Intelligent operation inspection method and platform for new energy power generation grid-connected interface device

Abstract

The invention discloses an intelligent operation and maintenance method and platform for a new energy power generation grid-connected interface device, and relates to the technical field of power systems and control thereof.The method comprises the steps that effective faults and invalid faults are distinguished by calling historical operation and maintenance data of the grid-connected interface device, and adversarial training is carried out based on the data; and generating an intelligent operation inspection model with a short-time memory function. And receiving grid-connected operation data, carrying out fault detection and validity judgment by using the model, and generating an operation detection fault list. And a model output is established to be in communication connection with the fault splitting controller and the anti-islanding controller, early warning is performed on the fault list, protection response of the related controller is triggered, and it is ensured that preset conditions are met. The technical problem of low fault identification and response efficiency in the existing new energy power generation operation inspection is solved, and the technical effects of improving the accuracy and response speed of fault identification and enhancing the operation reliability and safety of the new energy power generation grid-connected interface device are achieved through the intelligent operation inspection method.

Description

Technical Field

[0001] The present application relates to the field of power systems and control technologies thereof, and in particular to an intelligent operation and inspection method and platform for a new energy power generation grid-connected interface device. Background Art

[0002] With the rapid development of renewable energy, new energy generation (such as wind and solar energy) has gradually become an important part of power supply. However, the grid connection of these new energy sources faces challenges such as unstable power generation characteristics and complex operating environment. Traditional manual monitoring and manual fault response result in slow fault identification and low accuracy, which makes it difficult to meet the needs of efficient operation and maintenance. The lack of intelligent operation and inspection means makes the fault prevention and response mechanism inflexible. Therefore, it is an urgent need for the industry to develop an operation and inspection method based on intelligent technology to improve fault detection and response capabilities.

[0003] At the current stage, relevant technologies still have the technical problem of low efficiency in fault identification and response during operation and maintenance of renewable energy power generation. Summary of the invention

[0004] The present application provides an intelligent operation and inspection method and platform for a new energy power generation grid-connected interface device. It uses the historical operation and maintenance data of the grid-connected interface device to distinguish between valid faults and invalid faults, and conducts adversarial training based on the data to generate an intelligent operation and inspection model with short-term memory function. Receive grid-connected operation data, use the model to perform fault detection and effectiveness judgment, and generate an operation and inspection fault sheet. Establish a communication connection between the model output and the fault decoupling controller and the anti-islanding controller, and issue an early warning for the fault sheet, triggering the protection response of the relevant controllers to ensure that the preset conditions are met. This achieves the technical effect of improving the accuracy and response speed of fault identification and enhancing the reliability and safety of the operation of the new energy power generation grid-connected interface device through an intelligent operation and inspection method.

[0005] The present application provides an intelligent operation and inspection method for a new energy power generation grid-connected interface device, comprising: Call the historical operation and maintenance data of the grid-connected interface device to divide the valid fault data and the invalid fault data, wherein the need to execute a fault response is used as the dividing standard; conduct adversarial training based on the valid fault data and the invalid fault data to generate an intelligent operation and inspection model, wherein the intelligent operation and inspection model is determined through training decoupling and has a short-term memory function; receive the grid-connected operation data, perform fault detection and validity judgment in combination with the intelligent operation and inspection model, and determine the operation and inspection fault sheet; establish a communication connection between the output end of the intelligent operation and inspection model and the fault decoupling controller and the anti-islanding controller; perform a terminal display warning on the operation and inspection fault sheet, and trigger the fault decoupling controller and the anti-islanding controller to perform a protection response, wherein the protection response meets the preset trigger condition.

[0006] This application also provides an intelligent operation and inspection platform for a new energy power generation grid-connected interface device, including: A fault data division module, the fault data division module is used to call the historical operation and maintenance data of the grid-connected interface device, and divide the valid fault data and the invalid fault data, wherein the division standard is whether a fault response needs to be executed; an intelligent operation and inspection model production module, the intelligent operation and inspection model production module is used to perform adversarial training based on the valid fault data and the invalid fault data to generate an intelligent operation and inspection model, wherein the intelligent operation and inspection model is determined by training decoupling and has a short-term memory function; an operation and inspection fault sheet determination module, the operation and inspection fault sheet determination module is used to receive the grid-connected operation data, and perform fault detection and validity judgment in combination with the intelligent operation and inspection model to determine the operation and inspection fault sheet; a communication connection establishment module, the communication connection establishment module is used to establish a communication connection between the output end of the intelligent operation and inspection model and the fault decoupling controller and the anti-islanding controller; a protection response module, the protection response module is used to perform a terminal display warning on the operation and inspection fault sheet, and trigger the fault decoupling controller and the anti-islanding controller to perform a protection response, wherein the protection response meets the preset trigger condition.

[0007] The intelligent operation and inspection method and platform for the new energy power generation grid-connected interface device proposed in this application first distinguishes between valid faults and invalid faults by calling the historical operation and maintenance data of the grid-connected interface device, and conducts adversarial training based on the data to generate an intelligent operation and inspection model with short-term memory function. Receive grid-connected operation data, use the model to perform fault detection and effectiveness judgment, and generate an operation and inspection fault ticket. Establish a communication connection between the model output and the fault decoupling controller and the anti-islanding controller, and issue an early warning for the fault ticket, triggering the protection response of the relevant controllers to ensure that the preset conditions are met, and achieve the technical effect of improving the accuracy and response speed of fault identification and enhancing the reliability and safety of the operation of the new energy power generation grid-connected interface device through the intelligent operation and inspection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in this application to illustrate the operations performed by the platform according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0009] Figure 1 A flow chart of an intelligent operation and inspection method for a new energy power generation grid-connected interface device provided in an embodiment of the present application; Figure 2A schematic diagram of the structure of the intelligent operation and maintenance platform of the renewable energy power generation grid-connected interface device provided in the embodiment of the present application.

[0010] Explanation of reference numerals: fault data division module 10 , intelligent operation and inspection model production module 20 , operation and inspection fault ticket determination module 30 , communication connection establishment module 40 , protection response module 50 . DETAILED DESCRIPTION

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0012] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0013] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, platform, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0014] The embodiment of the present application provides an intelligent operation and inspection method for a new energy power generation grid-connected interface device, such as Figure 1 As shown, the method includes: Step S100, call the historical operation and maintenance data of the grid-connected interface device, and divide the valid fault data and invalid fault data, wherein the division standard is whether the fault response needs to be executed. Specifically, in the intelligent operation and maintenance scheme of the grid-connected interface device of the new energy power generation, the historical operation and maintenance data of the grid-connected interface device must first be called. The data comes from the local server database, cloud storage system or specific storage device, etc., and is obtained through the data interface or related query language (such as SQL). After acquisition, the data is classified according to whether the fault response needs to be executed. If the fault will cause a substantial threat to the normal operation of the device, the quality of power output, the safety of the equipment, or the stability of the connection with the power grid after the fault occurs, and immediate or specific measures need to be taken to deal with the repair, such as overvoltage faults that may cause damage to the equipment or impact the power grid and need to perform tripping protection and other measures, such fault data is divided into valid fault data; on the contrary, faults such as short-term communication signal interference have a small impact, can be recovered by themselves, or can be alleviated by the system's own adjustment, without special intervention, and such fault data is divided into invalid fault data. During the division process, it is necessary to check the fault event records one by one, including time, phenomenon, operating parameters and treatment measures. With the help of data analysis software, reference can be made to equipment technical manuals, industry standards and operation and maintenance experience documents to ensure that the division is accurate and reasonable, and further organize and annotate the divided data to prepare for subsequent intelligent operation and inspection model training and fault analysis.

[0015] Step S200, adversarial training is performed based on the valid fault data and the invalid fault data to generate an intelligent operation and inspection model, wherein the intelligent operation and inspection model is determined by training decoupling and has a short-term memory function. Specifically, in the intelligent operation and inspection scheme of the new energy power generation grid-connected interface device, adversarial training is performed based on valid fault data and invalid fault data to generate an intelligent operation and inspection model. First, it is determined that the generator and the discriminator are the initial training architecture, a random noise vector is introduced into the generator, and a generation learning rate is set for the invalid fault data. The generator outputs the first generated sample based on the invalid fault data and the noise vector. Then, the discriminator discriminates the first generated sample from the real valid fault data, determines the discrimination probability with the data being valid or invalid as the goal, and then performs iterative alternating training until convergence conditions such as the stability of the discriminator loss function are met. When the training architecture converges, decoupling is performed, and the trained discriminator is determined as the intelligent operation and inspection model. The model has a short-term memory function. For example, when faced with situations that require time-phase judgment, it can combine the characteristics of previous moments, such as using architectures such as recurrent neural networks or long short-term memory networks, to remember past information and combine it with current inputs to more accurately judge faults, adapt to the dynamic changes and time-phase characteristics of the device's operating status, and provide strong support for intelligent operation and maintenance.

[0016] In one possible implementation, adversarial training is performed based on the valid fault data and the invalid fault data to generate an intelligent operation and inspection model, wherein the intelligent operation and inspection model is determined by training decoupling and has a short-term memory function, and step S200 further includes step S210, determining an initial training architecture, wherein a generator and a discriminator are used as the initial training architecture. Specifically, in the training of the intelligent operation and inspection model of the renewable energy power generation grid-connected interface device, the selection of the generator and the discriminator as the initial training architecture is based on the principle of adversarial learning. The role of the generator is to learn the characteristic distribution of effective fault data so that it can generate similar samples. The discriminator is responsible for distinguishing whether the input data is real effective fault data or samples generated by the generator. The adversarial relationship prompts both parties to continuously improve their capabilities. The generator strives to generate more realistic samples to deceive the discriminator, while the discriminator continuously improves its identification ability, so that the final trained model can better understand and identify the characteristics of effective fault data. For example, in an actual renewable energy power generation grid-connected system, effective fault data may include various types of faults, such as overvoltage, overcurrent, short circuit, etc. The generator tries to generate similar false fault data by learning the patterns and characteristics of these fault data. The discriminator must accurately determine whether the received data is from a real fault scenario or a simulation of the generator. Through continuous training, the two gradually reach a balance, so that the model can accurately identify and judge effective faults.

[0017] Step S220, introduce a random noise vector, and set a generation learning rate for the invalid fault data, based on the generator outputting the first generated sample. Specifically, introducing a random noise vector into the generator, on the one hand, increases the diversity of the generated data. In actual new energy power generation systems, the manifestation of faults may be affected by many factors and have a certain degree of randomness. By adding random noise, the generator can learn a wider range of fault feature patterns, not just limited to existing invalid fault data. For example, the noise vector can simulate uncertain factors such as environmental interference and measurement errors that may occur in actual operation, so that the generated samples are closer to real complex situations. On the other hand, random noise helps the generator explore different solution spaces and avoid falling into local optimal solutions. If there is no noise, the generator will overfit the existing invalid fault data, resulting in the generated samples being too single and unable to adapt well to various unknown situations. Setting a generation learning rate for invalid fault data is to control the parameter update step size of the generator during the learning process. The setting of the generation learning rate needs to comprehensively consider multiple factors. First, it must work within the range of the current invalid samples to ensure that the generator can be based on the existing data. Effective learning and improvement. Through appropriate learning rates, the generator can gradually adjust its parameters to better simulate the characteristics of valid fault data. Generating learning rates will also prompt the generator to generate other invalid fault data. In order to ensure comprehensive coverage, there are some invalid fault modes that have not yet been discovered or recorded in the renewable energy power generation system. Through appropriate learning rate adjustment, the generator has the opportunity to generate these potential fault data, so that the discriminator can learn a wider range of fault characteristics and improve the detection effect of the discriminator. Based on the generator, the introduced random noise vector and the set generation learning rate, the generator starts working and outputs the first generated sample. The generator uses the invalid fault data as the basis and transforms and processes the data in combination with the random noise vector. Through the internal neural network structure or other algorithms, the data features are extracted and reorganized, and attempts to generate samples with valid fault data characteristics. The generator continuously adjusts its parameters according to the set learning rate, so that the output first generated sample gradually approaches the characteristics and distribution of the real valid fault data.

[0018] Step S230, based on the discriminator, discriminate the first generated sample and the valid fault data to determine the discrimination probability, wherein the validity or invalidity of the data is the discrimination target. Specifically, the discriminator uses the validity or invalidity of the data as the discrimination target, and its working principle is based on the analysis and recognition of data features. The discriminator usually adopts a machine learning algorithm, such as a neural network, etc. First, it extracts features from the input data, and then compares and analyzes the extracted features with the pre-learned valid fault data features and invalid fault data features. The discrimination probability is determined by calculating the probability that the input data belongs to valid fault data and invalid fault data. In the intelligent operation and inspection of the new energy power generation grid-connected interface device, the discriminator needs to accurately determine whether the received data is from a real valid fault scenario or a simulated sample generated by the generator. For example, the discriminator may analyze the data. Electrical parameter characteristics, such as the change pattern of current and voltage, abnormal power factor, etc., and other related monitoring data characteristics, such as the change trend of temperature and humidity, etc. If the characteristics are similar to the known valid fault data characteristics, the discriminator will give a higher validity probability; conversely, if the characteristics are closer to the invalid fault data or the false sample characteristics generated by the generator, the discriminator will give a lower validity probability. When the discriminator receives the first generated sample and the real valid fault data, it will analyze and judge the data one by one. The discriminator assigns a probability value between 0 and 1 to each input data, indicating the possibility that the data is valid fault data. The process of determining the discrimination probability involves Multi-dimensional analysis and statistical calculation of data features. For the first generated sample, the discriminator will determine the probability based on the degree of difference between it and the real valid fault data. If the first generated sample is very similar to the real valid fault data, the discrimination probability will be close to 1; if the difference is large, the discrimination probability will be close to 0. For real valid fault data, the discrimination probability should be close to 1, but because there may be some noise or abnormal conditions in actual situations, the discrimination probability may fluctuate to a certain extent, but it is generally high. The significance of determining the discrimination probability is to provide feedback information for subsequent training. The generator can adjust its parameters and generation strategy based on the feedback given by the discriminator to generate more realistic and effective fault data. At the same time, the discriminator can also further optimize its discrimination ability according to the samples continuously improved by the generator. Through continuous adjustment and optimization, the two are gradually improved in the confrontation, so that the intelligent operation and inspection model finally trained can more accurately identify valid fault data. Moreover, in the actual renewable energy power generation system, accurate determination of the discrimination probability is crucial to the accuracy of subsequent alarms, automated emergency control and other responses. For example, if abnormal data caused by an instantaneous impact is misjudged as a valid fault, unnecessary shutdown or emergency measures will be initiated, affecting power generation efficiency; and if a real fault is misjudged as invalid, it will cause equipment damage or system failure expansion.

[0019] Step S240, iterative alternating training to obtain a training architecture that meets the convergence conditions. Specifically, alternating training is the core process of adversarial training, including multiple rounds of training. In each round of training, the generator and the discriminator are alternately trained and optimized. First, the generator generates a batch of new samples using invalid fault data and random noise vectors. The discriminator discriminates the generated samples and the real valid fault data and gives the discrimination probability. According to the feedback from the discriminator, the generator adjusts its own parameters to generate samples that are closer to the real and valid fault data. The discriminator discriminates the updated generated samples again and further adjusts its own parameters according to the results to improve the accuracy of discrimination. This process is repeated continuously. The generator and the discriminator gradually improve their abilities in the competition with each other. The discriminator needs to continuously improve its discrimination ability to accurately distinguish. During the iterative alternating training process, the training effect is continuously monitored to obtain a training architecture that meets the convergence conditions. The setting of convergence conditions is usually based on some evaluation indicators and standards. Common methods include observing the loss function and accuracy of the discriminator. The loss function of the discriminator reflects the degree of error in judging the validity of the data. When the loss function gradually decreases and stabilizes at a lower level, it means that the discriminator's discrimination ability is gradually improving, and it is difficult to further reduce the error under the current training state. It can be considered that the training architecture is beginning to approach the convergence state. Accuracy is also an important evaluation indicator. Accuracy refers to the proportion of the discriminator's correct judgments on the validity of data. When the accuracy no longer improves significantly in multiple consecutive training rounds, or reaches a pre-set higher level, it can also be used as a sign of convergence of the training architecture. When the convergence conditions are met, the obtained training architecture is a relatively stable and mature intelligent operation and inspection model architecture. In the intelligent operation and inspection of the new energy power generation grid-connected interface device, the converged training architecture can be used for subsequent fault detection and judgment, providing a basis for accurate intelligent operation and inspection.

[0020] In one possible implementation, an initial training architecture is determined, wherein the generator and the discriminator are used as the initial training architecture, and step S210 further includes step S211, determining an operation and inspection dimension, wherein the operation and inspection dimension at least includes grid connection, power management, and power safety. Specifically, the operation and inspection dimension is determined to include at least grid connection, power management, and power safety. The grid connection dimension focuses on the connection status and performance between the grid-connected interface device and the grid, including aspects such as connection stability, compatibility, and power transmission quality. For example, ensuring that the interface device can be stably connected under different grid conditions to avoid problems such as connection interruption and power quality degradation is crucial to ensuring the smooth integration of new energy power generation into the grid. The power management dimension involves the management of power distribution, scheduling, and control. In a new energy power generation system, it is necessary to reasonably manage the output and distribution of power to meet the needs of the grid and the stability requirements of the system, including the power generation function. The power safety dimension focuses on ensuring the safety of the entire power generation and grid connection process, covering the safe operation of equipment, the safety protection of personnel, and the system's ability to respond to various faults and abnormal situations. For example, it prevents overvoltage, overcurrent and other faults from causing damage to equipment and personnel, ensures that effective safety measures can be taken in time when a fault occurs, and ensures the overall safety of the system. Determining the operation and inspection dimension provides a clear direction and focus for subsequent fault analysis and model training, so that the intelligent operation and inspection model can learn and identify fault characteristics and operating conditions in different dimensions in a more targeted manner.

[0021] Step S212, based on the operation and inspection dimension, divide the training phase, perform clustering of the valid fault data and the invalid fault data, and determine the dimensional training data. Specifically, the purpose of dividing the training phase based on the operation and inspection dimension is to analyze and train fault data from different aspects in a more targeted manner. Each operation and inspection dimension has its own unique fault mode and characteristics. Dividing the training process according to the dimension can enable the model to focus on learning the characteristics of valid fault data and invalid fault data under a specific dimension at each stage, thereby improving the efficiency and accuracy of training. For example, in the training phase of the grid connection dimension, the focus is on connection-related faults, such as poor contact and signal transmission failures; in the training phase of the power management dimension, the focus is on problems such as power generation power fluctuations and unreasonable power distribution; in the training phase of the power safety dimension, the focus is on the effectiveness of various safety protection mechanisms and safety response measures in fault situations. Valid fault data and invalid fault data are clustered separately. Clustering is an unsupervised learning method that uses the similarity of data to identify faults. The data is divided into different groups or clusters. Under each operation and maintenance dimension, the valid fault data is clustered. Fault data with similar characteristics can be classified into one category, which is convenient for the model to learn the common characteristics of these faults. For example, under the grid connection dimension, the connection interruption fault data caused by different reasons are clustered into one category. The model can better understand the various manifestations and characteristics of connection interruption faults. Clustering of invalid fault data can help distinguish different types of non-critical faults or abnormal conditions that can be recovered by themselves, and provide the model with more comprehensive operating status information. Through clustering operations, dimensional training data is determined. Dimensional training data is a representative data set obtained after clustering processing under each operation and maintenance dimension. They provide an accurate data basis for subsequent phased supervised training, enabling the model to learn the fault mode and normal operation characteristics under each operation and maintenance dimension in a more targeted manner.

[0022] Step S213, based on the dimensional training data, the initial training architecture is supervised and trained in stages. Specifically, based on the dimensional training data, the initial training architecture (generator and discriminator) is supervised and trained in stages. In each training stage, that is, for each operation and inspection dimension, the model is trained using the training data of the corresponding dimension. First, the dimensional training data is input into the generator, and the generator generates simulated samples according to the data features. Then, the discriminator discriminates the generated samples and the real valid fault data to determine whether it is a valid fault or an invalid fault. During the training process, the parameters of the generator and the discriminator are continuously adjusted according to the feedback of the discriminator, so that they can better identify and generate the fault data features under this dimension. For example, in the training stage of the grid connection dimension, the grid connection related features are used. The generator learns how to generate samples similar to grid connection faults, and the discriminator focuses on distinguishing these samples from real grid connection fault data. Through continuous iterative training, the performance of the generator and the discriminator in the grid connection dimension gradually improves. Each stage requires the convergence condition to be met, which is the key to ensuring the training effect. The convergence condition is usually based on some evaluation indicators, such as loss function value, accuracy, etc. For the training of each operation and inspection dimension, when the loss function value gradually decreases and stabilizes at a lower level, or the accuracy reaches a higher standard and no longer improves significantly, it means that the training of the model in this dimension has achieved good results, that is, the convergence condition is met. In order to achieve convergence, the model parameters and learning rate and other parameters need to be continuously adjusted during the training process. Through optimization algorithms, such as gradient descent, the model gradually finds the optimal parameter settings during the training process to minimize the loss function and improve the accuracy. Some regularization methods can be used to prevent the model from overfitting, improve the generalization ability of the model, and ensure that the model can accurately detect and judge faults when facing new data. Staged supervised training has many advantages, making the training process more targeted. Each stage focuses on a specific operation and inspection dimension, and can learn the fault characteristics under this dimension more deeply, avoiding the confusion of fault characteristics in different dimensions. At the same time, staged training is also more orderly. , training is carried out in sequence according to the importance and relevance of the operation and inspection dimensions, which helps the model to gradually establish comprehensive fault detection and judgment capabilities. Meeting the convergence conditions at each stage ensures that the model can achieve good training results in each dimension, improves the accuracy and reliability of the entire intelligent operation and inspection model, and provides more powerful support for the intelligent operation and inspection of the renewable energy power generation grid-connected interface device. The process of determining the operation and inspection dimensions, dividing the training stages, performing data clustering, and supervising the training in stages realizes the refined training of the intelligent operation and inspection model of the renewable energy power generation grid-connected interface device, and improves the model's ability to detect and judge faults under different operation and inspection dimensions, thereby better ensuring the safe and stable operation of the renewable energy power generation system.

[0023] In a possible implementation, iterative alternating training is performed to obtain a training architecture that meets a convergence condition, and step S240 further includes step S241, decoupling the training architecture to determine a trained discriminator. Specifically, decoupling refers to separating the originally interrelated and interacting parts of the training architecture so that they can function independently or be further analyzed and applied. In the adversarial training architecture of the intelligent operation and maintenance of the new energy power generation grid-connected interface device, decoupling mainly deals with the relationship between the generator and the discriminator. During the training process, the generator and the discriminator improve their respective capabilities through continuous confrontation and interaction, and jointly learn the characteristics and patterns of effective fault data. When the training reaches a certain level, in order to obtain a model that can be directly used for intelligent operation and maintenance, it is necessary to decouple the training architecture. The purpose of decoupling is to determine an independent model that can accurately judge the effectiveness of the fault. Through decoupling, the trained discriminator is separated from the close interaction with the generator, so that it can focus on the fault discrimination task of the new data, and is no longer disturbed by the generator in subsequent operation. The ability to distinguish between valid fault data and invalid fault data learned by the discriminator during the training process can be more clearly utilized to provide a reliable judgment basis for intelligent operation and maintenance. The decoupling operation involves training The connection and interaction mechanism between the generator and the discriminator in the architecture are adjusted. First, it is necessary to stop the iterative training process between the generator and the discriminator, analyze the structure and parameters of the discriminator, determine which parameters are directly related to the discrimination of fault data, and which are intermediate parameters or auxiliary parameters generated in the process of confrontation with the generator. For the parameters directly related to fault discrimination, they are retained and sorted and optimized so that they can be used as an independent intelligent operation and inspection model in the future. For auxiliary parameters or parameters that closely interact with the generator, they can be appropriately adjusted or discarded as needed. For example, in a neural network-based training architecture, decoupling may involve cutting off certain layer connections between the generator and the discriminator, or adjusting the network structure of the discriminator to adapt it to independent fault discrimination tasks. The parameters of the discriminator need to be recalibrated and optimized to ensure its stable and accurate performance after decoupling, including fine-tuning of parameters, reinitialization of certain parameters, or using some optimization algorithms to further optimize the parameters, so that the decoupled discriminator can better adapt to the actual intelligent operation and inspection application scenarios.

[0024] Step S242: Use the discriminator as the intelligent operation and inspection model. Specifically, during the training process, the discriminator is mainly responsible for distinguishing between real valid fault data and samples generated by the generator. Through continuous learning and adjustment, it has acquired a deep understanding of the characteristics of fault data and accurate discrimination capabilities. In the intelligent operation and maintenance of the new energy power generation grid-connected interface device, what we need is a model that can accurately determine whether the new data is a valid fault. After adversarial training with the generator, the discriminator has learned to extract key features from the data and judge the validity of the data based on the features. Therefore, it is a reasonable choice to use it as an intelligent operation and maintenance model. For example, in the actual operation of the grid-connected interface device, when new operating data is input into the intelligent operation and maintenance model, the discriminator can quickly and accurately determine whether the data represents a valid fault condition that needs attention and processing based on the valid fault data feature pattern learned during the training process. It can determine whether there is a fault and the type and severity of the fault by analyzing the change patterns of parameters such as current, voltage, and power, as well as other related monitoring data features, providing an important basis for subsequent fault handling and operation and maintenance decisions. After the discriminator is used as an intelligent operation and maintenance model, it can be directly applied to the new energy power generation grid-connected interface device. Real-time monitoring and fault detection, in actual operation, continuously receives operation data from the device, and analyzes and distinguishes the data in real time. Its advantage is that it can quickly and accurately identify effective faults, and issue early warning signals in time, so as to take corresponding measures for fault handling and maintenance. Compared with traditional fault detection methods, the intelligent operation and inspection model obtained based on adversarial training has higher accuracy and adaptability, can learn complex fault modes and data characteristics, and has better recognition ability for some potential faults that are difficult to detect directly. In addition, the model can be further optimized and adjusted with the continuous input and accumulation of new data. By learning new fault conditions and data patterns that appear in actual operation, it continuously improves its discrimination ability and accuracy, so as to better adapt to the changes in the operating state of the new energy power generation grid-connected interface device and various complex working conditions, and provide strong support for the safe and stable operation of the new energy power generation system. By decoupling the training architecture and using the discriminator as the intelligent operation and inspection model, the key transformation from adversarial training to practical application is realized, which provides an efficient and accurate tool and method for the intelligent operation and inspection of the new energy power generation grid-connected interface device.

[0025] Step S300, receiving grid-connected operation data, combining the intelligent operation and inspection model to detect faults and determine effectiveness, and determine the operation and inspection fault list. Specifically, in the intelligent operation and inspection scheme of the new energy power generation grid-connected interface device, firstly, the operation data, including electrical parameters and non-electrical parameters, are collected in real time through various sensors and monitoring equipment equipped by the device, and transmitted to the data receiving end through the communication network, and pre-processed, such as data cleaning, denoising, format conversion and timestamp marking, and then stored in the database or data storage system. Next, the stored grid-connected operation data is input into the fully trained and optimized intelligent operation and inspection model. The model extracts and analyzes the data features, compares it with the fault feature pattern learned during training, and detects whether there is a fault, such as monitoring whether the current value exceeds the threshold and judging the fault conditions such as overcurrent in combination with other relevant parameters. After detecting a possible fault, the validity is determined based on the duration of the fault, the degree of impact on the system operation, and the correlation with other parameters, so as to distinguish between the faults that really need to be handled and the short-term abnormal fluctuations. Finally, an operation and maintenance fault ticket is generated based on the fault detection and effectiveness judgment results of the intelligent operation and maintenance model, which contains detailed information such as the time, type, severity, location and related operating data parameters of the fault. It can be used to notify operation and maintenance personnel to carry out timely repairs and transmit them to relevant systems or departments to coordinate resources to handle faults and maintain equipment. It can also be transmitted through various methods such as displaying alarms on the monitoring interface, sending emails or text messages, and integrating with the operation and maintenance management system, providing accurate and timely information support to ensure the safe, stable operation and efficient maintenance of the power system.

[0026] In a possible implementation, grid-connected operation data is received, fault detection and effectiveness determination are performed in combination with the intelligent operation and inspection model, and an operation and inspection fault list is determined. Step S300 further includes step S310, and the grid-connected operation data includes electrical data and monitoring data. Specifically, electrical data is one of the core components of grid-connected operation data, which directly reflects the electrical operation status of the new energy power generation grid-connected interface device, and the electrical data includes basic parameters such as current, voltage, and power. Current data can reflect the flow of charge in the circuit, which is crucial for monitoring the circuit load and whether there are abnormal conditions such as overcurrent. Voltage data reflects the power quality of the power system and the operating voltage level of the equipment. Stable voltage is the key to ensuring the normal operation of the equipment. Too high or too low voltage may cause damage to the equipment or affect its performance. Power data includes active power and reactive power. Active power reflects the actual consumption or generation of electric energy, while reactive power is related to the power factor of the power system and has an important impact on the efficiency and stability of the power system. Electric data is collected in real time and is continuously updated over time. Through continuous monitoring and analysis of electric data, abnormal fluctuations and signs of faults in the power system can be discovered in a timely manner, providing an important basis for intelligent operation and inspection. The monitoring data covers a wider range of information. It is used to fully understand the operating environment and status of the grid-connected interface device, including parameters such as temperature and humidity. Temperature monitoring is very important for the thermal management of the equipment. Excessive temperature may cause the equipment to overheat and be damaged, especially in new energy power generation equipment. For example, solar photovoltaic panels may heat up under long-term sunlight exposure. Temperature monitoring is required to ensure that they operate within a safe temperature range. Humidity data reflects the humidity of the environment. Excessive humidity may affect the insulation performance of the equipment and increase the risk of leakage and short circuit. Vibration data can reflect the mechanical operating status of the equipment. For example, the vibration of the motor may indicate whether its bearings are worn or whether there are faults in other mechanical parts. The monitoring data also includes the equipment's operating status signals, communication status and other information. The monitoring data and electrical data complement each other and together constitute a data set that fully reflects the grid-connected operating status.

[0027] Step S320, based on the fault correlation, the grid-connected operation data is mapped for correlation, and multiple groups of data at the same timestamp are determined. Specifically, the fault correlation refers to the inherent connection and mutual influence degree that may exist between different types of data when a fault occurs. In the renewable energy power generation grid-connected system, there are often complex correlations between various data. For example, when an overcurrent fault occurs, not only will the current data show abnormal changes, but it is usually accompanied by a rise in temperature, because excessive current may cause the line to heat up. Similarly, abnormal vibration of the equipment may affect the stability of the electrical signal, which is then reflected in the electrical data. Analyzing the fault correlation requires in-depth research on historical fault data and operation data. By analyzing a large number of fault cases, the typical correlation patterns between different data when a fault occurs are found. For example, when an overvoltage fault occurs, the changes in humidity data, temperature data and other related monitoring data are statistically analyzed to establish a correlation model between them. Data analysis techniques, such as related Methods such as correlation analysis and causal analysis are used to quantify the degree of correlation between different data. Based on the fault correlation, the grid-connected operation data is mapped for correlation. First, a time base is determined, usually in timestamp units. According to the fault correlation model, other monitoring data associated with the electrical data at the same timestamp are found. For example, for abnormal current data at a certain moment, the corresponding temperature data, humidity data and vibration data of the equipment are found through correlation mapping. This is achieved through data processing algorithms. For example, database query languages ​​are used to filter out related data records according to timestamp conditions and associate them. In practical applications, it is necessary to establish a data association table or data structure to store and manage multiple sets of related data at different timestamps, so as to provide more comprehensive data support for subsequent fault analysis and help the intelligent operation and inspection model to more accurately determine the type and cause of the fault.

[0028] Step S330, standardize the multiple groups of data and transmit them to the intelligent operation and inspection model. Specifically, the purpose of standardizing the multiple groups of data is to make data of different types, different units and different ranges comparable and consistent, so that the intelligent operation and inspection model can better process and analyze these data. Different electrical data and monitoring data may have different dimensions and numerical ranges. For example, the current may be in amperes and the temperature may be in degrees Celsius. Their numerical ranges vary greatly. If standardization is not performed, the model may be affected by the scale of the data when learning and analyzing the data, resulting in neglect or misjudgment of certain data features. Standardization methods usually include data normalization and data standardization. Data normalization is to map the data to a specific interval. Through the standardized processing method, different data can be analyzed at the same scale, which improves the training efficiency and accuracy of the model. After standardization, The processed multiple sets of data are transmitted to the intelligent operation and maintenance model. During the transmission process, it is necessary to ensure the integrity and security of the data to prevent data loss or tampering. Data encryption technology can be used to encrypt the transmitted data to ensure data security. After receiving the data, the intelligent operation and maintenance model can use its internal algorithm and model structure to analyze and process the data, perform fault detection and effectiveness determination and other operations. For example, the model can extract features and recognize patterns from standardized electrical data and monitoring data to determine whether there is a fault, and determine the type and severity of the fault based on pre-trained knowledge and experience. The accurate transmission of data and the effective analysis of the intelligent operation and maintenance model together constitute the key link of the intelligent operation and maintenance of the new energy power generation grid-connected interface device, providing strong support for ensuring the safe and stable operation of the system.

[0029] Step S400, establish a communication connection between the output end of the intelligent operation and inspection model and the fault decoupling controller and the anti-islanding controller. Specifically, in the process of establishing the communication connection in the intelligent operation and inspection scheme of the new energy power generation grid-connected interface device, the output characteristics of the intelligent operation and inspection model must first be analyzed, including its output data format, transmission frequency and data volume, such as outputting fault information and severity level in the form of a structure containing a fault code and a timestamp. At the same time, the input requirements of the fault decoupling controller and the anti-islanding controller are clarified, such as the fault decoupling controller's reception requirements for the fault type and severity to determine the decoupling operation, and the anti-islanding controller's acquisition requirements for the grid operation status and fault condition information to determine the islanding risk. Then, according to the characteristics and requirements of the two, a suitable communication interface and protocol are selected, and system scalability and compatibility must be considered. Then, hardware connection and configuration are performed, and the wireless connection must be configured accordingly and the signal quality must be paid attention to. Finally, software programming and integration are performed, and a communication driver is developed or a library function is called in the intelligent operation and inspection model system, and a data check and retransmission mechanism is used to ensure reliability and real-time performance. The communication connection is successfully established through the above steps to ensure the timely and accurate transmission of fault information.

[0030] Step S500, the terminal displays an early warning for the operation and inspection fault sheet, triggering the fault separation controller and the anti-islanding controller to perform a protection response, wherein the protection response meets the preset triggering conditions. Specifically, in the intelligent operation and inspection scheme of the new energy power generation grid-connected interface device, when the intelligent operation and inspection model determines the operation and inspection fault sheet, the system will immediately generate an early warning message and display it on the terminal, including the fault time, type, severity, etc., and mark it with different colors according to the severity, and push it to the operation and maintenance personnel in a variety of ways. The terminal display interface is clearly designed, with functions such as classification display, operation buttons and links, real-time updates, query functions, and display of system topology and marking of fault locations. The preset triggering conditions of the protection response are set according to the system safety operation requirements and equipment characteristics. The fault separation controller is triggered when a serious fault is detected and the current or voltage threshold and duration conditions are met. The anti-islanding controller is triggered when the power grid is powered off or the relevant parameters are abnormal and the power mismatch reaches a certain level. After being triggered, the fault decoupling controller disconnects the connection switch between the grid-connected device and the grid and feeds back the action status. The anti-islanding controller detects the islanding status and takes measures such as stopping output. At the same time, it works in coordination with the intelligent operation and inspection system and the grid dispatching system. The intelligent operation and inspection system records the entire process information of the fault, generates a maintenance work order and notifies the operation and maintenance personnel. The fault decoupling controller and the anti-islanding controller report the fault and decoupling to the grid dispatching system to ensure the safe and stable operation of the power system and the reliable operation of the equipment.

[0031] In one possible implementation, a terminal display warning is performed on the operation and maintenance fault sheet, triggering the fault decoupling controller and the anti-islanding controller to perform a protection response, wherein the protection response satisfies the preset triggering conditions, and step S500 further includes step S510, receiving the operation and maintenance fault sheet, and determining a fault map based on the relative distribution position. Specifically, when the intelligent operation and maintenance system generates an operation and maintenance fault sheet, the relevant processing module will receive the fault sheet in a timely manner, and the receiving process includes reading and parsing the fault sheet data to obtain the fault information contained therein, such as the time when the fault occurred, the specific fault type, the severity of the fault, and possible fault location identification. For example, the system receives the fault sheet through a specific data interface or communication protocol, and uses a pre-written parsing program to convert the text or structured data in the fault sheet into a format that can be processed within the system. For fault location identification, if the equipment number, module location or coordinate information is provided in the fault ticket, the system will extract it for accurate marking in the fault map later, and use the relative distribution position data in the received fault information to construct the fault map. The relative distribution position can be the physical position of the grid-connected interface device in the entire new energy power generation system, or it can be the position in the logical topology structure. For example, if it is a large solar power field, the grid-connected interface devices are distributed in different photovoltaic array areas, then the fault map will be drawn according to the actual geographical location of these devices in the power field. In the logical topology structure, the fault map may show the connection relationship and relative position between the grid-connected interface device and other equipment. For each fault, it is represented by a specific icon or mark on the fault map, and its corresponding fault information is marked. The operation and maintenance personnel can intuitively see the distribution of the fault in the entire system, which is convenient for quickly understanding the overall situation and impact range of the fault.

[0032] Step S520, traverse the fault map, and determine the operation and maintenance strategy according to the fault correlation. Specifically, when traversing the fault map, the correlation between different faults must first be analyzed. Fault correlation may include time correlation, spatial correlation, and causal correlation, etc. Time correlation refers to the relationship between the temporal sequence and frequency of occurrence of multiple faults. For example, if similar communication faults occur successively in multiple grid-connected interface devices in the same area within a short period of time, there may be a common time-related factor, such as enhanced electromagnetic interference in a certain time period. Spatial correlation focuses on the aggregation and correlation of faults in spatial positions. If adjacent grid-connected interface devices have overcurrent faults at the same time, due to a sudden increase in load in the area or problems with the shared part of the line, causal correlation is to analyze whether one fault may be the cause or result of another fault. For example, an overheating fault of a device may cause communication abnormalities in its surrounding devices because overheating may affect the performance of the communication line. By analyzing and modeling a large amount of historical fault data, combined with professional knowledge and experience, the system can establish fault correlation. The knowledge base and analysis model of fault correlation are established to quickly and accurately identify and analyze the correlation between faults when traversing the fault map. According to the analysis results of fault correlation, the corresponding operation and maintenance strategy is determined. For simple single faults, if the cause is relatively clear, such as an overcurrent fault caused by damage to a certain component, the operation and maintenance strategy may be to directly replace the component and detect and adjust the relevant parameters. For multiple faults with time correlation, it is necessary to further investigate the changes in the system operating environment within a specific time period, such as whether new equipment is started or weather conditions change, and then formulate corresponding temporary response measures and long-term prevention strategies, such as adjusting the operating time of the equipment or adding shielding measures to reduce electromagnetic interference. For faults with strong spatial correlation, it is necessary to conduct a comprehensive inspection and maintenance of the equipment and lines in the relevant area, such as detecting and repairing the shared lines. If there is a causal correlation fault chain, it is necessary to solve the problem from the root cause and repair and adjust the affected equipment accordingly. The operation and maintenance strategy also includes determining the deployment of maintenance personnel, the preparation of required tools and spare parts, and the time arrangement of maintenance to ensure that the fault can be handled in a timely and effective manner, while minimizing the impact on the operation of the entire new energy power generation system.

[0033] Step S530, the operation and maintenance strategy is transmitted to the operation and maintenance terminal to perform maintenance management on the grid-connected interface device. Specifically, the determined operation and maintenance strategy is transmitted to the operation and maintenance terminal through a suitable communication method. The operation and maintenance terminal can be a mobile device or a fixed workstation for use by operation and maintenance personnel. The communication method can be wireless network or wired network connection. After the terminal receives the operation and maintenance strategy, it is displayed to the operation and maintenance personnel. The operation and maintenance personnel perform maintenance management on the grid-connected interface device according to the operation and maintenance strategy received on the operation and maintenance terminal. First, prepare the required tools and spare parts according to the strategy requirements and go to the fault site. At the site, according to the fault map and maintenance suggestions, the faulty equipment is inspected and diagnosed in detail to confirm the specific cause of the fault. If the cause of the fault is consistent with the analysis in the operation and maintenance strategy, it is repaired according to the predetermined maintenance plan, such as replacing damaged components, adjusting equipment parameters or repairing lines. During the maintenance process, the operation and maintenance personnel can record the maintenance process and results through the operation and maintenance terminal, including information such as the actual replaced component model, maintenance time, test data, etc. The record not only helps to summarize and evaluate the current fault maintenance, but also provides data support for subsequent fault analysis and operation and maintenance strategy optimization. After the maintenance is completed, the grid-connected interface device is tested and verified to ensure that it resumes normal operation, and the maintenance completion information is fed back to the intelligent operation and inspection system so that the system can update the fault status and equipment operation status, providing accurate information for subsequent monitoring and management.

[0034] In a possible implementation, a terminal displays an early warning for the operation and maintenance fault sheet, triggering the fault decoupling controller and the anti-islanding controller to perform a protection response, wherein the protection response satisfies a preset trigger condition, and step S500 further includes step S540, combining the historical operation and maintenance data to mine the operation and maintenance cycle. Specifically, first, a large amount of historical operation and maintenance data is analyzed in detail, and the data includes information such as the time when the equipment failed in the past, the type of failure, maintenance records, equipment operation time, environmental conditions, etc. By sorting and counting the data, the frequency and regularity of equipment failures can be understood. For example, the number of times different types of failures occur in different time periods, as well as the operating time interval after each maintenance of the equipment, etc., are counted, and the performance trend of the equipment over time is analyzed, such as the changes in indicators such as the efficiency and stability of the equipment. Through in-depth mining of historical operation and maintenance data, some potential patterns and periodic characteristics can be discovered. For example, it is found that similar failures occur in certain equipment at regular intervals. According to the analysis results of the historical operation and maintenance data, the operation and maintenance cycle of the equipment is mined. The operation and maintenance cycle refers to the periodic maintenance of the equipment. The time intervals and cycles for inspection and maintenance need to take into account a variety of factors, including the frequency of failures, the importance of the equipment, the cost of maintenance, and the impact on the operation of the power system. If a certain type of equipment fails frequently and has a greater impact on the system operation, then the operation and inspection cycle may be relatively short. On the contrary, if the failure rate is low and the maintenance is relatively simple, the operation and inspection cycle can be appropriately extended. For example, through analysis, it is found that a certain type of grid-connected interface device will have a more serious failure once every two months on average. Then the operation and inspection cycle can be initially set to about one month, so that inspection and maintenance can be carried out in time before the failure occurs to reduce the impact of the failure on the system. Combined with the recommended maintenance cycle and industry standards provided by the equipment manufacturer, the determined operation and inspection cycle can be further adjusted and optimized to ensure its rationality and scientificity.

[0035] Step S550, for the operation and inspection cycle, determine the cycle nodes, and perform the operation and inspection management based on the intelligent operation and inspection model. Specifically, after determining the operation and inspection cycle, further clarify the key nodes in the cycle. The cycle nodes can be divided according to equal time intervals, or can be determined according to the specific stage or state of the equipment operation. For example, for a one-month operation and inspection cycle, each week can be regarded as a node for key inspection; or set nodes according to different stages of equipment startup, operation, and shutdown, and pay special attention to a period of time after the equipment is started, because the equipment may be more prone to some startup-related failures at this time. Determine the nodes in combination with the key performance indicators and operating parameters of the equipment. For example, when the current, voltage and other parameters of the equipment reach a certain threshold or have abnormal fluctuations, treat it as a node for detailed inspection. By determining the cycle nodes, the equipment can be monitored and managed more specifically, and potential problems can be discovered in time. At each node of the operation and inspection cycle, the intelligent operation and inspection model is used for operation and inspection management. The intelligent operation and inspection model can perform real-time analysis and judgment on the current operating data of the equipment, compare it with historical data, and detect whether There are abnormal situations. For example, the model can judge whether the equipment has signs of failure based on the current current, voltage, temperature and other data combined with the previously learned fault modes. For some complex situations, such as when the operating parameters of the equipment are at the edge of the normal range or there are some unusual fluctuations, the intelligent operation and inspection model can use its deep learning capabilities to conduct a more in-depth analysis. It can comprehensively consider multiple factors, including the equipment's operating history, current environmental conditions, and its relationship with other equipment, to make accurate judgments. For regular fixed inspections, if errors are found in the data or the operation and inspection situation is more complicated, the intelligent operation and inspection model can combine its internal algorithms and model structures to conduct a more detailed analysis of the data to determine whether there are potential faults. For example, when it is detected that the temperature of the equipment has increased slightly but is still within the normal range, the model can judge whether further attention or preventive measures are needed based on whether subsequent failures have occurred in similar situations in history and other current relevant parameters, such as load conditions, ventilation conditions, etc.

[0036] Step S560, the front end sets a fuzzy detection level, and the fuzzy detection level is set with basic operation and inspection logic. Specifically, the fuzzy detection level is set at the front end to perform preliminary monitoring and screening of the daily operation status of the equipment. The function of the fuzzy detection level is to quickly determine whether the equipment is in a normal operation state without overly complex and detailed analysis, and to use some relatively simple but effective detection methods and rules to monitor the key parameters of the equipment in real time. For example, a simple threshold judgment rule is set. When a parameter of the equipment (such as current) exceeds or falls below a certain threshold range, the alarm mechanism of the fuzzy detection level is triggered. This threshold range can be determined based on the normal operating range of the equipment and a certain empirical value. It does not require precise analysis like the intelligent operation and inspection model, but it can quickly discover some obvious abnormal situations and set the basic operation and inspection logic for the fuzzy detection level. The basic operation and inspection logic mainly includes the processing methods and judgment criteria for different situations. For some common and relatively simple situations, such as slight fluctuations in the parameters of the equipment within a certain range, the basic operation and inspection logic can stipulate simple recording and observation, and no further action is needed immediately. If the parameter fluctuation exceeds a certain range but is still not serious, some early warning information may be triggered to remind the operation and maintenance personnel to pay attention to the equipment status. For some special situations, such as multiple abnormal parameter fluctuations in a short period of time or some situations that do not conform to the normal pattern, the basic operation and inspection logic will combine the historical operation data of the equipment for preliminary analysis. For example, if the equipment has frequent current fluctuations in a certain period of time and has failed in similar situations in history, the basic operation and inspection logic will recommend starting a more detailed inspection process or passing the relevant information to the intelligent operation and inspection model for further analysis. The basic operation and inspection logic also includes regular inspection arrangements for equipment and some basic safety inspection rules to ensure the basic reliability and stability of the equipment in daily operation.

[0037] Step S570, determine the periodic time zone for the operation and maintenance cycle, and perform operation and maintenance management based on the fuzzy detection level. Specifically, the periodic time zone is determined according to the operation and maintenance cycle and the operating characteristics of the equipment. The periodic time zone can be divided according to different time periods of the day, different dates of the week, or different seasons of the year. For example, for some equipment, the operating load is large during the day and the failure rate is relatively high, so the daytime can be divided into a time zone of focus; or in summer, due to the high ambient temperature, the equipment has difficulty in heat dissipation and may be more prone to failure, so summer is managed as a special periodic time zone. When determining the periodic time zone, the work arrangements and resource allocation of the operation and maintenance personnel need to be taken into account. For example, during time periods with fewer operation and maintenance personnel or during holidays, the frequency and method of detection can be appropriately adjusted to ensure that limited resources are used. It can effectively ensure the safe operation of the equipment. Combined with the actual use of the equipment and the user's electricity demand, the cycle time zone is reasonably determined to minimize the impact of faults on the power supply. In different cycle time zones, operation and inspection management is carried out based on fuzzy detection levels. Under normal daily operation, the general equipment is in a relatively stable operating state. At this time, continuous data feedback and complex analysis and judgment may consume a lot of resources and may not be necessary. Daily supervision mainly relies on the basic operation and inspection logic of the fuzzy detection level. The fuzzy detection level will monitor the key parameters of the equipment in real time according to the set rules, such as collecting the current, voltage and other data of the equipment at a certain interval (for example, 10 minutes) and comparing them with the preset thresholds. If the parameters are within the normal range, only simple records and regular summary analysis are required to understand the operating trends of the equipment. When entering some special periodic time zones, such as periods of high equipment operating load or seasons with harsh environmental conditions, the fuzzy detection level will increase the monitoring frequency and intensity. For example, the data collection frequency is shortened to once every 5 minutes, and it is more sensitive to parameter fluctuations. If abnormal changes in parameters are found, the fuzzy detection level will make preliminary judgments and processing based on the basic operation and inspection logic. If the situation is more complicated or there is suspected potential fault, the intelligent operation and inspection model will be notified in time for further analysis and judgment to ensure that possible problems can be discovered and handled in time to ensure the safe and stable operation of the equipment during the operation and inspection cycle.

[0038] In the above, refer to Figure 1 The intelligent operation and inspection method of the new energy power generation grid-connected interface device according to the embodiment of the present invention is described in detail. Figure 2 An intelligent operation and inspection platform for a new energy power generation grid-connected interface device according to an embodiment of the present invention is described.

[0039] The intelligent operation and inspection platform of the new energy power generation grid-connected interface device according to the embodiment of the present invention is used to solve the technical problem of low efficiency of fault identification and response in the existing new energy power generation operation and inspection. Through the intelligent operation and inspection method, the technical effect of improving the accuracy and response speed of fault identification and enhancing the reliability and safety of the operation of the new energy power generation grid-connected interface device is achieved. The intelligent operation and inspection platform of the new energy power generation grid-connected interface device includes: a fault data division module 10, an intelligent operation and inspection model production module 20, an operation and inspection fault sheet determination module 30, a communication connection establishment module 40, and a protection response module 50.

[0040] The fault data classification module 10 is used to call the historical operation and maintenance data of the grid-connected interface device and classify the valid fault data and the invalid fault data, wherein whether a fault response needs to be executed is used as a classification standard.

[0041] The intelligent operation and inspection model production module 20 is used to perform adversarial training based on the valid fault data and the invalid fault data to generate an intelligent operation and inspection model, wherein the intelligent operation and inspection model is determined by training decoupling and has a short-term memory function.

[0042] The operation and inspection fault sheet determination module 30 is used to receive the grid-connected operation data, perform fault detection and effectiveness determination in combination with the intelligent operation and inspection model, and determine the operation and inspection fault sheet.

[0043] The communication connection establishing module 40 is used to establish a communication connection between the output end of the intelligent operation and inspection model and the fault separation controller and the anti-islanding controller.

[0044] The protection response module 50 is used to display a terminal warning for the operation and inspection fault list, trigger the fault decoupling controller and the anti-islanding controller to perform a protection response, wherein the protection response meets a preset trigger condition.

[0045] The specific configuration of the intelligent operation and inspection model production module 20 will be described in detail below. As described above, adversarial training is performed based on the valid fault data and the invalid fault data to generate an intelligent operation and inspection model, wherein the intelligent operation and inspection model is determined by training decoupling and has a short-term memory function, and the intelligent operation and inspection model production module 20 further includes: an initial training architecture determination unit, the initial training architecture determination unit is used to determine the initial training architecture, wherein the generator and the discriminator are the initial training architecture; a learning rate production unit, the learning rate production unit is used to introduce a random noise vector, and set a generated learning rate for the invalid fault data, and output a first generated sample based on the generator; a discrimination probability determination unit, the discrimination probability determination unit is used to discriminate the first generated sample and the valid fault data based on the discriminator, and determine the discrimination probability, wherein the validity and invalidity of the data are the discrimination targets; a training architecture acquisition unit, the training architecture acquisition unit is used for iterative alternating training to obtain a training architecture that meets the convergence conditions.

[0046] Among them, an initial training architecture is determined, wherein the generator and the discriminator are used as the initial training architecture, and the initial training architecture determination unit further includes: an operation and inspection dimension determination subunit, the operation and inspection dimension determination subunit is used to determine the operation and inspection dimension, wherein the operation and inspection dimension at least includes grid connection, power management and power safety; a training stage division subunit, the training stage division subunit is used to divide the training stage based on the operation and inspection dimension, perform clustering of the valid fault data and clustering of the invalid fault data, and determine dimensional training data; a staged supervised training subunit, the staged supervised training subunit is used to perform staged supervised training on the initial training architecture based on the dimensional training data.

[0047] Among them, iterative alternating training is performed to obtain a training architecture that meets the convergence conditions, and the training architecture acquisition unit further includes: a discriminator determination subunit, the discriminator determination subunit is used to decouple the training architecture and determine the trained discriminator; an intelligent operation and inspection model designation subunit, the intelligent operation and inspection model designation subunit is used to use the discriminator as the intelligent operation and inspection model.

[0048] The specific configuration of the operation and maintenance fault sheet determination module 30 will be described in detail below. As described above, the grid-connected operation data is received, and fault detection and validity judgment are performed in combination with the intelligent operation and maintenance model to determine the operation and maintenance fault sheet. The operation and maintenance fault sheet determination module 30 further includes: a grid-connected operation data composition unit, the grid-connected operation data composition unit is used for the grid-connected operation data to include electrical data and monitoring data; a correlation mapping unit, the correlation mapping unit is used to perform correlation mapping on the grid-connected operation data based on the fault correlation degree, and determine multiple groups of data under the same timestamp; a standardization processing unit, the standardization processing unit is used to perform standardization processing on the multiple groups of data and transmit them to the intelligent operation and maintenance model.

[0049] The specific configuration of the protection response module 50 will be described in detail below. As described above, the terminal displays an early warning for the operation and maintenance fault sheet, triggering the fault decoupling controller and the anti-islanding controller to perform a protection response, wherein the protection response meets the preset triggering conditions, and the protection response module 50 further includes: a fault map determination unit, the fault map determination unit is used to receive the operation and maintenance fault sheet, and determine the fault map based on the relative distribution position; an operation and maintenance strategy determination unit, the operation and maintenance strategy determination unit is used to traverse the fault map, and determine the operation and maintenance strategy according to the fault correlation; and a maintenance management unit, the maintenance management unit is used to transmit the operation and maintenance strategy to the operation and maintenance terminal, and perform maintenance management on the grid-connected interface device.

[0050] Among them, the protection response module 50 further includes: an operation and maintenance cycle mining unit, which is used to mine the operation and maintenance cycle in combination with the historical operation and maintenance data; a cycle node determination unit, which is used to determine the cycle node for the operation and maintenance cycle, and perform operation and maintenance management based on the intelligent operation and maintenance model; a basic operation and maintenance logic setting unit, which is used to set a fuzzy detection level at the front end, and the fuzzy detection level is set with a basic operation and maintenance logic; a cycle time zone determination unit, which is used to determine the cycle time zone for the operation and maintenance cycle, and perform operation and maintenance management based on the fuzzy detection level.

[0051] The intelligent operation and inspection platform of the new energy power generation grid-connected interface device provided in the embodiment of the present invention can execute the intelligent operation and inspection method of the new energy power generation grid-connected interface device provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0052] Although the present application makes various references to certain modules in the platform according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be implemented; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0053] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An intelligent operation and inspection method for a new energy power generation grid-connected interface device, characterized in that: The method comprises: Calling historical operation and maintenance data of the grid-connected interface device to divide valid fault data into invalid fault data, wherein whether a fault response needs to be executed is used as a dividing standard; Performing adversarial training based on the valid fault data and the invalid fault data to generate an intelligent operation and inspection model, wherein the intelligent operation and inspection model is determined by training decoupling and has a short-term memory function; Receive grid-connected operation data, perform fault detection and effectiveness determination in combination with the intelligent operation and inspection model, and determine the operation and inspection fault list; Establishing a communication connection between the output end of the intelligent operation and inspection model and the fault separation controller and the anti-islanding controller; A terminal display warning is performed on the operation and inspection fault sheet, triggering the fault decoupling controller and the anti-islanding controller to perform a protection response, wherein the protection response meets a preset trigger condition.

2. The intelligent operation and inspection method of the new energy power generation grid-connected interface device according to claim 1, characterized in that: Performing adversarial training based on the valid fault data and the invalid fault data includes: Determine the initial training architecture, where the generator and the discriminator are the initial training architectures; Introducing a random noise vector, and setting a generation learning rate for the invalid fault data, and outputting a first generation sample based on the generator; Based on the discriminator, the first generated sample and the valid fault data are discriminated to determine the discrimination probability, wherein the validity and invalidity of the data are used as the discrimination target; Iterate and alternate training to obtain a training architecture that meets the convergence conditions.

3. The intelligent operation and inspection method of the new energy power generation grid-connected interface device according to claim 2, characterized in that: The method comprises: Determining operation and inspection dimensions, wherein the operation and inspection dimensions at least include grid connection, power management and power safety; Based on the operation and inspection dimension, the training phase is divided, the clustering of the valid fault data and the clustering of the invalid fault data are performed, and the dimension training data is determined; Based on the dimensional training data, the initial training architecture is subjected to supervised training in stages.

4. The intelligent operation and inspection method of the new energy power generation grid-connected interface device according to claim 2, characterized in that: After obtaining the training architecture that meets the convergence conditions, including: Decoupling the training architecture to determine a trained discriminator; The discriminator is used as the intelligent operation and inspection model.

5. The intelligent operation and inspection method of the new energy power generation grid-connected interface device according to claim 1, characterized in that: After receiving the grid-connected operation data, including: The grid-connected operation data includes electricity data and monitoring data; Based on the fault correlation, correlation mapping is performed on the grid-connected operation data to determine multiple groups of data with the same timestamp; The multiple groups of data are standardized and transmitted to the intelligent operation and inspection model.

6. The intelligent operation and inspection method of the new energy power generation grid-connected interface device according to claim 1, characterized in that: After the terminal displays an early warning for the operation and inspection fault ticket, the following steps are performed: Receiving the operation and inspection fault sheet, and determining a fault map based on relative distribution positions; Traversing the fault map, and determining an operation and maintenance strategy according to fault correlation; The operation and maintenance strategy is transmitted to the operation and maintenance terminal to perform maintenance management on the grid-connected interface device.

7. The intelligent operation and inspection method of the new energy power generation grid-connected interface device according to claim 1, characterized in that: After the terminal displays an early warning for the operation and inspection fault ticket, the following steps are performed: Combine the historical operation and maintenance data to mine the operation and inspection cycle; Determine the cycle nodes for the operation and inspection cycle, and manage the operation and inspection based on the intelligent operation and inspection model; The front end sets a fuzzy detection level, and the fuzzy detection level is set with basic operation and inspection logic; For the operation and inspection cycle, a cycle time zone is determined, and operation and inspection management is performed based on the fuzzy detection checkpoint.

8. Intelligent operation and inspection platform for new energy power generation grid-connected interface device, characterized in that: The platform is used to implement the intelligent operation and inspection method of the new energy power generation grid-connected interface device according to any one of claims 1 to 7, and the platform includes: A fault data classification module, the fault data classification module is used to call the historical operation and maintenance data of the grid-connected interface device to classify valid fault data and invalid fault data, wherein whether a fault response needs to be executed is used as a classification standard; An intelligent operation and inspection model production module, the intelligent operation and inspection model production module is used to perform adversarial training based on the valid fault data and the invalid fault data to generate an intelligent operation and inspection model, wherein the intelligent operation and inspection model is determined by training decoupling and has a short-term memory function; An operation and inspection fault sheet determination module, the operation and inspection fault sheet determination module is used to receive grid-connected operation data, perform fault detection and effectiveness determination in combination with the intelligent operation and inspection model, and determine the operation and inspection fault sheet; A communication connection establishing module, the communication connection establishing module is used to establish a communication connection between the output end of the intelligent operation and inspection model and the fault separation controller and the anti-islanding controller; A protection response module, wherein the protection response module is used to display a terminal warning for the operation and inspection fault list, trigger the fault decoupling controller and the anti-islanding controller to perform a protection response, wherein the protection response meets a preset trigger condition.

Images