Dynamic correction method and system for power curve of wind power plant

Abstract

The invention discloses a dynamic correction method and system for a power curve of a wind power plant, and the method achieves the precise dynamic correction of the power curve through the multi-dimensional real-time data collection and edge calculation preprocessing in combination with a deep learning algorithm and a reinforcement learning algorithm, greatly improves the accuracy and real-time performance of power prediction, and improves the accuracy and real-time performance of power prediction. The introduction of a digital twin system enhances the reliability of an optimization strategy, the block chain technology ensures the credibility and traceability of data, the federal element learning framework realizes cross-scene knowledge migration, the generalization ability and adaptability of the model are improved, the comprehensive optimization scheme significantly improves the power generation efficiency of the wind power plant, and the power generation efficiency of the wind power plant is improved. The equipment failure rate and the operation and maintenance cost are reduced, and the service life of the fan is prolonged; meanwhile, due to the self-adaptability and expandability of the system, the system can adapt to wind power plants of different geographic positions and scales, powerful technical support is provided for intelligent and efficient development of the wind power industry, and wider application of renewable energy sources is promoted.

Description

Technical Field

[0001] The invention belongs to the technical field of wind power generation, and in particular relates to a method and system for dynamically correcting a power curve of a wind farm. Background Art

[0002] The wind farm power curve is an important indicator that describes the relationship between wind speed and wind turbine output power. It directly affects the power generation efficiency and economic benefits of the wind farm. An accurate power curve is essential for the operation management, performance evaluation and optimization control of wind farms. The traditional power curve is based on wind turbine performance tests under ideal conditions. However, in actual operation, due to the complexity of wind conditions, equipment aging, environmental factors and other reasons, the actual power curve often deviates significantly from the theoretical curve.

[0003] At present, the wind power industry generally uses static correction methods to adjust the power curve. This method is usually based on historical data for periodic correction, or uses simple mathematical models for correction. However, these methods are difficult to adapt to the complex and changeable operating environment of wind farms and cannot reflect the dynamic changes in wind turbine performance in a timely manner. In addition, existing correction methods often only consider a single or limited influencing factor, and it is difficult to fully capture the multi-dimensional characteristics that affect power output. At the same time, traditional methods also have obvious deficiencies in data credibility, cross-scenario adaptability, and real-time optimization capabilities. Summary of the invention

[0004] The object of the present invention is to provide a method and system for dynamically correcting a power curve of a wind farm, so as to solve the technical problem that the power curve correction in the prior art is difficult to adapt to the complex and changeable operating environment of the wind farm.

[0005] In order to achieve the above object, the present invention adopts the following technical scheme: A method for dynamically correcting a wind farm power curve, characterized by comprising the following steps: Collecting real-time data of the wind farm through a multi-dimensional sensor network, wherein the multi-dimensional sensor network includes an ultrasonic anemometer, a fiber grating strain sensor, a temperature and humidity sensor, an air pressure sensor, and a lidar; Preprocessing the real-time data of the wind farm using edge computing nodes, including data cleaning, anomaly detection, short-term wind speed prediction, and equipment health status assessment; The pre-processed real-time data is transmitted to the central processing system for deep feature extraction and power prediction; Optimize wind turbine control strategy based on deep deterministic policy gradient algorithm; Combining power prediction and wind turbine control strategy, the digital twin system is used to simulate and verify the optimized wind turbine control strategy to obtain the optimal wind farm power curve; Use blockchain technology to record and verify the corrected power curve data, and dynamically correct the power curve based on the optimization and verification results; Continuously monitor the correction effect and perform closed-loop optimization. During the closed-loop optimization process, the federated meta-learning framework is used to achieve cross-scenario knowledge transfer.

[0006] Furthermore, the edge computing node adopts an anomaly detection algorithm combining the isolation forest and the local anomaly factor, and fuses the results of the two algorithms through an adaptively adjusted weight coefficient; The edge computing nodes include ARM processors, FPGA accelerators, and high-speed SSD storage to perform data preprocessing and short-term wind speed prediction tasks.

[0007] Furthermore, the central processing system uses an LSTM network combined with a multi-scale temporal attention mechanism to perform power prediction.

[0008] Furthermore, the deep deterministic policy gradient algorithm adopts a hybrid reward function, comprehensively considering normalized power output, equipment load, expected life and environmental impact factors, and the weight coefficient of each factor is dynamically adjusted.

[0009] Furthermore, the digital twin system adopts an adaptive model fusion method to combine the physical model and the data-driven model, and the fusion weight is dynamically adjusted according to real-time data and model performance.

[0010] Furthermore, the digital twin system includes a 3D modeling module, a physical simulation module and a data synchronization module, which are used to create a virtual twin of the wind farm and perform aerodynamic-structural coupling analysis.

[0011] Furthermore, the blockchain technology builds a consortium chain based on Hyperledger Fabric, including taking each wind turbine as a peer node, the central control system as an orderer node, and using a practical Byzantine fault-tolerant algorithm as a consensus mechanism.

[0012] Furthermore, the federated meta-learning framework combines Model-Agnostic Meta-Learning (MAML) and federated learning to control local model updates and global model updates through inner and outer loop learning rates, while considering the amount of data from different wind farms for weighting.

[0013] Furthermore, the closed-loop optimization includes real-time monitoring of power forecast accuracy, annual power generation, equipment utilization, and operation and maintenance costs, performing A / B testing to evaluate the effectiveness of different optimization strategies, and dynamically adjusting the weights of each objective in the multi-objective optimization framework.

[0014] In a second aspect, a wind farm power curve dynamic correction system is provided, including an acquisition module, a processing module, an optimization module, a verification module, a dynamic correction module, and a closed-loop optimization module, wherein: Acquisition module: used to collect real-time data of the wind farm through a multi-dimensional sensor network, wherein the multi-dimensional sensor network includes an ultrasonic anemometer, a fiber grating strain sensor, a temperature and humidity sensor, an air pressure sensor, and a lidar; Processing module: used to pre-process the real-time data of the wind farm using edge computing nodes, including data cleaning, anomaly detection, short-term wind speed prediction and equipment health status assessment; Optimization module: used to transmit the pre-processed real-time data to the central processing system, perform deep feature extraction and power prediction, obtain the wind farm power curve, optimize the wind farm power curve based on the deep deterministic policy gradient algorithm, and obtain the optimal wind farm power curve; Verification module: used to simulate and verify the optimized wind farm power curve using the digital twin system; Dynamic correction module: used to record and verify the corrected power curve data using blockchain technology, and dynamically correct the power curve according to the optimization and verification results; Closed-loop optimization module: used to continuously monitor the correction effect and perform closed-loop optimization. During the closed-loop optimization process, the federated meta-learning framework is used to achieve cross-scenario knowledge transfer.

[0015] Compared with the prior art, the present invention has the following beneficial technical effects: The present invention provides a method for dynamic correction of wind farm power curve, which realizes accurate dynamic correction of power curve through multi-dimensional real-time data collection and edge computing preprocessing, combined with deep learning and reinforcement learning algorithms, and greatly improves the accuracy and real-time performance of power prediction. The introduction of digital twin system enhances the reliability of optimization strategy, while blockchain technology ensures the credibility and traceability of data. The federated meta-learning framework realizes cross-scenario knowledge transfer and improves the generalization ability and adaptability of the model. This comprehensive optimization scheme significantly improves the power generation efficiency of wind farms, reduces equipment failure rate and operation and maintenance costs, and extends the service life of wind turbines. At the same time, the adaptability and scalability of the present invention enable it to adapt to wind farms of different geographical locations and sizes, providing strong technical support for the intelligent and efficient development of the wind power industry and promoting the wider application of renewable energy.

[0016] The present invention realizes a comprehensive, accurate and dynamic correction process of the wind farm power curve, significantly improving the overall performance and adaptability of the wind farm. This method not only solves the limitations of traditional static correction methods, but also realizes intelligent and efficient management of wind farms through the innovative application of technologies such as multi-source data fusion, deep learning, reinforcement learning, digital twins and blockchain.

[0017] Preferably, the central processing system first performs deep feature extraction to identify the key factors affecting wind turbine performance, and then uses an LSTM network combined with a multi-scale temporal attention mechanism to perform power prediction. The proposed multi-scale temporal attention mechanism can simultaneously capture short-term, medium-term and long-term temporal dependencies.

[0018] Preferably, the digital twin system performs simulation verification on the optimization results. The system first inputs the prediction results and optimization strategies into a high-fidelity three-dimensional model, and then the algorithm performs aerodynamic-structural coupling analysis to evaluate the impact of blade deformation on power generation efficiency, thereby improving the accuracy and reliability of simulation verification. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of a method for dynamic correction of a wind farm power curve according to an embodiment of the present invention; Figure 2 This is an architecture diagram of a wind farm power curve dynamic correction system according to Embodiment 1 of the present invention; Figure 3 This is a flow chart of a method for dynamically correcting a wind farm power curve according to Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0021] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] The present invention is further described in detail below in conjunction with the accompanying drawings: like Figure 1 As shown, a method for dynamically correcting a wind farm power curve comprises the following steps: Collecting real-time data of the wind farm through a multi-dimensional sensor network, wherein the multi-dimensional sensor network includes an ultrasonic anemometer, a fiber grating strain sensor, a temperature and humidity sensor, an air pressure sensor, and a lidar; Preprocessing the real-time data of the wind farm using edge computing nodes, including data cleaning, anomaly detection, short-term wind speed prediction, and equipment health status assessment; The pre-processed real-time data is transmitted to the central processing system for deep feature extraction and power prediction; Optimize wind turbine control strategy based on deep deterministic policy gradient algorithm; Combining power prediction and wind turbine control strategy, the digital twin system is used to simulate and verify the optimized wind turbine control strategy to obtain the optimal wind farm power curve; Use blockchain technology to record and verify the corrected power curve data, and dynamically correct the power curve based on the optimization and verification results; Continuously monitor the correction effect and perform closed-loop optimization. During the closed-loop optimization process, the federated meta-learning framework is used to achieve cross-scenario knowledge transfer.

[0024] Specifically, each step is described in detail below. Figure 2 As shown: 1. Multi-dimensional real-time data collection This step collects real-time data from the wind farm through a multi-dimensional sensor network. The system uses ultrasonic anemometers to measure wind speed and direction, fiber grating strain sensors to monitor the deformation of wind turbine blades, temperature and humidity sensors to record ambient temperature and humidity, air pressure sensors to measure atmospheric pressure, and lidar to scan the entire wind farm to obtain three-dimensional wind speed distribution data. These multi-source data provide comprehensive input for subsequent analysis and optimization.

[0025] In order to improve the accuracy and reliability of data collection, the present invention adopts adaptive sampling frequency technology. The system dynamically adjusts the sampling frequency of the sensor according to the changes in wind conditions:

[0026] in, is the real-time sampling frequency, is the base sampling frequency (usually 1Hz), is the adjustment coefficient, is the short-term wind speed standard deviation. When the wind speed fluctuates greatly, the system automatically increases the sampling frequency to ensure that the rapidly changing wind conditions are captured; when the wind speed is relatively stable, the sampling frequency is reduced to save computing resources. This method significantly improves data quality while optimizing system resource utilization.

[0027] 2. Edge computing preprocessing The edge computing node preprocesses the collected raw data. The system first performs data cleaning to remove outliers and noise to ensure data quality. In order to improve the accuracy and efficiency of anomaly detection, this paper introduces a new lightweight anomaly detection algorithm in the edge computing layer, combining the advantages of isolation forest and local anomaly factor:

[0028] Among them, AS(x) is the anomaly score, IF(x) is the isolation forest score, LOF(x) is the local anomaly factor score, and α is the weight coefficient, which is dynamically adjusted in an adaptive manner. This method can efficiently detect abnormal data on edge devices and lay the foundation for subsequent processing.

[0029] In addition, the present invention introduces a data compression algorithm based on wavelet transform in the edge computing node to reduce the amount of data transmission and improve real-time performance:

[0030]

[0031] Among them, DWT is discrete wavelet transform, IDWT is inverse discrete wavelet transform, and Threshold is threshold function. This algorithm reduces the amount of data by about 60% while retaining the key features of the data, significantly reducing the network bandwidth requirements and the computational burden of the central processing system.

[0032] Next, the algorithm performs short-term wind speed forecasting, predicting wind speed changes in the next 5-15 minutes, providing a basis for real-time control. The system also calculates the real-time power curve of a single machine and evaluates the health status of the equipment based on vibration analysis and temperature monitoring data.

[0033] 3. Central data processing and deep learning prediction The central processing system performs in-depth analysis and prediction on the preprocessed data. The system first performs in-depth feature extraction to identify the key factors affecting wind turbine performance. Then, the LSTM network is used in combination with a multi-scale time attention mechanism to predict power and obtain the wind farm power curve. The multi-scale time attention mechanism proposed in this invention can simultaneously capture short-term, medium-term and long-term time dependencies:

[0034] Where Q, K, and V represent query, key, and value matrices, respectively, and i represents different time scales. is the corresponding weight. This mechanism significantly improves the model's ability to capture features of different time scales, thereby improving the accuracy of power prediction.

[0035] Finally, the algorithm evaluates the confidence of the prediction results and provides a reliability reference for subsequent decision-making.

[0036] In order to further improve the prediction accuracy, this paper introduces a hybrid model based on the Transformer-based attention mechanism combined with a convolutional neural network (CNN):

[0037]

[0038]

[0039] Among them, MultiHeadAttention is a multi-head attention layer, CNN is a convolutional layer, and FullyConnected is a fully connected layer. This hybrid model can capture both the long-term dependencies and local features of the time series, and reduced the prediction error by about 15% in the experiment.

[0040] 4. Reinforcement Learning Control Optimization The system optimizes the wind turbine control strategy based on the Deep Deterministic Policy Gradient (DDPG) algorithm. First, the algorithm defines the state space including multi-dimensional features such as wind speed, wind direction, temperature, humidity, etc. Then, the system determines the action space, including pitch angle adjustment and yaw angle control.

[0041] In order to comprehensively consider the operational objectives of the wind farm, this paper proposes a new hybrid reward function for reinforcement learning control optimization:

[0042] Where R is the total reward, P is the normalized power output, L is the normalized equipment load, T is the normalized expected life, and E is the normalized environmental impact factor. , , , is a dynamically adjusted weight coefficient. This function comprehensively considers power generation efficiency, equipment life and environmental impact, and achieves a multi-objective balance.

[0043] Finally, the system performs strategy optimization and generates the optimal control strategy, that is, the optimal wind farm power curve.

[0044] The present invention also introduces a reinforcement learning algorithm based on Soft Actor-Critic (SAC) to improve the stability and sample efficiency of the control strategy:

[0045] in, is the policy function, is the reward function, is the policy entropy, is the temperature parameter. The SAC algorithm achieves a better balance between exploration and utilization by maximizing the weighted sum of expected return and policy entropy, making the control strategy more robust under complex wind conditions.

[0046] 5. Digital Twin Simulation Verification The digital twin system simulates and verifies the optimization results. The system first inputs the prediction results and optimization strategies into a high-fidelity 3D model. Then, the algorithm performs an aerodynamic-structural coupling analysis to evaluate the impact of blade deformation on power generation efficiency.

[0047] In order to fully utilize the advantages of physical models and data-driven models, the present invention implements an adaptive model fusion method:

[0048] in, For the fusion model, For the physical model, is a data-driven model, and β is an adaptive weight that is dynamically adjusted according to real-time data and model performance. This method makes full use of the interpretability of the physical model and the adaptability of the data-driven model to improve the accuracy and reliability of simulation verification.

[0049] 6. Blockchain data recording and verification The system uses blockchain technology to record and verify the corrected power curve data. The specific process includes: packaging the corrected data into transactions, the consensus algorithm verifies the transactions in the alliance chain network, the system records the verified transactions into blocks, and the smart contract automatically executes the data verification and power curve update process. This method ensures the immutability and traceability of the data.

[0050] 7. Dynamic Correction Implementation The system implements dynamic corrections based on the optimization and verification results. First, the system updates the wind turbine control parameters, including pitch angle and yaw angle. Then, the controller adjusts the wind turbine operating status to optimize power output. Finally, the system records the correction effect to provide a basis for subsequent optimization.

[0051] 8. Cross-scenario transfer learning optimization In order to improve the generalization ability of the model and realize cross-scenario knowledge transfer, this paper proposes a new federated meta-learning framework that combines Model-Agnostic Meta-Learning (MAML) and federated learning:

[0052] in, represents the global model parameters; represents the updated global model parameters; Represents the inner loop learning rate, used for local model update; Represents the outer loop learning rate, which is used for global model update; represents the total number of wind farms participating in federated learning; represents the index of a specific wind farm, ranging from 1 to K; represents the data volume of the k-th wind farm; Represents the total data volume of all wind farms, that is ; represents the loss function of the k-th wind farm; represents the gradient operator with respect to the parameter θ.

[0053] The framework first trains the model locally on multiple wind farms, then the system aggregates the local model parameters and updates the global model. The algorithm performs a meta-learning process to improve the model's ability to adapt quickly. Finally, the system distributes the optimized model to each wind farm. This approach protects data privacy and improves the model's generalization ability.

[0054] In order to solve the problem of large differences in data distribution of different wind farms, the present invention introduces domain adaptation technology:

[0055] in, is the task-related loss, For adversarial loss, By minimizing the difference in feature distribution between the source domain and the target domain, this method significantly improves the generalization ability of the model in new wind farms, reducing the prediction error after migration by about 20% on average.

[0056] 9. Closed-loop monitoring and continuous optimization The system continuously monitors the correction effect and optimizes it. Specifically, it monitors the power prediction accuracy, annual power generation, equipment utilization and operation and maintenance costs in real time; analyzes performance indicators and identifies areas that need to be optimized; regularly performs A / B tests to evaluate the effects of different optimization strategies; dynamically adjusts the weights of each goal in the multi-objective optimization framework; and updates model parameters and control strategies based on the optimization results. This closed-loop optimization mechanism ensures continuous improvement of system performance.

[0057] Example 2 See also Figure 3 , the system of the present invention includes the following main components: 1. Data Acquisition System The system consists of a variety of high-precision sensors, including ultrasonic anemometers, fiber grating strain sensors, temperature and humidity sensors, air pressure sensors, and lidar. These sensors are distributed at key locations in the wind farm to collect multi-dimensional data in real time.

[0058] 2. Edge computing system Each wind turbine is equipped with an edge computing node, which includes an ARM processor, FPGA accelerator and high-speed SSD storage. The system is responsible for tasks such as data preprocessing, short-term wind speed prediction and equipment health status assessment.

[0059] 3. Central Processing System This is the core of the entire system and includes the following modules: Deep learning module: LSTM combined with multi-scale temporal attention mechanism is used for power prediction; Reinforcement learning module: Optimize control strategy based on improved DDPG algorithm; Data preprocessing module: data cleaning, standardization and feature extraction; Cross-scenario learning module: implements federated meta-learning and improves model generalization capabilities.

[0060] 4. Application System It includes power curve correction module, control strategy optimization module, predictive maintenance module and decision support module. These modules convert the processing results into specific application outputs.

[0061] 5. Digital Twin System Comprising 3D modelling modules, physical simulation modules and data synchronization modules, the system creates a virtual twin of the wind farm for strategy verification and performance evaluation.

[0062] 6. Blockchain system Built on Hyperledger Fabric, it includes smart contracts and consensus mechanism modules. The system ensures the trusted storage and verification of data.

[0063] Each system is interconnected through a high-speed network to form a complete closed-loop control system. The data acquisition system transmits the original data to the edge computing system for preliminary processing, and then sends it to the central processing system for in-depth analysis and optimization. After the optimization results are verified by the digital twin system, the application system performs specific correction and optimization operations. The key data of the entire process is recorded and verified by the blockchain system to ensure the credibility and traceability of the data.

[0064] The present invention also provides a wind farm power curve dynamic correction system, comprising an acquisition module, a processing module, an optimization module, a verification module, a dynamic correction module, and a closed-loop optimization module, wherein: Acquisition module: used to collect real-time data of the wind farm through a multi-dimensional sensor network, wherein the multi-dimensional sensor network includes an ultrasonic anemometer, a fiber grating strain sensor, a temperature and humidity sensor, an air pressure sensor, and a lidar; Processing module: used to pre-process the real-time data of the wind farm using edge computing nodes, including data cleaning, anomaly detection, short-term wind speed prediction and equipment health status assessment; Optimization module: used to transmit the pre-processed real-time data to the central processing system, perform deep feature extraction and power prediction, obtain the wind farm power curve, optimize the wind farm power curve based on the deep deterministic policy gradient algorithm, and obtain the optimal wind farm power curve; Verification module: used to simulate and verify the optimized wind farm power curve using the digital twin system; Dynamic correction module: used to record and verify the corrected power curve data using blockchain technology, and dynamically correct the power curve according to the optimization and verification results; Closed-loop optimization module: used to continuously monitor the correction effect and perform closed-loop optimization. During the closed-loop optimization process, the federated meta-learning framework is used to achieve cross-scenario knowledge transfer.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims of the invention.

Claims

1. A method for dynamic correction of wind farm power curve, characterized in that: The following steps are involved: Collecting real-time data of the wind farm through a multi-dimensional sensor network, wherein the multi-dimensional sensor network includes an ultrasonic anemometer, a fiber grating strain sensor, a temperature and humidity sensor, an air pressure sensor, and a lidar; Preprocessing the real-time data of the wind farm using edge computing nodes, including data cleaning, anomaly detection, short-term wind speed prediction, and equipment health status assessment; The pre-processed real-time data is transmitted to the central processing system for deep feature extraction and power prediction; Optimize wind turbine control strategy based on deep deterministic policy gradient algorithm; Combining power prediction and wind turbine control strategy, the digital twin system is used to simulate and verify the optimized wind turbine control strategy to obtain the optimal wind farm power curve; Use blockchain technology to record and verify the corrected power curve data, and dynamically correct the power curve based on the optimization and verification results; Continuously monitor the correction effect and perform closed-loop optimization. During the closed-loop optimization process, the federated meta-learning framework is used to achieve cross-scenario knowledge transfer.

2. A method for dynamic correction of wind farm power curve according to claim 1, characterized in that: The edge computing node adopts an anomaly detection algorithm that combines isolation forest and local anomaly factor, and fuses the results of the two algorithms through adaptively adjusted weight coefficients; The edge computing nodes include ARM processors, FPGA accelerators, and high-speed SSD storage to perform data preprocessing and short-term wind speed prediction tasks.

3. A method for dynamic correction of wind farm power curve according to claim 1, characterized in that: The central processing system uses an LSTM network combined with a multi-scale temporal attention mechanism to perform power prediction.

4. A method for dynamic correction of wind farm power curve according to claim 1, characterized in that: The deep deterministic policy gradient algorithm adopts a hybrid reward function, comprehensively considers normalized power output, equipment load, expected life and environmental impact factors, and the weight coefficient of each factor is dynamically adjusted.

5. A method for dynamic correction of wind farm power curve according to claim 1, characterized in that: The digital twin system adopts an adaptive model fusion method to combine the physical model and the data-driven model, and the fusion weight is dynamically adjusted according to the real-time data and model performance.

6. A method for dynamic correction of wind farm power curve according to claim 1 or 5, characterized in that: The digital twin system includes a 3D modeling module, a physical simulation module and a data synchronization module, which are used to create a virtual twin of the wind farm and perform aerodynamic-structural coupling analysis.

7. A method for dynamic correction of wind farm power curve according to claim 1, characterized in that: The blockchain technology builds a consortium chain based on Hyperledger Fabric, including taking each wind turbine as a peer node, the central control system as an orderer node, and using a practical Byzantine fault-tolerant algorithm as a consensus mechanism.

8. A method for dynamic correction of wind farm power curve according to claim 1, characterized in that: The proposed federated meta-learning framework combines Model-Agnostic Meta-Learning (MAML) and federated learning, controls local model updates and global model updates through inner and outer loop learning rates, and considers the data volume of different wind farms for weighting.

9. A method for dynamic correction of wind farm power curve according to claim 1, characterized in that: The closed-loop optimization includes real-time monitoring of power prediction accuracy, annual power generation, equipment utilization and operation and maintenance costs, performing A / B testing to evaluate the effects of different optimization strategies, and dynamically adjusting the weights of each objective in the multi-objective optimization framework.

10. A wind farm power curve dynamic correction system, characterized in that: It includes acquisition module, processing module, optimization module, verification module, dynamic correction module and closed-loop optimization module, among which: Acquisition module: used to collect real-time data of wind farms through a multi-dimensional sensor network, wherein the multi-dimensional sensor network includes an ultrasonic anemometer, a fiber grating strain sensor, a temperature and humidity sensor, an air pressure sensor, and a lidar; Processing module: used to pre-process the real-time data of the wind farm using edge computing nodes, including data cleaning, anomaly detection, short-term wind speed prediction and equipment health status assessment; Optimization module: used to transmit the pre-processed real-time data to the central processing system, perform deep feature extraction and power prediction, obtain the wind farm power curve, optimize the wind farm power curve based on the deep deterministic policy gradient algorithm, and obtain the optimal wind farm power curve; Verification module: used to simulate and verify the optimized wind farm power curve using the digital twin system; Dynamic correction module: used to record and verify the corrected power curve data using blockchain technology, and dynamically correct the power curve according to the optimization and verification results; Closed-loop optimization module: used to continuously monitor the correction effect and perform closed-loop optimization. During the closed-loop optimization process, the federated meta-learning framework is used to achieve cross-scenario knowledge transfer.

Images