Photovoltaic power station real-time monitoring management system and method based on cloud platform

Abstract

The invention relates to a photovoltaic power station real-time monitoring management system and method based on a cloud platform, and belongs to the field of photovoltaic power station intelligent monitoring. The system adopts a three-layer architecture of an application layer, a platform layer and an edge layer deployed at a cloud end, the edge layer is composed of distributed edge nodes, data are acquired through a data acquisition interface, and local preprocessing is executed; the platform layer configures a data storage and machine learning model, and performs generating capacity prediction and equipment health degree evaluation; and the application layer provides a visual monitoring interface and an operation and maintenance management terminal, and supports multi-dimensional data display and alarm information pushing. And the edge layer and the platform layer adopt a calculation task dynamic allocation mechanism, so that the cloud load is reduced while the real-time performance is ensured. The full-life-cycle monitoring of the photovoltaic power station is realized through a pure software architecture, the system has the advantages of elastic expansion, cross-platform compatibility and remote maintenance, and the deployment efficiency and the operation and maintenance response speed of the monitoring system are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent monitoring of photovoltaic power stations, and relates to a remote monitoring and management system and method for photovoltaic power stations based on a cloud computing architecture. Background Art

[0002] Traditional photovoltaic power station monitoring systems generally adopt a local hardware deployment mode, and their core architecture relies on dedicated data acquisition devices and local server clusters. This hardware-binding mode results in high system deployment costs. Especially when accessing photovoltaic devices from different manufacturers, additional conversion modules often need to be configured due to incompatible communication protocols. More prominently, the expansion ability of existing systems is limited by the initial hardware resource configuration. When the scale of the power station expands or the device types are updated, complex hardware transformation and upgrading are often required, seriously restricting the flexible operation and maintenance needs of photovoltaic power stations.

[0003] Current mainstream photovoltaic power station operation and maintenance solutions have obvious shortcomings in data processing and analysis. Limited by the computing power of local servers, the operation and maintenance system is difficult to achieve real-time parsing and in-depth mining of massive device data, and often uses simple threshold comparison for anomaly judgment, unable to effectively identify potential problems such as device performance degradation and string mismatch. At the same time, the data processing process under the traditional architecture shows rigid characteristics and cannot dynamically adjust the computing resource allocation according to the business load, resulting in problems such as data loss or response delay in high-concurrency scenarios.

[0004] Although some technical solutions have tried to introduce cloud computing resources in recent years, most improvement solutions still have not completely got rid of hardware dependence. For example, some edge computing solutions still need to deploy physical gateway devices for data transfer, essentially forming a "hardware + cloud" hybrid architecture, which not only increases the system complexity but also introduces new single-point failure risks. And although the operation and maintenance solutions based on pure cloud reduce the local hardware requirements, due to the lack of an efficient edge-cloud coordination mechanism, operation risks such as control instruction loss and device status misjudgment are likely to occur during network fluctuations. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a real-time monitoring and management system and method for photovoltaic power stations based on a cloud platform. Through the collaborative work of the software-defined application layer, platform layer, and edge layer, a closed-loop system for photovoltaic device data acquisition, analysis, and visualization management is formed. At the same time, the system adopts a distributed architecture design, and all functional modules are deployed on cloud infrastructure to achieve virtualization of hardware resources and dynamic scheduling of computing tasks, thereby improving the operation and maintenance efficiency and system reliability of photovoltaic power stations.

[0006] To achieve the above object, on the one hand, the present invention provides a real-time monitoring and management system for photovoltaic power stations based on a cloud platform, and the system includes an edge layer, a platform layer, and an application layer deployed on the cloud.

[0007] The edge layer is composed of distributed virtual nodes. Each node obtains the operation data of photovoltaic devices through a data acquisition interface, performs data preprocessing, and then uploads it to the platform layer. The platform layer stores the operation data of photovoltaic devices in a time-series database, performs power generation prediction and equipment fault diagnosis through a machine learning model, and generates equipment maintenance strategies based on the diagnosis results. The application layer provides a visual monitoring interface and an operation and maintenance management terminal, dynamically displays the status information of the photovoltaic power station according to the stored data in the platform layer, generates operation and maintenance work orders according to the equipment maintenance strategies generated by the platform layer, and pushes multi-level alarm notifications.

[0008] Further, in the platform layer, the machine learning model includes a power generation prediction module, a fault diagnosis module, and an equipment health assessment module. The power generation prediction module generates the power generation power curve of the photovoltaic power station for the next 24 hours by analyzing historical power generation data and environmental parameters. The fault diagnosis module uses a clustering algorithm to analyze the operation data of photovoltaic strings, identifies abnormal deviations in the current-voltage characteristic curve, and locates the position of the performance degradation or faulty string. The equipment health assessment module generates maintenance priority suggestions based on the diagnosis results of the fault diagnosis module, combined with the equipment operation duration and environmental corrosion parameters.

[0009] Further, a two-way WebSocket communication connection is maintained between the application layer and the platform layer to enable real-time refreshing of the operation status information of the photovoltaic power station displayed on the visual monitoring interface.

[0010] The visual monitoring interface is constructed through a Web page that can be accessed across terminals. The visual monitoring interface displays a dynamically rendered 3D efficiency heat map of the photovoltaic array, a comparison dashboard of key indicators of multiple power stations, and the status of cluster devices based on a GIS map.

[0011] Further, a hierarchical security mechanism is adopted for data transmission between the edge layer, the platform layer, and the application layer, including enabling the TLS1.3 encryption protocol and message signature verification for data transmission from the edge layer to the platform layer to prevent data tampering; implementing OAuth 2.0 token authentication for service calls from the platform layer to the application layer to control API access permissions.

[0012] The control instructions between the edge layer, the platform layer, and the application layer are transmitted through a dedicated channel, and a response timeout threshold is set. When the instruction response time exceeds the set threshold, the system automatically terminates the current operation sequence and performs a status rollback to avoid equipment damage caused by instruction delay.

[0013] Furthermore, the virtual nodes constituting the edge layer run on the Kubernetes edge computing cluster, supporting automatic node scaling, shrinking, and fault migration; the platform layer deploys analysis services in the form of microservices through a containerized PaaS platform; the front end of the application layer uses WebAssembly technology to achieve cross-terminal rendering.

[0014] On the other hand, the present invention provides a real-time monitoring and management method for a photovoltaic power station based on a cloud platform, and the method includes:

[0015] First, construct a real-time monitoring and management system for a photovoltaic power station based on a cloud platform, and the system includes an edge layer, a platform layer, and an application layer deployed on the cloud; the edge layer is composed of distributed virtual nodes; the platform layer is configured with a distributed time series database and a machine learning model; the application layer is configured with a visual monitoring interface and an operation and maintenance management terminal;

[0016] Second, each virtual node in the edge layer obtains data through a data acquisition interface, and the acquired data includes current, voltage, temperature, and weather forecast, and a differential acquisition strategy is configured for different data types; the acquired original data stream is processed by sliding window filtering, and after the processing is completed, it is compressed and encapsulated into a message queue and uploaded to the platform layer. At the same time, the edge layer retains a backup of the original data for fault tracing;

[0017] Then, the platform layer writes the time series data into the distributed time series database, stores the data in slices according to the geographical partition of the photovoltaic power station, and sets a retention period for each slice; in addition, the platform layer periodically calls the machine learning model for power generation prediction and fault diagnosis, and generates a device maintenance strategy based on the diagnosis result;

[0018] Finally, the application layer displays a 3D efficiency heat map of the photovoltaic array, a comparison dashboard of key indicators of multiple power stations, and the status of cluster devices based on a GIS map in real time in the visual monitoring interface according to the data stored in the platform layer; at the same time, according to the device maintenance strategy generated by the platform layer, the application layer pushes multi-level alarm notifications through the operation and maintenance management terminal for device anomaly alarms.

[0019] Furthermore, the configuration of differential acquisition strategies for different data types includes: for current and voltage parameters, sampling is performed at a period of 1 second; for temperature parameters, acquisition is performed at a period of 5 seconds; the frequency deviation of data acquisition is within ±50 ms;

[0020] The sliding window filtering process for the collected original data stream includes: for the transient fluctuation characteristics of current and voltage signals, an adaptive window width moving average algorithm is used for filtering. The initial window width is set to 20 sampling points. When 3 consecutive sampling values exceed the ±5% fluctuation range of the nominal value of the PV power station components, the window width is dynamically reduced to 10 sampling points to improve the filtering response speed;

[0021] During the process of the edge layer uploading data to the platform layer, if the network delay between the edge layer and the platform layer is monitored to exceed the network delay threshold, the edge layer automatically activates the local decision-making mode, executes the inverter output power limit protection, and temporarily stores the operation log in the local non-volatile memory; after the network communication status is restored, the edge layer starts the data synchronization verification process, and uses the log comparison algorithm based on the hash chain to perform consistency verification on the operation log cached locally and the platform layer database; for the instruction records with version conflicts, the version stored in the platform layer is preferred, and a differential audit report is generated.

[0022] Further, the platform layer regularly calls the machine learning model for power generation prediction and fault diagnosis, including: performing power generation prediction through the power generation prediction module based on the LSTM network. The input parameters of the power generation prediction module include the historical 72-hour power generation data and the future 24-hour weather forecast information, and the relative error of the output result of the power generation prediction module is controlled within the range of ±5%; performing fault diagnosis through the fault diagnosis module based on the clustering algorithm, using the current-voltage characteristic curve of the PV string as the input data. When the fault diagnosis module detects that the current-voltage characteristic curve of the PV string deviates from the reference value by more than 15%, it determines that a fault has occurred, locates the fault location, and triggers the health assessment process to generate a maintenance priority matrix by combining the cumulative operation duration of the equipment and the environmental corrosion factor;

[0023] The platform layer also synchronously runs a rule engine to parse the data stream in real time. When the radiator temperature parameter of the inverter exceeds 75°C in 3 consecutive sampling cycles, an abnormal event record is generated and associated with the unique identification code of the corresponding inverter;

[0024] The refresh rules for the visualization monitoring interface to display the operation status information of the PV power station include: the three-dimensional efficiency heat map of the PV array updates the color gradient distribution with a period of 3 seconds; the device status indicators on the GIS map are dynamically updated according to the received heartbeat signals. When the signal is not received for 2 consecutive heartbeat cycles, it is marked as a warning state, and when the signal is not received for more than 5 heartbeat cycles, it is marked as an offline state.

[0025] Furthermore, this method is also applicable to the scenario of multi-power station cluster monitoring. When the number of photovoltaic power stations accessed by the management system exceeds 50, the horizontal expansion mechanism of the cloud computing center server is triggered, including: each server monitors its own resource utilization rate in real time. When the CPU occupancy rate of all servers continues to be higher than 80% for 5 minutes, the cloud computing center adds servers, and at the same time activates the container orchestration system at the edge layer, so that the newly added servers can complete service registration and access the service flow within 120 seconds.

[0026] The beneficial effects of the present invention are as follows: The present invention proposes a real-time monitoring and management system and method for photovoltaic power stations based on a cloud platform. Through a software-defined three-layer collaborative architecture and a full-cloud deployment mode, it effectively solves the core problems of traditional monitoring systems such as hardware dependence, insufficient scalability, and limited data processing capabilities, and realizes a comprehensive improvement in the operation and maintenance efficiency of photovoltaic power stations.

[0027] Through a fully virtualized edge layer architecture, the present invention uses distributed virtual nodes to replace traditional dedicated hardware devices, breaking through the protocol compatibility barrier of multi-vendor devices while reducing deployment costs, and significantly improving the access ability of heterogeneous devices.

[0028] Based on the machine learning model at the platform layer, the present invention realizes equipment fault diagnosis and equipment health assessment. It can not only accurately identify potential risks such as equipment performance degradation, but also dynamically generate maintenance priority strategies, greatly improving the accuracy and predictability of fault diagnosis.

[0029] Through a hierarchical security mechanism and an instruction response control strategy, combined with a local decision-making mode and a data synchronization verification mechanism in the network fluctuation scenario, the present invention effectively avoids the operation risks of traditional cloud solutions in the case of network instability while ensuring data integrity, and significantly enhances the system robustness.

[0030] Based on the horizontal expansion mechanism, the present invention can dynamically adjust the computing resource configuration of the cloud computing center according to the business load, achieve seamless horizontal expansion, and meet the resource requirements of photovoltaic power station cluster monitoring.

[0031] Through a cross-terminal visual interface and real-time interaction design, the present invention provides multi-dimensional dynamic data display and decision support for operation and maintenance personnel, forms a complete closed-loop management system from data collection to operation and maintenance response, and comprehensively optimizes the operation and maintenance efficiency and operation quality of photovoltaic power stations.

[0032] Other advantages, objectives and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail and preferably below in conjunction with the accompanying drawings, where:

[0034] Figure 1 It is a structural block diagram of a cloud platform management system for real-time monitoring of a photovoltaic power station provided by an embodiment of the present invention;

[0035] Figure 2 It is a data processing flow chart. Specific Embodiments

[0036] The following uses specific specific examples to illustrate the embodiments of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the drawings provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0037] Among them, the drawings are only used for exemplary illustration, showing only schematic diagrams, rather than physical diagrams, and should not be construed as a limitation on the present invention; in order to better illustrate the embodiments of the present invention, some components in the drawings will be omitted, enlarged, or reduced, and do not represent the dimensions of actual products; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0038] In the drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the drawings are only used for exemplary illustration and should not be construed as a limitation on the present invention. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0039] Embodiment 1

[0040] As Figure 1As shown, a cloud platform management system for real-time monitoring of a photovoltaic power station provided by the embodiment includes a software-defined application layer, a platform layer, and an edge layer, all of which are deployed on cloud infrastructure. Among them, the edge layer is composed of distributed virtual nodes, connects to the data sources of the photovoltaic power station through a standardized protocol interface, performs data collection, local data cleaning, and anomaly detection, and uploads the preprocessed data stream to the platform layer. Among them, the collected data includes current, voltage, temperature, and weather forecast, etc. The platform layer integrates data storage and analysis services, uses a time-series database to store device operation data, realizes power generation prediction and device fault diagnosis through a machine learning model, and generates device maintenance strategies based on the diagnosis results. The application layer provides a visual monitoring interface and an operation and maintenance management terminal, dynamically displays the operation status information of the power station, and pushes multi-level alarm notifications through the operation and maintenance management terminal according to the device maintenance strategy output by the platform layer.

[0041] The edge layer is composed of a cluster of distributed virtual nodes deployed in the cloud, and connects to devices such as inverters and meteorological sensors of the photovoltaic power station through a standardized protocol interface. Each virtual node real-time collects device operation parameters, and uses a sliding window algorithm to smooth the data stream, eliminating instantaneous noise interference to ensure that the quality of the original data meets the analysis requirements. In addition, the edge layer also automatically identifies abnormal events such as voltage mutations and temperature overlimits through a pre-set rule library, and generates structured alarm records with timestamps. In the edge layer, the generated data is compressed and encapsulated into a message queue for uploading after cleaning, and is transmitted to the platform layer through a service mesh. At the same time, the edge layer retains a backup of the original data for fault tracing.

[0042] In the platform layer, the embedded machine learning model includes: an LSTM network power generation prediction module based on historical data, a fault diagnosis module that identifies the performance degradation of the string level through a clustering algorithm, and a device health assessment module that calculates the health score by integrating the device operation duration and environmental parameters. Among them, the LSTM network power generation prediction module based on historical data generates a power generation power curve for the next 24 hours by analyzing historical power generation data and environmental parameters. The prediction result of the power generation prediction module is synchronized to the power grid dispatching system to assist in optimizing the power distribution strategy. The fault diagnosis module uses a clustering algorithm to analyze the operation data of the photovoltaic strings, identifies abnormal deviations in the current-voltage characteristic curves, and accurately locates the positions of the performance degradation or faulty strings. The diagnosis result of the fault diagnosis module is linked with the device health assessment module to generate maintenance priority suggestions by combining parameters such as operation duration and environmental corrosion.

[0043] When the platform layer detects device anomalies, it automatically generates an operation and maintenance work order containing fault location information and uploads it to the application layer. The application layer pushes it to the third-party operation and maintenance platform through the operation and maintenance management terminal. After the work order is dispatched, the processing progress is tracked in real time to form an operation and maintenance closed-loop management process. The alarm notifications pushed by the application layer implement a hierarchical push strategy. Critical device failures trigger multi-channel instant alarms such as text messages and APP pop-ups, while general anomalies are notified through in-site messages. In addition, the alarm information is accompanied by a historical case library for handling to assist operation and maintenance personnel in quickly formulating solutions.

[0044] The visual monitoring interface provided by the application layer dynamically displays the operation status information of the photovoltaic power station through a Web page accessible across terminals, which supports: rendering of a three-dimensional efficiency heat map of the photovoltaic array, a comparison dashboard of key indicators for multiple power stations, and monitoring of the status of cluster devices based on a GIS map. Among them, the dynamically rendered three-dimensional efficiency heat map of the photovoltaic array uses color gradient mapping technology to intuitively display the differences in power generation efficiency in each area, and also supports functions such as perspective rotation and sectional analysis; the comparison dashboard of key indicators for multiple power stations supports horizontal comparative analysis of core parameters such as power generation and equipment utilization rate; the monitoring of the status of cluster devices based on a GIS map can display the geographical distribution of the power station cluster, and clicking on the positioning point can penetrate to view the real-time operation parameters of a single power station.

[0045] A two-way WebSocket communication connection is maintained between the application layer and the platform layer, so that the data indicators displayed on the visual monitoring interface can be refreshed at the second level, supporting operation and maintenance personnel to view the changes in the operation status of the power station in real time, and formulating a preventive maintenance plan in combination with the health score output by the platform layer to reduce the risk of sudden device failures.

[0046] Two-way communication of data and control instructions is achieved between the edge layer, the platform layer, and the application layer through a service mesh, and the edge layer and the platform layer dynamically allocate computing tasks. When the network latency exceeds the set threshold, for example, exceeds the 200ms threshold, the edge layer automatically takes over the inverter for protection control, executes the emergency stop instruction and caches the operation log, and synchronizes the operation log to the cloud for verification after the network is restored. Among them, the data transmission between each layer adopts a hierarchical security mechanism, specifically: the data transmission from the edge layer to the platform layer enables the TLS1.3 encryption protocol and message signature verification to prevent data tampering; the service call from the platform layer to the application layer implements OAuth 2.0 token authentication to control the API access rights; critical control instructions are transmitted through an independent and dedicated channel, and a 300ms response timeout protection is set. When the instruction response time exceeds the 300ms set value, the system automatically terminates the current operation sequence and performs a status rollback to avoid device damage caused by instruction delay.

[0047] In this embodiment, the edge layer virtual nodes run on the Kubernetes edge computing cluster, supporting automatic node scaling and fault migration; the platform layer services are deployed in the containerized PaaS platform in the form of microservices, and each microservice can be independently upgraded without affecting the overall operation of the system; the application layer front-end uses WebAssembly technology to achieve cross-terminal rendering, ensuring a smooth interaction experience for both PC and mobile devices.

[0048] In summary, this embodiment provides a cloud platform-based management system for real-time monitoring of photovoltaic power stations. The system is a software-defined distributed architecture including an application layer, a platform layer, and an edge layer. The system is fully deployed on cloud infrastructure and has an edge-cloud collaboration mechanism for dynamic scheduling of computing tasks, significantly improving the operation and maintenance efficiency and system reliability of photovoltaic power stations.

[0049] Embodiment 2

[0050] As Figure 2 shown, this embodiment provides a method for real-time monitoring and management of a photovoltaic power station based on a cloud platform. The method includes:

[0051] 1. Configure a virtualized node cluster at the photovoltaic power station site to form an edge layer. Each node establishes a physical connection with a photovoltaic inverter and a meteorological monitoring device through an industrial communication protocol interface.

[0052] Configure differentiated acquisition strategies for different data types. Specifically, current and voltage parameters are sampled at a period of 1 second, and temperature parameters are acquired at a period of 5 seconds. The acquisition frequency deviation is controlled within ±50 ms.

[0053] 2. After data acquisition is completed, perform sliding window filtering on the original data stream. After processing, it is compressed and encapsulated into a message queue and uploaded to the platform layer. At the same time, the edge layer retains a backup of the original data for fault tracing.

[0054] Among them, for the transient fluctuation characteristics of current and voltage signals, a moving average algorithm with an adaptive window width is adopted. The initial window width is set to 20 sampling points. When 3 consecutive sampling values exceed the threshold fluctuation range, the window width is dynamically reduced to 10 sampling points to improve the filtering response speed.

[0055] The data is encapsulated in a lightweight binary encoding format, and the compression ratio is not less than 60%.

[0056] 3. After receiving the data, the platform layer writes the time-series data into a distributed time-series database, stores the data in shards according to the geographical partition of the photovoltaic power station, and sets the retention period of each shard to 30 days. During this process, if the network latency between the edge layer and the platform layer is detected to exceed the 200ms threshold, the edge layer automatically activates the local decision-making mode, executes the inverter output power limit protection, and temporarily stores the operation logs in the local non-volatile memory.

[0057] After the network communication status is restored, the edge layer starts the data synchronization verification process. Specifically, it uses a log comparison algorithm based on a hash chain to perform consistency verification on the operation logs cached locally and the platform layer database. For instruction records with version conflicts, the authoritative version stored in the platform layer is preferred, and a differential audit report is generated.

[0058] 4. The platform layer periodically calls the machine learning model for power generation prediction and fault diagnosis. Among them, the power generation prediction module based on the LSTM network is used for power generation prediction. The input parameters of the power generation prediction module include historical 72-hour power generation data and future 24-hour weather forecast information, and the relative error of the output result of the power generation prediction module is controlled within the range of ±5%. The fault diagnosis module based on the clustering algorithm is used for fault diagnosis, with the current-voltage characteristic curve of the photovoltaic string as the input data. When the fault diagnosis module detects that the current-voltage characteristic curve of the photovoltaic string deviates from the reference value by more than 15%, it determines that a fault has occurred, accurately locates the fault location, and triggers the health assessment process to generate a maintenance priority matrix in combination with the cumulative operation duration of the equipment and the environmental corrosion factor.

[0059] The rule engine running synchronously in the platform layer parses the data stream in real time. When the radiator temperature parameter of the inverter exceeds the preset threshold of 75°C within 3 consecutive sampling periods, an abnormal event record is generated and associated with the unique identification code of the corresponding inverter.

[0060] When the platform layer pushes a new fault warning event, the system automatically generates a maintenance task instruction containing geographical coordinates and sends a notification message to the specified terminal device through an encrypted communication channel.

[0061] 5. The design application layer provides a visual interaction interface, which provides a 3D efficiency heat map of the photovoltaic array, a comparison dashboard of key indicators of multiple power stations, and the status of cluster devices based on the GIS map.

[0062] Among them, the 3D efficiency heat map uses a physically based lighting rendering algorithm, and the color mapping rule is dynamically updated according to the real-time power generation efficiency value. When the operation and maintenance personnel select the coordinates of an abnormal power station on the GIS map interface, the system automatically retrieves the real-time data stream and synchronously displays the current operation parameters, health score, and pending alarm statistics information of the selected power station in the interaction interface.

[0063] A two-way WebSocket communication connection is maintained between the application layer and the platform layer. Based on the data stored in the platform layer, the three-dimensional heat map rendering engine updates the color gradient distribution at a period of 3 seconds. The device status indicators on the GIS map are dynamically updated according to the heartbeat signal reception. When no signal is received for 2 consecutive heartbeat cycles, it is marked as a warning state, and when no signal is received for more than 5 heartbeat cycles, it is marked as an offline state. Among them, the historical data query function is implemented by calling the sharded storage interface of the platform layer, and the time axis sliding operation triggers the data loading request for the corresponding time zone.

[0064] 6. In the scenario of multi-power station cluster monitoring, the platform layer implements a dynamic load balancing strategy. When the number of connected photovoltaic power stations exceeds 50 nodes, the horizontal expansion mechanism of the cloud computing center server is automatically triggered. Specifically, each server monitors its own resource utilization in real time. When the CPU occupancy rate of all servers continuously exceeds the 80% threshold for 5 minutes, the cloud computing center adds servers, and at the same time activates the container orchestration system of the edge layer, enabling the new servers to complete service registration and access the service flow within 120 seconds, so that the cloud computing center can provide the computing resources required by the system.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A real-time monitoring and management system for a photovoltaic power station based on a cloud platform, characterized in that The system includes an edge layer, a platform layer, and an application layer deployed in the cloud; the edge layer is composed of distributed virtual nodes, and each node obtains the operation data of photovoltaic devices through a data acquisition interface, and uploads the data to the platform layer after performing data preprocessing; the platform layer stores the operation data of photovoltaic devices through a time series database, performs power generation prediction and equipment fault diagnosis through a machine learning model, and generates equipment maintenance strategies based on the diagnosis results; the application layer provides a visual monitoring interface and an operation and maintenance management terminal, dynamically displays the status information of the photovoltaic power station according to the stored data of the platform layer, generates operation and maintenance work orders according to the equipment maintenance strategies generated by the platform layer, and pushes multi-level alarm notifications.

2. The system according to claim 1, wherein In the platform layer, the machine learning model includes a power generation prediction module, a fault diagnosis module, and an equipment health assessment module; the power generation prediction module generates a power generation power curve for the next 24 hours of the photovoltaic power station by analyzing historical power generation data and environmental parameters. The fault diagnosis module uses a clustering algorithm to analyze the operation data of photovoltaic strings, identify abnormal deviations in the current-voltage characteristic curve, and locate the positions of performance degradation or faulty strings; the equipment health assessment module generates maintenance priority suggestions according to the diagnosis results of the fault diagnosis module, combined with the equipment operation duration and environmental corrosion parameters.

3. The system according to claim 1, wherein A two-way WebSocket communication connection is maintained between the application layer and the platform layer, so that the operation status information of the photovoltaic power station displayed on the visual monitoring interface is refreshed in real time.

4. The system according to claim 1 or 3, characterized in that, The visual monitoring interface is constructed through a Web page that can be accessed across terminals, and the visual monitoring interface displays a dynamically rendered three-dimensional efficiency heat map of the photovoltaic array, a comparison dashboard of key indicators of multiple power stations, and the status of cluster devices based on a GIS map.

5. The system according to claim 1, wherein A hierarchical security mechanism is adopted for data transmission between the edge layer, the platform layer, and the application layer. Specifically, the data transmission from the edge layer to the platform layer enables the TLS1.3 encryption protocol and message signature verification to prevent data tampering; the service call from the platform layer to the application layer implements OAuth 2.0 token authentication to control the API access rights. The control instructions between the edge layer, the platform layer, and the application layer are transmitted through a dedicated channel, and a response timeout threshold is set. When the instruction response time exceeds the set threshold, the system automatically terminates the current operation sequence and performs a state rollback to avoid equipment damage caused by instruction delay.

6. The system according to claim 1, wherein The virtual nodes that make up the edge layer run in a Kubernetes edge computing cluster, supporting automatic expansion, contraction, and fault migration of nodes; the platform layer deploys analysis services in the form of microservices through a containerized PaaS platform; the front end of the application layer uses WebAssembly technology to achieve cross-terminal rendering.

7. A real-time monitoring and management method for a photovoltaic power station based on a cloud platform, characterized in that, The method includes: First, construct a real-time monitoring and management system for a photovoltaic power station based on a cloud platform. The system includes an edge layer, a platform layer, and an application layer deployed in the cloud; the edge layer is composed of distributed virtual nodes; the platform layer is configured with a distributed time series database and a machine learning model; the application layer is configured with a visual monitoring interface and an operation and maintenance management terminal. Secondly, each virtual node in the edge layer obtains data through the data acquisition interface. The collected data includes current, voltage, temperature, and weather forecast, and different data types are configured with different acquisition strategies. The original data stream collected is processed by sliding window filtering, and after the processing is completed, it is compressed and encapsulated into a message queue and uploaded to the platform layer. At the same time, the edge layer retains the original data backup for fault tracing. Then, the platform layer writes the time series data into a distributed time series database, stores the data in slices according to the geographical partition of the photovoltaic power station, and sets a retention period for each slice. In addition, the platform layer regularly calls the machine learning model for power generation prediction and fault diagnosis, and generates equipment maintenance strategies based on the diagnosis results. Finally, the application layer displays the 3D efficiency heat map of the photovoltaic array, the comparison dashboard of key indicators of multiple power stations, and the status of cluster devices based on the GIS map in real time in the visualization monitoring interface according to the data stored by the platform layer. At the same time, according to the equipment maintenance strategy generated by the platform layer, the application layer pushes multi-level alarm notifications through the operation and maintenance management terminal to perform equipment anomaly alarms.

8. The method according to claim 7, wherein The configuration of different acquisition strategies for different data types includes: for current and voltage parameters, sampling is performed at a period of 1 second; for temperature parameters, acquisition is performed at a period of 5 seconds; the frequency deviation of data acquisition is within ±50 ms. The sliding window filtering process for the original data stream collected includes: for the transient fluctuation characteristics of current and voltage signals, an adaptive window width moving average algorithm is used for filtering, and the initial window width is set to 20 sampling points. When 3 consecutive sampling values are detected to exceed the ±5% fluctuation range of the nominal value of the photovoltaic power station components, the window width is dynamically reduced to 10 sampling points to improve the filtering response speed to transient anomalies of the photovoltaic array. During the process of the edge layer uploading data to the platform layer, if it is monitored that the network delay between the edge layer and the platform layer exceeds the network delay threshold, the edge layer automatically activates the local decision-making mode, executes the inverter output power limit protection, and temporarily stores the operation log in the local non-volatile memory. After the network communication status is restored, the edge layer starts the data synchronization verification process, and uses a log comparison algorithm based on the hash chain to perform consistency verification on the operation log cached locally and the platform layer database. For the instruction records with version conflicts, the version stored in the platform layer is preferentially adopted, and a differential audit report is generated.

9. The method according to claim 7, characterized in that, The platform layer periodically calls the machine learning model for power generation prediction and fault diagnosis, including: performing power generation prediction through the power generation prediction module based on the LSTM network. The input parameters of the power generation prediction module include historical power generation data for the past 72 hours and weather forecast information for the next 24 hours. The relative error of the output result of the power generation prediction module is controlled within the range of ±5%; performing fault diagnosis through the fault diagnosis module based on the clustering algorithm, using the current-voltage characteristic curve of the photovoltaic string as the input data. When the fault diagnosis module detects that the current-voltage characteristic curve of the photovoltaic string deviates from the reference value by more than 15%, it determines that a fault has occurred, locates the fault position, and triggers the health assessment process to generate a maintenance priority matrix by combining the cumulative operation duration of the equipment and the environmental corrosion factor; The platform layer also synchronously runs a rule engine to parse the data stream in real time. When the radiator temperature parameter of the inverter exceeds 75°C within 3 consecutive sampling cycles, an abnormal event record is generated and associated with the unique identification code of the corresponding inverter; The refresh rules for the visualization monitoring interface to display the operation status information of the photovoltaic power station include: the three-dimensional efficiency heat map of the photovoltaic array updates the color gradient distribution every 3 seconds; The device status identifier on the GIS map is dynamically updated according to the received heartbeat signal. When the signal is not received for 2 consecutive heartbeat cycles, it is marked as a warning state, and when the signal is not received for more than 5 heartbeat cycles, it is marked as an offline state.

10. The method according to claim 7, wherein This method is applicable to the multi-power station cluster monitoring scenario. When the number of photovoltaic power stations connected to the management system exceeds 50, the horizontal expansion mechanism of the cloud computing center server is triggered, including: each server monitors its own resource utilization rate in real time. When the CPU occupancy rate of all servers continues to be higher than 80% for 5 minutes, the cloud computing center adds a server, and at the same time activates the container orchestration system of the edge layer to enable the newly added server to complete service registration and access the business flow within 120 seconds.

Images