New energy power generation device control method, device, apparatus and storage medium

Abstract

This application relates to a control method, device, equipment, and storage medium for new energy power generation equipment. The method includes: acquiring multi-dimensional operational data of the new energy power generation equipment; predicting the operational state of the new energy power generation equipment based on a target analysis model adapted to the data type of the multi-dimensional operational data, obtaining state prediction data; using the state prediction data as input data, performing global optimization scheduling processing with the optimization objectives of maximizing power generation and minimizing the failure rate of the new energy power generation equipment, obtaining optimized scheduling parameters; encapsulating the optimized scheduling parameters into scheduling control instructions; and using the scheduling control instructions to control the new energy power generation equipment. This method can accurately identify faults in new energy power generation equipment, effectively maintain the equipment, and shorten its lifespan degradation rate.

Description

Technical Field

[0001] This application relates to the field of new energy technology, and in particular to a control method, device, equipment and storage medium for new energy power generation equipment. Background Technology

[0002] As the global energy structure accelerates its transition to a low-carbon model, the installed capacity of new energy power generation equipment has experienced explosive growth. By the end of 2024, my country's installed wind and solar power capacity had exceeded 1.2 billion kilowatts, but the contradiction between traditional operation and maintenance models and large-scale development is becoming increasingly prominent.

[0003] Currently, the operation and maintenance of new energy power generation equipment largely relies on electrical parameters such as voltage and current collected by SCADA (Supervisory Control And Data Acquisition) systems for status assessment. However, under complex operating conditions, the analysis of electrical parameters is somewhat one-sided, making it difficult to accurately identify and diagnose various types of equipment faults, resulting in high equipment failure rates and shortened lifespans. Summary of the Invention

[0004] Therefore, it is necessary to provide a control method, device, equipment, and storage medium for new energy power generation equipment that can comprehensively identify and diagnose faults in new energy power generation equipment, effectively maintain the equipment, and shorten its lifespan degradation rate, in order to address the aforementioned technical problems.

[0005] In a first aspect, this application provides a control method for new energy power generation equipment, including:

[0006] Acquire multi-dimensional operational data of new energy power generation equipment;

[0007] Based on a target analysis model that is compatible with the data type of the multidimensional operational data, the operating status of the new energy power generation equipment is predicted to obtain status prediction data.

[0008] Using the state prediction data as input data, and with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, global optimization scheduling is performed to obtain optimized scheduling parameters.

[0009] The optimized scheduling parameters are encapsulated into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment.

[0010] In one embodiment, the prediction of the operating status of the new energy power generation equipment based on the target analysis model adapted to the data type of the multidimensional operational data, to obtain status prediction data, includes:

[0011] When the multidimensional operational data includes time-series data, the target analysis model that is adapted to the multidimensional operational data is determined to include a time-series neural network model;

[0012] The time series data is input into the time series neural network model to predict the fault components in the time series data, thereby obtaining state prediction data.

[0013] In one embodiment, the method further includes:

[0014] Extract the vibration prediction data and temperature prediction data from the state prediction data;

[0015] If the abnormal correlation between the vibration prediction data and the temperature prediction data meets the fault warning conditions, a fault warning message is output.

[0016] In one embodiment, the prediction of the operating status of the new energy power generation equipment based on the target analysis model adapted to the data type of the multidimensional operational data, to obtain status prediction data, includes:

[0017] When the multidimensional operational data includes visual data, the target analysis model that is adapted to the multidimensional operational data includes a defect detection model.

[0018] The visual data is input into the defect detection model to perform potential defect detection processing, identify potential defect areas of the new energy power generation equipment, and obtain state prediction data.

[0019] In one embodiment, the method further includes:

[0020] Extract environmental status data from the multidimensional operational data;

[0021] Determine a visual detection threshold that matches the environmental state data;

[0022] Based on the visual detection threshold, the detection boundary of the defect detection model is calibrated.

[0023] In one embodiment, the process of using the state prediction data as input data and performing global optimization scheduling with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment to obtain optimized scheduling parameters includes:

[0024] Determine the scheduling action space of the new energy power generation equipment;

[0025] The state prediction data is input into the reinforcement learning policy network under the scheduling action space constraint to obtain the action probability distribution;

[0026] With the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, the optimized scheduling parameters are obtained by sampling from the action probability distribution.

[0027] In one embodiment, the method further includes:

[0028] The multidimensional operating data is mapped to the digital twin model of the new energy power generation equipment to perform fault simulation and obtain fault simulation information;

[0029] The fault region represented by the fault simulation information is determined, and the fault region is highlighted in the digital twin model.

[0030] Secondly, this application also provides a control device for energy generation equipment, comprising:

[0031] The multi-dimensional data acquisition module is used to acquire multi-dimensional operating data of new energy power generation equipment;

[0032] The operation status prediction module is used to predict the operation status of the new energy power generation equipment based on a target analysis model that is compatible with the data type of the multidimensional operation data, and to obtain status prediction data.

[0033] The global optimization scheduling module is used to take the state prediction data as input data, and with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, perform global optimization scheduling processing to obtain optimized scheduling parameters.

[0034] The equipment scheduling and control module is used to encapsulate the optimized scheduling parameters into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment.

[0035] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0036] Acquire multi-dimensional operational data of new energy power generation equipment;

[0037] Based on a target analysis model that is compatible with the data type of the multidimensional operational data, the operating status of the new energy power generation equipment is predicted to obtain status prediction data.

[0038] Using the state prediction data as input data, and with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, global optimization scheduling is performed to obtain optimized scheduling parameters.

[0039] The optimized scheduling parameters are encapsulated into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment.

[0040] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0041] Acquire multi-dimensional operational data of new energy power generation equipment;

[0042] Based on a target analysis model that is compatible with the data type of the multidimensional operational data, the operating status of the new energy power generation equipment is predicted to obtain status prediction data.

[0043] Using the state prediction data as input data, and with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, global optimization scheduling is performed to obtain optimized scheduling parameters.

[0044] The optimized scheduling parameters are encapsulated into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment.

[0045] The aforementioned control method, device, equipment, and storage medium for new energy power generation equipment acquire multi-dimensional operational data of the new energy power generation equipment; based on a target analysis model adapted to the data type of the multi-dimensional operational data, the operating state of the new energy power generation equipment is predicted, resulting in state prediction data; using the state prediction data as input data, with the optimization objectives of maximizing power generation and minimizing failure rate of the new energy power generation equipment, global optimization scheduling is performed to obtain optimized scheduling parameters; the optimized scheduling parameters are encapsulated into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment. Multi-dimensional operational data can comprehensively reflect the operating state of the new energy power generation equipment. Through the adapted target analysis model, accurate prediction can be made for operational data of different modes, resulting in state prediction data. From the perspective of maximizing power generation and minimizing failure rate, global optimization scheduling can be performed, which can maintain stable equipment operation while minimizing equipment wear and tear. By controlling the operation of the new energy power generation equipment through scheduling control instructions, effective maintenance of the new energy power generation equipment can be achieved, slowing down the rate of equipment lifespan degradation. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is an application environment diagram of a new energy power generation equipment control method in one embodiment;

[0048] Figure 2 This is a flowchart illustrating a control method for new energy power generation equipment in one embodiment;

[0049] Figure 3 This is a flowchart illustrating the process of predicting the operating status of new energy power generation equipment and obtaining state prediction data based on a target analysis model adapted to the data type of multidimensional operational data in one embodiment.

[0050] Figure 4 This is a flowchart illustrating the process of predicting the operating status of new energy power generation equipment and obtaining state prediction data based on a target analysis model adapted to the data type of multidimensional operational data in another embodiment.

[0051] Figure 5 This is a flowchart illustrating how state prediction data is used as input data in one embodiment, and global optimization scheduling is performed with the optimization objectives of maximizing the power generation and minimizing the failure rate of new energy power generation equipment to obtain optimized scheduling parameters.

[0052] Figure 6 This is a flowchart illustrating the control method for new energy power generation equipment in another embodiment;

[0053] Figure 7 This is a flowchart illustrating the control method for new energy power generation equipment in yet another embodiment;

[0054] Figure 8 This is a structural block diagram of a new energy power generation equipment control device in one embodiment;

[0055] Figure 9 This is an internal structural diagram of a computer device in one embodiment;

[0056] Figure 10 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0058] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0059] The contradiction between traditional operation and maintenance models and the large-scale development of new energy power generation is becoming increasingly prominent. Monitoring is limited to a single dimension, relying on electrical parameters such as voltage and current collected by SCADA for condition assessment. This lack of collaborative analysis of multi-dimensional physical characteristics such as vibration, temperature field, and visual images leads to early-stage latent faults. Currently, the rate of missed detection for micro-cracks in wind turbine bearings and hidden cracks in photovoltaic modules remains high. Taking wind turbine gearboxes as an example, traditional systems only judge faults based on oil temperature, often ignoring the vibration characteristics changes in the early stages of gear wear, resulting in delayed fault detection. Fixed-threshold diagnostic algorithms are ill-suited to the nonlinear operating characteristics of new energy equipment. During varying operating conditions, the vibration, temperature, and other characteristic parameters of the equipment exhibit strong time-varying and non-stationary characteristics. Traditional algorithms, lacking a dynamic threshold update mechanism, are prone to feature space shifts and blurred diagnostic boundaries. In extreme environments, such as strong winds, abrupt changes in the physical characteristics of the equipment exceed the model training distribution range, causing a significant decrease in anomaly detection confidence and making it difficult to meet the needs of accurate diagnosis under complex operating conditions.

[0060] Data silos created by decentralized monitoring systems severely restrict collaborative management at the power plant cluster level. Different subsystems (wind turbine monitoring, photovoltaic power generation, and energy storage management) employ different communication protocols and sampling mechanisms, resulting in inconsistent spatiotemporal data benchmarks and heterogeneous feature dimensions, making it difficult to construct cross-device health status correlation models. Furthermore, the lack of deep fusion analysis of multi-source data prevents the formation of a power plant-level equipment health matrix and energy efficiency optimization strategies. Independent decision-making by each subsystem easily leads to power plant cluster oscillations and energy conversion efficiency losses. Manual inspection may not be suitable for the ultra-large-scale geographical distribution characteristics of new energy power plants. Large power plants have a high degree of spatial dispersion of equipment, and traditional inspection methods are limited by line-of-sight range and mobility efficiency, making it difficult to achieve high-frequency, full-coverage monitoring of equipment status. Moreover, in harsh environments, the efficiency of inspection operations and the accuracy of testing instruments decline simultaneously, leading to delayed detection of latent faults and failing to meet the real-time requirements of predictive maintenance.

[0061] With the advancements in Internet of Things (IoT) and sensor accuracy, breakthroughs in edge computing power, and the development of lightweight deep learning models, deploying systems with intelligent monitoring capabilities has become possible. However, current industry pain points are concentrated in three major areas: how to achieve spatiotemporal alignment and fusion of multi-source heterogeneous data, how to construct a cross-modal diagnostic model that adapts to operating conditions, and how to achieve accurate mapping between physical devices and virtual models through digital twins. These three technological bottlenecks restrict the improvement of operation and maintenance efficiency and monitoring accuracy of new energy power generation equipment.

[0062] The new energy power generation equipment control method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Terminal 102 acquires multi-dimensional operational data of the new energy power generation equipment; based on a target analysis model adapted to the data type of the multi-dimensional operational data, it predicts the operating status of the new energy power generation equipment, obtaining status prediction data; using the status prediction data as input data, and with the optimization objectives of maximizing power generation and minimizing failure rate of the new energy power generation equipment, it performs global optimization scheduling processing to obtain optimized scheduling parameters; the optimized scheduling parameters are encapsulated into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. Server 104 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0063] In one exemplary embodiment, such as Figure 2 As shown, a control method for new energy power generation equipment is provided, which can be applied to... Figure 1 The following steps are used as an example of the terminal in the example, including steps 202 to 208.

[0064] Step 202: Obtain multi-dimensional operation data of new energy power generation equipment.

[0065] New energy power generation equipment is equipment that can convert renewable energy sources such as solar and wind energy into electrical energy. The operational data of new energy power generation equipment reflects the status of various indicators during its operation. Multidimensional operational data indicates that the data type includes multiple dimensions, such as text, images, and videos. New energy power generation equipment can include photovoltaic arrays, wind turbines, etc. From the perspective of data source, multidimensional operational data can also include sensor data, meteorological and environmental data, power grid dispatch data, historical operation and maintenance records, etc.

[0066] For example, multi-dimensional operational data is acquired from a multi-source data fusion acquisition system for new energy power generation equipment. This system may include components that collect operational data of different types, such as electrical data acquisition components, visual data acquisition components, vibration sensing components, and temperature field monitoring components. Message middleware technology can be used to align and fuse operational data from different sources and with different sampling frequencies to obtain multi-dimensional operational data in a unified format, which can then be used for predicting the operational status of new energy power generation equipment.

[0067] In one feasible implementation, triaxial accelerometers are deployed as vibration sensing components in key areas such as the high-speed shaft of the wind turbine gearbox and the generator bearing housing. A charge amplifier amplifies and filters the weak vibration signals. By measuring the vibration signals of the equipment, the operational stability of the equipment and the wear condition of its mechanical components can be understood. When a fault occurs in the equipment, the vibration signal will change in both frequency and amplitude. After acquiring the vibration signal, wavelet packet decomposition processing can be performed, referring to Formula 1 below.

[0068] Formula 1: .

[0069] Where J is the number of decomposition layers, ψ is the wavelet basis function, and c is the decomposition coefficient.

[0070] In one feasible implementation, multiple temperature sensors are embedded in the backsheet of the photovoltaic module for temperature field monitoring. A fusion of embedded temperature sensors and an infrared thermal imager is used to perform three-dimensional monitoring of hotspot areas such as the photovoltaic inverter module and battery storage system cells. This is used to monitor temperature changes in the equipment, as temperature is a crucial indicator of equipment operating status; high temperatures may indicate overload, short circuit, or other malfunctions. The temperature data is processed using a non-uniformity correction algorithm to eliminate temperature drift caused by environmental interference. The correction formula is shown in Formula 2 below.

[0071] Formula 2: .

[0072] Where k is the correction coefficient, ∇T_ env For the ambient temperature gradient, T_ raw (x,y) represents the original temperature data, T_ corr (x,y) represents the corrected temperature data.

[0073] In one feasible implementation, a high-definition camera and LiDAR are used as data acquisition components, and the three-dimensional shape reconstruction of equipment such as wind turbine blades and photovoltaic supports is achieved through structured light scanning. After noise reduction processing, the LiDAR point cloud data is fitted with planar features using the Random Sample Consensus (RANSAC) algorithm, referring to the following formula 3.

[0074] Formula 3: .

[0075] Where ε is the distance threshold and p_i is the point cloud data.

[0076] Step 204: Based on the target analysis model adapted to the data type of multidimensional operational data, predict the operating status of new energy power generation equipment to obtain status prediction data.

[0077] The target analysis model is used to analyze multidimensional operational data and predict the operating status of new energy power generation equipment. Status prediction data represents the future operating status of new energy equipment. By employing an appropriate target analysis model for different types of operational data, predictions can be made that adapt to the characteristics of different data types, allowing for a more comprehensive prediction of the operating status of new energy power generation equipment from various perspectives.

[0078] For example, based on the prediction task type adapted to the multidimensional operational data, a trained target analysis model is constructed. The multidimensional operational data is then input into the corresponding target analysis model to predict the operating status of the new energy power generation equipment, obtaining state prediction data. Prediction task types may include regression, classification, and target detection. Multidimensional operational data within the current and historical time windows can be input into the trained target analysis model to output state prediction data. For instance, for a regression detection task, real-time acquired electrical data is input into the target classification model for regression prediction, outputting data such as power generation for the next time step.

[0079] Step 206: Using the state prediction data as input data, and with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, global optimization scheduling is performed to obtain the optimized scheduling parameters.

[0080] Global optimization scheduling refers to the operation of adjusting various operating parameters of new energy power generation equipment with the goal of achieving optimal overall performance. The objectives of global optimization include maximizing power generation and minimizing the failure rate. Optimization scheduling parameters are parameters used to indicate the optimization objectives of the operation and scheduling of new energy power generation equipment.

[0081] For example, a multi-objective function is set based on maximizing the power generation and minimizing the failure rate of new energy power generation equipment. State prediction data is used as input data, and multi-objective particle swarm optimization is performed to obtain optimized scheduling parameters. In the multi-objective function, the optimization objectives include the total power generation in the next scheduling cycle and the equipment failure risk index, where the equipment failure risk index is used to quantify the overall failure rate. The position vector of each particle can represent a set of operating strategy parameters. A preset number of particles are randomly initialized within the safe operating range of the equipment as the initial population. For each particle in the initial population, the value of the objective function can be calculated. Each particle records its historical best position, and particle velocity and position are updated. Iterative optimization is then performed to obtain the optimized scheduling parameters.

[0082] Step 208: Encapsulate the optimized scheduling parameters into scheduling control instructions.

[0083] Dispatch control commands are used to control new energy power generation equipment, enabling it to reach an operating state corresponding to the optimized dispatch objectives. For example, the dispatch control commands also include the equipment identification and verification information of the new energy power generation equipment. The verification information is used to verify the integrity of the dispatch control commands. The dispatch control commands can be encrypted and transmitted to the new energy power generation equipment using national cryptographic encryption protocols and TLS (Transport Layer Security) to achieve high-precision control of parameters such as wind turbine pitch angle and photovoltaic MPPT (Maximum Power Point Tracking) voltage, ensuring the real-time performance and accuracy of the command response.

[0084] In the aforementioned control method for new energy power generation equipment, multi-dimensional operating data of the new energy power generation equipment is acquired; based on a target analysis model adapted to the data type of the multi-dimensional operating data, the operating state of the new energy power generation equipment is predicted to obtain state prediction data; using the state prediction data as input data, with the optimization objectives of maximizing power generation and minimizing failure rate of the new energy power generation equipment, global optimization scheduling is performed to obtain optimized scheduling parameters; the optimized scheduling parameters are encapsulated into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment. Multi-dimensional operating data can comprehensively reflect the operating state of the new energy power generation equipment. Through the adapted target analysis model, accurate prediction can be made for operating data of different modes to obtain state prediction data. From the perspective of maximizing power generation and minimizing failure rate, global optimization scheduling can be performed to maintain stable equipment operation while minimizing equipment wear and tear. By controlling the operation of the new energy power generation equipment through scheduling control instructions, the new energy power generation equipment can be effectively maintained, and the equipment lifespan degradation rate can be slowed down.

[0085] In one exemplary embodiment, such as Figure 3 As shown, based on the target analysis model adapted to the data type of multidimensional operating data, the operating status of new energy power generation equipment is predicted to obtain status prediction data, including steps 302 to 304.

[0086] Step 302: When the multidimensional operational data includes time-series data, determine the target analysis model that is compatible with the multidimensional operational data, including the time-series neural network model.

[0087] Time-series data refers to data recorded in chronological order. It can be recorded by timestamps indicating the point in time the data was generated, and also includes the specific indicator values ​​corresponding to those times. Temporal neural network models are predictive models built specifically for the characteristics of time-series data. The temporal neural network model in this embodiment includes an improved LSTM-Transformer hybrid network, introduces a relative position encoding mechanism, and enhances the interaction capabilities of different modal features through a bidirectional attention mechanism, thereby achieving time-series prediction of device health status.

[0088] Step 304: Input the time series data into the time series neural network model to perform time series prediction, identify the fault components in the time series data, and obtain state prediction data.

[0089] The fault component in time series data refers to the data component that is predicted to have a certain probability of failure. During time series prediction, the time series data can be converted into a time series feature vector, and the fault component can be mapped to a component in the time series feature vector. First, the time series data can be processed using LSTM. Based on the current input time series data, the hidden state and cell state of the previous time step, the hidden state and cell state at the current time step are updated through input gates, forget gates, and output gates. Then, the LSTM output is used as the input to a Transformer. Position encoding is introduced during Transformer processing to preserve temporal information, as shown in Equation 4 below.

[0090] Formula 4: .

[0091] Where Transformer_inputs represents the Transformer inputs. The hidden state sequence of the LSTM. Let L be the positional encoding function and L be the sequence length. The single-head attention mechanism in the Transformer can be referenced in Equation 5 below.

[0092] Formula 5: .

[0093] Where Q (query), K (key), and V (value) are the input feature matrices, d k The dimension of the key vector. This is a scaling factor to prevent gradient vanishing.

[0094] The handling of multi-head attention mechanism in Transformer can be referenced in the following formula 6.

[0095] Formula 6: .

[0096] in, h represents the number of attention heads, the LSTM output dimension is 256, PositionalEncoding is the positional encoding function used to preserve temporal information; d_k=64 is the attention dimension, which enhances the ability to capture long sequence dependencies through a bidirectional attention mechanism.

[0097] The improved LSTM-Transformer hybrid network can automatically identify key fault indicators in time-series features such as vibration and temperature through a bidirectional attention mechanism. It assigns higher weights to high-frequency vibration components (such as gear meshing characteristic frequencies of 200-500Hz and bearing outer ring fault characteristic frequencies of 100-300Hz) and temperature gradient changes (such as temperature rise rate ≥2℃ / min per unit time). By learning the feature mapping relationship in historical fault cases, it constructs a target analysis model to accurately identify faults in new energy power generation equipment.

[0098] In one embodiment, after inputting time-series data into a time-series neural network model to predict fault components in the time-series data and obtain state prediction data, the method further includes: extracting vibration prediction data and temperature prediction data from the state prediction data; and outputting fault warning information when the abnormal correlation between the vibration prediction data and the temperature prediction data meets the fault warning conditions.

[0099] Vibration prediction data refers to data obtained by predicting the vibration state of new energy power generation equipment. Temperature prediction data refers to data obtained by predicting the temperature state of new energy power generation equipment. This embodiment takes wind power generation equipment as an example. In key mechanical components such as the high-speed shaft of the wind turbine gearbox, IEPE (Integrated Electronics Piezo-Electric) triaxial accelerometers can be deployed. A high-frequency sampling mechanism is used to capture the vibration characteristics of dynamic processes such as gear meshing and bearing operation. PT100 temperature sensors can be embedded in the stator windings. Combined with an infrared thermal imager at the top of the tower, a three-dimensional monitoring network of the temperature field of electrical and mechanical components is formed. Furthermore, lidar can be used to periodically scan the blades, tower, and other structures in three dimensions to obtain subtle changes in the equipment's geometry.

[0100] Anomaly correlation refers to the degree of correlation between anomalies in vibration and temperature. It is understandable that vibration and temperature anomalies are correlated, and judging the anomaly correlation degree allows for more accurate fault warnings. For example, fault warning conditions can include vibration-temperature anomaly coupling, equipment health status probability values, and anomaly correlation scores. Vibration-temperature anomaly coupling can include: vibration time-domain peak factor ≥6, kurtosis ≥5, corresponding part temperature T_corr ≥ the upper limit of the equipment's rated operating temperature (e.g., gearbox ≥90℃), and temperature gradient ∇T_env ≥1.5℃ / cm; or after a sustained temperature rise of ≥8℃ for 1 hour, the vibration fault characteristic frequency band energy exceeds 3 times the normal distribution standard deviation; or related components exhibit vibration and temperature anomalies simultaneously for more than 5 minutes.

[0101] The fault warning conditions can be set as follows: a fault warning is triggered and fault warning information is output when the above-mentioned arbitrary coupling state lasts for ≥8 minutes, the probability value of equipment health status is ≤30%, and the characteristic anomaly correlation score is ≥0.7. The output form of the fault warning information can include sound signals and light signals, etc.

[0102] In one embodiment, a maintenance window recommendation can be given based on the Remaining Useful Life (RUL) prediction algorithm. Based on the online evolution mechanism of Gaussian mixture model, the diagnostic threshold can be dynamically adjusted according to real-time operating conditions. When new data features deviate from the historical distribution, the system automatically starts the model fine-tuning process and updates parameters through incremental learning algorithm to ensure the adaptability of the diagnostic algorithm under changing load and environmental conditions, effectively reducing the false alarm rate when switching operating conditions.

[0103] For example, an improved LSTM-Transformer network receives time-series data on vibration and temperature, and calculates feature weights using the bidirectional attention mechanism shown in Equation 7.

[0104] Formula 7: .

[0105] Where f_i is the i-th feature and w_i is the weight corresponding to the i-th feature.

[0106] When the outlier score S(y) of new data exceeds the critical value m, the GMM (Gaussian Mixture Model) parameters can be updated. The formula for calculating the outlier score can be found in Formula 8 below.

[0107] Formula 8: .

[0108] Where h(y) is the path length of the isolated tree, and k is the scaling factor. A larger S(y) value increases the likelihood of an outlier. An outlier score threshold m is set; if S(y) ≥ m, the data point y is considered an outlier, and the outlier detection probability value P is... anom (y) = 1; if S(y) < m, it is determined to be a normal point, P anom (y) = 0.

[0109] In this embodiment, a time-series neural network model is used to predict time-series data, and combined with fault warning conditions and diagnostic threshold update mechanisms, the changing patterns of the equipment in different time periods can be accurately captured to obtain accurate prediction results.

[0110] In one exemplary embodiment, such as Figure 4 As shown, based on a target analysis model adapted to the data type of multidimensional operational data, the operating status of new energy power generation equipment is predicted to obtain status prediction data, including steps 402 to 404.

[0111] Step 402: In the case of multidimensional operational data including visual data, determine the target analysis model, including the defect detection model, that is compatible with the multidimensional operational data.

[0112] The defect detection model is used to predict the condition of new energy power generation equipment based on visual data. Visual data can include optical imaging data or image data based on other imaging principles. This embodiment takes photovoltaic power generation equipment as an example. A high-precision current / voltage monitoring device can be deployed on the DC side of the photovoltaic inverter to capture the harmonic characteristics of the current waveform in real time. An infrared thermal imaging monitoring network covers the entire array of components, achieving high-density scanning of temperature distribution, and can also acquire optical images of the photovoltaic panel surface. Edge computing nodes use signal processing algorithms such as wavelet analysis to extract feature parameters reflecting the health status of the inverter from the current waveform, perform temperature field calibration and feature enhancement on the infrared thermal imaging data, and identify abnormal areas exceeding normal operating temperatures. After spatiotemporal alignment and feature fusion, multi-source data forms a multi-dimensional feature vector containing electrical, temperature, and visual features. This vector not only retains the original features of each dimension but also constructs derived features reflecting parameter coupling relationships through cross-feature engineering, providing a more comprehensive input dimension for subsequent intelligent diagnosis. The defect detection model can significantly improve the accuracy and reliability of hot spot detection by integrating electrical features (such as equipment operating current, voltage fluctuations, and insulation resistance changes) with thermal image visual features (such as pixel grayscale distribution, regional contrast, and edge gradient information).

[0113] Step 404: Input the visual data into the defect detection model to perform potential defect detection processing, identify potential defect areas of the new energy power generation equipment, and obtain state prediction data.

[0114] Potential defect detection processing refers to the detection operation that predicts whether visual data will develop into defects. A potential defect region is an area that has not yet formed a defect at present but is predicted to have a high probability of developing into a defect in the future. It is understandable that the input visual data may not contain defects such as cracks, but the defect detection model can predict regions in the visual data that will develop into defects by combining operational data from other dimensions.

[0115] The defect detection model in this embodiment includes the YOLOv8 model. During the identification of potential defect regions, the input visual data is first scaled and padded, and pixel values ​​are normalized. The backbone network of the YOLOv8 model extracts depth features from targets such as hot spots, bird droppings, and fallen leaves, while the neck network fuses multi-scale features to identify targets of different sizes. The detection head densely predicts bounding boxes and categories that may contain the target at three scales. The coordinates are decoded, low-confidence predictions are filtered out, and only the most accurate bounding box is retained for each hot spot or foreign object. The coordinates of the bounding box are then mapped back to the original visual data, and the bounding box can serve as a marker for potential defect regions.

[0116] In one embodiment, after inputting visual data into a defect detection model for potential defect detection processing, identifying potential defect areas of new energy power generation equipment, and obtaining state prediction data, the method further includes: extracting environmental state data from multidimensional operating data; determining a visual detection threshold that matches the environmental state data; and calibrating the detection boundary of the defect detection model based on the visual detection threshold.

[0117] Environmental condition data refers to data characterizing the operating environment conditions of new energy power generation equipment. The visual inspection threshold is a threshold used to determine whether a region in the visual data is a potential defect area. For an image within the detection frame, if the similarity to an early-stage hotspot image exceeds the visual inspection threshold, it can be identified as a potential defect area. Furthermore, the visual inspection threshold can be dynamically adjusted through operating condition perception and adaptive adjustment to optimize the detection boundary.

[0118] The system can acquire real-time ambient temperature (-20℃~60℃ industrial scene range), light intensity (0~10000 lux), and real-time electrical data of the equipment through its built-in temperature sensor, light sensor, and electrical parameter acquisition unit. This data is then input into a pre-trained working condition threshold mapping model. This model is trained on a vast amount of historical working conditions (including extreme scenarios such as high-temperature exposure, low-temperature freezing, direct sunlight, and low-light conditions). Through multinomial fitting and gradient descent algorithms, it dynamically outputs visual detection thresholds that match the current working conditions. For example, in high-temperature environments, the background grayscale of thermal images is high, so the hot spot detection threshold can be increased by 15%~30% to avoid background interference. In low-light environments, the light redundancy threshold is reduced to enhance the distinction between hot spots and the background. This visual detection threshold is then combined with a defect detection model for boundary calibration. By calculating the spatial overlap between abnormal areas and bright areas in the thermal image, the initial detection box coordinates are corrected, eliminating boundary offsets caused by light reflection and temperature gradient interference, and accurately selecting the hot spot defect range.

[0119] In this embodiment, the potential defect area is predicted by the defect detection model, and the detection boundary related to the potential defect area is calibrated. This can effectively avoid the problem of missed detection and false detection that is prone to occur in the fixed threshold algorithm under extreme weather conditions, and realize the accurate identification and location of hot spot defects.

[0120] In one exemplary embodiment, such as Figure 5 As shown, state prediction data is used as input data, and the optimization objectives are to maximize the power generation and minimize the failure rate of new energy power generation equipment. Global optimization scheduling is performed to obtain optimized scheduling parameters, including steps 502 to 506.

[0121] Step 502: Determine the scheduling action space of the new energy power generation equipment.

[0122] This embodiment employs a reinforcement learning algorithm for global optimization scheduling. The scheduling action space refers to the range of values ​​for scheduling parameters involved in the actions output by reinforcement learning. The actions output by reinforcement learning are adjustments to the operating parameters of new energy power generation equipment. Since the operating parameters of the equipment have adjustment limits, the range of the scheduling action space can be defined by the performance and model of the new energy power generation equipment.

[0123] Step 504: Input the state prediction data into the reinforcement learning policy network under the scheduling action space constraint to obtain the action probability distribution.

[0124] The Markov decision process constructed in the reinforcement learning algorithm can be configured as follows: the state input is a state feature vector obtained by transforming the state prediction data, and the action space is fixed at a preset number of discrete actions. The weight coefficients in the reward function can be dynamically adjusted according to different requirements for power generation and failure rate. Both the reinforcement learning policy network and the evaluation network can be designed with an input layer dimension equal to the state feature vector dimension, followed by two hidden layers, each with 128 neurons, and the activation function being ReLU. The trained reinforcement learning policy network can output an action probability distribution for the input state feature vector, providing parameters to adjust the possible probability range of actions.

[0125] Step 506: Taking the maximization of power generation and the minimization of failure rate of new energy power generation equipment as the optimization objectives, sample from the action probability distribution to obtain the optimized scheduling parameters.

[0126] An action can be randomly sampled from the action probability distribution as the parameter adjustment action for new energy power generation equipment. This random sampling operation ensures a certain degree of exploratory nature. The parameters associated with the parameter adjustment action can be used as optimized scheduling parameters, or the parameters after executing the parameter adjustment action can be used as optimized scheduling parameters.

[0127] In this embodiment, global optimization scheduling is performed using reinforcement learning algorithms, which can respond to changes in state prediction data within a shorter time period and improve the timeliness of scheduling optimization for new energy power generation equipment.

[0128] In one exemplary embodiment, such as Figure 6 As shown, after encapsulating the optimized scheduling parameters into scheduling control instructions, the method further includes steps 602 to 604.

[0129] Step 602: Map the multi-dimensional operating data to the digital twin model of the new energy power generation equipment, perform fault simulation and deduction, and obtain fault deduction information.

[0130] Fault simulation refers to simulating equipment failure scenarios through simulation. Fault simulation information is information that characterizes the expected location of the failure after simulation.

[0131] In the case of wind power generation equipment, a three-dimensional digital twin model highly mapped to the physical equipment can be constructed based on laser point cloud and multiphysics simulation. This model integrates real-time data such as wind field distribution and equipment load, and extrapolates the temperature and stress field distributions of key components in real time. When local overheating or abnormal vibration is detected, fault prediction information is generated. Furthermore, multiphysics coupled simulation can predict fault development trends, providing spatiotemporal predictive support for operation and maintenance decisions.

[0132] When the new energy power generation equipment is photovoltaic (PV) equipment, a digital twin model of the PV array can integrate GIS (Geographic Information System) maps, weather forecasts, and equipment operation data to construct a virtual mapping system covering the entire field. This model uses photoelectric conversion efficiency simulation and heat loss analysis to evaluate the apparent efficiency and potential defects of each subarray in real time, predict hot spot defects, and generate fault prediction information. An optimization algorithm based on reinforcement learning aims to maximize overall power generation. Combining short-term weather forecasts and equipment health status, it dynamically optimizes the MPPT parameters of each inverter, achieving smooth power output control and efficiency improvement.

[0133] Step 604: Determine the fault area represented by the fault simulation information and highlight the fault area in the digital twin model.

[0134] Faults such as localized overheating, abnormal vibration, and hot spot defects can all be characterized using fault simulation information, and the defect locations can be recorded as fault areas. Digital twin models can synchronously display the vibration, temperature, and other states of physical equipment, visualize the equipment's thermal distribution through temperature color gradations (red-yellow-blue), display the frequency change trend of the vibration spectrum in the form of a waterfall plot, and highlight fault areas by selecting or zooming in.

[0135] In this embodiment, a digital twin model is added to map the operating status of the new energy power generation equipment, which intuitively displays the operating status of the equipment and improves the visualization of equipment maintenance.

[0136] In one exemplary embodiment, such as Figure 7 As shown, the method includes the following steps S1 to S3.

[0137] S1. Perform multi-source three-dimensional sensing of new energy equipment to acquire multi-dimensional operational data. Deploy a three-axis vibration sensor on the high-speed shaft of the wind turbine gearbox and embed a temperature sensor on the backsheet of the photovoltaic module. Perform wavelet packet decomposition on the vibration signal through edge nodes to extract time-domain, frequency-domain, and time-frequency-domain features, and finally generate feature vectors. Perform non-uniformity correction on the infrared thermal image to control temperature drift. Use voxel filtering downsampling on the lidar point cloud data to reduce the amount of computation while retaining key features.

[0138] S2. Multidimensional operational data is identified using a target analysis model to obtain state prediction data. An improved LSTM-Transformer network receives vibration and temperature time-series data, and feature weights are calculated using a bidirectional attention mechanism. FocalLoss is used to address sample imbalance, and a YOLOv8 model is combined for visual defect detection. When the abnormal score of new data exceeds a set threshold, the model parameters are dynamically updated. A Gaussian mixture model is used to dynamically update the abnormal detection threshold, effectively preventing model overfitting.

[0139] S3. Global scheduling optimization and digital twin-driven control based on state prediction data. A deep reinforcement learning algorithm is used to generate the field group control strategy, referring to Equations 9 and 10 below.

[0140] Formula 9: .

[0141] Formula 10: .

[0142] Wherein, s_t is the state space, including state features such as wind speed, a_t is the action space, r_t is the reward function, with the goal of maximizing power generation and minimizing failure rate, and γ is the discount factor. The three-dimensional digital twin model synchronizes the vibration, temperature, and other states of the physical equipment in real time, visualizes the thermal distribution of the equipment through temperature color gradation (red-yellow-blue), and displays the frequency change trend of the vibration spectrum in the form of a waterfall plot.

[0143] This embodiment constructs a technical architecture from a perception model to an analysis model and then to a control model, realizing intelligent operation of the entire process from equipment status monitoring to predictive maintenance; it designs a cross-modal feature fusion algorithm to solve the problem of joint representation of heterogeneous data such as vibration, temperature, and vision; and it develops a global optimization model based on digital twins to improve the energy efficiency management and fault early warning capabilities at the power plant level.

[0144] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0145] Based on the same inventive concept, this application also provides a new energy power generation equipment control device for implementing the aforementioned new energy power generation equipment control method. The solution provided by this device is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more embodiments of the new energy power generation equipment control device provided below can be found in the limitations of the new energy power generation equipment control method described above, and will not be repeated here.

[0146] In one exemplary embodiment, such as Figure 8 As shown, a new energy power generation equipment control device 800 is provided, including: a multi-dimensional data acquisition module 801, an operating status prediction module 802, a global optimization scheduling module 803, and an equipment scheduling control module 804, wherein:

[0147] The multidimensional data acquisition module 801 is used to acquire multidimensional operating data of new energy power generation equipment.

[0148] The operation status prediction module 802 is used to predict the operation status of new energy power generation equipment based on a target analysis model that is compatible with the data type of multidimensional operation data, and obtain status prediction data.

[0149] The global optimization scheduling module 803 is used to take the state prediction data as input data, and with the optimization objectives of maximizing the power generation and minimizing the failure rate of new energy power generation equipment, perform global optimization scheduling processing to obtain optimized scheduling parameters.

[0150] The equipment scheduling and control module 804 is used to encapsulate optimized scheduling parameters into scheduling control instructions; the scheduling control instructions are used to control new energy power generation equipment.

[0151] In an exemplary embodiment, the operating status prediction module 802 is further configured to: determine, when the multidimensional operating data includes time-series data, that the target analysis model adapted to the multidimensional operating data includes a time-series neural network model; input the time-series data into the time-series neural network model, predict the fault components in the time-series data, and obtain the status prediction data.

[0152] In an exemplary embodiment, the operating status prediction module 802 is further configured to: extract vibration prediction data and temperature prediction data from the status prediction data; and output fault warning information when the abnormal correlation between the vibration prediction data and the temperature prediction data meets the fault warning conditions.

[0153] In an exemplary embodiment, the operating status prediction module 802 is further configured to: determine, when the multidimensional operating data includes visual data, that the target analysis model adapted to the multidimensional operating data includes a defect detection model; input the visual data into the defect detection model, perform potential defect detection processing, identify potential defect areas of the new energy power generation equipment, and obtain status prediction data.

[0154] In an exemplary embodiment, the operating state prediction module 802 is further configured to: extract environmental state data from multidimensional operating data; determine a visual detection threshold that matches the environmental state data; and calibrate the detection boundary of the defect detection model based on the visual detection threshold.

[0155] In an exemplary embodiment, the global optimization scheduling module 803 is further configured to: determine the scheduling action space of the new energy power generation equipment; input the state prediction data into the reinforcement learning policy network under the constraints of the scheduling action space to obtain the action probability distribution; and sample from the action probability distribution to obtain optimized scheduling parameters with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment.

[0156] In an exemplary embodiment, the global optimization scheduling module 803 is further configured to: map multi-dimensional operating data to a digital twin model of the new energy power generation equipment, perform fault simulation and deduction, and obtain fault deduction information; determine the fault area represented by the fault deduction information, and highlight the fault area in the digital twin model.

[0157] Each module in the aforementioned new energy power generation equipment control device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0158] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores multi-dimensional operational data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a control method for a new energy power generation device.

[0159] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a control method for a new energy power generation device. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0160] Those skilled in the art will understand that Figure 9 and Figure 10The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0161] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0162] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0163] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0164] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0165] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0166] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A control method for new energy power generation equipment, characterized in that, The method includes: Acquire multi-dimensional operational data of new energy power generation equipment; Based on a target analysis model that is compatible with the data type of the multidimensional operational data, the operating status of the new energy power generation equipment is predicted to obtain status prediction data. Using the state prediction data as input data, and with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, global optimization scheduling is performed to obtain optimized scheduling parameters. The optimized scheduling parameters are encapsulated into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment.

2. The method according to claim 1, characterized in that, The target analysis model, based on a data type adapted to the multidimensional operational data, predicts the operating status of the new energy power generation equipment, obtaining status prediction data, including: When the multidimensional operational data includes time-series data, the target analysis model that is adapted to the multidimensional operational data is determined to include a time-series neural network model; The time series data is input into the time series neural network model to predict the fault components in the time series data, thereby obtaining state prediction data.

3. The method according to claim 2, characterized in that, The method further includes: Extract the vibration prediction data and temperature prediction data from the state prediction data; If the abnormal correlation between the vibration prediction data and the temperature prediction data meets the fault warning conditions, a fault warning message is output.

4. The method according to claim 1, characterized in that, The target analysis model, based on a data type adapted to the multidimensional operational data, predicts the operating status of the new energy power generation equipment, obtaining status prediction data, including: When the multidimensional operational data includes visual data, the target analysis model that is adapted to the multidimensional operational data includes a defect detection model. The visual data is input into the defect detection model to perform potential defect detection processing, identify potential defect areas of the new energy power generation equipment, and obtain state prediction data.

5. The method according to claim 4, characterized in that, The method further includes: Extract environmental status data from the multidimensional operational data; Determine a visual detection threshold that matches the environmental state data; Based on the visual detection threshold, the detection boundary of the defect detection model is calibrated.

6. The method according to claim 1, characterized in that, The process involves using the state prediction data as input data and taking the maximization of power generation and minimization of failure rate of the new energy power generation equipment as optimization objectives to perform global optimization scheduling processing, resulting in optimized scheduling parameters, including: Determine the scheduling action space of the new energy power generation equipment; The state prediction data is input into the reinforcement learning policy network under the scheduling action space constraint to obtain the action probability distribution; With the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, the optimized scheduling parameters are obtained by sampling from the action probability distribution.

7. The method according to any one of claims 1 to 6, characterized in that, The method further includes: The multidimensional operating data is mapped to the digital twin model of the new energy power generation equipment to perform fault simulation and obtain fault simulation information; The fault region represented by the fault simulation information is determined, and the fault region is highlighted in the digital twin model.

8. A control device for new energy power generation equipment, characterized in that, The device includes: The multi-dimensional data acquisition module is used to acquire multi-dimensional operating data of new energy power generation equipment; The operation status prediction module is used to predict the operation status of the new energy power generation equipment based on a target analysis model that is compatible with the data type of the multidimensional operation data, and to obtain status prediction data. The global optimization scheduling module is used to take the state prediction data as input data, and with the optimization objectives of maximizing the power generation and minimizing the failure rate of the new energy power generation equipment, perform global optimization scheduling processing to obtain optimized scheduling parameters. The equipment scheduling and control module is used to encapsulate the optimized scheduling parameters into scheduling control instructions; the scheduling control instructions are used to control the new energy power generation equipment.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images