Wind power energy-saving control method and system based on big data analysis

Abstract

The invention relates to the technical field of wind energy control systems, and discloses a wind power energy-saving control method and system based on big data analysis, and the system comprises a data collection and transmission layer, a data storage and management layer, a data analysis and prediction layer, an intelligent control layer, and an equipment full life cycle management layer. The data acquisition and transmission layer further comprises multi-source data acquisition system construction, edge computing data preprocessing and data transmission network construction. According to the intelligent control strategy, by means of a distributed cooperative control algorithm and an intelligent variable-pitch variable-speed control algorithm, the fan blade angle, the rotating speed and the power output are dynamically adjusted according to the real-time meteorological data, the power grid load requirement and the fan running state, wind power output is smoother through accurate regulation and control, and the wind power output efficiency is improved. The impact on the power grid caused by power fluctuation is effectively reduced, the wind power acceptance capability of the power grid is remarkably improved, and the stability and reliability of the operation of the power grid are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of wind energy control systems, and particularly relates to a wind power energy-saving control method and system based on big data analysis. Background Art

[0002] Wind power generation is a clean energy technology that uses wind power to drive the rotation of windmill blades, and then increases the rotation speed through a speed increaser to drive the generator to generate electricity.

[0003] However, due to the instability and intermittency of wind energy itself, the output power of wind power fluctuates greatly, making it difficult to meet the requirements of the power grid for stable power supply, and posing a great challenge to the safe and stable operation of the power grid. When the wind speed changes suddenly, the large change in wind power may cause voltage fluctuations in the power grid, affecting the normal operation of other electrical equipment. At the same time, the existing prediction models have poor adaptability to complex meteorological and geographical environments and insufficient prediction accuracy. Under complex terrain or variable meteorological conditions, key parameters such as wind speed and wind direction cannot be accurately predicted, resulting in the inability of the wind turbine to adjust its operating state in a timely manner to achieve efficient power generation, causing energy waste, and the energy-saving control strategy is relatively backward and cannot dynamically adjust the operating parameters of the wind turbine according to real-time meteorological conditions and power grid demands, further limiting the energy-saving potential of the wind power system. Summary of the Invention

[0004] The purpose of the present invention is to provide a wind power energy-saving control method and system based on big data analysis, which can effectively solve the problems in the background art.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] A wind power energy-saving control method and system based on big data analysis includes a data acquisition and transmission layer, a data storage and management layer, a data analysis and prediction layer, an intelligent control layer, and an equipment full-life cycle management layer;

[0007] The data acquisition and transmission layer further includes the construction of a multi-source data acquisition system, edge computing data preprocessing, and the construction of a data transmission network;

[0008] The data storage and management layer further includes the deepening of a hierarchical distributed storage architecture, the refinement of a data governance process, and a data security and rights management module;

[0009] The data analysis and prediction layer further includes the innovation of a multi-source data fusion algorithm, the optimization and expansion of a prediction model, and a model evaluation and update mechanism;

[0010] The intelligent control layer includes the optimization of a distributed cooperative control algorithm, the deepening of an intelligent pitch and variable speed control strategy, and control decision-making and straight-line optimization;

[0011] The full life cycle management layer of the device includes refinement of the fault diagnosis and prediction model, optimization of the formulation of the dynamic maintenance plan, improvement of the equipment health status evaluation system, and upgrade of the remote operation and maintenance and intelligent early warning system.

[0012] As a further preferred solution of the present invention, the multi-source data acquisition system is built for wind turbine data acquisition, meteorological data acquisition, and substation data acquisition. For wind turbine data acquisition, a magnetoelectric speed sensor is installed at the main shaft of the wind turbine to obtain the wind turbine speed. The output end of the generator is connected to an electronic power sensor to obtain active power, reactive power, and apparent power. Acceleration-type vibration sensors are deployed at key parts such as the wind turbine blades and nacelle to obtain equipment vibration signals.

[0013] The meteorological data acquisition uses meteorological detection equipment to obtain comprehensive meteorological data (wind speed and direction, temperature and humidity, atmospheric pressure).

[0014] The substation data acquisition is for an intelligent electricity meter to collect the voltage, current, and power factor of the substation.

[0015] The edge computing preprocessing is to deploy a high-performance industrial-grade edge computing gateway at the data acquisition point. The edge computing gateway adopts a distributed deployment method and is reasonably configured according to the equipment distribution. Preprocessing algorithms for data denoising, normalization, and outlier detection are configured. Wavelet transform is used for data denoising to remove noise interference in the data. The formula is as follows:

[0016] ,

[0017] where is the scale parameter ( ), which controls the stretching of the wavelet function; is the translation parameter, which controls the translation of the wavelet function; is the wavelet mother function, is its conjugate function. During the denoising process, usually the high-frequency coefficients are thresholded. Common threshold functions include the hard threshold function and the soft threshold function. The hard threshold function:

[0018] ,

[0019] where is the wavelet coefficient, is the threshold;

[0020] The soft threshold function:

[0021] ,

[0022] where is the sign function;

[0023] Normalization algorithm, which normalizes data of different types and magnitudes to a unified numerical range, maps the data to the [0, 1] range, and preserves the relative relationship of the data. The formula is as follows:

[0024] ,

[0025] where, is the original data, and are the minimum and maximum values in the dataset respectively, is the normalized data;

[0026] Outlier detection is based on the 3 sigma criterion of statistical analysis. Assuming that the data follows a normal distribution, in a normal distribution, the probability that the data falls within the interval is approximately 99.7%. Data outside this interval is considered abnormal data. The formula is as follows:

[0027] ,

[0028] where, is the number of data points, is the th data point. If the data point satisfies , then it is marked as abnormal data;

[0029] The edge computing device performs real-time processing and screening on the collected original data according to preset rules and algorithms, and only transmits key data and processed feature data, reducing the data transmission volume and improving the transmission efficiency;

[0030] For the construction of the data transmission network, for relatively concentrated devices such as wind turbines, wired Ethernet is used for data transmission. A fiber optic Ethernet backbone network is built inside the wind farm to achieve high-speed data interaction between devices. A redundant link design is adopted. When the main link fails, the standby link can be automatically switched to ensure the stability of data transmission. For relatively dispersed devices such as meteorological monitoring devices, 4G / 5G communication technology or Lora wireless communication technology is used for data transmission.

[0031] As a further preferred solution of the present invention, the hierarchical distributed storage architecture is deepened to use the Ceph distributed storage system to build the underlying storage foundation on the cloud computing platform. It can provide various storage methods such as object storage, block storage, and file storage to meet the storage requirements of different types of data. For real-time data, the Redis in-memory database is used as a cache, taking advantage of its high read-write speed and low latency characteristics to ensure the system's fast response to real-time data. For historical data, the distributed file system selects the Hadoop distributed file system and combines it with the Hbase distributed NoSQL database based on Hadoop to achieve efficient storage and fast retrieval of massive historical data;

[0032] The data governance process is refined as follows: for the operation data of wind turbines, the timestamp format is uniformly specified as the Iso8601 standard, and the device Id coding adopts a specific coding rule to ensure data consistency and identifiability. In the data cleaning link, data cleaning tools are used in combination with custom cleaning rules to process duplicate values, error values, and missing values in the data. For missing values, according to the characteristics and relevance of the data, the mean filling method is used for completion. At the same time, a data quality monitoring index system is established to regularly evaluate and report indexes such as data accuracy, integrity, and consistency, and timely discover and solve data quality problems;

[0033] The data security and permission management adopt data encryption technology to encrypt and store the data stored in the database and file system.

[0034] As a further preferred solution of the present invention, the multi-source data fusion algorithm is innovated to perform transfer learning on the data of different wind farms or different types of devices on the basis of the multi-source data fusion of the autoencoder and the generative adversarial network. The model parameters trained in one environment can be transferred to other environments, reducing the training time and data requirements;

[0035] The optimization and expansion of the prediction model is an adaptive dynamic prediction model based on Lstm and attention mechanism, which is optimized by combining with a reinforcement learning algorithm. Moreover, the reinforcement learning algorithm uses a deep q-network to construct a q-network. The input is the state space, and the output is the q-value of each action, representing the expected reward for taking a certain action in the current state. The model selects an action from the action space according to the current state at each time. After executing the action, it observes the new state and the obtained reward, stores the state, action, reward, and new state in the experience replay pool, and randomly samples a batch of data from the experience replay pool at regular intervals to train the q-network, updates the network parameters to maximize the cumulative reward. Subsequently, at each time, the current state is input into the Lstm and attention mechanism model for prediction to obtain the prediction result. The reward is calculated based on the prediction result and the actual value, and the state, action, reward, and new state are recorded. The parameters and prediction strategy of the model are adjusted regularly using the reinforcement learning algorithm, and the adjustment is performed at regular intervals of time or when the prediction error exceeds a certain threshold.

[0036] As a further preferred embodiment of the present invention, the optimization of the distributed cooperative control algorithm adopts a variant of the Paxos algorithm to ensure that the control instructions between each fan can be synchronized quickly and accurately. By optimizing the convergence speed and stability of the algorithm, the coordination time and error between the fans are reduced. At the same time, when adjusting the blade angle of the fan, each fan can quickly reach an agreement to avoid power loss caused by inconsistent coordination. At the same time, combined with the game theory idea, a power distribution game model between the fans is established, so that each fan can achieve the optimal power distribution of the entire wind farm while pursuing the maximization of its own power generation efficiency.

[0037] The intelligent pitch and variable speed control strategy is further refined by adopting an adaptive fuzzy sliding mode control strategy. According to the real-time wind speed, wind direction, and the operating state of the fan, the parameters of the fuzzy control rules and the sliding mode controller are dynamically adjusted. In different wind speed intervals, the fuzzy control rules are automatically adjusted to enable the blade angle and speed of the fan to more accurately adapt to the wind speed change.

[0038] The optimization of the control decision and execution adopts a real-time operating system. During the control execution process, a fault tolerance mechanism is introduced. When a fault occurs in the control execution unit of a certain fan, it can automatically switch to the standby unit to ensure the normal operation of the fan. At the same time, the control process is monitored and fed back in real time, and the control strategy is adjusted in a timely manner according to the actual operating conditions to achieve the optimal control of the fan.

[0039] As a further preferred embodiment of the present invention, the fault diagnosis and prediction model is refined to introduce a fusion model of the convolutional neural network Cnn and the recurrent neural network Rnn in deep learning to extract the local features of the data and process time series data.

[0040] The optimization of the dynamic maintenance plan formulation is based on the Reliability Centered Maintenance (RCM) method. By combining the importance assessment and risk assessment of equipment, a more scientific and reasonable dynamic maintenance plan is formulated. For critical equipment and high-risk equipment, the maintenance frequency and depth are increased; for non-critical equipment and low-risk equipment, the maintenance cost is appropriately reduced. At the same time, a maintenance resource optimization algorithm is introduced. According to the priority of maintenance tasks, equipment location, and the skill level of maintenance personnel, maintenance resources are reasonably allocated to improve maintenance efficiency and reduce maintenance costs.

[0041] The improvement of the equipment health status assessment system and the upgrade of the remote operation and maintenance and intelligent early warning system involve establishing a multi-level and multi-dimensional equipment health status assessment index system. In addition to considering the operating parameters, failure history, maintenance records, and environmental factors of the equipment, performance degradation indicators and remaining life prediction indicators of the equipment are introduced. The analytic hierarchy process (AHP) and the fuzzy comprehensive evaluation method are combined to comprehensively evaluate the health status of the equipment. For the key parameters of critical equipment, multiple early warning thresholds are set. When the parameters approach different thresholds, early warning information at different levels is sent to remind the operation and maintenance personnel to take timely measures. In addition, blockchain technology is introduced into the final data to improve the security and credibility of the data. Through the distributed ledger and encryption technology of the blockchain, the immutability and traceability of equipment operation data and operation and maintenance records are ensured.

[0042] A wind power energy-saving control method based on big data analysis includes the following steps:

[0043] S1. Data collection: At key parts such as the main shaft, generator output terminal, blades, and nacelle of the wind turbine, magnetoelectric speed sensors, electronic power sensors, and acceleration-type vibration sensors are installed to collect wind turbine speed, power, and vibration signal data in real time. At the weather station, ultrasonic wind speed and direction sensors, digital temperature and humidity sensors, piezoresistive pressure sensors, and solar radiation sensors are deployed to obtain wind speed, wind direction, temperature and humidity, and atmospheric pressure data. Through an intelligent electricity meter with an Rs485 communication interface and supporting the Modbus communication protocol, electrical parameters such as voltage, current, and power factor of the substation are collected. At the same time, a combination of video monitoring and intelligent monitoring is used to monitor the equipment status of the substation.

[0044] S2. Data transmission and preprocessing: The collected data is transmitted to the edge computing gateway through an adapted hardware interface and an optimized Mqtt communication protocol. The edge computing gateway is deployed in a distributed manner and is reasonably configured according to the equipment distribution. The gateway uses wavelet transform denoising, normalization algorithm, and 3 According to the principle, the data is denoised, normalized and outlier detected. After filtering out key data and feature data, it is transmitted to the data storage layer via wired Ethernet or 4g / 5g, and the real-time data is stored in the Redis memory database for quick response. The historical data is stored in the Hadoop distributed file system HDFS and the Hbase distributed Nosql database, and the Ceph distributed storage system is used to build the underlying storage foundation.

[0045] s3. Data analysis and prediction. Use autoencoders and generative adversarial networks combined with transfer learning and attention mechanisms to fuse multi-source data and extract key features. Build an adaptive dynamic prediction model based on LSTM and attention mechanisms. Combined with reinforcement learning algorithms, automatically adjust parameters and strategies according to prediction results and actual feedback to predict wind speed, wind direction, power generation and grid load demand. Regularly evaluate the model based on indicators such as mean square error (Mse), information gain, and Gini coefficient, and use model fusion technology to improve prediction accuracy.

[0046] s4. Intelligent control. The real-time control decision platform combines data analysis and prediction results to generate control instructions. It dynamically adjusts the wind turbine blade angle, speed and power output through the distributed collaborative control algorithm optimized by the distributed consistency algorithm, and the intelligent variable pitch and variable speed control algorithm of the adaptive fuzzy sliding mode control strategy and neural network adaptive control.

[0047] s5. Equipment life cycle management, using Cnn and Rnn fusion models combined with expert systems and knowledge graph technology to analyze equipment operation data, diagnose and predict faults, based on the reliability center maintenance Rcm method, combined with equipment importance and risk assessment, formulate dynamic maintenance plans, use maintenance resource optimization algorithms to reasonably allocate maintenance resources, and use the hierarchical analysis method AHP and fuzzy comprehensive evaluation method, combined with equipment performance degradation and remaining life prediction indicators, to evaluate the equipment health status.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. In the present invention, the intelligent control strategy of the present application uses a distributed collaborative control algorithm and an intelligent variable pitch and variable speed control algorithm to dynamically adjust the wind turbine blade angle, speed and power output according to real-time meteorological data, grid load demand and wind turbine operating status. Through precise regulation, the wind power output is smoother, which effectively reduces the impact of power fluctuations on the grid, significantly improves the grid's ability to accept wind power, and ensures the stability and reliability of grid operation.

[0050] 2. In the present invention, a model using Lstm and attention mechanism and reinforcement learning is utilized for accurate prediction. The fan parameters are dynamically adjusted through intelligent control. The model is also optimized through multi-source data fusion and transfer learning to improve power generation efficiency, reduce energy waste. In terms of reducing operation and maintenance costs, deep learning models are used to diagnose faults, and a maintenance plan is formulated based on Rcm to avoid increased energy consumption caused by faults and unreasonable maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a system flowchart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0052] To make the technical means, creative features, achieved purposes and functions of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0053] As Figure 1 shown, a wind power energy-saving control method and system based on big data analysis provided by the present invention includes a data acquisition and transmission layer, a data storage and management layer, a data analysis and prediction layer, an intelligent control layer, and an equipment full life cycle management layer;

[0054] The data acquisition and transmission layer further includes the construction of a multi-source data acquisition system, edge computing data preprocessing, and the construction of a data transmission network;

[0055] The data storage and management layer further includes the deepening of a hierarchical distributed storage architecture, the refinement of a data governance process, and a data security and rights management module;

[0056] The data analysis and prediction layer further includes the innovation of a multi-source data fusion algorithm, the optimization and expansion of a prediction model, and a model evaluation and update mechanism;

[0057] The intelligent control layer includes the optimization of a distributed cooperative control algorithm, the deepening of an intelligent pitch and variable speed control strategy, and control decision-making and straight-line optimization;

[0058] The equipment full life cycle management layer includes the refinement of a fault diagnosis and prediction model, the optimization of the formulation of a dynamic maintenance plan, the improvement of an equipment health status evaluation system, and the upgrade of a remote operation and maintenance and intelligent warning system.

[0059] The construction of a multi-source data acquisition system is for the data acquisition of wind turbines, meteorological data, and substation data. For the data acquisition of wind turbines, a magnetoelectric speed sensor is installed at the main shaft of the fan to obtain the fan speed. The output end of the generator is connected to an electronic power sensor to obtain active power, reactive power, and apparent power. Acceleration-type vibration sensors are deployed at key parts such as the fan blades and nacelle to obtain equipment vibration signals;

[0060] Meteorological data collection uses meteorological detection equipment to obtain comprehensive meteorological data (wind speed and direction, temperature and humidity, atmospheric pressure);

[0061] Substation data collection is for intelligent electricity meters to collect the voltage, current, and power factor of the substation;

[0062] Edge computing preprocessing is to deploy high-performance industrial-grade edge computing gateways at the data collection points. The edge computing gateways adopt a distributed deployment method and are reasonably configured according to the device distribution. Configure data denoising, normalization, and outlier detection preprocessing algorithms. Data denoising uses wavelet transform to remove noise interference in the data. The formula is as follows:

[0063] ,

[0064] Among them, is the scale parameter ( ), which controls the stretching of the wavelet function; is the translation parameter, which controls the translation of the wavelet function; is the mother wavelet function, is its conjugate function. During the denoising process, usually, threshold processing is performed on the high-frequency coefficients. Common threshold functions include the hard threshold function and the soft threshold function. The hard threshold function:

[0065] ,

[0066] Among them, is the wavelet coefficient, is the threshold;

[0067] The soft threshold function:

[0068] ,

[0069] Among them, is the sign function;

[0070] The normalization algorithm normalizes data of different types and magnitudes to a unified numerical interval, maps the data to the [0, 1] interval, and retains the relative relationship of the data. The formula is as follows:

[0071] ,

[0072] Among them, is the original data, and are the minimum and maximum values in the dataset respectively, is the normalized data;

[0073] Outlier detection is based on 3 of statistical analysis Criterion, assuming the data follows a normal distribution. In a normal distribution, the probability that the data falls within the interval is approximately 99.7%. Data outside this interval is considered abnormal data. The formula is as follows:

[0074] ,

[0075] where is the number of data points, is the th data point. If the data point satisfies , it is marked as abnormal data;

[0076] Edge computing devices perform real-time processing and screening on the collected raw data according to preset rules and algorithms, only transmit key data and processed feature data, reduce the data transmission volume, and improve the transmission efficiency;

[0077] Data transmission network construction. For relatively concentrated devices such as wind turbines, wired Ethernet is used for data transmission. A fiber optic Ethernet backbone network is built inside the wind farm to achieve high-speed data interaction between devices. Redundant link design is adopted. When the main link fails, the backup link can be automatically switched to ensure the stability of data transmission. For relatively dispersed devices such as meteorological monitoring devices, 4G / 5G communication technology or Lora wireless communication technology is used for data transmission.

[0078] The hierarchical distributed storage architecture is deepened to use the Ceph distributed storage system to build the underlying storage foundation on the cloud computing platform. It can provide multiple storage methods such as object storage, block storage, and file storage to meet the storage needs of different types of data. For real-time data, the Redis in-memory database is used as a cache, taking advantage of its high read and write speed and low latency characteristics to ensure the system's fast response to real-time data. For historical data, the distributed file system selects the Hadoop distributed file system and combines it with the Hbase, a distributed NoSQL database based on Hadoop, to achieve the efficient storage and fast retrieval of massive historical data;

[0079] The data governance process is refined as follows: for the operation data of wind turbines, the timestamp format is uniformly specified as the Iso8601 standard, and the device ID coding adopts specific coding rules to ensure data consistency and identifiability. In the data cleaning process, data cleaning tools are used in combination with custom cleaning rules to process duplicate values, error values, and missing values in the data. For missing values, according to the characteristics and correlations of the data, the mean filling method is used for completion. At the same time, a data quality monitoring index system is established to regularly evaluate and report indexes such as data accuracy, integrity, and consistency, and timely discover and solve data quality problems;

[0080] Data security and permission management adopt data encryption technology to encrypt and store the data stored in the database and file system.

[0081] As a further preferred solution of the present invention, the multi-source data fusion algorithm is innovated as follows: based on the fusion of multi-source data by an autoencoder and a generative adversarial network, transfer learning is performed on the data of different wind farms or different types of devices. The model parameters trained in one environment can be transferred to other environments, reducing training time and data requirements, quickly adapting to the data of different wind farms or devices, improving data fusion efficiency and accuracy, and then optimizing the prediction and control models to assist the efficient power generation of wind turbines and achieve energy conservation;

[0082] Optimization and expansion of the prediction model: an adaptive dynamic prediction model based on Lstm and attention mechanism is optimized in combination with a reinforcement learning algorithm. And the reinforcement learning algorithm uses a deep q network to construct a q network. The input is the state space, and the output is the q value of each action, representing the expected reward for taking a certain action in the current state. The model selects an action from the action space according to the current state at each time. After executing the action, the new state and the obtained reward are observed, and the state, action, reward, and new state are stored in the experience replay pool. A batch of data is randomly sampled from the experience replay pool regularly to train the q network, update the network parameters to maximize the cumulative reward. Subsequently, at each time, the current state is input into the Lstm and attention mechanism model for prediction to obtain the prediction result. The reward is calculated based on the prediction result and the actual value, and the state, action, reward, and new state are recorded. The parameters and prediction strategies of the model are adjusted regularly using the reinforcement learning algorithm, and the adjustment is performed every certain number of time steps or when the prediction error exceeds a certain threshold.

[0083] The optimization of the distributed cooperative control algorithm adopts a variant of the Paxos algorithm to ensure that the control instructions between each fan can be synchronized quickly and accurately. By optimizing the convergence speed and stability of the algorithm, the coordination time and error between the fans are reduced. At the same time, when adjusting the blade angle of the fans, each fan can quickly reach an agreement, avoiding power loss caused by inconsistent coordination. Meanwhile, combining the idea of game theory, a power distribution game model between the fans is established, enabling each fan to achieve the optimal power distribution of the entire wind farm while pursuing the maximization of its own power generation efficiency. Through an adaptive dynamic prediction model based on Lstm and attention mechanism, combined with a reinforcement learning algorithm, it can accurately predict wind speed, wind direction, power generation, and grid load demand, etc. According to these prediction results, the intelligent control layer adopts a distributed cooperative control algorithm and an intelligent pitch and variable speed control strategy to dynamically adjust the blade angle, speed, and power output of the fans. In different wind speed intervals, the fuzzy control rules are automatically adjusted to keep the fans always operating under the best conditions, maximizing the capture of wind energy, improving power generation efficiency, and reducing energy waste;

[0084] The intelligent pitch and variable speed control strategy is further optimized by adopting an adaptive fuzzy sliding mode control strategy. According to the real-time wind speed, wind direction, and the operating state of the fan, the fuzzy control rules and the parameters of the sliding mode controller are dynamically adjusted. In different wind speed intervals, the fuzzy control rules are automatically adjusted to enable the blade angle and speed of the fan to more accurately adapt to the wind speed change;

[0085] The optimization of control decision-making and execution adopts a real-time operating system. During the control execution process, a fault tolerance mechanism is introduced. When a fault occurs in the control execution unit of a certain fan, it can automatically switch to the standby unit to ensure the normal operation of the fan. At the same time, the control process is monitored and fed back in real time, and the control strategy is adjusted in a timely manner according to the actual operating conditions to achieve the optimal control of the fan.

[0086] The fault diagnosis and prediction model is refined by introducing a fusion model of the convolutional neural network Cnn and the recurrent neural network Rnn in deep learning to extract the local features of the data and process time series data;

[0087] The optimization of the dynamic maintenance plan formulation is based on the reliability-centered maintenance (Rcm) method. Combining the importance assessment and risk assessment of the equipment, a more scientific and reasonable dynamic maintenance plan is formulated. For key equipment and high-risk equipment, the maintenance frequency and depth are increased; for non-key equipment and low-risk equipment, the maintenance cost is appropriately reduced. Meanwhile, a maintenance resource optimization algorithm is introduced. According to the priority of the maintenance tasks, the equipment location, and the skill level of the maintenance personnel, the maintenance resources are reasonably allocated to improve the maintenance efficiency and reduce the maintenance cost. Combining expert system and knowledge graph technology, a comprehensive analysis of the equipment operation data is carried out to accurately diagnose and predict faults in advance, avoiding downtime losses and additional energy consumption caused by sudden equipment failures and reducing unnecessary energy waste;

[0088] Improve the equipment health status evaluation system and upgrade the remote operation and maintenance and intelligent early warning system. Establish a multi-level and multi-dimensional equipment health status evaluation index system. In addition to considering the operating parameters, fault history, maintenance records and environmental factors of the equipment, introduce the performance degradation index and remaining life prediction index of the equipment. Adopt the combination of the Analytic Hierarchy Process (Ahp) and the fuzzy comprehensive evaluation method to comprehensively evaluate the health status of the equipment. For the key parameters of key equipment, set multiple warning thresholds. When the parameters approach different thresholds, send warning messages at different levels to remind the operation and maintenance personnel to take measures in a timely manner. In addition, introduce blockchain technology into the final data to improve the security and credibility of the data. Through the distributed ledger and encryption technology of the blockchain, ensure the immutability and traceability of the equipment operation data and operation and maintenance records.

[0089] A wind power energy-saving control method based on big data analysis, comprising the following steps:

[0090] S1. Data collection: Install magnetoelectric speed sensors, electronic power sensors, and acceleration-type vibration sensors at key parts such as the main shaft, generator output terminal, blades, and nacelle of the wind turbine to collect wind turbine speed, power, and vibration signal data in real time. Deploy ultrasonic wind speed and direction sensors, digital temperature and humidity sensors, piezoresistive pressure sensors, and solar radiation sensors at the weather station to obtain wind speed, wind direction, temperature and humidity, and atmospheric pressure data. Collect electrical parameters such as voltage, current, and power factor of the substation through an intelligent electric meter with an Rs485 communication interface and supporting the Modbus communication protocol. At the same time, adopt the combination of video monitoring and intelligent monitoring to monitor the equipment status of the substation.

[0091] S2. Data transmission and preprocessing: Transmit the collected data to the edge computing gateway through the adapted hardware interface and the optimized Mqtt communication protocol. The edge computing gateway is deployed in a distributed manner and is reasonably configured according to the equipment distribution. The gateway uses wavelet transform denoising, normalization algorithm and 3 criteria to perform denoising, normalization and outlier detection processing on the data. After screening out the key data and characteristic data, transmit them to the data storage layer through wired Ethernet or 4G / 5G, and store the real-time data in the Redis in-memory database for quick response. The historical data is stored in the Hadoop Distributed File System (Hdfs) and the Hbase Distributed NoSQL Database, and the Ceph distributed storage system is used to build the underlying storage foundation.

[0092] S3. Data analysis and prediction: Using autoencoders, generative adversarial networks, combined with transfer learning and attention mechanisms, integrating multi-source data, extracting key features, constructing an adaptive dynamic prediction model based on LSTM and attention mechanisms, combined with reinforcement learning algorithms, automatically adjusting parameters and strategies according to prediction results and actual feedback, predicting wind speed, wind direction, power generation, and grid load demand, regularly evaluating the model based on indicators such as mean squared error (MSE), information gain, and Gini coefficient, and using model fusion technology to improve prediction accuracy.

[0093] S4. Intelligent control: The real-time control decision-making platform combines data analysis and prediction results to generate control instructions, through a distributed collaborative control algorithm optimized by a distributed consensus algorithm, as well as an intelligent pitch and variable speed control algorithm with an adaptive fuzzy sliding mode control strategy and neural network adaptive control, dynamically adjusting the angles, speeds, and power outputs of the fan blades.

[0094] S5. Equipment full life cycle management: Using a CNN and RNN fusion model combined with expert systems and knowledge graph technology, analyzing equipment operation data, diagnosing and predicting faults, based on the reliability-centered maintenance (RCM) method, combined with equipment importance and risk assessment, formulating a dynamic maintenance plan, using a maintenance resource optimization algorithm to reasonably allocate maintenance resources, and evaluating the equipment health status through the analytic hierarchy process (AHP) and fuzzy comprehensive evaluation method, combined with equipment performance degradation and remaining life prediction indicators.

[0095] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A wind power energy-saving control method and system based on big data analysis, characterized in that: It includes data collection and transmission layer, data storage and management layer, data analysis and prediction layer, intelligent control layer and equipment full life cycle management layer; The data collection and transmission layer also includes the construction of a multi-source data collection system, edge computing data preprocessing, and data transmission network construction; The data storage and management layer also includes the deepening of the layered distributed storage architecture, the refinement of the data governance process, and the data security and rights management module; The data analysis and prediction layer also includes innovation of multi-source data fusion algorithms, optimization and expansion of prediction models, and model evaluation and update mechanisms; The intelligent control layer includes distributed collaborative control algorithm optimization, intelligent pitch and speed control strategy deepening, and control decision and straight line optimization; The equipment life cycle management layer includes the refinement of fault diagnosis and prediction models, optimization of dynamic maintenance plan formulation, improvement of equipment health status assessment system, and upgrade of remote operation and maintenance and intelligent early warning system.

2. The wind power energy-saving control system based on big data analysis according to claim 1 is characterized in that: The multi-source data collection system is constructed for wind turbine data collection, meteorological data collection and substation data collection. The wind turbine data collection installs a magnetoelectric speed sensor on the main shaft of the wind turbine to obtain the wind turbine speed, connects an electronic power sensor to the generator output end to obtain active power, reactive power and apparent power, and deploys acceleration vibration sensors at key locations such as wind turbine blades and nacelles to obtain equipment vibration signals; The meteorological data collection uses meteorological detection equipment to obtain comprehensive meteorological data (wind speed and direction, temperature and humidity, atmospheric pressure); The substation data collection is that the smart meter collects the voltage, current and power factor of the substation; The edge computing preprocessing is to deploy a high-performance industrial-grade edge computing gateway at the data collection point. The edge computing gateway adopts a distributed deployment mode and is reasonably configured according to the distribution of equipment. Data denoising, normalization and outlier detection preprocessing algorithms are configured. Data denoising uses wavelet transformation to remove noise interference in the data. The formula is as follows: ， in, is the scale parameter ( ), controls the expansion and contraction of the wavelet function; is the translation parameter, which controls the translation of the wavelet function; is the wavelet mother function, is its conjugate function. In the denoising process, high-frequency coefficients are usually subjected to threshold processing. Commonly used threshold functions include hard threshold function and soft threshold function. Hard threshold function: ， in, are the wavelet coefficients, is the threshold value; Soft threshold function: ， in, is a sign function; The normalization algorithm normalizes data of different types and magnitudes to a unified numerical interval, maps the data to the [0,1] interval, and retains the relative relationship of the data. The formula is as follows: ， in, is the original data, and are the minimum and maximum values ​​in the data set, respectively. is the normalized data; Outlier detection based on statistical analysis The data are assumed to follow a normal distribution. In a normal distribution, the data fall within The probability of the interval is about 99.7%. Data outside this interval is considered abnormal data. The formula is as follows: , in, is the number of data points, It is data points, if the data points satisfy , It is marked as abnormal data; Edge computing devices process and filter the collected raw data in real time according to preset rules and algorithms, and only transmit key data and processed feature data, thus reducing the amount of data transmission and improving transmission efficiency. The data transmission network is constructed by using wired Ethernet for data transmission for relatively concentrated equipment such as wind turbines, and a fiber-optic Ethernet backbone network is constructed inside the wind farm to achieve high-speed data interaction between devices. A redundant link design is adopted, and when the main link fails, the backup link can be automatically switched to ensure the stability of data transmission. For relatively dispersed equipment such as meteorological monitoring equipment, 4G / 5G communication technology or Lora wireless communication technology is used for data transmission.

3. A wind power energy-saving control system based on big data analysis according to claim 2, characterized in that: The layered distributed storage architecture is deepened to build the underlying storage foundation on the cloud computing platform using the Ceph distributed storage system, which can provide multiple storage methods such as object storage, block storage and file storage to meet the storage requirements of different types of data. For real-time data, the Redis memory database is used as a high-speed cache, and its high read and write speed and low latency characteristics are used to ensure the system's rapid response to real-time data. For historical data, the distributed file system uses the Hadoop distributed file system, combined with Hbase, a distributed Nosql database based on Hadoop, to achieve efficient storage and rapid retrieval of massive historical data; The data governance process is refined as follows: for the operation data of wind turbines, the timestamp format is uniformly stipulated as ISO8601 standard, and the equipment ID coding adopts specific coding rules to ensure the consistency and identifiability of the data; in the data cleaning stage, the data cleaning tool is used in combination with the custom cleaning rules to process the duplicate values, error values ​​and missing values ​​in the data. For the missing values, the mean filling method is used to fill them according to the characteristics and relevance of the data. At the same time, a data quality monitoring indicator system is established to regularly evaluate and report on the accuracy, completeness, consistency and other indicators of the data, so as to timely discover and solve data quality problems; The data security and authority management adopts data encryption technology to encrypt and store the data stored in the database and file system.

4. The wind power energy-saving control system based on big data analysis according to claim 3 is characterized in that: The multi-source data fusion algorithm is innovative in that it performs transfer learning on data from different wind farms or different types of equipment based on the fusion of multi-source data by autoencoders and generative adversarial networks. It can migrate model parameters trained in one environment to other environments, thus reducing training time and data requirements. The prediction model is optimized and expanded, and an adaptive dynamic prediction model based on LSTM and attention mechanism is optimized in combination with a reinforcement learning algorithm. In addition, the reinforcement learning algorithm uses a deep q network to construct a q network, with the input being the state space and the output being the q value of each action, which represents the expected reward for taking a certain action in the current state. The model selects an action from the action space according to the current state at each time, and after executing the action, observes the new state and the reward obtained, and stores the state, action, reward, and new state in an experience replay pool. A batch of data is randomly sampled from the experience replay pool regularly to train the q network, and the network parameters are updated to maximize the cumulative reward. Subsequently, at each time, the current state is input into the LSTM and attention mechanism model for prediction to obtain a prediction result, and the reward is calculated based on the prediction result and the actual value, and the state, action, reward, and new state are recorded. The parameters and prediction strategy of the model are adjusted regularly using the reinforcement learning algorithm, and adjustments are made every a certain number of times or when the prediction error exceeds a certain threshold.

5. The wind power energy-saving control system based on big data analysis according to claim 4 is characterized in that: The distributed collaborative control algorithm optimization adopts a variant of the Paxos algorithm to ensure that the control instructions between the wind turbines can be synchronized quickly and accurately. By optimizing the convergence speed and stability of the algorithm, the coordination time and error between the wind turbines are reduced. At the same time, when adjusting the angle of the wind turbine blades, the wind turbines can quickly reach a consensus to avoid power loss caused by inconsistent coordination. At the same time, combined with the idea of ​​game theory, a power allocation game model between wind turbines is established, so that each wind turbine can achieve the optimal power allocation of the entire wind farm while pursuing the maximization of its own power generation efficiency. The intelligent variable pitch and variable speed control strategy further adopts an adaptive fuzzy sliding mode control strategy, dynamically adjusts the fuzzy control rules and the parameters of the sliding mode controller according to the real-time wind speed, wind direction and wind turbine operating status, and automatically adjusts the fuzzy control rules in different wind speed ranges, so that the blade angle and speed of the wind turbine can more accurately adapt to wind speed changes; The control decision and execution optimization adopts a real-time operating system. During the control execution process, a fault tolerance mechanism is introduced. When a control execution unit of a fan fails, it can automatically switch to a backup unit to ensure the normal operation of the fan. At the same time, the control process is monitored and fed back in real time, and the control strategy is adjusted in time according to the actual operating conditions to achieve optimal control of the fan.

6. The wind power energy-saving control system based on big data analysis according to claim 5 is characterized in that: The fault diagnosis and prediction model is refined into a fusion model that introduces the convolutional neural network Cnn and the recurrent neural network Rnn in deep learning, so as to extract local features of the data and process time series data; The dynamic maintenance plan formulation optimization is based on the reliability center maintenance Rcm method, combined with the importance assessment and risk assessment of the equipment, to formulate a more scientific and reasonable dynamic maintenance plan. For key equipment and high-risk equipment, the maintenance frequency and depth are increased; for non-key equipment and low-risk equipment, the maintenance cost is appropriately reduced. At the same time, the maintenance resource optimization algorithm is introduced to reasonably allocate maintenance resources according to the priority of the maintenance task, the location of the equipment and the skill level of the maintenance personnel, so as to improve the maintenance efficiency and reduce the maintenance cost. The equipment health status assessment system is improved, and the remote operation and maintenance and intelligent early warning systems are upgraded to establish a multi-level and multi-dimensional equipment health status assessment indicator system. In addition to considering the equipment's operating parameters, fault history, maintenance records and environmental factors, the equipment's performance degradation indicators and remaining life prediction indicators are introduced. A combination of the analytic hierarchy process (AHP) and the fuzzy comprehensive evaluation method is used to comprehensively assess the equipment's health status. For key parameters of key equipment, multiple early warning thresholds are set. When the parameters are close to different thresholds, different levels of early warning information are issued to remind the operation and maintenance personnel to take timely measures. In addition, the final data introduces blockchain technology to improve data security and credibility. Through the distributed ledger and encryption technology of blockchain, the equipment operation data and operation and maintenance records are ensured to be tamper-proof and traceable.

7. A wind power energy-saving control method based on big data analysis according to claims 1-6, characterized in that: The following steps are involved: s1. Data collection: magnetoelectric speed sensors, electronic power sensors, and acceleration-type vibration sensors are installed at key locations such as the main shaft, generator output end, blades, and nacelle of wind turbines to collect wind turbine speed, power, and vibration signal data in real time. Ultrasonic anemometers, digital temperature and humidity sensors, piezoresistive air pressure sensors, and solar radiation sensors are deployed at meteorological stations to obtain data on wind speed, wind direction, temperature and humidity, and atmospheric pressure. The voltage, current, power factor, and other electrical parameters of the substation are collected through smart meters that have an RS485 communication interface and support the Modbus communication protocol. At the same time, a combination of video monitoring and intelligent monitoring is used to monitor the status of substation equipment. s2. Data transmission and preprocessing: The collected data is transmitted to the edge computing gateway through the adapted hardware interface and the optimized Mqtt communication protocol. The edge computing gateway adopts distributed deployment and is reasonably configured according to the distribution of devices. The gateway uses wavelet transform denoising, normalization algorithm and 3 According to the principle, the data is denoised, normalized and outliers are detected. After filtering out key data and feature data, it is transmitted to the data storage layer via wired Ethernet or 4g / 5g. Real-time data is stored in the Redis memory database for quick response. Historical data is stored in the Hadoop distributed file system HDFS and Hbase distributed Nosql database. The Ceph distributed storage system is used to build the underlying storage foundation. s3. Data analysis and prediction: using autoencoders, generative adversarial networks combined with transfer learning and attention mechanisms, integrating multi-source data, extracting key features, building an adaptive dynamic prediction model based on LSTM and attention mechanisms, combining reinforcement learning algorithms, automatically adjusting parameters and strategies according to prediction results and actual feedback, predicting wind speed, wind direction, power generation and grid load demand, regularly evaluating models based on mean square error Mse information gain, Gini coefficient and other indicators, and using model fusion technology to improve prediction accuracy; s4. Intelligent control: The real-time control decision platform combines data analysis and prediction results to generate control instructions; the distributed collaborative control algorithm optimized by the distributed consistency algorithm, as well as the intelligent variable pitch and variable speed control algorithm of the adaptive fuzzy sliding mode control strategy and neural network adaptive control, dynamically adjusts the wind turbine blade angle, speed and power output; s5. Equipment life cycle management, using Cnn and Rnn fusion models combined with expert systems and knowledge graph technology to analyze equipment operation data, diagnose and predict faults, based on the reliability center maintenance Rcm method, combined with equipment importance and risk assessment, formulate dynamic maintenance plans, use maintenance resource optimization algorithms to reasonably allocate maintenance resources, and use the hierarchical analysis method AHP and fuzzy comprehensive evaluation method, combined with equipment performance degradation and remaining life prediction indicators, to evaluate the equipment health status.

Images