Distributed photovoltaic fault positioning and diagnosis method based on multi-terminal data fusion

Abstract

The invention discloses a distributed photovoltaic fault positioning and diagnosis method based on multi-terminal data fusion, which belongs to the technical field of photovoltaic power generation system monitoring and comprises the following steps: acquiring real-time operating parameters of an inverter, acquiring actually measured power and calculating a theoretical power interval; judging whether the actually measured power exceeds a theoretical power interval or not, and generating a collaborative sampling instruction; collecting combiner box branch current and environment monitoring data in response to the instruction; fusing multi-end data to generate a multi-dimensional associated feature set; performing feature comparison type primary diagnosis to obtain a primary fault area identifier, then starting secondary diagnosis to extract a dynamic deviation rate and environmental response of a target area branch, and generating a positioning diagnosis feature set; and correcting and outputting accurate fault coordinates and fault type identifiers by combining the consistency characteristics of the combiner box and the power flow traceability characteristics of the multi-dimensional associated characteristics. According to the method, cooperative triggering, feature hierarchical diagnosis and closed-loop optimization strategies are adopted, and fault coordinates can be accurately positioned and fault types can be identified in a complex operation environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of photovoltaic power generation system monitoring technology, and particularly to a distributed photovoltaic fault positioning and diagnosis method. BACKGROUND

[0002] A distributed photovoltaic system collects electric energy through a plurality of photovoltaic components arranged in a distributed manner via a current collection box and finally merges into a power grid through an inverter. Since such a system is small in scale and widely distributed, its fault monitoring mainly relies on inverter operating parameters or a small amount of sensor data.

[0003] The prior art usually adopts single data source threshold alarm or fixed frequency multi-end data superposition analysis. Some schemes trigger manual inspection after judging the abnormal power of the inverter through a rule model. A few schemes introduce current collection box data, but only perform simple branch power comparison. Environmental data is usually processed independently as an auxiliary reference.

[0004] The above method has the following technical defects: the fault positioning stays at the string level and cannot accurately locate the fault point; the recognition rate of component attenuation or transient connection failure with gentle power fluctuation is low; full-quantity high-frequency sampling leads to data redundancy and resource waste; and environmental mutations easily cause false alarms and interfere with operation and maintenance decisions. SUMMARY

[0005] To solve the above problems, the present application provides a distributed photovoltaic fault positioning and diagnosis method based on multi-end data fusion, which adopts a cooperative triggering, feature hierarchical diagnosis and closed-loop optimization strategy, and can accurately locate the fault coordinates and identify the fault type under complex operating environment.

[0006] The above object can be achieved by the following scheme:

[0007] A distributed photovoltaic fault positioning and diagnosis method based on multi-end data fusion, comprising collecting real-time operating parameters of an inverter, obtaining measured power, and calculating a theoretical power interval; judging whether the measured power exceeds the theoretical power interval to generate a cooperative sampling instruction; in response to the cooperative sampling instruction, collecting current collection box branch current data and environmental monitoring data to generate a multi-end synchronous sampling data set; fusing the real-time operating parameters of the inverter, the current collection box branch current data and the environmental monitoring data to generate a multi-dimensional correlation feature set; performing feature comparison type primary diagnosis based on the multi-dimensional correlation feature set to obtain a preliminary fault region identifier, and starting secondary diagnosis based on the preliminary fault region identifier to extract the dynamic deviation rate of the target region branch and the environmental response abnormality degree to generate a positioning diagnosis feature set; combining the cross-current collection box consistency features and power flow tracing features of the multi-dimensional correlation feature set to correct and output accurate fault coordinates and fault type identifier.

[0008] Optionally, the generation of the collaborative sampling instruction includes: collecting real-time operating parameters of the inverter and generating historical operating data; establishing a correlation between environmental parameters and theoretical power based on the historical operating data and determining the theoretical power value corresponding to the environmental parameters; calculating the theoretical power range based on the theoretical power value; and generating a collaborative sampling instruction when the measured power continuously exceeds the theoretical power range.

[0009] Optionally, the collected environmental monitoring data includes: acquiring local irradiance data at the string level through spatially distributed illumination sensors; the number of spatially distributed illumination sensors is spatially matched according to the physical distribution of the combiner box branches; and acquiring local temperature data of the same area through temperature sensors mounted on the combiner box body.

[0010] Optionally, generating the multi-dimensional correlation feature set includes: calculating the real-time power of each branch and comparing it with the historical operating data to generate branch power contribution deviation features; combining the environmental monitoring data and the real-time power of each branch to analyze the time-series response characteristics of different branches under the same environmental conditions to generate branch-environment coupling response anomaly; comparing the branch power output of adjacent combiner boxes under the same environmental conditions to extract cross-combiner box consistency features; calculating the real-time difference and rate of change between the inverter input power and the sum of the output power of the associated combiner boxes to generate power flow tracing features; and comprehensively extracting the branch power contribution deviation features, the branch-environment coupling response anomaly, the cross-combiner box consistency features, and the power flow tracing features to finally generate the multi-dimensional correlation feature set.

[0011] Optionally, the initiation of secondary diagnosis based on the preliminary fault area identifier includes: selecting a predetermined number of branches with the highest dynamic power deviation rate from the preliminary fault area identifier to generate a set of suspicious branches; filtering the branch-environment coupling response anomalies of the set of suspicious branches to verify the physical correlation between power changes and real-time environmental parameters; using the cross-combiner box consistency feature to eliminate global environmental fluctuation interference and confirm the existence of local specific faults; determining the fault level based on the power flow tracing feature and outputting a component-level fault type identifier or a combiner box internal fault type identifier.

[0012] Optionally, the physical correlation between the power change and real-time environmental parameters includes: when a sudden increase in irradiance is detected, determining that the power of the suspected branch collection does not rise synchronously or decreases in the opposite direction as an abnormal response; when a local abnormal increase in temperature is detected and accompanied by a persistently low power, associating it with a hot spot fault type identifier.

[0013] Optionally, the step of correcting and outputting accurate fault coordinates and fault type identifiers includes: generating a fault type identifier inside the combiner box when the output power of the combiner box is continuously lower than the total power of its subordinate branches and the rate of change of the power difference is greater than a predetermined mutation ratio; and outputting a component-level fault type identifier when the deviation characteristics of the branch power contribution are continuously deviated from the threshold and the power flow difference is in a stable state.

[0014] Optionally, the method further includes: generating a sampling strategy optimization instruction based on the precise fault coordinates and the fault type identifier; dynamically adjusting the regular sampling frequency of the devices within the preliminary fault area identifier, increasing the sampling density for implicit attenuation branches, and decreasing the sampling density for fault-free areas; and optimizing the trigger sensitivity parameter of the collaborative sampling instruction based on historical diagnostic results.

[0015] Optionally, the dynamic adjustment of the regular sampling frequency of the equipment within the initial fault area includes: for occasional poor contact faults, automatically activating high-frequency sampling signals during periods of strong winds indicated by meteorological warnings or under conditions of rapid environmental changes; and extending the regular sampling interval for combiner box branches that have no fault records within a continuous preset period.

[0016] Based on the same inventive concept, this invention also provides a distributed photovoltaic fault location and diagnosis system based on multi-terminal data fusion. The system includes: an interval calculation module for collecting real-time operating parameters of the inverter, obtaining measured power, and calculating a theoretical power range; a collaborative triggering module for determining whether the measured power exceeds the theoretical power range and generating a collaborative sampling command; a multi-terminal acquisition module for responding to the collaborative sampling command, collecting combiner box branch current data and environmental monitoring data, and generating a multi-terminal synchronous sampling dataset; a feature fusion module for fusing the real-time operating parameters of the inverter, the combiner box branch current data, and the environmental monitoring data to generate a multi-dimensional associated feature set; a hierarchical diagnosis module for performing feature comparison-based primary diagnosis based on the multi-dimensional associated feature set to obtain a preliminary fault area identifier, and initiating secondary diagnosis with the preliminary fault area identifier to extract the dynamic deviation rate and environmental response anomaly degree of the target area branch, generating a location diagnosis feature set; and a closed-loop optimization module for combining the cross-combiner box consistency feature and power flow tracing feature of the multi-dimensional associated feature set to correct and output accurate fault coordinates and fault type identifiers.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] Through multi-terminal data fusion and hierarchical diagnosis mechanism, this invention can effectively distinguish between environmental interference and real faults, achieve accurate positioning from the component level to the combiner box level, and significantly reduce the false alarm rate and the missed detection rate.

[0019] Based on an adaptive sampling strategy that considers fault history and real-time environmental risks, this method significantly reduces invalid data collection in stable regions while increasing monitoring density in high-fault-risk regions, thus balancing computational load and diagnostic timeliness.

[0020] By combining seasonal attenuation coefficients with local microclimate monitoring, the diagnostic threshold is dynamically adjusted to overcome the impact of environmental fluctuations on the theoretical power model, ensuring the stability and reliability of fault identification. Furthermore, the diagnostic results directly drive the iterative updates of sampling strategies and model parameters, thereby continuously improving the system's sensitivity to latent faults and providing continuous optimization technical support for operation and maintenance decisions.

[0021] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a distributed photovoltaic fault location and diagnosis method based on multi-terminal data fusion according to an embodiment of the present invention.

[0024] Figure 2 This is a multi-dimensional feature time-series curve diagram of an embodiment of the present invention.

[0025] Figure 3 This is a scatter plot of the branch physical response anomaly criteria in an embodiment of the present invention.

[0026] Figure 4 This is a multi-feature radar map of suspicious branches according to an embodiment of the present invention.

[0027] Figure 5 This is a schematic diagram of the structure of a distributed photovoltaic fault location and diagnosis method system based on multi-terminal data fusion according to an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0029] Reference Figure 1 One embodiment of the present invention proposes a method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion. It adopts a collaborative triggering, feature-based hierarchical diagnosis and closed-loop optimization strategy, which can accurately locate fault coordinates and identify fault types in complex operating environments.

[0030] The method described in this embodiment specifically includes:

[0031] Collect real-time operating parameters of the inverter, obtain the measured power, and calculate the theoretical power range;

[0032] Specifically, real-time operating parameters of the inverter, such as active power, reactive power, voltage, current, ambient temperature, and irradiance, are collected at fixed intervals and stored for subsequent analysis. By analyzing historical operating data and combining it with current ambient temperature and irradiance, the theoretical power range that the inverter should achieve under current environmental conditions is determined.

[0033] Determine whether the measured power exceeds the theoretical power range, and generate a collaborative sampling command; Specifically, the measured power collected in real time is compared with the calculated theoretical power range. If the measured power consistently exceeds the theoretical range, it indicates that the inverter may have an abnormality or fault. In this case, the system automatically generates a collaborative sampling command to further collect and analyze relevant data to assist in fault location and diagnosis.

[0034] In response to the collaborative sampling command, the current data of the combiner box branch and environmental monitoring data are collected to generate a multi-terminal synchronous sampling dataset;

[0035] Specifically, upon responding to the coordinated sampling command, three types of data acquisition are performed simultaneously: acquiring the real-time input power and DC voltage parameters of the inverter; measuring the current values ​​of each branch of the combiner box and converting them into real-time power; and acquiring spatially distributed light sensor data and combiner box body temperature. All data are appended with precise timestamps to ensure consistency in spatiotemporal dimensions.

[0036] By integrating the real-time operating parameters of the inverter, the branch current data of the combiner box, and the environmental monitoring data, a multi-dimensional correlation feature set is generated. Specifically, a multi-dimensional correlation feature set is constructed by integrating three types of data sources. First, the deviation between the real-time power of each branch and the historical average value is calculated. Then, the physical correlation between the branch power change rate and the environmental parameter change rate is analyzed. At the same time, the output power difference value of adjacent combiner boxes under the same environmental conditions is extracted, and the difference between the inverter input power and the total power of the associated combiner boxes and its changing trend are monitored.

[0037] Based on the multi-dimensional associated feature set, a feature comparison-based primary diagnosis is performed to obtain a preliminary fault area identifier. Then, based on the preliminary fault area identifier, a secondary diagnosis is initiated to extract the dynamic deviation rate and environmental response anomaly degree of the target area branch and generate a location diagnosis feature set.

[0038] Specifically, the feature comparison-based primary diagnosis includes: calculating the "branch power contribution deviation feature" of each branch in the multi-dimensional correlation feature set, and setting a preliminary diagnosis threshold (e.g., twice the standard deviation of historical data); identifying all branches exceeding this threshold as preliminary fault areas. Alternatively, sorting the feature values ​​of all branches and selecting the N branches with the highest ranking as preliminary fault areas. The secondary diagnosis stage performs in-depth verification on suspicious areas: verifying whether power changes conform to the physical laws of environmental parameter changes; using cross-regional consistency features to eliminate global environmental fluctuation interference; and determining the fault level attribute by combining the stability of power flow difference.

[0039] By combining the cross-junction box consistency features and power flow tracing features of the multi-dimensional associated feature set, accurate fault coordinates and fault type identifiers are corrected and output.

[0040] Specifically, when the output power of the combiner box is continuously lower than the sum of the power of its subordinate branches and the difference changes drastically, it is determined to be an internal fault of the combiner box; when the branch power is continuously abnormal but the power flow difference remains stable, it is determined to be a component-level fault, and finally outputs a fault type identifier with physical location coordinates.

[0041] By fusing data from multiple sources, the method effectively overcomes misjudgments caused by environmental interference, achieving precise positioning from the component level to the combiner box level. This significantly improves the accuracy and efficiency of distributed photovoltaic fault diagnosis. The hierarchical diagnosis mechanism greatly reduces the computational load, while the adaptive optimization module continuously improves the system's response sensitivity. The closed-loop design enables diagnostic results to directly drive the optimization of the sampling strategy, ensuring fault detection capabilities while achieving dynamic allocation of system resources, providing highly reliable support for operation and maintenance decisions.

[0042] Optionally, the generated co-sampling instruction includes: Collect real-time operating parameters of the inverter and generate historical operating data;

[0043] Specifically, real-time operating parameters of the inverter, such as active power, reactive power, voltage, current, ambient temperature, and irradiance, are collected at fixed intervals. These parameters and their corresponding timestamps are then stored in a historical database in a time-series format to generate historical operating data. The historical operating data includes measured power data.

[0044] Based on the real-time operating parameters, establish the correlation between environmental parameters and theoretical power, and determine the theoretical power value corresponding to the environmental parameters; Specifically, a correlation is established between environmental parameters and theoretical power. This correlation is obtained by training with historical operating data, including different irradiance levels. (Unit: W / m²) and temperature Theoretical power value under (unit: °C) combination (Unit: kW); A piecewise linear regression model is used during training, and the formula is: , where the coefficient , , It was obtained by fitting historical data using the least squares method.

[0045] The theoretical power range is calculated based on the theoretical power value. When the measured power continuously exceeds the proportion of the theoretical power range, a collaborative sampling command is generated.

[0046] Specifically, based on real-time environmental parameters (irradiance) ,temperature Query the correlation and calculate the theoretical power range. , To allow for a fluctuation range (default value 0.15), interpolation via table lookup is used. A seasonal attenuation coefficient is introduced. (Based on historical power generation data for the same month at the power station, values ​​range from 1.0 to 0.9, with higher values ​​used in winter and lower values ​​used in summer), dynamically adjusting the boundary values ​​of the theoretical power range. , Updated monthly based on the component aging model. The component aging model can be adopted... The model, or an industry-recognized linear degradation model, is updated monthly based on the module's operational years or cumulative operating time to reflect the trend of module performance decline over time. When the measured power... Continuously exceeding the revised theoretical power range , When the preset duration (e.g., 3 minutes) is reached, the specific trigger condition is determined to be met. (Power abnormally low) or If the state of (abnormally high power) lasts for 3 minutes, a collaborative sampling command will be generated immediately.

[0047] For example, when a photovoltaic power station has an irradiance of 800W / m² and a temperature of 35℃, a query for related relationships yields... Calculate the theoretical interval The actual measured inverter power was 56kW for 5 minutes, and the 56kW power remained. If the triggering condition is not met within the specified interval, no instruction will be generated. Summer Correction The interval is then updated to The 56kW output exceeds the corrected range and meets the generation conditions. By dynamically adjusting the power reference using a seasonal coefficient, the theoretical power overestimation caused by high summer temperatures is avoided, accurately capturing abnormal component performance, significantly reducing the false alarm rate caused by changes in environmental factors, and ensuring sensitivity to latent faults.

[0048] Optionally, the collected environmental monitoring data includes:

[0049] Local irradiance data at the string level are acquired using spatially distributed illumination sensors; Specifically, spatially distributed light sensors are deployed at the photovoltaic string level, with each sensor having a measurement radius of... ( Local irradiance within 2 meters The data is transmitted to the central processing unit via a wireless sensor network.

[0050] The number of spatially distributed illumination sensors is spatially matched based on the physical distribution of the combiner box branches. Specifically, the number of sensors deployed Satisfy the formula ,in K represents the number of branches under the junction box, and K is the spatial distribution density coefficient (value 3-5). This represents the rounding up function, ensuring that each branch string is covered by a sensor.

[0051] Local temperature data for the same area is obtained by a temperature sensor mounted on the main body of the junction box.

[0052] Specifically, a temperature sensor is installed on the surface of the junction box's outer casing to directly measure the local temperature. The installation location should be chosen around the area of ​​the ventilation holes. The final generated environmental monitoring dataset contains the irradiance vectors corresponding to each branch. and junction box area temperature value .

[0053] For example, a junction box connects 8 branch circuits, according to... Calculation requires layout One light sensor is installed at the center point of strings 1-4 in branch circuits and at the center point of strings 5-8 in branch circuits, with a measurement radius covering all strings. A temperature sensor is mounted on the heat dissipation grille on the side wall of the combiner box to measure... By deploying spatial matching, branch-level irradiance differences can be accurately captured, avoiding environmental data distortion caused by monitoring large areas with a single sensor. This provides a true and reliable environmental parameter benchmark for subsequent fault diagnosis, effectively identifying latent faults caused by local shading or heat island effects.

[0054] Optionally, generating a multi-dimensional association feature set includes: Calculate the real-time power of each branch and compare it with the historical operating data to generate branch power contribution deviation characteristics; Specifically, firstly, based on the collected branch current data from the combiner box, the real-time power of each branch is calculated in conjunction with the system standard voltage value, which is obtained by measuring the inverter input voltage. The real-time power of each branch is then compared with the corresponding historical average power in the historical operating data. The historical operating data stores historical statistical values ​​under the same environmental conditions. The branch power contribution deviation characteristic is generated by calculating the deviation percentage. The deviation percentage formula is as follows: ,in Indicates a branch Real-time power, , For branch current, For the system standard voltage , Extracted from real-time inverter parameters; Indicates a branch The historical average power was obtained by querying the historical operating database.

[0055] By combining the environmental monitoring data with the real-time power of each branch, the time-series response characteristics of different branches under the same environmental conditions are analyzed, and the branch-environment coupling response anomaly degree is generated. Specifically, by combining environmental monitoring data, including irradiance and temperature, the temporal response characteristics of different branches under the same environmental conditions are analyzed. Irradiance represents light intensity, and temperature represents ambient heat, both collected from distributed sensors. The ratio of power change rate to irradiance change rate is calculated and compared with the expected response model to generate the branch-environment coupling response anomaly degree. The anomaly degree formula is as follows: ,in Indicates a branch The rate of change of power is calculated from time-series power data; This represents the rate of change in irradiance, calculated from a time series of environmental monitoring data. This represents the expected response coefficient, obtained through training on historical irradiance-power correlations. These correlations establish a linear relationship between irradiance and theoretical power based on historical data. For example... Figure 2 As shown, this method achieves comprehensive perception and anomaly detection of branch operation status through time-series fusion analysis of multi-dimensional features such as real-time power, historical average power, theoretical power, irradiance and temperature, providing a rich feature basis for subsequent fault diagnosis.

[0056] By comparing the branch power output of adjacent combiner boxes under the same environmental conditions, cross-combiner box consistency characteristics are extracted.

[0057] Specifically, by comparing the branch power output of adjacent combiner boxes under the same environmental conditions, cross-combiner box consistency characteristics are extracted. This is achieved by calculating the normalized difference in average power between adjacent combiner boxes, using the following characteristic formula: ,in Indicates junction box The average power is calculated by summing the power of the subordinate branches; Indicates adjacent combiner boxes The average power is calculated using the same method.

[0058] Calculate the real-time difference and rate of change between the inverter input power and the sum of the output power of the associated combiner box, and generate power flow tracing characteristics; Specifically, the real-time difference and rate of change between the inverter input power and the sum of the output power of the associated combiner box are calculated to generate power flow tracing characteristics. The difference formula is as follows: ,in

[0059] This represents the inverter input power, which is collected from the inverter's real-time operating parameters. This represents the total output power of the associated combiner boxes, obtained by summing the power of each combiner box branch; the rate of change formula is... It is calculated using the time derivative.

[0060] The branch power contribution deviation characteristics, the branch-environment coupling response anomaly, the cross-combiner box consistency characteristics, and the power flow tracing characteristics are comprehensively extracted to finally generate a multi-dimensional associated feature set.

[0061] Specifically, four types of feature data streams are aligned under a unified time reference, and a timestamp synchronization mechanism is established based on the collaborative sampling trigger time. Secondly, a feature fusion matrix is ​​constructed, and power contribution deviation features are organized according to the branch dimension. Coupling response anomaly degree Organize cross-junction box consistency features according to junction box dimension Organize power flow source tracing characteristics according to system hierarchy and rate of change Finally, each feature dimension is multiplied by its corresponding weight to generate a multi-dimensional related feature vector. The weighting coefficients are dynamically adjusted based on feature confidence: they are increased when environmental stability is high. Weights, enhanced during environmental abrupt changes Weighting, with emphasis during global fluctuations .

[0062] For example, in a distributed photovoltaic power station, the system detects inverter power fluctuations and triggers collaborative sampling; data is collected from the combiner box. and Branch current and environmental irradiance data; calculated real-time power of branch 1 as 850W, historical average power as 900W, and generated deviation characteristics. Combined with a sudden 10% increase in irradiance, the power of branch 1 only increased by 2%, while the expected response coefficient... The anomaly score is 0.8. Compare adjacent junction boxes and The average power outputs are 5kW and 5.2kW respectively, exhibiting consistent characteristics. Simultaneously, the inverter input power is calculated to be 10kW, and the total combiner box power is 10.5kW. The difference is... rate of change Subsequently, the system aligns the aforementioned features under a unified time reference to construct a multi-dimensional associated feature set: using the co-sampling time t0 as the reference, all feature time series are aligned, and then a feature weighted fusion algorithm is used to generate multi-dimensional associated feature vectors. .

[0063] After generating a multi-dimensional correlation feature set, the primary diagnosis identifies branch 1 as a suspicious area, and the secondary diagnosis confirms a local fault. Through multi-dimensional feature cross-validation, environmental interference and actual faults are effectively distinguished, avoiding misjudging changes in irradiance as component problems. At the same time, poor contact faults are quickly located, improving system maintenance efficiency and reducing downtime losses.

[0064] Optionally, initiating secondary diagnostics based on the initial fault area identifier includes: From the initial fault area identifier, a predetermined number of branches with the highest dynamic power deviation rate are selected to generate a set of suspicious branches; Specifically, the dynamic power deviation rate of all branches is extracted from the generated preliminary fault area identifiers. Sort in descending order and select the first few. These branches constitute a suspicious branch set. The value is configured to be 3-5 depending on the system size.

[0065] The branch-environment coupling response anomalies of the suspected branch set are filtered to examine the physical correlation between power changes and real-time environmental parameters; Specifically, the branch-environment coupling response anomaly degree is calculated for each branch within the suspected branch set. When a sudden change in environmental parameters is detected, a physical correlation test is performed: if the real-time irradiance suddenly increases... Then the power change rate of the suspected branch is required. , The historical minimum response coefficient is set to 0.65; if this condition is not met, it is marked as an abnormal response. If an abnormal increase in local temperature is detected... If the power consistently falls below 15% of the theoretical value, a hot spot fault flag will be displayed. Figure 3 As shown, the abnormal branch and hot spot fault identification method based on the dual verification mechanism of physical laws effectively distinguishes between environmental noise and real equipment faults.

[0066] By leveraging the cross-junction box consistency characteristics, global environmental fluctuations can be eliminated, confirming the existence of localized specific faults. Specifically, the generated cross-junction box consistency feature is invoked. When the whole An average value greater than 0.1 indicates the presence of environmental fluctuation interference. In this case, if the junction box of the suspected branch... If the value is less than 50% of the global mean, it is confirmed as a localized specific fault.

[0067] Based on the power flow tracing characteristics, the fault level is determined, and a component-level fault type identifier or a combiner box internal fault type identifier is output.

[0068] Specifically, based on the power flow source characteristics The value and rate of change :when and Output junction box internal fault indicator when kW / s; when and Output component-level fault indicators at kW / s. For example... Figure 4 As shown, through a multi-level feature cross-validation mechanism, multi-dimensional indicators are calculated respectively. The radar chart intuitively displays the performance of each suspicious branch in different feature dimensions, which facilitates comprehensive judgment and accurate location of abnormal branches.

[0069] For example, a primary diagnostic identification area for a photovoltaic system It contains 12 branches, and the 3 branches with the highest dynamic power deviation rate are selected. This constitutes a suspicious branch set. The examination revealed that when the irradiance suddenly increased by 15%, the branch... Power increased by only 5%. Exceeding the threshold of 0.6; cross-junction box consistency characteristics show a global mean of 0.12, but branch... The C-value of the junction box is only 0.05, eliminating environmental interference; power flow tracing characteristics. and When the power flow rate reaches kW / s and meets the component-level fault conditions, the system outputs a component fault identifier. Through a dual filtering mechanism of environmental response verification and cross-regional consistency comparison, local component degradation faults masked by global irradiance fluctuations are effectively identified, avoiding misjudgments caused by rapid cloud movement. At the same time, power flow stability verification ensures that fault location is not affected by instantaneous measurement noise, significantly improving the reliability of diagnostic results.

[0070] Optionally, the physical correlation between the power change and real-time environmental parameters includes: When a sudden increase in irradiance is detected, the abnormal response is determined to be that the power of the suspected branch does not increase synchronously or decreases in the opposite direction. Specifically, the operational steps for verifying the physical correlation between power changes and real-time environmental parameters are as follows: First, continuously monitor the irradiance sequence in the environmental monitoring data. and temperature sequence Data is collected in real time through distributed light and temperature sensors. When the system detects a sudden increase in irradiance, the determination condition is... ( (Take 100 W / m²·min), and simultaneously obtain the real-time power change rate of each branch within the suspected branch set. If a branch satisfies (Power decrease) or ( If the minimum response coefficient is 0.65 (obtained through historical data statistics), then the branch is marked as an abnormal response.

[0071] When a localized abnormal increase in temperature is detected, accompanied by a persistently low power, an associated hot spot fault type identifier is generated.

[0072] Specifically, when the temperature sensor detects a local temperature difference... ( (The average temperature of adjacent areas) and the power of the corresponding branch in that area continuously meets the requirements. ( (Obtained by querying the irradiance-temperature-power correlation from current environmental parameters) Exceeding the time window ( When the time is 3 minutes, the hot spot fault type identifier will be automatically associated.

[0073] For example, a photovoltaic power station detected a sudden increase in irradiance from 500W / m² to 850W / m² at 10:15. Suspicious side road The power change rate was only 0.2 kW / min, which is below the threshold. kW / min, system flags abnormal response. Simultaneously, temperature monitoring shows string... The local temperature reached 65℃, the average temperature of the adjacent area was 53℃, and the corresponding branch power of 4.2kW ​​was consistently lower than the theoretical value of 5.1kW, triggering a hot spot fault identification in the system. Through a dual verification mechanism based on physical laws, the system effectively identifies component failures under sudden irradiation scenarios, accurately distinguishes between genuine hot spot faults and ordinary temperature fluctuations, significantly reducing the false alarm rate caused by sensor drift and providing maintenance personnel with clear fault handling guidelines.

[0074] Optionally, the step of correcting and outputting accurate fault coordinates and fault type identifiers includes:

[0075] When the output power of the combiner box is detected to be continuously lower than the total power of its subordinate branches, and the rate of change of the power difference is greater than the predetermined mutation ratio, an internal fault type identifier of the combiner box is generated.

[0076] Specifically, monitor the output power of the combiner box in real time. This data is collected through the built-in meter in the combiner box, and the total power of all branches under the combiner box is calculated simultaneously. The power of a single branch , , The measured value is from the branch current sensor. The system standard voltage is used; calculate the power difference. A fault type identifier for the combiner box is generated when two conditions are met simultaneously: Condition one is... > 0 continues beyond the time window ( (Take 5 minutes), achieved through time-series data continuity detection; Condition 2 is the rate of change of power difference. ( The predetermined mutation ratio (0.2 kW / min) was obtained through differential calculation.

[0077] When the branch power contribution deviation characteristic is detected to continuously deviate from the threshold and the power flow difference is in a stable state, a component-level fault type identifier is output.

[0078] Specifically, when the generated branch power contribution deviation characteristic is detected... Continuously exceeding the threshold ( (Take three times the historical statistical standard deviation), exceeding the time window. ( Take 10 minutes), and the inverter-combiner box power difference in the power flow tracing characteristics. ( Indicates the inverter input power. This represents the total output power of the associated combiner box. for rate of change, When the stability threshold is set to 0.05 kW / min, output the component-level fault type identifier.

[0079] For example, a photovoltaic power station monitored the output power of combiner box CB-03. The total power of its five subordinate branches power difference Lasting for 6 minutes; simultaneously, the rate of change of power difference If the power output exceeds a preset threshold of 0.2 kW / min, the system generates an internal fault type identifier for the combiner box. In another scenario, the power contribution deviation characteristics of branch B7 are... Lasting 12 minutes (threshold) ), while the inverter input power Total power of associated combiner box Difference And the rate of change When the kW / min is less than the stable threshold, the system outputs a component-level fault type identifier. Through a dual verification mechanism of power flow dynamic characteristics and time persistence, it effectively distinguishes between combiner box terminal loosening faults and component aging faults, avoids misjudging the heat loss of connectors as component power attenuation, and eliminates the interference of temporary cloud cover on diagnostic results, significantly improving the accuracy of fault handling by maintenance personnel.

[0080] Optionally, the method further includes:

[0081] Based on the precise fault coordinates and the fault type identifier, a sampling strategy optimization instruction is generated;

[0082] Specifically, based on the precise fault coordinates and fault type identifiers, branches that repeatedly experience component-level faults are marked as implicit attenuation branches, and sampling strategy optimization instructions are generated.

[0083] The regular sampling frequency of the equipment within the initial fault area is dynamically adjusted, the sampling density is increased for implicit attenuation branches, and the sampling density is decreased for fault-free areas.

[0084] Specifically, the regular sampling frequency of equipment within the initial fault area is dynamically adjusted, and the sampling interval for implicit attenuation branches is increased from... (Default 5 minutes) shortened to (1 minute), the sampling density is increased by modifying the configuration parameters of the data acquisition unit; for combiner box branches with no fault records for M consecutive diagnostic cycles (M = 30 days), the sampling interval is extended to... (10 minutes).

[0085] The trigger sensitivity parameters of the collaborative sampling command are optimized based on historical diagnostic results.

[0086] Specifically, the failure rate is calculated based on the historical diagnostic results database. and false alarm rate ,when Adjust the power fluctuation trigger threshold proportionally. The formula for adjustment is: Where k is the adjustment coefficient of 0.2; when Increased proportionally .

[0087] For example, a photovoltaic system diagnosed branch L09 with three module-level faults, marking it as a branch with latent degradation, and increasing the sampling frequency from 5 minutes / time to 1 minute / time; simultaneously, the branches under combiner box CB-12 had been fault-free for 45 consecutive days, and the sampling interval was extended to 10 minutes / time. Historical diagnostic analysis showed a false negative rate of 6% last month, and the original trigger threshold... Downgraded to By dynamically allocating resources, the probability of missed detections in high-fault-risk areas is significantly reduced, while invalid data collection in stable areas is reduced. This results in a decrease in the overall system's computational load and an improvement in the timeliness of fault detection, forming a self-optimizing diagnostic closed loop.

[0088] Optionally, the dynamic adjustment of the regular sampling frequency of the devices within the initial fault area identification includes:

[0089] In response to occasional poor contact faults, the high-frequency sampling signal is automatically activated during periods of strong winds indicated by weather warnings or under conditions of rapid environmental change.

[0090] Specifically, the steps for dynamically adjusting the regular sampling frequency of equipment within the initial fault area are as follows: First, establish a meteorological early warning interface to receive strong wind warning signals. When the wind speed forecast value... ( (take 8 m / s) or detect it using a real-time wind speed sensor At that time, the high-frequency sampling signal is automatically activated, and the sampling interval of the affected area is changed from the normal... Compress to ( , This instruction is implemented by writing to the configuration register of the data acquisition unit.

[0091] For combiner box branches that have no fault records within a continuous preset period, extend their regular sampling interval.

[0092] Specifically, for continuous The junction box branch with no fault records in each diagnostic cycle ( (Using a preset value of 30 days), calculate its power stability index. ( For the standard deviation of branch power, (These are power averages, calculated based on historical 30-day data). ( When the value is 0.98, the sampling interval extension mechanism is triggered, and the sampling interval is set to... .

[0093] For example, after a meteorological observatory issues a yellow gale warning, a photovoltaic system automatically adjusts the regional... The sampling interval for the 12 inner branches was adjusted to 30 seconds; during strong winds, the branch... Three transient current interruptions were detected, confirming a poor contact fault. Meanwhile, the combiner box... Subordinate branch roads for 35 consecutive days The index reached 0.99, and the sampling interval was extended to 10 minutes. The environmental risk perception mechanism accurately captures occasional contact faults caused by strong winds, avoiding the missed detection problems of traditional fixed sampling modes. Simultaneously, it reduces the monitoring resource consumption of stable branches, achieving synergistic optimization of system operating efficiency and fault diagnosis accuracy.

[0094] Based on the same inventive concept, such as Figure 5 As shown, the present invention also provides a distributed photovoltaic fault location and diagnosis method system based on multi-terminal data fusion, the system comprising: The interval calculation module is used to collect real-time operating parameters of the inverter, obtain the measured power, and calculate the theoretical power range. The collaborative triggering module is used to determine whether the measured power exceeds the theoretical power range and generate a collaborative sampling command; The terminal acquisition module is used to respond to the collaborative sampling command, collect the branch current data of the combiner box and environmental monitoring data, and generate a multi-terminal synchronous sampling dataset. The feature fusion module is used to fuse the real-time operating parameters of the inverter, the branch current data of the combiner box, and the environmental monitoring data to generate a multi-dimensional associated feature set. The hierarchical diagnosis module is used to perform feature comparison-based primary diagnosis based on the multi-dimensional associated feature set, obtain preliminary fault area identification, and initiate secondary diagnosis with the preliminary fault area identification to extract the dynamic deviation rate and environmental response anomaly degree of the target area branch and generate a localization diagnosis feature set. The closed-loop optimization module is used to combine the cross-junction box consistency features and power flow tracing features of the multi-dimensional associated feature set to correct and output accurate fault coordinates and fault type identifiers.

[0095] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0096] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion, characterized in that, The method includes: Collect real-time operating parameters of the inverter, obtain the measured power, and calculate the theoretical power range; Determine whether the measured power exceeds the theoretical power range, and generate a collaborative sampling command; In response to the collaborative sampling command, the current data of the combiner box branch and environmental monitoring data are collected to generate a multi-terminal synchronous sampling dataset; By integrating the real-time operating parameters of the inverter, the branch current data of the combiner box, and the environmental monitoring data, a multi-dimensional correlation feature set is generated. Based on the multi-dimensional associated feature set, a feature comparison-based primary diagnosis is performed to obtain a preliminary fault area identifier. Then, based on the preliminary fault area identifier, a secondary diagnosis is initiated to extract the dynamic deviation rate and environmental response anomaly degree of the target area branch and generate a location diagnosis feature set. By combining the cross-junction box consistency features and power flow tracing features of the multi-dimensional associated feature set, accurate fault coordinates and fault type identifiers are corrected and output.

2. The method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion according to claim 1, characterized in that, The generated collaborative sampling instructions include: Collect real-time operating parameters of the inverter and generate historical operating data; Based on the historical operating data, establish the correlation between environmental parameters and theoretical power, and determine the theoretical power value corresponding to the environmental parameters; The theoretical power range is calculated based on the theoretical power value. When the measured power continuously exceeds the theoretical power range, a collaborative sampling command is generated.

3. The method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion according to claim 1, characterized in that, The collected environmental monitoring data includes: Local irradiance data at the string level are acquired using spatially distributed illumination sensors; The number of spatially distributed illumination sensors is spatially matched based on the physical distribution of the combiner box branches. Local temperature data for the same area is obtained by a temperature sensor mounted on the main body of the junction box.

4. The method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion according to claim 2, characterized in that, The generation of the multi-dimensional association feature set includes: Calculate the real-time power of each branch and compare it with the historical operating data to generate branch power contribution deviation characteristics; By combining the environmental monitoring data with the real-time power of each branch, the time-series response characteristics of different branches under the same environmental conditions are analyzed, and the branch-environment coupling response anomaly degree is generated. By comparing the branch power output of adjacent combiner boxes under the same environmental conditions, cross-combiner box consistency characteristics are extracted. Calculate the real-time difference and rate of change between the inverter input power and the sum of the output power of the associated combiner box, and generate power flow tracing characteristics; The branch power contribution deviation characteristics, the branch-environment coupling response anomaly, the cross-combiner box consistency characteristics, and the power flow tracing characteristics are comprehensively extracted to finally generate a multi-dimensional associated feature set.

5. The method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion according to claim 4, characterized in that, The initiation of secondary diagnostics based on the initial fault area identification includes: From the initial fault area identifier, a predetermined number of branches with the highest dynamic power deviation rate are selected to generate a set of suspicious branches; The branch-environment coupling response anomalies of the suspected branch set are filtered to examine the physical correlation between power changes and real-time environmental parameters; By leveraging the cross-junction box consistency characteristics, global environmental fluctuations can be eliminated, confirming the existence of localized specific faults. Based on the power flow tracing characteristics, the fault level is determined, and a component-level fault type identifier or a combiner box internal fault type identifier is output.

6. The method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion according to claim 5, characterized in that, The physical correlation between the test power change and real-time environmental parameters includes: When a sudden increase in irradiance is detected, the abnormal response is determined to be that the power of the suspected branch does not increase synchronously or decreases in the opposite direction. When a localized abnormal increase in temperature is detected, accompanied by a persistently low power, an associated hot spot fault type identifier is generated.

7. The method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion according to claim 5, characterized in that, The process of correcting and outputting accurate fault coordinates and fault type identifiers includes: When the output power of the combiner box is detected to be continuously lower than the total power of its subordinate branches, and the rate of change of the power difference is greater than the predetermined mutation ratio, an internal fault type identifier of the combiner box is generated. When the branch power contribution deviation characteristic is detected to continuously deviate from the threshold and the power flow difference is in a stable state, a component-level fault type identifier is output.

8. The method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion according to claim 1, characterized in that, Also includes: Based on the precise fault coordinates and the fault type identifier, a sampling strategy optimization instruction is generated; The regular sampling frequency of the equipment within the initial fault area is dynamically adjusted, the sampling density is increased for implicit attenuation branches, and the sampling density is decreased for fault-free areas. The trigger sensitivity parameters of the collaborative sampling command are optimized based on historical diagnostic results.

9. A method for fault location and diagnosis of distributed photovoltaic systems based on multi-terminal data fusion according to claim 8, characterized in that, The dynamic adjustment of the regular sampling frequency of the devices within the initial fault area marker includes: In response to occasional poor contact faults, the high-frequency sampling signal is automatically activated during periods of strong winds indicated by weather warnings or under conditions of rapid environmental change. For combiner box branches that have no fault records within a continuous preset period, extend their regular sampling interval.

10. A distributed photovoltaic fault location and diagnosis system based on multi-terminal data fusion, applied to the distributed photovoltaic fault location and diagnosis method based on multi-terminal data fusion as described in any one of claims 1-9, characterized in that, The system includes: The interval calculation module is used to collect real-time operating parameters of the inverter, obtain the measured power, and calculate the theoretical power range. The collaborative triggering module is used to determine whether the measured power exceeds the theoretical power range and generate a collaborative sampling command; The multi-terminal acquisition module is used to respond to the collaborative sampling command, acquire the branch current data of the combiner box and environmental monitoring data, and generate a multi-terminal synchronous sampling dataset. The feature fusion module is used to fuse the real-time operating parameters of the inverter, the branch current data of the combiner box, and the environmental monitoring data to generate a multi-dimensional associated feature set. The hierarchical diagnosis module is used to perform feature comparison-based primary diagnosis based on the multi-dimensional associated feature set, obtain preliminary fault area identification, and initiate secondary diagnosis with the preliminary fault area identification to extract the dynamic deviation rate and environmental response anomaly degree of the target area branch and generate a localization diagnosis feature set. The closed-loop optimization module is used to combine the cross-junction box consistency features and power flow tracing features of the multi-dimensional associated feature set to correct and output accurate fault coordinates and fault type identifiers.

Images