Photovoltaic module dynamic control system and method

Abstract

The present application relates to the field of photovoltaic technology, and particularly relates to a photovoltaic module dynamic control system and method. The method first collects thermal imaging, vibration and electrical three types of multi-dimensional parameters, constructs a time-space fusion data system through time-space alignment; then establishes a time-space evolution model containing multi-causal evolution characteristics, generates an inversion path and calculates a confidence value through reverse iteration when abnormal; verifies and identifies single or composite hot spot causes through layer induction relationship, and finally outputs maintenance control information containing accurate causes and positions. The present application breaks through the limitation of single electrical parameter, realizes passive alarm to active and accurate diagnosis, restores the composite fault cause and effect chain, captures early hot spots and arc hazards in advance, avoids unnecessary module shutdown, improves operation and maintenance efficiency and system safety, and prolongs the service life of the module.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic technology, and in particular to a dynamic control system and method for photovoltaic modules. Background Technology

[0002] Driven by the global energy transition and carbon neutrality goals, photovoltaic (PV) power generation, as a highly promising green energy technology, has seen continuous expansion in installed capacity, with its application scenarios expanding from centralized large-scale ground-mounted power plants to deeply penetrating industrial and commercial rooftops and distributed residential systems. As the integration of PV systems continues to improve, the industry's requirements for energy conversion efficiency and long-term operational reliability are gradually upgrading. The management and control dimension is shifting from the overall power plant level down to individual PV modules, with refined management and control becoming a core trend, aiming to achieve efficient energy harvesting and end-to-end security.

[0003] To address the safety challenges of photovoltaic (PV) strings in complex environments, module-level power electronics technology, centered on safe shutdown and basic data monitoring, has gradually become the industry mainstream. This technology involves adding a control module to the module, relying on power line carrier communication or simple electrical parameter discrimination logic to periodically collect module voltage, current, and temperature data. Upon identifying anomalies, the module connection is disconnected, controlling the system voltage within a safe threshold, thereby improving operational safety. For example, the dynamic control method for PV modules disclosed in Chinese patent CN110970918A enhances system reliability by regulating the output voltage through a module shutdown device.

[0004] However, during the operation of photovoltaic systems, issues such as shading, high-resistance circuit contacts, and circuit breaks accompanied by arcing can easily lead to hot spots. Although the formation mechanisms of hot spots with different causes are quite different, they all disrupt the normal power generation state of the modules, resulting in highly similar abnormal characteristics in the monitored electrical parameters, making them difficult to distinguish directly with conventional monitoring. While the aforementioned control technologies can capture abnormal signals such as voltage, current, and temperature, they struggle to accurately identify the specific cause of the hot spots. To mitigate safety risks, most control systems adopt a conservative strategy of uniformly and forcibly shutting down the module circuits. After the photovoltaic modules are forcibly shut down, maintenance personnel must inspect and repair each faulty module individually. However, the handling methods for different causes of hot spots vary greatly. For example, shading can be resolved simply by cleaning the module, high-resistance contacts require checking for cold solder joints or loose nodes, and arcing faults necessitate the emergency replacement of damaged cables or modules. Because the cause of the fault is unclear, maintenance personnel often face the problem of excessive loads or insufficient tools requiring resupply, increasing maintenance pressure and reducing repair efficiency. Summary of the Invention

[0005] This invention provides a dynamic control system and method for photovoltaic modules to solve the problems of unclear hot spot causes, difficulty for maintenance personnel to prepare suitable tools in advance, and low maintenance efficiency.

[0006] To solve the above-mentioned technical problems, this application provides the following technical solution: The dynamic control method for photovoltaic modules includes the following steps: Multidimensional parameters are collected, including thermal imaging data of the photovoltaic panel surface, vibration data of each preset position point of the photovoltaic panel, and electrical data of the component circuit. Based on the collection time, the spatial coordinate system established with the fixed position and orientation of the photovoltaic panel, and the topological connection relationship of the component circuit, the multidimensional parameters are aligned in time and space to obtain spatiotemporal fusion data. From the spatiotemporal fusion data, the numerical evolution sequence of each multidimensional parameter within the preset abnormal parameter range on the time axis, as well as the coupling law between different multidimensional parameters, are extracted to construct a spatiotemporal evolution model containing the evolution characteristics corresponding to the causes of multiple hot spots. When any current multidimensional parameter is within the range of abnormal parameters, the multidimensional parameter sequence within the preset retrospective time window is obtained as the actual observation sequence. In the spatiotemporal evolution model space, based on the current multidimensional parameters, the inverse iterative analysis along the time axis according to the coupling law is performed to obtain the inversion path reflecting the formation process of potential hot spots. The inversion path is processed into a predicted preceding sequence. By performing time-point similarity feature residual calculation on the predicted preceding sequence and the actual observation sequence, the confidence value corresponding to the inversion path is obtained. Inversion paths with confidence values ​​exceeding a preset confidence threshold are selected as reference inversion paths. The hot spot genesis corresponding to the unique reference inversion path is taken as the final gene. Alternatively, the triggering order of each hot spot gene and the transmission time constant of the spatiotemporal evolution model are extracted from each reference inversion path. The evolution time difference is obtained by measuring the numerical variation of parameters in each dimension in the actual observation sequence. The numerical deviation between the transmission time constant and the evolution time difference is compared to verify the hierarchical induction relationship between the prior hot spot genesis and the subsequent hot spot genesis. The combination of hot spot genesis that satisfies the hierarchical induction relationship is determined as the final gene. Maintenance control information is generated based on the final cause, corresponding thermal imaging data, or component circuitry.

[0007] The basic principle and beneficial effects of this invention are as follows: By simultaneously acquiring three types of multi-dimensional parameters—thermal imaging, vibration, and electrical parameters—and aligning time and space based on a unified spatiotemporal coordinate system, this invention constructs a spatiotemporal fusion data system that comprehensively reflects the physical essence of hot spots. This solves the problem of difficulty in distinguishing hot spots with similar characteristics due to different causes. Furthermore, it captures multi-dimensional physical characteristics of hot spots, overcoming the technical limitations of monitoring only electrical parameters. Based on this, the invention establishes a spatiotemporal evolution model containing multiple hot spot formation and evolution characteristics. Starting from the current abnormal state, it iteratively generates an inversion path along the time axis. Confidence values ​​are calculated using the residuals of similarity features at each time point, solving the problem of difficulty in identifying the cause of hot spots from abnormal data. This also improves upon passive anomaly alarms by enabling proactive and accurate cause determination, effectively avoiding unnecessary component shutdowns and power generation losses caused by temporary anomalies. Meanwhile, this invention, through a hierarchical induction relationship verification mechanism, extracts the triggering order and model propagation time constant of hotspot formation in each reference inversion path, and verifies the causal relationship by comparing the evolution time difference of the actual observation sequence. This solves the problem of handling multiple superimposed faults and also achieves complete reconstruction of the fault causal chain, avoiding misjudging composite causes as multiple independent faults, and significantly improving the diagnostic accuracy under complex operating conditions. Finally, through the synergistic effect of the above-mentioned multimodal data fusion, inverse inversion quantitative verification, and composite cause causal analysis, this invention outputs maintenance control information containing accurate causes, hotspot location coordinates, and component circuit identification, solving the core problems of blind troubleshooting and insufficient tool preparation by maintenance personnel. This achieves standardization of maintenance processes and, based on accurate fault cause determination, significantly improves maintenance efficiency and reduces the workload of maintenance personnel.

[0008] Furthermore, most hot spot development is a gradual process. Currently, most technologies only detect hot spots when they are severe and electrical parameters are significantly abnormal, making it difficult to track subtle parameter changes during the gradual development of hot spots and to continuously identify weak anomalies. Various damage characteristics accumulate over a long period and then erupt due to environmental interference such as temperature, humidity, and light. By the time an alarm is triggered due to significant changes in electrical parameters, the latent damage accumulated in the early stages has been masked by the sudden anomaly, resulting in incomplete identification of the fault cause and inadequate fault tracing. In addition, this invention relies on multi-dimensional parameter spatiotemporal coupling analysis to capture early weak features of hot spots. It can diagnose hot spots in advance when high-resistance contacts only produce a small temperature rise and weak vibration and when electrical parameters have not changed, allowing for timely intervention to avoid permanent damage to devices and extend the service life of components. For arc faults that are prone to causing fires, it relies on high-frequency vibration combined with thermal imaging and electrical data collaborative discrimination to complete identification and device shutdown at the fault initiation stage, avoiding fire hazards.

[0009] In summary, this invention solves the problems of traditional photovoltaic hot spot monitoring relying solely on single electrical parameters, difficulty in distinguishing similar fault characteristics, incomplete tracing of composite faults, and delayed fault alarms by integrating multimodal and multidimensional parameter spatiotemporal fusion modeling, fault inverse inversion confidence verification, and hot spot hierarchical causal relationship verification. It also achieves accurate prediction of early-stage minor faults in photovoltaic modules, accurate identification of the causes of single and composite hot spots, and accurate fault location, effectively avoiding unnecessary module shutdowns and power plant safety hazards, standardizing photovoltaic operation and maintenance processes, and significantly improving fault repair efficiency and overall operational safety of photovoltaic modules.

[0010] Furthermore, all data segments within the preset abnormal parameter range are selected from the spatiotemporal fusion data, and numerical evolution sequences corresponding to thermal imaging data, vibration data, and electrical data are generated according to the acquisition time sequence. The correlation between different multidimensional parameters within the same time interval is calculated to obtain the coupling law between different multidimensional parameters. The numerical evolution sequence features and coupling law features corresponding to each preset hot spot cause are integrated to construct a spatiotemporal evolution model containing the evolution features corresponding to multiple hot spot causes.

[0011] This invention filters anomalous parameter data fragments and constructs multi-parameter time-series evolution sequences and cross-dimensional coupling rules, eliminating interference from massive amounts of normal data and accurately focusing on the entire process of hot spot anomaly evolution. This invention overcomes the limitations of single-parameter analysis, deeply mining the implicit correlation features of thermal imaging, vibration, and electrical parameters, and completing modeling based on multi-dimensional coupling rules, enabling the capture of early, weak anomaly precursors. For initial hot spot anomalies caused by slight dust accumulation and hidden solder joints in photovoltaic modules, it can accurately identify multi-parameter coordinated small-amplitude perturbation features, constructing a dedicated evolutionary feature system adapted to various hot spot causes, providing high-precision model support for subsequent fault inversion and cause tracing, and significantly improving the modeling accuracy and completeness of hot spot diagnosis.

[0012] Furthermore, the multidimensional parameters corresponding to photovoltaic modules at the same acquisition time point and the same spatial coordinate position are integrated into multidimensional parameter integrated data after time and space alignment using a unified spatiotemporal coordinate system. In the spatiotemporal evolution model space, the current multidimensional parameter integrated data is used as the initial data set for inverse iteration. The variation constraints of each multidimensional parameter within each time step are determined according to the coupling law, where the time step is the acquisition sampling interval of the multidimensional parameters. The multidimensional parameter integrated data corresponding to each preceding time point is solved step by step in reverse along the time axis to generate at least one complete multidimensional parameter change trajectory as the inversion path. The multidimensional parameter integrated data corresponding to all preceding time points in the inversion path are arranged in chronological order and processed into a prediction preceding sequence.

[0013] This invention achieves spatiotemporal alignment and integration of multi-dimensional parameters, forming standardized multi-dimensional parameter integrated data, thus solving the problems of spatiotemporal misalignment and parameter mismatch in traditional data. This invention uses a fixed acquisition time step to constrain the inverse iterative logic, standardizing the boundary conditions for parameter changes and effectively avoiding the problems of distorted fault inversion trajectories and data disorder. For complex operation and maintenance scenarios involving densely arranged multi-component photovoltaic arrays and intertwined parameter interference, it can stably generate continuous and realistic fault evolution inversion trajectories, outputting standardized prediction preorder sequences to ensure the accuracy and stability of subsequent residual comparison, confidence calculation, and fault source tracing results.

[0014] Furthermore, thermal imaging data, vibration data, and electrical data under various preset hotspot causes are collected. After temporal and spatial alignment, a historical training dataset is constructed for training the spatiotemporal evolution model. The feature residuals of thermal imaging data, vibration data, and electrical data in the predicted preceding sequence and the actual observation sequence are calculated at each corresponding acquisition time point. From the historical training dataset, the ratio of inter-class variance to intra-class variance of each multi-dimensional parameter under each hotspot cause is calculated to obtain the feature discrimination of each multi-dimensional parameter. The contribution of each multi-dimensional parameter to hotspot cause identification is determined according to the magnitude of the feature discrimination. After normalizing the contribution, the influence weight corresponding to each dimension of data is obtained. The feature residuals of each dimension of data are multiplied by the corresponding influence weights and summed to obtain the comprehensive residual. The reciprocal of the comprehensive residual is normalized to obtain the confidence value corresponding to the inversion path.

[0015] This invention quantifies the ratio of inter-class to intra-class variance of each parameter based on historical training datasets, accurately defining the fault feature discrimination of parameters in different dimensions and achieving adaptive matching of the influence weights of data in each dimension. This invention obtains accurate confidence values ​​through differentiated weighted residual calculation, effectively mitigating environmental noise interference caused by fluctuations in light intensity, temperature, and humidity, and amplifying the identification weight of core fault parameters. For scenarios with high parameter noise and easily masked abnormal features in complex outdoor photovoltaic applications, this invention effectively solves the problems of distorted confidence in inversion paths and misjudgments or omissions in fault identification, significantly improving the anti-interference capability and accuracy of hot spot cause identification.

[0016] Furthermore, by comparing the integrated multidimensional parameter data corresponding to two adjacent preceding time points in the reference inversion path, the overall change amplitude is obtained, and the comparison result between the overall change amplitude and the preset threshold is obtained. Based on the comparison result, the mutation point of the integrated multidimensional parameter data is determined. Based on the mutation point of the integrated multidimensional parameter data, each reference inversion path is segmented and analyzed. The starting time point of each segment is determined as the initial perturbation occurrence time point of the corresponding hot spot cause. All hot spot causes are arranged in chronological order according to the initial perturbation occurrence time points to obtain the triggering order of each hot spot cause. The time parameter corresponding to each hot spot cause is extracted from the spatiotemporal evolution model as the propagation time constant of the hot spot cause.

[0017] This invention identifies abrupt change points by comparing the overall change magnitude of parameters with a preset threshold, enabling refined segmented analysis of the inversion path and overcoming the limitation of overall time-series analysis of abnormal multidimensional data in breaking down multi-stage faults. By segmenting and locating the initial disturbance occurrence time, sorting out the triggering sequence of hot spot formation, and extracting the standardized propagation time constant of each cause, the hierarchical evolution process of progressive and superimposed hot spots can be accurately deconstructed. For multi-stage fault scenarios involving loosening, overheating, and high resistance superposition in photovoltaic modules, the timing triggering logic and evolution rate of each fault can be clearly distinguished, providing accurate time-series data support for verifying the causal relationship of composite faults.

[0018] Furthermore, from the actual observation sequence, the corresponding acquisition time when the change amplitude of the integrated data of multidimensional parameters corresponding to the cause of each hot spot exceeds the preset threshold of the change amplitude is extracted; the interval between the acquisition times corresponding to two adjacent hot spot causes is calculated to obtain the evolution time difference; the difference between the transmission time constant corresponding to the pre-existing hot spot cause and the evolution time difference is obtained to obtain the numerical deviation; when the numerical deviation falls within the preset value range, it is determined that the pre-existing hot spot cause has a cascading induced relationship with the subsequent hot spot cause.

[0019] This invention quantifies the numerical deviation between the evolution time difference and the propagation time constant of adjacent faults, using a preset value range as the criterion for judgment. It standardizes the identification of the cascading causal relationships between hot spots, reducing the shortcomings in distinguishing between independent faults and related composite faults. This invention can accurately verify the inducing effect of a preceding fault on a subsequent fault, effectively avoiding the problems of mis-splitting and mis-judging composite faults. For chain fault scenarios such as photovoltaic shading dust accumulation leading to high-resistivity heating and circuit aging leading to abnormal arcing, it accurately reconstructs the fault causal transmission chain, significantly improving the diagnostic accuracy of composite hot spot causes under complex operating conditions.

[0020] Furthermore, the causes of hot spots include shading, circuit disconnection, and high-resistance contact; the method also includes the following steps: S10: Integrating data and binding spatial coordinates based on multi-dimensional parameters, establishing a one-to-one correspondence between the final cause, thermal imaging data, and component circuits for maintenance; obtaining the maintenance location and corresponding component circuit from the thermal imaging data based on the maintenance correlation, and generating maintenance control information; disconnecting the component circuit from the maintenance control information; obtaining the circuit in the disconnected component circuit whose final cause includes high-resistance contact as the test connection circuit; obtaining the maintenance location corresponding to each test connection circuit from the maintenance correlation as the test connection location; and determining the test connection location based on the test connection location in space. The coordinates in the reference system are used to integrate adjacent test connection positions into a test connection range; S20: The test connection circuit corresponding to the outermost test connection position in the test connection range is obtained as the priority test connection circuit, the priority test connection circuit is connected, and the time corresponding to the preset traceability time window is waited. If the priority test connection circuit is not disconnected after the wait ends, the test connection position corresponding to the priority test connection circuit is removed from the test connection range; return to S20; if the priority test connection circuit is disconnected again after the wait ends, the test connection circuit corresponding to the test connection position in the test connection range, the corresponding final cause and the repair position are updated to update the maintenance control information.

[0021] This invention establishes an accurate correlation between the final cause, thermal imaging data, and component circuitry for troubleshooting, achieving a one-to-one match between fault location, fault type, and faulty circuit. Temperature transmission in photovoltaic fault areas can interfere with surrounding normal circuits, easily leading to widespread misdiagnosis and mishandling. This invention utilizes a high-resistance contact fault detection mechanism, employing a sequential reconnection and secondary disconnection verification process from the outside in, to automatically identify and exclude normal circuits, achieving automated screening of benign circuits. In scenarios involving suspected faults in contiguous modules, it effectively eliminates fault-free circuits, pinpoints the true fault location, significantly reduces the scope of repair, minimizes ineffective repair work, achieves accurate location and efficient repair, and significantly improves photovoltaic operation and maintenance efficiency.

[0022] Further, in step S20, a traceability time window is set for each disconnected component circuit; the transfer time constant of the adjacent component circuits of the disconnected component circuit during maintenance is obtained, and the transfer time constant is extracted from the spatiotemporal evolution model; when the final cause includes circuit disconnection or high-resistance contact, the hot spot position corresponding to the final cause is obtained from the thermal imaging data, at least one adjacent component circuit that is spatially adjacent to the hot spot position is obtained, and the coupling law corresponding to the adjacent component circuit is extracted from the spatiotemporal evolution model as a reference law; a corresponding benchmark coupling law is preset for each coupling law, and if the difference between the reference law and the benchmark coupling law exceeds the preset coupling change threshold, the duration of the traceability time window of the adjacent component circuit is shortened.

[0023] This invention extracts the coupling patterns of adjacent circuits based on a spatiotemporal evolution model. By referencing the degree of difference between the coupling patterns and a baseline coupling pattern, it dynamically and adaptively adjusts the traceability time window duration. The heat in the photovoltaic fault area accelerates the aging of surrounding normal circuits. Fixed traceability durations have poor adaptability, computational redundancy, and inaccurate judgments. This invention uses the differences in coupling patterns to assess the degree of circuit aging, achieving multi-purpose reuse of parameter rules. The dynamically adjusted traceability duration can accurately adapt to circuit conditions with different aging states and different coupling strengths, balancing the flexibility and accuracy of fault verification, effectively reducing redundant computational load on equipment, and improving operational adaptability and judgment efficiency.

[0024] Furthermore, when the final cause includes both prior occlusion and subsequent circuit disconnection, the following steps are executed: S30: Based on the hot spot position in the maintenance control information, analyze the occlusion position and its spatial movement trajectory corresponding to the occlusion, and predict the first prediction range of the occlusion position within the next traceability time window based on the positive evolution characteristics of shadow occlusion in the spatiotemporal evolution model; S31: Identify the component circuit whose hot spot position is within the first prediction range as an aging sensitive circuit, and disconnect the aging sensitive circuit; wait for the duration corresponding to a traceability time window; S32: After the wait ends, use the hot spot position obtained during the waiting process as a reference position. When the proportion of the hot spot position within the first prediction range in the reference position is greater than the preset accuracy threshold, and the vibration amplitude of the vibration data corresponding to each position within the first prediction range is less than the preset minimum vibration amplitude, connect each aging sensitive circuit.

[0025] This invention addresses the complex fault of initial shading followed by circuit disconnection. It predicts the shading movement range based on shadow evolution characteristics, proactively identifying aging-sensitive circuits and disconnecting them for protection. Vibration and shock from shading can exacerbate damage to aging circuits, inducing high-resistance contact or sudden circuit disconnection. This invention predicts the subsequent vibration impact area through trajectory prediction, disconnecting aging circuits in advance to avoid sudden circuit damage and secondary faults caused by vibration impact. After the shading subsides and vibration disappears, circuit connectivity is restored systematically using both point percentage and vibration amplitude as criteria. This effectively avoids abrupt damage to aging circuits caused by residual vibration and current surges, while also reducing the negative impact of fault management on surrounding circuits, balancing circuit protection performance and module power generation stability. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the dynamic control method for photovoltaic modules. Figure 2 A schematic diagram of the process for regional testing and verification of photovoltaic hot spots with multiple causes; Figure 3 A flowchart illustrating the process of predicting photovoltaic shading and controlling sensitive circuits. Detailed Implementation

[0027] The following will describe the concept and technical effects of the present invention clearly and completely with reference to embodiments, so as to fully understand the purpose, features and effects of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the scope of protection of the present invention. Example 1 like Figure 1 As shown, the dynamic control method for photovoltaic modules includes the following steps: Multidimensional parameters are collected, including thermal imaging data of the photovoltaic panel surface, vibration data of each preset position point of the photovoltaic panel, and electrical data of the component circuit. Based on the collection time, the spatial coordinate system established with the fixed position and orientation of the photovoltaic panel, and the topological connection relationship of the component circuit, the multidimensional parameters are aligned in time and space to obtain spatiotemporal fusion data.

[0028] Specifically, small, fixed infrared thermal imaging probes are installed at pre-set measurement points on the outer side of the photovoltaic panel. These probes face vertically towards the panel surface, continuously receiving infrared radiation energy from the photovoltaic surface. The built-in photoelectric sensing element converts the infrared radiation signal into an analog electrical signal, which is then processed by an internal signal conditioning and analog-to-digital conversion module to generate digital thermal imaging data with spatial location and timestamps. Strain gauge vibration sensing components are installed at various monitoring points on the photovoltaic panel surface. The strain gauges are tightly fitted to the photovoltaic panel surface and frame. When the photovoltaic module undergoes mechanical deformation and vibration, the strain gauges deform synchronously with the panel surface, resulting in a change in resistance. This change is converted into an analog voltage signal by a matching Wheatstone bridge, and then digitized by an onboard AD converter, ultimately outputting standardized vibration data with location and time stamps. The component electrical data is acquired using a layered acquisition method. The string busbar side is equipped with a multi-channel integrated Hall effect sampling module, which collects the string loop current through magnetic isolation induction and combines it with the voltage at the voltage divider terminal. After internal circuit shaping, the string-level digital electrical data is output. The individual component junction box contains an onboard sampling unit consisting of a manganese-copper sampling resistor, a voltage divider circuit, and an onboard AD sampling chip. This unit collects analog current through the resistor voltage drop and analog voltage through the voltage divider circuit. The onboard AD chip performs local analog-to-digital conversion to obtain high-precision electrical data for the individual component. All three parameters carry a unified timing sequence, spatial coordinates, and device number, providing a complete and well-organized raw data foundation for subsequent spatiotemporal alignment and data fusion.

[0029] A globally unified spatial coordinate system is pre-established based on the factory installation reference orientation of the photovoltaic modules. The origin of the coordinate system is selected from the fixed corner point of the first module at the edge of the array, and the axis of the coordinate system is aligned with the length and width of the module arrangement. The factory-preset hardware topology wiring relationship of the modules is simultaneously recorded as the basis for circuit topology connection. The acquisition timestamp of each parameter is used as the time series reference. Combined with the three constraints of spatial coordinates and circuit topology, multi-dimensional parameter spatiotemporal alignment is completed. After alignment, the data is collected and organized to form spatiotemporal fusion data, realizing the one-to-one binding of parameters at the same time and spatial location. The preset traceability time window, preset abnormal parameter range, and preset information threshold are all statistically calibrated based on the long-term historical operation statistics of photovoltaic modules in this region. The specific calculation method is set by the operation and maintenance personnel. Extreme values ​​and critical values ​​are statistically analyzed from a large number of historical normal and fault samples to determine the reasonable values ​​of each preset parameter.

[0030] From the spatiotemporal fusion data, the numerical evolution sequence of each multidimensional parameter within the preset abnormal parameter range on the time axis, as well as the coupling law between different multidimensional parameters, are extracted to construct a spatiotemporal evolution model containing the evolution characteristics corresponding to the causes of multiple hot spots.

[0031] All data segments falling within the preset abnormal parameter range are selected from the spatiotemporal fusion data. Specifically, the abnormal interval derived from historical fault samples is used as the preset abnormal parameter range. Redundant data falling within the normal parameter range are removed from the spatiotemporal fusion data, and data segments falling within the abnormal interval are retained. The data is then split into numerical evolution sequences (each with its own independent time axis numerical evolution sequence) according to the acquisition time sequence for thermal imaging data, vibration data, and electrical data.

[0032] The degree of correlation between different multidimensional parameters within the same time interval is calculated to obtain the coupling rules between them. Specifically, Pearson correlation analysis is selected as the correlation analysis algorithm. The input of this algorithm is two sets of synchronously changing time series data of parameters within the same continuous observation period. The operation process sequentially completes the mean calculation, covariance calculation, and standardization conversion, and finally outputs the correlation coefficient with a value between a fixed interval. The magnitude of the correlation coefficient distinguishes the strength of the parameter correlation, thereby obtaining the coupling rules between different parameters. More specifically, the mean and covariance are calculated item by item for each corresponding time point, and cross-time series misalignment is prohibited; the standardization conversion uniformly standardizes the output interval to [-1, 1], and the positive and negative signs of the coefficients distinguish the direction of correlation, and the absolute value represents the tightness of coupling. An additional preset classification threshold is used to classify the coupling rules into three categories: strong coupling, weak coupling, and no coupling, based on the correlation coefficient. The multi-parameter coupling rules are determined based on the classification results.

[0033] The numerical evolution sequence features and coupling law features corresponding to each preset hotspot cause are integrated to construct a spatiotemporal evolution model containing the evolution features corresponding to each of the multiple hotspot causes. The evolution sequence features and coupling law features matched to three preset hotspot causes—occlusion, circuit disconnection, and high-resistance contact—are summarized and integrated to build the spatiotemporal evolution model. The data source used for constructing the spatiotemporal evolution model is historical spatiotemporal fusion data corresponding to multiple types of faults. The modeling method is feature label binding and aggregation, which is used to constrain the boundary of inverse iteration parameter changes and assist in fault inversion path deduction.

[0034] When any current multidimensional parameter falls within the range of outliers, the multidimensional parameter sequence within a preset retrospective time window is acquired as the actual observation sequence. In the spatiotemporal evolution model space, based on the current multidimensional parameters, an inverse iterative analysis is performed along the time axis according to the coupling rules to obtain the inversion path reflecting the formation process of potential hotspots. This inversion path is then processed into a predicted preceding sequence. By performing time-point similarity feature residual calculations on the predicted preceding sequence and the actual observation sequence, the confidence value corresponding to the inversion path is obtained.

[0035] Specifically, the multidimensional parameters corresponding to photovoltaic modules at the same acquisition time and spatial coordinate location are integrated into multidimensional parameter integrated data after time and space alignment using a unified spatiotemporal coordinate system. In the spatiotemporal evolution model space, the current multidimensional parameter integrated data is used as the initial data set for inverse iteration.

[0036] The constraints on the changes of each multidimensional parameter within each time step are determined according to the coupling law, where the time step is the sampling interval for acquiring the multidimensional parameters. Specifically, the inherent sampling interval of the hardware sensor is used as a fixed time step, and reasonable floating boundaries of each parameter under each time step, i.e., parameter change constraints, are delineated based on the obtained coupling law.

[0037] The process involves progressively solving the integrated multidimensional parameter data corresponding to each preceding time point along the time axis in reverse order, generating at least one complete multidimensional parameter change trajectory as the inversion path. The integrated multidimensional parameter data corresponding to all preceding time points in the inversion path are then arranged chronologically and processed into a predicted preceding sequence. Specifically, starting from the current acquisition moment, the process proceeds backward along the time axis in the historical direction, solving the corresponding integrated multidimensional parameter data for each preceding time point, and continuously stitching together the data change trajectories at each moment to obtain at least one inversion path. All historical multidimensional parameter integrated data contained in the inversion path are then rearranged in chronological order from earliest to latest to obtain the predicted preceding sequence. This step achieves misalignment correction of multi-source data from densely arranged photovoltaic modules, adapting to on-site conditions where parameters from multiple modules interfere with each other.

[0038] Thermal imaging, vibration, and electrical data under various preset hotspot causes were collected and, after temporal and spatial alignment, a historical training dataset was constructed for training the spatiotemporal evolution model. The feature residuals of thermal imaging, vibration, and electrical data at each corresponding acquisition time point were calculated between the predicted preceding sequence and the actual observation sequence. Specifically, for the same acquisition time point in both sequences, the measured and predicted values ​​of the same dimension parameters for thermal imaging, vibration, and electrical data were extracted, and the difference between the two types of data was used as the original deviation at a single time point. All time point deviations of the same dimension within the entire time window were summarized and processed, and a unified feature standardization conversion was performed to finally form the overall feature residual for that dimension.

[0039] From the historical training dataset, the ratio of inter-class variance to intra-class variance of each multi-dimensional parameter under each hot spot cause is statistically calculated to obtain the feature discrimination of each multi-dimensional parameter. The contribution of each multi-dimensional parameter to hot spot cause identification is determined based on the feature discrimination. After normalizing the contribution, the influence weight corresponding to each dimension of data is obtained. The feature residuals of each dimension of data are multiplied by their corresponding influence weights and summed to obtain the comprehensive residual. The reciprocal of the comprehensive residual is normalized to obtain the confidence value corresponding to the inversion path. In this embodiment, the ratio of inter-class variance to intra-class variance is used to characterize the feature discrimination. The higher the ratio, the more significant the difference of the parameter under different fault conditions and the smaller the fluctuation of data for the same type of fault. The discrimination of the three types of parameters—thermal imaging, vibration data, and electrical data—is horizontally ranked; the higher the value, the stronger the contribution to fault identification. The weights are converted using extreme value normalization. Specifically, the maximum and minimum values ​​of the three class discrimination indices are selected as the benchmark. The conversion is completed by dividing the difference between the single-class parameter discrimination indices and the minimum value by the maximum and minimum difference. The result falling in the range of 0 to 1 is the influence weight of the corresponding dimension. The higher the discrimination indices, the greater the weight.

[0040] Inversion paths with confidence values ​​exceeding a preset confidence threshold are selected as reference inversion paths. The hotspot genesis corresponding to the unique reference inversion path is taken as the final genesis. Alternatively, the triggering order of each hotspot genesis and the propagation time constant of the spatiotemporal evolution model are extracted from each reference inversion path. The evolution time difference is obtained by measuring the numerical variation of parameters in each dimension in the actual observation sequence. The numerical deviation between the propagation time constant and the evolution time difference is compared to verify the hierarchical induction relationship between the pre-existing hotspot genesis and the subsequent hotspot genesis. The combination of hotspot genesis that satisfies the hierarchical induction relationship is determined as the final genesis.

[0041] In the specific online monitoring process, when any real-time multidimensional parameter falls within the aforementioned preset abnormal parameter range, all multidimensional parameters falling within the preset traceability time window (set by the maintenance personnel) from the fault tracing back to the current time are extracted and summarized into the actual observation sequence. Within the established spatiotemporal evolution model space, starting from the current multidimensional parameters, a time-series reverse iterative deduction is carried out according to the coupling law constraints. This deduction generates multiple inversion paths that can reconstruct the entire fault occurrence process and organizes them into a prediction preorder sequence. The data from the same collection time in the prediction preorder sequence and the actual observation sequence are matched one by one. The independent feature residuals of each dimension are calculated item by item by relying on the similarity comparison method. The variable correspondence between the single-dimensional residual, the weighted conversion of the sub-item score, the sum of the sub-item scores, and the normalized mapping to generate the confidence value is established, realizing the transformation of the multidimensional residual into the final confidence value.

[0042] Inversion paths with confidence values ​​higher than a preset confidence threshold are selected as reference inversion paths: when only a single reference inversion path exists, the hotspot origin associated with it is directly determined as the final origin. When multiple reference inversion paths exist, the sequence of hotspot origin triggering and the propagation time constants corresponding to various origins stored in the spatiotemporal evolution model are extracted from each path. The evolution time difference is calculated by combining the variation amplitude of each parameter in the actual observation sequence. The degree of deviation between the propagation time constant and the evolution time difference is compared to verify the hierarchical induced relationship between the origins before and after. The combination of origins that satisfies the induced correlation is taken as the final origin.

[0043] Among them, the propagation time constant is determined based on the statistical analysis of long-term operation and maintenance historical monitoring data of the same type of photovoltaic modules. The operation and maintenance personnel summarize the time interval from the occurrence of the cause of various faults such as shading, circuit disconnection, and high resistance contact to the abnormal manifestation of parameters, and statistically analyze the time average of multiple sets of samples as the benchmark value. Finally, the benchmark value corresponding to the causes of various hot spots is determined as the propagation time constant, and the obtained value is stored in the spatiotemporal evolution model for retrieval and use.

[0044] The deviation is obtained by subtracting the preset transmission time constant of the corresponding fault type from the evolution time difference; the deviation is compared with the interval boundary based on the allowable deviation interval defined in advance by relying on historical operation and maintenance data: if the deviation falls within the interval, the two are considered to match, and it is determined that the faults before and after have a hierarchical induced relationship; if the deviation exceeds the interval range, the induced association is excluded.

[0045] Maintenance control information is generated based on the final cause, corresponding thermal imaging data, or component circuitry.

[0046] Specifically, the thermal imaging data collected by the infrared probe is bound to a preset global coordinate system for the measurement points. Abnormal high-temperature areas are delineated based on the grayscale differences of the thermal imaging pixels. These delineated areas are then converted into horizontal and vertical coordinates in the coordinate system to determine the spatial location of the hot spot. A one-to-one mapping ledger between spatial coordinates and individual components and branches is pre-established, recording the coordinate range of each component's border and its corresponding circuit number. By searching the ledger using the coordinates of the hot spot, if the coordinates fall within the defined coordinate range of a component, the individual component and its corresponding series branch can be identified. Combined with the previously determined final fault cause, standardized maintenance control information is generated by integrating the hot spot location and the faulty circuit number.

[0047] This embodiment also includes a photovoltaic module dynamic control system that uses the above-described photovoltaic module dynamic control method.

[0048] Example 2 The difference between this embodiment and Embodiment 1 lies only in that the integrated multidimensional parameter data corresponding to two adjacent preceding time points in the reference inversion path are compared to obtain the overall change amplitude. The comparison result of the overall change amplitude and a preset threshold is obtained, and the mutation point of the integrated multidimensional parameter data is determined based on the comparison result. According to the mutation point of the integrated multidimensional parameter data, each reference inversion path is segmented and analyzed, that is, the single reference inversion path is segmented with the mutation point as the boundary, and the starting time point of each segment is determined as the initial perturbation occurrence time point of the corresponding hot spot cause. All hot spot causes are arranged in chronological order according to the initial perturbation occurrence time points to obtain the triggering order of each hot spot cause. The time parameter corresponding to each hot spot cause is extracted from the spatiotemporal evolution model as the propagation time constant of the hot spot cause.

[0049] In this embodiment, the preset threshold for determining abrupt changes is determined by maintenance personnel based on historical fault mutation samples of components of the same model. The change amplitude at each point along each path is compared with the preset threshold, and points where the change amplitude exceeds the preset threshold are designated as parameter mutation points.

[0050] All causes are sorted from earliest to latest according to their initial disturbance time points to determine the triggering order of the entire set of hot spot causes; the propagation time constants corresponding to various hot spot causes are still retrieved from the spatiotemporal evolution model. The propagation time constants are determined by relying on the statistical average of historical fault interval durations, following the method of Example 1.

[0051] From the actual observation sequence, the corresponding acquisition time when the change amplitude of the integrated data of multidimensional parameters corresponding to the cause of each hot spot exceeds the preset threshold of the change amplitude is extracted; among them, a large number of samples of the natural fluctuation amplitude of multidimensional parameters under fault-free steady-state operation and the parameter mutation amplitude induced by various hot spot disturbances are collected. The operation and maintenance personnel remove the normal small disturbance data caused by wind and sand and diurnal temperature changes, take the upper limit of steady-state fluctuation as the benchmark, and fix the preset threshold of change amplitude based on the benchmark.

[0052] The evolution time difference is obtained by calculating the time interval between the acquisition times corresponding to the causes of two adjacent hot spots in the time series. The difference between the propagation time constant corresponding to the cause of the prior hot spot and the evolution time difference is obtained to obtain the numerical deviation. The preset range of the numerical deviation is defined by the summary of historical operation and maintenance fault data. When the numerical deviation falls within the preset range, it can be determined that the cause of the prior hot spot can induce the generation of the subsequent hot spot, and the two have a hierarchical induced correlation. In this way, the effective combination of causes is selected as the final cause of the fault.

[0053] Example 3 The only difference between this embodiment and Embodiment 1 is that the causes of the hot spot include shielding, circuit disconnection, and high-resistance contact; such as Figure 2 As shown, the following regional testing and verification steps for photovoltaic multi-gene hot spots are also included: S10: Based on multi-dimensional parameter integration data and spatial coordinate binding, establish a one-to-one correspondence between the final cause, thermal imaging data, and component circuits for maintenance. Based on this maintenance correlation, obtain the maintenance location and corresponding component circuit from the thermal imaging data, and generate maintenance control information. Disconnect the component circuit from the maintenance control information; obtain the circuit with high-resistance contact as the final cause among the disconnected component circuits as the test connection circuit. Obtain the maintenance location corresponding to each test connection circuit from the maintenance correlation as the test connection location. Based on the coordinates of the test connection location in the spatial coordinate system, integrate adjacent test connection locations into a test connection range.

[0054] The system uses multi-dimensional parameter integration data with spatial coordinates as the core link to establish a maintenance correlation mapping ledger that uniquely corresponds to the final hot spot cause, the coordinate position of the thermal imaging hot spot, and the component circuit number. This ledger is built based on the photovoltaic module's factory topology and global spatial coordinate system. When obtaining the maintenance location corresponding to each test connection circuit from the maintenance correlation relationship as the test connection location, circuits whose final fault cause includes high-resistance contact are defined as test connection circuits, and the spatial maintenance location corresponding to each test connection circuit in the ledger is retrieved simultaneously.

[0055] S20: Obtain the test connection circuit corresponding to the outermost test connection position in the test connection range as the priority test connection circuit, connect the priority test connection circuit, and wait for the duration corresponding to the preset traceability time window. If the priority test connection circuit is not disconnected after the wait ends, that is, the priority test connection circuit has not been disconnected twice, it proves that the priority test connection circuit has no hidden fault, then remove the test connection position corresponding to the priority test connection circuit from the test connection range; return to S20, and continue to iterate and verify the remaining circuit.

[0056] If the priority test connection circuit is disconnected again after the waiting period ends, it is determined that there is an associated hidden fault in the area. Then, the test circuits, matching hot spot cause types and maintenance location information corresponding to all points within the current test connection range are synchronously updated to the maintenance control information to complete the fault information.

[0057] In step S20, a traceability time window is set for each disconnected component circuit; the propagation time constant of the adjacent component circuits of the disconnected component circuit during maintenance is obtained, wherein the propagation time constant is extracted from the spatiotemporal evolution model. When the final cause includes circuit disconnection or high-resistance contact, the high-temperature hot spot region is identified based on the grayscale difference of thermal imaging data, the accurate spatial coordinates of the hot spot are calculated, the spatial coordinate system ledger is retrieved, and at least one set of adjacent component circuits spatially adjacent to the hot spot location is matched and obtained; at the same time, the parameter coupling law corresponding to the adjacent circuit is extracted from the spatiotemporal evolution model as a reference law for fault determination.

[0058] A corresponding benchmark coupling rule is preset for each coupling rule. The benchmark coupling rule is trained and calibrated using massive amounts of normal steady-state operating data, representing the fixed correlation characteristics of thermal imaging, vibration, and electrical parameters under fault-free circuit conditions. The overall difference between the reference coupling rule and the benchmark coupling rule is quantified by comparing parameters one by one. This quantification method is a weighted summation of multi-dimensional parameter deviations. This embodiment presets a coupling change threshold, which is statistically calibrated based on the extreme values ​​of coupling feature offsets under historical fault conditions. When the difference between the reference rule and the benchmark rule exceeds the coupling change threshold, it is determined that the adjacent circuit fault evolution rate is faster and the disturbance is more significant. The traceability time window duration of the corresponding adjacent component circuit is automatically shortened, achieving adaptive optimization of the detection cycle and improving the accuracy of latent fault detection.

[0059] like Figure 3 As shown, when the final cause includes both prior shading and subsequent circuit disconnection, the following photovoltaic shading prediction and sensitive circuit control steps are performed: S30: Based on the hot spot location in the maintenance control information, i.e. the hot spot coordinates marked in the maintenance control information, the actual shading location of the photovoltaic panel is obtained by reverse analysis. Combined with the point change characteristics of continuous time-series thermal imaging data, the spatial movement trajectory of the shading area is fitted. The pre-trained forward evolution features of shadow shading in the spatiotemporal evolution model are retrieved. These forward evolution features of shadow shading are trained by a large number of time-series samples of shading faults and can characterize the evolution law of shadow with light, environment and time. Based on the forward evolution features of shadow shading, the diffusion range of the shading area in the next traceback time window is predicted and recorded as the first prediction range.

[0060] S31: Compare the first prediction range with the global component coordinate ledger, identify the component circuits whose coordinates fall completely within the prediction range, define them as aging sensitive circuits that are sensitive to shading disturbances, actively perform power-off disconnection operations on all aging sensitive circuits, and continuously wait for a complete traceability time window to reserve the fault evolution observation period.

[0061] S32: After the waiting period ends, the hot spot position obtained during the waiting process is used as the reference position. When the proportion of hot spot positions in the first prediction range in the reference position is greater than the preset accuracy threshold, and the vibration amplitude of the vibration data corresponding to each position in the first prediction range is less than the preset minimum vibration amplitude, the aging sensitive circuits are connected.

[0062] Specifically, after the waiting period ends, all hotspot locations continuously collected during the observation period are aggregated into a reference location set. In this step, maintenance personnel preset a prediction accuracy threshold and a minimum vibration amplitude threshold. The prediction accuracy threshold is used to determine the effectiveness of trajectory prediction; the minimum vibration amplitude threshold is used to determine that there is no external force disturbance on the panel surface. The proportion of hotspot locations within the first prediction range to all reference locations is counted. When the proportion is greater than the preset accuracy threshold, it proves that the shading trajectory prediction is accurate and effective. At the same time, the real-time vibration data amplitude corresponding to all points within the first prediction range is less than the preset minimum vibration amplitude, excluding external vibration interference such as wind and sand, equipment shaking, etc., and determining that the current shading state is stable and without secondary disturbances. After meeting both conditions, all previously disconnected aging sensitive circuits are automatically reconnected to complete fault verification and circuit reset.

[0063] The above are merely embodiments of the present invention. The invention is not limited to the fields covered by these embodiments. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are able to access all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A dynamic control method for photovoltaic modules, characterized in that, Includes the following steps: Multidimensional parameters are collected, including thermal imaging data of the photovoltaic panel surface, vibration data of each preset position point of the photovoltaic panel, and electrical data of the component circuit. Based on the collection time, the spatial coordinate system established with the fixed position and orientation of the photovoltaic panel, and the topological connection relationship of the component circuit, the multidimensional parameters are aligned in time and space to obtain spatiotemporal fusion data. From the spatiotemporal fusion data, the numerical evolution sequence of each multidimensional parameter within the preset abnormal parameter range on the time axis and the coupling law between different multidimensional parameters are extracted to construct a spatiotemporal evolution model containing the evolution characteristics corresponding to the causes of multiple hot spots. When any current multidimensional parameter is within the range of abnormal parameters, the multidimensional parameter sequence within the preset traceback time window is obtained as the actual observation sequence. In the spatiotemporal evolution model space, based on the current multidimensional parameters, the inversion path reflecting the formation process of potential hot spots is obtained by iterative analysis along the time axis according to the coupling law. The inversion path is then processed into a predicted preceding sequence. By performing time-point similarity feature residual calculation on the predicted preceding sequence and the actual observation sequence, the confidence value corresponding to the inversion path is obtained. Inversion paths with confidence values ​​exceeding a preset confidence threshold are selected as reference inversion paths. The hot spot genesis corresponding to the unique reference inversion path is taken as the final gene. Alternatively, the triggering order of each hot spot gene and the transmission time constant of the spatiotemporal evolution model are extracted from each reference inversion path. The evolution time difference is obtained by measuring the numerical variation of parameters in each dimension in the actual observation sequence. The numerical deviation between the transmission time constant and the evolution time difference is compared to verify the hierarchical induction relationship between the prior hot spot genesis and the subsequent hot spot genesis. The combination of hot spot genesis that satisfies the hierarchical induction relationship is determined as the final gene. Maintenance control information is generated based on the final cause, corresponding thermal imaging data, or component circuitry.

2. The photovoltaic module dynamic control method according to claim 1, characterized in that: All data segments within the preset anomaly parameter range are selected from the spatiotemporal fusion data, and numerical evolution sequences corresponding to thermal imaging data, vibration data, and electrical data are generated according to the acquisition time sequence. The correlation between different multidimensional parameters within the same time interval is calculated to obtain the coupling law between different multidimensional parameters. The numerical evolution sequence features and coupling law features corresponding to each preset hot spot cause are integrated to construct a spatiotemporal evolution model containing the evolution features corresponding to multiple hot spot causes.

3. The dynamic control method for photovoltaic modules according to claim 2, characterized in that: The multidimensional parameters corresponding to photovoltaic modules at the same acquisition time point and the same spatial coordinate position are integrated into multidimensional parameter integrated data after time and space alignment through a unified spatiotemporal coordinate system. In the spatiotemporal evolution model space, the current multidimensional parameter integrated data is used as the initial data set for reverse iteration. According to the coupling law, the change constraints of each multidimensional parameter within each time step are determined, where the time step is the acquisition sampling interval of the multidimensional parameters. The multidimensional parameter integrated data corresponding to each preceding time point is solved step by step along the time axis in reverse, generating at least one complete multidimensional parameter change trajectory as the inversion path; the multidimensional parameter integrated data corresponding to all preceding time points in the inversion path are arranged in chronological order and processed into a predicted preceding sequence.

4. The photovoltaic module dynamic control method according to claim 1, characterized in that: Thermal imaging data, vibration data, and electrical data under various preset hot spot causes were collected and, after temporal and spatial alignment, a historical training dataset was constructed for training the spatiotemporal evolution model. The feature residuals of thermal imaging data, vibration data, and electrical data in the predicted preceding sequence and the actual observation sequence at each corresponding acquisition time point were calculated. From the historical training dataset, the ratio of inter-class variance to intra-class variance of each multidimensional parameter under each hot spot cause is calculated to obtain the feature discrimination of each multidimensional parameter. The contribution of each multidimensional parameter to the identification of hot spot causes is determined based on the magnitude of feature discrimination. The contribution is then normalized to obtain the influence weight of each dimension of data. The feature residuals of each dimension of data are multiplied by their corresponding influence weights and then summed to obtain the comprehensive residual. The reciprocal of the comprehensive residual is normalized to obtain the confidence value corresponding to the inversion path.

5. The dynamic control method for photovoltaic modules according to claim 3 or 4, characterized in that: By comparing the integrated multidimensional parameter data corresponding to two adjacent preceding time points in the reference inversion path, the overall change amplitude is obtained. The comparison result of the overall change amplitude and the preset threshold is obtained, and the mutation point of the integrated multidimensional parameter data is determined according to the comparison result. Based on the mutation point of the integrated multidimensional parameter data, each reference inversion path is segmented and analyzed. The starting time point of each segment is determined as the initial perturbation occurrence time point of the corresponding hot spot cause. All hot spot causes are arranged in the order of the initial perturbation occurrence time points to obtain the triggering order of each hot spot cause. The time parameter corresponding to each hot spot cause is extracted from the spatiotemporal evolution model as the propagation time constant of the hot spot cause.

6. The dynamic control method for photovoltaic modules according to claim 5, characterized in that: From the actual observation sequence, the corresponding acquisition time when the change amplitude of the integrated data of multidimensional parameters corresponding to the cause of each hot spot exceeds the preset threshold of the change amplitude is extracted; the time interval between the acquisition times corresponding to two adjacent hot spot causes is calculated to obtain the evolution time difference; the difference between the transmission time constant corresponding to the pre-existing hot spot cause and the evolution time difference is obtained to obtain the numerical deviation; when the numerical deviation falls within the preset value range, it is determined that the pre-existing hot spot cause has a stratified induction relationship with the subsequent hot spot cause.

7. The dynamic control method for photovoltaic modules according to claim 1, characterized in that: The causes of hot spots include shielding, circuit disconnection, and high-resistance contact; it also includes the following steps: S10: Based on the multi-dimensional parameter integration data binding spatial coordinates, establish a one-to-one correspondence between the final cause, thermal imaging data, and component circuits for maintenance correlation. Based on the maintenance correlation, obtain the maintenance location and corresponding component circuit from the thermal imaging data, and generate maintenance control information; disconnect the component circuit in the maintenance control information; obtain the circuit with high resistance contact in the final cause of the disconnected component circuit as the test connection circuit; obtain the maintenance location corresponding to each test connection circuit as the test connection location from the maintenance correlation; and integrate adjacent test connection locations into the test connection range based on the coordinates of the test connection locations in the spatial coordinate system. S20: Obtain the test connection circuit corresponding to the outermost test connection position in the test connection range as the priority test connection circuit, connect the priority test connection circuit, wait for the duration corresponding to the preset traceability time window, if the priority test connection circuit is not disconnected after the wait ends, remove the test connection position corresponding to the priority test connection circuit from the test connection range; return to S20; if the priority test connection circuit is disconnected again after the wait ends, update the maintenance control information with the test connection circuit corresponding to the test connection position in the test connection range, the corresponding final cause and maintenance position.

8. The dynamic control method for photovoltaic modules according to claim 7, characterized in that: In step S20, a traceability time window is set for each disconnected component circuit; the transfer time constant of the adjacent component circuits of the disconnected component circuit during the maintenance process is obtained, and the transfer time constant is extracted from the spatiotemporal evolution model; when the final cause includes circuit disconnection or high resistance contact, the hot spot position corresponding to the final cause is obtained from the thermal imaging data, at least one adjacent component circuit that is spatially adjacent to the hot spot position is obtained, and the coupling law corresponding to the adjacent component circuit is extracted from the spatiotemporal evolution model as a reference law; A corresponding reference coupling rule is preset for each coupling rule. If the difference between the reference rule and the reference coupling rule exceeds the preset coupling change threshold, the traceability time window of the adjacent component circuit is shortened.

9. The dynamic control method for photovoltaic modules according to claim 8, characterized in that: When the final cause includes both pre-position blocking and subsequent circuit disconnection, perform the following steps: S30: Based on the hot spot location in the maintenance control information, analyze the occlusion location and its spatial movement trajectory corresponding to the occlusion. Based on the positive evolution characteristics of shadow occlusion in the spatiotemporal evolution model, predict the first prediction range of the occlusion location within the next traceback time window. S31: Identify component circuits whose hot spot locations are within the first prediction range as aging-sensitive circuits, disconnect the aging-sensitive circuits; wait for the duration corresponding to a traceability time window; S32: After the waiting period ends, the hot spot position obtained during the waiting process is used as the reference position. When the proportion of hot spot positions in the first prediction range in the reference position is greater than the preset accuracy threshold, and the vibration amplitude of the vibration data corresponding to each position in the first prediction range is less than the preset minimum vibration amplitude, the aging sensitive circuits are connected.

10. A dynamic control system for photovoltaic modules, characterized in that, The photovoltaic module dynamic control method according to any one of claims 1-9 was used.

Images