Method and system for monitoring real-time operation state of photovoltaic power station

Abstract

The application provides a method and system for real-time monitoring of the operation state of a photovoltaic power station, and relates to the technical field of photovoltaic power station monitoring. The method integrates a multi-modal sensor array in a photovoltaic module junction box, and collects sensor data of the photovoltaic power station in real time. The sensor data is pre-processed locally by an edge computing node, and different priority data is transmitted to the cloud based on a dynamic hybrid communication protocol. The sensor data is fused in the cloud to construct a four-dimensional digital twin model, and spatiotemporal continuous weather prediction, component aging and dust evolution data are fused. The fault mode of the photovoltaic power station is identified, and the string topology of the photovoltaic power station is dynamically optimized in combination with the real-time operation state, equipment health data and fault mode identification result of the photovoltaic power station. By integrating distributed sensors, edge computing modules and intelligent alarm mechanisms, the application can effectively solve the problems of delay, insufficient intelligent analysis and response lag in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power station monitoring, and in particular to a method and system for real-time monitoring of the operating state of a photovoltaic power station. BACKGROUND

[0002] With the rapid development of renewable energy, photovoltaic power generation, as an important technology, is widely used in various scenarios. However, existing photovoltaic power station monitoring systems mostly rely on centralized data processing platforms, and most systems rely on data acquisition modules and data transmission networks on site.

[0003] However, the existing photovoltaic power station monitoring system has the problems of single sensing dimension, delayed decision response, and insufficient adaptability. It mainly relies on electrical parameter monitoring and cannot capture physical state changes such as component micro-cracks and dust accumulation. The abnormal monitoring and early warning mechanism often relies on preset rules and fails to fully utilize data mining technology to identify potential fault patterns from massive data. The fixed networking topology used is difficult to cope with dynamic scenarios such as shadow movement and component aging, and cannot automatically adjust the operation strategy according to real-time environmental changes or system load. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a method and system for real-time monitoring of the operating state of a photovoltaic power station based on a distributed sensor network and edge computing, aiming to improve data transmission efficiency, enhance intelligent analysis capability, and realize dynamic optimization management of the power station. By integrating distributed sensors, edge computing modules, and intelligent alarm mechanisms, the present application can effectively solve the problems of delay, insufficient intelligent analysis, and delayed response in the prior art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] Based on the above purpose, in the first aspect, the present application provides a method for real-time monitoring of the operating state of a photovoltaic power station, comprising the following steps:

[0007] Integrating a multi-modal sensor array in the junction box of a photovoltaic module to collect real-time sensor data of the photovoltaic power station;

[0008] Performing local preprocessing of the sensor data by an edge computing node deployed on site of the photovoltaic power station, and transmitting different priority data to the cloud based on a dynamic hybrid communication protocol;

[0009] Fusing the sensor data in the cloud to construct a four-dimensional digital twin model, and fusing spatiotemporally continuous weather prediction, component aging, and dust evolution data;

[0010] A transfer learning framework is used to learn from historical fault data, identify fault modes of photovoltaic power plants, and dynamically optimize the string topology of photovoltaic power plants by combining the real-time operating status of photovoltaic power plants, equipment health data and fault mode identification results.

[0011] The optimized string topology of the photovoltaic power station is presented through a visual interface, and the health status of the power station and control commands are fed back in real time. The control commands are sent to the edge computing nodes on site in real time.

[0012] As a further aspect of the present invention, the real-time acquisition of sensor data from the photovoltaic power station includes electrical parameters, physical state, and environmental data; when performing local preprocessing on the sensor data, it includes anomaly detection, data compression, and priority marking; and the data of different priorities are transmitted to the cloud via LoRa Mesh, 5G slicing, or satellite link.

[0013] As a further aspect of the present invention, the multimodal sensor array includes a graphene quantum dot sensor, a piezoelectric vibration sensor, and a polarized light sensor, which simultaneously detect the component's operating voltage, leakage current, microcracks, dust accumulation thickness, surface temperature, polarization angle offset, and microcrack depth.

[0014] As a further aspect of the present invention, control commands are sent to the edge computing nodes in real time, and the edge computing nodes perform the following steps:

[0015] Run a lightweight ResNet-8 model to complete real-time fault classification within a preset time.

[0016] Sparse coding is used to compress multidimensional sensor data and data priority is divided according to fault confidence. Emergency alarm data is marked as priority level, and the trigger data is transmitted directly to the cloud through LoRa Mesh, 5G slicing or satellite link.

[0017] As a further aspect of the present invention, when transmitting data of different priorities to the cloud based on the dynamic hybrid communication protocol, the dynamic hybrid communication protocol satisfies the following:

[0018] LoRa Mesh network coverage radius ≥1km, single node power consumption ≤0.3W;

[0019] The end-to-end latency of the 5G slicing network is ≤8ms, supporting concurrent access of 1000 terminals per square kilometer;

[0020] The satellite link adopts dual-mode redundancy of Beidou-3 and Starlink, and automatically switches when communication is interrupted, with a switching delay of ≤3s.

[0021] As a further aspect of the present invention, the sensor data is fused in the cloud to construct a four-dimensional digital twin model, including the following steps:

[0022] A three-dimensional power station model is established according to a spatial dimension, and physical parameters and electrical connection relationships of each component are integrated;

[0023] Meteorological forecast data of a preset time period, including irradiance fluctuation, wind speed change and precipitation probability, are fused according to a time dimension;

[0024] According to a state dimension, a component aging index is calculated through EL image degradation analysis, and a remaining service life is predicted;

[0025] According to an environment dimension, a dust accumulation growth model is constructed based on PM2.5 concentration and rainfall frequency, and a cleaning period is predicted.

[0026] As a further scheme of the present application, when a fault mode of the photovoltaic power station is identified by using a transfer learning framework to learn historical fault data, the implementation of the transfer learning framework includes:

[0027] A feature vector knowledge base containing several historical fault modes is constructed;

[0028] When an unknown anomaly is detected, a graph convolution network (GCN) is used to match topological similarity, and zero-shot detection is realized;

[0029] Each week, the data of each power station is aggregated through federated learning to update the global model parameters, while the local privacy data is preserved.

[0030] As a further scheme of the present application, the string topology structure of the photovoltaic power station is dynamically optimized, including the following steps:

[0031] The irradiance distribution map of the power station is scanned in real time, and the area with a power generation efficiency lower than a power generation threshold is identified;

[0032] The components in the shadow area are switched to an independent MPPT channel through a matrix switch array to form a dynamic subarray;

[0033] Based on deep deterministic policy gradient, Pareto optimality is realized among power generation efficiency, heat loss and equipment life;

[0034] The optimization strategy is recalculated every 60 minutes, and the networking reconstruction response time is ≤30s.

[0035] As a further scheme of the present application, the method for monitoring the operation state of the photovoltaic power station in real time further includes a self-calibration mechanism:

[0036] Temperature drift errors are eliminated through mutual checking algorithms of adjacent sensor data every 20ms;

[0037] Offline calibration is performed every morning, and a laser interferometer is used to correct the reference value of micro-crack detection;

[0038] When the dust thickness is greater than 3 g / m2, the scheduled work order of the unmanned aerial vehicle cleaning system is automatically triggered.

[0039] In a second aspect, the application further provides a system for real-time monitoring of the operation state of a photovoltaic power station, comprising the following components:

[0040] A data acquisition module for acquiring sensor data of the photovoltaic power station in real time by integrating a multi-modal sensor array in the junction box of the photovoltaic module;

[0041] An edge computing module for local preprocessing of the sensor data by an edge computing node deployed on site of the photovoltaic power station, and transmitting data of different priorities to the cloud based on a dynamic hybrid communication protocol;

[0042] A model construction module for fusing the sensor data in the cloud, constructing a four-dimensional digital twin model, and fusing spatiotemporally continuous meteorological prediction, module aging and dust evolution data;

[0043] A fault identification module for learning from historical fault data using a transfer learning framework to identify fault modes of the photovoltaic power station, and dynamically optimizing the string topology of the photovoltaic power station in combination with the real-time operation state, equipment health data and fault mode identification results of the photovoltaic power station;

[0044] A visual display module for presenting the optimized string topology of the photovoltaic power station through a visual interface, and feeding back the health state and control instructions of the power station in real time, with the control instructions being issued to the edge computing node on site in real time.

[0045] As a further scheme of the application, the model construction module further comprises a cloud data processing platform receiving data from the edge computing node, fusing the sensor data and constructing a four-dimensional digital twin model, and the cloud data processing platform comprises the following functional modules:

[0046] A spatial dimension modeling module for establishing a three-dimensional power station model, integrating physical parameters and electrical connection relationships of each module;

[0047] A time dimension modeling module for fusing meteorological prediction data, including irradiance fluctuation, wind speed variation and precipitation probability;

[0048] A state dimension modeling module for calculating a module aging index through EL image degradation analysis to predict the remaining life;

[0049] An environmental dimension modeling module for constructing a dust growth model based on PM2.5 concentration and rainfall frequency to predict the cleaning period.

[0050] As a further scheme of the present application, the fault identification module comprises a string topology optimization unit for dynamically optimizing the string topology structure of the photovoltaic power station according to the real-time operation state of the power station, equipment health data and fault mode identification results, and the string topology optimization unit comprises:

[0051] Scan the irradiance distribution map of the power station, and identify the area with power generation efficiency lower than the power generation threshold;

[0052] Switch the shadow area component to an independent MPPT channel through the matrix switch array to form a dynamic subarray;

[0053] Based on the deep deterministic policy gradient (DDPG) algorithm, the Pareto optimality is realized among power generation efficiency, heat loss and equipment life;

[0054] The optimization strategy is recalculated every 60 minutes, and the networking reconstruction response time is ≤30s.

[0055] As a further scheme of the present application, the system for real-time monitoring of the operation state of the photovoltaic power station further comprises a self-calibration mechanism, and the self-calibration mechanism performs the following operations:

[0056] Eliminate temperature drift error through mutual checking algorithm of adjacent sensor data every 20ms;

[0057] Perform offline calibration every morning, and use a laser interferometer to correct the reference value of micro-crack detection;

[0058] When the dust thickness is >3g / m², an appointment work order of the unmanned aerial vehicle cleaning system is automatically triggered.

[0059] Compared with the prior art, the method and system for real-time monitoring of the operation state of the photovoltaic power station have the following beneficial effects:

[0060] 1. The present application can monitor the electrical parameters, physical state and environmental data of the photovoltaic power station in real time through the integrated multi-modal sensor array, such as voltage, leakage current, micro-crack, dust thickness, surface temperature, polarization angle offset and hidden crack depth. Through multi-dimensional data collection, comprehensive monitoring of the operation state of the photovoltaic power station is provided, which helps to accurately identify potential faults or abnormalities. By deploying edge computing nodes, local data preprocessing and real-time fault classification are carried out on site, which not only effectively reduces data transmission delay, but also avoids bandwidth pressure caused by large-scale data upload. The edge nodes perform hierarchical transmission according to the priority of the data, ensuring that critical data (such as emergency alarm data) can be quickly transmitted to the cloud, improving fault response speed and enhancing the intelligence and reliability of the system.

[0061] 2.The application can identify historical failure modes of photovoltaic power stations and realize zero-sample detection. Through intelligent identification of failure modes, potential failures can be predicted based on historical data and operating conditions, early warning can be given, and the occurrence of sudden failures can be reduced to ensure the long-term stable operation of the power station. Based on real-time power station operating data and failure mode identification results, the application can dynamically optimize the string topology structure of the power station to improve the power generation efficiency of the photovoltaic power station. By using deep deterministic policy gradient, an optimal balance between power generation efficiency, heat loss and equipment life can be found, and a Pareto optimal string topology configuration is realized, thereby further improving the overall economic benefit of the photovoltaic power station.

[0062] 3.The application can also provide a real-time feedback visualization interface to intuitively present the optimized photovoltaic power station string topology structure, operating health status and real-time control instructions, making it easy for users to view the power station status and issue control instructions in real time, improving the operability and management convenience of the system. By monitoring the operating data of the photovoltaic power station in real time, the system can timely trigger an alarm when a failure or anomaly occurs, and through priority sorting, the alarm information can be quickly transmitted to relevant personnel or systems, ensuring that the photovoltaic power station can be processed as soon as possible when a failure occurs, reducing the impact of the failure on the operation of the power station, reducing system downtime, and improving the reliability of the power station.

[0063] 4.The string topology optimization and failure mode identification mechanism of the application not only improves the power generation efficiency of the photovoltaic power station, but also reduces the heat loss of the equipment and prolongs the service life of the equipment. Through intelligent optimization and management, the system improves the power generation benefit of the photovoltaic power station while reducing the loss of the equipment, thereby prolonging the service life of the power station equipment, which has high long-term economic value; combined with edge computing and cloud big data processing, real-time data processing and control decision-making can be realized on the spot of the photovoltaic power station, and the self-calibration mechanism and failure identification module of the system enable the power station to automatically optimize the system and give early warnings without human intervention, greatly reducing the labor cost and improving the intelligent level of system operation.

[0064] In summary, the method and system for monitoring the operating state of a photovoltaic power station in real time provided by the application can efficiently, accurately and intelligently monitor the operating state of the power station, predict and prevent failures, optimize the operating configuration of the power station, ensure the stable and efficient operation of the photovoltaic power station, and improve the economic and environmental benefits of photovoltaic power generation.

[0065] These aspects or other aspects of the present application will be more apparent in the following description of the embodiments. It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the following will briefly introduce the drawings required by the exemplary embodiments or the related art description. The drawings are used to provide further understanding of the present application, and constitute a part of the specification. The drawings together with the embodiments of the present application are used to explain the present application and do not constitute a limitation of the present application. In the drawings:

[0067] Figure 1 A flow chart of a method for monitoring a running state of a photovoltaic power station in real time according to an embodiment of the present application.

[0068] Figure 2 A flow chart of constructing a four-dimensional digital twin model in a method for monitoring a running state of a photovoltaic power station in real time according to an embodiment of the present application.

[0069] Figure 3 A flow chart of implementing a transfer learning framework in a method for monitoring a running state of a photovoltaic power station in real time according to an embodiment of the present application.

[0070] Figure 4 A flow chart of dynamically optimizing a string topology of a photovoltaic power station in a method for monitoring a running state of a photovoltaic power station in real time according to an embodiment of the present application.

[0071] Figure 5 A flow chart of an edge computing node performing in a method for monitoring a running state of a photovoltaic power station in real time according to an embodiment of the present application. DETAILED DESCRIPTION

[0072] The present application will be further described below in conjunction with the drawings and specific embodiments. It should be noted that the following described embodiments or technical features can be combined with each other to form new embodiments without conflict.

[0073] In order to make the objectives, technical solutions and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the drawings and in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0074] It should be noted that all the expressions of "first" and "second" in the embodiments of the present application are used to distinguish two non-identical entities or non-identical parameters with the same name. It can be seen that "first" and "second" are only used for the convenience of description and should not be understood as a limitation of the embodiments of the present application. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, the process, method, system, product or device inherently includes other steps or units.

[0075] With reference to the drawings, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative work belong to the protection scope of the present application.

[0076] The flowcharts shown in the drawings are only exemplary, not necessarily including all the contents and operations / steps, and not necessarily executed in the described order. For example, some operations / steps can be decomposed, combined or partially merged, so that the actual execution order can be changed according to the actual situation.

[0077] Some embodiments of the present application will be described in detail below with reference to the drawings. The following embodiments and features in the embodiments can be combined with each other without conflict.

[0078] In view of the delay problem, insufficient intelligent analysis and response lag of the existing photovoltaic power station monitoring system, the present application proposes a method for real-time monitoring of the operating state of a photovoltaic power station, aiming to improve data transmission efficiency, enhance intelligent analysis capability, and realize dynamic optimization management of the power station. The above problems are solved by integrating distributed sensors, edge computing modules and intelligent alarm mechanisms.

[0079] Referring to Figure 1 The embodiments of the present application provide a method for real-time monitoring of the operating state of a photovoltaic power station, which comprises the following steps:

[0080] Step S10, integrating a multi-modal sensor array in a photovoltaic module junction box to collect sensor data of the photovoltaic power station in real time.

[0081] In this step, the real-time collection of sensor data of the photovoltaic power station includes electrical parameters, physical states and environmental data.

[0082] The multi-modal sensor array includes a graphene quantum dot sensor, a piezoelectric vibration sensor and a polarized light sensor, which synchronously detect the working voltage, leakage current, micro-cracks, dust thickness, surface temperature, polarization angle offset and hidden crack depth of the component. Specifically, in the sensor array integrated in the photovoltaic module junction box, the functions of each sensor are as follows:

[0083] Graphene quantum dot sensor: used for high-precision detection of electrical parameters of the photovoltaic component, including current, voltage, etc.

[0084] Piezoelectric vibration sensor: detects physical states such as micro-cracks, component structure vibration and mechanical stress.

[0085] Polarization light sensor: used to detect the physical properties of the photovoltaic module surface, such as dust thickness, surface temperature and polarization angle offset.

[0086] When synchronously detecting the working state of the photovoltaic module, the graphene quantum dot sensor, the piezoelectric vibration sensor and the polarization light sensor in the above multi-modal sensor array are used to synchronously detect the working voltage and leakage current of the module, micro-cracks, dust thickness, surface temperature, polarization angle offset and hidden crack depth. The micro-cracks are used to determine whether the module is damaged due to external stress, the dust thickness affects the light absorption efficiency of the module, the surface temperature is used to detect the overheating condition of the module, the polarization angle offset is used to detect the influence of external environmental factors on the photovoltaic module, and the hidden crack depth is used to evaluate the health status of the module by detecting the depth of the micro-cracks. The sensor array can provide multi-dimensional data about the health of the module, the power generation efficiency and the environmental influence for the power station managers by real-time monitoring of the above parameters of the photovoltaic module, so that potential faults or performance degradation can be found in time. For example, the voltage of the photovoltaic module normally works at 30V, if the voltage of a certain module is lower than the preset threshold, the system will determine that the module may have a fault and issue an alarm.

[0087] In step S20, the edge computing node deployed on site of the photovoltaic power station performs local preprocessing on the sensor data, and transmits different priority data to the cloud based on a dynamic hybrid communication protocol.

[0088] In this step, when the sensor data is preprocessed locally, it includes anomaly detection, data compression and priority marking; different priority data is transmitted to the cloud through LoRa Mesh, 5G slice or satellite link. When performing anomaly detection, the edge node determines whether the collected electrical parameters and physical state are abnormal through a model, for example, an alarm is triggered when the voltage fluctuation exceeds the threshold; when performing data compression, the edge computing node compresses the data due to the large amount of data generated by the sensor, only important information is retained, the data transmission amount is reduced, and the bandwidth consumption is reduced; when performing priority marking, the edge node can ensure that emergency data (such as device fault alarm) is uploaded to the cloud in priority, so as to ensure timely response.

[0089] In this embodiment, when different priority data is transmitted to the cloud based on the dynamic hybrid communication protocol, the dynamic hybrid communication protocol meets the following conditions:

[0090] The coverage radius of the LoRa Mesh network is ≥1km, and the power consumption of a single node is ≤0.3W;

[0091] The end-to-end delay of the 5G slice network is ≤8ms, and 1000 terminal concurrent accesses per square kilometer are supported;

[0092] The satellite link adopts a Beidou third-generation and Starlink dual-mode redundancy, which automatically switches when the communication is interrupted, and the switching delay is less than 3s.

[0093] In this embodiment, the edge computing node realizes local processing and preliminary analysis of data, which can reduce the transmission amount of data and ensure the priority transmission of critical information. Using a dynamic hybrid communication protocol, the system can select the appropriate transmission method according to the priority of different data, thereby optimizing the use of network resources. For example, when a device fails, the edge node detects the fault data and immediately marks it as high priority, and uploads the data to the cloud through the 5G slice network. Other data (such as environmental temperature) is uploaded regularly through the LoRa Mesh network. The edge computing node can perform real-time data preprocessing on site, reduce data transmission load, improve response speed, and ensure that faults can be quickly handled and alarms can be sent. Using a dynamic hybrid communication protocol, the appropriate transmission method is selected according to the priority of the data, maximizing communication efficiency and ensuring that important data can be quickly transmitted to the cloud. Through the combination of LoRa Mesh, 5G slice and satellite link, the system can still ensure stable data transmission even in unstable network environment, improving the reliability and stability of the system.

[0094] Step S30, fuse the sensor data in the cloud to build a four-dimensional digital twin model, and fuse the spatio-temporal continuous weather forecast, component aging and dust evolution data.

[0095] In this step, referring to Figure 2 as shown, the sensor data is fused in the cloud to build a four-dimensional digital twin model, including the following steps:

[0096] Step S301, establish a three-dimensional power station model according to the spatial dimension, integrate the physical parameters and electrical connection relationship of each component.

[0097] In this embodiment, a three-dimensional model of the power station is created in the cloud, which contains the physical location and spatial distribution of all components in the photovoltaic power station; the physical parameters (size, model, installation angle and direction) of each photovoltaic component are modeled according to the actual installation situation; through the electrical connection relationship, the electrical circuit and transmission path between the components of the power station are established, simulating the electrical connection and data flow of each component inside the power station, providing a comprehensive and three-dimensional view of the power station through the three-dimensional power station model, which helps to understand the physical layout of the power station and the electrical interconnection relationship between components. For example, taking a certain photovoltaic power station as an example, the power station contains 100 photovoltaic components, and the installation angle and direction of each component are different. The three-dimensional model obtains the actual data of the components, draws the accurate position of the components, and labels the electrical connection relationship between each component and the power station combiner box. Through this model, the electrical performance and spatial distribution of each component can be monitored in real time.

[0098] Step S302, fuse the weather forecast data of the preset time period according to the time dimension, including irradiance fluctuation, wind speed change and precipitation probability.

[0099] In this embodiment, the weather data is integrated into the digital twin model. By cooperating with the weather service provider, the weather data within a preset time period is obtained, including:

[0100] Irradiance fluctuation: the expected hourly radiation intensity affects the power generation efficiency of the photovoltaic module.

[0101] Wind speed change: the change of wind speed has an impact on the heat dissipation performance of the photovoltaic module, wind damage, etc.

[0102] Precipitation probability: precipitation will affect the cleanliness and power generation efficiency of the photovoltaic module.

[0103] Among them, the time dimension weather data can help predict the impact of different weather conditions on the power station. For example, if the wind speed is high in a certain time period, the system can predict the possible wind damage or insufficient heat dissipation of the photovoltaic module, and then make necessary warning or maintenance arrangements.

[0104] For example, assume that the weather data for the next 24 hours is: irradiance fluctuation range is 800-1200 W / m², wind speed change range is 5-15 m / s, and precipitation probability is 20%. By integrating these data, the system can predict the power generation situation in the next 24 hours and timely adjust the cleaning plan or alarm the area with high wind speed.

[0105] Step S303, calculate the component aging index according to the state dimension through EL image degradation analysis, and predict the remaining life.

[0106] In this embodiment, by obtaining the electroluminescence (EL) image of the photovoltaic module, the degradation state of the module is analyzed; image processing technology is used to analyze the degradation of the electroluminescence (EL) image, to detect microcracks, hot spots, defects and other problems inside the module; according to these degradation indicators, the aging index of the module is calculated, and the remaining life is predicted combined with historical data and degradation trend. Among them, the EL image degradation analysis can identify microcracks, hot spots and other degradation phenomena by analyzing the electroluminescence response of the photovoltaic module. Through these degradation information, combined with the design life and operating state of the module, the system can provide accurate prediction of the remaining service life.

[0107] For example: if the EL image analysis of the photovoltaic module shows that there are multiple microcracks, and after degradation analysis, the system calculates that the aging index of the module is 0.85, and the remaining life is expected to be 5 years. Based on this information, the maintenance or replacement work of the module can be planned in advance by the operation and maintenance personnel.

[0108] Step S304, according to the environmental dimension, a dust accumulation growth model is constructed based on PM2.5 concentration and rainfall frequency to predict the cleaning period.

[0109] In this embodiment, by monitoring the PM2.5 concentration and rainfall frequency in the air, a dust accumulation growth model is constructed to estimate the impact of dust accumulation on the performance of the photovoltaic module, wherein the PM2.5 concentration directly affects the surface dust of the photovoltaic module, and higher PM2.5 concentration means more dust accumulation; the rainfall frequency affects the cleaning period of the dust, and the rainfall can help to remove part of the dust and reduce the cleaning frequency. The above data system can calculate the optimal cleaning time window, and the environmental dimension data and the dust accumulation growth model are combined to predict the surface dust of the photovoltaic module, and the system can dynamically adjust the cleaning strategy. For example, when the PM2.5 concentration rises, the system can adjust the cleaning plan in advance.

[0110] For example, if the monitored PM2.5 concentration is 150 μg / m³ and the rainfall probability is low, the system predicts that the dust accumulation of the photovoltaic module in this area will reach the preset cleaning standard, and suggests cleaning in the next month. If the rainfall is frequent, the cleaning plan can be delayed to reduce maintenance costs.

[0111] The embodiment of the present application fuses sensor data in the cloud to construct a four-dimensional digital twin model, making the operation of the photovoltaic power station more intelligent and automated. By establishing a three-dimensional power station model, the physical structure and electrical connection relationship of the photovoltaic power station are comprehensively described. By integrating weather forecast data, the system can predict the impact of weather on the power generation efficiency and safety of the power station, and timely adjust the operation strategy. Based on EL image degradation analysis, the system can accurately predict the aging degree and remaining life of the component, which is helpful for long-term operation and maintenance management of the power station. Through the dust accumulation growth model, the cleaning period is predicted combined with environmental data to optimize the cleaning plan of the power station and reduce unnecessary maintenance costs. The present application provides a more efficient and intelligent operation and maintenance management scheme for photovoltaic power stations, which can significantly improve the operation efficiency and reliability of the power station.

[0112] Step S40, learning historical fault data using a transfer learning framework to identify fault modes of the photovoltaic power station, and dynamically optimizing the string topology of the photovoltaic power station combined with the real-time operation state of the photovoltaic power station, the equipment health data and the fault mode identification result.

[0113] In this step, referring to Figure 3 As shown in the figure, when learning historical fault data using a transfer learning framework to identify fault modes of the photovoltaic power station, the implementation of the transfer learning framework includes:

[0114] Step S401, a feature vector knowledge base containing several historical fault modes is constructed.

[0115] In this embodiment, by collecting a large amount of historical fault data, including different types of fault modes (such as short circuit, poor contact, component aging, system failure, etc.), a feature vector is extracted for each fault mode, including abnormal fluctuations in real-time monitoring data such as voltage, current, temperature, and radiation, and then the fault data is preprocessed to generate feature vectors for different fault modes, and a knowledge base is constructed. Each feature vector of each fault mode represents the trend of changes in different parameters (such as temperature, power, etc.) under that mode. By constructing a fault mode feature library, it is possible to quickly match historical fault modes when an anomaly is detected, improving the accuracy of fault diagnosis.

[0116] Step S402, when an unknown anomaly is detected, the topological similarity is matched by a graph convolutional network (GCN) to realize zero-shot detection.

[0117] When an unknown anomaly occurs in a power station, the current state of the power station is first captured in real time by sensor data and a monitoring system. Then, the topological structure of the power station is modeled using a graph convolutional network (GCN), and the current state is matched with historical fault modes to find potential fault modes. The GCN model can effectively associate anomalies with fault modes by learning the topological structure between components in the power station, and can still perform zero-shot detection even when there is not enough labeled data.

[0118] For example, when the power station detects an abnormal current fluctuation in a component, the GCN model matches the topological position of the component (e.g., which other components it is connected to) and historical fault data to find that it is a zero-shot detection result of a "poor contact" fault mode, and a fault alarm is triggered in a timely manner.

[0119] Step S403, aggregate the data of each power station by federated learning every week to update the global model parameters while preserving local privacy data.

[0120] In this embodiment, the photovoltaic power station uses a federated learning mechanism that allows different power stations to train data locally and transmit updated model parameters to the cloud for aggregation; federated learning data aggregation is performed regularly every week to ensure continuous optimization of the model of each power station, while protecting local data privacy and not requiring sensitive data to be uploaded to the cloud. The federated learning framework can update and optimize the model by sharing and aggregating local model parameters while ensuring data privacy, improving the accuracy and robustness of the global model.

[0121] For example, there are 5 photovoltaic power stations participating in federated learning, each power station trains a fault diagnosis model using local sensor data and uploads the parameters to the central server for aggregation. The server optimizes the global fault detection model by aggregating the model parameters, thereby improving the recognition ability of unknown fault patterns.

[0122] In this embodiment, referring to Figure 4 The dynamic optimization of the string topology of the photovoltaic power station includes the following steps:

[0123] Step S411, real-time scanning of the irradiance distribution map of the power station to identify areas with power generation efficiency below the power generation threshold.

[0124] The irradiance distribution map of the power station is monitored in real time by sensors of the photovoltaic power station, and image processing or sensor data analysis techniques are used to identify areas with power generation efficiency below the preset threshold. The area has a decrease in power generation efficiency due to shading, failure or component damage. Through the irradiance distribution map, the power generation of each area of the power station can be monitored in real time. When the power generation efficiency of an area is below the threshold, the system can quickly identify and issue a fault warning. For example, if a sensor of a power station finds that the irradiance of an area is low, resulting in a power generation efficiency of only 70%, which is below the set threshold of 90%. After the system identifies this anomaly, it automatically triggers a maintenance plan to check whether the photovoltaic components in the area have a fault or shading.

[0125] Step S412, switching the components in the shadow area to an independent MPPT channel through a matrix switch array to form a dynamic sub-array.

[0126] For areas with uneven irradiance distribution, the matrix switch array technology is used to switch the components in the area to an independent Maximum Power Point Tracking (MPPT) channel, which can avoid the impact of the shadow area on the overall power generation efficiency of the power station, and maximize the power generation performance of each area. Through the matrix switch array, the power station can dynamically switch the components in different areas to an independent MPPT channel, ensuring that the components in the shadow area do not affect the overall power generation efficiency of the power station.

[0127] Step S413, achieving Pareto optimality between power generation efficiency, heat loss and device lifetime based on deep deterministic policy gradient.

[0128] Step S414, recompute the optimization strategy every 60 minutes, and the network reconstruction response time is ≤30s.

[0129] The optimization strategy of the power station is recalculated and adjusted every 60 minutes, which can ensure that the topology of the power station is optimized according to the latest environmental changes, equipment states and power generation data, and the response time of network reconstruction is required to be less than 30 seconds, so that the power station can timely adjust and adapt to external condition changes.

[0130] For example, when the external environment of the power station changes during operation (such as cloudy weather), the system will recalculate and optimize the topology every 60 minutes to ensure that the reconstruction response is completed within 30 seconds to continue to maintain the best power generation efficiency.

[0131] The application can identify historical failure modes of photovoltaic power stations and realize zero-sample detection. Through intelligent identification of failure modes, potential failures can be predicted according to historical data and operating states, early warning can be given, and the occurrence of sudden failures can be reduced to ensure long-term stable operation of the power station. Based on real-time power station operation data and failure mode identification results, the application can dynamically optimize the string topology structure of the power station to improve the power generation efficiency of the photovoltaic power station. By using deep deterministic policy gradient, the optimal balance between power generation efficiency, heat loss and equipment life can be found to realize the Pareto optimal string topology configuration, thereby further improving the overall economic benefit of the photovoltaic power station.

[0132] Step S50, the optimized string topology structure of the photovoltaic power station is presented through a visual interface, and the health status of the power station and control instructions are fed back in real time. The control instructions are sent to the edge computing node in the field in real time.

[0133] In this step, referring to Figure 5 The control instructions are sent to the edge computing node in the field in real time, and the edge computing node performs the following steps:

[0134] Step S501, a lightweight ResNet-8 model is run to complete real-time classification of faults within a preset time;

[0135] Step S502, multi-dimensional sensing data is compressed by using sparse coding, and data priority is divided according to fault confidence, wherein emergency alarm data is marked as a priority level, and data is triggered to be transmitted directly to the cloud through LoRa Mesh, 5G slice or satellite link.

[0136] The application can also provide a visual interface for real-time feedback, intuitively presenting the optimized photovoltaic power station string topology, operation health status and real-time control instructions, so that users can easily check the power station status and issue control instructions in real time, improving the operability and management convenience of the system. By monitoring the operation data of the photovoltaic power station in real time, the alarm can be triggered in time when a fault or anomaly occurs, and the alarm information can be quickly transmitted to the relevant personnel or system through priority sorting, so as to ensure that the photovoltaic power station can be processed as soon as possible when a fault occurs, reduce the influence of the fault on the operation of the power station, reduce the downtime of the system, and improve the reliability of the power station.

[0137] In some embodiments, the method for monitoring the operation state of the photovoltaic power station in real time further comprises a self-calibration mechanism.

[0138] Every 20 ms, the temperature drift error is eliminated through mutual checking algorithm of adjacent sensor data;

[0139] Offline calibration is performed every morning, and a laser interferometer is used to correct the reference value of micro-crack detection;

[0140] When the dust thickness is greater than 3g / m2, the reservation work order of the unmanned aerial vehicle cleaning system is automatically triggered.

[0141] The method for monitoring the operation state of the photovoltaic power station in real time can monitor the electrical parameters, physical state and environmental data of the photovoltaic power station in real time through the integrated multi-modal sensor array, such as voltage, leakage current, micro-crack, dust thickness, surface temperature, polarization angle offset and hidden crack depth. Through multi-dimensional data collection, comprehensive monitoring of the operation state of the photovoltaic power station is provided, which helps to accurately identify potential faults or abnormalities. By deploying edge computing nodes, local data preprocessing and real-time fault classification are performed on site, which not only effectively reduces the data transmission delay, but also avoids the bandwidth pressure caused by large-scale data upload. The edge nodes perform hierarchical transmission according to the priority of the data, ensuring that critical data (such as emergency alarm data) can be quickly transmitted to the cloud, improving the fault response speed and enhancing the intelligence and reliability of the system.

[0142] It should be noted that the above-described figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the application, and are not for limiting purposes. It is easy to understand that the processes shown in the above-described figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be executed synchronously or asynchronously, for example, in multiple modules.

[0143] It should be understood that although the above steps are described in a certain order, these steps are not necessarily executed in the above order. Unless explicitly stated herein, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, part of the steps of the present embodiment can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately executed with at least part of other steps or steps or stages in other steps.

[0144] Embodiments of the present application also provide a system for monitoring the operation state of a photovoltaic power station in real time, which is used to execute the aforementioned method for monitoring the operation state of a photovoltaic power station in real time, and comprises:

[0145] A data acquisition module is configured to acquire sensor data of the photovoltaic power station in real time by integrating a multi-modal sensor array in a photovoltaic component junction box.

[0146] An edge computing module is configured to perform localized preprocessing on the sensor data through an edge computing node deployed on site of the photovoltaic power station, and transmit data of different priorities to the cloud based on a dynamic hybrid communication protocol.

[0147] A model construction module is configured to fuse the sensor data in the cloud, and construct a four-dimensional digital twin model, and fuse spatiotemporal continuous weather forecast, component aging and dust evolution data.

[0148] A fault identification module is configured to learn historical fault data using a transfer learning framework, identify fault modes of the photovoltaic power station, and dynamically optimize the string topology of the photovoltaic power station in combination with real-time operation state, equipment health data and fault mode identification results of the photovoltaic power station.

[0149] A visual display module is configured to present the optimized string topology of the photovoltaic power station through a visual interface, and feed back the health state of the power station and control instructions in real time, and the control instructions are fed back to the edge computing node on site in real time.

[0150] The model construction module further comprises a cloud data processing platform that receives data from the edge computing node, fuses the sensor data and constructs a four-dimensional digital twin model, and the cloud data processing platform comprises the following functional modules:

[0151] A spatial dimension modeling module is configured to establish a three-dimensional power station model, and integrate physical parameters and electrical connection relationships of each component.

[0152] A time dimension modeling module is configured to fuse weather forecast data, including irradiance fluctuation, wind speed variation and precipitation probability.

[0153] State dimension modeling module: calculate the aging index through EL image degradation analysis, and predict the remaining life;

[0154] Environmental dimension modeling module: based on PM2.5 concentration and rainfall frequency, build dust accumulation growth model to predict cleaning period.

[0155] Among them, the fault identification module includes a string topology optimization unit for dynamically optimizing the string topology structure of the photovoltaic power station according to the real-time operation state of the power station, the equipment health data and the fault mode identification result, and the string topology optimization unit comprises:

[0156] Scan the irradiance distribution map of the power station, and identify the area with power generation efficiency lower than the power generation threshold;

[0157] Switch the shadow area component to an independent MPPT channel through a matrix switch array to form a dynamic subarray;

[0158] Based on the deep deterministic policy gradient (DDPG) algorithm, the Pareto optimality is realized among the power generation efficiency, heat loss and equipment life;

[0159] The optimization strategy is recalculated every 60 minutes, and the networking reconstruction response time is less than or equal to 30 seconds.

[0160] The system for monitoring the real-time operation state of the photovoltaic power station further comprises a self-calibration mechanism, and the self-calibration mechanism performs the following operations:

[0161] Every 20ms, the temperature drift error is eliminated through mutual checking algorithm of adjacent sensor data;

[0162] Offline calibration is performed every morning, and a laser interferometer is used to correct the reference value of micro-crack detection;

[0163] When the dust thickness is greater than 3g / m2, the pre-order work order of the unmanned aerial vehicle cleaning system is automatically triggered.

[0164] The string topology optimization and fault mode identification mechanism of the present application can not only improve the power generation efficiency of the photovoltaic power station, but also reduce the heat loss of the equipment and prolong the service life of the equipment. Through intelligent optimization and management, the system improves the power generation benefit of the photovoltaic power station while reducing the loss of the equipment, thereby prolonging the service life of the power station equipment, which has high long-term economic value; combined with edge computing and cloud big data processing, real-time data processing and control decision can be realized on the spot of the photovoltaic power station, and the self-calibration mechanism and the fault identification module of the system enable the power station to automatically optimize the system and give fault warning without human intervention, greatly reducing the labor cost and improving the intelligent level of system operation.

[0165] Through the above detailed steps, the system for monitoring the operation state of a photovoltaic power station in real time of the present application is used to execute the steps of the method for monitoring the operation state of a photovoltaic power station in real time in the above embodiments, which will not be described here again. The method and system for monitoring the operation state of a photovoltaic power station in real time provided by the present application can efficiently, accurately and intelligently monitor the operation state of the power station, predict and prevent faults, optimize the operation configuration of the power station, ensure the stable and efficient operation of the photovoltaic power station, and improve the economic and environmental benefits of photovoltaic power generation.

[0166] The above is the exemplary embodiment disclosed by the present application, but it should be noted that various changes and modifications can be made without departing from the scope of the embodiments disclosed by the present application defined by the claims. The functions, steps and / or acts of the method claims described herein need not be performed in any particular order. Furthermore, although the elements of the embodiments disclosed by the present application can be described or claimed in individual form, unless explicitly restricted, they can also be implemented in multiple forms.

[0167] It should be understood that, as used herein, the singular forms "a", "an" and "the" are intended to include plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising", when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. The embodiment number of the embodiments disclosed by the present application is only for description, and does not represent the advantages and disadvantages of the embodiments.

[0168] Those skilled in the art should understand that the above discussion of any embodiment is only exemplary and is not intended to imply that the scope of the embodiments disclosed by the present application (including claims) is limited to these examples; under the idea of the embodiments of the present application, the technical features of the above embodiments or different embodiments can also be combined, and there are many other changes of the different aspects of the embodiments of the present application as described above. In order to be brief, they are not provided in detail. Therefore, any omissions, modifications, equivalent replacements, improvements, etc. made in the spirit and principles of the embodiments of the present application shall be included in the protection scope of the embodiments of the present application.

Claims

1. A method for monitoring the operating state of a photovoltaic power plant in real time, characterized in that, The method comprises the following steps: Integrating a multi-modal sensor array in the junction box of a photovoltaic assembly to collect sensor data of the photovoltaic power station in real time; Performing local preprocessing on the sensor data through an edge computing node deployed on site of the photovoltaic power station, and transmitting data of different priorities to the cloud based on a dynamic hybrid communication protocol; Fusing the sensor data in the cloud to build a four-dimensional digital twin model, and fusing spatiotemporally continuous weather forecast, component aging and dust evolution data; Learning historical fault data using a transfer learning framework to identify fault modes of the photovoltaic power station, and dynamically optimizing the string topology of the photovoltaic power station in combination with real-time operation status, equipment health data and fault mode identification results of the photovoltaic power station; Presenting the optimized string topology of the photovoltaic power station through a visual interface, and feeding back the health status and control instructions of the power station in real time, wherein the control instructions are transmitted to the edge computing node on site in real time; The edge computing node performs the following steps: Running a lightweight ResNet-8 model to complete real-time classification of faults within a preset time; Compressing multi-dimensional sensor data using sparse coding, and dividing data priorities according to fault confidence, wherein emergency alarm data is marked as a priority level to trigger straight-through transmission of data to the cloud through LoRa Mesh, 5G slicing or satellite link; When transmitting data of different priorities to the cloud based on the dynamic hybrid communication protocol, the dynamic hybrid communication protocol satisfies the following conditions: The coverage radius of the LoRa Mesh network is ≥1km, and the power consumption of a single node is ≤0.3W; The end-to-end delay of the 5G slicing network is ≤8ms, and 1000 concurrent terminals per square kilometer are supported; The satellite link adopts dual-mode redundancy of Beidou-3 and Starlink, and automatically switches when communication is interrupted, with a switching delay ≤3s; When fusing the sensor data in the cloud to build a four-dimensional digital twin model, the following steps are included: Building a three-dimensional power station model according to the spatial dimension, integrating physical parameters and electrical connection relationships of each component; Fusing weather forecast data of a preset time period according to the time dimension, including irradiance fluctuation, wind speed variation and precipitation probability; According to the state dimension, calculating the component aging index through EL image degradation analysis to predict the remaining life; According to the environmental dimension, building a dust growth model based on PM2.5 concentration and rainfall frequency to predict the cleaning period; When dynamically optimizing the string topology of the photovoltaic power station, the following steps are included: Real-time scanning of the irradiance distribution map of the power station to identify areas with power generation efficiency lower than a threshold; Switching components in the shadow area to independent MPPT channels through a matrix switch array to form a dynamic sub-array; Based on deep deterministic policy gradient, achieving Pareto optimality among power generation efficiency, heat loss and equipment life; Recomputing the optimization strategy every 60 minutes, and the network reconstruction response time is ≤30s.

2. The method for monitoring the operation state of a photovoltaic power station in real time according to claim 1, characterized in that, The sensor data of the photovoltaic power station includes electrical parameters, physical states and environmental data; when the sensor data is locally preprocessed, it includes anomaly detection, data compression and priority marking; different priority data is transmitted to the cloud through LoRa Mesh, 5G slice or satellite link.

3. The method for monitoring the operation state of a photovoltaic power station in real time according to claim 2, characterized in that, The multi-modal sensor array includes a graphene quantum dot sensor, a piezoelectric vibration sensor and a polarized light sensor, which synchronously detects the component operating voltage, leakage current, micro-crack, dust thickness, surface temperature, polarization angle offset and hidden crack depth.

4. The method for monitoring the operation state of a photovoltaic power station in real time according to claim 1, characterized in that, When learning historical fault data using a transfer learning framework to identify fault modes of the photovoltaic power station, the implementation of the transfer learning framework includes: A feature vector knowledge base containing several historical fault modes is constructed; When an unknown anomaly is detected, the topological similarity is matched through a graph convolution network to realize zero-shot detection; Each week, the global model parameters are updated through federated learning to aggregate data from each power station while preserving local privacy data.

5. The method for monitoring the operation state of a photovoltaic power station in real time according to claim 4, characterized in that, The method for monitoring the operating state of the photovoltaic power station in real time also includes a self-calibration mechanism: Every 20 ms, the temperature drift error is eliminated through mutual checking algorithm of adjacent sensor data; Every morning, offline calibration is performed to correct the reference value of micro-crack detection using a laser interferometer; When the dust thickness is greater than 3 g / m², the pre-order work order of the unmanned aerial vehicle cleaning system is automatically triggered.

6. A system for monitoring the operating state of a photovoltaic power plant in real time, characterized in that it comprises: The system for monitoring the operating state of the photovoltaic power station in real time includes: A data acquisition module for acquiring sensor data of the photovoltaic power station in real time by integrating a multi-modal sensor array in the junction box of the photovoltaic component; An edge computing module for locally preprocessing the sensor data through an edge computing node deployed on site of the photovoltaic power station, and transmitting different priority data to the cloud based on a dynamic hybrid communication protocol; A model construction module for fusing the sensor data in the cloud to construct a four-dimensional digital twin model, and fusing spatiotemporal continuous weather prediction, component aging and dust evolution data; A fault identification module for learning historical fault data using a transfer learning framework to identify fault modes of the photovoltaic power station, and dynamically optimizing the string topology structure of the photovoltaic power station based on the real-time operating state, equipment health data and fault mode identification results of the photovoltaic power station; A visual display module for presenting the optimized string topology structure of the photovoltaic power station through a visual interface, and feeding back the health state and control instructions of the power station in real time, and the control instructions are sent to the edge computing node on site in real time.

Images