Wind power generation operation monitoring and control system and methods

Abstract

This invention discloses a wind power generation operation monitoring and control system and method, belonging to the field of new energy power generation technology. The system includes an analysis module, an optimization module, a control module, and a feedback module. The analysis module receives wind power equipment operation data and uses a generative adversarial network (GAN) and Transformer architecture to achieve cross-domain data mapping, time-series prediction, and state assessment. The optimization module combines prediction results with equipment damage changes, introducing uncertainty constraints and a multi-objective weighted mechanism to generate operation control parameters. The control module controls equipment operation based on these parameters and executes local emergency responses in abnormal situations. The feedback module dynamically updates model parameters using the control results and optimizes maintenance strategies through cloud-based reinforcement learning. This invention can improve operational stability, maintenance efficiency, and fault prediction capabilities, and is suitable for the intelligent operation management of wind power equipment.

Description

Technical Field

[0001] This invention belongs to the field of new energy power generation technology, and particularly relates to a wind power generation operation monitoring and control system and method. Background Technology

[0002] As a crucial component of clean and renewable energy, wind power has seen accelerated deployment globally in recent years. Wind turbines are typically located in wind farms with complex geographical environments and variable climates, making their operation susceptible to various factors such as wind speed fluctuations, temperature and humidity changes, and terrain disturbances. This results in highly nonlinear and uncertain operating conditions. Against this backdrop, monitoring the operational status, health assessment, predictive control, and fault early warning of wind power equipment have become critical to ensuring the safe, stable, and efficient operation of wind power systems.

[0003] In existing technologies, the condition monitoring and predictive control of wind power equipment generally suffers from the following problems: Traditional data-driven models often rely on data from specific wind turbines or wind farms for training, lacking adaptability to multiple wind turbines and cross-wind farms, and failing to effectively handle differences in operating characteristics between wind turbines, affecting model transferability and versatility. Some methods only rely on static features or short-term windows for state judgment, ignoring the long-term dependence, seasonal fluctuations, and potential trends during wind power equipment operation, resulting in limited prediction accuracy and anomaly detection capabilities. Fixed threshold or rule-driven anomaly identification methods are difficult to adapt to complex environmental changes and lack mechanisms for dynamically adjusting thresholds based on prediction uncertainty and state assessment, easily leading to false alarms or missed alarms. Traditional operation control parameters are mostly optimized based on a single objective (such as optimal power generation efficiency), without fully considering multi-dimensional constraints and objectives such as equipment fatigue damage, load balancing, and uncertain disturbances, making it difficult to achieve system-level operational optimization. Some solutions only focus on forward prediction and offline optimization, failing to build an edge-cloud collaborative feedback learning path and lacking the ability to update models and control strategies online as operation evolves.

[0004] Therefore, there is an urgent need to provide an intelligent monitoring and decision-making system that integrates cross-wind turbine operation data feature extraction, time series modeling in complex environments, dynamic anomaly detection, multi-objective robust control, and edge-cloud adaptive feedback optimization capabilities to improve the predictive operation and maintenance level and the overall life cycle efficiency of wind power generation equipment. Summary of the Invention

[0005] This invention is proposed based on the above-mentioned needs, aiming to provide a wind power generation operation monitoring and control system and method with cross-device data adaptation, time-series prediction and intelligent decision-making capabilities, so as to improve the overall operation and maintenance intelligence level and operation efficiency of wind farms.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] On the one hand, the present invention provides a wind power generation operation monitoring and control system, including:

[0008] An analysis module is used to receive operating data from wind power generation equipment and perform time-series prediction, anomaly detection, and equipment status assessment on the operating data. Specifically: the analysis module includes a domain adaptation submodule based on an adversarial generative network with constraints that preserve fault-sensitive features, used to map the operating data of the target wind turbine across domains to data with a distribution consistent with the operating data of the reference wind turbine while retaining fault-sensitive features; the analysis module also includes a time-series prediction submodule based on a Transformer decoder structure combined with a dynamic attention allocation mechanism, used to perform time-series feature encoding, dynamically increase the weight of anomaly-related features during decoding, output prediction results, and generate equipment status assessment information based on the prediction results.

[0009] The optimization module receives prediction results and equipment status assessment information, introduces uncertainty constraints in the multi-objective optimization solution, and dynamically adjusts the weights of the multi-objectives according to the equipment fatigue damage change rate in order to solve and output the operating control parameters.

[0010] The control module is used to receive the operation control parameters, perform operation control operations on the wind power generation equipment, and perform local control actions when the abnormality detection result indicates that there is an emergency abnormality, and output the execution result data.

[0011] The feedback module is used to update the analysis module and optimization module based on the execution result data, and to dynamically optimize the maintenance time window and maintenance resource scheduling through the cloud maintenance decision module based on reinforcement learning.

[0012] Preferably, the domain adaptation submodule employs a cyclic consistent generative adversarial network that introduces fault-sensitive feature weighted constraints into the loss function to perform cross-domain mapping on the operating data of the target wind turbine and output time series data that is consistent with the distribution of the reference wind turbine operating data and retains the fault-sensitive features.

[0013] Preferably, the time-series prediction submodule receives the time-series data and wind field environmental condition vector output by the domain adaptation submodule. Based on the model parameters pre-trained in the energy, meteorology and environment fields and fine-tuned by migration from historical data of the target wind field, it performs time-series feature encoding on the time-series data. During the decoding process, it uses a dynamic attention allocation mechanism to increase the weight of abnormal correlation features and outputs the prediction results within the prediction time window. The prediction results include power generation, bearing temperature rise and power curve deviation information, and equipment status assessment information is generated based on the prediction results.

[0014] Preferably, the analysis module includes a dynamic threshold update step when performing anomaly detection, the dynamic threshold update step including:

[0015] Calculate the mean error between the predicted and measured values ​​within the sliding time window;

[0016] Calculate the standard deviation of the error between the predicted and measured values ​​within the same time window;

[0017] A new anomaly detection threshold is generated by weighting the model output prediction uncertainty measure, mean error, and standard error.

[0018] The new anomaly detection threshold is compared with the real-time predicted value to output the anomaly detection result, and the equipment status assessment information is updated based on the anomaly detection result.

[0019] Preferably, the optimization module includes the following steps when solving for the operating control parameters:

[0020] Receive the prediction results and corresponding equipment status assessment information output by the analysis module;

[0021] In the optimization solution, the confidence interval of the prediction result is introduced as an uncertainty constraint, and the equipment operating state parameters in the equipment state assessment information are used as operating constraints.

[0022] Based on the uncertainty constraints and the equipment operation constraints, a multi-objective weighted optimization operation is performed on the power generation, unit load distribution and equipment fatigue damage, and the weight coefficients of each optimization objective are dynamically adjusted according to the equipment fatigue damage change rate.

[0023] Output the operating control parameters corresponding to each wind power generation device.

[0024] Preferably, the optimization module includes the following steps when calculating the operation control parameters of the entire wind farm:

[0025] Periodically receive the prediction results and corresponding equipment status assessment information of all wind power generation equipment in the wind farm;

[0026] In the optimization solution, the confidence interval of the prediction results of each wind power generation device is used as an uncertainty constraint, and the equipment operation status parameters in the equipment status assessment information are used as equipment operation constraints.

[0027] The optimization objectives include the total power generation of the entire wind farm, the load distribution balance of the units, and the fatigue damage index of the equipment. The weight coefficients of the optimization objectives of the entire wind farm are dynamically adjusted according to the rate of change of the load distribution balance of each unit.

[0028] Output the operating control parameters corresponding to each wind power generation device in the entire wind farm.

[0029] Preferably, the control module includes the following steps after executing the operation control parameters:

[0030] Receive the execution result data corresponding to the operation control parameters;

[0031] Calculate the deviation between the operation control parameters and the actual response of the equipment based on the execution result data;

[0032] When the deviation exceeds the preset abnormal threshold, update the local control rule parameters and generate new operating control parameters;

[0033] The updated operation control parameters will be applied to subsequent operation control operations.

[0034] Preferably, the feedback module is connected to the cloud-based maintenance decision module, which includes the following steps:

[0035] Periodically receive long-term operating data and equipment status assessment information;

[0036] Calculate the equipment performance degradation trend and failure probability based on long-term operating data;

[0037] The maintenance time window is selected based on the performance degradation trend and the probability of failure, and maintenance tasks are generated.

[0038] Based on the results of maintenance tasks, reinforcement learning is used to adaptively adjust the maintenance time window and maintenance resource scheduling parameters.

[0039] Preferably, the prediction result, the anomaly detection result, and the device status assessment information are transmitted via a message queue, and the communication and feedback mechanism includes the following steps:

[0040] The message queue is managed based on device identifier, event type and timestamp. The optimization module and the control module obtain messages and perform corresponding operations according to preset subscription rules, and dynamically adjust the message subscription priority according to the confidence interval of the prediction result.

[0041] After completing the control or optimization operation, the control module and the feedback module feed back the execution result data to the analysis module through a message queue for model parameter updates. When the anomaly detection result exceeds the dynamic threshold, the model parameters or anomaly detection threshold are updated.

[0042] On the other hand, the present invention provides a wind power generation operation monitoring and control method, applied to the wind power generation operation monitoring and control system as described above, comprising the following steps:

[0043] Step 1: Receive the operating data of the wind power generation equipment, and perform time-series prediction, anomaly detection, and equipment status assessment on the operating data;

[0044] Step 2: Through the domain adaptation submodule, based on the generative adversarial network and introducing fault-sensitive feature preservation constraints, the operating data of the target wind turbine is mapped to data that is consistent with the distribution of the operating data of the reference wind turbine and retains the fault-sensitive features, and the prediction results, anomaly detection results and equipment status assessment information are output.

[0045] Step 3: Receive the prediction results and the equipment status assessment information, solve for the operation control parameters based on the confidence interval of the prediction results and the equipment status assessment information, and output the operation control parameters;

[0046] Step 4: Receive the operation control parameters, perform operation control operations on the wind power generation equipment, and when the abnormal detection result indicates that there is an emergency abnormality, perform local control actions and output the execution result data;

[0047] Step 5: Receive the execution result data and use the execution result data to update the model parameters or anomaly detection thresholds in the analysis module and optimization module.

[0048] The beneficial effects of this invention are as follows: This invention achieves refined management of the operating status of wind power equipment through a modular architecture. Specifically, the analysis module integrates Generative Adversarial Networks (GANs) and Transformer structures, enabling the extraction of cross-domain temporal features from multi-source heterogeneous data. This achieves faithful modeling of fault-sensitive features and accurate status assessment, significantly improving the accuracy and robustness of anomaly identification and time-series prediction. The optimization module introduces a dynamic weighting mechanism based on prediction confidence constraints and fatigue-driven mechanisms, coordinating power output, load allocation, and fatigue accumulation control in multi-objective optimization, improving the system's safety and operating efficiency under complex conditions. The control module triggers rapid local responses based on anomaly detection results, possessing proactive safety protection capabilities and effectively addressing sudden anomaly risks. The feedback module achieves adaptive model updates based on control execution results and optimizes maintenance and repair rhythms and resource scheduling through cloud-based reinforcement learning, realizing dynamic evolution of operation and maintenance strategies and full lifecycle management of equipment. The overall system integrates prediction, optimization and control functions, with a clear closed-loop technical path and close collaboration between modules. It has significant advantages such as improving the operational stability of wind power equipment, reducing operation and maintenance costs, and enhancing fault prediction capabilities, and has good engineering promotion value and industrial application prospects. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the 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. Wherein:

[0050] Figure 1 This is a schematic diagram of the overall system structure of the present invention;

[0051] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, this is an embodiment of the present invention, which provides a wind power generation operation monitoring and control system, including:

[0054] (1) Analysis Module

[0055] This module is used to receive operating data from wind power generation equipment and perform time-series prediction, anomaly detection, and equipment status assessment on the operating data. Specifically: the analysis module includes a domain adaptation submodule based on an adversarial generative network with constraints that preserve fault-sensitive features, used to map the target wind turbine's operating data across domains to data with a distribution consistent with the reference wind turbine's operating data while retaining fault-sensitive features; the analysis module also includes a time-series prediction submodule based on a Transformer decoder structure combined with a dynamic attention allocation mechanism, used to perform time-series feature encoding, dynamically increase the weight of anomaly-related features during decoding, output prediction results, and generate equipment status assessment information based on the prediction results.

[0056] Furthermore, to improve the generalization performance and sensitive feature preservation capability of cross-domain data alignment, the domain adaptation submodule employs a cyclic consistent generative adversarial network (GAN) that introduces fault-sensitive feature weighted constraints into the loss function. This network performs cross-domain mapping on the operating data of the target wind turbine and outputs time-series data that is consistent with the distribution of the reference wind turbine operating data while preserving fault-sensitive features. The loss function can be expressed as:

[0057] ;

[0058] in: To standardize the fight against losses, For cycle consistency loss, The weighted constraint for KL divergence, which addresses the differences in the distribution of sensitive channels, takes the following form:

[0059] ;

[0060] in: , The reference wind turbine and the mapped wind turbine are respectively in the 1st... Probability distribution along the fault-related feature dimension As an experience-weighted factor, , This is a hyperparameter that can be dynamically set based on the performance of the validation set.

[0061] Specifically, the time-series prediction submodule receives time-series data and wind field environmental condition vectors output by the domain adaptation submodule. Based on model parameters pre-trained in the energy, meteorology, and environment fields and fine-tuned by migration from historical data of the target wind field, it performs time-series feature encoding on the time-series data. During the decoding process, it uses a dynamic attention allocation mechanism to increase the weight of abnormal correlation features and outputs prediction results within the prediction time window. The prediction results include power generation, bearing temperature rise, and power curve deviation information. Based on the prediction results, it generates equipment status assessment information.

[0062] During the attention allocation phase, an abnormal saliency enhancement factor is introduced to weight the standard attention mechanism, as shown below:

[0063] ;

[0064] in: The standard attention score is calculated from the inner product of the query vector and the key vector. These are outlier scoring factors extracted based on statistical measures such as prediction residuals and variance. To amplify the weighting coefficients, data-driven learning can be used.

[0065] The analysis module includes a dynamic threshold update step when performing anomaly detection. The dynamic threshold update step includes:

[0066] Calculate the mean error between the predicted and measured values ​​within the sliding time window;

[0067] Calculate the standard deviation of the error between the predicted and measured values ​​within the same time window;

[0068] A new anomaly detection threshold is generated by weighting the model output prediction uncertainty measure, mean error, and standard error.

[0069] The calculation method for the new threshold can be expressed as follows:

[0070] ;

[0071] in: This represents the mean prediction error within the sliding window. The standard deviation of the error; A measure of uncertainty for the model’s predicted output, such as based on Bayesian output or Dropout estimation; , , Weighting coefficients, satisfying It can be dynamically adjusted according to the scenario.

[0072] The new anomaly detection threshold is compared with the real-time predicted value to output the anomaly detection result, and the equipment status assessment information is updated based on the anomaly detection result.

[0073] (2) Optimization module

[0074] The optimization module is used to receive prediction results and equipment status assessment information, introduce uncertainty constraints in the multi-objective optimization solution, and dynamically adjust the multi-objective weights according to the equipment fatigue damage change rate to solve and output the operation control parameters. The optimization module preferably includes an operation sub-module and a weight adjustment sub-module to jointly realize collaborative operation control optimization based on safety, economy and reliability.

[0075] Specifically, the optimization module includes the following steps when solving for the operating control parameters:

[0076] First, the system receives the prediction results and corresponding equipment status assessment information output by the analysis module. The prediction results include the power generation, bearing temperature rise, and power curve deviation within the future time window. The equipment status assessment information includes the current fatigue life level of the unit, load variation index, and operating temperature rise redundancy margin, etc. The prediction results are based on the decoding output of the time series large model and include corresponding confidence intervals to express the degree of uncertainty of the prediction results.

[0077] In the optimization solution, the confidence interval of the predicted result is introduced as an uncertainty constraint term to limit the stability of the control solution within the range of predicted value variation. The confidence constraint can be defined as:

[0078] ;

[0079] in: To predict expected power generation, Given its confidence interval half-width, The confidence level threshold is used as the basis for determining the operating status parameters. At the same time, the equipment operating status parameters in the equipment status assessment information are used as operating constraints, such as the maximum temperature rise threshold and the fatigue cumulative damage rate not exceeding the limit. These parameters are then embedded into the optimization model in the form of inequalities to ensure that the optimization results meet the requirements for safe operation of the fan.

[0080] In constructing the objective function, the following multi-objective weighted form is adopted:

[0081] ;

[0082] in: For the predicted power generation under the current optimized control strategy; For unit load fluctuation; For equipment fatigue damage indicators; weighting coefficient , , It can be dynamically adjusted based on the current rate of change of fatigue damage in the equipment to meet the following requirements: ;

[0083] To achieve dynamic weight adjustment, the weight adjustment submodule introduces the following update rules:

[0084] ;

[0085] in: This is a weighted sensitivity coefficient used to amplify the impact of fatigue damage changes on the control strategy. When the equipment enters the fatigue acceleration phase, the weights of safety-related objectives are automatically increased.

[0086] In the calculation of control parameters for the entire wind farm, the optimization module is further extended to a centralized scheduling mode, periodically receiving the prediction results and status assessment information of all wind power generation equipment within the wind farm; at this time, the following is introduced into the optimization objective:

[0087] Maximize the total power generation of the entire wind farm;

[0088] Minimize the load distribution uniformity of units within the wind farm (e.g., the difference between maximum and minimum loads);

[0089] Targets include controlling the average fatigue index throughout the game.

[0090] Its overall objective function is in the form of:

[0091] ;

[0092] The corresponding full wind farm operation control parameters are output through distributed solving or cloud-based centralized optimization strategies.

[0093] Finally, the optimization module outputs the corresponding operating control parameters for the wind power generation equipment, which are then executed by the control module to ensure that power generation efficiency is improved and the equipment lifespan is extended while ensuring equipment safety.

[0094] (3) Control module

[0095] Used to receive the operation control parameters, perform operation control operations on the wind power generation equipment, and execute local control actions when the abnormality detection result indicates that there is an emergency abnormality, and output execution result data;

[0096] Specifically, after executing the operating control parameters, the control module includes the following steps:

[0097] Receive the execution result data corresponding to the operation control parameters;

[0098] The execution result data includes wind turbine operation feedback information, such as active power feedback value, speed, pitch angle, cooling system operating status, etc. The data is collected by the underlying sensors and transmitted to the control module in real time through the industrial fieldbus to ensure the accuracy of control response evaluation.

[0099] Calculate the deviation between the operation control parameters and the actual response of the equipment based on the execution result data;

[0100] The deviation calculation adopts a dynamic error analysis method based on a sliding time window, which compares the target control value with the real-time feedback value, and establishes a multi-level tolerance evaluation interval in combination with environmental disturbance factors to dynamically identify the degree of operational deviation.

[0101] When the deviation exceeds the preset abnormal threshold, update the local control rule parameters and generate new operating control parameters;

[0102] The update of local control rule parameters includes adjustments to proportional-integral-derivative (PID) parameters, adaptive hysteresis compensation coefficients, or fuzzy control weights. The control module automatically selects the adjustment strategy based on the type of deviation to enhance the response capability to sudden anomalies and improve control stability.

[0103] The updated operation control parameters will be applied to subsequent operation control operations.

[0104] This parameter will be incorporated into the next control cycle through a real-time scheduling mechanism, covering key aspects of wind turbine equipment such as pitch control, excitation, and power output, and continuously monitoring its feedback effect. Furthermore, the control module will record control commands, feedback results, and deviation information and push them to the feedback module for subsequent model updates and status evaluation.

[0105] (4) Feedback module

[0106] This is used to update the analysis module and optimization module based on the execution result data, and to dynamically optimize the maintenance time window and maintenance resource scheduling through the cloud maintenance decision module based on reinforcement learning.

[0107] The feedback module is connected to the cloud-based maintenance decision module, which includes the following steps:

[0108] Periodically receive long-term operating data and equipment status assessment information;

[0109] Calculate the equipment performance degradation trend and failure probability based on long-term operating data;

[0110] The probability of failure can be modeled based on an exponential decay model, as shown in the following formula:

[0111] ;

[0112] in: Indicates time Predicted failure probability of time-sensitive equipment; The prior historical failure probability; This is a performance trend decay factor; For the equipment in time The rate of performance degradation.

[0113] The maintenance time window is selected based on the performance degradation trend and the probability of failure, and maintenance tasks are generated.

[0114] When generating a maintenance task scheduling plan, the feedback module can construct the scheduling objective function as follows:

[0115] ;

[0116] in: To maintain the task scheduling objective function; For the first The time cost corresponding to each maintenance task; For the task The urgency weighting coefficient; This represents the total number of all maintenance tasks that need to be scheduled.

[0117] Based on the results of maintenance tasks, reinforcement learning is used to adaptively adjust the maintenance time window and maintenance resource scheduling parameters.

[0118] In reinforcement learning scheduling optimization, the system uses the following reward function for policy training:

[0119] ;

[0120] in: This refers to the improvement in equipment reliability after performing maintenance tasks; The resource overhead corresponding to the current schedule; , To adjust the ratio between reliability gain and cost.

[0121] Through the above methods, the feedback module can dynamically update the analysis and optimization module with execution data, and realize intelligent scheduling of maintenance resources based on long-term trend prediction, thereby effectively improving the operation and maintenance efficiency and safety of wind power equipment throughout its entire life cycle.

[0122] The prediction results, the anomaly detection results, and the device status assessment information are transmitted via a message queue. The communication and feedback mechanism includes the following steps:

[0123] The message queue is managed based on device identifier, event type and timestamp. The optimization module and the control module obtain messages and perform corresponding operations according to preset subscription rules, and dynamically adjust the message subscription priority according to the confidence interval of the prediction result.

[0124] After completing the control or optimization operation, the control module and the feedback module feed back the execution result data to the analysis module through a message queue for model parameter updates. When the anomaly detection result exceeds the dynamic threshold, the model parameters or anomaly detection threshold are updated.

[0125] like Figure 2 The above is another embodiment of the present invention, which provides a wind power generation operation monitoring and control method, applied to the wind power generation operation monitoring and control system described above, and includes the following steps:

[0126] Step 1: Receive the operating data of the wind power generation equipment, and perform time-series prediction, anomaly detection, and equipment status assessment on the operating data;

[0127] Step 2: Through the domain adaptation submodule, based on the generative adversarial network and introducing fault-sensitive feature preservation constraints, the operating data of the target wind turbine is mapped to data that is consistent with the distribution of the operating data of the reference wind turbine and retains the fault-sensitive features, and the prediction results, anomaly detection results and equipment status assessment information are output.

[0128] Step 3: Receive the prediction results and the equipment status assessment information, solve for the operation control parameters based on the confidence interval of the prediction results and the equipment status assessment information, and output the operation control parameters;

[0129] Step 4: Receive the operation control parameters, perform operation control operations on the wind power generation equipment, and when the abnormal detection result indicates that there is an emergency abnormality, perform local control actions and output the execution result data;

[0130] Step 5: Receive the execution result data and use the execution result data to update the model parameters or anomaly detection thresholds in the analysis module and optimization module.

[0131] In summary, the wind power generation operation monitoring and control system and method provided by this invention constructs an end-to-cloud collaborative overall solution around four core aspects: wind power equipment operation status perception, intelligent prediction, optimized control, and adaptive feedback. The system incorporates several advanced technologies in areas such as cross-domain data adaptation, time-series modeling, anomaly identification, and operational decision-making. These include a generative adversarial network with constraints based on fault-sensitive features, a Transformer decoder structure combined with dynamic attention mechanisms, a multi-objective optimization strategy integrating uncertainty and fatigue-driving factors, and a maintenance resource scheduling method based on reinforcement learning, forming a complete closed-loop intelligent operation and maintenance system. This system not only significantly improves the accuracy of wind farm operation status identification and the reliability of control response but also possesses excellent self-learning and generalization capabilities, making it suitable for wind power scenarios of various scales and operating conditions. This technology has significant advantages in ensuring power generation efficiency, extending equipment lifespan, and reducing operation and maintenance costs, and has broad promotional value and application prospects.

[0132] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0133] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0134] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A wind power generation operation monitoring and control system, characterized in that, include: An analysis module is used to receive operating data from wind power generation equipment and perform time-series prediction, anomaly detection, and equipment status assessment on the operating data. Specifically: the analysis module includes a domain adaptation submodule based on an adversarial generative network with constraints that preserve fault-sensitive features, used to map the operating data of the target wind turbine across domains to data with a distribution consistent with the operating data of the reference wind turbine while retaining fault-sensitive features; the analysis module also includes a time-series prediction submodule based on a Transformer decoder structure combined with a dynamic attention allocation mechanism, used to perform time-series feature encoding, dynamically increase the weight of anomaly-related features during decoding, output prediction results, and generate equipment status assessment information based on the prediction results. The optimization module receives prediction results and equipment condition assessment information, introduces uncertainty constraints in the multi-objective optimization solution, dynamically adjusts the weights of the multi-objectives according to the rate of change of equipment fatigue damage, and solves and outputs the operating control parameters. The control module is used to receive the operation control parameters, perform operation control operations on the wind power generation equipment, and execute local control actions when an abnormality detection result indicates that there is an emergency abnormality, and output the execution result data. The feedback module is used to update the analysis module and optimization module based on the execution result data, and to dynamically optimize the maintenance time window and maintenance resource scheduling through the cloud maintenance decision module based on reinforcement learning. The domain adaptation submodule employs a cyclic consistent generative adversarial network that introduces fault-sensitive feature weighted constraints into the loss function. This network is used to perform cross-domain mapping on the operating data of the target wind turbine and output time-series data that is consistent with the distribution of the reference wind turbine operating data and retains the fault-sensitive features. The time-series prediction submodule receives time-series data and wind field environmental condition vectors output by the domain adaptation submodule. Based on model parameters pre-trained in the energy, meteorology, and environment fields and fine-tuned by migration from historical data of the target wind field, it performs time-series feature encoding on the time-series data. During the decoding process, it uses a dynamic attention allocation mechanism to increase the weight of abnormal correlation features and outputs prediction results within the prediction time window. The prediction results include power generation, bearing temperature rise, and power curve deviation information. Based on the prediction results, it generates equipment status assessment information.

2. The wind power generation operation monitoring and control system according to claim 1, characterized in that, The analysis module includes a dynamic threshold update step when performing anomaly detection. The dynamic threshold update step includes: Calculate the mean error between the predicted and measured values ​​within the sliding time window; Calculate the standard deviation of the error between the predicted and measured values ​​within the same time window; A new anomaly detection threshold is generated by weighting the model output prediction uncertainty measure, mean error, and standard error. The new anomaly detection threshold is compared with the real-time predicted value to output the anomaly detection result, and the equipment status assessment information is updated based on the anomaly detection result.

3. The wind power generation operation monitoring and control system according to claim 1, characterized in that, The optimization module includes the following steps when solving for the operating control parameters: Receive the prediction results and corresponding equipment status assessment information output by the analysis module; In the optimization solution, the confidence interval of the prediction result is introduced as an uncertainty constraint, and the equipment operating state parameters in the equipment state assessment information are used as operating constraints. Based on the uncertainty constraints and the equipment operation constraints, a multi-objective weighted optimization operation is performed on the power generation, unit load distribution and equipment fatigue damage, and the weight coefficients of each optimization objective are dynamically adjusted according to the equipment fatigue damage change rate. Output the operating control parameters corresponding to each wind power generation device.

4. The wind power generation operation monitoring and control system according to claim 3, characterized in that, The optimization module includes the following steps when calculating the operational control parameters of the entire wind farm: Periodically receive the prediction results and corresponding equipment status assessment information of all wind power generation equipment in the wind farm; In the optimization solution, the confidence interval of the prediction results of each wind power generation device is used as an uncertainty constraint, and the equipment operation status parameters in the equipment status assessment information are used as equipment operation constraints. The optimization objectives include the total power generation of the entire wind farm, the load distribution balance of the units, and the fatigue damage index of the equipment. The weight coefficients of the optimization objectives of the entire wind farm are dynamically adjusted according to the rate of change of the load distribution balance of each unit. Output the operating control parameters corresponding to each wind power generation device in the entire wind farm.

5. The wind power generation operation monitoring and control system according to claim 1, characterized in that, The control module, upon receiving the operation control parameters, includes the following steps: Receive the execution result data corresponding to the operation control parameters; Calculate the deviation between the operation control parameters and the actual response of the equipment based on the execution result data; When the deviation exceeds the preset abnormal threshold, update the local control rule parameters and generate new operating control parameters; The updated operation control parameters will be applied to subsequent operation control operations.

6. The wind power generation operation monitoring and control system according to claim 1, characterized in that, The feedback module is connected to the cloud-based maintenance decision module, which includes the following steps: Periodically receive long-term operating data and equipment status assessment information; Calculate the equipment performance degradation trend and failure probability based on long-term operating data; The maintenance time window is selected based on the performance degradation trend and the probability of failure, and maintenance tasks are generated. Based on the results of maintenance tasks, reinforcement learning is used to adaptively adjust the maintenance time window and maintenance resource scheduling parameters.

7. The wind power generation operation monitoring and control system according to claim 1, characterized in that, The prediction results, the anomaly detection results, and the device status assessment information are transmitted via a message queue. The communication and feedback mechanism includes the following steps: The message queue is managed based on device identifier, event type and timestamp. The optimization module and the control module obtain messages and perform corresponding operations according to preset subscription rules, and dynamically adjust the message subscription priority according to the confidence interval of the prediction result. After completing the control or optimization operation, the control module and the feedback module feed back the execution result data to the analysis module through a message queue for model parameter updates. When the anomaly detection result exceeds the dynamic threshold, the model parameters or anomaly detection threshold are updated.

8. A wind power generation operation monitoring and control method, applied to the wind power generation operation monitoring and control system as described in any one of claims 1-7, characterized in that, Includes the following steps: Step 1: Receive the operating data of the wind power generation equipment, and perform time-series prediction, anomaly detection, and equipment status assessment on the operating data; Step 2: Through the domain adaptation submodule, based on the generative adversarial network and introducing fault-sensitive feature preservation constraints, the operating data of the target wind turbine is mapped to data that is consistent with the distribution of the operating data of the reference wind turbine and retains the fault-sensitive features, and the prediction results, anomaly detection results and equipment status assessment information are output. Step 3: Receive the prediction results and the equipment status assessment information, solve for the operation control parameters based on the confidence interval of the prediction results and the equipment status assessment information, and output the operation control parameters; Step 4: Receive the operation control parameters, perform operation control operations on the wind power generation equipment, and when the abnormal detection result indicates that there is an emergency abnormality, perform local control actions and output the execution result data; Step 5: Receive the execution result data and use the execution result data to update the model parameters or anomaly detection thresholds in the analysis module and optimization module.

Images