to conduct in-depth correlation analysis of the temperature changes of large components in the wind farm to identify potential temperature anomalies.

[0087] Specifically, the construction of the association analysis model involves selecting appropriate algorithms and parameters. In this case, the input features of the model include the temperature, statistical characteristics, and correlation indicators of large components, while the output features are the predictions of temperature anomalies. During model training, appropriate parameters such as learning rate, number of iterations, and regularization coefficient need to be set to ensure that the model can learn effective patterns from the data. Parameter optimization can be achieved through methods such as cross-validation and grid search to find the best model parameters.

[0088] Preferably, in order to improve the performance and adaptability of the association analysis model, multiple machine learning algorithms can be used for comparison and selection. For example, in addition to the random forest model, support vector machines, neural networks, or gradient boosting machines can also be considered. Each algorithm has its own unique advantages and applicable scenarios. Choosing a suitable algorithm can significantly improve the prediction accuracy of the model.

[0089] Furthermore, feature selection and dimensionality reduction techniques, such as principal component analysis or autoencoders, can be introduced to reduce noise and improve the generalization ability of the model. After the model training is completed, the model parameters can be continuously monitored and updated to adapt to changes in wind farm operating conditions to ensure the long-term effectiveness of the model.

[0090] In some embodiments, the association analysis model is a random forest model.

[0091] It should be noted that the association analysis model is a random forest model, which is an integrated learning method that improves the accuracy and robustness of prediction by building multiple decision trees and voting or averaging their results. Random forest performs well in processing large data sets and data with high-dimensional features, and is suitable for complex nonlinear relationship modeling.

[0092] Specifically, the construction of a random forest model involves the generation and parameter setting of multiple decision trees. Each decision tree randomly extracts samples from the original data set during training, and each split node only considers a randomly selected subset of features. Model parameters include the number of trees, the maximum number of features considered for each split, the maximum depth of the tree, etc.

[0093] Furthermore, the settings of these parameters need to be adjusted according to the characteristics of the data and the performance requirements of the model. For example, the greater the number of trees, the greater the complexity of the model and the training time, but it may improve the accuracy and generalization ability of the model.

[0094] Preferably, in order to optimize the performance of the random forest model, multiple strategies can be used. First, the complexity of the model and the training efficiency can be balanced by adjusting the number of decision trees. Second, feature selection mechanisms, such as feature selection based on information gain or Gini impurity, can be introduced to improve the predictive ability of the model. In addition, overfitting can be controlled by adjusting the maximum depth of the tree or the minimum number of leaf node samples.

[0095] In actual operation, parameter optimization techniques such as grid search and random search can be used to automatically find the optimal parameter combination. In order to further improve the robustness of the model, model fusion techniques such as bagging or boosting can be introduced to reduce the model's sensitivity to outliers and noise.

[0096] In some embodiments, the specific process of constructing the random forest model in step S3 includes:

[0097] Step S301: collect the original data and calculate the temperature T of the i-th component at time t. i (t), statistical characteristics and correlation indicators as a sample feature, and the temperature anomaly label Y i (t) is used as the sample label; the samples in the original data set are divided into a training sample set and a test sample set;

[0098] Step S302: using the training sample set data as the input of the random forest model, and using a genetic algorithm to select the optimal parameter combination of the random forest model;

[0099] Step S303: Use the optimal parameter combination to input the test sample set data into the random forest model, and calculate the error between the model output data and the sample label until the evaluation indicators of the model on the training sample set and the test sample set meet the set requirements;

[0100] Step S304: Save the trained model and parameter combination.

[0101] Specifically, step S301 involves organizing the data in the original data set into a format suitable for random forest model training. This includes taking the temperature, statistical characteristics and correlation indicators of each component at a specific moment as features, and taking the temperature anomaly label as the target variable. It is a standard practice in machine learning to divide data into training sample sets and test sample sets to evaluate the performance of the model during the training process. Parameter settings include determining the size and proportion of the sample set, and whether to perform data enhancement. For example, 70% of the data can be used as a training set and 30% as a test set.

[0102] Further, in step S302, the training sample set data is used to train the random forest model, and a genetic algorithm is used to select the optimal model parameters. A genetic algorithm is an optimization algorithm that simulates the natural selection process, and it searches for the optimal solution through iterative selection, crossover and mutation operations. In this step, it is necessary to define the parameters of the genetic algorithm, such as population size, number of iterations, crossover rate and mutation rate.

[0103] More specifically, step S303 involves using the optimal parameter combination to predict the test sample set data and calculating the error between the model output and the sample label. This step is the key to evaluating the performance of the model. Common evaluation indicators include accuracy, recall, and F1 score. The evaluation indicators of the model on the training set and the test set need to meet the preset requirements to ensure the generalization ability of the model.

[0104] Furthermore, in order to improve the prediction accuracy and robustness of the model, the following refinements or alternatives can be taken. First, in the data preprocessing stage, more data cleaning and feature engineering steps can be introduced, such as missing value processing, outlier detection, and feature normalization. Second, in the model training stage, multiple ensemble learning methods can be used for comparison, such as gradient boosting trees or extreme random trees, to find the most suitable model for the current data set.

[0105] In addition, model regularization techniques, such as L1 or L2 regularization, can be introduced to reduce the risk of overfitting of the model. In the model evaluation stage, cross-validation can be used to more comprehensively evaluate the performance of the model, and more evaluation indicators can be used to comprehensively evaluate the predictive ability of the model.

[0106] In some embodiments, the process of selecting the optimal parameter combination of the random forest model using a genetic algorithm includes:

[0107] 1) Define the parameter space; determine the random forest model parameters to be optimized and the corresponding value range;

[0108] 2) Define the evaluation index and fitness function: Select the accuracy A as the evaluation index to measure the performance of the random forest model under given parameters; the fitness function is the inverse of the error, that is,

[0109]

[0110] Among them, Parameters represents the model parameters; Error Rate represents the error rate of the model;

[0111] 3) Iterative optimization: The optimization process is repeated according to the specified number of iterations. In each iteration, the algorithm will adjust the parameters according to the previous evaluation results until the specified stopping condition is reached.

[0112] 4) Obtaining the best parameters: After the optimization is completed, the best parameter combination is selected from the obtained results, which corresponds to the random forest model with the best performance index.

[0113] It should be noted that the process of selecting the optimal parameter combination of the random forest model using a genetic algorithm is an optimization method that finds the optimal model parameters by simulating natural selection and genetic mechanisms.

[0114] Specifically, the implementation of the genetic algorithm includes defining parameter space, evaluation indicators and fitness function, iterative optimization and obtaining the best parameters. Parameter space refers to the parameters and their value ranges that need to be optimized in the random forest model, such as the number of trees, the maximum depth of the tree, the minimum number of leaf node samples, etc. The evaluation indicator usually selects the accuracy of the model, and the fitness function can be defined as the inverse of the error, that is, the higher the value of the fitness function, the better the performance of the model. Iterative optimization is to continuously adjust the parameters according to the evaluation results of the fitness function through multiple iterations until the stopping condition is reached. The best parameter combination refers to the optimal parameter combination selected according to the fitness function after the genetic algorithm runs.

[0115] Preferably, in order to improve the efficiency and effect of the genetic algorithm, the following refinement measures or alternatives can be taken. First, when defining the parameter space, prior knowledge of the parameters can be introduced to narrow the search range and improve the search efficiency. Second, in the selection of evaluation indicators and fitness functions, in addition to accuracy, other indicators such as the area under the AUC-ROC curve, F1 score, etc. can also be considered to more comprehensively evaluate the model performance.

[0116] Furthermore, in the iterative optimization process, a variety of crossover and mutation strategies can be used, such as single-point crossover, multi-point crossover, uniform mutation, etc., to increase the algorithm's exploration ability and avoid premature convergence. When obtaining the best parameters, an elite strategy can be used, that is, retaining a portion of the best individuals and directly inheriting them to the next generation to ensure the convergence speed and stability of the algorithm.

[0117] In some embodiments, the objective function of the random forest model training process includes a loss function and a regularization term, and the specific formula is:

[0118]

[0119] Among them, L(Θ) is the loss function, Θ is the model parameter, N is the number of samples, l is the loss function, y i is the true label of the sample, is the predicted label, λ is the regularization coefficient, and R(Θ) is the regularization term.

[0120] It should be noted that the objective function of the random forest model training process includes a loss function and a regularization term, which is to improve the prediction accuracy of the model while preventing the model from being too complex and causing overfitting. The loss function measures the difference between the model's predicted value and the actual value, while the regularization term is used to penalize the complexity of the model.

[0121] Specifically, the loss function in the objective function can be mean square error, cross entropy, etc. These functions define the cost of model prediction errors. The model parameter θ is the parameter set of all decision trees in the random forest. The number of samples N represents the total number of data points in the training set, and the loss of each sample is is the model's predicted label for the i-th sample and the true label y i The regularization coefficient λ is a hyperparameter used to control the influence of the regularization term R(θ) on the objective function. The regularization term can be the L1 norm or L2 norm of the parameter, which is used to control the complexity of the model.

[0122] Preferably, in order to more effectively train the random forest model and optimize the objective function, the following refinements or alternatives can be taken. First, in the selection of the loss function, the most suitable loss function can be selected according to the characteristics of the specific problem. For example, for classification problems, the cross entropy loss can be used; for regression problems, the mean square error or mean absolute error loss can be used. Second, the setting of the regularization coefficient λ can be determined by cross-validation to find the optimal regularization strength.

[0123] Furthermore, the choice of regularization terms can also be adjusted according to the complexity of the model and the characteristics of the data. For example, for models with a large number of parameters, L2 regularization can be used to prevent overfitting. In actual operation, early stopping technology can also be introduced to stop training when the performance on the validation set no longer improves, in order to avoid unnecessary calculations and overfitting.

[0124] The above-mentioned various embodiments of the present invention have the following beneficial effects: First, by collecting the temperature data of large components of the wind farm and calculating the correlation index, accurate data support can be provided for the temperature anomaly analysis of the wind farm. Secondly, by using the time series analysis method to process the temperature data, the time resolution of the data can be matched with the operating cycle of the wind farm, so as to more accurately identify the components with temperature anomalies. In addition, by setting the temperature threshold and adopting the correlation analysis method, the identification process of the temperature anomaly can be further refined and the accuracy of the identification can be improved. Furthermore, by constructing an association analysis model, especially a random forest model, and using a genetic algorithm for parameter optimization, the efficiency and accuracy of temperature anomaly identification can be improved. This method can not only identify potential temperature anomalies, but also provide more comprehensive information for fault prediction and health management of wind farms by analyzing the temperature correlation between large components.

[0125] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Under the concept of the present invention, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other changes in different aspects of the present invention as above, which are not provided in detail for the sake of simplicity. Although the present invention is described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing the temperature correlation of large components in a wind farm, characterized in that: include: Step S1, collecting temperature data of large components of the wind farm, setting parameters for temperature correlation analysis, calculating correlation indicators between the temperatures of the large components and saving them to the original data set; Step S2: identifying components with abnormal temperatures based on the temperature data of large components of the wind farm; using the statistical characteristics of the temperature data, including mean, variance and range, to extract features from the temperature data and save them to the original data set; Step S3: construct an association analysis model and use the data in the original data set for model training and parameter optimization; This enables the model to correlate temperature variations in large wind farm components to identify potential temperature anomalies.

2. The method for temperature correlation analysis of large components in a wind farm according to claim 1, characterized in that: Among the parameters of temperature correlation analysis, the time series analysis method is used to process the temperature data so that the time resolution of the temperature data matches the operation cycle of the wind farm.

3. The method for analyzing temperature correlation of large components in a wind farm according to claim 1, characterized in that: The step S2 includes: Step S201, calculating the statistical characteristics of temperature according to the temperature data of the major components of the wind farm; Step S202: Identify components with abnormal temperatures by setting temperature thresholds; Step S203: using a correlation analysis method to calculate the correlation between the temperatures of the large components based on the Pearson correlation coefficient; Step S204: Save the calculated correlation index and the temperature data together in the original data set.

4. The method for temperature correlation analysis of large components in a wind farm according to claim 3, characterized in that: The temperature T of the i-th component at time t during temperature anomaly identification i (t) is calculated as: Among them, T i (t) is the temperature of the i-th component at time t; is the mean temperature of component i; Θ i Anomaly threshold set for component i; i Indicates whether the i-th component is abnormal at time t.

5. The method for temperature correlation analysis of large components in a wind farm according to claim 3, characterized in that: The process for determining temperature correlations between large components is: 1) For n large wind farm components, calculate the temperature correlation of all components; Among them, r ij is the correlation coefficient between component i and component j; T i (t) and T i (t) is the temperature of component i and component j at time t; and is the mean temperature of components i and j; 2) According to the correlation coefficient r ij , determine the temperature correlation between large components.

6. The method for temperature correlation analysis of large components in a wind farm according to claim 1, characterized in that: In step S3, by constructing an association analysis model, a genetic algorithm is used to select the optimal parameter combination in the model, and training is performed to establish a model in which input features include the temperature of large components, statistical characteristics and correlation indicators, and the output feature is temperature anomaly; based on the temperature correlation between large components, the temperature anomalies of large components in the wind farm are identified.

7. The method for analyzing temperature correlation of large components in a wind farm according to claim 6, characterized in that: The association analysis model is a random forest model.

8. The method for analyzing temperature correlation of large components in a wind farm according to claim 7, characterized in that: The specific process of building a random forest model in step S3 includes: Step S301: collect the original data and calculate the temperature T of the i-th component at time t. i (t), statistical characteristics and correlation indicators as a sample feature, and the temperature anomaly label Y i (t) is used as the sample label; the samples in the original data set are divided into a training sample set and a test sample set; Step S302: using the training sample set data as the input of the random forest model, and using a genetic algorithm to select the optimal parameter combination of the random forest model; Step S303: Use the optimal parameter combination to input the test sample set data into the random forest model, and calculate the error between the model output data and the sample label until the evaluation indicators of the model on the training sample set and the test sample set meet the set requirements; Step S304: Save the trained model and parameter combination.

9. The method for temperature correlation analysis of large components in a wind farm according to claim 1, characterized in that: The process of selecting the optimal parameter combination of the random forest model using genetic algorithm includes: 1) Define the parameter space; determine the random forest model parameters to be optimized and the corresponding value range; 2) Define the evaluation index and fitness function: Select the accuracy A as the evaluation index to measure the performance of the random forest model under given parameters; the fitness function is the inverse of the error, that is, Among them, Parameters represents the model parameters; Error Rate represents the error rate of the model; 3) Iterative optimization: The optimization process is repeated according to the specified number of iterations. In each iteration, the algorithm will adjust the parameters according to the previous evaluation results until the specified stopping condition is reached. 4) Obtaining the best parameters: After the optimization is completed, the best parameter combination is selected from the obtained results, which corresponds to the random forest model with the best performance index.

10. The method for analyzing temperature correlation of large components in a wind farm according to claim 8, characterized in that: The objective function of the random forest model training process includes a loss function and a regularization term, and the specific formula is: Among them, L(Θ) is the loss function, Θ is the model parameter, N is the number of samples, l is the loss function, yi is the true label of the sample, is the predicted label, λ is the regularization coefficient, and R(Θ) is the regularization term.

Images