A multi-scene new energy station maintenance operation method and system and a medium

Abstract

The present application relates to the technical field of new energy station operation and maintenance, and discloses a multi-scene new energy station maintenance and operation method, system and medium. The present application trains and generates fault early warning models and health assessment models corresponding to different scene types for different scene types of new energy stations. Real-time monitoring data and scene type information of each station are obtained, and corresponding fault early warning models and health assessment models are called according to the scene type information. Fault identification and health assessment are performed on the real-time monitoring data, and early warning is performed according to the fault identification result and the health assessment result. According to the early warning result, a maintenance scheme is generated and distributed in combination with operation and maintenance resource information. The present application solves the technical problems that the existing technology cannot provide accurate guidance for maintenance work due to high false alarm rate and missed alarm rate caused by using a general fault early warning model, and that maintenance and operation do not form a whole-process closed-loop management.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance technology for new energy power stations, specifically to a method, system, and medium for the inspection and maintenance of new energy power stations in multiple scenarios. Background Technology

[0002] With the large-scale development of the new energy industry, wind and solar power stations in various scenarios, such as mountain wind power, photovoltaic power in coal mining subsidence areas, and agricultural-solar hybrid power, have become an important part of new energy installed capacity. Containerized substations are widely used in distributed and remote new energy power stations due to their advantages such as short construction cycle, flexible deployment, and controllable cost. Currently, most wind and solar power stations in China have achieved remote access to the centralized control center for operational data, but on-site maintenance and operation still mainly adopt the traditional model of "regular inspections + emergency repairs after failures".

[0003] The current maintenance and repair technology for new energy power plants has the following main shortcomings: First, there is poor scenario adaptability. Existing operation and maintenance technologies are mostly developed for wind and solar power stations of a single type or in a single geographical scenario, making it difficult to adapt to the differentiated operation and maintenance needs of multiple scenario clusters such as mountain wind power, mountain photovoltaic in coal mining subsidence areas, conventional mountain photovoltaic, and flatland agricultural photovoltaic. Under different scenarios, equipment is affected by factors such as geological subsidence, mountain shading, agricultural operations, and temperature and humidity differences, resulting in significant differences in its fault evolution patterns and environmental risk characteristics. When using a general fault early warning model, the false alarm rate and false negative rate are high, making it difficult to provide accurate guidance for maintenance work.

[0004] Secondly, there are blind spots in the safety management of unmanned substations. For containerized substations without perimeter walls or on-site personnel, current technology can only remotely upload electrical operation data, failing to deeply integrate perimeter security, environmental monitoring, equipment status, and fire protection linkage with the entire maintenance process. There is a lack of automatic verification and access control for maintenance personnel entering and exiting the substation; abnormal environmental conditions and security intrusions cannot be linked to maintenance tasks; emergency response capabilities are insufficient, posing significant safety hazards.

[0005] Third, the maintenance and repair process lacks a complete closed loop. Existing centralized control center monitoring systems are mostly limited to data display and passive fault alarms, failing to achieve a complete closed-loop management process encompassing "intelligent fault early warning—intelligent generation and assignment of maintenance tasks—on-site maintenance safety management and work guidance—maintenance result verification—model iterative optimization." Maintenance task assignment heavily relies on manual experience, lacking intelligent optimization based on the site's geographical environment, maintenance personnel skill tags, spare parts inventory, and road conditions. This results in slow maintenance response, high maintenance costs, and frequent repetitive maintenance issues.

[0006] Fourth, there is insufficient collaborative capability for clustered operation and maintenance. For maintenance work involving multiple power stations, existing technologies struggle to achieve unified scheduling and collaborative management of maintenance resources, personnel, and spare parts across different power stations. Maintenance personnel in remote areas such as mountainous regions and coal mining subsidence areas spend a lot of time traveling back and forth, and spare parts are not allocated in a timely manner, affecting the power generation efficiency and fault handling speed of the power stations. Summary of the Invention

[0007] This invention provides a method, system, and medium for the maintenance and operation of new energy power stations in multiple scenarios, in order to solve the technical problems in the prior art where the use of a general fault early warning model leads to a high false alarm rate and a high missed alarm rate, which makes it impossible to provide accurate guidance for maintenance work, and where maintenance and operation have not formed a closed-loop management of the entire process.

[0008] In a first aspect, the present invention provides a multi-scenario maintenance and repair method for new energy power stations, comprising: training and generating fault early warning models and health assessment models corresponding to different scenario types for new energy power stations; acquiring real-time monitoring data and scenario type information of each power station; calling the corresponding fault early warning model and health assessment model according to the scenario type information; performing fault identification and health assessment on the real-time monitoring data; and issuing early warnings based on the fault identification results and health assessment results; and generating and dispatching maintenance plans based on the early warning results and in conjunction with maintenance resource information.

[0009] This invention overcomes the limitation of general-purpose models, which cannot adapt to the differentiated operation and maintenance needs of various scenarios, by training and generating dedicated fault early warning models and health assessment models for different types of new energy power plants. Each scenario-specific model can accurately identify the evolution patterns of equipment faults and environmental risk characteristics under different geographical environments, effectively reducing false alarm and missed alarm rates and providing precise guidance for maintenance work. Simultaneously, by acquiring power plant scenario type information and calling the corresponding dedicated models for fault identification and health assessment, this invention achieves differentiated and precise early warnings for different scenarios. Based on this, it automatically generates and dispatches maintenance plans according to the early warning results and operation and maintenance resource information, thus streamlining the key link from early warning to maintenance task dispatch and realizing closed-loop management of maintenance operations. This avoids the uncertainty and lag of manual task dispatch, significantly improving maintenance response speed and operation and maintenance efficiency. Furthermore, the maintenance plans generated based on precise early warnings provide clear task guidance for maintenance personnel, reducing on-site survey and plan development time, and effectively reducing the risk of equipment fault escalation and power generation loss due to untimely maintenance.

[0010] In one optional implementation, the different scenario types include four categories: mountain wind power, mountain photovoltaic power in coal mining subsidence areas, conventional mountain photovoltaic power, and flatland agricultural-photovoltaic complementary photovoltaic power. This implementation specifically divides the different scenario types of new energy power plants into four typical scenarios: mountain wind power, mountain photovoltaic power in coal mining subsidence areas, conventional mountain photovoltaic power, and flatland agricultural-photovoltaic complementary photovoltaic power. This ensures that the scenario type classification accurately corresponds to the real geographical environment. The unique environmental risk factors of each scenario can be learned by the model as differentiated training features in the corresponding scenario's dedicated model, thereby ensuring that the fault warning model and health assessment model under each scenario can accurately capture the unique fault evolution patterns of that scenario. Compared with the existing technology that uses the same general model indiscriminately for different geographical scenarios, this implementation, through the accurate classification and corresponding modeling of the above four scenario types, significantly improves the accuracy of equipment fault warnings under different geographical environments, reduces the false alarm rate and missed alarm rate caused by scenario misjudgment, and thus provides accurate early warning guidance for maintenance work, effectively reducing ineffective maintenance or missed maintenance caused by inaccurate warnings.

[0011] In one optional implementation, the fault warning model and health assessment model are trained and generated using the following method: Data sets for each scenario are collected separately, with each dataset including electrical parameters, environmental parameters, scenario-specific additional data, and historical operation and maintenance data; fault warning models and health assessment models for each scenario are trained using machine learning algorithms based on the datasets for each scenario. This implementation ensures that the model training data comprehensively covers multi-dimensional factors affecting equipment operating status by separately collecting electrical parameters, environmental parameters, scenario-specific additional data, and historical operation and maintenance data for each scenario. Scenario-specific additional data refers to disturbance factors unique to a specific geographical scenario, while historical operation and maintenance data reflects the degradation patterns and fault histories of the equipment. These multi-dimensional data collectively constitute a complete profile of the equipment fault evolution patterns in each scenario. Based on this, dedicated models for each scenario are independently trained using the datasets for each scenario. The data from each scenario are not mixed or overlapped, avoiding the feature ambiguity problem caused by mutual interference between different scenario data during general model training. This allows the dedicated model for each scenario to accurately learn the unique fault evolution patterns and environmental risk characteristics of that scenario, effectively improving the accuracy of fault warning and health assessment, and reducing false alarm and false negative rates.

[0012] In one optional implementation, generating a maintenance plan includes: determining the fault level and assigning work according to the fault level; selecting matching maintenance personnel based on their real-time location, skills, qualifications, and on-duty status; matching spare parts based on spare parts inventory information and route access status; and performing multi-constraint weighted calculations based on the fault level, maintenance personnel selection results, spare parts matching results, and route access status to output the optimal maintenance plan. This implementation method achieves a tiered response mechanism by assigning tasks according to fault severity, prioritizing emergency faults and scheduling general faults, thus improving the timeliness of handling high-risk faults. By screening and matching personnel based on their real-time location, skills, qualifications, and on-duty status, and combining spare parts inventory and route accessibility with material and feasible routes, the feasibility of the maintenance plan is ensured from three dimensions: personnel accessibility, material availability, and route accessibility. Furthermore, by incorporating fault severity, personnel matching, spare parts matching, and route accessibility into a unified multi-constraint weighted calculation model for comprehensive optimization, the optimal global allocation of maintenance resources is achieved. This avoids maintenance delays caused by mismatched personnel skills, spare parts shortages, or route interruptions, effectively shortening maintenance response time and work preparation time, reducing operation and maintenance costs, and improving the scheduling efficiency and accuracy of clustered operation and maintenance.

[0013] In one optional implementation, the early warning system includes: compressing the trained fault early warning model and health assessment model into a lightweight model and deploying it at the field station; the field station uses the lightweight model to perform on-site inference on real-time monitoring data, triggering local linkage control and uploading alarm signals when an emergency fault is identified; and deploying the trained fault early warning model and health assessment model as a high-precision model at the central control center, where the central control center uses the high-precision model to analyze the uploaded data and achieve tiered early warning. This implementation achieves millisecond-level on-site identification and local linkage control of emergency faults by compressing the trained model into a lightweight model and deploying it at the field station, allowing the field station to perform on-site inference on real-time monitoring data, thus avoiding delays in response caused by data upload latency. Simultaneously, deploying the full-precision model at the central control center allows for in-depth analysis of the uploaded data and tiered early warning, ensuring accurate identification of complex faults. Through the division of labor and collaboration of the above two-level architecture, both the real-time nature of emergency response and the accuracy of early warning are guaranteed.

[0014] In one optional implementation, the method further includes: after the maintenance is completed, the central control center verifies the maintenance effect, confirms that the fault has been eliminated, and archives the maintenance task in a closed loop; the fault data and maintenance data are used as new samples to regularly retrain and iteratively optimize the fault early warning model and the health assessment model. This implementation method achieves closed-loop management of the entire maintenance task process, from early warning and dispatch to execution and result verification, by having the centralized control center verify the maintenance effect after the maintenance is completed and confirm that the fault has been eliminated. This avoids the shortcomings of traditional models, such as the untraceability of maintenance task execution status and the inability to effectively verify results. At the same time, by using the fault data and maintenance data as new samples, the fault early warning model and health assessment model are regularly retrained and iteratively optimized. This allows the model to continuously learn from newly occurring fault cases and actual maintenance feedback, constantly updating its understanding of the evolution of equipment faults in different scenarios. This enables the model's early warning capability to continuously improve with the accumulation of maintenance data, solving the technical problem that static models cannot adapt to dynamic changes in equipment status. This further reduces the false alarm rate and false negative rate, providing data and model support for the continuous optimization of maintenance strategies and preventive maintenance throughout the entire life cycle.

[0015] Secondly, this invention provides a multi-scenario new energy power station maintenance and operation system, comprising: a model management module, deployed in the central control center, for storing fault warning models and health assessment models corresponding to different scenario types; a power station edge computing module, deployed at the power station, for acquiring real-time monitoring data and power station scenario type information of each power station, calling the corresponding fault warning model and health assessment model according to the scenario type information, performing fault identification and health assessment on the real-time monitoring data, and issuing warnings based on the fault identification results and health assessment results; and a task scheduling module, deployed in the central control center, for generating and dispatching maintenance plans based on the warning results and combined with operation and maintenance resource information.

[0016] In one optional implementation, the site-side edge computing module includes: a lightweight early warning unit, used to acquire real-time monitoring data and preprocess it, then perform on-site inference on the preprocessed data, and trigger local linkage control and upload alarm signals when an emergency fault is detected.

[0017] In one optional implementation, it further includes: an intelligent sensing terminal, deployed on the equipment side of the site, used to collect real-time monitoring data and transmit it to the edge computing module at the site; the intelligent sensing terminal includes an electrical status sensing unit, an environmental sensing unit, and a perimeter security and access control unit.

[0018] Thirdly, the present invention provides a computer-readable storage medium storing computer instructions, which are used to cause a computer to execute the multi-scenario new energy power station maintenance and repair method described in the first aspect or any of its corresponding embodiments. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the first type of maintenance and repair method for multi-scenario new energy power stations according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the second process of the multi-scenario maintenance and operation method for new energy power stations according to an embodiment of the present invention; Figure 3 This is a functional structure block diagram of a multi-scenario new energy power station maintenance and repair system according to an embodiment of the present invention; Figure 4 This is a detailed architecture deployment diagram of the multi-scenario new energy power station maintenance and repair system according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0023] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0024] The existing maintenance and operation technologies for new energy power stations have the following problems: First, poor scenario adaptability. Existing technologies are mostly developed for single scenarios and cannot adapt to the differentiated maintenance needs of multi-scenario clusters. Equipment in different scenarios is affected by factors such as geological subsidence, mountainous terrain, agricultural operations, and temperature and humidity differences, resulting in significant differences in fault evolution patterns and environmental risk characteristics. General fault early warning models have high false alarm and false negative rates, failing to provide accurate guidance for maintenance work. Second, blind spots exist in the management and control of unattended substations. Existing technologies only achieve remote uploading of electrical operation data and do not integrate perimeter security, environmental monitoring, etc. The system lacks a deep integration of equipment status, fire alarm linkage, and maintenance processes, but it also lacks automatic verification and access control for personnel. Security anomalies cannot be linked to maintenance tasks to trigger emergency response. Third, maintenance has not formed a closed-loop process. Existing systems mostly remain at the stage of data display and passive alarm. The dispatch of maintenance tasks relies heavily on human experience, resulting in slow response, high costs, and frequent repetitive maintenance. Fourth, the clustered maintenance collaboration capability is insufficient. Existing technology cannot achieve unified scheduling and collaborative management of resources, personnel, and spare parts across sites. Personnel travel to and from remote sites takes a long time, and spare parts are not dispatched in a timely manner. Based on this, the present invention provides a method, system, and medium for the maintenance and operation of new energy power stations in multiple scenarios, in order to solve the technical problems in the prior art, such as the inability to accurately identify the evolution law of equipment failure and environmental risk characteristics in different geographical scenarios due to the use of a general fault early warning model, resulting in high false alarm and false alarm rates and failure to provide accurate guidance for maintenance work; the lack of comprehensive security control of unmanned container substations in terms of security, environment, access control and maintenance linkage; slow response and high cost due to the reliance on manual dispatch of maintenance tasks; and the inability to uniformly schedule and coordinate maintenance resources of multiple power station clusters.

[0025] This invention addresses four typical scenarios: mountain wind power, mountain photovoltaic power in coal mining subsidence areas, conventional mountain photovoltaic power, and flatland agricultural photovoltaic power. It constructs scenario-specific equipment fault early warning models and health assessment models, and combines a two-level early warning mechanism of edge-side on-site inference and in-depth analysis at the central control center. This enables differentiated and precise operation and maintenance for different geographical scenarios and different equipment types, and solves the technical problems of poor scenario adaptability and low early warning accuracy of existing general models.

[0026] This invention also constructs a comprehensive management and maintenance linkage system for unattended containerized substations, deeply integrating electrical condition monitoring, environmental perception, perimeter security, access control, and fire protection linkage with the entire maintenance process. It achieves closed-loop safety management throughout the entire process, including personnel identity verification, remote access control, and security linkage during operations, thus solving the technical problems of large blind spots and high safety risks in the maintenance and management of containerized substations without walls or on-site personnel.

[0027] This invention also proposes a closed-loop management method for the entire intelligent maintenance and operation process of multi-site cluster collaboration. It connects the entire chain of intelligent fault early warning, intelligent generation and dispatch of maintenance tasks, standardized control of on-site maintenance, verification of maintenance results and iterative optimization of models. Combining multi-dimensional information, it realizes intelligent collaborative scheduling of maintenance personnel, spare parts and tools across sites, and solves the technical problems of slow maintenance response, low resource utilization and high operation and maintenance costs in existing technologies.

[0028] This invention also designs a mobile maintenance terminal architecture adapted to complex geographical scenarios, supporting offline operations, remote expert assistance, and standardized operation guidance functions. It adapts to the maintenance operation needs of areas with weak network signals, such as mountainous areas and coal mining subsidence areas, and realizes full-process data traceability and controllability of maintenance operations, ensuring the standardization and continuity of maintenance operations in complex scenarios.

[0029] According to an embodiment of the present invention, a method for maintenance and operation of new energy power stations in multiple scenarios is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0030] This embodiment provides a method for the inspection and maintenance of new energy power stations in multiple scenarios. Figure 1 This is a flowchart of a multi-scenario maintenance and operation method for new energy power stations according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: For new energy power stations of different scenario types, respectively train and generate fault early warning models and health assessment models corresponding to different scenario types.

[0031] Among them, scenario type refers to the geographical environment category of the new energy power station. Different scenario types have different environmental risk characteristics and failure evolution patterns, such as the complex terrain and sudden wind speed changes of mountain wind power, the geological subsidence risk of photovoltaic power in coal mining subsidence areas, the terrain shading of conventional mountain photovoltaic power, and the agricultural operations and temperature and humidity fluctuations of flatland agro-photovoltaic complementary photovoltaic power. The fault early warning model is a machine learning model used to identify fault types and predict the probability of fault occurrence of equipment; the health assessment model is a machine learning model used to quantitatively assess the health status of equipment, outputting a health score reflecting the degree of equipment aging and remaining service life.

[0032] This step trains models independently for each scenario type. The training processes for each scenario do not interfere with each other and the data is not mixed. This generates a fault warning model and a health assessment model specific to each scenario. In other words, there is a one-to-one correspondence between the model and the scenario type. Each scenario has its own exclusive model, rather than all scenarios sharing a common model.

[0033] The purpose of this step is to bind the model to the scenario type, providing a model basis for subsequent differentiated early warning by calling the corresponding exclusive model according to the scenario type. This enables fault identification and health assessment to be based on the exclusive model that matches the current site scenario, thereby avoiding false alarms and missed alarms caused by the inability of the general model to identify the unique environmental risk characteristics under the specific scenario.

[0034] Step S102: Obtain real-time monitoring data and site scenario type information for each site; call the corresponding fault warning model and health assessment model according to the scenario type information; perform fault identification and health assessment on the real-time monitoring data; and issue a warning based on the fault identification results and health assessment results.

[0035] The aforementioned real-time monitoring data refers to the equipment operation status data collected in real time by various sensors deployed on the equipment side of the site; the site scene type information is used to identify the scene category to which the current site belongs.

[0036] This step selects the model corresponding to the current site's scenario type from among the multiple scenario-specific models trained in step S101. That is, it uses the scenario-specific model for processing, rather than using the same model indiscriminately for all sites. Then, it comprehensively judges the fault identification results and health assessment results before outputting a warning message.

[0037] The purpose of this step is to achieve accurate matching between scenarios and models through scenario type information. Each site uses only a dedicated model that matches its scenario type for fault identification and health assessment, ensuring that the analysis basis is consistent with the actual scenario characteristics of the site and avoiding misjudgments caused by using unmatched models. At the same time, the early warning information is a comprehensive judgment result that integrates the two dimensions of fault identification and health assessment, rather than a simple judgment based on a single dimension, thereby improving the accuracy and reliability of the early warning.

[0038] Step S103: Based on the early warning results and combined with the operation and maintenance resource information, generate and dispatch a maintenance plan.

[0039] This step, based on the warning information output in step S102, triggers the maintenance process and determines the priority of this maintenance task. Maintenance resource information refers to various resource data related to the execution of the maintenance task, including the real-time location, skills and qualifications, and on-duty status of maintenance personnel, spare parts inventory information, and the accessibility of routes between sites. This information collectively constitutes the constraints and decision-making basis for generating the maintenance plan.

[0040] This step, based on the faulty equipment, fault type, and warning level indicated by the warning results, and combined with operation and maintenance resource information, performs comprehensive matching and optimization calculations from multiple dimensions such as personnel accessibility, material availability, and route drivability. The final output includes a maintenance execution plan that includes the assignment of maintenance personnel, a list of spare parts, and feasible routes.

[0041] The core of this step lies in automatically generating and dispatching maintenance plans by combining operation and maintenance resource information. This achieves rapid response and precise connection from early warning to maintenance, avoiding the lag and uncertainty of manual dispatch in the traditional model. It enables maintenance tasks to be rationally planned and dispatched for execution as soon as an early warning is generated, thereby effectively shortening maintenance response time and improving operation and maintenance efficiency.

[0042] The multi-scenario maintenance and operation method for new energy power plants provided in this embodiment overcomes the shortcomings of general models that cannot adapt to the differentiated operation and maintenance needs of multiple scenarios by training and generating dedicated fault early warning models and health assessment models for new energy power plants of different scenario types. Each scenario-specific model can accurately identify the evolution patterns of equipment faults and environmental risk characteristics under different geographical environments, effectively reducing the false alarm rate and missed alarm rate of early warnings, and providing precise guidance for maintenance work. At the same time, this invention achieves differentiated and accurate early warning for different scenarios by acquiring the scenario type information of the power plant and calling the corresponding dedicated model for fault identification and health assessment. Based on this, maintenance plans are automatically generated and dispatched according to the early warning results and operation and maintenance resource information, opening up the key link from early warning to maintenance task dispatch, realizing closed-loop management of maintenance and operation, avoiding the uncertainty and lag of manual task dispatch, and significantly improving maintenance response speed and operation and maintenance efficiency. Furthermore, the maintenance plans generated based on accurate early warnings can provide clear task guidance for maintenance personnel, reduce the time for on-site inspection and plan formulation, and effectively reduce the risk of equipment failure expansion and power generation loss due to untimely maintenance.

[0043] This embodiment provides a method for the inspection and maintenance of new energy power stations in multiple scenarios. Figure 2 This is a flowchart of a multi-scenario maintenance and operation method for new energy power stations according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: For new energy power stations of different scenario types, respectively train and generate fault early warning models and health assessment models corresponding to different scenario types.

[0044] The different scenario types include four categories: mountain wind power, mountain photovoltaic power in coal mining subsidence areas, conventional mountain photovoltaic power, and flatland agricultural photovoltaic power.

[0045] The fault warning model and health assessment model are trained and generated according to the following steps: Step S2011: Collect datasets for each scenario. Each dataset includes electrical parameters, environmental parameters, scenario-specific additional data, and historical operation and maintenance data.

[0046] For example, a dataset was collected from one mountain wind farm and five photovoltaic power stations in different scenarios.

[0047] In one optional implementation, the electrical parameters include at least one of voltage, current, power, winding temperature, and partial discharge value; the environmental parameters include at least one of ambient temperature and humidity, water depth, SF6 concentration, mountain subsidence monitoring data, illumination data, and wind speed data; the additional scenario data includes at least one of geological data of the coal mining area, mountain shading data, and agricultural-solar complementary farmland operation time sequence data; and the historical operation and maintenance data includes at least one of historical fault records, maintenance and replacement parts records, and historical downtime data.

[0048] Data sets were categorized and summarized into three types based on the power station: mountain wind power dataset, coal mining subsidence area mountain photovoltaic dataset, ordinary mountain photovoltaic dataset, and flatland agricultural-photovoltaic complementary dataset. Missing values ​​and anomalous jump data were removed, and the data were normalized.

[0049] In step S2012, the fault warning model and health assessment model for each scenario are generated by training the datasets for each scenario through machine learning algorithms.

[0050] A machine learning model combining time series prediction and classification was selected as the base model. The model input features X include: real-time electrical quantities of a single device, on-site environmental data, geographical attributes of the site, seasonal climate data, and surrounding environmental disturbance data (including coal mining vibration, tree and crop obstruction); the training labels Y include: historical fault types of the equipment, fault occurrence time, and equipment aging health score (0~100 points, the lower the score, the higher the fault risk).

[0051] Each dataset is split into training, validation, and test sets in a 7:2:1 ratio. Each dataset is trained independently for four scenarios, with a dedicated model trained for each type of site. Data sets are not mixed. The validation set is used to iteratively optimize hyperparameters, and the test set is used to verify the model accuracy. Once the model accuracy is greater than 92%, the trained models are deployed separately.

[0052] The model output includes: fault warning output (fault category and fault occurrence probability, where fault category includes short circuit, overheating and gas leakage) and equipment health assessment output (0~100 health score, divided into three levels: 80~100 is normal, 50~79 is sub-healthy, and less than 50 is a high-risk fault warning).

[0053] During daily operation, new faults and new operational data generated during each maintenance loop are added to the sample library monthly, and the model is automatically iterated, retrained, and optimized quarterly.

[0054] In one optional implementation, a complete high-parameter model is deployed at the central control center, and the edge computing module at the field station trims network weights and compresses parameters to obtain a lightweight sub-model adapted to the edge hardware computing power. The lightweight model is then distributed to the edge computing units of the corresponding field stations, while the complete high-precision model is deployed to the central control center's management platform.

[0055] In one optional implementation, the skill tags, qualification information, spare parts inventory information, geographical information and road data of all maintenance personnel in the cluster are entered into the central control platform to complete the system initialization configuration.

[0056] In one alternative implementation, new fault data and new operational data generated during each maintenance loop are added to the sample library monthly, and the model is automatically iterated, retrained, and optimized quarterly.

[0057] Step S202: Obtain real-time monitoring data and site scenario type information for each site; call the corresponding fault warning model and health assessment model according to the scenario type information; perform fault identification and health assessment on the real-time monitoring data; and issue a warning based on the fault identification results and health assessment results.

[0058] Specifically, step S202 above includes the following steps: Step S2021: Obtain real-time monitoring data and site scenario type information for each site.

[0059] The intelligent sensing terminal of the container substation collects electrical, environmental and security data in real time. The SCADA system of the station collects the operating data of wind turbines, photovoltaic strings and inverters in real time. All data is transmitted to the edge computing module at the station end for preprocessing in real time.

[0060] Step S2022: Based on the scenario type information, select the corresponding fault warning model and health assessment model from multiple preset scenario-specific models.

[0061] Acquire real-time monitoring data and site scenario type information for each site, and select the corresponding fault warning model and health assessment model from multiple preset scenario-specific models based on the scenario type information.

[0062] Step S2023: Input the real-time monitoring data into the fault early warning model and the health assessment model, and perform inference calculations on the input data through the fault early warning model and the health assessment model respectively to obtain the fault identification result and the health assessment result.

[0063] The edge computing module at the site uses a lightweight model deployed locally to perform in-situ reasoning on the preprocessed data, identify abnormal equipment states and environmental risks. If an emergency fault (including equipment short circuit, fire and illegal intrusion) is detected, it immediately triggers local linkage control (including security linkage control, fire linkage control and emergency shutdown control of faulty equipment), and uploads the alarm signal to the central control center.

[0064] Step S2024: Based on the fault identification results and health assessment results, a comprehensive judgment is made and a warning message is output. The warning includes: Step a1: Compress the trained fault warning model and health assessment model into a lightweight model and deploy it at the field station. The field station uses the lightweight model to perform local inference on the real-time monitoring data. When an emergency fault is detected, it triggers local linkage control and uploads alarm signals. Step a2: The fault early warning model and health assessment model generated during training are deployed as high-precision models in the central control center. The central control center uses the high-precision models to analyze the uploaded data and realize hierarchical early warning.

[0065] For example, tiered early warning systems can include general, important, and urgent alerts. Corresponding tiered alarm information can also be generated for security and environmental anomalies at containerized substations.

[0066] Step S203: Based on the early warning results and combined with operation and maintenance resource information, generate and dispatch a maintenance plan.

[0067] The maintenance plan includes: Step b1: Determine the fault level and assign work according to the fault level.

[0068] After receiving an alarm signal, the central control platform automatically matches the faulty equipment, fault type and corresponding site scenario, and retrieves the equipment's historical maintenance records, standardized operating instructions and spare parts demand list.

[0069] The fault level is determined based on the early warning results, and work is assigned according to the fault level. Emergency faults (including fire and short circuit) are assigned immediately, important faults (equipment sub-health and high risk) are handled within a time limit, and general faults (minor abnormalities) are included in the periodic maintenance plan.

[0070] Step b2: Select matching maintenance personnel based on their real-time location, skills and qualifications, and on-duty status.

[0071] Based on the real-time location, skills and qualifications, and on-duty status of maintenance personnel, on-duty personnel with the corresponding equipment maintenance qualifications and skills tags are selected. Commuting mileage is calculated by combining the personnel's real-time GPS location with the geographical coordinates of the site, and feasible routes are selected based on the route traffic status.

[0072] Step b3: Match spare parts based on spare parts inventory information and route access status.

[0073] Based on spare parts inventory information and route access status, the system matches spare parts, checks the spare parts inventory of the nearest site, and automatically allocates spare parts from other sites within the cluster when parts are in short supply.

[0074] Step b4: Perform multi-constraint weighted calculation based on fault level, maintenance personnel screening results, spare parts matching results, and route access status to output the optimal maintenance plan.

[0075] A multi-constraint weighted calculation is performed based on the fault level, maintenance personnel selection results, spare parts matching results, and route accessibility. Weights are set for fault level, distance, personnel skill matching, spare parts inventory, road accessibility, and personnel on-duty status. The optimal maintenance plan is output based on the highest overall weighted score. The optimal maintenance plan includes maintenance personnel assignment, maintenance work order content, spare parts list, optimal maintenance route, work safety measures, and planned processing time.

[0076] For example, the following weightings are used to generate a solution step by step: fault level weight (maximum 40%), distance weight (25%), personnel skill matching weight (15%), spare parts inventory weight (10%), road accessibility weight (5%), and personnel on-duty status weight (5%). The first step is to classify faults by severity: emergency faults (fire / short circuit, dispatch immediately), major faults (equipment in sub-health or high risk, handle within 24 hours), and minor faults (minor abnormalities, included in periodic maintenance). The second step is to screen on-the-job personnel who have the corresponding equipment maintenance qualifications and whose skill tags match; The third step is to calculate the commuting mileage by combining the real-time GPS location of personnel and the geographical coordinates of the station, and to screen feasible routes by combining data on mountain road restrictions and rain and snow closures. The fourth step is to check the inventory of spare parts at the nearest site, and if a part is missing, automatically allocate spare parts from other power stations within the cluster. The fifth step involves automatically outputting the combination with the highest weighted score: maintenance personnel, work order details, spare parts list, optimal driving route, safety measures, and planned working hours.

[0077] In one optional implementation, for cluster maintenance needs across multiple stations, the cross-station resource collaborative management module coordinates and allocates idle maintenance personnel and spare parts from surrounding stations to achieve optimal resource utilization, while dispatching maintenance work orders to the mobile maintenance terminals of the corresponding maintenance personnel.

[0078] In one optional implementation, the method of this embodiment further includes: after the maintenance is completed, the central control center verifies the maintenance effect, confirms that the fault has been eliminated, and archives the maintenance task in a closed loop; the fault data and maintenance data are used as new samples to regularly retrain and iteratively optimize the fault early warning model and health assessment model, continuously improve the model's early warning accuracy, and reduce the false alarm rate and false alarm rate.

[0079] In one optional implementation, the method of this embodiment further includes: on-site safety management and operation execution. After the maintenance personnel arrive at the container substation, they undergo facial recognition identity verification via a mobile maintenance terminal. The central control platform verifies the work order information and personnel qualifications. After verification, the platform remotely authorizes the opening of the substation access control. At the same time, the personnel's entry and exit times are automatically recorded, and the entire process is recorded and stored via video surveillance.

[0080] During maintenance, the perimeter security and environmental monitoring equipment of the container substation operate in real time. If unauthorized personnel intrusion or environmental anomalies are detected, a local audible and visual alarm is immediately triggered, and the signal is uploaded to the central control center. The central control center then notifies the on-site maintenance personnel immediately and simultaneously activates the emergency response plan.

[0081] Maintenance personnel use mobile maintenance terminals to complete maintenance tasks according to work order steps, and upload maintenance data, on-site photos and operation videos in real time. For complex faults, they can obtain remote technical guidance from experts in the central control center through the remote expert assistance module.

[0082] In one optional implementation, the method of this embodiment further includes: verification of maintenance results and closed-loop archiving of work orders. After the maintenance work is completed, the maintenance personnel submit the maintenance results through a mobile maintenance terminal. The central control platform verifies the maintenance effect by combining real-time equipment operation data and information transmitted from the field. After confirming that the fault has been eliminated, the work order is archived in a closed loop.

[0083] In an optional implementation, the method of this embodiment further includes: storing fault data and maintenance data in a database. The operation and maintenance data lifecycle management module stores the fault data, maintenance data, and handling results in the database as new sample data for model training.

[0084] In one optional implementation, the method of this embodiment further includes: generating operation and maintenance analysis reports. Based on full-volume operation and maintenance data, the centralized control center platform periodically generates operation and maintenance analysis reports for each site, optimizes maintenance strategies and inspection cycles, and achieves preventative operation and maintenance throughout the entire lifecycle.

[0085] This embodiment also provides a multi-scenario new energy power station maintenance and operation system, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the system / device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0086] This embodiment provides a multi-scenario new energy power station maintenance and operation system, such as Figure 3 As shown, it includes: The model management module 301 is deployed in the central control center and is used to store fault warning models and health assessment models corresponding to different scenario types. The edge computing module 302 at the site is deployed at the site to acquire real-time monitoring data and site scenario type information of each site. Based on the scenario type information, it calls the corresponding fault warning model and health assessment model to perform fault identification and health assessment on the real-time monitoring data, and issues warnings based on the fault identification results and health assessment results. The task scheduling module 303, deployed in the central control center, is used to generate and dispatch maintenance plans based on early warning results and operation and maintenance resource information.

[0087] In one alternative implementation, the control center is equipped with a control center management platform, including a model management module and a task scheduling module.

[0088] The model management module is used to store fault warning models and health assessment models corresponding to different scenario types.

[0089] The task scheduling module is used to generate and dispatch maintenance plans based on the early warning results and maintenance resource information. For example, based on the early warning results, combined with the site's geographical information, maintenance personnel skill tags, real-time location, spare parts inventory, and route access status, it intelligently generates maintenance work orders, optimal maintenance routes, and personnel schedules, completing work order dispatch, full-process tracking, closed-loop management, and verification.

[0090] In one optional implementation, the edge computing module at the site is deployed at the site, for example, at the site of one wind farm and five photovoltaic power stations, to acquire real-time monitoring data and site scenario type information of each site, call the corresponding fault warning model and health assessment model according to the scenario type information, perform fault identification and health assessment on the real-time monitoring data, and issue warnings based on the fault identification results and health assessment results.

[0091] In one optional implementation, the station-side edge computing module includes: The data acquisition and preprocessing unit is used to interface with the site's SCADA system, environmental monitoring equipment, and security equipment to complete the acquisition, cleaning, and normalization of multi-source heterogeneous data. The lightweight early warning unit is used to acquire real-time monitoring data and preprocess it. Then, it performs on-site inference on the preprocessed data, and triggers local linkage control and uploads alarm signals when an emergency fault is detected. This enables rapid on-site identification of equipment abnormalities and millisecond-level alarms without waiting for instructions from the central control center, which greatly improves the speed of emergency response. The local linkage control unit is used to receive commands from the central control center or alarm signals from the edge side to realize local linkage control of security, fire protection, and access control of the container substation, as well as emergency shutdown operation of faulty equipment. The on-site communication support unit is used to provide wireless communication coverage for the maintenance site. LoRa relay nodes are added for mountainous areas with no signal to ensure real-time data interaction between mobile maintenance terminals and the central control center and edge computing modules.

[0092] In one optional implementation, the system further includes: an intelligent sensing terminal, deployed on the site equipment side, used to collect real-time monitoring data and transmit it to the site-end edge computing module; the intelligent sensing terminal includes: The electrical status sensing unit consists of a voltage sensor, a current sensor, a power sensor, a partial discharge sensor, and an equipment temperature sensor. It is used to collect real-time electrical operating status data of high and low voltage equipment, switchgear, transformer substations, and inverters in the substation. The environmental sensing unit consists of temperature and humidity sensors, water immersion sensors, SF6 gas leak sensors, and smoke sensors, and is used to monitor the environmental conditions and safety risks inside the container in real time. The perimeter security and access control unit consists of a face recognition access control host and access controller inside the box, as well as a perimeter intrusion radar, AI video surveillance camera and electronic fence. It is used to realize the intrusion detection, identity verification of maintenance personnel and full-process control of access permissions of the substation perimeter. Further, it may include a fire-fighting linkage unit, which consists of an aerosol fire extinguishing device and a fire alarm controller, used to link with the environmental sensing unit, perimeter security and access control unit to realize automatic fire alarm and emergency response.

[0093] In one optional implementation, the centralized control center platform also includes a container substation full-dimensional control module, which is used to receive substation operation data, security data and environmental data uploaded by the site, and realize remote status monitoring, personnel access management, security linkage and emergency response command issuance.

[0094] In one optional implementation, the centralized control center management platform also includes a full lifecycle management module for operation and maintenance data, which stores the operation data, fault data, maintenance data and personnel operation data of the entire site, providing data support for model optimization and operation and maintenance strategy adjustment.

[0095] In one optional implementation, the central control center platform also includes a cross-site resource collaborative management module, which is used to manage the maintenance personnel, tools and spare parts of all sites within the cluster in a unified manner, so as to realize the unified scheduling, sharing and optimal configuration of resources across sites.

[0096] In one alternative implementation, the system further includes: Mobile maintenance terminal, carried by maintenance personnel, includes: The work order receiving and execution module is used to receive maintenance work orders dispatched by the central control center, display maintenance task details, standardized operation instructions, equipment drawings and historical fault records, and support step check-in, data upload, problem feedback and result submission during the maintenance process; The identity verification and access control module supports facial recognition identity verification for maintenance personnel, as well as online submission of access control applications for temporary maintenance tasks, which are then remotely approved by the central control center before access is granted. The remote expert assistance module is used to support real-time video transmission from the maintenance site, enabling remote video calls with experts in the central control center to obtain remote operation guidance. The offline operation module is used in areas with poor network signal, such as mountainous areas and coal mining subsidence areas, to support offline caching of work orders, work instructions and equipment drawings. The data is automatically synchronized when the network is restored after the operation is completed.

[0097] In one alternative implementation, some structures or steps in the above implementations may be replaced by the following: Regarding alternative deployment methods for edge computing units: merge the edge computing functions of multiple adjacent sites and deploy them in a central site. This is suitable for cluster scenarios where the sites are close together, thereby reducing hardware deployment costs.

[0098] Regarding alternative communication methods for intelligent sensing terminals: In addition to Ethernet, 4G, or 5G communication methods, LoRa wireless communication technology is used to achieve data transmission between intelligent sensing terminals and edge computing modules at the site, especially in mountainous areas with no signal.

[0099] Regarding alternative forms of mobile maintenance terminals: In addition to handheld smart terminals, mobile maintenance terminals can also use wearable devices such as smart safety helmets or AR glasses to achieve hands-free maintenance operations and remote assistance.

[0100] Regarding alternative model deployment architectures: For small-scale sites, all lightweight models will be deployed in the central control center, with only data acquisition and command execution functions retained at the site, in order to simplify on-site hardware configuration.

[0101] The multi-scenario maintenance and operation method and system for new energy power stations provided by this invention have the following beneficial effects: This invention constructs dedicated fault warning and health assessment models for four typical operating scenarios, solving the core problems of poor scenario adaptability and high false alarm and missed alarm rates of existing general models. Through a two-level early warning architecture that combines edge-side local inference with in-depth analysis at the central control center, the accuracy of equipment fault warning can be improved by more than 30%, and the missed alarm rate can be reduced by more than 80%, realizing the transformation from "passive emergency repair" to "preventive maintenance", significantly reducing unplanned equipment downtime and increasing power generation at the plant.

[0102] This invention deeply integrates electrical monitoring, environmental sensing, perimeter security, access control, fire protection linkage, and the entire maintenance process of containerized substations, solving the problem of blind spots in security management in substations without walls or on-site personnel. Through the identification of maintenance personnel and remote authorization access control, the entire process of maintenance entry is traceable. Through the linkage between security and maintenance tasks, safety risks can be prevented in real time during maintenance, reducing the substation safety incident rate by more than 90%, and comprehensively ensuring the safety of maintenance operations and station equipment.

[0103] This invention constructs a closed-loop management system for the entire maintenance and repair process, consisting of "early warning-dispatch-execution-verification-optimization," which solves the problems of existing technologies relying on manual experience for dispatching orders, non-standardized maintenance processes, and untimely handling. Through the intelligent maintenance task scheduling module, maintenance response time can be reduced by more than 50%, and maintenance operation time can be reduced by more than 30%. Through mobile maintenance terminals, standardized control of the maintenance process and remote expert assistance are realized, significantly reducing the maintenance rework rate and improving the quality of maintenance operations.

[0104] This invention, through a cross-site resource collaborative management module, coordinates and schedules maintenance personnel, tools, and spare parts for one wind farm and five photovoltaic power stations within a cluster, solving the problems of untimely resource allocation and high personnel travel costs at remote sites. Through cross-site collaborative maintenance, it can reduce the number of maintenance personnel by more than 30%, reduce the safety stock of spare parts by more than 25%, and reduce equipment failure losses through preventive maintenance, thereby reducing the total life cycle maintenance cost by more than 20%.

[0105] The mobile maintenance terminal of this invention supports offline operation mode, adapts to the maintenance operation needs of areas with poor network signals such as mountainous areas and coal mining subsidence areas, and ensures that maintenance data is not lost and the operation process is not interrupted. Through the storage of full operation and maintenance data and iterative optimization of models, it realizes continuous self-improvement of operation and maintenance capabilities and forms a virtuous cycle of operation and maintenance.

[0106] Figure 4 This is a detailed architecture deployment diagram of the multi-scenario new energy power station maintenance and operation system provided in this embodiment of the invention. (See diagram below.) Figure 4 As shown, the system includes a central control platform, a field edge computing module, intelligent sensing terminals, a communication network, and mobile maintenance terminals.

[0107] The centralized control center platform includes a model management module and a task scheduling module; the field edge computing module includes a lightweight early warning unit.

[0108] The intelligent sensing terminal is deployed on the equipment side of the site to collect real-time monitoring data and transmit it to the edge computing module at the site through the communication network; the edge computing module at the site interacts bidirectionally with the control platform of the central control center through the communication network; the mobile maintenance terminal receives maintenance work orders dispatched by the control platform of the central control center through the communication network and provides feedback on maintenance results.

[0109] The multi-scenario new energy power station maintenance and operation system provided in this embodiment of the invention can execute the multi-scenario new energy power station maintenance and operation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments above, and will not be repeated here.

[0110] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0111] The following is a detailed reference. Figure 5 The diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from memory 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device. The processor 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0112] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0113] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a memory 508, or installed from a ROM 502. When the computer program is executed by the processor 501, it performs the functions defined in the multi-scenario new energy power station maintenance and repair method of the embodiments of the present invention.

[0114] Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0115] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the multi-scenario maintenance and repair method for new energy power stations shown in the above embodiments is implemented.

[0116] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0117] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for overhauling and maintaining new energy power stations in multiple scenarios, characterized in that, The method includes: For new energy power stations with different scenario types, fault early warning models and health assessment models corresponding to different scenario types are trained and generated respectively. Acquire real-time monitoring data and site scenario type information of each site, call the corresponding fault early warning model and health assessment model according to the scenario type information, perform fault identification and health assessment on the real-time monitoring data, and issue early warnings based on the fault identification results and health assessment results; Based on the early warning results and combined with operation and maintenance resource information, a maintenance plan is generated and dispatched.

2. The multi-scenario new energy power station maintenance and repair method according to claim 1, characterized in that, The different scenario types include four categories: mountain wind power, mountain photovoltaic power in coal mining subsidence areas, conventional mountain photovoltaic power, and flatland agricultural photovoltaic power.

3. The multi-scenario maintenance and operation method for new energy power stations according to claim 1, characterized in that, The fault early warning model and health assessment model are trained and generated according to the following method: Data sets were collected for each scenario, and each type of dataset included electrical parameters, environmental parameters, scenario-specific additional data, and historical operation and maintenance data. Using datasets from different scenarios, machine learning algorithms are used to train and generate corresponding fault warning models and health assessment models.

4. The multi-scenario new energy power station maintenance and repair method according to claim 1, characterized in that, The maintenance plan includes: Determine the fault level and assign tasks according to the fault level. Based on the real-time location, skills and qualifications, and on-duty status of maintenance personnel, suitable maintenance personnel are selected. Match spare parts based on spare parts inventory information and route access status; The optimal maintenance plan is output by performing multi-constraint weighted calculations based on the fault level, the screening results of maintenance personnel, the matching results of spare parts and components, and the road access status.

5. The multi-scenario new energy power station maintenance and repair method according to claim 1, characterized in that, The warning includes: The fault early warning model and health assessment model generated during training are compressed into a lightweight model and deployed at the field station. The field station uses the lightweight model to perform local inference on real-time monitoring data, and when an emergency fault is detected, it triggers local linkage control and uploads alarm signals. The fault early warning model and health assessment model generated during training are deployed as high-precision models in the central control center. The central control center uses the high-precision models to analyze the uploaded data and realize hierarchical early warning.

6. The multi-scenario maintenance and operation method for new energy power stations according to claim 5, characterized in that, Also includes: After the maintenance is completed, the central control center verifies the maintenance effect and archives the maintenance task in a closed loop after confirming that the fault has been eliminated. Using the fault data and maintenance data as new samples, the fault early warning model and health assessment model are periodically retrained and iteratively optimized.

7. A multi-scenario maintenance and operation system for new energy power stations, characterized in that, The system includes: The model management module, deployed in the central control center, is used to store fault warning models and health assessment models corresponding to different scenario types. The edge computing module at the site is deployed at the site and is used to acquire real-time monitoring data and site scenario type information of each site. Based on the scenario type information, it calls the corresponding fault warning model and health assessment model to perform fault identification and health assessment on the real-time monitoring data, and issues warnings based on the fault identification results and health assessment results. The task scheduling module, deployed in the central control center, is used to generate and dispatch maintenance plans based on early warning results and operation and maintenance resource information.

8. The multi-scenario new energy power station maintenance and operation system according to claim 7, characterized in that, The station-end edge computing module includes: The lightweight early warning unit is used to acquire the real-time monitoring data, preprocess it, perform on-site reasoning on the preprocessed data, and trigger local linkage control and upload alarm signals when an emergency fault is detected.

9. The multi-scenario new energy power station maintenance and operation system according to claim 7, characterized in that, Also includes: The intelligent sensing terminal is deployed on the equipment side of the site to collect the real-time monitoring data and transmit it to the edge computing module at the site. The intelligent sensing terminal includes an electrical status sensing unit, an environmental sensing unit, and a perimeter security and access control unit.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which are used to cause the computer to execute the multi-scenario new energy power station maintenance and repair method according to any one of claims 1 to 6.

Images