A method and system for multi-source data integration in wind power load testing systems

Abstract

This invention discloses a method and system for multi-source data integration in wind power load testing systems, relating to the field of data processing technology. The method includes: determining N load scenarios; traversing the N load scenarios to construct a Markov load network; developing a data integration module within the wind power load testing system, performing supervised training based on the Markov load network, and establishing a direct data interface connection between the data integration module and a distributed sensor array; generating test targets through the wind power load testing system, controlling the distributed sensor array to perform load test sampling based on a front-end configured differential frequency sampler, and determining multi-source sampling data; transmitting the multi-source sampling data back via the data interface and combining it with the data integration module to perform data integration processing based on load scenario changes, determining the integrated load data, and storing it. This solves the technical problem of low multi-source data integration efficiency in existing wind power load testing systems, achieving the technical effect of improving data integration efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to a method and system for integrating multi-source data for wind power load testing systems. Background Technology

[0002] Load testing is a crucial step in assessing the structural reliability and operational safety of wind turbines during operation and maintenance. Existing wind power load testing systems typically rely on distributed sensors to collect load data from different parts of the turbine. However, due to the complex structure of wind turbines and the variability of load conditions, data collected from different sensors suffers from spatiotemporal asynchrony and inconsistent formats, making multi-source data integration difficult. Furthermore, existing data processing methods lack effective modeling of dynamic load changes, failing to accurately reflect the load response characteristics of the turbine under different operating conditions, thus limiting the in-depth analysis and application of test data. Especially in complex wind environments, load changes in wind turbines are not only related to external factors such as wind speed and direction, but also closely related to the turbine's control attitude and operating mode. Traditional data processing methods struggle to fully capture the dynamic correlations between these multi-dimensional factors, resulting in insufficient reliability and accuracy of test results. Summary of the Invention

[0003] This application provides a method and system for integrating multi-source data for wind power load testing systems, which solves the technical problem of low efficiency in integrating multi-source data in existing wind power load testing systems.

[0004] In view of the above problems, this application provides a method and system for integrating multi-source data for wind power load testing systems.

[0005] A first aspect of this application provides a method for integrating multi-source data for a wind power load testing system, the method comprising:

[0006] For wind turbines, N load scenarios are determined based on the load of key components and the turbine's control attitude. A Markov load network is constructed by traversing these N load scenarios, where load scenarios are nodes and connections are established based on the integration of scenario transitions between nodes. A data integration module is developed within the wind power load testing system, using the Markov load network as a benchmark for supervised training, and a direct data interface is established between the data integration module and the distributed sensor array. Test targets are generated through the wind power load testing system, and the distributed sensor array is controlled to perform load testing sampling based on the differential frequency sampler configured at the front end, determining multi-source sampling data. The multi-source sampling data is returned via the data interface and combined with the data integration module for data integration processing based on load scenario transitions, determining and storing the integrated load data, where transition analysis is performed based on the upper-level scenario nodes.

[0007] A second aspect of this application provides a multi-source data integration system for a wind power load testing system, the system comprising:

[0008] The system comprises the following components: a scenario determination unit, used to determine N load scenarios for wind turbines based on the load of key components and the turbine's control attitude; a load network construction unit, used to traverse the N load scenarios and construct a Markov load network, where load scenarios are nodes and connections are established based on the integration of scenario changes between nodes; a training unit, used to develop a data integration module within the wind power load testing system, performing supervised training based on the Markov load network, and establishing a direct data interface connection between the data integration module and the distributed sensor array; a sampling unit, used to generate test targets through the wind power load testing system, controlling the distributed sensor array to perform load test sampling based on the differential frequency sampler configured at the front end, and determining multi-source sampling data; and a data processing unit, used to return the multi-source sampling data through the data interface and combine it with the data integration module to perform data integration processing based on load scenario changes, determining and storing the integrated load data, wherein change analysis is performed based on the upper-level scenario nodes.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] First, for wind turbine generators, N load scenarios are determined based on the load of key components and the generator's control attitude. Next, a Markov load network is constructed by traversing these N load scenarios, with load scenarios as nodes and transitions between nodes defined by an integration method based on scenario changes. Further, a data integration module is developed within the wind power load testing system, using the Markov load network as a benchmark for supervised training, and a direct data interface is established between the data integration module and the distributed sensor array. Then, the wind power load testing system generates test targets, and based on the differential frequency sampler configured at the front end, controls the distributed sensor array to perform load testing sampling, determining multi-source sampling data. Finally, the multi-source sampling data is returned via the data interface and combined with the data integration module for data integration processing based on load scenario changes, determining and storing the integrated load data, where transition analysis is performed based on the upper-level scenario nodes. This solves the technical problem of low multi-source data integration efficiency in existing wind power load testing systems, achieving a significant improvement in data integration efficiency. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of a multi-source data integration method for a wind power load testing system provided in an embodiment of this application.

[0013] Figure 2 This is a schematic diagram of the structure of a multi-source data integration system for a wind power load testing system provided in an embodiment of this application.

[0014] Figure labeling: Scene determination unit 11, payload network construction unit 12, training unit 13, sampling unit 14, data processing unit 15. Detailed Implementation

[0015] This application solves the technical problem of low efficiency in multi-source data integration in existing wind power load testing systems by providing a method and system for integrating multi-source data for wind power load testing systems.

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0017] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0018] Example 1, as Figure 1 As shown, this application provides a method for integrating multi-source data for a wind power load testing system, wherein the method includes:

[0019] For wind turbine units, N load scenarios are determined based on the loads of key components and the control attitude of the unit.

[0020] In wind turbine units, different operating conditions (such as wind speed, wind direction, blade angle, and unit attitude) can significantly affect the load distribution of key components (such as blades, hub, tower, main shaft, and gearbox). Therefore, in order to comprehensively analyze the load characteristics of the unit under different operating conditions, it is necessary to determine N representative load scenarios based on the loads of key components and the unit's control attitude. Specifically, firstly, load data for key components of the wind turbine are collected using a distributed sensor array, including blade bending moment, hub torque, tower axial force, and gearbox shear force. Simultaneously, control attitude parameters such as yaw angle, pitch angle, and rotational speed are acquired. The collected data is preprocessed to remove outliers and synchronized with the time data to ensure accuracy and consistency. Next, a multi-dimensional feature vector is constructed based on the relationship between load and control attitude, combining load parameters with attitude parameters to form a complete dataset. Then, clustering algorithms (such as K-means or DBSCAN) are used to perform cluster analysis on the multi-dimensional data, automatically dividing it into N load scenarios, covering typical operating conditions as shown in Table 1, including low wind speed partial load conditions, medium wind speed high load conditions, high wind speed extreme yaw conditions, braking and shutdown conditions, and wind shear rapid response conditions.

[0021] Table 1 (Typical Load Conditions):

[0022] .

[0023] Traverse the N load scenarios to construct a Markov load network, where load scenarios are nodes and the connection relationship is based on the integration method of scenario changes between nodes.

[0024] A Markov load network is constructed by traversing N load scenarios. Specifically, each load scenario of the wind turbine is defined as a network node, and each load scenario represents the operating state of the turbine under specific conditions. Based on the N load scenarios, a network graph containing N nodes is constructed. The connection relationship between nodes is based on the transition between scenarios. That is, the transition between load scenarios forms the edges connecting these nodes. The transition of each scenario can be caused by changes in the external environment (such as wind speed, wind direction, etc.), turbine control strategies (such as yaw angle, pitch angle adjustment), or turbine operating status (such as fault, start-up and shutdown). For example, transitioning from a low wind speed state to a medium wind speed state, or switching from a normal operating state to a braking and shutdown state.

[0025] Furthermore, constructing a Markov payload network includes:

[0026] The communication protocol and data format for interactive multi-source data are defined, and a unified data standard is set, wherein the unified data standard can be any communication protocol and data format; using the unified data standard as a general integration standard, the N load scenarios are traversed, and the transition analysis between scenario groups and the transition analysis of integration methods are performed respectively to construct the Markov load network.

[0027] In constructing a Markov load network, the first step is to define a unified data standard to ensure smooth communication and integration of multi-source data. This unified data standard involves data communication protocols and data formats, ensuring that data generated by sensors, control systems, simulation models, and other devices from different sources can be exchanged and processed in a consistent manner. Specifically, common communication protocols (such as Modbus and CAN) and standardized data formats (such as JSON, XML, CSV, and Protocol Buffers) can be used. Based on the unified data standard, N load scenarios are traversed, and transition analysis is performed between each scenario. Each load scenario represents the state of the turbine under specific operating conditions, and the transitions between different scenarios reflect the influence of turbine control strategies, external wind force changes, or other factors. For example, wind speed changes and blade angle adjustments may cause the wind turbine to shift from one load scenario to another. Through the analysis of historical data or the processing of real-time data, the system can identify the transition patterns between these load scenarios, forming a transition matrix that represents the changing trends and transition probabilities between scenarios. Based on the load scenario transitions, further analysis of the integration method transitions is required. Integration method transition analysis aims to dynamically integrate multi-source data to ensure that the data accurately reflects changes in load characteristics during state transitions between different scenarios. Based on load scenario transitions, further integration method transition analysis is required. This analysis aims to ensure that the data accurately reflects changes in load characteristics during state transitions between different scenarios by dynamically integrating multi-source data. The choice of integration method can be based on various approaches, such as weighted averaging, interpolation, or more advanced machine learning methods, to ensure accurate reflection of the unit's load characteristics under different scenarios. Finally, based on scenario transition analysis and integration method transition analysis, a Markov load network is constructed. The nodes of the Markov load network represent different load scenarios, and the connections between nodes are determined by the transition probabilities between scenarios. These connections not only reflect the dynamic transitions between different load scenarios but also improve the accuracy and efficiency of state transitions between scenarios through data integration analysis. The weight of each edge in the network represents the transition probability between scenarios, reflecting the possibility of transitioning from one load scenario to another.

[0028] Furthermore, we will conduct transition analysis between scenario groups and transition analysis of integration methods, including:

[0029] Using the unified data standard as the general integration standard, starting with distributed source data, and guided by the data requirements based on the test objectives, the data integration method for the first payload scenario is determined by introducing integration constraints. The data integration method includes at least unified integration, differential relaxation integration, and key management integration. For the first payload scenario and the other N-1 payload scenarios, scenario change characteristics are identified and integration change characteristics based on the data integration method are mined to integrate and determine the first payload network.

[0030] Specifically, a unified data standard is used as a general integration standard to ensure efficient interaction of multi-source data across different systems and devices. The unified data standard refers to the use of consistent protocols and structures in data communication and format processing, such as standardized data formats like JSON and XML, and communication protocols like CAN, ETH, Modbus, and fiber optic communication, to ensure data consistency during acquisition, transmission, and storage. Data integration starts with distributed source data, which refers to multi-source sensor data from different parts of the wind turbine, such as blade stress sensors, main shaft torque sensors, and tower vibration sensors. These data have diversity in time, space, and physical quantities.

[0031] During data integration, data requirements are determined based on the test objectives of the wind power load testing system. These objectives may include load response analysis of wind turbines under specific wind speeds, loads, or control strategies. Guided by these data requirements, integration constraints are introduced. These constraints refer to specific requirements that must be met during data processing and analysis, such as data consistency, real-time performance, integrity, and security, ensuring high reliability and validity of the integrated data. Based on this, the data integration method for the first load scenario is determined. The data integration methods include at least three types: unified integration, differential relaxation integration, and key management integration. Unified integration refers to standardizing data from different sensors or systems to address issues such as inconsistent data formats and different sampling frequencies. For example, aligning time-series data from different sensors ensures comparison and analysis on the same timeline. Differential relaxation integration is a dynamic adjustment strategy based on data differences, primarily used to handle subtle differences between multi-source data. By introducing relaxation factors, the sensitivity of data fusion is adjusted to adapt to dynamic changes under different load scenarios. For example, when wind speed changes drastically, differential relaxation integration can automatically relax the tolerance for differences in certain data, avoiding data anomalies caused by instantaneous fluctuations. Key management integration focuses on the security of data transmission and storage. It uses key encryption technology to ensure the confidentiality and integrity of data during transmission, preventing data from being illegally tampered with or leaked. It is especially suitable for scenarios involving remote data transmission or distributed systems.

[0032] After determining the data integration method for the first load scenario, the transition relationship between this scenario and the remaining N-1 load scenarios is further analyzed. Scenario transition characteristics refer to the key factors that cause the wind turbine to shift from one operating condition to another. These characteristics may include drastic changes in wind speed, adjustments in the turbine's yaw angle, and changes in blade pitch angle. For example, a sudden increase in wind speed from 8 m / s to 15 m / s may cause the turbine to switch from a partial load state to a full load state; this change falls under the category of scenario transition characteristics. Based on identifying these characteristics, the integration transition characteristics based on the data integration method are further explored. Integration transition characteristics refer to how the data integration method adjusts to adapt to the new scenario requirements during the transition between different load scenarios. For example, when wind speed changes drastically, the system may switch from a standard unified integration mode to a more flexible differential relaxation integration mode to better cope with the uncertainty brought about by data fluctuations.

[0033] Finally, by integrating scenario transition features and integrated transition features, the system determines the structure of the first load network. The first load network is a directed graph network based on a Markov model. Nodes in the network represent different load scenarios, the connections between nodes represent transition paths between scenarios, and the weights of the connections represent the transition probabilities between scenarios. In this network, scenario transitions are based not only on changes in the physical state of the wind turbines but also on dynamic adjustments to the data integration method, ensuring that the data processing strategy can flexibly adapt to changes in the scenarios.

[0034] Furthermore, the data integration method includes at least unified integration, differential relaxation integration, and key management integration, including:

[0035] Guided by the testing objective, a differential relaxation is determined, wherein the differential relaxation is determined based on data value; using the differential relaxation as a benchmark, differential transformation processing is performed on the standardized multi-source data determined after unification and integration to determine the effective data; sensitivity determination is performed on the effective data, sensitive data parts are located and key encryption processing is performed.

[0036] Specifically, the differential relaxation is determined based on the test objectives, which typically include load response analysis of wind turbines under specific operating conditions and structural safety assessments in extreme environments. Differential relaxation is a parameter that measures the tolerance range of data differences and is used to adjust the sensitivity to differences during data integration. This differential relaxation is determined based on data value, which refers to the importance of data in achieving the test objectives. Data value is usually quantified by indicators such as data contribution, frequency of change, and probability of anomaly detection. For example, for critical load data (such as main shaft torque or tower stress data), the differential relaxation is set lower to improve the accuracy of data integration; while for auxiliary data (such as ambient temperature or humidity data), the differential relaxation can be appropriately relaxed to improve processing efficiency. After determining the differential relaxation, it is used as a benchmark to perform differential transformation processing on the standardized multi-source data after unified integration to filter out effective data with practical significance. The purpose of unified integration is to standardize data from different sensors, solve problems such as inconsistent formats and different sampling frequencies, and ensure that the data has a unified structure and timestamp. Differential transformation processing dynamically adjusts the data based on differential relaxation. It primarily calculates the difference between data points and compares it to a relaxation threshold, filtering out invalid data with fluctuations below the threshold and retaining only data with significant changes or key characteristics. This processing not only effectively reduces data redundancy and improves data processing efficiency but also highlights data change characteristics under critical operating conditions. After obtaining valid data, a sensitivity assessment is performed to locate sensitive data portions. Sensitive data determination is mainly based on data importance, sensitivity level classification, and potential security risk assessment. For sensitive data involving core wind turbine control strategies, extreme load values ​​of key components, or external communication interactions, the system will automatically mark it. After locating sensitive data, the system will activate key management integration to encrypt the sensitive data portion. Key encryption uses a dynamic key distribution mechanism combined with symmetric encryption (such as AES) or asymmetric encryption (such as RSA) algorithms to ensure the confidentiality and integrity of data during transmission and storage. Encryption not only effectively prevents data from being tampered with or leaked during transmission but also enables fine-grained control over data access permissions, ensuring that sensitive data is only available within authorized scopes.

[0037] Furthermore, the introduction of integrated constraints includes:

[0038] The load types are categorized, including static loads, dynamic loads, steady-state loads, and unsteady-state loads; the data access types are categorized, including sensitive data and non-sensitive data; the load types and data access types are traversed to determine integration constraints, wherein the integration constraints are determined based on personalized integration requirements of the load types and data access types.

[0039] Introducing integrated constraints is a crucial step in ensuring data processing efficiency and security. First, load types need to be scientifically categorized based on the operating characteristics of wind turbines, primarily including static loads, dynamic loads, steady-state loads, and unsteady-state loads. Static loads refer to the constant forces experienced by the turbine under unchanging or slowly changing operating conditions, such as the self-weight of the tower or the support load of fixed structures. Dynamic loads fluctuate continuously with time, wind speed, and rotational speed, such as the aerodynamic changes experienced by the blades during rotation. Steady-state loads typically refer to loads that remain relatively constant or fluctuate regularly during stable turbine operation, with typical scenarios including torque loads under constant wind speeds. Unsteady-state loads, on the other hand, are characterized by drastic fluctuations over short periods, often occurring during sudden wind speed changes, emergency braking, or fault conditions, exhibiting high uncertainty and suddenness. Simultaneously, data access types also need to be categorized, primarily including sensitive data and non-sensitive data. Sensitive data typically involves key information such as core operating parameters of the unit, control commands, and fault diagnosis results. Leakage or tampering of this type of data could lead to serious security risks. Non-sensitive data, on the other hand, includes routine monitoring data such as ambient temperature and wind speed. Even if anomalies occur, the impact on system security is relatively limited.

[0040] After classifying load types and data access types, the system will traverse all possible load and data combinations to determine personalized integration constraints. The formulation of integration constraints aims to meet the specific data processing needs under different combinations, ensuring that data processing maintains both high efficiency and high security and reliability during integration. For example, for combinations of dynamic loads and sensitive data, due to frequent data changes and the involvement of core control information of the wind turbine, the integration process must meet the requirements of high real-time performance, low latency, and strong consistency. Simultaneously, encryption and access control mechanisms must be introduced to prevent data leakage and unauthorized access. For combinations of static loads and non-sensitive data, integration constraints can be relatively relaxed, allowing for batch data processing and delayed transmission to reduce system resource consumption and improve data processing efficiency. In this process, integration constraints not only consider the characteristics of the data itself but also need to be dynamically adjusted according to the real-time operating status of the wind turbine to achieve adaptive optimization of data processing.

[0041] Furthermore, the formulation of integration constraints must consider data processing priorities and security strategies. Sensitive data typically has a higher processing priority and must be encrypted and integrity-verified at the initial stage of data integration, while non-sensitive data can be integrated in subsequent processing stages. For non-steady-state load scenarios, the system needs to have rapid response capabilities to adjust integration strategies in a timely manner to cope with sudden data fluctuations and ensure data real-time performance and reliability. In steady-state load scenarios, data smoothing or redundant data compression strategies can be adopted to further optimize data storage and transmission efficiency. Ultimately, through a comprehensive traversal and analysis of load types and data access types, the system can form a complete set of integration constraint rules, providing accurate, efficient, and secure technical support for data acquisition, processing, and analysis of wind turbines under different operating conditions.

[0042] A data integration module is developed within the wind power load testing system. Supervised training is performed using the Markov load network as a benchmark, and a direct data interface is established between the data integration module and the distributed sensor array.

[0043] Developing a data integration module within the wind power load testing system is a crucial step in achieving efficient fusion and intelligent analysis of multi-source data. This module undergoes supervised training based on a constructed Markov load network to improve the accuracy and robustness of load scenario identification and data processing. The Markov load network, as the core model, establishes the dynamic transition relationships of wind turbines under different load scenarios based on the changing patterns of wind turbines under various load scenarios. This is achieved by defining each load scenario as a node and the transition probabilities between scenarios as edge weights. Building upon this foundation, the data integration module is trained using a supervised learning algorithm. Utilizing historical multi-source load data as training samples, it matches the actually collected load data with known load scenarios, identifies the mapping relationship between data features and scenario transition patterns, and gradually optimizes the classification accuracy and scenario switching prediction capabilities of the integrated model.

[0044] To ensure the data integration module can process multi-source data from various components of the wind turbine in real time and efficiently, the system further establishes a direct data interface connection between the data integration module and the distributed sensor array. The distributed sensor array includes stress sensors, acceleration sensors, temperature sensors, and vibration monitoring devices deployed on key structures such as blades, hubs, main shafts, and towers, used to collect load data of the unit under different operating conditions in real time. The direct data interface design ensures that sensor data can be directly transmitted to the data integration module without intermediate conversion, reducing data latency and transmission loss, and improving the real-time performance and stability of data processing. This data interface supports multiple communication protocols, such as CAN, Modbus, and Ethernet, to adapt to the access requirements of different sensing devices, while also possessing high bandwidth and low latency characteristics to ensure reliable transmission of high-frequency sampled data.

[0045] The wind power load testing system generates test targets and controls a distributed sensor array to perform load testing sampling based on the differential frequency sampler configured at the front end, thereby determining multi-source sampling data.

[0046] In a wind power load testing system, test objectives are first generated using pre-defined rules built into the system. These objectives are typically determined based on the wind turbine's operating status, environmental conditions, and historical load data analysis results, aiming to evaluate the turbine's load response characteristics and structural safety under specific operating conditions. These test objectives may include load assessments for specific wind speed ranges, stress testing during turbine start-up and shutdown, and load variation analysis under extreme weather conditions, ensuring coverage of various critical operating conditions the turbine may face.

[0047] Once the test target is generated, the system automatically configures the front-end differential frequency sampler to achieve accurate acquisition of load data within different frequency ranges. The differential frequency sampler is a key device for optimizing signal sampling efficiency and quality. It dynamically adjusts the sampling frequency according to the needs of the test target, capturing both high-frequency and low-frequency signal components and avoiding frequency aliasing during data sampling. For example, in scenarios where the unit experiences severe load fluctuations or transient shocks, the differential frequency sampler automatically increases the sampling frequency to capture rapidly changing load data within a short period; while in steady-state operation scenarios, the sampling frequency can be appropriately reduced to decrease data redundancy and storage pressure.

[0048] After the differential frequency sampler is configured, the system controls a distributed sensor array to perform load testing sampling. The distributed sensor array is deployed at key locations on the wind turbine, including the blade root, hub, main shaft, tower midsection, and foundation, encompassing various types of monitoring equipment such as stress-strain sensors, accelerometers, torque sensors, and temperature sensors. The system efficiently links with these sensors through the differential frequency sampler to collect load data from the turbine under different operating conditions in real time, ensuring broad data coverage and high accuracy. During sampling, the system automatically calibrates sensor data, performs noise filtering and anomaly detection, ensuring the validity and reliability of the collected data. Finally, the system performs preliminary processing on the multi-source sampling data acquired from the distributed sensor array, including time synchronization, data format standardization, and preliminary data integrity verification, forming a structured load dataset.

[0049] Furthermore, controlling the distributed sensing array to perform load test sampling includes:

[0050] Guided by the test objective, the sampling requirements based on the target load scenario are determined; a timestamp constraint for distributed sampling is introduced, and in combination with the sampling requirements, the differential frequency sampler generates a sensing sampling command; the sensing sampling command responds to the target sensor to acquire the multi-source sampling data, wherein the target sensor belongs to the distributed sensing array.

[0051] Specifically, based on the generated test objectives, the sampling requirements for the target load scenarios are determined. Test objectives may include structural response under specific wind speed ranges, stress distribution under extreme load conditions, or dynamic load changes during turbine start-up and shutdown. Based on these objectives, the system analyzes the data requirements under different load scenarios, determining parameters such as sampling frequency, sampling duration, data accuracy, and target monitoring points. For example, when conducting extreme load tests under high wind speed conditions, the system requires a higher sampling frequency and finer time resolution to capture rapidly changing stress and vibration signals; while in steady-state operation scenarios, the sampling frequency can be appropriately reduced, focusing on long-term trend data. After determining the sampling requirements, the system introduces a distributed sampling timestamp constraint to ensure that data from different sensors can be accurately synchronized in time. The timestamp constraint is the core of the distributed sampling system, aiming to solve the problem of inconsistent timing of multi-source data and ensure that data collected by all sensors are comparable under the same time reference. The system uses global clock or GPS time synchronization technology to assign a high-precision timestamp to each sampled data point to achieve cross-sensor time alignment, which is particularly suitable for complex dynamic system environments such as wind turbines. Based on timestamp constraints and sampling requirements, the differential frequency sampler generates specific sensing sampling instructions. According to the operating characteristics and sampling frequency requirements of different sensors, the differential frequency sampler dynamically adjusts the sampling strategy, generating instructions including parameters such as sampling frequency, sampling duration, and data format. These sampling instructions can flexibly adapt to different types of sensors, such as stress sensors, accelerometers, torque sensors, and temperature sensors, to ensure the integrity and diversity of the sampled data. For example, when the system detects a rapid change in wind speed, the differential frequency sampler automatically adjusts the sampling frequency, increasing the sampling density of key sensors (such as blade root stress sensors) to capture details of transient load changes. The generated sensing sampling instructions are directly sent to the target sensors in the distributed sensor array. The selection of these target sensors depends on the test target and sampling requirements. After receiving the sampling instructions, the sensors collect data according to preset parameters, recording multi-source data such as load, vibration, and temperature in real time, along with timestamp information to ensure accurate data synchronization.

[0052] According to the data interface, the multi-source sampling data is returned and combined with the data integration module to perform data integration processing based on load scenario changes, determine the load integration data and store it, wherein the change analysis is based on the upper-level scenario node.

[0053] In wind power load testing systems, the backhaul and integration of multi-source sampling data based on data interfaces are crucial for efficient management and intelligent analysis of load data. First, multi-source sampling data collected by the distributed sensor array is transmitted back to the system's data integration module via a high-efficiency data interface. This interface not only supports high-speed data transmission but also boasts multi-protocol compatibility, adapting to different types of sensor data, including various physical quantities such as stress, strain, acceleration, torque, and temperature, while ensuring data integrity and real-time performance. During data transmission, the system automatically adds timestamps, sensor identifiers, and sampling parameters to ensure accurate temporal consistency and traceability in subsequent processing.

[0054] After the multi-source sampled data is transmitted back to the data integration module, the system combines it with the existing Markov load network model to perform data integration processing based on load scenario changes. Specifically, the data integration module first preprocesses the transmitted data, including data cleaning (removing outliers and noise), time alignment (synchronizing multi-source data based on timestamps), and format standardization. Subsequently, based on the scenario change path of the Markov load network, the system identifies the load scenario to which the current data belongs and its position in the historical scenario chain, and analyzes the dynamic change patterns between scenarios. During this process, the system uses the superior scenario node as a reference to conduct change analysis. The superior scenario node refers to the preceding scenario in the Markov chain of the current load scenario, that is, the state node of the unit under previous operating conditions. By comparing the data characteristics of the current data with those of the superior scenario, the system can capture subtle changes in load characteristics during scenario changes and identify potential abnormal patterns or trends, such as stress concentration and fatigue accumulation.

[0055] Based on scenario change analysis, the system deeply integrates sampled data from different sources and time periods to generate high-quality load integration data. Data integration processing includes not only simple data summarization but also advanced processing methods such as dimensionality reduction, feature extraction, and differential analysis to ensure that the integrated data comprehensively reflects the actual operating status of the wind turbine under different load scenarios. For example, when the wind turbine switches from a medium-wind-speed scenario to a high-wind-speed scenario, the system automatically adjusts the data integration strategy, increasing the weight of blade root stress and tower vibration data to more accurately reflect the structure's response characteristics under extreme conditions. After data integration, the system standardizes and stores the load integration data in formats including structured databases and distributed data warehouses to support subsequent rapid retrieval and in-depth analysis.

[0056] Furthermore, the data integration processing based on load scenario changes includes:

[0057] Using the Markov payload network, the upper-level scene node of the upper-level payload scene is located, and the lower-level scene node is determined based on the target payload scene; the connection relationship between the upper-level scene node and the lower-level scene node is determined, and a target integration method based on scene change is determined; the multi-source sampling data is integrated and processed using the target integration method, wherein the data integration module has short-term memory based on the upper-level scene node.

[0058] Specifically, based on the current target load scenario, the system locates its upper and lower level scenario nodes in the Markov load network. The upper level scenario node represents the operating conditions experienced by the unit before the current scenario, reflecting the path and characteristics of the load's historical evolution. The lower level scenario node indicates the next stage of operation the unit may transition to, predicting future load change trends. After locating the upper and lower level scenario nodes, the system further analyzes the connection relationship between these two nodes to determine the target integration method based on scenario changes. This connection relationship not only reflects the transition probability between load scenarios but also includes the dynamic characteristics of scenario changes, such as load fluctuation amplitude, transition rate, and stress concentration changes. Based on these characteristics, the system intelligently matches the most suitable data integration strategy. For example, when scenario changes involve drastic load fluctuations (such as switching from low wind speed to extreme high wind speed conditions), the system will choose a highly sensitive data integration method, focusing on capturing instantaneous load changes and local stress concentration phenomena. In steady-state scenario transitions, the system may adopt a smoothing or trend analysis-based integration method to reduce data noise and improve long-term trend recognition capabilities.

[0059] After determining the target integration method, the system performs deep integration processing on multi-source sampling data based on this method. During this process, the data integration module fully utilizes its built-in short-term memory capability based on higher-level scenario nodes to enhance the temporal correlation and historical sensitivity of data processing. The short-term memory mechanism allows the system to retain and reference key data features and load change patterns related to the higher-level scenario while performing current data integration, ensuring that data integration not only focuses on current transient data but also captures deeper information related to historical operating condition changes. Finally, the data processed by the target integration method is standardized into load integrated data, including the fusion results of multi-source information such as stress, strain, vibration, and temperature, supplemented by metadata such as scenario transition paths, timestamps, and processing logs to ensure data traceability and reusability.

[0060] In summary, the embodiments of this application have at least the following technical effects:

[0061] First, for wind turbine generators, N load scenarios are determined based on the load of key components and the generator's control attitude. Next, a Markov load network is constructed by traversing these N load scenarios, with load scenarios as nodes and transitions between nodes defined by an integration method based on scenario changes. Further, a data integration module is developed within the wind power load testing system, using the Markov load network as a benchmark for supervised training, and a direct data interface is established between the data integration module and the distributed sensor array. Then, the wind power load testing system generates test targets, and based on the differential frequency sampler configured at the front end, controls the distributed sensor array to perform load testing sampling, determining multi-source sampling data. Finally, the multi-source sampling data is returned via the data interface and combined with the data integration module for data integration processing based on load scenario changes, determining and storing the integrated load data, where transition analysis is performed based on the upper-level scenario nodes. This solves the technical problem of low multi-source data integration efficiency in existing wind power load testing systems, achieving a significant improvement in data integration efficiency.

[0062] Example 2, based on the same inventive concept as the multi-source data integration method for the wind power load testing system in the foregoing examples, such as... Figure 2 As shown, this application provides a multi-source data integration system for wind power load testing systems, wherein the system includes:

[0063] The system comprises the following components: a scenario determination unit 11, which determines N load scenarios for wind turbines based on the load of key components and the control attitude of the turbine; a load network construction unit 12, which traverses the N load scenarios to construct a Markov load network, wherein load scenarios are nodes and the connection relationship is based on the integration method of scenario changes between nodes; a training unit 13, which develops a data integration module within the wind power load testing system, performs supervised training based on the Markov load network, and establishes a direct data interface connection between the data integration module and the distributed sensor array; a sampling unit 14, which generates test targets through the wind power load testing system, controls the distributed sensor array to perform load test sampling based on the differential frequency sampler configured at the front end, and determines multi-source sampling data; and a data processing unit 15, which transmits the multi-source sampling data back through the data interface and performs data integration processing based on load scenario changes in conjunction with the data integration module, determines the integrated load data, and stores it, wherein change analysis is performed based on the upper-level scenario nodes.

[0064] Furthermore, the load network construction unit 12 is used to perform the following method:

[0065] The communication protocol and data format for interactive multi-source data are defined, and a unified data standard is set, wherein the unified data standard can be any communication protocol and data format; using the unified data standard as a general integration standard, the N load scenarios are traversed, and the transition analysis between scenario groups and the transition analysis of integration methods are performed respectively to construct the Markov load network.

[0066] Furthermore, the load network construction unit 12 is used to perform the following method:

[0067] Using the unified data standard as the general integration standard, starting with distributed source data, and guided by the data requirements based on the test objectives, the data integration method for the first payload scenario is determined by introducing integration constraints. The data integration method includes at least unified integration, differential relaxation integration, and key management integration. For the first payload scenario and the other N-1 payload scenarios, scenario change characteristics are identified and integration change characteristics based on the data integration method are mined to integrate and determine the first payload network.

[0068] Furthermore, the load network construction unit 12 is used to perform the following method:

[0069] Guided by the testing objective, a differential relaxation is determined, wherein the differential relaxation is determined based on data value; using the differential relaxation as a benchmark, differential transformation processing is performed on the standardized multi-source data determined after unification and integration to determine the effective data; sensitivity determination is performed on the effective data, sensitive data parts are located and key encryption processing is performed.

[0070] Furthermore, the load network construction unit 12 is used to perform the following method:

[0071] The load types are categorized, including static loads, dynamic loads, steady-state loads, and unsteady-state loads; the data access types are categorized, including sensitive data and non-sensitive data; the load types and data access types are traversed to determine integration constraints, wherein the integration constraints are determined based on personalized integration requirements of the load types and data access types.

[0072] Furthermore, the sampling unit 14 is used to perform the following method:

[0073] Guided by the test objective, the sampling requirements based on the target load scenario are determined; a timestamp constraint for distributed sampling is introduced, and in combination with the sampling requirements, the differential frequency sampler generates a sensing sampling command; the sensing sampling command responds to the target sensor to acquire the multi-source sampling data, wherein the target sensor belongs to the distributed sensing array.

[0074] Furthermore, the data processing unit 15 is configured to perform the following method:

[0075] Using the Markov payload network, the upper-level scene node of the upper-level payload scene is located, and the lower-level scene node is determined based on the target payload scene; the connection relationship between the upper-level scene node and the lower-level scene node is determined, and a target integration method based on scene change is determined; the multi-source sampling data is integrated and processed using the target integration method, wherein the data integration module has short-term memory based on the upper-level scene node.

[0076] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0077] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0078] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for integrating multi-source data for wind power load testing systems, characterized in that, The method includes: For wind turbine units, N load scenarios are determined based on the loads of key components and the control attitude of the unit. Traverse the N load scenarios to construct a Markov load network, where load scenarios are nodes and the connection relationship is based on the integration method of scenario changes between nodes. A data integration module is developed within the wind power load testing system. Supervised training is performed using the Markov load network as a benchmark, and a direct data interface is established between the data integration module and the distributed sensor array. The wind power load testing system generates test targets, and controls the distributed sensor array to perform load test sampling based on the differential frequency sampler configured at the front end, thereby determining multi-source sampling data. According to the data interface, the multi-source sampling data is returned and combined with the data integration module to perform data integration processing based on load scenario changes, determine the load integration data and store it, wherein the change analysis is based on the upper-level scenario node. The construction of the Markov payload network includes: Communication protocols and data formats for interacting with multi-source data, and setting a unified data standard, wherein the unified data standard is any communication protocol and data format; Using the unified data standard as the general integration standard, the N load scenarios are traversed, and the transition analysis between scenario groups and the transition analysis of integration methods are performed respectively to construct the Markov load network. The analysis of changes between scene groups and the analysis of changes in integration methods include: Using the unified data standard as the general integration standard, starting with distributed source data, and guided by the data requirements based on the test objectives, the data integration method for the first load scenario is determined by introducing integration constraints. The data integration method includes at least unified integration, differential relaxation integration, and key management integration. For the first load scenario and the remaining N-1 load scenarios, identify scenario change features and mine integration change features based on the data integration method to integrate and determine the first load network.

2. The multi-source data integration method for a wind power load testing system as described in claim 1, characterized in that, The data integration method includes at least unified integration, differential relaxation integration, and key management integration, including: The differential relaxation is determined based on the test objective, wherein the differential relaxation is determined based on the data value; Based on the differential relaxation, differential transformation is performed on the standardized multi-source data determined after unification and integration to determine the effective data; The valid data is subjected to sensitivity assessment, the sensitive data portion is located, and key encryption processing is performed.

3. The multi-source data integration method for a wind power load testing system as described in claim 1, characterized in that, The introduced integrated constraints include: The load types are classified, including static loads, dynamic loads, steady-state loads, and unsteady-state loads. Data access types are categorized, including sensitive data and non-sensitive data; Traverse the load type and the data access type to determine integration constraints, wherein the integration constraints are determined based on personalized integration requirements based on the load type and the data access type.

4. The multi-source data integration method for a wind power load testing system as described in claim 1, characterized in that, Controlling the distributed sensing array to perform load test sampling includes: Based on the testing objectives, determine the sampling requirements for the target load scenario; By introducing a timestamp constraint for distributed sampling and combining it with the sampling requirements, the difference frequency sampler generates a sensing sampling instruction. The sensing sampling command responds to the target sensor to acquire the multi-source sampling data, wherein the target sensor belongs to the distributed sensing array.

5. The multi-source data integration method for a wind power load testing system as described in claim 4, characterized in that, The data integration and processing based on load scenario changes includes: Using the Markov payload network, locate the upper-level scene node of the upper-level payload scene, and determine the lower-level scene node using the target payload scene; For the connection relationship between the upper-level scene node and the lower-level scene node, determine the target integration method based on scene changes; The multi-source sampling data is integrated and processed using the target integration method, wherein the data integration module has short-term memory based on the upper-level scene node.

6. A multi-source data integration system for wind power load testing systems, characterized in that, A method for integrating multi-source data in a wind power load testing system according to any one of claims 1-5, the system comprising: The scenario determination unit is used to determine N load scenarios for wind turbine units based on the loads of key components and the control attitude of the unit. The load network construction unit is used to traverse the N load scenarios and construct a Markov load network, wherein the load scenarios are nodes and the connection relationship is transformed by the integration method based on the scenario changes between nodes. The training unit is used to develop a data integration module within the wind power load testing system, perform supervised training based on the Markov load network, and establish a direct data interface connection between the data integration module and the distributed sensor array. The sampling unit is used to generate test targets through the wind power load testing system, control the distributed sensor array to perform load test sampling based on the differential frequency sampler configured at the front end, and determine multi-source sampling data; The data processing unit is used to transmit the multi-source sampling data back according to the data interface and combine it with the data integration module to perform data integration processing based on load scenario changes, determine the load integration data and store it, wherein the change analysis is based on the upper-level scenario node.

Images