Wind power plant equipment health degree evaluation method and system based on multi-source data fusion and adaptive dynamic modeling

Abstract

The invention provides a wind power plant equipment health degree assessment method and system based on multi-source data fusion and adaptive dynamic modeling, and the method comprises the steps: collecting multi-source data, carrying out the preprocessing of the data, taking a timestamp as an index, carrying out the fusion of data from different sources through a data fusion technology, intercepting sequence data in a fixed time period p, and carrying out the segmentation of the sequence data in the fixed time period p; carrying out dimensionality reduction on the sequence data by using P CA, and extracting main components; training a dynamic LS < TM > sequence model, calculating a sliding step length q according to a deviation index, setting p = p + q, and returning to training; and outputting a prediction result. An adaptive dynamic LS < TM > model is constructed through multi-source data and a sliding time window with variable width, real-time judgment of fan faults is realized in combination with a dynamic threshold algorithm, data-driven equipment health degree evaluation and degradation trend prediction are supported, an attention mechanism is added, important moments in a time sequence are captured, the convergence speed is increased, and the prediction accuracy is improved. The accuracy of the model is improved, and generating capacity optimization and ecological compatibility decision making are supported.

Description

Technical Field

[0001] This application relates to the field of health assessment of wind farm equipment, and particularly to a method and system for assessing the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling. Background Art

[0002] In China, photovoltaic power station equipment has a large quantity and occupies a large area. Moreover, photovoltaic power stations are generally located in areas such as deserts, gobi, wastelands, and mountain slopes, where the natural environment is poor, resulting in a poor network environment for the equipment. Therefore, intelligent devices such as drones are required for inspections. When a drone conducts an inspection, it is necessary to retrieve the data of the photovoltaic power station to determine whether there are any abnormalities. A large amount of data is generated during the operation of a photovoltaic power station, and there are significant differences in the data formats between different devices. However, the computing power of a drone is limited and it is unable to process so much data in different formats in a timely manner. Therefore, it is necessary to first normalize and analyze the data formats through an edge computing device in the photovoltaic power station and transmit the analysis results to the intelligent device.

[0003] With the large-scale development of wind farms, the assessment of equipment health has become the key to ensuring the efficient operation of wind farms. Traditional methods often rely on a single data source or a static model and are difficult to cope with the complex and changeable wind farm environment.

[0004] Therefore, designing a method for assessing the health of wind farm equipment that combines multi-source data fusion and adaptive dynamic modeling methods, improving the accuracy and real-time performance of the assessment, reducing the risk of errors caused by human subjectivity, and saving manpower and material resources is an important research content for those skilled in the art. Summary of the Invention

[0005] In view of this, it is necessary to provide a method and system for assessing the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling. By constructing an adaptive dynamic LSTM model through multi-source data and a sliding time window with variable width, the time series data related to abnormalities is analyzed, improving the performance of the model and the accuracy of prediction.

[0006] In a first aspect, an embodiment of the present application provides a method for assessing the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling, the method comprising:

[0007] S1: Collect the operation and monitoring data with timestamps from different devices, the devices including photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video monitoring equipment, and drones;

[0008] S2: Preprocess the data and perform time series data processing;

[0009] S3: Indexed by timestamps, use data fusion technology to fuse data from different sources to generate sequence data of the comprehensive health index of wind farm equipment;

[0010] S4: Intercept the sequence data within a fixed time period p, use PCA to reduce the dimension of the sequence data, extract the main components, reduce the feature dimension, and divide it into a training set, a validation set, and a test set according to a preset ratio;

[0011] S5: Construct the extracted main components into a dynamic LSTM sequence model according to the time step and train it; verify the time series data of the validation set through the dynamic LSTM sequence model, evaluate the model performance, and adjust the hyperparameters; finally evaluate the generalization ability of the model on the test set;

[0012] S6: Evaluate the dynamic LSTM sequence model through the mean square error evaluation index. When the threshold index is not met, calculate the sliding step q according to the deviation index, set p = p + q, and return to S4;

[0013] S7: Otherwise, load the dynamic LSTM sequence model, perform anomaly detection on the time series data to be detected through the dynamic LSTM sequence model, analyze whether there is an abnormal situation, and output the prediction result.

[0014] Optionally, in an implementation manner of the first aspect of the present invention, the S2: preprocess the data and perform time series data processing, and its steps include:

[0015] S2.1: Data collection and integration, clarify the data sources, uniformly store the data from different sources, ensure the same format, process missing values, duplicate values, and outliers, and ensure data integrity;

[0016] S2.2: Data format standardization, unify the formats of different data sources, convert the data into a type suitable for analysis, and encode categorical data;

[0017] S2.3: Data alignment and matching, align the timestamps of different time series data, unify the coordinate systems of geospatial data, and match the entities in different data sources;

[0018] S2.4: Feature engineering, extract useful features from the original data, select the features most helpful for model prediction, reduce redundancy, and perform standardization or normalization processing on numerical features;

[0019] S2.5: Data quality assessment, perform data consistency checks to ensure the data consistency of different data sources, perform integrity checks: check whether the data is complete, process missing values, verify the accuracy of the data, and correct incorrect data;

[0020] S2.6: Sort the different time series data according to the timestamps.

[0021] Optionally, in an implementation manner of the first aspect of the present invention, the S3: Using the timestamp as an index, adopt a data fusion technology to fuse the data from different sources to generate the comprehensive health index sequence data of the wind farm equipment, including:

[0022] Use a Convolutional Neural Network (CNN) to extract the features of the image data, and use a Recurrent Neural Network (RNN) to extract the features of the time series data;

[0023] Adopt feature splicing to fuse the data from different sources;

[0024] Use the fused data to train a deep learning model to generate the comprehensive health index sequence data of the wind farm equipment.

[0025] Optionally, in an implementation manner of the first aspect of the present invention, the S4: Intercept the sequence data within a fixed time period p, use PCA to reduce the dimension of the sequence data, extract the main components, reduce the feature dimension, and divide it into a training set, a validation set, and a test set according to a preset ratio, including:

[0026] Determine the start point and the end point of the time period p;

[0027] Use the sliding window method to intercept the sequence data between the start point and the end point;

[0028] Convert the intercepted time series data into a matrix form suitable for PCA with each row as a sample and each column as a feature;

[0029] Perform Z-score normalization processing on the matrix data;

[0030] Apply the PCA algorithm to reduce the dimension of the intercepted sequence data, and use the Elbow Plot method to select the optimal number of the first k principal components.

[0031] Optionally, in an implementation manner of the first aspect of the present invention, the architecture of the dynamic LSTM sequence model includes an input layer, an LSTM layer, an attention mechanism layer, a fully connected layer, and an output layer;

[0032] Among them, the input layer receives and preprocesses the input time series data;

[0033] The LSTM layer extracts the features of the sequence data and generates a hidden state sequence;

[0034] The attention mechanism layer assigns weights to the hidden states of each time step to generate a weighted representation, enhancing the model's attention to important time steps;

[0035] The fully connected layer selects an activation function according to the task type and maps the context vector to the target space;

[0036] The output layer generates the final prediction result.

[0037] Optionally, in an implementation manner of the first aspect of the present invention, the step S6: evaluating the LSTM model through the mean square error evaluation index, and when the threshold index is not met, calculating the sliding step size q according to the deviation index, includes:

[0038] Evaluating the LSTM model through the mean square error evaluation index, where the error evaluation formula is M SE :

[0039]

[0040] where, y i represents the true value, represents the predicted value, n represents the number of samples, ε represents the error coefficient, and when y i = 0, ε = 0.001, when y i ≠ 0, ε = 0;

[0041] Calculating the sliding step size q according to the deviation index, the formula is:

[0042]

[0043] where, δ represents the threshold, [] represents the rounding operation, and η is the adjustment coefficient.

[0044] Optionally, in an implementation manner of the first aspect of the present invention, the loss function is:

[0045]

[0046] where, N is the number of samples, y i,c is the one-hot encoding of the true label, is the predicted probability that the i-th sample predicted by the model belongs to the category C, C is the number of categories, if the sample belongs to the category C, then y i,c = 1, otherwise 0, α t,i is the attention weight of the time step t to the i-th input, and λ is the regularization coefficient.

[0047] In a second aspect, an embodiment of the present application provides a wind farm equipment health assessment system based on multi-source data fusion and adaptive dynamic modeling, which is applied to the wind farm equipment health assessment method based on multi-source data fusion and adaptive dynamic modeling as described in the first aspect. The system includes:

[0048] Data acquisition module: It acquires the operation and monitoring data with timestamps from different devices, and the devices include photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video surveillance equipment, and unmanned aerial vehicles;

[0049] Data preprocessing module: It preprocesses the data and performs time series data processing;

[0050] Data fusion module: Indexed by timestamps, it uses data fusion technology to fuse data from different sources and generates the comprehensive health index sequence data of wind farm equipment;

[0051] Data processing module: It intercepts the sequence data within a fixed time period p, uses PCA to reduce the dimension of the sequence data, extracts the main components, reduces the feature dimension, and divides them into a training set, a validation set, and a test set according to a preset ratio;

[0052] Model training module: It constructs a dynamic LSTM sequence model based on the extracted main components according to the time step and conducts training; it validates the time series data of the validation set through the dynamic LSTM sequence model, evaluates the model performance, and adjusts the hyperparameters; it finally evaluates the generalization ability of the model on the test set;

[0053] Model evaluation module: It evaluates the dynamic LSTM sequence model through the mean square error evaluation index. When the threshold index is not met, it calculates the sliding step q according to the deviation index, sets p = p + q, and returns to S4;

[0054] Anomaly detection module: Otherwise, it loads the dynamic LSTM sequence model, conducts anomaly detection on the time series data to be detected through the dynamic LSTM sequence model, analyzes whether there is an abnormal situation, and outputs the prediction result.

[0055] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0056] A processor;

[0057] A memory for storing executable instructions of the processor;

[0058] Wherein, when the processor is configured to execute the instructions, it implements the method for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling as described in the first aspect.

[0059] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, and the computer-readable storage medium stores a program, and the program instructs the device to execute the method for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling as described in the first aspect.

[0060] The technical solution provided by the present invention provides a method and system for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling, collects multi-source data, pre-processes the data, uses timestamps as indexes, fuses data from different sources using data fusion technology, intercepts sequence data within a fixed time period p, uses PCA to reduce the dimension of the sequence data, and extracts the main components; trains a dynamic LSTM sequence model, and when the threshold index is not met, calculates the sliding step q according to the deviation index, sets p=p+q, and returns to the training step; outputs the prediction result. An adaptive dynamic LSTM model is constructed through multi-source data and a sliding time window with variable width, and the time series data related to the anomaly is analyzed, thereby improving the performance of the model and the accuracy of the prediction.

[0061] Beneficial effects:

[0062] (1) Considering that PV stations involve multiple types of equipment, the data generated by different devices are processed by data fusion, which improves the monitoring capability of the comprehensive operating status of the entire PV station.

[0063] (2) An adaptive dynamic LSTM model is constructed with a sliding time window of variable width to analyze the time series data related to anomalies, thereby improving the performance of the model and the accuracy of prediction.

[0064] (3) An attention mechanism is added to assign weights to different parts of the input sequence, capture important moments in the time series, speed up convergence, and improve the accuracy of the model.

[0065] (4) Based on the auto-encoding algorithm, the operating characteristic parameters of the whole machine are extracted, and combined with the dynamic threshold algorithm, real-time identification of fan faults is achieved, supporting data-driven equipment health assessment and degradation trend prediction.

[0066] (5) Integrate grid load data, environmental parameters (such as atmospheric boundary layer model) and operation and maintenance logs to establish a dynamic correlation model between wind farm efficiency and environmental adaptability to support power generation optimization and ecological compatibility decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A schematic diagram of a module of a wind farm equipment health assessment method based on multi-source data fusion and adaptive dynamic modeling provided in one embodiment of the present application.

[0068] Figure 2 A schematic diagram of a wind farm equipment health assessment system module based on multi-source data fusion and adaptive dynamic modeling provided in one embodiment of the present application.

[0069] Figure 3 A schematic diagram of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0070] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application.

[0071] It should be noted that in the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs. The terms used in the specification of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application.

[0072] It should be noted that in the embodiments of the present application, terms such as "first" and "second" are only used for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order. Features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to mean as an example, illustration or explanation. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present relevant concepts in a specific manner.

[0073] All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts are within the scope of protection of the present application.

[0074] Embodiment 1

[0075] This application provides a method and system for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling, including: S1: Collect operation and monitoring data with timestamps from different devices, including photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video surveillance equipment, and drones; S2: Preprocess the data and perform time series data processing; S3: Using the timestamp as an index, employ data fusion technology to fuse data from different sources to generate sequence data of comprehensive health indicators for wind farm equipment; S4: Intercept the sequence data within a fixed time period p, use PCA to reduce the dimension of the sequence data, extract the main components, reduce the feature dimension, and divide it into a training set, a validation set, and a test set according to a preset ratio; S5: Construct a dynamic LSTM sequence model with the extracted main components according to the time step and train it; Verify the performance of the model by using the dynamic LSTM sequence model for the time series data of the validation set, adjust the hyperparameters; Finally evaluate the generalization ability of the model on the test set; S6: Evaluate the dynamic LSTM sequence model through the mean square error evaluation index. When the threshold index is not met, calculate the sliding step q according to the deviation index, set p = p + q, and return to S4; S7: Otherwise, load the dynamic LSTM sequence model, perform anomaly detection on the time series data to be detected through the dynamic LSTM sequence model, analyze whether there is an abnormal situation, and output the prediction result.

[0076] Construct an adaptive dynamic LSTM model through multi-source data and a sliding time window with variable width, combine with a dynamic threshold algorithm to realize real-time discrimination of fan faults, support data-driven equipment health evaluation and degradation trend prediction, and add an attention mechanism to capture important moments in the time series, accelerate the convergence speed, improve the accuracy of the model, and support power generation optimization and ecological compatibility decision-making.

[0077] Figure 1 Schematic diagram of the process of the method for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling provided by an embodiment of this application.

[0078] As Figure 1 shown, a method for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling includes:

[0079] S1: Collect operation and monitoring data with timestamps from different devices, including photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video surveillance equipment, and drones.

[0080] It can be understood that in this embodiment, data from different devices is obtained. The devices here may include photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video surveillance equipment, and drones, etc.

[0081] Multi-source heterogeneous data integration: Integrate multi-modal data sources such as SCADA real-time operation data, meteorological sensors (temperature / pressure / wind speed), video surveillance (AI / AR / PTZ), and equipment status detection systems to build a spatio-temporal unified database.

[0082] S2: Preprocess the data and perform time series data processing.

[0083] It can be understood that in this embodiment, the step of S2: preprocessing the data and performing time series data processing includes:

[0084] S2.1: Data collection and integration, clarify the data sources, store the data from different sources uniformly, ensure the same format, handle missing values, duplicate values, and outliers to ensure data integrity;

[0085] S2.2: Data format standardization, unify the formats of different data sources, convert the data into a type suitable for analysis, and encode categorical data;

[0086] S2.3: Data alignment and matching, align the timestamps of different time series data, unify the coordinate systems of geospatial data, and match the entities in different data sources;

[0087] S2.4: Feature engineering, extract useful features from the original data, select the features most helpful for model prediction, reduce redundancy, and perform standardization or normalization on numerical features;

[0088] S2.5: Data quality assessment, perform data consistency checks to ensure the data consistency of different data sources, perform integrity checks: check whether the data is complete, handle missing values, verify the accuracy of the data, and correct incorrect data;

[0089] S2.6: Sort different time series data by timestamp.

[0090] Specifically, preprocess the data. This mainly includes missing value processing, outlier detection and processing, data standardization and normalization, and time series data processing.

[0091] Missing value processing: In a photovoltaic power station, different devices may generate different data types, so care must be taken to handle data missing due to equipment failures or communication problems. Identify and process the missing data. Fill in the missing values, and the interpolation method used here is a relatively conventional method.

[0092] Outlier Detection and Handling: Outliers may be caused by equipment failures, sensor errors, or other abnormal situations. Methods for handling outliers include deleting outliers, replacing them with appropriate values, or filling them using interpolation methods. These are also relatively conventional methods.

[0093] Time Series Data Processing: When dealing with time series data, ensure the correctness of timestamps and perform time alignment. Since this analysis mainly involves whether there are abnormalities in the photovoltaic power station, the timestamp is an important parameter.

[0094] Data Standardization and Normalization: Data standardization and normalization are key steps in data preprocessing. The main purpose is to make data with different features have the same scale, so as to better adapt to model training and improve the performance of algorithms.

[0095] The goal of data normalization is to scale the data to a specific range, usually [0,1]. The method adopted here is the L1 norm, also called the ("sparse rule operator" (Lasso regularization)). The L1 norm is the sum of the absolute values of the elements in the vector, and is also known as the Manhattan norm because it measures the Manhattan distance between vector elements at two points. For an n-dimensional vector x = [x1, x2,... x n ,

[0096] S3: Using the timestamp as an index, adopt data fusion technology to fuse data from different sources to generate sequence data of the comprehensive health index of wind farm equipment.

[0097] It can be understood that in this embodiment, the S3: Using the timestamp as an index, adopt data fusion technology to fuse data from different sources to generate sequence data of the comprehensive health index of wind farm equipment, including:

[0098] Use the convolutional neural network CNN to extract image data features, and use the recurrent neural network RNN to extract time series data features;

[0099] Adopt feature splicing to fuse data from different sources;

[0100] Use the fused data to train a deep learning model to generate sequence data of the comprehensive health index of wind farm equipment.

[0101] Specifically, adopt data fusion technology (such as Kalman filtering, deep learning fusion model) to fuse data from different sources to generate a comprehensive health index. Kalman filtering is applicable to linear systems, while deep learning fusion models can handle non-linear relationships.

[0102] S4: intercepting the sequence data within a fixed time period p, using PCA to reduce the dimension of the sequence data, extracting the main components, reducing the feature dimensions, and dividing the sequence data into a training set, a validation set, and a test set according to a preset ratio.

[0103] Specifically, principal component analysis (PCA) is a commonly used dimensionality reduction technique that can extract main features from high-dimensional data and reduce redundant information. LSTM (Long Short-Term Memory Network) is a deep learning model suitable for time series modeling that can capture long-term dependencies in data. Combining PCA with LSTM can reduce model complexity while improving the accuracy and efficiency of time series prediction. This method has broad application prospects in finance, meteorology, medical care and other fields. Use PCA to reduce the dimensionality of standardized data, extract the main components, and reduce the feature dimensions. The reduced dimensionality data is constructed into a sequence suitable for LSTM input according to the time step.

[0104] It can be understood that, in this embodiment, the S4: intercepting the sequence data within a fixed time period p, using PCA to reduce the dimension of the sequence data, extracting the main components, reducing the feature dimensions, and dividing it into a training set, a validation set, and a test set according to a preset ratio, includes:

[0105] Determine the starting and ending points of time period p;

[0106] Using a sliding window method to intercept the sequence data between the starting point and the end point;

[0107] The intercepted time series data is converted into a matrix form suitable for PCA, with each row as a sample and each column as a feature;

[0108] Perform Z-score standardization on matrix data;

[0109] The PCA algorithm is applied to reduce the dimension of the intercepted sequence data, and the ElbowPlot method is used to select the optimal number of the first k principal components.

[0110] S5: construct the extracted main components into a dynamic LSTM sequence model according to the time step and train it; verify the time series data of the validation set through the dynamic LSTM sequence model, evaluate the model performance, and adjust the hyperparameters; finally evaluate the generalization ability of the model on the test set.

[0111] It can be understood that in this embodiment, the architecture of the dynamic LSTM sequence model includes an input layer, an LSTM layer, an attention mechanism layer, a fully connected layer, and an output layer;

[0112] Wherein, the input layer receives and preprocesses input time series data;

[0113] The LSTM layer extracts the features of the sequence data and generates a hidden state sequence;

[0114] The attention mechanism layer assigns weights to the hidden states at each time step, generates a weighted representation, and enhances the model's attention to important time steps;

[0115] The fully connected layer selects an activation function according to the task type and maps the context vector to the target space;

[0116] The output layer generates the final prediction result.

[0117] Specifically, a multi-layer LSTM network is constructed, with the input being the features extracted by PCA and the output being the prediction result. A dynamic adjustment mechanism, such as an adaptive learning rate, Dropout, etc., is introduced to improve the generalization ability of the model. The loss function (such as mean squared error, cross entropy, etc.) is selected according to the task requirements.

[0118] Specifically, the architecture of the dynamic LSTM sequence model includes an input layer, an LSTM layer, an attention mechanism layer, a fully connected layer, and an output layer. Among them, the shape of the input data is (batch_size, time_steps, feature_dim), where: batch_size: the batch size. time_steps: the time step. feature_dim: the feature dimension of each time step.

[0119] LSTM layer: The core component, used to capture long-term dependencies in the time series. Parameter settings: units: the number of neurons in the hidden layer, which determines the complexity of the model. return_sequences: whether to return the output of each time step (True for multi-step prediction, False for single-step prediction). dropout: to prevent overfitting, randomly discard some neurons.

[0120] Attention mechanism layer: The attention mechanism layer is a multi-head attention mechanism, which is an extension of the self-attention mechanism. By calculating multiple attention heads in parallel, it captures the information in different subspaces of the input data. Project Query, Key, and Value into different subspaces through multiple linear transformations. Calculate the attention separately in each subspace. Concatenate the outputs of multiple attention heads and obtain the final output through a linear transformation.

[0121] Fully connected layer: Maps the LSTM output to the target dimension, usually used for classification or regression tasks. Output layer: Select an activation function according to the task (such as softmax for classification, linear for regression).

[0122] Select time-series data related to preprocessing and anomalies, such as light intensity, temperature, current, etc. These data may show obvious changes when anomalies occur. Label the anomaly data, which can be binary labels (normal / anomalous) or multi-class labels (different types of anomalies).

[0123] from tensorflow.keras.models import Sequential

[0124] from tensorflow.keras.layers import LSTM,Dense

[0125] model = Sequential()

[0126] model.add(LSTM(units = 50, activation ='relu', input_shape = (n_steps, n_features)))

[0127] model.add(Dense(units = 1)) # Output layer

[0128] model.compile(optimizer = 'adam', loss ='mse') # Mean Squared Error (MSE) in regression tasks

[0129] Construct the training set time-series data into a format suitable for LSTM input, usually a three-dimensional array, including the number of samples, the number of time steps, and the number of features. Use the training set to train the LSTM model and monitor the performance of the model on the validation set.

[0130] Here, the evaluation of LSTM uses three methods: confusion matrix, accuracy, and recall.

[0131] The confusion matrix is a two-dimensional table used to show the performance of a classification model on different classes. For binary classification problems, the confusion matrix contains four terms: True Positive (TP): Positive samples are correctly classified as positive classes. True Negative (TN): Negative samples are correctly classified as negative classes. False Positive (FP): Negative samples are misclassified as positive classes. False Negative (FN): Positive samples are misclassified as negative classes.

[0132] Predicted as positive class Predicted as negative class Actually positive class TP FN Actually negative class FP TN

[0133] Recall is a metric used to measure the ability of a model to capture positive examples. For a binary classification problem, recall is defined as:

[0134]

[0135] In the anomaly detection of a photovoltaic power station, a high recall means that the model can better capture the actually occurring abnormal events and reduce the false negative rate.

[0136] Accuracy is a metric for evaluating the performance of a classification model. It refers to the proportion of correctly classified samples among all samples. For a binary classification problem, the formula for calculating accuracy is:

[0137]

[0138] In the problem of anomaly detection, accuracy is not the only evaluation metric. Because in an extremely imbalanced dataset (where the number of normal samples is much larger than that of abnormal samples), the model may tend to predict all samples as normal, resulting in a high accuracy but being unable to effectively capture anomalies. In such cases, other metrics such as recall are more meaningful.

[0139] Model adjustment: The LSTM model can capture the patterns of normal data by learning temporal patterns, so as to identify abnormal situations that do not conform to the normal pattern during testing. During the model training process, it is necessary to continuously adjust and optimize the model according to the specific scenario to improve the accuracy and robustness of anomaly detection.

[0140] S6: Evaluate the dynamic LSTM sequence model through the mean squared error evaluation metric. When the threshold metric is not met, calculate the sliding step size q according to the deviation index, set p = p + q, and return to S4.

[0141] It can be understood that in this embodiment, S6: Evaluate the LSTM model through the mean squared error evaluation metric. When the threshold metric is not met, calculate the sliding step size q according to the deviation index, including:

[0142] Evaluate the LSTM model through the mean squared error evaluation metric, where the error evaluation formula is M SE :

[0143]

[0144] where, y i represents the true value, represents the predicted value, n represents the number of samples, ε represents the error coefficient, and when y i = 0, ε = 0.001, when y i ≠0, ε = 0;

[0145] Calculate the sliding step size q according to the deviation index, and the formula is:

[0146]

[0147] where δ represents the threshold, [] represents the rounding operation, and η is the adjustment coefficient.

[0148] Specifically, the loss function is:

[0149]

[0150] where N is the number of samples, y i,c is the one-hot encoding of the true label, is the predicted probability that the i-th sample predicted by the model belongs to class C, C is the number of classes. If the sample belongs to class C, then y i,c = 1, otherwise 0, α t,i is the attention weight of time step t to the i-th input, and λ is the regularization coefficient.

[0151] It can be understood that in this embodiment, through a rolling time window with variable width, the data for building the model is continuously updated, the old data is removed, and new data is continuously added, avoiding the problem that the model does not adapt to the data statistical law.

[0152] S7: Otherwise, load the dynamic LSTM sequence model, perform anomaly detection on the time-series data to be detected through the dynamic LSTM sequence model, analyze whether there is an abnormal situation, and output the prediction result.

[0153] Extract the whole-machine operation characteristic parameters based on the auto-encoding algorithm, and combine with the dynamic threshold algorithm to realize real-time discrimination of fan faults, and support data-driven equipment health assessment and degradation trend prediction. Cross-system collaborative analysis: Integrate grid load data, environmental parameters (such as atmospheric boundary layer models) and operation and maintenance logs to establish a dynamic association model of wind farm efficiency and environmental adaptability, and support power generation optimization and ecological compatibility decision-making.

[0154] Specifically, collect data in real time to ensure that real-time data is obtained from various devices (photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video monitoring equipment, drones, etc.). These data may include information related to the operation status of the photovoltaic station, such as light intensity, temperature, current, etc.

[0155] Perform preprocessing steps similar to the training data on the real-time collected data, including missing value processing, outlier detection and processing, data standardization and normalization, and time-series data processing. Ensure that the real-time data can be in the same data format as the data during model training.

[0156] Use the trained LSTM model to predict real-time data. Ensure that the time series format of the input data is correct, and use the probability values or class labels output by the model to determine whether there are anomalies. If the model is designed for multi-class anomaly detection, different types of anomalies can also be identified.

[0157] Anomaly judgment and handling: Based on the probability values or class labels output by the model, set a threshold, and samples exceeding this threshold are determined to be anomalies. According to actual requirements, different handling methods can be adopted, such as recording anomaly information, issuing alarms, and performing automatic repairs.

[0158] Multi-class anomaly handling: If the model is designed for multi-class anomaly detection, different types of anomalies can be identified according to the class labels output by the model. Different handling strategies can be adopted for different types of anomalies.

[0159] Regularly evaluate the performance of the model and monitor its performance in actual applications. If the performance of the model degrades or drifts, consider retraining the model or adjusting the model parameters.

[0160] Compare the results output by the model with the set anomaly threshold. If it exceeds the threshold, an anomaly alarm is triggered, and the anomaly information is transmitted to the intelligent devices related to the inspection tour. The alarm can also be implemented in various ways, including but not limited to: sending emails or text messages to notify relevant personnel. Trigger the automated system to execute specific repair or isolation measures. Display warning messages on the monitoring interface so that operators can notice in time.

[0161] Considering that the photovoltaic power station involves various devices, the data generated by different devices is processed through data fusion, which improves the monitoring ability of the comprehensive operation status of the entire photovoltaic power station.

[0162] Adopt data standardization and normalization methods to ensure that the data generated by different devices is compared on the same scale. This helps to eliminate the dimensional differences between the data of different devices and improves the stability and convergence speed of the model.

[0163] By adopting deep learning models such as LSTM, complex patterns in time series data can be better captured, thus improving the accurate detection of abnormal situations. Evaluation metrics of the model such as confusion matrix, accuracy, and recall rate help to evaluate the performance of the model.

[0164] The establishment of the anomaly alarm system enables relevant personnel to be notified in time when anomalies are discovered and necessary measures are taken. This enhances the real-time response ability to abnormal situations in the photovoltaic power station.

[0165] Data from different devices in a photovoltaic power station is obtained through intelligent devices, including photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video surveillance equipment, and unmanned aerial vehicles, etc. The collected data is preprocessed, including missing value processing, outlier detection and processing, data standardization and normalization, and time series data processing. Among them, missing values are filled using interpolation methods, outliers can be deleted or replaced, and time series data ensures that the timestamps are correct and aligned.

[0166] The long short-term memory (LSTM) model is used to analyze the time series data related to anomalies. The architecture of the LSTM model includes an input layer, an LSTM layer, an attention layer, and an output layer, which are used to capture the patterns in the time series data, especially the obvious changes shown during the occurrence of anomalies.

[0167] The trained model is used to predict real-time data to determine whether there are anomalies in the photovoltaic station. When an anomaly is detected, the anomaly alarm system is triggered to respond to the anomaly situation by sending notifications or executing automated measures.

[0168] Embodiment 2

[0169] As Figure 2 shown, this application provides a schematic diagram of the modules of a wind farm equipment health assessment system based on multi-source data fusion and adaptive dynamic modeling. This application provides a wind farm equipment health assessment system based on multi-source data fusion and adaptive dynamic modeling, which is applied to the wind farm equipment health assessment method based on multi-source data fusion and adaptive dynamic modeling as described in Embodiment 1, and includes: a data acquisition module 11, a data preprocessing module 12, a data fusion module 13, a data processing module 14, a model training module 15, a model evaluation module 16, and an anomaly detection module 17.

[0170] Specifically, in this embodiment, the data acquisition module 11 is used to collect the operation and monitoring data with timestamps from different devices, and the devices include photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video surveillance equipment, and unmanned aerial vehicles.

[0171] Specifically, in this embodiment, the data preprocessing module 12 is used to preprocess the data and perform time series data processing.

[0172] Specifically, in this embodiment, the data fusion module 13 is used to index by timestamp and adopt data fusion technology to fuse data from different sources to generate the sequence data of the comprehensive health index of wind farm equipment.

[0173] Specifically, in this embodiment, the data processing module 14 is configured to intercept the sequence data within a fixed time period p, perform dimensionality reduction on the sequence data using PCA, extract the main components, reduce the feature dimensions, and divide them into a training set, a validation set, and a test set according to a preset ratio.

[0174] Specifically, in this embodiment, the model training module 15 is configured to construct a dynamic LSTM sequence model based on the extracted main components according to the time step and perform training; verify the time series data of the validation set through the dynamic LSTM sequence model, evaluate the model performance, and adjust the hyperparameters; finally evaluate the generalization ability of the model on the test set.

[0175] Specifically, in this embodiment, the model evaluation module 16 is configured to evaluate the dynamic LSTM sequence model through the mean square error evaluation index. When the threshold index is not met, calculate the sliding step q according to the deviation index, set p = p + q, and return to S4.

[0176] Specifically, in this embodiment, the anomaly detection module 17 is configured to otherwise, load the dynamic LSTM sequence model, perform anomaly detection on the time series data to be detected through the dynamic LSTM sequence model, analyze whether there is an abnormal situation, and output the prediction result.

[0177] Figure 3 This is an electronic device provided by an embodiment of the present application. As Figure 3 shown, the electronic device at least includes the following parts: a processor 101, a memory 100, a communication interface 103, and a bus 102.

[0178] In the embodiment of the present application, the memory 100 is used to store executable instructions of the processor 101, and the processor 101 is configured to implement a wind farm equipment health assessment system module based on multi-source data fusion and adaptive dynamic modeling as Figure 2 shown when executing the instructions.

[0179] In the embodiment of the present application, a computer-readable storage medium includes instructions for instructing the device to execute the method of the first aspect. For example, the instructions instruct the device to execute the wind farm equipment health assessment method based on multi-source data fusion and adaptive dynamic modeling shown in the Figure 1 process steps.

[0180] The program operating in the electronic device according to an embodiment of the present application may be a program for controlling a central processing unit (CPU) or the like to implement the functions of the above-described embodiments related to a solution of the present invention (a program for causing a computer to function). Then, the information processed by these devices is temporarily stored in a random access memory (RAM) during its processing, and then stored in various ROMs such as a read-only memory (FlashROM), a hard disk drive (HDD), etc., and read, corrected, and written by the CPU as needed.

[0181] It should be noted that a part of the electronic device of the above-described embodiment can also be implemented by a computer. In this case, the program for implementing the control function can be recorded on a computer-readable recording medium, and implemented by reading the program recorded on the recording medium into the computer and executing it.

[0182] It should be noted that the "computer" mentioned here refers to a computer built into an electronic device, which is a computer including hardware such as an OS and peripheral devices. In addition, the "computer-readable recording medium" refers to a removable medium such as a floppy disk, a magneto-optical disk, a ROM, a CD-ROM, or a storage device such as a hard disk built into a computer.

[0183] Moreover, the "computer-readable recording medium" may include: a medium that dynamically stores a program for a short time, such as a communication line in the case of transmitting a program via a network such as the Internet or a communication line such as a telephone line; a medium that stores a program for a fixed time, such as a volatile memory inside a computer of a server or a client in this case. In addition, the above program may be a program for implementing a part of the above functions, and may also be a program that can implement the above functions by combining with a program already recorded in a computer.

[0184] In addition, the electronic device in the above-described embodiment can also be implemented as an aggregate (device group) composed of multiple devices. Each device constituting the device group may have all or part of the functions or functional blocks of the electronic device of the above-described embodiment. As the device group, it is sufficient to have all the functions or functional blocks of the electronic device.

[0185] Those of ordinary skill in the art of this technology should recognize that the above embodiments are only used to illustrate the present application, rather than to limit the present application. As long as it is within the scope of the spirit of the present application, appropriate changes and variations made to the above embodiments fall within the scope of protection required by the present application.

Claims

1. A method for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling, characterized in that, The method includes: S1: Collect the operation and monitoring data with timestamps from different devices, where the devices include photovoltaic modules, electrical appliances, detection devices, meteorological sensors, video surveillance devices, and unmanned aerial vehicles; S2: Preprocess the data and perform time series data processing; S3: Using the timestamp as an index, adopt data fusion technology to fuse data from different sources to generate sequence data of the comprehensive health index of wind farm equipment; S4: Intercept the sequence data within a fixed time period p, use PCA to reduce the dimension of the sequence data, extract the main components, reduce the feature dimension, and divide it into a training set, a validation set, and a test set according to a preset ratio; S5: Construct the extracted main components into a dynamic LSTM sequence model according to the time step and train it; verify the time series data of the validation set through the dynamic LSTM sequence model, evaluate the model performance, and adjust the hyperparameters; finally evaluate the generalization ability of the model on the test set; S6: Evaluate the dynamic LSTM sequence model through the mean square error evaluation index. When the threshold index is not met, calculate the sliding step q according to the deviation index, set p = p + q, and return to S4; S7: Otherwise, load the dynamic LSTM sequence model, perform anomaly detection on the time series data to be detected through the dynamic LSTM sequence model, analyze whether there is an abnormal situation, and output the prediction result.

2. The method for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling according to claim 1, wherein The S2: Preprocess the data and perform time series data processing, and its steps include: S2.1: Data collection and integration, clarify the data sources, store the data from different sources uniformly, ensure the same format, process missing values, duplicate values, and outliers to ensure data integrity; S2.2: Data format standardization, unify the formats of different data sources, convert the data into a type suitable for analysis, and encode categorical data; S2.3: Data alignment and matching, align the timestamps of different time series data, unify the coordinate systems of geospatial data, and match the entities in different data sources; S2.4: Feature engineering, extract useful features from the original data, select the features that are most helpful for model prediction, reduce redundancy, and perform standardization or normalization processing on numerical features; S2.5: Data quality assessment, perform data consistency checks to ensure the data consistency of different data sources, perform integrity checks: check whether the data is complete, process missing values, verify the accuracy of the data, and correct the error data; S2.6: Sort different time series data according to timestamps.

3. The method for evaluating the health degree of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling according to claim 1, characterized in that, The S3: Using the timestamp as an index, adopt data fusion technology to fuse data from different sources to generate sequence data of the comprehensive health index of wind farm equipment, including: Use a convolutional neural network CNN to extract image data features, and use a recurrent neural network RNN to extract time series data features; Adopt feature splicing to fuse data from different sources; Use the fused data to train a deep learning model to generate sequence data of the comprehensive health index of wind farm equipment.

4. A method for evaluating the health degree of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling according to claim 1, characterized in that, S4: Intercept the sequence data within a fixed time period p, use PCA to reduce the dimension of the sequence data, extract the main components, reduce the feature dimension, and divide it into a training set, a validation set, and a test set according to a preset ratio, including: Determine the start point and end point of the time period p; Use the sliding window method to intercept the sequence data between the start point and the end point; Convert the intercepted time series data into a matrix form suitable for PCA with each row as a sample and each column as a feature; Perform Z-score normalization processing on the matrix data; Apply the PCA algorithm to reduce the dimension of the intercepted sequence data, and use the ElbowPlot method to select the optimal number of the first k principal components.

5. A method for evaluating the health status of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling according to claim 4, characterized in that The architecture of the dynamic LSTM sequence model includes an input layer, an LSTM layer, an attention mechanism layer, a fully connected layer, and an output layer; Among them, the input layer receives and preprocesses the input time series data; The LSTM layer extracts the features of the sequence data and generates a hidden state sequence; The attention mechanism layer assigns weights to the hidden states at each time step to generate a weighted representation, enhancing the model's attention to important time steps; The fully connected layer selects an activation function according to the task type and maps the context vector to the target space; The output layer generates the final prediction result.

6. The method for evaluating the health degree of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling according to claim 5, wherein S6: Evaluate the LSTM model through the mean squared error evaluation index. When the threshold index is not satisfied, calculate the sliding step size q according to the deviation index, including: Evaluate the LSTM model using the mean squared error evaluation metric, where the error evaluation formula is M SE : Among them, y i represents the true value, represents the predicted value, n represents the number of samples, ε represents the error coefficient, and when y i = 0, ε = 0.001, when y i ≠ 0, ε = 0; Calculate the sliding step size q according to the deviation index, and the formula is: Among them, δ represents the threshold, [] represents the rounding operation, and η is the adjustment coefficient.

7. A method for evaluating the health of wind farm equipment based on multi-source data fusion and adaptive dynamic modeling according to claim 5, characterized in that The loss function is: where N is the number of samples, and y i,c is the one - hot encoding of the true label, is the predicted probability that the i - th sample predicted by the model belongs to class C, where C is the number of classes. If the sample belongs to class C, then y i,c = 1, otherwise 0, and α t,i is the attention weight of time step t for the i - th input, and λ is the regularization coefficient.

8. A wind farm equipment health assessment system based on multi-source data fusion and adaptive dynamic modeling, which is applied to the wind farm equipment health assessment method based on multi-source data fusion and adaptive dynamic modeling according to any one of claims 1 to 7, and is characterized in that, The system includes: Data acquisition module: Collect the operation and monitoring data with timestamps from different devices, and the devices include photovoltaic modules, electrical equipment, detection equipment, meteorological sensors, video monitoring equipment, and unmanned aerial vehicles; Data preprocessing module: Preprocess the data and perform time series data processing; Data fusion module: Indexed by timestamps, use data fusion technology to fuse data from different sources to generate sequence data of the comprehensive health status index of wind farm equipment; Data processing module: Intercept the sequence data within a fixed time period p, use PCA to reduce the dimension of the sequence data, extract the main components, reduce the feature dimension, and divide it into a training set, a validation set, and a test set according to a preset ratio; Model training module: Construct the extracted main components into a dynamic LSTM sequence model according to the time step and train it; verify the time series data of the validation set through the dynamic LSTM sequence model, evaluate the model performance, and adjust the hyperparameters; finally evaluate the generalization ability of the model on the test set; Model evaluation module: Evaluate the dynamic LSTM sequence model through the mean squared error evaluation index. When the threshold index is not satisfied, calculate the sliding step size q according to the deviation index, set p = p + q, and return to S4; Anomaly detection module: Otherwise, load the dynamic LSTM sequence model, perform anomaly detection on the time series data to be detected through the dynamic LSTM sequence model, analyze whether there is an abnormal situation, and output the prediction result.

9. An electronic device, characterized in that, Comprising: A processor; A memory for storing processor-executable instructions; Wherein, when the processor is configured to execute the instructions, it implements the wind farm equipment health assessment method based on multi-source data fusion and adaptive dynamic modeling according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program, and the program instructs the device to execute the wind farm equipment health assessment method based on multi-source data fusion and adaptive dynamic modeling according to any one of claims 1 to 7.

Images