Photovoltaic power station monitoring alarm method, device, equipment and storage medium

Abstract

The invention discloses a photovoltaic power station monitoring alarm method, device and equipment and a storage medium. The method comprises the following steps: obtaining the physical structure importance degree and the fault influence degree of each device in the photovoltaic power station; judging whether each piece of equipment is a key node or not by taking the physical structure importance degree and the fault influence range of the equipment as judgment indexes; carrying out data acquisition at acquisition frequencies of different preset frequencies for the equipment of the key node and the non-key node; matching a transmission path meeting a first constraint condition for each device in the photovoltaic power station, transmitting the collected data of each device to an edge node pointed by the corresponding transmission path, extracting feature data, and uploading the feature data to a cloud; and matching the feature data extracted by each device in the photovoltaic power station with a fault detection module which meets a second constraint condition and is arranged in the cloud, obtaining whether each device has a fault or not and the fault type and grade, and generating a warning prompt. The data acquisition strategy of the photovoltaic power station can be optimized, and the alarm accuracy can be improved.

Description

Technical Field

[0002] This application relates to the technical field of photovoltaic power station monitoring and warning, and specifically relates to a method, device, equipment and storage medium for monitoring and warning of a photovoltaic power station. Background Art

[0003] With the continuous growth of the global demand for renewable energy, photovoltaic power stations, as an important part of green energy, have been continuously expanding in scale and complexity. In order to ensure the efficient operation of photovoltaic power stations, monitoring and fault warning technologies have received extensive attention. These technologies can effectively improve the operation and maintenance efficiency of power stations, reduce downtime, and provide important guarantees for the stability and sustainability of energy supply by collecting and analyzing real-time data of equipment in photovoltaic power stations.

[0004] In the prior art, the monitoring of photovoltaic power stations mainly relies on periodic manual inspections and automatic alarm systems based on fixed thresholds. Specifically, technicians usually deploy a sensor network to collect the operating parameters of equipment such as photovoltaic arrays and inverters, and upload the collected data to a central control platform for analysis. At the same time, whether there are abnormal situations is judged through preset static thresholds, and once data beyond the threshold range is detected, the alarm mechanism will be triggered. In addition, recently, the technology of collaborating edge nodes with the cloud has gradually emerged, and some data processing tasks are offloaded to edge nodes for execution to improve data processing speed and response efficiency.

[0005] However, the above traditional methods have obvious limitations. On the one hand, due to collecting data at a fixed frequency and not distinguishing between key nodes and ordinary nodes, the data volume is huge and the transmission efficiency is low, thus affecting the real-time performance of fault detection. On the other hand, the alarm mechanism based on static thresholds is difficult to adapt to complex and changeable actual working conditions, and false alarms or missed alarms are likely to occur, further increasing the operation and maintenance costs. In addition, in practical applications, due to the failure to fully consider possible abnormal problems in the data transmission process, such as data loss, noise interference or transmission delay, etc., the data that has been preliminarily processed by edge nodes may still have deviations, thus affecting the final data detection accuracy.

[0006] Therefore, how to optimize the data collection strategy and improve the alarm accuracy has become an urgent problem to be solved. Summary of the Invention

[0007] In order to optimize the data collection strategy of a photovoltaic power station and improve the alarm accuracy, this application provides a method, device, equipment and storage medium for monitoring and warning of a photovoltaic power station.

[0008] In a first aspect, this application provides a method for monitoring and warning of a photovoltaic power station, including: Regularly perform physical topology modeling and historical fault correlation analysis on the photovoltaic power station to obtain the physical connection relationship and fault propagation correlation relationship between devices in the photovoltaic power station; Based on the obtained physical connection relationship and fault propagation correlation relationship, determine the physical structure importance degree and fault impact degree of each device; use the physical structure importance degree and fault impact degree of the device as judgment indicators to dynamically judge whether each device is a key node; for the device judged as a key node, collect data at the acquisition frequency of the first preset frequency, and for the device judged as a non-key node, collect data at the second preset frequency less than the first preset frequency; For each device in the photovoltaic power station, match a transmission path that meets the first constraint condition, and transmit the collected data of each device to the edge node pointed to by the corresponding transmission path; the first constraint condition is that the comprehensive score based on the distance of each device from the pointed edge node, the load of the pointed edge node, and the frequency of transmission to the pointed edge node is greater than the first preset score; use the lightweight model built in the pointed edge node to extract the characteristic data of the device and upload it to the cloud; Match the characteristic data extracted for each device in the photovoltaic power station with a fault detection module built in the cloud that meets the second constraint condition; the second constraint condition is that the comprehensive score based on the load of the pointed fault detection module and the frequency of transmission to the pointed fault detection module is greater than the second preset score; use the matched fault detection model to obtain whether each device has a fault and the fault type and level; Generate a warning prompt according to whether each device has a fault and the fault type and level obtained.

[0009] By adopting the above solution, obtain the physical connection relationship and fault propagation correlation relationship between devices, dynamically judge key nodes based on the physical structure importance degree and fault impact degree, set different data acquisition frequencies for different nodes, optimize the data acquisition strategy of the photovoltaic power station, and improve the detection efficiency; and match the transmission path and fault detection module that meet the constraint conditions, which not only ensures the accuracy of data transmission but also the flexibility of fault detection, and improves the accuracy of alarm.

[0010] Preferably, it further includes: Obtain the structural information of each device in the photovoltaic power station; According to the preset device structure information - device key monitoring parameter - sensor information mapping table, obtain the device key monitoring parameter information and sensor information corresponding to each device; the sensor information includes sensor type and sensor deployment information; the preset device structure information - device key monitoring parameter - sensor information mapping table is obtained by statistical analysis of the historical structural information of each device and the historical key monitoring parameters and historical sensor information that are matched and verified to meet the preset requirements for fault location accuracy and location timeliness; Control the actual sensor deployment and operation according to the acquired sensor information; According to the acquired key monitoring parameter information and sensor information corresponding to each device, collect and obtain the key parameters of the corresponding characteristics and quantities of each device.

[0011] By adopting the above solution, determine the key monitoring parameters and sensor information of devices with different structures based on a preset mapping table, so as to collect accurate and timely key parameters, providing a reliable data basis for subsequent fault detection and warning.

[0012] Preferably, it further includes: Obtain environmental data for each device in the photovoltaic power station; Match a transmission path that meets the first constraint condition under the first environmental basis for each device in the photovoltaic power station; the first constraint condition under the first environmental basis is that under the first environmental basis, according to the distance of each device from the edge node it points to, the load of the edge node it points to, and the frequency of transmission to the edge node it points to, the comprehensive score is greater than the first preset score; the first environmental basis refers to the environmental conditions with a similarity greater than the first preset similarity to the current environmental data; Match the feature data extracted for each device in the photovoltaic power station with a built-in fault detection module in the cloud that meets the second constraint condition under the second environmental basis; the second constraint condition under the second environmental basis is that under the second environmental basis, according to the load of the fault detection module it points to and the frequency of transmission to the fault detection module it points to, the comprehensive score is greater than the second preset score; the second environmental basis refers to the environmental conditions with a similarity greater than the second preset similarity to the current environmental data.

[0013] By adopting the above solution, considering the influence of environmental factors on detection, filter out the constraint conditions under the similar environmental basis according to the similarity of the current environmental data, and then dynamically adjust the selection conditions of the transmission path and the fault detection module, improving the adaptability and accuracy of data transmission and fault detection.

[0014] Preferably, it further includes: When the number of transmission paths that meet the first constraint condition for each device in the photovoltaic power station is multiple, randomly select a preset number of transmission paths as a verification group; Establish communication transmissions between the edge nodes pointed to in the verification group, and mutually transmit the feature data of the devices extracted by each edge node; Calculate the feature similarity between the feature data of the devices extracted by every two edge nodes, Eliminate the feature data of devices whose feature similarity with the feature data of other edge node extraction devices is greater than the preset feature similarity, fuse the feature data of the remaining edge node extraction devices that are retained, and upload the fused feature data of the devices to the cloud.

[0015] By adopting the above solution, multiple transmission paths that meet the constraint conditions are obtained and a verification group is generated. The feature data extracted from the multiple transmission paths are mutually verified and fused, effectively reducing data redundancy, reducing the transmission burden, improving the data processing efficiency, and ensuring that the data received by the cloud is more representative and accurate.

[0016] Preferably, it further includes: When collecting data from the device determined to be a key node at the acquisition frequency of the first preset frequency, collect the image data of the device determined to be a key node at the same acquisition frequency; when collecting data from the device determined not to be a key node at the acquisition frequency of the second preset frequency less than the first preset frequency, collect the image data of the device determined not to be a key node at the same acquisition frequency; For the electrical digital data collected from each device, compare the similarity of the electrical digital data of the device predicted by the lightweight model built in the edge node according to the historical direction. If the similarity is lower than the third preset similarity, then correspondingly transmit the synchronously collected image data to the currently pointed edge node; combine the electrical digital data and image data collected from each device, and use the lightweight model built in the pointed edge node to extract the feature data of the device and upload it to the cloud.

[0017] By adopting the above solution, compare the similarity between the actually collected electrical digital data and the model predicted data, judge that the currently collected electrical data may be abnormal, and then transmit the synchronously collected image data to the cloud together, and use multimodal fusion to improve the accuracy of fault detection.

[0018] Preferably, it further includes: When extracting the feature data of the device using the lightweight model built into the pointed edge node and uploading it to the cloud, the feature data extracted from the electrical digital data of the device predicted according to the lightweight model built into the pointed edge node is synchronously uploaded to the cloud; when it is determined at the current moment that the pointed edge node is the same as the pointed edge node at the previous moment, for the electrical digital data collected for each device, compare the similarity of the electrical digital data of the device predicted according to the lightweight model built into the pointed edge node at the previous moment. If the similarity is not lower than the third preset similarity, there is no need to transmit the electrical digital data collected by the current device, and directly use the electrical digital data of the device predicted according to the lightweight model built into the pointed edge node at the previous moment as the electrical digital data collected at the current moment, and correspondingly match the built-in fault detection module in the cloud that meets the second constraint condition, and use the matched fault detection model to obtain whether each device is faulty and the type and level of the fault.

[0019] By adopting the above solution, the electrical digital data of the device is predicted using the lightweight model built into the edge node, and the similarity of the predicted data is uploaded to the cloud. There is no need to transmit the actual collected data, and the predicted data is directly used for fault detection, effectively reducing the data transmission volume, reducing the network load, improving the data processing efficiency at the same time, and ensuring the real-time and accuracy of fault detection.

[0020] Preferably, it further includes: According to the obtained fault detection results of each device, regularly determine the fault frequency of each device, and use the determined fault frequency of each device as the first feedback data; According to the obtained fault detection results of each device, determine whether there is a standby device that allows immediate switching for each faulty device as the second feedback data; According to the first feedback data and the second feedback data, correspondingly adjust the weights of the physical structure importance degree and the fault impact degree of each device; among them, the lower the fault frequency of each device, the lower the corresponding weight of the fault impact degree; for each faulty device with a standby device that allows immediate switching, correspondingly optimize the weights of the physical structure importance degree and the fault impact degree, avoiding unnecessary resource allocation.

[0021] By adopting the above solution, the fault frequency of the device and whether there is a standby device that allows immediate switching for the faulty device are used as feedback data, optimizing the physical structure importance degree and the weight In a second aspect, the present application provides a photovoltaic power station monitoring and warning system, including: A photovoltaic device relationship acquisition module, which is used to regularly perform physical topology modeling and historical fault correlation analysis on the photovoltaic power station, and obtain the physical connection relationship and fault propagation correlation relationship between the devices in the photovoltaic power station; The photovoltaic device data acquisition module is used to determine the physical structure importance degree and fault influence degree of each device based on the obtained physical connection relationship and fault propagation association relationship; taking the physical structure importance degree and fault influence degree of the device as judgment indicators, dynamically judge whether each device is a key node; for the device judged as a key node, data is collected at the acquisition frequency of the first preset frequency, and for the device judged as a non-key node, data is collected at the acquisition frequency of the second preset frequency which is less than the first preset frequency; The photovoltaic device feature extraction module is used to match a transmission path that meets the first constraint condition for each device in the photovoltaic power station, and transmit the collected data of each device to the edge node pointed to by the corresponding transmission path; the first constraint condition is that the comprehensive score based on the distance of each device from the pointed edge node, the load of the pointed edge node, and the frequency of transmission to the pointed edge node is greater than the first preset score; the lightweight model built in the pointed edge node is used to extract the feature data of the device and upload it to the cloud; The photovoltaic device fault detection module is used to match the fault detection module built in the cloud that meets the second constraint condition for the feature data extracted from each device in the photovoltaic power station; the second constraint condition is that the comprehensive score based on the load of the pointed fault detection module and the frequency of transmission to the pointed fault detection module is greater than the second preset score; the matching fault detection model is used to obtain whether each device has a fault and the fault type and level; The photovoltaic device fault warning module is used to generate a warning prompt according to whether each device has a fault and the fault type and level obtained.

[0022] By adopting the above scheme, the optimization of the data acquisition strategy of the photovoltaic power station is realized, the necessity of data acquisition is optimized, and the accuracy of fault detection and warning is improved.

[0023] In a third aspect, the present application provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method as described above.

[0024] In a fourth aspect, the present application provides a computer device, which includes a memory, a processor, and a program stored and executable on the memory. When the program is executed by the processor, it implements the steps of the method as described above.

[0025] In summary, the present application has the following beneficial effects: 1. By regularly obtaining the physical connection relationships and fault propagation correlation relationships among various devices, dynamically determining key nodes, and adopting a differential data acquisition frequency, unnecessary data transmission volume is reduced, and data processing efficiency is improved; matching edge nodes and fault detection modules that meet the constraint conditions, reducing data error transmission, using lightweight models built into the edge nodes for feature extraction, and using the fault detection modules to accurately detect faults in photovoltaic power station devices; 2. Based on the device data corresponding to the accuracy and timeliness of fault location of each device that met requirements in history, constructing a device structure - key monitoring parameter - sensor information mapping table to guide the deployment and operation of sensors, collecting key parameters with corresponding characteristics and quantities, and ensuring the comprehensiveness and accuracy of data acquisition; 3. Considering environmental factors, using similarity analysis technology to match edge nodes and fault detection modules in an environment base similar to the current environmental data, improving the adaptability and accuracy of data transmission and fault detection. Description of the Drawings

[0026] Figure 1 It is a flowchart of the photovoltaic power station monitoring and warning method described in the specific embodiment; Figure 2 It is a schematic structural diagram of the photovoltaic power station monitoring and warning system described in the specific embodiment. Detailed Embodiment

[0027] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0028] As Figure 1 shown, an embodiment of the present application discloses a photovoltaic power station monitoring and warning method, and the specific steps include: S1. Regularly obtain the relationships among various devices in the photovoltaic power station.

[0029] To optimize data acquisition in the photovoltaic power station, avoid unnecessary data acquisition, and reduce data calculation volume, analyze the relationships among various devices in the photovoltaic power station, including: physical connection relationships and actual operation correlation relationships, etc.

[0030] Specifically, regularly perform physical topology modeling on the photovoltaic power station to obtain the physical connection relationships (such as electrical connection topologies) among various devices (such as: photovoltaic module arrays, inverters, busbar boxes, and grid interfaces, etc.) in the photovoltaic power station.

[0031] Based on the constructed physical topology, clarify the physical connections and dependencies between devices, thereby obtaining and analyzing historical fault data, identifying the fault propagation paths and patterns between devices in each photovoltaic power station, constructing a fault correlation network. For example, the nodes in the network represent devices, the edges represent the fault propagation relationships between devices, and the weights of the edges reflect the frequency or intensity of fault propagation, so as to obtain the fault propagation correlation relationships between devices in the photovoltaic power station.

[0032] S2. Based on the obtained physical connection relationships and fault propagation correlation relationships, dynamically determine whether each device in the photovoltaic power station is a key node, and determine the acquisition frequency of each device in the photovoltaic power station according to the judgment results.

[0033] Specifically, the judgment indicators for whether each device is a key node include: the importance degree of the physical structure, the degree of fault impact, etc. The weights of each indicator can be dynamically adjusted according to the device fault feedback data.

[0034] Adopt a neural network algorithm or set the judgment rules for the importance degree of the physical structure to determine the importance degree of the physical structure of each device in the photovoltaic power station, which is reflected in the form of a score. For example: set the judgment rules for the importance degree of the physical structure: according to the obtained physical connection relationships, divide the electrical correlation groups. For the devices in the same group, the more physical connection edges or the devices belonging to the physical connection center, the higher the corresponding score of the importance degree of the physical structure. For example, divide the photovoltaic module arrays under the same combiner box or inverter into the same group, and assign the first score to the physical structure importance degree of the photovoltaic modules at the center and edge of the photovoltaic module array in the same group, and the score of the importance degree of the physical structure of the remaining photovoltaic modules is lower than the first score; or use the constructed first neural network model, the input of the model is the physical connection relationships obtained regularly, and the output is the importance degree of the physical structure of each device, which is trained and generated by the physical connection relationships of each device in the historical photovoltaic power station and the scores of the importance degree of the physical structure of each device in the historically labeled photovoltaic power station.

[0035] Correspondingly, adopt a neural network algorithm or set the judgment rules for the degree of fault impact to determine the degree of fault impact of each device in the photovoltaic power station, which is displayed in the form of a score. For example: use the constructed second neural network model, the input of the model is the fault propagation correlation relationships obtained regularly, and the output is the degree of fault impact of each device, which is trained and generated by the fault propagation correlation relationships of each device in the historical photovoltaic power station and the scores of the degree of fault impact of each device in the historically labeled photovoltaic power station.

[0036] Using the importance degree of the physical structure of the device and the impact degree of the fault as judgment indicators, dynamically judge whether each device is a key node; that is, regularly calculate the weighted scores of the importance degree of the physical structure of each device and the impact degree of the fault of each device, and compare whether the comprehensive score after weighted calculation is greater than the preset comprehensive score. If it is greater, it is judged as a key node, otherwise it is a non-key node.

[0037] In addition, in order to more accurately judge key nodes, the judgment indicators for whether each device is a key node may also include: device service life, environmental exposure degree, etc., and set the scoring rules for the device service life and the scoring rules for the environmental exposure degree accordingly; correspondingly, regularly obtain the service life of each device, and determine the score of the service life of each device according to the service life scoring rules; regularly obtain the environment where each device is located, and determine the score of the environmental exposure degree of each device according to the scoring rules of the environmental exposure degree; calculate the weighted comprehensive score and judge whether it is a key node.

[0038] Obtain the judgment results of each device, and perform data collection on the devices judged as key nodes at the acquisition frequency of the first preset frequency, and perform data collection on the devices judged as non-key nodes at the acquisition frequency of the second preset frequency less than the first preset frequency; the first preset frequency and the second preset frequency can be set manually.

[0039] S3. Match a transmission path that meets the first constraint condition for each device in the photovoltaic power station, and transmit the collected data of each device to the edge node pointed to by the corresponding transmission path, and upload it to the cloud after feature extraction.

[0040] Considering the situation that data transmission errors may occur due to edge node failures, which may further lead to incorrect detection results, select edge nodes with good matching performance and cross-switch edge nodes for data transmission to avoid the above situation. Specifically, the following measures are taken: First, match a transmission path that meets the first constraint condition for each device in the photovoltaic power station, and transmit the collected data of each device to the edge node pointed to by the corresponding transmission path; where the first constraint condition is that the comprehensive score based on the distance of each device from the pointed edge node, the load of the pointed edge node, and the frequency of transmission to the pointed edge node (or the continuous frequency of transmission to the pointed edge node) is greater than the first preset score; the calculation formula for the corresponding comprehensive score is as follows: F i =αF(d i )+βF(L i )+γF(f i ) In the formula, F(d i) is the distance scoring function from each device to the edge node it points to. Different scores are judged according to different distance ranges. For example, the first distance range, the second distance range, and the third distance range (increasing gradually) correspond to the judgments of 90, 70, and 50; F(L i ) is the load scoring function of the edge node it points to. Different scores are judged according to different load ranges. For example, the first load range, the second load range, and the third load range (increasing gradually) correspond to the judgments of 90, 70, and 50; F(f i ) is the scoring function of the frequency of the same device transmitting data to the edge node it points to. Different scores are judged according to different frequency ranges. For example, the first frequency range, the second frequency range, and the third frequency range (increasing gradually) correspond to the judgments of 90, 70, and 50; different scores are judged according to different frequency ranges; α, β, and γ are weights.

[0041] Secondly, use the lightweight model built in the edge node it points to to extract the feature data of the device and upload it to the cloud; specifically, algorithms such as fast Fourier transform (FFT) or wavelet transform (WT) can be used to extract key features from the collected data.

[0042] S4. Match the feature data extracted from each device in the photovoltaic power station with the fault detection module built in the cloud that meets the second constraint condition, complete the fault detection, and output whether each device has a fault and the fault type and level.

[0043] Considering that the detection result may be incorrect due to the fault of the fault detection module, select a fault detection module with good matching performance and cross-switch the edge nodes for data transmission to avoid the above situation. Specifically, the following measures are taken: First, match the feature data extracted from each device in the photovoltaic power station with the fault detection module built in the cloud that meets the second constraint condition; among them, the second constraint condition is that the comprehensive score based on the load of the pointed fault detection module and the frequency of transmitting data to the pointed fault detection module is greater than the second preset score; the calculation formula for the corresponding comprehensive score is as follows: F j =σF(L j )+ρF(f j ) In the formula, F(L j ) is the load scoring function of the pointed fault detection module. Different scores are judged according to different load ranges. For example, the fourth load range, the fifth load range, and the sixth load range (increasing gradually) correspond to the judgments of 90, 70, and 50; F(f j ) is the scoring function of the frequency of the same device transmitting feature data to the pointed fault detection module. Different scores are judged according to different frequency ranges. For example, the fourth frequency range, the fifth frequency range, and the sixth frequency range (increasing gradually) correspond to the judgments of 90, 70, and 50; σ and ρ are weights.

[0044] Secondly, use the matched fault detection model to obtain whether each device has a fault, as well as the fault type and level; the fault detection model adopts a deep neural network S5. Generate a warning prompt according to whether each device has a fault, as well as the fault type and level obtained

[0045] In a specific embodiment, considering photovoltaic modules with different structures, a mapping table can be constructed through historical data to guide the deployment and operation of sensors, ensuring the comprehensiveness and accuracy of data collection, and achieving more accurate and timely positioning. The method includes: Obtain the structural information of each device in the photovoltaic power station; the structural information of each device includes: the type of photovoltaic module array (series array, parallel array, hybrid structure, BL style, etc.), the structural type of a single photovoltaic module (fixed inclination angle, single-axis tracking, bifacial module, etc.), centralized inverter or string inverter, etc.

[0046] According to the preset mapping table of device structural information - device key monitoring parameters - sensor information, obtain the device key monitoring parameter information and sensor information corresponding to each device. Among them, the key monitoring parameter information includes: electrical characteristic data such as current and voltage, temperature, inclination, etc.; the sensor information includes the sensor type (temperature sensor, voltage / current sensor, inclination angle sensor, etc.), sensor deployment information (such as deployment location); among them, the preset mapping table of device structural information - device key monitoring parameters - sensor information is obtained through statistical analysis of the structural information of each historical device and the historical key monitoring parameters and historical sensor information that are matched and verified to meet the preset requirements for fault location accuracy and positioning timeliness; for example: series-parallel type photovoltaic array - branch voltage deviation - voltage sensor, every two adjacent photovoltaic branches are used as a detection unit, and the connection point between two adjacent photovoltaic modules in each branch is regarded as a node, and voltage sensors are connected to two parallel nodes in the same detection unit.

[0047] Control the actual sensor deployment and operation according to the obtained sensor information; specifically, remotely control the sensors actually deployed at the corresponding types and positions of each device to operate according to the obtained sensor type and deployment location.

[0048] Collect and obtain the key parameters of the corresponding characteristics and quantities of each device according to the device key monitoring parameter information and sensor information corresponding to each device obtained, so as to ensure the comprehensiveness and accuracy of data collection, and provide a reliable data basis for subsequent fault detection and alarm.

[0049] In a specific embodiment, considering that environmental factors can affect the accuracy of data transmission and data detection, edge nodes and detection modules with good performance under the same environmental conditions are selected to improve the accuracy of data transmission and data detection. The method further includes: Obtain environmental data for each device in the photovoltaic power station; specifically, collect environmental data using environmental sensors around the location of each device, including temperature, humidity, wind speed, etc.

[0050] Match a transmission path that meets the first constraint condition under the first environmental basis for each device in the photovoltaic power station; the first constraint condition under the first environmental basis is that under the first environmental basis, the comprehensive score based on the distance from each device to the edge node it points to, the load of the edge node it points to, and the frequency of transmission to the edge node it points to is greater than the first preset score; where the first environmental basis refers to environmental conditions with a similarity greater than the first preset similarity to the current environmental data, and the calculation formula for the corresponding comprehensive score is as follows: F i =αF(d i , τ)+βF(L i , τ)+γF(f i , τ) In the formula, F(d i , τ) is the distance scoring function of each device from the edge node it points to under the first environmental basis. Different scores are judged for different distance ranges under the first environmental basis. For example, for the first distance range, the second distance range, and the third distance range (increasing gradually) under the first environmental basis, the corresponding scores are 85, 75, and 55; F(L i , τ) is the load scoring function of the edge node pointed to under the first environmental basis. Different scores are judged for different load ranges under the first environmental basis. For example, for the first load range, the second load range, and the third load range (increasing gradually) under the first environment, the corresponding scores are 90, 80, and 60; F(f i , τ) is the scoring function of the frequency of the same device transmitting data to the edge node it points to under the first environmental basis. Different scores are judged for different frequency ranges under the first environmental basis. For example, for the first frequency range, the second frequency range, and the third frequency range (increasing gradually), the corresponding scores are 80, 60, and 50.

[0051] Match a built-in fault detection module in the cloud that meets the second constraint condition under the second environmental basis for the feature data extracted from each device in the photovoltaic power station; the second constraint condition under the second environmental basis is that under the second environmental basis, the comprehensive score based on the load of the pointed fault detection module and the frequency of transmission to the pointed fault detection module is greater than the second preset score; the second environmental basis refers to environmental conditions with a similarity greater than the second preset similarity to the current environmental data, and the calculation formula for the corresponding comprehensive score is as follows: F j = σF(L j , τ) + ρF(f j , τ) In the formula, F(L j , τ) is the load scoring function of the fault detection module pointed to under the second environmental basis. Different scores are judged corresponding to different load ranges. For example, the fourth load range, the fifth load range, and the sixth load range (increasing gradually) under the second environmental basis correspond to the judgments of 95, 80, and 60; F(f j , τ) is the scoring function of the frequency of transmitting the feature data of the same device to the pointed fault detection module under the second environmental basis. Different scores are judged corresponding to different frequency ranges. For example, under the second environmental basis, the fourth frequency range, the fifth frequency range, and the sixth frequency range (increasing gradually) correspond to the judgments of 80, 60, and 55; σ and ρ are weights.

[0052] In a specific embodiment, considering that when there are multiple transmission paths to be matched, how to further verify and screen out the transmission path with better transmission performance and higher transmission accuracy, so as to ensure the accuracy of the extracted features, the method includes: When the number of transmission paths matched by each device in the photovoltaic power station that meet the first constraint condition is multiple, randomly select a preset number of transmission paths as the verification group; in this embodiment, the preset number is 3.

[0053] Establish communication transmissions between the edge nodes pointed to in the verification group, and transmit the feature data of the devices extracted by each edge node to each other. The corresponding edge node will store the special feature data transmitted corresponding to the edge nodes in the verification group for subsequent query and retrieval. To avoid occupying too much storage resources of the edge nodes, the stored content will be deleted regularly.

[0054] Calculate the feature similarity between the feature data of the devices extracted by every two edge nodes, eliminate the feature data of the devices whose feature similarity with the feature data of the devices extracted by other edge nodes is greater than the preset feature similarity, fuse the feature data of the devices extracted from the remaining edge nodes, and upload the fused feature data of the devices to the cloud or randomly select the feature data of the devices extracted by one edge node from the feature data of the devices extracted from the remaining edge nodes or select the feature data of the devices extracted by the edge node with a higher comprehensive score from the feature data of the devices extracted from the remaining edge nodes.

[0055] In a specific embodiment, in order to further improve the accuracy of fault detection of each device in the photovoltaic power station and balance the data processing efficiency and the accuracy of data detection, set conditions and obtain multi-modal data under specific conditions and perform fault detection based on the collected multi-modal data. The method includes: Collect image data. Specifically, when collecting data from devices determined to be critical nodes at a first preset frequency, collect the image data of the devices determined to be critical nodes at the same collection frequency; when collecting data from devices determined to be non-critical nodes at a second preset frequency less than the first preset frequency, collect the image data of the devices determined to be non-critical nodes at the same collection frequency.

[0056] For the electrical digital data collected from each device (i.e., the collected electrical data, such as current, voltage, etc.), compare the similarity of the electrical digital data of the device predicted by the lightweight model built into the edge node pointed to according to the history. If the similarity is lower than the third preset similarity, it indicates that there may be a large error in the currently collected electrical digital data. Therefore, select to synchronously transmit the image data to assist in fault detection, and correspondingly transmit the synchronously collected image data to the currently pointed edge node; combine the electrical digital data and image data collected from each device, and use the lightweight model built into the pointed edge node to extract the feature data of the device and upload it to the cloud.

[0057] In addition, in order to further reduce the data transmission volume and network load while ensuring the accuracy of data transmission, the method further includes: When using the lightweight model built into the pointed edge node to extract the feature data of the device and upload it to the cloud, synchronously upload the feature data extracted from the electrical digital data of the device predicted by the lightweight model built into the pointed edge node to the cloud; when it is determined that the edge node pointed to at the current moment is the same as the edge node pointed to at the previous moment, for the electrical digital data collected from each device, compare the similarity of the electrical digital data of the device predicted by the lightweight model built into the pointed edge node at the previous moment. If the similarity is not lower than the third preset similarity, there is no need to transmit the electrical digital data collected from the current device, and directly use the electrical digital data of the device predicted by the lightweight model built into the pointed edge node at the previous moment as the electrical digital data collected at the current moment, and correspondingly match the fault detection module built into the cloud that meets the second constraint condition, and use the matched fault detection model to obtain whether each device has a fault and the fault type and level.

[0058] In a specific embodiment, a feedback mechanism is introduced to optimize the data collection strategy, and the method further includes: According to the obtained fault detection results of each device, regularly determine the fault frequency of each device, and use the determined fault frequency of each device as the first feedback data.

[0059] According to the obtained fault detection results of each device, determine whether there is a standby device that allows immediate switching for each faulty device as the second feedback data, which can be selected to be determined synchronously with the determination of the fault frequency.

[0060] Adjust the weights of the physical structure importance and the failure impact degree of each device according to the first feedback data and the second feedback data; that is, when dynamically judging whether each device is a key node by using the physical structure importance and the failure impact degree of the device as judgment indicators, adjust the weight of the failure impact degree of each device according to the first feedback data. The lower the failure frequency of each device, the lower the corresponding weight of the failure impact degree. Adjust the weights of the physical structure importance of each device according to the second feedback data. For each failed device with an immediately switchable standby device, the weight of the physical structure importance is correspondingly reduced.

[0061] As Figure 2 shown, this embodiment specifically discloses a photovoltaic power station monitoring and warning system, including: A photovoltaic device relationship acquisition module 101, configured to periodically perform physical topology modeling and historical fault correlation analysis on a photovoltaic power station, and acquire the physical connection relationship and the fault propagation correlation relationship between devices in the photovoltaic power station; A photovoltaic device data acquisition module 102, configured to determine the physical structure importance and the failure impact degree of each device based on the acquired physical connection relationship and the fault propagation correlation relationship; use the physical structure importance and the failure impact degree of the device as judgment indicators to dynamically judge whether each device is a key node; for the devices determined to be key nodes, perform data acquisition at a first preset acquisition frequency, and for the devices determined not to be key nodes, perform data acquisition at a second preset acquisition frequency less than the first preset frequency; A photovoltaic device feature extraction module 103, configured to match a transmission path that meets the first constraint condition for each device in the photovoltaic power station, and transmit the acquired data of each device to the edge node pointed to by the corresponding transmission path; the first constraint condition is that the comprehensive score based on the distance of each device from the pointed edge node, the load of the pointed edge node, and the frequency of transmission to the pointed edge node is greater than the first preset score; use the lightweight model built in the pointed edge node to extract the feature data of the device and upload it to the cloud; A photovoltaic device fault detection module 104, configured to match the feature data extracted for each device in the photovoltaic power station with a fault detection module built in the cloud that meets the second constraint condition; the second constraint condition is that the comprehensive score based on the load of the pointed fault detection module and the frequency of transmission to the pointed fault detection module is greater than the second preset score; use the matched fault detection model to obtain whether each device fails and the fault type and level; A photovoltaic device fault warning module 105, configured to generate a warning prompt according to whether each device fails and the fault type and level obtained.

[0062] In a specific embodiment, the photovoltaic device data acquisition module 102 in the system is further configured to obtain the structural information of each device in the photovoltaic power station; according to the preset mapping table of device structural information - device key monitoring parameters - sensor information, obtain the device key monitoring parameter information and sensor information corresponding to each device; the sensor information includes the sensor type and the sensor deployment information; the preset mapping table of device structural information - device key monitoring parameters - sensor information is obtained by statistical analysis of the historical structural information of each device and the historical key monitoring parameters and historical sensor information that are matched and verified to meet the preset requirements for fault location accuracy and location timeliness; control the actual sensor deployment and operation according to the obtained sensor information. According to the obtained device key monitoring parameter information and sensor information corresponding to each device, collect and obtain the key parameters of the corresponding characteristics and quantities of each device.

[0063] In a specific embodiment, the photovoltaic device data acquisition module 102 in the system is further configured to obtain the environmental data of each device in the photovoltaic power station. The photovoltaic device feature extraction module 103 is further configured to match a transmission path that meets the first constraint condition under the first environmental basis for each device in the photovoltaic power station; the first constraint condition under the first environmental basis is that under the first environmental basis, the comprehensive score based on the distance of each device from the edge node it points to, the load of the edge node it points to, and the frequency of transmission to the edge node it points to is greater than the first preset score; the first environmental basis refers to the environmental conditions with a similarity greater than the first preset similarity to the current environmental data. The photovoltaic device fault detection module 103 is further configured to match the built-in fault detection module in the cloud that meets the second constraint condition under the second environmental basis for the feature data extracted from each device in the photovoltaic power station; the second constraint condition under the second environmental basis is that under the second environmental basis, the comprehensive score based on the load of the fault detection module it points to and the frequency of transmission to the fault detection module it points to is greater than the second preset score; the second environmental basis refers to the environmental conditions with a similarity greater than the second preset similarity to the current environmental data.

[0064] In a specific embodiment, the photovoltaic device feature extraction module 103 in the system is further configured to, when there are multiple transmission paths that match and meet the first constraint condition for each device in the photovoltaic power station, randomly select a preset number of transmission paths as a verification group; establish communication transmissions between the edge nodes pointed to in the verification group, and mutually transmit the feature data of the devices extracted by each edge node; calculate the feature similarity between the feature data of the devices extracted by every two edge nodes, eliminate the feature data of the devices whose feature similarity with the feature data of the devices extracted by other edge nodes is greater than the preset feature similarity, fuse the feature data of the remaining devices extracted by each edge node, and upload the fused feature data of the devices to the cloud.

[0065] In a specific embodiment, the photovoltaic device data acquisition module 102 in the system is further configured to, when collecting data of the devices determined to be key nodes at the acquisition frequency of the first preset frequency, collect image data of the devices determined to be key nodes at the same acquisition frequency; when collecting data of the devices determined not to be key nodes at the acquisition frequency of the second preset frequency less than the first preset frequency, collect image data of the devices determined not to be key nodes at the same acquisition frequency; The photovoltaic device feature extraction module 103 is further configured to, for the electrical digital data collected by each device, compare the similarity of the electrical digital data of the device predicted according to the lightweight model built in the historically pointed edge node. If the similarity is lower than the third preset similarity, the synchronously collected image data is correspondingly transmitted to the currently pointed edge node; combining the electrical digital data and the image data collected by each device, the feature data of the device is extracted by using the lightweight model built in the pointed edge node and uploaded to the cloud.

[0066] It is further configured to, when extracting the feature data of the device by using the lightweight model built in the pointed edge node and uploading it to the cloud, synchronously upload the feature data extracted from the electrical digital data of the device predicted according to the lightweight model built in the pointed edge node to the cloud; when it is determined that the currently pointed edge node is the same as the edge node pointed to at the previous moment, for the electrical digital data collected by each device, compare the similarity of the electrical digital data of the device predicted according to the lightweight model built in the pointed edge node at the previous moment. If the similarity is not lower than the third preset similarity, there is no need to transmit the electrical digital data collected by the current device, and directly use the electrical digital data of the device predicted according to the lightweight model built in the pointed edge node at the previous moment as the electrical digital data collected at the current moment, and correspondingly match and satisfy the second constraint condition and the built-in fault detection module in the cloud, and use the matched fault detection model to obtain whether each device has a fault and the fault type and level.

[0067] A specific embodiment, the system further includes: a photovoltaic device failure feedback data acquisition module 106, configured to regularly determine the failure frequency of each device according to the acquired failure detection result of each device, and use the determined failure frequency of each device as the first feedback data; according to the acquired failure detection result of each device, determine whether there is a standby device that allows immediate switching for each failed device as the second feedback data; The photovoltaic device data acquisition module 102 is further configured to correspondingly adjust the weights of the physical structure importance degree and the failure impact degree of each device according to the first feedback data and the second feedback data; wherein, the lower the failure frequency of each device, the lower the corresponding weight of the failure impact degree; for each failed device with a standby device that allows immediate switching, the weight of the physical structure importance degree is correspondingly reduced.

[0068] An embodiment of the present application also discloses a computer-readable storage medium.

[0069] Specifically, the computer-readable storage medium stores a computer program that can be loaded and executed by a processor such as the above-mentioned photovoltaic power station monitoring and warning method. The computer-readable storage medium includes, for example: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0070] An embodiment of the present application also discloses a computer device.

[0071] Specifically, the computer device includes a memory and a processor, and the memory stores a computer program that can be loaded and executed by the processor such as the above-mentioned photovoltaic power station monitoring and warning method.

[0072] The above are all preferred embodiments of the present application. The protection scope of the present application is not limited by this. Any feature disclosed in this specification (including the abstract and drawings), unless specifically described, can be replaced by other equivalent or similar-purpose alternative features. That is, unless specifically described, each feature is only an example of a series of equivalent or similar features.

Claims

1. A method for monitoring and alarming of a photovoltaic power station, characterized in that, Including: Regularly conduct physical topology modeling and historical fault correlation analysis on the photovoltaic power station to obtain the physical connection relationship and fault propagation correlation relationship between devices in the photovoltaic power station; Based on the obtained physical connection relationship and fault propagation correlation relationship, determine the physical structure importance degree and fault influence degree of each device; using the physical structure importance degree and fault influence degree of the device as judgment indicators, dynamically judge whether each device is a key node; Collect data for devices judged to be key nodes at the acquisition frequency of the first preset frequency, and collect data for devices judged not to be key nodes at the acquisition frequency of the second preset frequency less than the first preset frequency; Match a transmission path that meets the first constraint condition for each device in the photovoltaic power station, and transmit the collected data of each device to the edge node pointed to by the corresponding transmission path; the first constraint condition is that the comprehensive score based on the distance of each device from the pointed edge node, the load of the pointed edge node, and the frequency of transmission to the pointed edge node is greater than the first preset score; use the lightweight model built in the pointed edge node to extract the characteristic data of the device and upload it to the cloud; Match a fault detection module built in the cloud that meets the second constraint condition for the characteristic data extracted from each device in the photovoltaic power station; the second constraint condition is that the comprehensive score based on the load of the pointed fault detection module and the frequency of transmission to the pointed fault detection module is greater than the second preset score; use the matched fault detection model to obtain whether each device has a fault and the fault type and level; Generate a warning prompt according to whether each device has a fault and the fault type and level obtained; 2. The photovoltaic power station monitoring and warning method according to claim 1, characterized in that, Also including: Obtain the structure information of each device in the photovoltaic power station; According to the preset device structure information - device key monitoring parameter - sensor information mapping table, obtain the device key monitoring parameter information and sensor information corresponding to each device; the sensor information includes the sensor type and sensor deployment information; the preset device structure information - device key monitoring parameter - sensor information mapping table is obtained through statistical analysis of the historical structure information of each device and the historical key monitoring parameters and historical sensor information that are verified to meet the preset requirements for fault location accuracy and location timeliness; Control the actual sensor deployment and operation according to the obtained sensor information; Collect and obtain the corresponding characteristics and the number of key parameters of each device according to the device key monitoring parameter information and sensor information corresponding to each device obtained; 3. The photovoltaic power station monitoring and warning method according to claim 1, wherein Also including: Obtain the environmental data of each device in the photovoltaic power station; Match a transmission path that meets the first constraint condition under the first environmental basis for each device in the photovoltaic power station; The first constraint condition under the first environmental basis is that under the first environmental basis, the comprehensive score based on the distance of each device from the pointed edge node, the load of the pointed edge node, and the frequency of transmission to the pointed edge node is greater than the first preset score; the first environmental basis refers to the environmental conditions with a similarity greater than the first preset similarity to the current environmental data; The feature data extracted for each device in the photovoltaic power station is matched with the built-in fault detection module in the cloud that meets the second constraint condition under the second environmental basis; the second constraint condition under the second environmental basis is that under the second environmental basis, the comprehensive score of the load of the pointed fault detection module and the frequency transmitted to the pointed fault detection module is greater than the second preset score; the second environmental basis refers to the environmental conditions with a similarity greater than the second preset similarity to the current environmental data.

4. The photovoltaic power station monitoring and warning method according to claim 1, wherein It further includes: When there are multiple transmission paths for each device in the photovoltaic power station that match and meet the first constraint condition, a preset number of transmission paths are randomly selected as the verification group; For the edge nodes pointed to in the verification group, establish communication transmissions with each other and transmit the feature data of the devices extracted by each edge node to each other; Calculate the feature similarity between the feature data of the devices extracted by every two edge nodes; Eliminate the feature data of the devices whose feature similarity with the feature data of the devices extracted by other edge nodes is greater than the preset feature similarity, fuse the remaining feature data of the devices extracted by each edge node, and upload the fused feature data of the devices to the cloud.

5. The photovoltaic power station monitoring and warning method according to claim 1, characterized in that, It further includes: When collecting data for the devices determined to be key nodes at the first preset frequency, collect the image data of the devices determined to be key nodes at the same collection frequency; When collecting data for the devices determined to be non-key nodes at the second preset frequency less than the first preset frequency, collect the image data of the devices determined to be non-key nodes at the same collection frequency; For the electrical digital data collected for each device, compare the similarity of the electrical digital data of the device predicted according to the lightweight model built in the historical pointed edge node. If the similarity is lower than the third preset similarity, then transmit the synchronously collected image data to the currently pointed edge node; combine the electrical digital data and image data collected for each device, and use the lightweight model built in the pointed edge node to extract the feature data of the device and upload it to the cloud.

6. The photovoltaic power station monitoring and warning method according to claim 5, wherein It further includes: When using the lightweight model built in the pointed edge node to extract the feature data of the device and upload it to the cloud, synchronously upload the feature data extracted from the electrical digital data of the device predicted according to the lightweight model built in the pointed edge node to the cloud; When it is determined that the currently pointed edge node is the same as the pointed edge node at the previous moment, for the electrical digital data collected for each device, compare the similarity of the electrical digital data of the device predicted according to the lightweight model built in the pointed edge node at the previous moment. If the similarity is not lower than the third preset similarity, there is no need to transmit the electrical digital data collected for the current device, and directly use the electrical digital data of the device predicted according to the lightweight model built in the pointed edge node at the previous moment as the electrical digital data collected at the current moment, and correspondingly match the built-in fault detection module in the cloud that meets the second constraint condition, and use the matched fault detection model to obtain whether each device is faulty and the type and level of the fault.

7. The photovoltaic power station monitoring and warning method according to claim 1, characterized in that It further includes: According to the obtained fault detection results of each device, regularly determine the fault frequency of each device, and use the determined fault frequency of each device as the first feedback data; According to the obtained fault detection results of each device, determine whether there is a standby device that allows immediate switching for each faulty device as the second feedback data; Correspondingly adjust the weight of the physical structure importance and the weight of the fault impact degree of each device according to the first feedback data and the second feedback data; among them, the lower the fault frequency of each device, the lower the corresponding weight of the fault impact degree; if there is a standby device that allows immediate switching for each faulty device, the weight of the physical structure importance is correspondingly reduced.

8. A photovoltaic power station monitoring and warning system, characterized in that, It includes: A photovoltaic device relationship acquisition module, which is used to regularly perform physical topology modeling and historical fault correlation analysis on a photovoltaic power station to obtain the physical connection relationship and fault propagation correlation relationship between devices in the photovoltaic power station; A photovoltaic device data acquisition module, which is used to determine the physical structure importance and fault impact degree of each device based on the obtained physical connection relationship and fault propagation correlation relationship; using the physical structure importance and fault impact degree of the device as judgment indicators, dynamically judge whether each device is a key node; Data of devices judged as key nodes are collected at the acquisition frequency of the first preset frequency, and data of devices judged as non-key nodes are collected at the acquisition frequency of the second preset frequency less than the first preset frequency; A photovoltaic device feature extraction module, which is used to match a transmission path that meets the first constraint condition for each device in the photovoltaic power station, and transmit the collected data of each device to the edge node pointed to by the corresponding transmission path; the first constraint condition is that the comprehensive score based on the distance of each device from the pointed edge node, the load of the pointed edge node, and the frequency of transmission to the pointed edge node is greater than the first preset score; use the lightweight model built in the pointed edge node to extract the feature data of the device and upload it to the cloud; A photovoltaic device fault detection module, which is used to match the feature data extracted from each device in the photovoltaic power station with a fault detection module built in the cloud that meets the second constraint condition; the second constraint condition is that the comprehensive score based on the load of the pointed fault detection module and the frequency of transmission to the pointed fault detection module is greater than the second preset score; use the matched fault detection model to obtain whether each device has a fault and the fault type and level; A photovoltaic device fault warning module, which is used to generate a warning prompt according to whether each device has a fault and the fault type and level obtained; 9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method described in any one of claims 1 to 7.

10. A computer device, characterized in that, The computer device includes a memory, a processor, and a program stored and executable on the memory. When the program is executed by the processor, it implements the steps of the method described in any one of claims 1 to 7.

Images