Wind power plant edge calculation data cleaning and real-time transmission optimization method and system

Abstract

The invention relates to the field of wind power plants, in particular to a wind power plant edge calculation data cleaning and real-time transmission optimization method and system. The method comprises the following steps: forming a multi-source heterogeneous data stream by collecting SCADA operation data, CMS vibration monitoring data and meteorological environment data of a wind turbine generator in real time; performing standardization processing on the data at an edge computing node, constructing a multi-dimensional feature vector, and generating a fusion data set by adopting a self-adaptive fusion algorithm based on a dynamic weight; and constructing a fault prediction model based on the deep dual-channel convolutional neural network, and outputting a health state evaluation value and a fault risk level in real time. And when the risk level exceeds a threshold value, generating an early warning signal. The system also performs data compression transmission optimization according to the real-time bandwidth state, and dynamically adjusts the weight of the monitoring parameter to realize the closed-loop optimization of the preventive maintenance strategy.

Description

Technical Field

[0001] The present invention relates to the field of wind farms, and in particular to a method and system for optimizing wind farm edge computing data cleaning and real-time transmission. Background Art

[0002] With the rapid development of the wind power industry, the amount of operational data generated by wind turbines has exploded. This data includes heterogeneous data from multiple sources, such as SCADA operational data, CMS vibration monitoring data, and meteorological and environmental data. Real-time collection and processing of this data is crucial for the efficient operation and maintenance of wind farms. However, existing technologies face numerous challenges in processing this data. First, the fusion of heterogeneous data from multiple sources is difficult. Differences in timestamps, sampling frequencies, and data formats between different data sources make direct data fusion and analysis difficult. Second, traditional data processing methods are typically performed in the cloud, which not only increases data transmission latency but also places high demands on network bandwidth. This makes it difficult to ensure real-time and complete data transmission, especially in unstable or bandwidth-constrained environments. Furthermore, existing technologies lack the ability to accurately assess the health status and fault risk level of wind turbines in real time, making preventive maintenance difficult. Finally, existing technologies lack effective utilization of edge computing node resources and are unable to dynamically adjust data processing strategies based on real-time bandwidth availability and computing resource load. This results in low data processing efficiency and fails to meet the requirements of real-time wind farm operation and maintenance. Therefore, there is an urgent need for a wind farm data processing and fault warning method and system that can address these challenges. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for wind farm edge computing data cleaning and real-time transmission optimization to solve the problems in the above-mentioned background technology.

[0004] To achieve the above objectives, the following technical solutions are adopted.

[0005] A method for cleaning and optimizing real-time transmission of wind farm edge computing data includes the following steps: collecting SCADA operation data, CMS vibration monitoring data, and meteorological environment data of wind turbines in real time to form a multi-source heterogeneous data stream; standardizing the multi-source heterogeneous data at the edge computing node to construct a multi-dimensional feature vector with a timestamp; performing cross-source feature correlation analysis on the multi-dimensional feature vector using a dynamic weight-based adaptive fusion algorithm to generate a spatiotemporally aligned fused data set; constructing a fault prediction model based on a deep dual-channel convolutional neural network, and supervising the model using historical fault samples to obtain a multi-level fault probability prediction function; inputting the fused data set into the trained fault prediction model to output health status assessment values and fault risk levels of the unit's transmission chain components in real time; generating a warning signal containing the fault type, location information, and repair suggestions when the fault risk level exceeds a dynamic warning threshold; and optimizing the transmission of the cleaned data using a data priority queue based on the real-time bandwidth status of the edge computing node; dynamically adjusting monitoring parameter weights according to the warning results, and iteratively updating the newly added fault samples to the fault prediction model to achieve closed-loop optimization of the preventive maintenance strategy.

[0006] Optionally, the adaptive fusion algorithm based on dynamic weights includes: establishing a cross-source data association matrix based on the Pearson correlation coefficient, calculating the dynamic weight coefficient of each data source and the target fault type; using a sliding time window to perform time domain alignment on the multidimensional feature vector, and generating a fused feature vector with time and space consistency through weighted superposition; performing outlier removal and noise filtering on the fused data stream to form a standardized input sequence; and dynamically adjusting the calculation granularity of the fusion algorithm according to the computing resource load of the edge node.

[0007] Optionally, the deep dual-channel convolutional neural network includes: the first channel is used to process SCADA time series data, and an LSTM network with an attention mechanism is used to extract equipment operating status features; the second channel is used to process CMS high-frequency vibration signals, and a multi-layer wavelet convolution kernel is used to extract frequency domain fault features; the dual-channel output is feature spliced through a fully connected layer, and a residual network is introduced to optimize feature transfer efficiency; the output layer combines the Softmax function and the Sigmoid function to generate a multi-dimensional fault probability distribution and risk level label.

[0008] Optionally, the generation of the dynamic warning threshold includes: constructing a dynamic threshold curve based on seasonal wind speed changes according to the health status assessment value, and triggering a three-level warning mechanism when the assessment value continuously exceeds the threshold curve and the duration reaches a preset window; the warning signal uses a fuzzy logic algorithm to integrate the failure probability, equipment criticality and current maintenance resource availability to generate a differentiated emergency response strategy; and adjusting the slope parameter of the threshold curve in real time to adapt to the calculation delay of the edge node.

[0009] Optionally, the closed-loop optimization of the preventive maintenance strategy includes: constructing a maintenance knowledge base containing an equipment life prediction model based on historical response data of early warning signals; dynamically optimizing the unit maintenance cycle and spare parts replacement strategy based on the maintenance records and remaining life prediction results in the knowledge base; and generating a distributed maintenance task scheduling queue in combination with the geographic location information of the edge nodes.

[0010] Optionally, the iterative update of the model includes: when the confidence of a new fault sample exceeds a preset threshold, using an online incremental learning algorithm to update the model parameters; for low-confidence samples, triggering a manual labeling module and conducting adversarial training with historical data, and using the redundant computing resources of edge nodes to fine-tune the model; and establishing a model version management mechanism to be compatible with the environmental differences of different wind farms.

[0011] Optionally, the real-time transmission optimization also includes: dividing the transmission channels according to the urgency of the data priority queue, using a lossless compression algorithm for high-priority warning data, and using a lossy compression algorithm for low-frequency monitoring data; dynamically allocating bandwidth resources between edge nodes and cloud servers, and adjusting the data segment size based on transmission delay feedback.

[0012] Optionally, it also includes: pushing warning results to mobile terminals in the form of visual charts, supporting user-defined warning rules; optimizing warning thresholds and model parameters through reinforcement learning algorithms based on user feedback data; establishing a data synchronization mechanism between edge nodes and the cloud to achieve a closed loop of human-machine collaborative decision-making.

[0013] Wind farm edge computing data cleaning and real-time transmission optimization system, including:

[0014] The data acquisition module is deployed on the edge computing node to obtain SCADA operation data, CMS vibration monitoring data, and meteorological environment data in real time;

[0015] The data cleaning module uses a dynamic weight allocation algorithm to remove outliers, filter noise, and align time domains on multi-source heterogeneous data to generate standardized data streams;

[0016] The fusion analysis module performs feature fusion and fault prediction on the cleaned data based on a deep dual-channel convolutional neural network, and outputs health status assessment values and risk levels;

[0017] The transmission optimization module divides data into priority queues according to bandwidth status and adopts differentiated compression strategies for real-time data transmission;

[0018] The maintenance decision module builds a preventive maintenance knowledge base based on early warning results and dynamically generates equipment maintenance task scheduling plans;

[0019] The visual interaction module is used to display warning results and receive user feedback, and to iteratively update model parameters through the edge-cloud collaborative mechanism.

[0020] Optionally, the fusion analysis module also includes: a cross-source data association unit, which is used to dynamically calculate the data source weight through the Pearson correlation coefficient; the transmission optimization module integrates a bandwidth perception unit, which dynamically adjusts the compression algorithm level according to the real-time network load; the maintenance decision module has a built-in life prediction algorithm, which combines geographic information to generate an optimized spare parts logistics path.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] This invention provides a method for wind farm edge computing data cleansing and real-time transmission optimization. By standardizing and fusing heterogeneous multi-source data at edge computing nodes, this method addresses the difficulties and low processing efficiency of data fusion in existing technologies. This method utilizes an adaptive fusion algorithm based on dynamic weights to dynamically adjust weights based on the correlation between the data source and the target fault type, generating a spatiotemporally aligned fused dataset, thereby improving the accuracy and reliability of data fusion. Furthermore, a fault prediction model constructed based on a deep dual-channel convolutional neural network (DCNN) outputs real-time health status assessments and fault risk levels for turbine drive train components, enabling accurate prediction and dynamic early warning of wind turbine faults. Furthermore, this method uses data priority queues to optimize the compression and transmission of cleaned data based on the real-time bandwidth status of the edge computing node, effectively improving the real-time and integrity of data transmission and reducing network bandwidth requirements. This invention provides improvements and supplements in the specific implementation of the adaptive fusion algorithm, the architectural design of the DCNN, the generation mechanism of dynamic early warning thresholds, the closed-loop optimization of preventive maintenance strategies, iterative model updates, real-time transmission optimization strategies, and interactive visualization. These improvements not only improve the accuracy and reliability of fault prediction, but also enhance the intelligence and practicality of the system, enabling it to better adapt to the complex and changeable operating environment of wind farms, and providing strong technical support for the efficient operation and maintenance of wind farms. DETAILED DESCRIPTION

[0023] The present invention will be described in detail below with reference to and in combination with the embodiments. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0024] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise indicated, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0025] This invention provides a method and system for optimizing wind farm edge computing data cleansing and real-time transmission. This method aims to achieve efficient processing and fault warning of multi-source heterogeneous data from wind farms through edge computing technology, while also optimizing the data transmission process and improving the system's real-time performance and reliability. The following is a detailed implementation of this method and system.

[0026] First, various operational data from wind turbines is collected in real time, including SCADA operational data, CMS vibration monitoring data, and meteorological and environmental data. These data come from a wide range of sources, covering the mechanical, electrical, and environmental parameters of the wind turbine and comprehensively reflecting the operational status of the wind turbine. SCADA operational data primarily includes key operating parameters such as wind turbine power, speed, and temperature. These data are collected in real time through the wind turbine's SCADA system, and the sampling frequency can be set based on actual needs, such as once per second. CMS vibration monitoring data is obtained through vibration sensors installed on key components of the wind turbine. These sensors can monitor the vibration of the components in real time, and the sampling frequency is typically high, such as hundreds of times per second, to capture subtle changes in vibration signals. Meteorological and environmental data, including wind speed, wind direction, temperature, and air pressure, are collected through meteorological monitoring stations installed at wind farms. These data are crucial for analyzing the operating environment of wind turbines and predicting potential failures.

[0027] The collected heterogeneous data from multiple sources differs in time and format, necessitating standardization. Specifically, data from each data source undergoes preprocessing, including data cleaning, format conversion, and timestamp alignment. Data cleaning primarily removes outliers and noise. For example, vibration data is filtered using a sliding average filter to remove abnormal vibration signals caused by sensor failure or external interference. Format conversion converts data from different sources into a unified format for subsequent processing. For example, wind speed units in meteorological data can be converted from meters per second to kilometers per hour to align with the units of other data sources. Timestamp alignment aligns data from different sources in chronological order to accurately analyze relationships between them. For example, a sliding time window algorithm is used to align SCADA data, CMS data, and meteorological data based on timestamps. This allows data at the same time point to be mapped, forming multidimensional feature vectors with timestamps. These multidimensional feature vectors contain various operating status information about the wind turbine at different time points, providing a foundation for further data fusion and fault prediction.

[0028] Next, an adaptive fusion algorithm based on dynamic weights is used to perform cross-source feature correlation analysis on the multidimensional feature vectors to generate a fused dataset. The core of this algorithm lies in dynamically calculating the weight coefficients of each data source based on the degree of correlation between the different data sources and the target fault type, and then fusing the multi-source data through weighted superposition. Specifically, a cross-source data correlation matrix based on the Pearson correlation coefficient is first established to calculate the correlation between each data source and the target fault type. For example, by calculating the Pearson correlation coefficient between power parameters in SCADA data and gearbox faults, the importance of power parameters in predicting gearbox faults is determined. Then, based on the calculated correlation coefficient, a sliding time window is used to perform time-domain alignment on the multidimensional feature vectors, and a weighted superposition is performed to generate a fused feature vector with spatiotemporal consistency. During the weighted superposition process, data sources with a high correlation with the target fault type are assigned a higher weight, while data sources with a lower correlation are assigned a lower weight, thereby achieving effective fusion of multi-source data. Furthermore, the fused data stream is subjected to outlier removal and noise filtering to further improve the data quality and reliability, forming a standardized input sequence that provides a high-quality data foundation for subsequent fault prediction model training. The calculation granularity of the fusion algorithm is dynamically adjusted according to the computing resource load of the edge node. For example, a high-precision fusion algorithm is used when computing resources are sufficient, and the calculation accuracy is appropriately reduced to improve the processing speed when resources are tight.

[0029] A fault prediction model is constructed based on a deep dual-channel convolutional neural network. The model is supervised and trained using historical fault samples to obtain a multi-level fault probability prediction function. The deep dual-channel convolutional neural network consists of two channels: the first channel processes SCADA time series data, employing an LSTM network with an attention mechanism to extract equipment operating status features; the second channel processes CMS high-frequency vibration signals, employing multi-layer wavelet convolution kernels to extract frequency-domain fault features. Specifically, in the first channel, the LSTM network effectively processes the long-term and short-term dependencies of time series data. By introducing the attention mechanism, the model can focus more on fault-related time series features, thereby improving fault prediction accuracy. For example, when processing wind turbine power fluctuation data, the attention mechanism can enable the model to focus on key features such as sudden power changes, which are often closely related to equipment failures. In the second channel, multi-layer wavelet convolution kernels perform multi-scale analysis of the CMS high-frequency vibration signals, extracting fault features in different frequency bands. For example, wavelet convolution kernels can be used to extract low-frequency vibration signal features, which may be related to the equipment's low-frequency failure modes. They can also extract high-frequency vibration signal features, which may be related to the equipment's high-frequency failure modes. The fully connected layer performs feature concatenation on the dual-channel outputs, integrating the features extracted from the two channels to output a multidimensional fault probability distribution. For example, the fault probability distribution output by the model may include multiple dimensions such as the gearbox failure probability and the generator failure probability. The fault probability value for each dimension reflects the likelihood of a wind turbine failure in the corresponding component. The output layer combines the Softmax function with the Sigmoid function to generate a multidimensional fault probability distribution and risk level labels. The Softmax function is used to normalize the output value into a probability distribution, and the Sigmoid function is used to generate a binary probability value for each fault type, thereby more accurately assessing the fault risk level.

[0030] The fused dataset is fed into the trained fault prediction model, which outputs the health status assessment value and fault risk level of the unit's drive train components in real time. When the fault risk level exceeds the dynamic warning threshold, an early warning signal is generated, including the fault type, location information, and repair recommendations. The generation of the dynamic warning threshold involves constructing a dynamic threshold curve based on seasonal wind speed variations based on the health status assessment value. When the assessment value continuously exceeds the threshold curve for a period of time exceeding a preset window, a three-level early warning mechanism is triggered. The early warning signal uses a fuzzy logic algorithm to integrate the fault probability, equipment criticality, and the current availability of maintenance resources to generate a differentiated emergency response strategy. For example, for high-risk faults, detailed fault location information and emergency repair recommendations are generated; for low-risk faults, general maintenance recommendations are generated. The slope parameter of the threshold curve is adjusted in real time to accommodate the computational delay of edge nodes, ensuring the real-time and accuracy of the early warning system.

[0031] Data transmission optimization and maintenance strategy

[0032] Based on the real-time bandwidth status of edge computing nodes, data priority queues are used to compress and optimize transmission of cleaned data. Specifically, transmission channels are divided according to the urgency of the data priority queue. A lossless compression algorithm is used for high-priority warning data to ensure the complete transmission of critical information; a lossy compression algorithm is used for low-frequency monitoring data to reduce data transmission volume. Bandwidth resources are dynamically allocated between edge nodes and cloud servers, and data segment size is adjusted based on transmission delay feedback. For example, when bandwidth is sufficient, the data segment size is increased to improve transmission efficiency, and when bandwidth is limited, the data segment size is reduced to reduce transmission delay.

[0033] Based on the early warning results, the weights of monitoring parameters are dynamically adjusted, and newly added fault samples are iteratively updated to the fault prediction model to achieve closed-loop optimization of the preventive maintenance strategy. Specifically, based on the historical response data of the early warning signals, a maintenance knowledge base containing equipment life prediction models is constructed. Based on the maintenance records and remaining life prediction results in the knowledge base, the unit maintenance cycle and spare parts replacement strategy are dynamically optimized. For example, for components that frequently fail, the maintenance cycle is shortened and spare parts are replaced in advance; for components that operate stably, the maintenance cycle is extended to reduce maintenance costs. The geographical location information of the edge nodes is combined to generate a distributed maintenance task scheduling queue to achieve efficient allocation and execution of maintenance tasks.

[0034] System architecture and functional modules

[0035] The present invention also provides a wind farm edge computing data cleaning and real-time transmission optimization system, which includes the following modules:

[0036] Data Acquisition Module: Deployed on edge computing nodes, it is used to obtain real-time SCADA operation data, CMS vibration monitoring data, and meteorological and environmental data. This module uses multiple sensors and data interfaces to ensure the real-time and integrity of the data.

[0037] Data Cleansing Module: This module uses a dynamic weighting algorithm to remove outliers, filter noise, and perform time-domain alignment on heterogeneous multi-source data, generating standardized data streams. This module dynamically adjusts data processing strategies based on the reliability and relevance of data sources to improve data quality.

[0038] Fusion Analysis Module: This module uses a deep dual-channel convolutional neural network to perform feature fusion and fault prediction on cleaned data, outputting health status assessment values and risk levels. This module dynamically calculates data source weights through a cross-source data association unit to improve the accuracy of fault prediction.

[0039] Transmission Optimization Module: This module prioritizes data queues based on bandwidth availability and employs differentiated compression strategies for real-time data transmission. This module integrates a bandwidth-aware unit that dynamically adjusts the compression algorithm level based on real-time network load, ensuring efficient and reliable data transmission.

[0040] Maintenance Decision Module: This module builds a preventive maintenance knowledge base based on early warning results and dynamically generates equipment maintenance task scheduling plans. This module incorporates a built-in lifespan prediction algorithm and combines geographic information to generate optimized spare parts logistics routes, achieving closed-loop optimization of maintenance strategies.

[0041] Visual Interaction Module: This module displays warning results and receives user feedback, iteratively updating model parameters through edge-cloud collaboration. This module supports user-defined warning rules and optimizes warning thresholds and model parameters through reinforcement learning algorithms, achieving a closed-loop collaborative decision-making process.

[0042] Through the collaborative operation of the above modules, the present invention achieves efficient processing of multi-source heterogeneous data in wind farms, fault warning, and optimized data transmission, thereby improving the efficiency and reliability of wind farm operations and maintenance. This system not only monitors the operating status of wind turbines in real time but also uses intelligent algorithms to predict potential faults, enabling proactive maintenance measures and reducing both failure rates and repair costs. Furthermore, by optimizing the data transmission process, the system reduces network bandwidth usage, improves real-time performance, and enhances responsiveness, providing strong technical support for intelligent wind farm operations and maintenance.

[0043] It is understood from common technical knowledge that the present invention may be implemented by other embodiments that do not depart from its spirit or essential features. Therefore, the embodiments disclosed above are, in all respects, merely illustrative and not exclusive. All modifications within the scope of the present invention or equivalent to the scope of the present invention are intended to be encompassed by the present invention.

Claims

1. A wind farm edge computing data cleaning and real-time transmission optimization method, characterized in that: The following steps are involved: Real-time collection of SCADA operation data, CMS vibration monitoring data, and meteorological environment data from wind turbines to form multi-source heterogeneous data streams; The multi-source heterogeneous data is standardized at the edge computing node to construct a multi-dimensional feature vector with a timestamp; a dynamic weight-based adaptive fusion algorithm is used to perform cross-source feature correlation analysis on the multi-dimensional feature vector to generate a spatiotemporally aligned fusion data set; a fault prediction model is constructed based on a deep dual-channel convolutional neural network, and the model is supervised and trained using historical fault samples to obtain a multi-level fault probability prediction function; the fusion data set is input into the trained fault prediction model to output the health status assessment value and fault risk level of the unit's transmission chain components in real time; When the fault risk level exceeds the dynamic warning threshold, a warning signal including the fault type, location information and repair suggestions is generated; Based on the real-time bandwidth status of the edge computing node, a data priority queue is used to compress and optimize the transmission of the cleaned data; the monitoring parameter weights are dynamically adjusted according to the early warning results, and the newly added fault samples are iteratively updated to the fault prediction model to achieve closed-loop optimization of the preventive maintenance strategy.

2. The wind farm edge computing data cleaning and real-time transmission optimization method according to claim 1 is characterized in that: The adaptive fusion algorithm based on dynamic weights includes: establishing a cross-source data association matrix based on the Pearson correlation coefficient, calculating the dynamic weight coefficients of each data source and the target fault type; using a sliding time window to align the multidimensional feature vectors in the time domain, and generating a fused feature vector with spatiotemporal consistency through weighted superposition; performing outlier removal and noise filtering on the fused data stream to form a standardized input sequence; and dynamically adjusting the calculation granularity of the fusion algorithm according to the computing resource load of the edge node.

3. The wind farm edge computing data cleaning and real-time transmission optimization method according to claim 1 is characterized in that: The deep dual-channel convolutional neural network includes: the first channel is used to process SCADA time series data, and an LSTM network with an attention mechanism is used to extract equipment operating status features; the second channel is used to process CMS high-frequency vibration signals, and a multi-layer wavelet convolution kernel is used to extract frequency domain fault features; the dual-channel output is feature-concatenated through a fully connected layer, and a residual network is introduced to optimize feature transfer efficiency; the output layer combines the Softmax function and the Sigmoid function to generate a multi-dimensional fault probability distribution and risk level label.

4. The wind farm edge computing data cleaning and real-time transmission optimization method according to claim 1 is characterized in that: The generation of the dynamic warning threshold includes: constructing a dynamic threshold curve based on seasonal wind speed changes according to the health status assessment value; triggering a three-level warning mechanism when the assessment value continuously exceeds the threshold curve and the duration reaches a preset window; the warning signal uses a fuzzy logic algorithm to integrate the failure probability, equipment criticality and current maintenance resource availability to generate a differentiated emergency response strategy; and adjusting the slope parameter of the threshold curve in real time to adapt to the calculation delay of the edge node.

5. The wind farm edge computing data cleaning and real-time transmission optimization method according to claim 1 is characterized in that: The closed-loop optimization of the preventive maintenance strategy includes: building a maintenance knowledge base containing equipment life prediction models based on historical response data of early warning signals; dynamically optimizing unit maintenance cycles and spare parts replacement strategies based on maintenance records and remaining life prediction results in the knowledge base; and generating a distributed maintenance task scheduling queue based on the geographic location information of edge nodes.

6. The wind farm edge computing data cleaning and real-time transmission optimization method according to claim 1 is characterized in that: The iterative update of the model includes: when the confidence of a new fault sample exceeds a preset threshold, the model parameters are updated using an online incremental learning algorithm; for low-confidence samples, the manual labeling module is triggered and adversarial training is performed with historical data, and the redundant computing resources of edge nodes are used to fine-tune the model; and a model version management mechanism is established to accommodate the environmental differences of different wind farms.

7. The wind farm edge computing data cleaning and real-time transmission optimization method according to claim 1 is characterized in that: The real-time transmission optimization also includes: dividing transmission channels according to the urgency of data priority queues, using a lossless compression algorithm for high-priority warning data, and a lossy compression algorithm for low-frequency monitoring data; dynamically allocating bandwidth resources between edge nodes and cloud servers, and adjusting data segment size based on transmission delay feedback.

8. The wind farm edge computing data cleaning and real-time transmission optimization method according to claim 1 is characterized in that: Also includes: Push warning results to mobile terminals in the form of visual charts, and support user-defined warning rules; Based on user feedback data, the warning threshold and model parameters are optimized through reinforcement learning algorithms; a data synchronization mechanism between edge nodes and the cloud is established to achieve a closed loop of human-machine collaborative decision-making.

9. A wind farm edge computing data cleaning and real-time transmission optimization system, based on the wind farm edge computing data cleaning and real-time transmission optimization method according to any one of claims 1 to 8, characterized in that: include: The data acquisition module is deployed on the edge computing node to obtain SCADA operation data, CMS vibration monitoring data, and meteorological environment data in real time; The data cleaning module uses a dynamic weight allocation algorithm to remove outliers, filter noise, and align time domains on multi-source heterogeneous data to generate standardized data streams; The fusion analysis module performs feature fusion and fault prediction on the cleaned data based on a deep dual-channel convolutional neural network, and outputs health status assessment values and risk levels; The transmission optimization module divides data into priority queues according to bandwidth status and adopts differentiated compression strategies for real-time data transmission; The maintenance decision module builds a preventive maintenance knowledge base based on early warning results and dynamically generates equipment maintenance task scheduling plans; The visual interaction module is used to display warning results and receive user feedback, and to iteratively update model parameters through the edge-cloud collaborative mechanism.

10. The wind farm edge computing data cleaning and real-time transmission optimization system according to claim 9 is characterized in that: The fusion analysis module also includes: a cross-source data association unit for dynamically calculating the data source weight through the Pearson correlation coefficient; the transmission optimization module integrates a bandwidth perception unit to dynamically adjust the compression algorithm level according to the real-time network load; the maintenance decision module has a built-in life prediction algorithm and combines geographic information to generate an optimized spare parts logistics path.