A fault location method for long-term photovoltaic arrays based on irradiance correction

Abstract

This invention relates to the field of solar photovoltaic power generation operation and maintenance technology, and discloses a fault location method for long-term photovoltaic arrays based on irradiance correction. The method first acquires operational data, meteorological data, and topology data of the long-term photovoltaic array. After quality inspection, time synchronization, and spatial mapping, the irradiance received by each string is obtained. Comparable operating condition intervals are divided according to irradiance, temperature, season, or solar altitude angle. A dynamic irradiance correction benchmark is established or updated based on healthy samples, and a corrected power deviation trajectory is formed. After deducting the common aging component of the entire array and the individual aging difference component of each string, the fault residual is obtained. The fault string is determined based on the fault residual, and the fault location result is output. This application can reduce false alarms and missed alarms in fault location of long-term photovoltaic arrays, and improve the reliability and traceability of the location results.

Description

Technical Field

[0001] This invention relates to the field of solar photovoltaic power generation operation and maintenance technology, specifically a fault location method for long-term photovoltaic arrays based on irradiance correction. Background Technology

[0002] As photovoltaic power plants age, photovoltaic modules are subjected to prolonged exposure to outdoor environments such as high temperatures, ultraviolet radiation, humidity, wind and sand, dust accumulation, salt spray, and diurnal temperature variations. Consequently, their output performance, temperature response characteristics, and inter-string consistency change over time. Current photovoltaic array fault location methods typically collect operational data such as string current, voltage, power, irradiance, and temperature, and determine string abnormalities based on fixed performance ratio thresholds, module nameplate parameters, initial calibration parameters, or comparisons with adjacent strings. For example, Chinese invention patent application CN107395119B discloses a photovoltaic array fault location method that performs preliminary diagnosis based on abnormal changes in output power of each path and determines the location of faulty branches and components by combining changes in the conduction state of bypass diodes. Similarly, Chinese invention patent application CN108649892B discloses a defect diagnosis method and device for photovoltaic power plants, which determines the PR values ​​of the power plant, power generation units, and components based on photovoltaic power plant operation test data, and diagnoses and locates defects layer by layer based on a PR evaluation threshold model. The above-mentioned solutions can achieve a certain degree of fault detection and location for photovoltaic arrays or photovoltaic power plants. However, they mainly rely on abnormal output power, fixed physical parameters, PR threshold models, or layer-by-layer performance comparisons, and do not adequately consider the impact of natural aging drift of components under long-term service conditions on the correction benchmark and deviation judgment. For long-term service photovoltaic arrays, different strings within the same power generation unit may have different aging differences due to orientation, tilt angle, ventilation conditions, local dust accumulation, local shading, batch differences of components, or different maintenance and replacement records. If the factory nameplate parameters, initial calibration values, or fixed performance ratio thresholds are still used as long-term judgment criteria, the power deviation after irradiation correction will simultaneously include... The fault location includes recoverable disturbances such as common aging of the entire array, individual aging differences in strings, dust accumulation or shading, as well as local fault components such as hot spots, microcracks, poor contact, and abnormal bypass diodes. Since the above components may all manifest as a decrease in output in terms of electrical performance, healthy aging strings can easily be falsely reported as faults, and true early faults can easily be masked by the aging background and missed. Therefore, it is necessary to propose a fault location method that can dynamically update the correction benchmark under comparable irradiance and temperature conditions, and distinguish between aging components and fault residuals from string electrical deviations, so as to reduce false alarms and missed alarms in fault location of long-term photovoltaic arrays and improve the reliability and traceability of fault location results. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a fault location method for long-term photovoltaic arrays based on irradiance correction, which solves the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A fault location method for long-term photovoltaic arrays based on irradiance correction includes: Acquire operational data, meteorological data, and topology data of long-term photovoltaic arrays and conduct quality inspections to form a series of string-by-string operational samples; The time synchronization and spatial mapping of the sequential running samples are performed to obtain the irradiance of each sequential sample; Comparable operating condition intervals are divided according to irradiance, temperature, season or solar altitude angle. Within the same comparable operating condition interval, dynamic irradiance correction benchmarks are established or updated based on healthy samples that have passed the sample consistency screening. Based on the dynamic irradiance correction benchmark, the corrected charge deviation of each string is determined and a deviation trajectory is formed; The fault residual is obtained by subtracting the common aging component of the entire array and the individual aging difference component of the string from the deviation trajectory. Based on the starting point of the change in the fault residual, the historical tolerance range, the operating condition performance, and the continuous occurrence, the fault sequence is determined, and the fault location result is output.

[0005] In one possible implementation, the operational data, meteorological data, and topological data include: The operating data is acquired by the inverter, combiner box, or string monitoring unit; The operational data includes string number, string power, string current, string voltage, acquisition time, inverter, combiner box, and data quality identifier. The meteorological data is obtained from meteorological stations, total radiation meters, irradiance measuring points, or temperature measuring points. The meteorological data includes reference irradiance, component backsheet temperature, ambient temperature, collection time, and measurement point number. The topology data is obtained from the power plant ledger, monitoring platform, combiner box configuration table or string-level monitoring device mapping table; The topology data includes the generator unit number, inverter number, combiner box number, string number, and the corresponding relationship between components or segments.

[0006] In one possible implementation, quality inspection includes: Verify missing fields, physical range of values, continuity of timestamps, matching relationship between data sources and topology data, and self-consistency between power consumption, irradiance, and temperature; Data that fails the quality inspection are marked as abnormal samples, and these abnormal samples are excluded from the current dynamic irradiance correction benchmark update and fault confirmation.

[0007] In one possible implementation, time synchronization and spatial mapping include: Align multi-source runtime records to a unified time window; Data with collection periods shorter than the unified time window are merged according to the data field properties, and the values ​​within the corresponding time window are determined for data with collection periods longer than the unified time window. During historical periods of clear, stable, and unobstructed skies, a spatial mapping coefficient between the reference irradiance and the actual irradiance level of each string is established based on the string installation area, array orientation, tilt angle, reference irradiance measurement point location, and historical output records. Based on the spatial mapping coefficient, the reference irradiance is converted into the irradiance of each string in the corresponding string.

[0008] In one possible implementation, a spatial mapping coefficient is established between the reference irradiance and the actual irradiance levels of each string, including: When multiple irradiation measurement points are set up in the power plant, a reference irradiation measurement point is selected according to the principle of the area where the string is located, the junction box to which it belongs, the orientation of the component, or the closest distance, and the mapping relationship between the reference irradiation measurement point and the actual output of the string is maintained. In the event of an anomaly at the reference irradiance measurement point, switch to a nearby irradiance measurement point, or use a reference irradiance reconstructed from historical clear-sky day data, and set a data quality limitation flag for samples using alternative irradiance data.

[0009] In one possible implementation, comparable operating condition intervals are divided, including: The string power after time synchronization and spatial mapping is paired with the string irradiance sequence, and divided into a comparable working condition sample set according to irradiance range, temperature range, season, solar altitude angle or combination of conditions. Read the operation and maintenance event logs and set the samples corresponding to cleaning, rainfall, or occlusion clearing as the recovery observation samples; After component replacement, wiring maintenance, or bypass diode maintenance, the historical baseline version or aging status identifier of the corresponding string should be re-established or saved in segments.

[0010] In one possible implementation, consistency screening of samples within the same comparable operating condition range includes: Verify the data quality, string status, historical deviation status, and operation and maintenance event status of samples within the same comparable operating condition range. A healthy sample is one that simultaneously meets the following criteria: power, irradiance, temperature, timestamp, and topology relationship. The sample must not be in a fault candidate state, fault confirmation state, or data restriction state. The historical deviation of the sample must not exceed the historical tolerance range. Furthermore, the sample must not be in a period of maintenance, power restriction, communication abnormality, cleaning recovery period, or shading cleaning recovery period. Samples that do not meet the above verification criteria are excluded from the dynamic irradiance correction benchmark update process.

[0011] In one possible implementation, establishing or updating a dynamic irradiance correction baseline includes: Based on the unit irradiance level of the healthy sample set within the retrospective window, a dynamic irradiance correction benchmark is formed for the corresponding comparable operating condition range. The historical allowable range is determined based on the normal residual fluctuation range under the same historical operating conditions, sensor accuracy, inverter or combiner box metering accuracy, acquisition cycle and on-site calibration results. If a historical healthy sample is identified as a faulty or abnormal sample, the historical healthy sample is removed from the healthy sample set, and a baseline version and deviation trajectory for the affected time period are regenerated.

[0012] In one possible implementation, the deviation trajectory is formed by: Within each comparable operating condition range, the measured power, irradiance of each string, spatial mapping coefficient, and dynamic irradiance correction benchmark of each string are read, and the power of each string is converted to the same comparison caliber. The converted string power is compared with the healthy power level in the comparable operating range to obtain the corrected power deviation. Arrange the corrected power deviations of the same series of strings within the same or similar comparable operating conditions in chronological order to form a deviation trajectory.

[0013] In one possible implementation, deducting the common aging component of the entire array and the individual aging difference component of each string includes: Read the corrected power deviation of multiple strings in the same power generation unit, the same combiner box, or the same comparable area, and determine the overall downward shift component that is common to multiple healthy strings, changes slowly over time, and is consistent in direction in different comparable operating conditions as the common aging component of the entire array. For the remaining deviation after deducting the common aging component of the entire array, the deviation that changes slowly over a long period of time relative to the common level of the entire array, has the same direction, and has a similar direction of change in multiple comparable operating condition intervals is determined as the individual aging difference component of the string. The remaining deviation after deducting the common aging component of the entire array and the individual aging difference component of the string is determined as the fault residual.

[0014] In one possible implementation, the fault sequence is determined based on the starting point of the fault residual change, the historical tolerance band, the operating condition performance, and the frequency of occurrence, including: When the fault residual in the same string changes from the stable range to a continuous expansion, or exceeds the historical allowable range and fails to recover to the normal range in subsequent observation windows, or expands relative to the historical same operating conditions under high irradiance, high current, high temperature, local shading sensitive periods or specific solar altitude angles, the string is marked as a fault candidate string. After the string is marked as a fault candidate string, new samples of the string within the corresponding comparable operating condition range are excluded from the dynamic irradiance correction benchmark update. The candidate fault strings are continuously confirmed through multiple observation windows and multiple comparable operating condition intervals, and the fault range is determined based on the topological relationship between the generator unit, inverter, combiner box, string and component.

[0015] In one possible implementation, the fault extent is determined based on the topological relationships between the generator unit, inverter, combiner box, string, and modules, including: With the availability of string segment monitoring, string-level optimizer, current and voltage curves, or bypass diode conduction information, read the current and voltage curves of the faulty string during a clear and stable period or during on-site testing, and compare them with historical curves under the same operating conditions or curves of adjacent healthy strings. If there are steps, peaks, valleys, or slope changes in the current-voltage curve that correspond to the conduction of the bypass diode, local mismatch of the component, or local hot spots, the corresponding change position is matched with the mapping relationship of the component or bypass diode section in the string to determine the component section or bypass section.

[0016] In one possible implementation, the fault location results are output, including: The location reliability is generated based on the fault location results, data quality identifiers, deducted common aging components, individual aging differences in the string, fault residual out-of-bounds magnitude, consistency under multiple operating conditions, number of continuous confirmations, location level, and abnormal degradation status. Output the fault string number, the generator unit to which it belongs, the inverter to which it belongs, the combiner box to which it belongs, the fault candidate start time, the fault confirmation time, the operating condition range involved, the residual out-of-bounds range, the starting point of the residual change, the number of continuous windows, the deviation performance under different operating conditions, the data quality identifier, the confidence level, the fault nature prompt, the suggested review method, and the suggested handling priority.

[0017] Compared with existing technologies, this invention provides a fault location method for long-term photovoltaic arrays based on irradiance correction, which has the following advantages: 1. This invention performs quality inspection, time synchronization, and spatial mapping on the operational, meteorological, and topological data of long-term photovoltaic arrays. It divides comparable operating condition intervals according to irradiance, temperature, season, or solar altitude angle, enabling comparison of the power output of each string under similar irradiance and temperature conditions. By establishing or updating a dynamic irradiance correction benchmark based on healthy samples after consistency screening, the correction benchmark can be adapted to the current service status of the photovoltaic array. By forming a corrected power deviation trajectory and deducting the common aging component of the entire array and the individual aging difference component of each string to obtain the fault residual, continuous confirmation and fault location are performed based on the fault residual. This reduces false alarms and missed alarms caused by static nameplate parameters, initial site calibration values, or fixed performance ratio thresholds in long-term service scenarios, improving the reliability and traceability of fault location results.

[0018] 2. This invention sets abnormal sample identifiers for data that fails quality inspection, and performs alternative data activation, pause processing, downgraded output, or recalculation processing for situations such as abnormal reference irradiance, abnormal temperature sensor, discontinuous timestamps, communication interruption, or insufficient sample quantity. Simultaneously, after cleaning, rain, shading removal, component replacement, wiring repair, or bypass diode maintenance, relevant samples are restored for observation, segmented storage, or a new baseline version is established. This reduces the improper participation of abnormal data, recoverable disturbed samples, and samples with inconsistent states before and after maintenance in baseline updates or fault confirmation, thereby improving the stability of the fault location process and the consistency of result verification for long-term photovoltaic arrays under complex operating conditions. Attached Figure Description

[0019] Figure 1 This is a diagram of the fault location system architecture of the present invention; Figure 2 This is a flowchart of the long-term photovoltaic array fault location method of the present invention; Figure 3 This is a schematic diagram of the dynamic correction benchmark update of the present invention; Figure 4 This is a schematic diagram of the layer-by-layer separation of the deviation component in this invention; Figure 5 This is a schematic diagram illustrating the fault residual determination and continuous verification of the present invention; Figure 6 This is a schematic diagram of the hierarchical positioning and confidence output of the present invention; Figure 7 This is the closed-loop diagram of anomaly handling and recalculation in this invention.

[0020] Among them: 100, Fault Location System; 110, Data Access Module; 120, Synchronization Mapping Module; 130, Operating Condition Collection Module; 140, Baseline Update Module; 150, Component Separation Module; 160, Fault Location Module; 170, Confidence Output Module; 180, Anomaly Degradation Module. Detailed Implementation

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

[0022] This invention provides a fault location method for long-term photovoltaic arrays based on irradiance correction.

[0023] The long-term service photovoltaic array described in this application can be a photovoltaic array that has been in operation for many years and whose module output performance, temperature response characteristics, or inter-string consistency have changed with the service process. This photovoltaic array can be set up in ground-mounted centralized photovoltaic power stations, mountain photovoltaic power stations, agricultural photovoltaic complementary power stations, fishery photovoltaic complementary power stations, or industrial and commercial rooftop distributed power stations. The long-term service period can be determined based on the power station commissioning archives, module warranty data, historical power generation records, and operation and maintenance records. For example, for existing photovoltaic power stations that have been in operation for more than 5 years, the historical operation records can usually reflect the slow degradation trend of module output performance. For photovoltaic power stations that have been in operation for more than 10 years, the differences in aging between strings, differences in local maintenance and replacement, and the cumulative impact of the environment are usually easier to reflect. Therefore, 5 years and 10 years are used to describe the common engineering scope in long-term service scenarios. The values ​​are based on the power station commissioning years, module warranty degradation curves, and historical operation records. In actual implementation, they can be adjusted according to the specific operating status of the power station.

[0024] In existing photovoltaic array fault location processes, data such as string current, voltage, charge, irradiance, and temperature are typically collected. Then, based on fixed performance ratio thresholds, component nameplate parameters, initial calibration parameters, or comparisons with adjacent strings, it is determined whether a particular string is abnormal. This method can complete basic monitoring in newly built power plants or those with mild aging. However, in long-term service power plants, photovoltaic modules are constantly exposed to high temperatures, ultraviolet radiation, humidity, wind and sand, dust accumulation, salt spray, or diurnal temperature variations. As a result, the module conversion efficiency and electrical performance will slowly change over time. Furthermore, different modules within the same power generation unit may also experience variations due to differences in orientation, tilt angle, ventilation conditions, and local conditions. Dust accumulation, localized shading, batch differences in components, or different maintenance and replacement records can all lead to varying degrees of aging differences. If the factory nameplate parameters or initial calibration values ​​are continued to be used as the sole long-term calibration benchmark, the power deviation after irradiation correction will simultaneously include common aging of the entire array, slow aging differences of individual strings, recoverable disturbances such as dust accumulation or shading, as well as local fault components such as hot spots, microcracks, poor contact, and abnormal bypass diodes. Since these components may all manifest as a decrease in output in terms of power performance, a single fixed threshold may easily misjudge naturally decaying healthy strings as faults, or may absorb real early faults into the overall aging background and fail to report them.

[0025] Based on this, the implementation method of this application addresses the problem that static calibration benchmarks in long-term photovoltaic arrays are difficult to reflect the current service status of the modules. It proposes to establish an irradiance calibration benchmark that is dynamically updated with the service progress under comparable operating conditions, and to avoid faulty samples, abnormal samples, or recoverable disturbance samples from entering the health benchmark update process through sample consistency screening. On this basis, the power deviation after irradiance correction is separated into a common aging component of the entire array, a slowly varying aging difference component of individual strings, and a local fault residual. This enables the fault location results to distinguish between natural aging, individual attenuation differences, and local abnormal changes, thereby reducing false alarms and missed alarms caused by long-term static benchmark distortion, and making the output results have traceable confidence and data quality identification.

[0026] The core processing procedure of this application's implementation method is as follows: A series-by-series operation sample is formed from power plant operation data, meteorological data, and topology data under a unified time caliber. This series-by-series operation sample includes at least the following fields: series number, time period power, series current, series voltage, reference irradiance, temperature, timestamp, topology relationship, and data quality identifier. The reference irradiance is converted into an irradiance suitable for each series through time synchronization and spatial mapping. Comparable operating condition intervals are divided according to irradiance, temperature, season, or solar altitude angle. Within the same operating condition interval, healthy samples with qualified data quality and no signs of failure are selected, and these are established and updated. Dynamic irradiance correction benchmark; based on the dynamic irradiance correction benchmark, the corrected deviation trajectory of each string is obtained. First, the slow downward shift component common to multiple strings is deducted, and then the slow, monotonous, and consistent aging difference of a single string relative to the common level is deducted. The remaining part is used as the fault residual; based on the starting inflection point, tolerance band out-of-bounds, operating condition dependence, and continuous occurrence of the fault residual, the fault string and its hierarchical position are determined, and the confidence level, data quality label, and operation and maintenance review suggestions are output. Through the above processing, the risk of fault samples contaminating the health benchmark can be reduced, and the situation of misjudging long-term natural decay as local faults can be reduced.

[0027] In this application, the irradiance correction benchmark refers to the healthy electrical charge level corresponding to a unit irradiance within a given operating condition range. This benchmark is not a fixed nameplate parameter, nor is it a one-time calibration value at the initial stage of station construction. Instead, it is updated based on healthy samples selected through sample consistency screening during long-term service. The spatial mapping coefficient refers to the proportional relationship between the reference irradiance and the actual irradiance level of each string during a period of clear skies, no obstruction, and normal acquisition channels. This spatial mapping coefficient is used to reconstruct the irradiance of each string from the reference irradiance and is not directly used as a fault deviation. The operating condition range refers to a set of comparable samples divided according to conditions such as irradiance energy, component temperature, ambient temperature, season, or solar altitude angle. This is used to avoid including samples from weak light, high temperature, low temperature, low solar altitude angle in the early morning, and noon. Samples with significant differences, such as high solar altitude angle, are directly mixed and compared; the common aging component of the entire array refers to the overall downward shift component that is shared by multiple strings, changes slowly over time, and exhibits consistent behavior across multiple operating conditions; the individual slowly varying aging difference component refers to the continuous offset of a single string relative to the common aging level of the entire array, which is typically slow, monotonous, and consistent in operating conditions; the fault residual refers to the remaining deviation from its own historical trajectory after deducting the common aging component of the entire array and the individual slowly varying aging difference component, used to identify localized progressive faults or relatively sudden faults; the allowable band refers to the permissible deviation range determined based on the historical normal fluctuations, sensor accuracy, sampling period, and field calibration results of the same string within the same operating condition range, used to distinguish between normal fluctuations and abnormal residual expansion.

[0028] The thresholds, intervals, window lengths, continuous confirmation counts, sample quantity conditions, and anomaly judgment conditions involved in the embodiments of this application can be configured or updated based on equipment calibration results, component model information, inverter and combiner box rated parameters, irradiance sensor and temperature sensor accuracy, historical operation records, field test results, operation and maintenance event records, and power plant false alarm tolerance. The data quality threshold can be determined by the acquisition equipment range, string design capacity, inverter rated input range, and historical normal operation boundaries. The operating condition division threshold can be determined by irradiance variation characteristics, component power temperature coefficient, sensor accuracy, and sample distribution. The allowable band can be determined by historical normal residual fluctuations, sensor errors, and measurement errors within the same string in the same operating condition interval. The continuous confirmation conditions can be determined by cloud obstruction, cleaning recovery, communication anomaly recovery, and operation and maintenance response cycles. The above thresholds and intervals can be formed through field calibration in the early stage of power plant commissioning and corrected based on historical samples that have passed quality inspection during long-term operation.

[0029] like Figure 1 As shown, the fault location system 100 provided in this application embodiment may include a data access module 110, a synchronization mapping module 120, a working condition collection module 130, a reference update module 140, a component separation module 150, a fault location module 160, a confidence output module 170, and an anomaly degradation module 180. The data access module 110 can communicate with inverters, combiner boxes, string monitoring units, weather stations, total radiation meters, temperature sensors, and power plant monitoring platforms to acquire photovoltaic array operation data, meteorological data, and topology data. The photovoltaic array operation data includes at least the following fields: string number, string power, string current, string voltage, acquisition time, associated inverter, associated combiner box, and data quality identifier. The meteorological data includes at least the following fields: reference irradiance, ambient temperature, module backsheet temperature, acquisition time, and measurement point number. The topology data includes at least the following fields: power generation unit number, inverter number, combiner box number, string number, and module or segment correspondence. These fields provide a data foundation for subsequent time alignment, irradiance correction, operating condition aggregation, and fault location. The synchronization mapping module 120 is used to perform time grid alignment on data from different acquisition devices, and establish a spatial mapping relationship between the reference irradiance and the actual irradiance level of each string based on the string installation location, array orientation, tilt angle, irradiance measurement point location, and historical clear-sky operation records, thereby obtaining the irradiance applicable to each string; the operating condition collection module 130 is used to divide the time-aligned and spatially mapped operating samples into a comparable operating condition sample set according to irradiance, temperature, season, solar altitude angle, or a combination thereof; the benchmark update module 140 is used to screen healthy samples with qualified data quality and no signs of failure from the comparable operating condition sample set, and establish or update the dynamic irradiance correction benchmark and allowable band accordingly; the component separation module 150 is used to separate the common aging of the entire array based on the corrected deviation trajectory of each string. The system includes components such as chemical composition, individual gradual aging difference components, and fault residuals. The fault location module 160 is used to determine the fault string and its hierarchical location based on the starting inflection point, allowable band out-of-bounds, operating condition dependence, and continuous occurrence of the fault residuals. When current and voltage curves, string segment monitoring data, or bypass diode conduction information are available, it can further assist in determining component segments or bypass segments. The confidence output module 170 is used to output the fault string number, fault hierarchical location, initial judgment of fault nature, location confidence, data quality identifier, and operation and maintenance review suggestions. The anomaly degradation module 180 is used to perform pause update, re-acquisition, enable alternative data, degradation output, manual review prompts, or recalculation processing when there are abnormal reference irradiance, abnormal temperature sensor, missing power, time asynchrony, insufficient samples, communication interruption, or state conflict. The execution entity in this application embodiment can be at least one of the following: a centralized monitoring platform for power plants, field edge devices, inverter control units, combiner box monitoring units, string-level monitoring units, or string-level optimizers. The above modules can be deployed in the same processing device or can be implemented collaboratively by multiple processing devices. For centralized photovoltaic power plants, data access, reference updates, component separation, and result output can be performed by the centralized monitoring platform, while time synchronization, spatial mapping, and anomaly handling can be performed collaboratively by the field edge devices and the centralized monitoring platform. For photovoltaic power plants with unstable communication conditions or complex terrain, the field edge devices can first complete local data alignment, spatial mapping, and fault candidate identification. After communication is restored, the centralized monitoring platform can then perform a full-site joint aging review, confidence correction, and result recalculation. For photovoltaic power plants with string-level optimizers or segmented monitoring capabilities, the fault location module 160 can read the segmented charge, current and voltage curves, or bypass diode conduction information within the string after confirming the faulty string, so as to further converge the fault range from the string to the module segment or bypass segment.

[0030] like Figure 2As shown, the method provided in this application embodiment may include steps S101 to S110; each number is used to facilitate the description of the processing flow of the method and does not mean that each processing action must be executed in the listed order; without affecting the data dependency and technical effect, some processing in data quality inspection, anomaly identification and status identifier update can be executed in parallel with the main process. In step S101, multi-source operating data is acquired and verified. When the photovoltaic power station is in normal grid-connected operation and the acquisition channel is available, the data access module 110 acquires the operating data, meteorological data, and topology data of each string or branch within the same statistical period. The operating data includes at least the string number, cumulative power generation during the time period, DC current, DC voltage, acquisition time, inverter, combiner box, and data quality identifier. The meteorological data includes at least the reference irradiance, module backsheet temperature, ambient temperature, acquisition time, and measurement point number. The topology data includes at least the power generation unit number, Fields include inverter number, combiner box number, string number, and module or segment correspondence; string power can be cumulative power over a period, daily power, or branch power, and its source can be the inverter, combiner box, or string monitoring unit; reference irradiance can be cumulative irradiance energy over a period or cumulative daily irradiance, and its source can be a weather station, total radiation meter, nearby radiation measuring points, or irradiance measurement information that has been confirmed on-site as an alternative; temperature can be module backsheet temperature, ambient temperature, nearby measuring point temperature, or daily average temperature; topology data can come from power plant ledgers, monitoring platforms, combiner box configuration tables, or mapping tables of string-level monitoring equipment. The data access module 110 performs field unification and quality checks on the above data. The quality checks may include checking for missing fields, whether values ​​are within the physically reachable range, whether timestamps are continuous, whether the data source matches the topology data, and whether there are inconsistencies between power consumption, irradiance, and temperature. The physically reachable range can be determined based on the inverter's rated input range, the combiner box's measurement range, the string design capacity, the module's open-circuit voltage, the module's short-circuit current, the irradiance sensor's range, and the temperature sensor's range. Inconsistencies can be determined based on historical normal operating boundaries, the changing trends of adjacent healthy strings within the same combiner box, the correspondence between irradiance and string power consumption, and on-site calibration results. For example, when the reference irradiance is low and there is no energy storage reaction... In the case of abnormal data transmission or metering records, if the power consumption of a certain string during a certain period is higher than the historical normal boundary of the same area, then the sample can be marked as a sample to be reviewed; for example, when the temperature sensor remains unchanged within a continuous acquisition cycle, or when the difference between the temperature sensor and adjacent temperature measurement points exceeds a preset range, the corresponding temperature sample will not be directly included in the temperature condition classification; the number of continuous acquisition cycles can be determined based on the acquisition cycle, the temperature sensor resolution, and the historical temperature change records of the same area; the preset range of the difference between adjacent temperature measurement points can be determined based on the temperature sensor accuracy, the historical temperature difference of the same area, and the on-site calibration results; data that fails the quality inspection can be marked as abnormal samples, will not participate in the current dynamic irradiance correction benchmark update, and will not be directly used as the basis for confirming deterministic faults; In one possible implementation, the data access module 110 can acquire raw data at a minute-level acquisition cycle and generate corresponding statistical data at statistical cycles of 15 minutes, 30 minutes, 1 hour, or 1 day. The aforementioned acquisition and statistical cycles can be determined based on the inverter interface capabilities, combiner box monitoring capabilities, meteorological station sampling cycles, and power plant operation and maintenance accuracy requirements. Among these, 15 minutes and 30 minutes are common time-period statistical cycles in photovoltaic power plant monitoring systems, inverters, or combiner boxes, and are suitable for string-level trend tracking. The 1-hour cycle is suitable for scenarios with limited communication resources but still requiring daily operating condition data collection. The 1-day cycle is suitable for existing power plants that only have daily electricity generation and daily cumulative irradiance data. For power plants with current and voltage curve acquisition capabilities, current and voltage curve data can be used as an enhanced data source and, after passing quality inspection, for subsequent fault section location, but are not considered as the minimum necessary data for implementing this method.

[0031] In step S102, time synchronization and spatial mapping are performed. The synchronization mapping module 120 aligns the multi-source operation records from the inverter, combiner box, weather station, temperature sensor, and monitoring platform to a unified time window. The unified time window can be configured according to the capabilities of the acquisition equipment and the power plant monitoring requirements, and the value can be 15 minutes, 30 minutes, 1 hour, or 1 day. The above values ​​match the acquisition cycle of common inverters, combiner boxes, weather stations, and power plant monitoring systems. In actual implementation, the values ​​can be adjusted according to the on-site acquisition cycle, communication conditions, and operation and maintenance accuracy requirements. For data with an acquisition period shorter than the unified time window, the synchronization mapping module 120 can merge the data according to the cumulative value, average value, or end value within the unified time window. Specifically, the cumulative value can be used for power and irradiance, while the average value or end value can be used for current, voltage, and temperature. The specific method can be determined based on the nature of the data fields. For data with an acquisition period longer than the unified time window, the synchronization mapping module 120 can determine the value within the corresponding time window according to the principle of time proximity, effective retention time, or on-site configuration rules. After alignment, the time-synchronized operation record should include at least the string number, unified time window, merged string power, string current, string voltage, matched reference irradiance, temperature, and data quality identifier. If the timestamp offset of multi-source data exceeds the allowable range, the anomaly degradation module 180 can trigger realignment; time periods that cannot be realigned will not participate in fault confirmation; the allowable range of timestamp offset can be determined based on the unified time window length, the clock synchronization accuracy of each acquisition device, the statistical results of communication delay, and the time calibration rules of the power plant monitoring system; for example, when the unified time window is 15 minutes, the allowable range of timestamp offset can be less than a portion of the unified time window, and the specific proportion can be determined by the on-site communication delay, the clock error of the acquisition device, and the time calibration rules of the operation and maintenance platform; through this processing, the risk of misjudgment caused by mismatched matching of power, irradiance, and temperature data can be reduced; After time synchronization is completed, the synchronization mapping module 120 establishes a spatial mapping coefficient between the reference irradiance and the actual irradiance level of each string based on historical data of clear, stable, and unobstructed skies. The historical period of clear, stable, and unobstructed skies can be determined by the variation range of the reference irradiance, the variation trend of neighboring irradiance measuring points, records of rapid cloud changes, rainfall records, shading records, inverter power-limiting status, and maintenance events. For example, if the reference irradiance changes relatively stably within the continuous acquisition window, the variation trend of neighboring irradiance measuring points is consistent, no rainfall, clearing, shading clearing, or power-limiting events are recorded, and the output change direction of adjacent healthy strings is consistent, then this period can be used as a candidate period for establishing the spatial mapping coefficient. The judgment condition of relatively stable changes can be determined by the sampling frequency of the irradiance sensor, the historical clear-sky daily irradiance variation curve, on-site meteorological station records, and the consistency of output of adjacent strings. The synchronous mapping module 120 can determine the proportional relationship between the reference irradiance and the actual irradiance level of the string within a confirmed healthy, clear, stable, and unobstructed historical period, combining the string installation area, array orientation, tilt angle, reference irradiance measurement point location, and historical output records, thus forming the spatial mapping coefficient corresponding to the string. This spatial mapping coefficient is used to correct the difference between the measurement results of the reference irradiance measurement point and the actual irradiance level of the corresponding string, and to convert the reference irradiance into an irradiance applicable to the corresponding string. The spatial mapping record includes at least the string number, reference irradiance measurement point number, unified time window, spatial mapping coefficient, applicable season, applicable solar altitude angle range, number of samples established, validity period, and data quality identifier. For photovoltaic arrays with seasonal shading, mountain shadows, different tilt angles, or different orientations, the spatial mapping coefficient can be maintained separately by season, solar altitude angle, installation area, or validity period. When events such as clearing vegetation, removing obstructions, or adjusting the module arrangement change the irradiance conditions, the spatial mapping coefficient of the affected string can be re-established. When a power station has multiple irradiance measurement points, the synchronization mapping module 120 can select the irradiance measurement point that matches the corresponding string as a reference irradiance measurement point according to the string's location, its junction box, component orientation, or proximity. It maintains the mapping relationship between the reference irradiance measurement point and the actual output of the string in the historical health samples of the same string. When the reading of a reference irradiance measurement point exceeds the range of the irradiance sensor, exhibits continuous drift, or is significantly inconsistent with neighboring irradiance measurement points and historical clear-sky day conditions, the anomaly degradation module 180 marks the reference irradiance measurement point as abnormal. For reference irradiance measurement points marked as abnormal, a temporary switch to a neighboring irradiance measurement point can be made, or a reference irradiance amount reconstructed based on historical clear-sky day data can be used. Samples using alternative irradiance data should have a data quality limitation indicator, and the confidence level of the corresponding result should be reduced in subsequent confidence level outputs. This processing method is suitable for mountainous photovoltaic power stations, multi-orientation rooftop photovoltaic power stations, or scenarios with unevenly distributed irradiance measurement points.

[0032] like Figure 3 As shown, in step S103, operating condition collection is performed; the operating condition collection module 130 pairs the string power after time synchronization and spatial mapping with the string-by-string irradiance sequence, and divides them into multiple comparable operating condition sample sets according to irradiance range, temperature range, season, solar altitude angle or a combination thereof; the comparable operating condition sample set includes at least the following fields: operating condition number, irradiance range, temperature range, applicable season, solar altitude angle range, string number, statistical period, number of samples, number of valid samples, data quality identifier and maintenance event identifier; through the above fields, the establishment of subsequent dynamic irradiance correction benchmark, the generation of correction deviation trajectory, and the judgment of fault residual can all be carried out under similar operating conditions; The irradiance range can be determined based on the measurement error in the low-light segment, the nonlinear characteristics of the module efficiency change with irradiance, the accuracy of the irradiance sensor, and the distribution of on-site samples. For example, the irradiance operating conditions can be divided into ranges from 150W / m² to 250W / m², with an optimal range width of approximately 200W / m². The basis for setting this range is that if the irradiance range is too narrow, the number of healthy samples available for long-term service power plants under the same operating conditions may be insufficient; if the irradiance range is too wide, the differences in module efficiency under low-light, medium-irradiance, and high-irradiance conditions can easily be mixed and compared, thus affecting the accuracy of subsequent deviation judgments. An interval width of approximately 200W / m² can achieve a good engineering balance between the number of samples and the comparability of operating conditions. In actual implementation, it can be adjusted according to the accuracy of the irradiance sensor, the acquisition cycle, the number of historical samples, the solar resource conditions at the power plant location, and the requirements for operation and maintenance accuracy. The temperature range can be determined based on the module power temperature coefficient, temperature sensor accuracy, historical temperature distribution, and the number of field samples. For example, the module temperature or ambient temperature operating conditions can be divided according to a temperature step size of 5℃ to 10℃. The basis for setting this range is that the output power of photovoltaic modules is significantly affected by temperature. If the temperature range is too wide, the temperature difference within the same set of operating conditions will have a significant impact on the comparison of power output. If the temperature range is too narrow, the number of usable samples may be insufficient. A temperature step size of 5℃ to 10℃ can usually keep the temperature influence within the same temperature range within a calibrable and comparable range. In actual implementation, the specific step size can be determined by combining the module power temperature coefficient, temperature sensor error, historical temperature distribution, and field test results. For power plants with daily electricity consumption as the basic statistical period, the operating condition collection module 130 can classify comparable operating conditions according to the daily cumulative irradiance and the daily average module temperature. For power plants with relatively complete time period data, the operating condition collection module 130 can classify comparable operating conditions according to the time period irradiance and the time period module backsheet temperature. For power plants with obvious seasonal shading, mountain shadows, or multi-directional layout, the operating condition collection module 130 can further subdivide comparable operating conditions by combining the season, solar altitude angle, or installation area to avoid samples under different irradiation conditions being directly mixed and compared. When forming the operating condition sample set, the operating condition collection module 130 can also read the operation and maintenance event records. The operation and maintenance event records include at least the following fields: event type, event occurrence time, involved power generation unit, involved inverter, involved combiner box, involved string, event duration, event processing status, and record source. The event type can include cleaning, rainfall, shading cleaning, component replacement, inverter power limiting, maintenance shutdown, combiner box maintenance, and sensor calibration. If a cleaning or rainfall event occurs within a certain statistical period, the samples within that statistical period and its adjacent recovery period can be set as recovery observation samples to determine whether the subsequent residuals belong to recoverable disturbances caused by dust accumulation or shading, rather than being directly used as the basis for confirming unrecoverable hardware failures. If a string undergoes component replacement, wiring maintenance, or bypass diode maintenance, the historical baseline version and aging status identifier corresponding to that string can be re-established or saved in segments after maintenance to avoid mixing samples before and after maintenance for the same aging trajectory. Through the above-mentioned working condition aggregation processing, subsequent irradiance correction benchmark updates, deviation trajectory generation, aging component separation, and fault residual judgment can all be based on a comparable working condition sample set, thereby reducing the interference of different irradiation conditions, temperature conditions, seasonal conditions, or maintenance events on fault location results.

[0033] In step S104, a dynamic irradiance correction benchmark is established and updated. The benchmark update module 140 reads historical healthy samples and current quality-qualified samples in each operating condition interval, and determines the samples that can participate in the benchmark update according to the sample consistency screening rules. The sample consistency screening rules can be judged from four aspects: data quality, string status, residual status, and operation and maintenance events. Among them, in terms of data quality, the power, irradiance, temperature, timestamp, and topology relationship in the sample must all pass the quality inspection. In terms of string status, the string to which the sample belongs must not be in a fault candidate state, fault confirmation state, or data restricted state. In terms of residual status, the sample must not have a starting inflection point, must not exceed the historical allowable zone, and must not show an operating condition dependence inconsistent with natural aging. In terms of operation and maintenance events, the sample must not be in an operating state that would affect the output stability, such as maintenance, power limitation, communication abnormality, cleaning recovery period, or shading cleaning recovery period. The samples that pass the above screening can be included in the healthy sample set of the operating condition interval for subsequent establishment or updating of the dynamic irradiance correction benchmark. The health sample set includes at least the following fields: operating condition number, string number, statistical period, unit irradiance power, reference irradiance, string irradiance, temperature, sample source, data quality identifier, string status identifier, maintenance event identifier, and whether it participates in the baseline update. Through these fields, the baseline update module 140 can identify the operating condition, sample source, operating status, and availability of the sample, avoiding the inclusion of faulty samples, abnormal samples, or recoverable disturbance samples in the health baseline. The baseline update module 140 can form a dynamic irradiance correction baseline for the operating condition range based on the unit irradiance energy level of the healthy sample set within the backtracking window. The dynamic irradiance correction baseline includes at least the following fields: operating condition number, irradiance range, temperature range, healthy energy level per unit irradiance, allowable band, number of healthy samples, excluded samples, screening rule version, baseline version, update time, and applicable status. The backtracking window can be set from several months to about one year, and its setting basis includes the natural aging rate of the components, the data collection cycle, the number of samples under the same operating condition, and the power plant operation and maintenance cycle. The natural aging of components is usually reflected slowly on a monthly or yearly scale. If the backtracking window is too short, it is easily affected by short-term factors such as weather, shading, cleaning, or communication anomalies. If the backtracking window is too long, it may not be able to reflect the current health level after long-term service in a timely manner. Therefore, in actual implementation, an appropriate backtracking window can be selected within the range of several months to about one year according to the number of power plant samples, the component degradation rate, and operation and maintenance requirements. If it is subsequently discovered that some historical samples are actually faulty or abnormal samples, the baseline update module 140 can remove the sample from the historical healthy sample set according to the recalculation instruction, and regenerate the baseline version and deviation trajectory for the affected time period. Through the baseline version record and sample exclusion record, the health baseline on which a fault judgment is based can be traced during operation and maintenance review, avoiding the interpretation of fault location results due to the lack of clarity in the baseline update process. The allowable band can be determined based on the historical normal residual fluctuation amplitude under the same operating conditions, sensor accuracy, inverter or combiner box metering accuracy, acquisition cycle, and on-site calibration results. The historical normal residual fluctuation amplitude under the same operating conditions can be determined from historical samples of the same string within the same operating condition range, excluding those under maintenance, power limiting, cleaning recovery, obstruction clearing, and fault candidate states. If the number of historical samples for the same string is insufficient, the historical fluctuation amplitude of adjacent healthy strings within the same combiner box can be used as a temporary reference, and recalculated after subsequent samples are added. For example, the allowable band can be... The value is set to 2 to 3 times the historical normal residual fluctuation range, or it can be configured by the operation and maintenance personnel based on the on-site test results. The basis for setting it to 2 to 3 times is that when the tolerance band is too narrow, short-term cloud shadows, sampling errors, or sensor fluctuations can easily trigger false alarms; when the tolerance band is too wide, early fault residuals may not be able to be displayed in time. Using 2 to 3 times the historical normal residual fluctuation range can achieve a good engineering balance between suppressing false alarms and retaining early fault sensitivity. In actual implementation, it can be adjusted according to the equipment accuracy, historical sample stability, and on-site false alarm tolerance. For existing power plants that only have daily electricity generation, daily cumulative irradiance, and daily average temperature, the benchmark update module 140 can maintain the dynamic irradiance correction benchmark with a statistical cycle of 1 day. At this time, the operating condition range can be determined by the daily cumulative irradiance level and the daily average temperature level. The healthy sample set can be composed of daily samples that are of qualified quality and not in the fault candidate state in the past few months to about 1 year. The positioning range of this method can usually be down to the string level, but it can still reduce the misjudgment caused by using a static benchmark through dynamic benchmark updates and aging component separation.

[0034] As shown in the figure, in step S105, the corrected deviation trajectory is calculated; the component separation module 150 reads the measured power of each string, the irradiance of each string, the spatial mapping coefficient and the dynamic irradiance correction benchmark in each operating condition interval, and converts the power of each string to the same comparison caliber; the converted string power is compared with the healthy power level in the operating condition interval to obtain the corrected power deviation for each string, each operating condition and each statistical period; the corrected power deviation is used for subsequent aging component separation and fault residual judgment, and is not directly used as the sole basis for confirming the fault; The component separation module 150 can arrange the corrected power deviations of the same string within the same or similar operating condition range into a deviation trajectory according to time sequence. The deviation trajectory includes at least the following fields: string number, statistical period, operating condition number, measured power, irradiance received by each string, dynamic irradiance correction benchmark version, corrected power deviation, sample quantity, data quality identifier, and maintenance event identifier. Through the above fields, the operating condition, benchmark version, and data status on which the deviation is based can be recorded, so that subsequent processing can not only determine whether the power is too low within a single statistical period, but also observe the direction of the deviation, the rate of change, and the correspondence with irradiance, temperature, or maintenance events.

[0035] like Figure 4 As shown, in step S106, the aging component is separated. The component separation module 150 can divide the deviation in the deviation trajectory into the original correction deviation, the common aging component of the entire array, the individual aging difference component of the string, and the fault residual. First, within the same processing cycle or the same observation window, the component separation module 150 reads the corrected power deviation of multiple strings in the same power generation unit, the same combiner box, or the same comparable area, and identifies the overall downward shift component that is common to multiple healthy strings. This overall downward shift component is spatially manifested as multiple healthy strings changing in the same direction, temporally manifested as a slow change, and in different comparable operating condition intervals, it is manifested as a consistent direction, which can be used as the common aging component of the entire array. The multiple healthy strings can refer to strings in the same comparable string set that have passed data quality inspection and are not in a fault candidate state, fault confirmation state, or data restricted state. The number of healthy strings participating in the joint aging judgment can be determined based on the number of strings connected to the combiner box, the number of historical health samples, and the on-site false alarm tolerance. For example, the number of healthy strings can be required to exceed half of the total number of strings in the same comparable string set, or a higher proportion can be adopted according to the configuration of the power plant operation and maintenance platform. This setting can reduce the impact of individual abnormal strings on the joint aging judgment. The common aging component of the entire array mainly reflects the common offset of the entire array or a comparable area caused by long-term service, environmental effects, or overall component degradation. For recoverable factors such as overall dust accumulation, snow accumulation, cleaning recovery, or changes in shading, the component separation module 150 can separately identify them in conjunction with maintenance event records, meteorological records, and residual recovery status, instead of directly treating them as unrecoverable aging results. After deducting the common aging component of the entire array from the corrected power deviation of each string, the component separation module 150 can reduce the situation where all strings are mistakenly identified as large-area failures when they uniformly decline with the service life. Subsequently, the component separation module 150 further processes the remaining deviation after deducting common aging for each string. If the remaining deviation of a certain string shows a long-term slow change with a basically consistent direction relative to the common level of the entire array, and shows a similar direction of change in multiple operating conditions such as high irradiance, low irradiance, high temperature, and low temperature, then the remaining deviation can be regarded as the individual aging difference component of the string. This individual aging difference component of the string can be caused by string orientation, ventilation conditions, local dust accumulation, component batch differences, installation location differences, or local operating environment differences. The rate of change of the aging difference component of individual modules can be determined by combining the module warranty documents, the module model of the project, historical power generation records, and on-site sampling results. For crystalline silicon photovoltaic modules, the first-year degradation rate and subsequent annual degradation rates can usually be obtained from the module warranty documents or manufacturer technical data. In specific power plants, it can also be corrected by combining the historical power generation degradation records of the same model of modules. Therefore, the rate of aging change can be different for different power plants, different module models, or different operating environments, and a uniform fixed value is not used. If the residual deviation of a string suddenly increases from a certain statistical period or a certain observation window, or only significantly increases under high irradiance, high current, high temperature or a specific solar position, then the deviation will not be treated as an individual aging difference component of the string, but will be retained as a candidate fault residual. The candidate fault residual is used to subsequently determine whether there are hot spots, microcracks, poor contact, bypass diode abnormalities, local shading sensitivity or other local abnormalities. In the case of multiple orientations, tilt angles, shading environments, or installation areas in the same power plant, the common aging levels of different areas may be different. The component separation module 150 can first divide the comparable string sets according to the topology, array orientation, tilt angle, junction box affiliation, installation area, or historical spatial mapping relationship, and then determine the common aging component of each comparable string set. In this way, the deviations formed under different irradiation conditions can be avoided by directly mixing and processing them.

[0036] In one possible implementation, if the power plant has auxiliary data such as equivalent series resistance, equivalent parallel resistance, temperature coefficient change, or current-voltage curves, the component separation module 150 can also use the above auxiliary data to correct the individual aging difference components of the string or the location confidence. The above auxiliary data can be updated during periods of clear skies, stable temperatures, or when the current-voltage curve quality is acceptable. If the equivalent series resistance increases slowly and the string power deviation also shows a long-term slow decrease, this trend can support the judgment of individual aging differences of the string. If the electrical parameters change abruptly in the short term, accompanied by a fault residual inflection point or residual expansion under high current conditions, this trend can support the judgment of fault residuals. When the above auxiliary data is not available, this method can still perform string-level fault location based on string power, irradiance, temperature, and timestamps.

[0037] like Figure 5 As shown, in step S107, the fault residual is determined; the fault location module 160 takes the remaining deviation after deducting the common aging component of the entire array and the individual aging difference component of the string as the fault residual; the fault location module 160 does not directly confirm the fault based on the residual exceeding the limit in a single statistical period, but makes a comprehensive judgment by combining the starting point of residual change, historical tolerance band, performance under different working conditions and continuous occurrence. The starting point of residual change can be manifested as follows: the fault residual of the same series changes from a stable state to a continuously expanding state starting from a certain statistical period or a certain observation window; the historical tolerance zone can be the allowable range formed by the historical normal fluctuations of the same series within the same operating condition interval; when the fault residual exceeds the historical tolerance zone and does not recover to the normal range in subsequent observation windows, it can be considered that the series has residual over-limit phenomenon; the performance under different operating conditions can include: the fault residual is significantly expanded relative to the historical state under the same operating conditions under high radiation, high current, high temperature, sensitive periods of local shading, or specific solar altitude angle, while it is weaker under other operating conditions; the above performance is different from the slow, monotonous and basically consistent natural aging under multiple operating conditions, and can be used as the basis for judging local anomalies; The criteria for judging the starting point of residual change, residual exceeding the limit, and performance under different operating conditions can be determined based on the historical stable range of the same string, the allowable range under the same operating condition, the component power temperature coefficient, the sample distribution during high irradiation periods, sensor accuracy, measurement accuracy, and on-site fault verification records. If a string simultaneously meets at least one or more of the following criteria: continuous expansion of residual, exceeding the historical allowable range, and abnormal amplification under specific operating conditions, the fault location module 160 can mark the string as a candidate fault string. After being marked as a candidate fault string, new samples of the string within the corresponding operating condition range will no longer participate in the dynamic irradiance correction benchmark update to avoid faulty samples affecting the health benchmark. The fault location module 160 can also output fault nature prompts based on the performance of the fault residual. If the fault residual is more pronounced under high irradiance and high temperature conditions, and continues to exceed the historical allowable range after deducting the aging component, it can output a review suggestion for hot spots, microcracks, or local contact anomalies. If the fault residual has a strong correlation with a certain solar position, a certain time period, or local shading conditions, and falls back to the historical allowable range after cleaning, rainfall, or shading removal, it can output shading, dust accumulation, or other recoverable disturbance prompts, rather than directly confirming it as a hardware fault. If the fault residual continues to expand in multiple operating condition ranges, and there is no obvious recovery relationship with maintenance events such as cleaning, rainfall, maintenance, or shading removal, the fault confirmation priority of this string can be increased. The above fault nature prompts are used to guide on-site review and do not replace infrared detection, electrical testing, or manual inspection results.

[0038] like Figure 5 As shown, in step S108, hierarchical location is continuously confirmed and performed; the fault location module 160 continuously confirms the fault candidate string through multiple observation windows and multiple operating condition intervals; the observation window can be several days to two weeks, or it can be multiple continuous sampling windows; the basis for setting this time range includes the duration of cloud cover, the recovery time of short-term communication anomalies, the period of short-term dust or bird droppings impact, the cleaning recovery period, and the operation and maintenance response period; the observation window should be longer than the common recovery time of occasional disturbances to reduce misjudgments caused by instantaneous disturbances; at the same time, it should be shorter than the backtracking window used to identify aging trends in order to detect local faults in a timely manner; In one possible implementation, the fault location module 160 may require that a fault candidate string repeatedly exhibits anomalies within approximately three consecutive observation windows and deviates in the same direction at least within two comparable operating condition intervals before it is confirmed as a fault string. The basis for setting approximately three consecutive observation windows and at least two comparable operating condition intervals is that a single observation window or a single operating condition is easily affected by cloud shadows, bird droppings, sampling errors, communication misalignments, or temporary obstructions, while multiple observation windows and multiple operating conditions simultaneously exhibiting deviations in the same direction can improve the stability of fault confirmation. In actual implementation, the above continuous confirmation conditions can be adjusted according to the field false alarm tolerance, data acquisition cycle, equipment accuracy, and maintenance response requirements. When performing hierarchical location, the fault location module 160 can determine the fault range based on the topological relationship between the power generation unit, inverter, combiner box, string and component. If multiple combiner boxes in the same power generation unit show similar common decline, it can be preferentially judged as regional aging, environmental disturbance, abnormal irradiation measurement point or limited power generation state, and verified by the abnormal degradation module 180. If only a few strings under a combiner box have abnormal residual faults, while other strings under the same combiner box are operating stably, the fault range can be converged to the specific string under that combiner box. If there are no obvious abnormalities in the adjacent strings of the same string, but the string continuously exceeds the historical allowable range in multiple operating condition intervals, the specific fault string number can be output. With string segment monitoring, string-level optimizer, current-voltage curves, or bypass diode conduction information, the fault location module 160 can further determine the component segment or bypass section. Specifically, the fault location module 160 can read the current-voltage curves of the faulty string during a clear and stable period or during on-site testing, and compare them with historical curves under the same operating conditions or curves of adjacent healthy strings. If the curves show steps, peaks, valleys, or slope changes corresponding to bypass diode conduction, component local mismatch, or local hot spots, the corresponding change positions can be matched with the mapping relationship of the component or bypass diode section within the string, and the possible bypass section or component segment can be output. If the current-voltage curve data is missing, the acquisition conditions are unstable, or the curve quality inspection fails, this enhanced location processing is not performed, the fault location result can be maintained at the string level, and the location level is limited in the data quality identifier. like Figure 6As shown, in step S109, the confidence level and handling result are output; the confidence output module 170 generates the location confidence level based on the fault location result, data quality identifier, deducted common aging component, individual aging difference of the string, fault residual out-of-bounds range, consistency under multiple operating conditions, number of continuous confirmations, location level, and abnormal degradation status; the confidence level can be output in a graded manner, such as high confidence, medium confidence, and low confidence; or it can be output using a preset score range; the determination of the confidence level should be traceable to the data, operating conditions, residuals, windows, and abnormal handling records involved in the judgment; High confidence can correspond to situations where fault residuals significantly exceed the historical tolerance range, multiple operating condition intervals deviate in the same direction, multiple observation windows provide continuous confirmation, and the data quality is good with clear topological positioning; medium confidence can correspond to situations where fault residuals have exceeded the limits but the sample size is limited, some operating conditions are not sufficiently confirmed, or there is a possibility of recoverable disturbances; low confidence can correspond to situations where there is abnormal irradiance, abnormal temperature, time asynchrony, insufficient samples, substitution data is used for judgment, or the positioning level is limited. The fault location record output by the confidence output module 170 includes at least the following fields: fault string number, associated generator unit, associated inverter, associated combiner box, optional component or bypass section number, fault candidate start time, fault confirmation time, affected operating condition range, residual boundary crossing magnitude, residual change start point, number of continuous windows, deviation performance under different operating conditions, data quality identifier, confidence level, fault nature prompt, suggested review method, and suggested handling priority. For high-confidence faults, a fixed-point investigation or handling work order can be output. For medium-confidence faults, suggestions for infrared verification, electrical testing, or continuous observation can be output. For low-confidence or data-limited situations, prompts for manual takeover and supplementary data collection can be output, and the reason for the limitation can be explained in the fault location record.

[0039] like Figure 7 As shown, in step S110, exception handling and recalculation are performed; the exception degradation module 180 can receive exception events at any processing stage and trigger corresponding processing according to the exception type; the exception events may include abnormal reference irradiance, abnormal temperature sensor, missing power or electrical quantity, discontinuous timestamp, asynchronous time of multi-source data, communication interruption, insufficient sample quantity, operation and maintenance event conflict, and abnormal data pairing relationship, etc. If the reference irradiance reading exceeds the range of the irradiance sensor, exhibits continuous drift, or is significantly inconsistent with the status of neighboring irradiance measurement points and historical clear-sky days, the anomaly degradation module 180 can use the reference irradiance from neighboring irradiance measurement points at the same station, irradiance measurement points in the same region, or a reference irradiance reconstructed based on historical clear-sky day data as alternative data. If no reliable alternative data exists, the irradiance correction judgment for the corresponding time period is suspended, and a data restriction flag is output. If the temperature sensor experiences a sudden change, its reading remains unchanged for multiple consecutive acquisition cycles, or the difference between it and neighboring temperature measurement points exceeds a preset range, the temperature of neighboring measurement points, ambient temperature, or daily average temperature can be used as alternative data, and the confidence level of the corresponding fault location result is reduced. The conditions for enabling the above alternative data can be determined based on the sensor range, sensor accuracy, consistency of neighboring measurement points, historical clear-sky day records, historical temperature change records, and on-site calibration results. If the proportion of missing power or electrical quantity exceeds a set limit, or if the timestamps are discontinuous, the abnormal degradation module 180 can exclude the missing samples from the dynamic irradiance correction benchmark update and fault confirmation process; when the remaining samples are insufficient to support benchmark update or fault confirmation, a supplementary sampling request can be output; the missing proportion threshold and sample quantity requirements can be determined based on the collection cycle, statistical cycle, backtracking window length, number of samples under the same operating condition, number of samples required for benchmark update, and the real-time requirements of the operation and maintenance system for alarms; for scenarios of creating a new benchmark or benchmark recalculation, the benchmark update module 140 can update the benchmark only after the number of healthy samples reaches the preset minimum number of samples; the preset minimum number of samples can be determined by the collection cycle, backtracking window length, frequency of occurrence under the same operating condition, and historical sample stability; If multi-source data is out of time or out of order, the anomaly degradation module 180 can trigger realignment. During periods when realignment cannot be completed, the data will not participate in the pairing calculation of power, irradiance, and temperature, nor will it participate in fault confirmation. If the pairing relationship between power and irradiance in the same string deviates significantly from the long-term spatial mapping relationship, the anomaly degradation module 180 can first review the acquisition channel, irradiance measurement point, time synchronization status, and topology relationship before determining whether to enter the fault candidate process. If the fault residual falls back to the historical tolerance range after rainfall, cleaning, or shading removal, the corresponding fault candidate can be rolled back to a recoverable disturbance, and a cleaning, shading, or dust accumulation processing prompt can be output. If communication is interrupted, the field edge device can retain the local judgment result. After communication is restored, the centralized monitoring platform can combine the retransmitted data for review, summary, and result recalculation. When performing replacement, pause, degradation output, manual review prompts, or recalculation processing, the anomaly degradation module 180 can generate an anomaly control record. The anomaly control record includes at least the following fields: trigger time, anomaly type, involved power generation unit, involved inverter, involved combiner box, involved string, involved data source, alternative data source, pause processing step, recovery conditions, recalculation scope, processing result, and data quality identifier. Through these fields, the source, impact range, and processing result of the anomaly event can be recorded, enabling subsequent fault location results to be traced back to the corresponding data state and processing action. After data recovery, sensor calibration, or sample replenishment, the anomaly degradation module 180 can trigger the re-execution of the affected links and their subsequent links. For example, during the anomaly of the reference irradiance sensor, a nearby irradiance measurement point is used as substitute data. After the sensor is restored and calibration is completed, the system can re-execute time synchronization, spatial mapping, operating condition collection, dynamic irradiance correction benchmark update, deviation trajectory generation, and fault residual judgment for the statistical period involved during the anomaly, and replace the original data-limited results with the recalculated results. Through anomaly control recording and recalculation processing, this method can maintain traceability in the event of sensor anomalies, communication anomalies, or insufficient samples, avoiding the output of deterministic fault conclusions without evidence based on temporary substitute data.

[0040] In this embodiment, a state feedback relationship can be established between the update of the dynamic irradiance correction benchmark and the fault determination. After the fault location module 160 identifies a certain string as a fault candidate, new samples of that string within the corresponding operating condition range are suspended from participating in the dynamic irradiance correction benchmark update to avoid abnormal samples being included in the healthy sample set. If subsequent continuous confirmation results indicate that the string is a recoverable disturbance and the fault residual recovers to the historical allowable range after cleaning, rainfall, or shading removal, the benchmark update module 140 can allow the quality qualified samples of that string to re-enter the healthy sample set after the recovery observation period ends. If component replacement, wiring repair, or bypass diode maintenance is completed after fault confirmation, the system can save the pre-repair and post-repair samples in segments and re-establish the benchmark version or aging status identifier for the post-repair string.

[0041] The state feedback relationship can be realized through state feedback records. The state feedback records include at least the string number, the generating unit to which it belongs, the inverter to which it belongs, the combiner box to which it belongs, the trigger time, the trigger reason, the fault candidate state, the recovery observation period, the maintenance event type, the baseline update state, whether the sample is allowed to enter the healthy sample set, and the recalculation state. Through the above fields, the correspondence between the string state and the baseline update can be recorded, reducing the risk of fault samples contaminating the healthy baseline and reducing the situation where repaired strings continue to be affected by abnormal trajectories before maintenance.

[0042] For example, in a ground-mounted centralized photovoltaic power station, each power generation unit includes multiple combiner boxes, and each combiner box is connected to multiple strings. The power station centralized monitoring platform, as the main body of the fault location system 100, obtains the cumulative power, current and voltage of each string over a period of time through the inverter and combiner boxes, obtains the reference irradiance through the weather station, obtains the module temperature through the module backplane temperature sensor, and reads the hierarchical relationship between power generation units, combiner boxes and strings from the power station ledger. The data access module 110 obtains complete data fields and continuous timestamps within the continuous acquisition period. The synchronization mapping module 120 aligns each data source to a unified time window and maintains the spatial mapping coefficient of each string using recent clear and stable unshaded historical periods. The operating condition collection module 130 divides the samples into comparable operating condition intervals with high irradiance and medium temperature.

[0043] The baseline update module 140 reads the health samples from the most recent few months to about one year within the operating condition range. After excluding samples that are under maintenance, power restriction, cleaning and recovery period, or in a fault candidate state, it updates the dynamic irradiance correction baseline and allowable band. The backtracking window of several months to about one year can be determined based on the natural aging rate of the components, the data acquisition cycle, the number of samples under the same operating condition, and the power plant operation and maintenance cycle. When the component separation module 150 finds that multiple strings have a common deviation that slowly decreases with the service life after correction, it first deducts the common deviation as the common aging component of the entire array. For strings that are still slowly low after deducting the common aging and whose change direction is consistent in multiple operating condition ranges, their deviation is further processed as the individual aging difference component of the string. If another string shows a sudden increase in fault residual under high irradiance and high temperature conditions starting from a certain observation window, and this continues to exceed the limits of multiple observation windows... Based on its own historical tolerance range, the fault location module 160 can mark the string as a fault candidate and suspend the participation of the string sample in the benchmark update of the corresponding operating condition range. Subsequently, if the string deviates in the same direction in multiple operating condition ranges and the continuous confirmation conditions are met, the fault location module 160 can confirm it as a fault string and locate the corresponding combiner box and string number according to the topology relationship. The confidence output module 170 combines the fault residual out-of-bounds magnitude, the number of continuous confirmations, the data quality status and the location level to output a fault work order and suggest on-site infrared verification and electrical testing. If the component replacement or wiring repair is completed after on-site verification, the system records the maintenance event and re-establishes the status label for the repaired sample. Therefore, this method does not directly judge the string power as low based on a fixed threshold, but outputs the fault location result after dynamic benchmark update, aging component deduction and continuous confirmation.

[0044] Under abnormal operating conditions, if a mountain photovoltaic power station experiences a short-term sharp fluctuation in irradiance during rapidly moving cloud formations, and the pairing relationship between the reference irradiance and the power of multiple strings is unstable, the corresponding time period can be marked as a meteorological disturbance sample. This sample does not participate in the spatial mapping coefficient update and health baseline update. If the fault residual of a string exceeds the limit during this time period, but recovers to within the allowable range during subsequent clear and stable periods, it is not directly confirmed as a hardware fault. If the same string continues to exceed the limit in subsequent clear and stable operating conditions, it enters the fault candidate confirmation process. For low power due to prolonged lack of cleaning, if the fault residual falls back to within the historical allowable range after rainfall or cleaning events, it can be reverted to a recoverable disturbance, and a dust accumulation or shading cleaning prompt will be output. If the fault residual does not recover after cleaning and continues to expand under high irradiance conditions, fault confirmation will continue. Through this processing, misjudgments caused by cloud shadows, dust accumulation, shading, and short-term measurement errors can be reduced.

[0045] The method described in this application can be executed by a device with processing capabilities. This device may include a processor, a memory, a communication interface, a data acquisition interface, and an output interface. The data stored in the memory includes at least program instructions, operating data, meteorological data, topology relationships, spatial mapping records, operating condition division rules, dynamic irradiance correction benchmarks, allowable bands, string status identifiers, anomaly control records, and historical location results, among other fields or data tables. When the processor executes the program instructions in the memory, it can perform data access, time synchronization, spatial mapping, operating condition aggregation, benchmark updating, component separation, fault location, confidence output, and anomaly handling. The communication interface can communicate with inverters, combiner boxes, weather stations, temperature sensors, string monitoring units, operation and maintenance platforms, alarm systems, or work order systems. The output interface can output fault location reports, alarm messages, review suggestions, or work order fields. The aforementioned device is used to illustrate that this method can be implemented by a specific processing device and does not require each module to be completed by a single device.

[0046] Through the above implementation methods, this application can correct string power deviations in long-term photovoltaic arrays by placing them under comparable irradiance and temperature conditions, and form a closed-loop process through health sample screening, dynamic benchmark updating, aging component separation, fault residual confirmation, and anomaly recalculation. As a result, false alarms and false negatives caused by static nameplate parameters or fixed performance ratio thresholds in long-term service scenarios can be reduced, and the fault location results can simultaneously reflect the faulty string, data quality, location reliability, and operation and maintenance review recommendations.

[0047] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0048] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fault location method for long-term photovoltaic arrays based on irradiance correction, characterized in that, include: Acquire operational data, meteorological data, and topology data of long-term photovoltaic arrays and conduct quality inspections to form a series of string-by-string operational samples; The time synchronization and spatial mapping of the sequential running samples are performed to obtain the irradiance of each sequential sample; Comparable operating condition intervals are divided according to irradiance, temperature, season or solar altitude angle. Within the same comparable operating condition interval, dynamic irradiance correction benchmarks are established or updated based on healthy samples that have passed the sample consistency screening. Based on the dynamic irradiance correction benchmark, the corrected charge deviation of each string is determined and a deviation trajectory is formed; The fault residual is obtained by subtracting the common aging component of the entire array and the individual aging difference component of the string from the deviation trajectory. Based on the starting point of the change in the fault residual, the historical tolerance range, the operating condition performance, and the continuous occurrence, the fault sequence is determined, and the fault location result is output.

2. The method for fault location of a long-term photovoltaic array based on irradiance correction according to claim 1, characterized in that, Operational data, meteorological data, and topological data, including: The operating data is acquired by the inverter, combiner box, or string monitoring unit; The operational data includes string number, string power, string current, string voltage, acquisition time, inverter, combiner box, and data quality identifier. The meteorological data is obtained from meteorological stations, total radiation meters, irradiance measuring points, or temperature measuring points. The meteorological data includes reference irradiance, component backsheet temperature, ambient temperature, collection time, and measurement point number. The topology data is obtained from the power plant ledger, monitoring platform, combiner box configuration table or string-level monitoring device mapping table; The topology data includes the generator unit number, inverter number, combiner box number, string number, and the corresponding relationship between components or segments.

3. The method for fault location of a long-term photovoltaic array based on irradiance correction according to claim 1, characterized in that, Quality inspection includes: Verify missing fields, physical range of values, continuity of timestamps, matching relationship between data sources and topology data, and self-consistency between power consumption, irradiance, and temperature; Data that fails the quality inspection are marked as abnormal samples, and these abnormal samples are excluded from the current dynamic irradiance correction benchmark update and fault confirmation.

4. The method for fault location of a long-term photovoltaic array based on irradiance correction according to claim 1, characterized in that, Time synchronization and spatial mapping include: Align multi-source runtime records to a unified time window; Data with collection periods shorter than the unified time window are merged according to the data field properties, and the values ​​within the corresponding time window are determined for data with collection periods longer than the unified time window. During historical periods of clear, stable, and unobstructed skies, a spatial mapping coefficient between the reference irradiance and the actual irradiance level of each string is established based on the string installation area, array orientation, tilt angle, reference irradiance measurement point location, and historical output records. Based on the spatial mapping coefficient, the reference irradiance is converted into the irradiance of each string in the corresponding string.

5. A fault location method for long-term photovoltaic arrays based on irradiance correction according to claim 4, characterized in that, Establish spatial mapping coefficients between reference irradiance and the actual irradiance levels of each string, including: When multiple irradiation measurement points are set up in the power plant, a reference irradiation measurement point is selected according to the principle of the area where the string is located, the junction box to which it belongs, the orientation of the component, or the closest distance, and the mapping relationship between the reference irradiation measurement point and the actual output of the string is maintained. In the event of an anomaly at the reference irradiance measurement point, switch to a nearby irradiance measurement point, or use a reference irradiance reconstructed based on historical clear-sky day data, and set a data quality limitation flag for samples using alternative irradiance data.

6. The method for fault location of a long-term photovoltaic array based on irradiance correction according to claim 1, characterized in that, Dividing comparable operating condition ranges includes: The string power after time synchronization and spatial mapping is paired with the string irradiance sequence, and divided into a comparable working condition sample set according to irradiance range, temperature range, season, solar altitude angle or combination of conditions. Read the operation and maintenance event logs and set the samples corresponding to cleaning, rainfall, or occlusion clearing as the recovery observation samples; After component replacement, wiring maintenance, or bypass diode maintenance, the historical baseline version or aging status identifier of the corresponding string should be re-established or saved in segments.

7. The method for fault location of a long-term photovoltaic array based on irradiance correction according to claim 1, characterized in that, Within the same comparable operating condition range, samples are screened for consistency, including: Verify the data quality, string status, historical deviation status, and operation and maintenance event status of samples within the same comparable operating condition range. A healthy sample is one that simultaneously meets the following criteria: power, irradiance, temperature, timestamp, and topology relationship. The sample must not be in a fault candidate state, fault confirmation state, or data restriction state. The historical deviation of the sample must not exceed the historical tolerance range. Furthermore, the sample must not be in a period of maintenance, power restriction, communication abnormality, cleaning recovery period, or shading cleaning recovery period. Samples that do not meet the above verification criteria are excluded from the dynamic irradiance correction benchmark update process.

8. The method for fault location of a long-term photovoltaic array based on irradiance correction according to claim 1, characterized in that, Establish or update dynamic irradiance correction benchmarks, including: Based on the unit irradiance level of the healthy sample set within the retrospective window, a dynamic irradiance correction benchmark is formed for the corresponding comparable operating condition range. The historical allowable range is determined based on the normal residual fluctuation range under the same historical operating conditions, sensor accuracy, inverter or combiner box metering accuracy, acquisition cycle and on-site calibration results. If a historical healthy sample is identified as a faulty or abnormal sample, the historical healthy sample is removed from the healthy sample set, and a baseline version and deviation trajectory for the affected time period are regenerated.

9. A fault location method for long-term photovoltaic arrays based on irradiance correction according to claim 1, characterized in that, The formation of the deviation trajectory includes: Within each comparable operating condition range, the measured power, irradiance of each string, spatial mapping coefficient, and dynamic irradiance correction benchmark of each string are read, and the power of each string is converted to the same comparison caliber. The converted string power is compared with the healthy power level in the comparable operating range to obtain the corrected power deviation. Arrange the corrected power deviations of the same series of strings within the same or similar comparable operating conditions in chronological order to form a deviation trajectory.

10. A fault location method for long-term photovoltaic arrays based on irradiance correction according to claim 1, characterized in that, After deducting the common aging component of the entire array and the individual aging difference component of each string, the following is included: Read the corrected power deviation of multiple strings in the same power generation unit, the same combiner box, or the same comparable area, and determine the overall downward shift component that is common to multiple healthy strings, changes slowly over time, and is consistent in direction in different comparable operating conditions as the common aging component of the entire array. For the remaining deviation after deducting the common aging component of the entire array, the deviation that changes slowly over a long period of time relative to the common level of the entire array, has the same direction, and has a similar direction of change in multiple comparable operating condition intervals is determined as the individual aging difference component of the string. The remaining deviation after deducting the common aging component of the entire array and the individual aging difference component of the string is determined as the fault residual.

11. A fault location method for long-term photovoltaic arrays based on irradiance correction according to claim 1, characterized in that, The fault sequence is determined based on the starting point of the fault residual change, the historical tolerance band, the operating condition performance, and the frequency of occurrence, including: When the fault residual in the same string changes from the stable range to a continuous expansion, or exceeds the historical allowable range and fails to recover to the normal range in subsequent observation windows, or expands relative to the historical same operating conditions under high irradiance, high current, high temperature, local shading sensitive periods or specific solar altitude angles, the string is marked as a fault candidate string. After the string is marked as a fault candidate string, new samples of the string within the corresponding comparable operating condition range are excluded from the dynamic irradiance correction benchmark update. The candidate fault strings are continuously confirmed through multiple observation windows and multiple comparable operating condition intervals, and the fault range is determined based on the topological relationship between the generator unit, inverter, combiner box, string and component.

12. The method for fault location of a long-term photovoltaic array based on irradiance correction according to claim 11, characterized in that, The fault range is determined based on the topological relationships between the generator unit, inverter, combiner box, string, and module, including: With the availability of string segment monitoring, string-level optimizer, current and voltage curves, or bypass diode conduction information, read the current and voltage curves of the faulty string during a clear and stable period or during on-site testing, and compare them with historical curves under the same operating conditions or curves of adjacent healthy strings. If there are steps, peaks, valleys, or slope changes in the current-voltage curve that correspond to the conduction of the bypass diode, local mismatch of the component, or local hot spots, the corresponding change position is matched with the mapping relationship of the component or bypass diode section in the string to determine the component section or bypass section.

13. The method for fault location of a long-term photovoltaic array based on irradiance correction according to claim 1, characterized in that, Output the fault location results, including: The location reliability is generated based on the fault location results, data quality identifiers, deducted common aging components, individual aging differences in the string, fault residual out-of-bounds magnitude, consistency under multiple operating conditions, number of continuous confirmations, location level, and abnormal degradation status. Output the fault string number, the generator unit to which it belongs, the inverter to which it belongs, the combiner box to which it belongs, the fault candidate start time, the fault confirmation time, the operating condition range involved, the residual out-of-bounds range, the starting point of the residual change, the number of continuous windows, the deviation performance under different operating conditions, the data quality identifier, the confidence level, the fault nature prompt, the suggested review method, and the suggested handling priority.

Images