Intelligent operation and maintenance and fault diagnosis system and method for distributed photovoltaic power station

Abstract

This invention discloses an intelligent operation and maintenance and fault diagnosis system and method for distributed photovoltaic power stations, including multiple intelligent photovoltaic modules deployed at the power station site, on-site data acquisition and communication terminals, and a remote intelligent diagnostic cloud platform. The intelligent photovoltaic modules integrate embedded sensor arrays and module-level voltage monitoring modules. This invention, through multi-level deep perception and intelligent fusion, achieves a fundamental transformation from post-maintenance to pre-warning, and from extensive management to precise intervention. Specifically, the system deeply integrates a passive surface acoustic wave sensor array inside the photovoltaic modules, acting like giving each module continuous self-sensing nerve endings, enabling in-situ, real-time, and wireless monitoring of key physical states such as temperature and stress. This overcomes the bottleneck of traditional operation and maintenance, which relies solely on external electrical parameters and cannot perceive the microscopic physical changes inside the modules.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, and more specifically, to an intelligent operation and maintenance and fault diagnosis system and method for distributed photovoltaic power plants. Background Technology

[0002] With the acceleration of global energy structure transformation, photovoltaic power generation, as a major force in clean energy, has seen its installed capacity grow rapidly. However, the operation and maintenance management of photovoltaic power plants, especially large-scale distributed photovoltaic power plants, is facing increasingly severe challenges.

[0003] Traditional operation and maintenance models heavily rely on periodic manual inspections and aggregated electrical data such as DC-side voltage, current, and power provided by string-level inverters. This model suffers from significant lag and inefficiency. Manual inspections are inefficient, costly, and fail to detect internal micro-defects such as microcracks in solar cells, poor solder joints, and early aging of packaging materials. String-level monitoring can only reflect the overall performance anomalies of the entire string, failing to pinpoint the faulty component or identify the specific type and root cause of the fault.

[0004] For example, when the system alarms and shows a decrease in the power of a certain branch, maintenance personnel cannot distinguish whether it is caused by the common hot spot effect due to dust obstruction or partial shading, or by the more hidden and harmful potential-induced degradation, internal cracks in the cells, or corrosion at the connection points. This leads to passive maintenance response, long repair cycles, and a large amount of potential power generation loss and safety hazards that cannot be detected and eliminated in a timely manner. Therefore, professionals in this field have provided intelligent operation and maintenance and fault diagnosis systems and methods for distributed photovoltaic power stations to solve the above-mentioned problems. Summary of the Invention

[0005] In view of the problems existing in the prior art, the purpose of this invention is to provide an intelligent operation and maintenance and fault diagnosis system and method for distributed photovoltaic power plants.

[0006] To solve the above problems, the present invention adopts the following technical solution;

[0007] The intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations includes multiple intelligent photovoltaic modules deployed at the power station site, on-site data acquisition and communication terminals, and a remote intelligent diagnosis cloud platform. The intelligent photovoltaic modules integrate embedded sensor arrays and module-level voltage monitoring modules. The embedded sensor array consists of multiple passive surface acoustic wave (SAW) sensors arranged in a predetermined topology within the module laminated package. The resonant frequencies of the passive SAW sensors are sensitive to temperature, stress, and strain, and are used to characterize the physical state of local areas inside the module. Each passive SAW sensor has a unique radio frequency code. The module-level voltage monitoring module is connected in parallel with the busbar or bypass diode of the photovoltaic module string to measure and report the real-time voltage to ground or voltage to string of the intelligent photovoltaic module with high precision.

[0008] The field data acquisition and communication terminal integrates a main control processing module, which is electrically connected to a radio frequency excitation and reading module, a branch electrical data acquisition module, a multi-mode communication gateway module, a positioning and timing module, and a field environment monitoring module.

[0009] The core of the radio frequency excitation and reading module is a radio frequency reader / writer. During the inspection process, the radio frequency reader / writer transmits a specific frequency band excitation radio frequency signal to the embedded sensor array of the smart photovoltaic module through an antenna, and receives the response signal containing frequency offset information returned by the passive surface acoustic wave sensor. The radio frequency excitation and reading module also includes a signal conditioning and calculation unit. The signal conditioning and calculation unit is connected to the radio frequency reader / writer and is used to filter and amplify the response signal, and calculate the real-time temperature, stress or strain physical quantities corresponding to each sensor.

[0010] The branch electrical data acquisition module is used to collect the voltage, current and power electrical parameters of the DC side of the photovoltaic module string or string inverter;

[0011] The multi-mode communication gateway module is used to establish a data backhaul link with the remote intelligent diagnostic cloud platform and a local communication link with the on-site mobile inspection equipment. The multi-mode communication gateway module includes a cellular network unit, a low-power wide area network unit and a local wireless unit. The main control processing module dynamically selects or aggregates at least one communication link according to data bandwidth requirements and network conditions.

[0012] The positioning and timing module is used to provide accurate spatial location labels and unified timestamps for the various types of data collected;

[0013] The on-site environmental monitoring module is used to collect macroscopic environmental parameters at the power plant site. The on-site environmental monitoring module includes an ambient temperature sensor, an ambient humidity sensor, an irradiance sensor, and a wind speed and direction sensor.

[0014] The main control processing module is used to coordinate the work of each module, preprocess, associate, package and cache the collected component-level voltage data, branch electrical data, physical quantity data and environmental data locally, and upload them to the remote intelligent diagnostic cloud platform through the multi-mode communication gateway module.

[0015] The remote intelligent diagnostic cloud platform is remotely connected to the multi-mode communication gateway module of the field data acquisition and communication terminal. The remote intelligent diagnostic cloud platform is used to receive and store all field data. The remote intelligent diagnostic cloud platform contains a component physical field digital twin engine and a micro fault prediction algorithm model.

[0016] The remote intelligent diagnostic cloud platform is remotely connected to the multi-mode communication gateway module of the field data acquisition and communication terminal. The remote intelligent diagnostic cloud platform is used to receive and store all field data. The remote intelligent diagnostic cloud platform is internally deployed with a component physical field digital twin engine and a multi-source fusion fault diagnosis algorithm model.

[0017] The component physical field digital twin engine is used to reconstruct and visualize the two-dimensional temperature field distribution map, stress field distribution map and strain field distribution map inside the component based on the physical quantity data reported by all passive surface acoustic wave sensors in a single smart photovoltaic module.

[0018] The multi-source fusion fault diagnosis algorithm model is used to fuse and analyze multi-dimensional data to identify abnormal voltage differences, physical field anomalies, and environmental performance correlation anomalies, and diagnose a variety of faults including but not limited to potential-induced decay, hot spot effect, dust and snow obstruction, cell microcracks, solder strip poor soldering, and packaging aging.

[0019] As a further description of the above technical solution: the predetermined topology of the embedded sensor array is grid-like, and the sensor placement positions cover at least the interconnection solder strip area of ​​the photovoltaic cell, the stress concentration area at the edge of the cell, and the junction of the module frame and the laminate.

[0020] As a further description of the above technical solution: the radio frequency excitation and reading module also includes a directional scanning antenna array, which is connected to the radio frequency reader. The beam pointing and width of the directional scanning antenna array can be electrically controlled and adjusted to achieve precise radio frequency energy projection and signal reception for specific smart photovoltaic modules or module arrays during inspection, thereby improving the reading success rate and reducing interference from neighboring areas.

[0021] As a further description of the above technical solution: the field data acquisition and communication terminal integrates an edge computing module, which is connected to the main control processing module. The edge computing module is used to perform preliminary fault diagnosis logic before data is uploaded.

[0022] As a further description of the above technical solution: the multi-source fusion fault diagnosis algorithm model includes a voltage difference analysis sub-model. The voltage difference analysis sub-model compares the module-level voltage data reported by different smart photovoltaic modules within the same photovoltaic module string. When it is detected that the voltage of a certain module is significantly lower than the average voltage of other modules in the string and the difference exceeds the first preset threshold, it is initially diagnosed as series mismatch or severe shading.

[0023] When an abnormally low voltage to ground is detected in the component, combined with the high humidity conditions reported by the on-site environmental monitoring module, a preliminary diagnosis of potential-induced degradation risk is made.

[0024] As a further description of the above technical solution: the remote intelligent diagnostic cloud platform is also remotely connected to a mobile inspection device, which is a drone or a ground robot. The mobile inspection device is equipped with an airborne version of the radio frequency excitation and reading module and a positioning and timing module. The mobile inspection device communicates with the field data acquisition and communication terminal through a local wireless unit, receives inspection path instructions, and transmits back the read sensor array data.

[0025] As a further description of the above technical solution: the remote intelligent diagnostic cloud platform also includes an intelligent inspection path planning module. The intelligent inspection path planning module dynamically generates and issues path instructions to the mobile inspection equipment based on the preliminary diagnostic results and fault probability output by the multi-source fusion fault diagnosis algorithm model.

[0026] The intelligent operation and maintenance and fault diagnosis method for distributed photovoltaic power stations includes the following steps:

[0027] S1. First, multiple smart photovoltaic modules are deployed at the photovoltaic power station site. The embedded sensor array inside each module is encapsulated in a laminate with a predetermined topology.

[0028] The on-site data acquisition and communication terminal is installed near the power station combiner box or inverter. It is connected to the DC side of the photovoltaic module string through the branch electrical data acquisition module, and an initial communication link is established with the remote intelligent diagnostic cloud platform through the multi-mode communication gateway module.

[0029] After the system is powered on, the positioning and timing module obtains the precise geographical location and unified time reference. The main control processing module of the field data acquisition and communication terminal starts self-testing and coordinates the radio frequency excitation and reading module to perform the first full reading of the passive surface acoustic wave sensors of all smart photovoltaic modules within the communication range, obtain the initial temperature, stress, and strain physical quantity baseline data, and upload them together with the initial module-level voltage, branch electrical parameters and environmental parameters to the remote intelligent diagnostic cloud platform to complete the construction of the system's digital profile.

[0030] S2, the branch electrical data acquisition module continuously collects voltage, current and power data of each photovoltaic branch at a first sampling frequency such as per minute; the on-site environmental monitoring module simultaneously collects ambient temperature, humidity, irradiance, wind speed and wind direction data;

[0031] The main control processing module schedules the radio frequency excitation and reading modules to work according to the preset inspection plan or cloud platform instructions. When a fixed terminal is used, the antenna array can be directionally scanned to align with each smart photovoltaic module in sequence, transmit excitation radio frequency signals and receive responses. The signal conditioning and calculation unit calculates the real-time physical quantities of each sensor.

[0032] When mobile inspection equipment is deployed, the intelligent inspection path planning module of the remote intelligent diagnostic cloud platform generates inspection instructions, controls the mobile device to fly or move along the planned path, and its airborne radio frequency module reads data from the components along the route.

[0033] The main control processing module collects component physical quantity data, component-level voltage data, branch electrical data and environmental data, and uses the location tags and timestamps provided by the positioning and timing module to perform spatiotemporal correlation and alignment to form a structured data packet. The edge computing module performs preliminary cleaning, compression and local caching of the data packet, and performs primary anomaly marking according to preset simple rules such as voltage sudden drop to zero and physical quantity exceeding absolute threshold.

[0034] S3. The field data acquisition and communication terminal uploads the pre-processed data packets to the remote intelligent diagnostic cloud platform through the communication link selected by the multi-mode communication gateway module.

[0035] After receiving the data, the cloud platform stores the received multi-source heterogeneous data into a time-series database and a relational database to establish a complete power plant operation history archive. For each smart photovoltaic module, the module's physical field digital twin engine calls the real-time temperature, stress, and strain data reported by all passive surface acoustic wave sensors inside the module and uses spatial interpolation algorithms to reconstruct a high-resolution two-dimensional temperature field distribution map, stress field distribution map, and strain field distribution map inside the module. These distribution maps are displayed on the cloud platform's visualization interface in the form of thermal maps or contour maps, intuitively presenting the microscopic non-uniformity of the module's internal state.

[0036] S4. The multi-source fusion fault diagnosis algorithm model of the remote intelligent diagnostic cloud platform is activated, and the aggregated global data is fused, analyzed, and deeply mined. The diagnostic process is as follows:

[0037] Voltage difference analysis and series mismatch diagnosis: The voltage difference analysis sub-model compares the module-level voltage reported by all smart photovoltaic modules in real time for the same photovoltaic module string. If the voltage of a certain module is detected to be continuously and significantly lower than the average voltage of other modules in the same string, and the difference exceeds the first preset threshold, then based on the physical field distribution map of the module, it is preliminarily diagnosed as series mismatch or severe shading such as bird droppings or local dust accumulation. If the voltage of the module to ground is abnormally low, and the ambient humidity reported by the on-site environmental monitoring module is continuously higher than the second preset threshold, then it is diagnosed as an increased risk of potential-induced degradation and an early warning is issued.

[0038] Diagnostic correlation between physical field anomalies and microscopic defects: Analysis of the distribution map generated by the component physical field digital twin engine;

[0039] Hot spot effect diagnosis: If a clear high-temperature "island" area appears in the temperature field distribution map, and the stress or strain field corresponding to the area also shows abnormality, and the output current of the component is lower than expected, then the hot spot effect is diagnosed, and the precise location of the hot spot in the component is located.

[0040] Diagnosis of microcracks and mechanical stress: If the stress and strain field distribution map shows a high gradient stress concentration zone at the edge of the cell or in a specific area, while the temperature field is not obviously abnormal, it suggests that there may be microcracks in the cell or changes in stress distribution caused by aging of the encapsulation material. Combined with the trend analysis of historical strain data, the risk of microcrack propagation can be assessed.

[0041] Diagnosis of solder ribbon defects or detachment: If the sensor detects an abnormally high temperature point or interruption of temperature conduction in the interconnect solder ribbon area, and the voltage of the component fluctuates intermittently, a solder ribbon defect or detachment is suspected.

[0042] The power and efficiency data at the branch level are correlated with real-time irradiance and ambient temperature data to form a model. If, under the same irradiance and temperature conditions, the efficiency of a branch continues to deviate from the theoretical value or the average value of the power plant exceeds the third preset threshold, the cloud platform will further call the physical field and voltage data of all components under that branch for drill-down analysis to locate the root cause. For example, general dust blockage may manifest as a slight and uniform increase in temperature of all components, a synchronous decrease in efficiency, or a serious failure of a key component in the branch.

[0043] S5. The multi-source fusion fault diagnosis algorithm model, by integrating the above analysis steps, generates a structured diagnostic report containing the following elements:

[0044] Fault / Abnormality Type: Identify the nature of the fault, such as hot spot, PID risk, microcrack, obstruction, etc.

[0045] Fault location: pinpoint the fault location to the power station, string, and specific component number, and mark the coordinates of the abnormal area on the component's internal diagram;

[0046] Severity Level: Based on the impact on power generation, safety risks, and fault development trends, fault severity levels are classified as emergency, important, general, and warning.

[0047] Confidence level: Provides the confidence probability of this diagnosis;

[0048] Root cause analysis and maintenance recommendations: Analyze possible causes of the failure and generate specific maintenance work order recommendations, such as: "Perform hot spot inspection and cleaning of component 5 in string 3 of XX array", "Activate PID repair function for components in YY area", "Arrange for drones to conduct detailed infrared and EL retests on ZZ branch".

[0049] When a high-risk fault that could cause a fire is diagnosed, such as a severe hot spot or a partial short circuit, the cloud platform will generate a work order and send a power reduction or emergency shutdown command to the inverter in the corresponding string through a security protocol to eliminate the safety risk. When a minor series mismatch is diagnosed, it can suggest that the inverter enable the optimizer function or adjust the MPPT operating point to mitigate the mismatch loss and improve the overall energy efficiency of the system.

[0050] S6. For high-priority faults, the cloud platform automatically generates maintenance work orders and dispatches them to the mobile terminal. At the same time, the intelligent inspection path planning module dynamically adjusts the subsequent inspection path of the mobile inspection equipment based on the diagnostic results, prioritizing high-risk components for high-frequency, close-range verification and testing to achieve precise maintenance.

[0051] On-site maintenance personnel perform inspection, repair, or cleaning tasks according to work orders and feed the results back to the cloud platform via mobile terminals. The cloud platform compares the "diagnostic prediction" with the "actual processing result" to form a closed loop. This feedback data is used to continuously train and optimize the multi-source fusion fault diagnosis algorithm model to improve its diagnostic accuracy and generalization ability.

[0052] Compared with the prior art, the advantages of this invention are:

[0053] This invention achieves a fundamental transformation from post-maintenance to pre-warning and from extensive management to precise intervention through multi-level deep perception and intelligent integration. Specifically, the system deeply integrates a passive surface acoustic wave sensor array into the photovoltaic module, which is like giving each module a nerve ending that continuously senses itself. This enables in-situ, real-time, and wireless monitoring of key physical states such as temperature and stress, breaking through the bottleneck of traditional operation and maintenance that relies solely on external electrical parameters and cannot perceive the microscopic physical changes inside the module.

[0054] Based on this, the system uses on-site data acquisition and communication terminals to collect and spatiotemporally correlate component-level physical data, branch-level electrical data and power station-level environmental data from multiple sources, thus constructing an unprecedented panoramic data base.

[0055] The remote intelligent diagnostic cloud platform serves as the system's intelligent brain. Its component physical field digital twin engine can dynamically reconstruct discrete sensor data into an intuitive and visual internal physical field distribution map, making hidden dangers such as hot spots and microcracks visible. At the same time, its multi-source fusion fault diagnosis algorithm model can accurately distinguish and diagnose the specific types and causes of more than ten complex faults, such as potential-induced decay, solder strip cold solder joints, and package aging, through deep correlation analysis of electrical anomalies, physical field distortions and environmental causes, greatly improving the accuracy of diagnosis and early warning capabilities.

[0056] Furthermore, the system integrates an edge computing module to achieve local preliminary diagnosis and real-time alarms, reducing cloud load and communication latency. The optional mobile inspection enhancement solution uses drones or robots equipped with directional scanning antennas to conduct precise close-range re-inspections of high-risk targets based on intelligent cloud-planned paths. This forms a three-dimensional and automated inspection system with wide-area coverage of fixed monitoring networks, routine general inspections, and on-demand deployment of mobile inspection equipment for precise verification. Ultimately, while comprehensively improving the power generation reliability of the power station and extending the life of components, it significantly reduces operation and maintenance costs, safety risks, and power generation losses, realizing intelligent, precise, and unmanned closed-loop management of photovoltaic power station operation and maintenance. Attached Figure Description

[0057] Figure 1 This is a schematic diagram illustrating the basic framework principle of operation and maintenance and fault diagnosis of the present invention;

[0058] Figure 2 This is a schematic diagram illustrating the construction principle of the modular unit of the present invention;

[0059] Figure 3 This is a schematic diagram of the principle of the field data acquisition and communication terminal of the present invention;

[0060] Figure 4 This is a schematic diagram of the principle of the remote intelligent diagnostic cloud platform of the present invention.

[0061] Explanation of the labels in the diagram:

[0062] 1. Intelligent photovoltaic module; 101. Embedded sensor array; 1011. Passive surface acoustic wave sensor; 102. Module-level voltage monitoring module; 2. Field data acquisition and communication terminal; 201. Main control processing module; 202. RF excitation and reading module; 2021. RF reader / writer; 2022. Signal conditioning and processing unit; 2023. Directional scanning antenna array; 203. Branch electrical data acquisition module; 204. Multimode communication gateway module; 2041. Cellular network unit; 2042. Low-power wide area network unit 2043, Local Area Wireless Unit; 205, Positioning and Timing Module; 206, On-site Environmental Monitoring Module; 2061, Ambient Temperature Sensor; 2062, Ambient Humidity Sensor; 2063, Irradiance Sensor; 2064, Wind Speed ​​and Direction Sensor; 207, Edge Computing Module; 3, Remote Intelligent Diagnosis Cloud Platform; 301, Component Physical Field Digital Twin Engine; 302, Multi-Source Fusion Fault Diagnosis Algorithm Model; 3021, Voltage Difference Analysis Sub-model; 303, Intelligent Inspection Path Planning Module; 4, Mobile Inspection Equipment. Detailed Implementation

[0063] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention;

[0064] Example 1:

[0065] Please see Figures 1-4 In this invention, the intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations includes multiple intelligent photovoltaic modules 1 deployed at the power station site, a field data acquisition and communication terminal 2, and a remote intelligent diagnosis cloud platform 3. The intelligent photovoltaic module 1 integrates an embedded sensor array 101 and a module-level voltage monitoring module 102. The embedded sensor array 101 consists of multiple passive surface acoustic wave sensors 1011 arranged in a predetermined topology within the module laminated package. The resonant frequency of the passive surface acoustic wave sensors 1011 is sensitive to temperature, stress, and strain, and is used to characterize the physical state of local areas inside the module. Each passive surface acoustic wave sensor 1011 has a unique radio frequency code. The module-level voltage monitoring module 102 is connected in parallel with the busbar or bypass diode of the photovoltaic module string, and is used to measure and report the real-time voltage to ground or voltage to string of the intelligent photovoltaic module 1 with high precision.

[0066] The field data acquisition and communication terminal 2 integrates a main control processing module 201, which is electrically connected to a radio frequency excitation and reading module 202, a branch electrical data acquisition module 203, a multi-mode communication gateway module 204, a positioning and timing module 205, and a field environment monitoring module 206.

[0067] The core of the RF excitation and reading module 202 is the RF reader 2021. The RF reader 2021 is used to transmit a specific frequency band excitation RF signal to the embedded sensor array 101 of the smart photovoltaic module 1 through an antenna during the inspection process, and to receive the response signal containing frequency offset information returned by the passive surface acoustic wave sensor 1011. The RF excitation and reading module 202 also includes a signal conditioning and calculation unit 2022. The signal conditioning and calculation unit 2022 is connected to the RF reader 2021 and is used to filter and amplify the response signal, and calculate the real-time temperature, stress or strain physical quantities corresponding to each sensor.

[0068] The branch electrical data acquisition module 203 is used to acquire the voltage, current and power electrical parameters of the DC side of the photovoltaic module string or string inverter;

[0069] The multi-mode communication gateway module 204 is used to establish a data backhaul link with the remote intelligent diagnostic cloud platform 3 and a local communication link with the on-site mobile inspection equipment 4. The multi-mode communication gateway module 204 includes a cellular network unit 2041, a low-power wide area network unit 2042 and a local wireless unit 2043. The main control processing module 201 dynamically selects or aggregates at least one communication link according to data bandwidth requirements and network conditions.

[0070] The positioning and timing module 205 is used to provide accurate spatial location labels and unified timestamps for various types of data collected;

[0071] The on-site environmental monitoring module 206 is used to collect macroscopic environmental parameters at the power plant site. The on-site environmental monitoring module 206 includes an ambient temperature sensor 2061, an ambient humidity sensor 2062, an irradiance sensor 2063, and a wind speed and direction sensor 2064.

[0072] The main control processing module 201 is used to coordinate the work of each module, preprocess, associate, package and cache the collected component-level voltage data, branch electrical data, physical quantity data and environmental data locally, and upload them to the remote intelligent diagnostic cloud platform 3 through the multi-mode communication gateway module 204;

[0073] The remote intelligent diagnostic cloud platform 3 is remotely connected to the multi-mode communication gateway module 204 of the field data acquisition and communication terminal 2. The remote intelligent diagnostic cloud platform 3 is used to receive and store all field data. The remote intelligent diagnostic cloud platform 3 contains a component physical field digital twin engine 301 and a micro fault prediction algorithm model.

[0074] The remote intelligent diagnostic cloud platform 3 is remotely connected to the multi-mode communication gateway module 204 of the field data acquisition and communication terminal 2. The remote intelligent diagnostic cloud platform 3 is used to receive and store all field data. The remote intelligent diagnostic cloud platform 3 is internally deployed with a component physical field digital twin engine 301 and a multi-source fusion fault diagnosis algorithm model 302.

[0075] The component physical field digital twin engine 301 is used to reconstruct and visualize the two-dimensional temperature field distribution map, stress field distribution map and strain field distribution map inside the component based on the physical quantity data reported by all passive surface acoustic wave sensors 1011 in a single smart photovoltaic module 1.

[0076] The multi-source fusion fault diagnosis algorithm model 302 is used to fuse and analyze multi-dimensional data. By identifying abnormal voltage differences, physical field anomalies and environmental performance correlation anomalies, it diagnoses a variety of faults including but not limited to potential-induced decay, hot spot effect, dust and snow obstruction, cell microcracks, solder strip poor soldering and packaging aging.

[0077] The radio frequency excitation and reading module 202 also includes a directional scanning antenna array 2023, which is connected to the radio frequency reader 2021. The beam pointing and width of the directional scanning antenna array 2023 can be electrically controlled and adjusted to achieve precise radio frequency energy projection and signal reception on a specific smart photovoltaic module 1 or module array during inspection, thereby improving the reading success rate and reducing interference from neighboring cells.

[0078] The predetermined topology of the embedded sensor array 101 is a grid, and its sensor placement positions at least cover the interconnection solder strip area of ​​the photovoltaic cells, the stress concentration area at the edge of the cells, and the junction of the module frame and the laminate.

[0079] The field data acquisition and communication terminal 2 integrates an edge computing module 207, which is connected to the main control processing module 201. The edge computing module 207 is used to perform preliminary fault diagnosis logic before data is uploaded.

[0080] The multi-source fusion fault diagnosis algorithm model 302 includes a voltage difference analysis sub-model 3021. Within the same photovoltaic module string, the voltage difference analysis sub-model 3021 compares the module-level voltage data reported by different smart photovoltaic modules 1. When it is detected that the voltage of a certain module is significantly lower than the average voltage of other modules in the string and the difference exceeds the first preset threshold, it is initially diagnosed as series mismatch or severe shading.

[0081] When an abnormally low voltage to ground was detected in the component, combined with the high humidity conditions reported by the on-site environmental monitoring module 206, a preliminary diagnosis of potential-induced degradation risk was made.

[0082] In this invention, at the power plant site, each smart photovoltaic module 1 operates continuously. Inside, a passive surface acoustic wave sensor array 1011 is arranged in a grid-like topology in key areas such as the cell solder strips and edges. Its resonant frequency changes with the local temperature and stress state inside the module.

[0083] Meanwhile, the component-level voltage monitoring module 102 measures the voltage performance of the component in the string in real time. The field data acquisition and communication terminal 2 starts the inspection process according to the preset cycle. Its radio frequency excitation and reading module 202 transmits radio frequency signals through the antenna to excite and read the response signals carrying frequency offset information returned by each sensor. After being processed by the signal conditioning and calculation unit 2022, it is calculated into specific temperature and stress values.

[0084] Meanwhile, the branch electrical data acquisition module 203 collects the current and voltage of the entire string, the on-site environmental monitoring module 206 records the macroscopic conditions such as irradiance, temperature and humidity at that time, and the main control processing module 201 associates and packages all the data with the timestamp and location information provided by the positioning and timing module 205. During this process, the integrated edge computing module 207 can perform preliminary diagnosis. For example, using the rules of the voltage difference analysis sub-model 3021, it compares the voltage reported by each component in the same string in real time. If a component's voltage is found to be significantly low, it is immediately marked as "suspected mismatch / shading". If, combined with environmental data, it is found that high humidity conditions and low voltage to ground occur at the same time, it is marked as "PID risk".

[0085] The preprocessed data is uploaded to the remote intelligent diagnostic cloud platform 3 via the multi-mode communication gateway module 204, such as using 4G / 5G cellular network. After the cloud platform receives the data, the component physical field digital twin engine 301 first performs spatial interpolation and rendering on the temperature data of all sensors of the specified component to generate a high-resolution two-dimensional temperature field cloud map inside the component. The operation and maintenance personnel can intuitively see whether there are local high temperature areas, i.e. potential hot spots.

[0086] Subsequently, the multi-source fusion fault diagnosis algorithm model 302 initiates in-depth analysis. For example, when the algorithm identifies an isolated high-temperature "hot spot" in the temperature field cloud map of a certain component, and the branch current of the component is slightly lower than the average level of the same string, and the component-level voltage is not significantly abnormal, and the environmental irradiance is uniform, it can be comprehensively diagnosed as "micro hot spot caused by internal resistance increase due to microcracks in the battery cell", rather than a hot spot caused by leaf shading. The platform then generates a diagnostic report containing the precise location of the faulty component, the fault type, the severity, and handling suggestions, and pushes it to the handheld terminal of the operation and maintenance personnel to guide them to carry out precise repairs, thereby greatly improving the operation and maintenance efficiency and the power generation reliability of the power station.

[0087] Example 2:

[0088] Please see Figure 1 ,2 In addition to 3, the remote intelligent diagnostic cloud platform 3 is also remotely connected to the mobile inspection device 4. The mobile inspection device 4 is a drone or a ground robot. The mobile inspection device 4 is equipped with an airborne version of the radio frequency excitation and reading module 202 and a positioning and timing module 205. The mobile inspection device 4 communicates with the field data acquisition and communication terminal 2 through the local wireless unit 2043, receives the inspection path instructions and transmits back the read sensor array data.

[0089] The remote intelligent diagnostic cloud platform 3 also includes an intelligent inspection path planning module 303. Based on the preliminary diagnostic results and fault probability output by the multi-source fusion fault diagnosis algorithm model 302, the intelligent inspection path planning module 303 dynamically generates and issues path instructions to the mobile inspection device 4 to prioritize the inspection of high-fault-risk components or areas.

[0090] In this invention, the remote intelligent diagnostic cloud platform 3 continuously analyzes the data reported by the fixed terminal based on the multi-source fusion fault diagnosis algorithm model 302, and dynamically evaluates the health status and fault risk probability of components in each area. When the cloud platform identifies multiple potential-induced decay (PID) risk warnings in a certain area, or speculates through the physical field model that there may be potential soldering defects in several components, the intelligent inspection path planning module 303 will be automatically started.

[0091] The module dynamically generates an optimal inspection path based on the constraints of failure probability, geographical location, and equipment battery life. The instructions are sent from the cloud to the field data acquisition and communication terminal 2 on site, and then sent to the standby mobile inspection equipment 4, such as a drone, through its local wireless unit 2043, such as Wi-Fi or a private radio frequency link.

[0092] After receiving the command, the UAV equipped with the miniaturized radio frequency excitation and reading module 202 and the positioning and timing module 205 autonomously flies to the target area. Utilizing its onboard directional scanning antenna array 2023, the UAV can perform a precise "point-to-point" radio frequency scan on the high-risk smart photovoltaic module 1 specified by the ground command while hovering or flying at low speed. This effectively overcomes the problems of signal obstruction or weak signal reading at long distances that may exist with fixed antennas, and reads detailed physical state data of the sensor array inside the target module with a higher signal-to-noise ratio and success rate.

[0093] This high-precision data is transmitted back to the cloud in real time via a local link established between the drone and the ground terminal. The remote intelligent diagnostic cloud platform 3 integrates and compares the "special inspection" data obtained by the mobile inspection with the "general inspection" data reported by the fixed terminal, thereby verifying or correcting the preliminary diagnosis and generating a more accurate fault assessment.

[0094] For example, after a drone performs a close-range, detailed scan of a component suspected of having a poor solder joint, its stress field distribution map may clearly show the abnormal stress concentration at the solder joint, thereby upgrading the fault diagnosis from a "probabilistic hidden danger" to a "deterministic fault".

[0095] By combining "fixed monitoring with wide-area coverage" with "mobile inspection with precise verification", a three-dimensional inspection mode is achieved, enabling on-demand, in-depth, and automated diagnosis of the power plant's health status. It is particularly suitable for large power plants, complex installation environments such as corrugated steel roofs, and critical scenarios requiring extremely high diagnostic confidence, significantly reducing the cost and safety risks of manual inspection.

[0096] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and its improved concepts, should be covered within the scope of protection of the present invention.

Claims

1. An intelligent operation and maintenance and fault diagnosis system for a distributed photovoltaic power station, comprising a plurality of intelligent photovoltaic modules (1) deployed at the power station site, a site data acquisition and communication terminal (2), and a remote intelligent diagnosis cloud platform (3), characterized in that: The intelligent photovoltaic module (1) integrates an embedded sensor array (101) and a module-level voltage monitoring module (102). The embedded sensor array (101) consists of multiple passive surface acoustic wave sensors (1011) arranged in a predetermined topology within the module laminated package. The resonant frequency of the passive surface acoustic wave sensors (1011) is sensitive to temperature, stress, and strain, and is used to characterize the physical state of local areas inside the module. Each passive surface acoustic wave sensor (1011) has a unique radio frequency code. The module-level voltage monitoring module (102) is connected in parallel with the busbar or bypass diode of the photovoltaic module string, and is used to measure and report the real-time voltage to ground or voltage to string of the intelligent photovoltaic module (1) with high precision. The field data acquisition and communication terminal (2) integrates a main control processing module (201), which is electrically connected to a radio frequency excitation and reading module (202), a branch electrical data acquisition module (203), a multi-mode communication gateway module (204), a positioning and timing module (205), and a field environment monitoring module (206). The core of the radio frequency excitation and reading module (202) is the radio frequency reader (2021). The radio frequency reader (2021) is used to transmit a specific frequency band excitation radio frequency signal to the embedded sensor array (101) of the smart photovoltaic module (1) through an antenna during the inspection process, and receive the response signal containing frequency offset information returned by the passive surface acoustic wave sensor (1011). The radio frequency excitation and reading module (202) also includes a signal conditioning and calculation unit (2022). The signal conditioning and calculation unit (2022) is connected to the radio frequency reader (2021) to filter and amplify the response signal, and calculate the real-time temperature, stress or strain physical quantities corresponding to each sensor. The branch electrical data acquisition module (203) is used to acquire the voltage, current and power electrical parameters of the DC side of the photovoltaic module string or string inverter; The multi-mode communication gateway module (204) is used to establish a data backhaul link with the remote intelligent diagnostic cloud platform (3) and a local communication link with the on-site mobile inspection equipment (4). The multi-mode communication gateway module (204) includes a cellular network unit (2041), a low-power wide area network unit (2042), and a local wireless unit (2043). The main control processing module (201) dynamically selects or aggregates at least one communication link according to data bandwidth requirements and network conditions. The positioning and timing module (205) is used to provide accurate spatial location labels and unified timestamps for the various types of data collected; The on-site environmental monitoring module (206) is used to collect macroscopic environmental parameters at the power plant site. The on-site environmental monitoring module (206) includes an ambient temperature sensor (2061), an ambient humidity sensor (2062), an irradiance sensor (2063), and a wind speed and direction sensor (2064). The main control processing module (201) is used to coordinate the work of each module, preprocess, associate, package and cache the collected component-level voltage data, branch electrical data, physical quantity data and environmental data locally, and upload them to the remote intelligent diagnostic cloud platform (3) through the multi-mode communication gateway module (204). The remote intelligent diagnostic cloud platform (3) is remotely connected to the multi-mode communication gateway module (204) of the field data acquisition and communication terminal (2). The remote intelligent diagnostic cloud platform (3) is used to receive and store all field data. The remote intelligent diagnostic cloud platform (3) contains a component physical field digital twin engine (301) and a micro fault prediction algorithm model. The remote intelligent diagnostic cloud platform (3) is remotely connected to the multi-mode communication gateway module (204) of the field data acquisition and communication terminal (2). The remote intelligent diagnostic cloud platform (3) is used to receive and store all field data. The remote intelligent diagnostic cloud platform (3) is internally deployed with a component physical field digital twin engine (301) and a multi-source fusion fault diagnosis algorithm model (302). The component physical field digital twin engine (301) is used to reconstruct and visualize the two-dimensional temperature field distribution map, stress field distribution map and strain field distribution map inside the component based on the physical quantity data reported by all passive surface acoustic wave sensors (1011) in a single smart photovoltaic module (1). The multi-source fusion fault diagnosis algorithm model (302) is used to fuse and analyze multi-dimensional data to identify abnormal voltage differences, physical field anomalies and environmental performance correlation anomalies, and diagnose a variety of faults including but not limited to potential-induced decay, hot spot effect, dust and snow blockage, cell microcracks, solder strip poor soldering and packaging aging.

2. The intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations according to claim 1, characterized in that: The predetermined topology of the embedded sensor array (101) is grid-like, and its sensor placement positions cover at least the interconnection ribbon area of ​​the photovoltaic cells, the stress concentration area at the edge of the cells, and the junction of the module frame and the laminate.

3. The intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations according to claim 1, characterized in that: The radio frequency excitation and reading module (202) also includes a directional scanning antenna array (2023), which is connected to the radio frequency reader (2021). The beam pointing and width of the directional scanning antenna array (2023) can be electrically controlled and adjusted to achieve precise radio frequency energy projection and signal reception on specific smart photovoltaic modules (1) or module arrays during inspection, thereby improving the reading success rate and reducing neighboring area interference.

4. The intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations according to claim 1, characterized in that: The field data acquisition and communication terminal (2) integrates an edge computing module (207), which is connected to the main control processing module (201). The edge computing module (207) is used to perform preliminary fault diagnosis logic before data is uploaded.

5. The intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations according to claim 1, characterized in that: The multi-source fusion fault diagnosis algorithm model (302) includes a voltage difference analysis sub-model (3021). The voltage difference analysis sub-model (3021) compares the module-level voltage data reported by different smart photovoltaic modules (1) within the same photovoltaic module string. When it is detected that the voltage of a certain module is significantly lower than the average voltage of other modules in the string and the difference exceeds the first preset threshold, it is initially diagnosed as series mismatch or severe shading. When an abnormally low voltage to ground is detected in the component, combined with the high humidity conditions reported by the on-site environmental monitoring module (206), a preliminary diagnosis is made as a risk of potential-induced degradation.

6. The intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations according to claim 1, characterized in that: The remote intelligent diagnostic cloud platform (3) is also remotely connected to a mobile inspection device (4). The mobile inspection device (4) is a drone or a ground robot. The mobile inspection device (4) is equipped with an airborne version of the radio frequency excitation and reading module (202) and a positioning and timing module (205). The mobile inspection device (4) communicates with the field data acquisition and communication terminal (2) through the local wireless unit (2043), receives the inspection path instructions, and transmits the read sensor array data back.

7. The intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations according to claim 6, characterized in that: The remote intelligent diagnostic cloud platform (3) also includes an intelligent inspection path planning module (303). The intelligent inspection path planning module (303) dynamically generates and sends path instructions to the mobile inspection device (4) based on the preliminary diagnostic results and fault probability output by the multi-source fusion fault diagnosis algorithm model (302).

8. The intelligent operation and maintenance and fault diagnosis method for distributed photovoltaic power stations according to claim 1, applicable to the intelligent operation and maintenance and fault diagnosis system for distributed photovoltaic power stations described in any one of claims 1-7, includes the following steps: S1. First, multiple smart photovoltaic modules (1) are deployed at the photovoltaic power station site. The embedded sensor array (101) inside each module is encapsulated in a laminate with a predetermined topology. The on-site data acquisition and communication terminal (2) is installed near the power station combiner box or inverter, and connected to the DC side of the photovoltaic module string through the branch electrical data acquisition module (203). An initial communication link is established with the remote intelligent diagnostic cloud platform (3) through the multi-mode communication gateway module (204). After the system is powered on, the positioning and timing module (205) obtains the accurate geographical location and unified time reference. The main control processing module (201) of the field data acquisition and communication terminal (2) starts self-testing and coordinates the radio frequency excitation and reading module (202) to perform the first full reading of the passive surface acoustic wave sensor (1011) of all smart photovoltaic modules (1) within the communication range, obtain the baseline data of initial temperature, stress and strain physical quantities, and upload them together with the initial module-level voltage, branch electrical parameters and environmental parameters to the remote intelligent diagnostic cloud platform (3) to complete the construction of the system digital profile. S2, the branch electrical data acquisition module (203) continuously collects the voltage, current and power data of each photovoltaic branch at a first sampling frequency such as every minute; the on-site environmental monitoring module (206) simultaneously collects the ambient temperature, humidity, irradiance and wind speed and direction data; The main control processing module (201) schedules the radio frequency excitation and reading module (202) to work according to the preset inspection plan or cloud platform instructions. When a fixed terminal is used, the directional scanning antenna array (2023) can be aligned with each smart photovoltaic module (1) in sequence, transmit excitation radio frequency signals and receive responses. The signal conditioning and calculation unit (2022) calculates the real-time physical quantities of each sensor. When a mobile inspection device (4) is deployed, the intelligent inspection path planning module (303) of the remote intelligent diagnostic cloud platform (3) generates inspection instructions, controls the mobile device to fly or move along the planned path, and its airborne radio frequency module reads data from the components it passes through. The main control processing module (201) collects component physical quantity data, component-level voltage data, branch electrical data and environmental data, and uses the location tags and timestamps provided by the positioning and timing module (205) to perform spatiotemporal correlation and alignment to form a structured data packet. The edge computing module (207) performs preliminary cleaning, compression and local caching of the data packet, and performs primary anomaly marking according to preset simple rules such as voltage sudden drop to zero and physical quantity exceeding absolute threshold. S3. The field data acquisition and communication terminal (2) uploads the pre-processed data packets to the remote intelligent diagnostic cloud platform (3) through the communication link selected by the multi-mode communication gateway module (204). After receiving the data, the cloud platform stores the received multi-source heterogeneous data into a time-series database and a relational database to establish a complete power plant operation history archive. For each smart photovoltaic module (1), the module physical field digital twin engine (301) calls the real-time temperature, stress, and strain data reported by all passive surface acoustic wave sensors (1011) inside it, and uses spatial interpolation algorithm to reconstruct the high-resolution two-dimensional temperature field distribution map, stress field distribution map, and strain field distribution map inside the module. These distribution maps are displayed on the cloud platform visualization interface in the form of thermal maps or contour maps, intuitively presenting the micro-uniformity of the internal state of the module. S4. The multi-source fusion fault diagnosis algorithm model (302) of the remote intelligent diagnosis cloud platform (3) is started, and the global data is fused, analyzed and deeply mined. The diagnosis process is as follows: Voltage difference analysis and series mismatch diagnosis: The voltage difference analysis sub-model (3021) compares the module-level voltage reported by all smart photovoltaic modules (1) in real time for the same photovoltaic module string. If the voltage of a certain module is detected to be continuously and significantly lower than the average voltage of other modules in the same string, and the difference exceeds the first preset threshold, then combined with the physical field distribution map of the module, it is preliminarily diagnosed as series mismatch or severe shading such as bird droppings or local dust accumulation. If the voltage of the module to ground is abnormally low, and the ambient humidity reported by the on-site environmental monitoring module (206) is continuously higher than the second preset threshold, then it is diagnosed as an increased risk of potential-induced degradation and an early warning is issued. Diagnosis of the correlation between physical field anomalies and microscopic defects: Analysis of the distribution map generated by the component physical field digital twin engine (301); Hot spot effect diagnosis: If a clear high-temperature "island" area appears in the temperature field distribution map, and the stress or strain field corresponding to the area also shows abnormality, and the output current of the component is lower than expected, then the hot spot effect is diagnosed, and the precise location of the hot spot in the component is located. Diagnosis of microcracks and mechanical stress: If the stress and strain field distribution map shows a high gradient stress concentration zone at the edge of the cell or in a specific area, while the temperature field is not obviously abnormal, it suggests that there may be microcracks in the cell or changes in stress distribution caused by aging of the encapsulation material. Combined with the trend analysis of historical strain data, the risk of microcrack propagation can be assessed. Diagnosis of solder ribbon defects or detachment: If the sensor detects an abnormally high temperature point or interruption of temperature conduction in the interconnect solder ribbon area, and the voltage of the component fluctuates intermittently, a solder ribbon defect or detachment is suspected. The power and efficiency data at the branch level are correlated with real-time irradiance and ambient temperature data to form a model. If, under the same irradiance and temperature conditions, the efficiency of a branch continues to deviate from the theoretical value or the average value of the power plant exceeds the third preset threshold, the cloud platform will further call the physical field and voltage data of all components under that branch for drill-down analysis to locate the root cause. For example, general dust blockage may manifest as a slight and uniform increase in temperature of all components, a synchronous decrease in efficiency, or a serious failure of a key component in the branch. S5. Multi-source fusion fault diagnosis algorithm model (302) Based on the above analysis steps, a structured diagnostic report containing the following elements is generated: Fault / Abnormality Type: Identify the nature of the fault, such as hot spot, PID risk, microcrack, obstruction, etc. Fault location: pinpoint the fault location to the power station, string, and specific component number, and mark the coordinates of the abnormal area on the component's internal diagram; Severity Level: Based on the impact on power generation, safety risks, and fault development trends, fault severity levels are classified as emergency, important, general, and warning. Confidence level: Provides the confidence probability of this diagnosis; Root cause analysis and maintenance recommendations: Analyze possible causes of the failure and generate specific maintenance work order recommendations, such as: "Perform hot spot inspection and cleaning on component 5 of string 3 of array XX", "Activate PID repair function on components in YY area", "Arrange for drone to conduct detailed infrared and EL retesting on ZZ branch"; When a high-risk fault that could cause a fire is diagnosed, such as a severe hot spot or a partial short circuit, the cloud platform will generate a work order and send a power reduction or emergency shutdown command to the inverter in the corresponding string through a security protocol to eliminate the safety risk. When a minor series mismatch is diagnosed, it can suggest that the inverter enable the optimizer function or adjust the MPPT operating point to mitigate the mismatch loss and improve the overall energy efficiency of the system. S6. For high-priority faults, the cloud platform automatically generates maintenance work orders and dispatches them to the mobile terminal. At the same time, the intelligent inspection path planning module (303) dynamically adjusts the subsequent inspection path of the mobile inspection equipment (4) according to the diagnosis results, giving priority to high-risk components for high-frequency and close-range verification and testing, so as to achieve precise operation and maintenance. On-site maintenance personnel perform inspection, repair or cleaning work according to the work order, and feed back the processing results to the cloud platform through mobile terminals. The cloud platform compares the "diagnosis prediction" with the "actual processing result" to form a closed loop. These feedback data are used to continuously train and optimize the multi-source fusion fault diagnosis algorithm model (302) to improve its diagnostic accuracy and generalization ability.

Images