A distributed photovoltaic intelligent operation and maintenance system

Abstract

The present invention relates to the field of photovoltaic intelligent operation and maintenance technology, and discloses a distributed photovoltaic intelligent operation and maintenance system. The method collects operating parameters such as voltage output, current intensity, temperature, and ambient light of multiple photovoltaic units in real time; then, based on these parameters, the performance status of each photovoltaic unit is classified to identify high-efficiency operation areas, performance degradation areas, and potential fault areas; then, rules are extracted from the maintenance strategy database to construct a mapping relationship between performance partitions and maintenance strategies; then, key partitions are determined based on partition priority weights and system influencing factors, and the matching degree of maintenance strategy combinations is optimized; then, the real-time operating status of the system is checked based on the optimized combination, maintenance demand changes are identified, and the mapping relationship is updated; finally, the maintenance goals and execution order are determined based on the updated mapping relationship, and a structured operation and maintenance management library is generated to guide subsequent maintenance activities. This method can improve operation and maintenance efficiency and accuracy, and ensure stable and efficient operation of the photovoltaic system.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic intelligent operation and maintenance, and specifically to a distributed photovoltaic intelligent operation and maintenance system. Background Art

[0002] Distributed photovoltaic systems, thanks to their localized consumption and reduced line losses, are increasingly being used in energy transitions. However, their operation and maintenance (O&M) management presents numerous challenges. Distributed photovoltaic units are typically deployed on residential rooftops, industrial parks, public buildings, and other locations. These geographically dispersed locations and diverse environmental conditions make traditional O&M models difficult to effectively cover.

[0003] Traditional operations and maintenance rely heavily on regular manual inspections, which not only consume significant manpower and resources but also have fixed inspection cycles, making it difficult to detect operational anomalies in PV cells in real time. While some areas have introduced simple monitoring devices, the data collected is often limited to a single parameter, such as voltage or current. It fails to integrate environmental factors like temperature and light with the device's own status, resulting in a one-sided assessment of PV cell performance.

[0004] Existing performance evaluation methods often use a holistic approach, treating the entire PV system as a single entity. This makes it difficult to distinguish the actual status of different units. In reality, PV units within the same system can exhibit varying states of high efficiency, performance degradation, or potential failure due to differences in component aging, dust coverage, and shadowing. Holistic evaluations fail to accurately pinpoint problematic units, leading to imbalanced allocation of maintenance resources.

[0005] Maintenance strategies also face significant shortcomings. Current maintenance plans are often based on empirical assumptions and lack a dynamic connection to the actual performance of PV units. For example, areas of declining performance are treated with the same maintenance strategy as those operating at high efficiency, or potential failure areas are addressed with a delayed approach. This leads to a disconnect between maintenance measures and actual needs.

[0006] Existing O&M systems lack a precise mechanism for identifying critical areas. Different PV units in a system have varying impacts on overall operation. While some units may be experiencing performance degradation, their impact on the overall system output is minimal. However, some potentially faulty units may directly threaten system stability. Failure to distinguish critical zones based on priority and impact factors can easily lead to delayed maintenance of important units and over-maintenance of less important units, impacting the economic efficiency and stability of system operation. Furthermore, once maintenance strategies are established, they remain fixed for a long time, making it difficult to dynamically adjust to changes in PV unit performance, further exacerbating O&M limitations. Summary of the Invention

[0007] The purpose of the present invention is to provide a distributed photovoltaic intelligent operation and maintenance system to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides a distributed photovoltaic intelligent operation and maintenance system, the method comprising:

[0009] Data collection step: real-time acquisition of operating parameters of multiple photovoltaic units in the distributed photovoltaic system, the operating parameters including voltage output data, current intensity data, temperature measurement data, and ambient light data;

[0010] Partition analysis step: based on the operating parameters, classifying the performance status of each photovoltaic unit to identify performance partitions, wherein the performance partitions include a high-efficiency operation zone, a performance degradation zone, and a potential failure zone;

[0011] Mapping construction step: extracting maintenance rule information from a maintenance policy database according to the performance partitions, and constructing a mapping relationship between the performance partitions and maintenance policies;

[0012] Combination optimization step: extracting key partitions from the performance partitions, the key partitions being determined based on partition priority weights and system impact factors, and optimizing the combination matching degree of the maintenance strategies;

[0013] Status update step: Based on the optimized maintenance strategy combination, the real-time operating status of the distributed photovoltaic system is verified, changes in maintenance requirements are identified, and the mapping relationship between the performance partitions and the maintenance strategies is updated;

[0014] Management output step: Based on the updated mapping relationship, determine the objectives and execution sequence of the maintenance operations, and generate a structured operation and maintenance management library to guide subsequent maintenance activities.

[0015] Preferably, the data collection step is implemented by:

[0016] For any photovoltaic unit in the distributed photovoltaic system, obtaining monitoring sensor data corresponding to the operating parameters;

[0017] Using a data preprocessing model to perform noise filtering and outlier removal on the monitoring sensor data;

[0018] Identifying key parameter change trends and related parameter association features in the monitoring sensor data to form a standardized data sample set;

[0019] The key parameter change trends and related parameter association characteristics in the standardized data sample set are analyzed respectively, and the trend analysis area, change analysis area and association characteristic area corresponding to the standardized data sample set are obtained in sequence.

[0020] Preferably, the implementation method of obtaining the trend analysis area, the change analysis area and the associated feature area corresponding to the standardized data sample set further includes:

[0021] Merging the key parameter change trends and related parameter association features in the standardized data sample set according to the photovoltaic unit identification to obtain multiple merged result sets;

[0022] extracting trend matching pairs from the merged result set, and comparing the trend matching pairs with a historical operation database to obtain a trend analysis area;

[0023] Extracting the change rate threshold of the change analysis feature and the correlation strength value of the association feature in the merged result set, dividing the merged result set according to the change rate threshold and the correlation strength value to obtain the change analysis area and the association feature area.

[0024] Preferably, the implementation method of classifying the performance status of each photovoltaic unit in the partition analysis step further includes:

[0025] Priority evaluation is performed on the performance partitions, the operating time data and fault frequency data of the photovoltaic units in the performance partitions are analyzed, the performance partitions are fitted according to the operating time data and fault frequency data, and a mapping relationship between the performance partitions and actual maintenance needs is constructed.

[0026] Preferably, the mapping construction step is implemented by:

[0027] Retrieving maintenance rule information and historical maintenance case data in the maintenance policy database to generate a plurality of unlabeled maintenance policy identification results, wherein the unlabeled maintenance policy identification results represent maintenance rule information and historical maintenance case data that are not associated with parameter features in the performance partition;

[0028] It is determined whether the plurality of unlabeled maintenance strategy identification results are valid maintenance strategy identification results. If they are valid maintenance strategy identification results, the valid maintenance strategy identification results are regarded as target maintenance strategies in the maintenance strategy database.

[0029] Preferably, the implementation of constructing the mapping relationship between the performance partitions and the maintenance strategies includes:

[0030] A mapping relationship is established between the performance partition and the maintenance strategy database using the information representation of the trend analysis area, the change parsing area and the associated feature area in the performance partition, the description information of the maintenance strategy database and the maintenance strategy category.

[0031] Preferably, the implementation of the combination optimization step includes:

[0032] Performing cluster evaluation processing on the performance partition and the maintenance policy database according to the maintenance type, policy priority and policy function, and setting the core cluster center after the cluster evaluation processing as the key partition;

[0033] Extracting core feature items of the key partitions, evaluating similarity matching between the core feature items, and setting common sequence information related to the similarity matching between the core feature items;

[0034] Using the public sequence information related to the similarity matching between the core feature items, the maintenance strategy items present in the public sequence information are extracted, and the longest matching sequence between the maintenance strategy items is set, and the matching length value of the longest matching sequence is set as the combined matching degree between the maintenance strategy items;

[0035] The maintenance strategy bias combination is set based on the combination matching degree between the maintenance strategy items and the time distribution characteristics of the maintenance strategy items.

[0036] Preferably, the implementation of the status updating step includes:

[0037] Extracting the time distribution characteristics of each maintenance policy item from the maintenance policy bias combination; setting the target maintenance path of the maintenance policy bias combination according to the time period information corresponding to the time distribution characteristics of each maintenance policy item;

[0038] Fitting the target maintenance path of each maintenance strategy item in the maintenance strategy bias combination to obtain a fitted target maintenance path, and setting the occurrence probability value of the fitted target maintenance path in each time period information as the combined occurrence probability of the maintenance strategy bias combination;

[0039] The occurrence probability of the maintenance strategy bias combination is compared with the historical maintenance case references in the maintenance strategy database, the difference value changes are identified, and the historical maintenance cases in the maintenance strategy database are classified and updated according to the difference value changes.

[0040] Preferably, the management output step is implemented as follows:

[0041] Based on the updated mapping relationship between the performance partitions and the maintenance strategies, maintenance target information in the distributed photovoltaic system is extracted. The maintenance target information is sorted according to the occurrence frequency characteristics of the maintenance target information to obtain the target and execution sequence of the maintenance operation. The maintenance target information and the execution sequence are combined in a structured storage format to obtain the operation and maintenance management library.

[0042] Preferably, the method further comprises:

[0043] Data interaction step: in the state updating step, the updated mapping relationship data is transferred to the management output step;

[0044] In the management output step, a feedback signal is generated using the updated mapping relationship data, and the feedback signal is transmitted to the data collection step to start a new round of data collection;

[0045] Through the interaction of the feedback signal and the real-time operating parameters, closed-loop operation and maintenance management of the distributed photovoltaic system is achieved.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] This distributed photovoltaic intelligent operation and maintenance method achieves comprehensive awareness of the operating status of distributed photovoltaic systems through multi-dimensional data collection. Real-time data on voltage output, current intensity, temperature measurement, and ambient light intensity captures both the operating characteristics of the photovoltaic units themselves and external environmental factors. This overcomes the limitations of traditional monitoring, which relies on single, incomplete data, and enables operation and maintenance activities to be based on complete status information.

[0048] Categorizing performance status enables refined management of photovoltaic cells. By identifying high-efficiency operating zones, performance degradation zones, and potential failure zones, the previously ambiguous system status is transformed into clear zoning results, enabling operators to intuitively understand the performance differences between different cells. This zoning approach avoids overly broad judgments about the system as a whole, allowing cells in different states to receive targeted attention and reducing operational deviations caused by misjudgment of status.

[0049] The mapping relationship between maintenance strategies and performance zones allows maintenance measures to be tailored to the characteristics of each zone. Rules extracted from the maintenance strategy database are aligned with the performance characteristics of each zone, changing the traditional maintenance practice of rigid strategies that are disconnected from actual conditions. Different maintenance rules correspond to different zones, ensuring that areas of performance degradation receive appropriate intervention measures, potential failure areas receive appropriate preventive measures, and efficient operation areas maintain appropriate attention, ensuring that maintenance activities are more aligned with actual needs.

[0050] The extraction of key partitions and the combined optimization of maintenance strategies further enhance the targeted nature of operations and maintenance. Key partitions, determined based on partition priority weights and system impact factors, focus on units with significant impact on system operations, avoiding ineffective investment of maintenance resources in non-critical areas. Optimizing the matching degree of maintenance strategy combinations enables multiple strategies to coordinate with each other, adapting to the complex conditions of key partitions and reducing the limitations of a single strategy in complex scenarios.

[0051] The status update step enables dynamic adjustment of operations and maintenance management. By continuously checking the system's real-time operating status, changes in maintenance requirements can be promptly identified, and the mapping between performance zones and maintenance strategies can be updated. This dynamic adjustment mechanism can address the natural degradation of PV unit performance over time and state fluctuations caused by environmental changes. It avoids the inapplicability caused by long-term fixed strategies and ensures that maintenance strategies are always synchronized with the actual system status.

[0052] The generation of a structured O&M management library provides clear action guidelines for maintenance activities. Defined maintenance operation objectives and execution sequences transform abstract strategies into concrete action plans, enabling O&M personnel to carry out their work in a sequential manner and reducing blind spots during the maintenance process. This structured management approach integrates the results of zoning information, strategy matching, and dynamic adjustment to form a coherent O&M system that covers the entire process from status perception to action execution, making the O&M management of distributed PV systems more organized and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a working principle diagram of the distributed photovoltaic intelligent operation and maintenance system according to the present invention;

[0054] Figure 2 Flowchart for partitioning the standardized data sample set;

[0055] Figure 3 Flowchart of classification processing for partition analysis steps;

[0056] Figure 4 A flowchart of the status update steps. DETAILED DESCRIPTION

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] See also Figure 1 The present invention provides a distributed photovoltaic intelligent operation and maintenance system, the method comprising:

[0059] Data collection step, partition analysis step, mapping construction step, combinatorial optimization step, state update step and management output step. Specifically, it includes the following steps:

[0060] In the data collection step, the system obtains the operating parameters of multiple photovoltaic units in the distributed photovoltaic system in real time. The operating parameters include voltage output data, current intensity data, temperature measurement data and ambient light data. These data are collected by sensors deployed on each photovoltaic unit to ensure the continuity and integrity of the data.

[0061] The zoning analysis step classifies the performance status of each PV unit based on the collected operating parameters. By setting performance indicator thresholds, the PV units are divided into high-efficiency operation areas, performance degradation areas, and potential fault areas, thereby achieving a preliminary assessment of the overall operating status of the system.

[0062] Based on the above performance partitioning results, the mapping construction step extracts maintenance rule information that matches the characteristics of each partition from the maintenance strategy database, establishes a mapping relationship between performance partitions and maintenance strategies, and provides structured support for subsequent decision-making.

[0063] The combinatorial optimization step extracts key partitions from the identified performance partitions. The determination of key partitions comprehensively considers the partition priority weights and the factors affecting the overall operation of the system. The optimization algorithm is used to improve the combined matching between the maintenance strategy and the key partitions.

[0064] The status update step verifies the real-time operating status of the distributed photovoltaic system based on the optimized maintenance strategy combination, identifies changes in maintenance requirements due to factors such as environmental changes or equipment aging, and dynamically updates the mapping relationship between performance partitions and maintenance strategies to maintain the timeliness of system response.

[0065] The management output step determines the target objects and execution order of specific maintenance operations based on the updated mapping relationship, and generates a structured operation and maintenance management library. The library contains information such as maintenance tasks, execution time, and required resources, which is used to guide on-site maintenance personnel to carry out operations and realize the intelligence and standardization of the operation and maintenance process.

[0066] Example 1: See Figure 2 For each photovoltaic unit in a distributed photovoltaic system, monitoring sensor data corresponding to its operating parameters is first acquired. These operating parameters include voltage output data, current intensity data, temperature measurement data, and ambient light data. Each type of data is collected in real time by the corresponding sensor deployed on the photovoltaic unit. The sensor data collection frequency is set according to the operating characteristics of the photovoltaic system to ensure that changes in key operating conditions are captured. The collected raw data may contain noise interference or abnormal fluctuations, so preprocessing is required to improve data quality.

[0067] The data preprocessing model is used to filter noise and remove outliers from monitoring sensor data. This model is based on a variety of data processing techniques, such as sliding window filtering, wavelet transforms, and statistically based outlier detection methods. These methods effectively remove random noise from the data while identifying and removing outliers caused by sensor failures or external interference. During this process, the main trends in the data are preserved, avoiding the loss of critical information due to oversmoothing.

[0068] After preprocessing, the system identifies trends in key parameters within the sensor data. For example, voltage output may show a slow decline over time, potentially related to module aging; current intensity may fluctuate periodically with changes in light intensity, reflecting the system's response to the external environment. The system also analyzes correlations between different parameters, such as the negative correlation between temperature and output power, or the positive correlation between light intensity and current output. These trends and correlations together provide crucial insights into the operating status of the PV cell.

[0069] Based on the above analysis results, the system converts the processed data into a standardized data sample set. The standardization process includes steps such as data format unification, dimensional normalization, and timestamp alignment to ensure that all data is comparable and consistent in subsequent analysis. The standardized data sample set includes information on the changing trends of key parameters and the correlation characteristics between related parameters, providing a data foundation for subsequent performance partitioning.

[0070] The system analyzes the key parameter trends and associated features within the standardized data sample set. During the analysis, trend information and associated features associated with the same PV unit are merged to generate multiple merged result sets. Each merged result set contains the operating status characteristics of the PV unit over a specific time period, facilitating subsequent trend matching and feature segmentation.

[0071] During trend matching, the system extracts trend matching pairs from the merged result set. These trend matching pairs represent the changing patterns of the PV unit's operating parameters. For example, if the voltage output of a PV unit shows a downward trend over several consecutive days, this trend pattern may match a known fault pattern. These extracted trend matching pairs are compared with records in the historical operation database to identify trend patterns that match known operating conditions, thereby generating trend analysis areas. The identification of trend analysis areas can be used to determine whether the current operating status is within the normal range or whether early signs of performance degradation have occurred.

[0072] During the feature segmentation process, the system further extracts change analysis features and correlation features from the merged result set. Change analysis features focus on the rate of change of parameters, such as the magnitude of change in current intensity per unit time. The system sets a change rate threshold. When the rate of change of a parameter exceeds this threshold, it is placed in the change analysis area. The identification of change analysis areas helps to detect sudden anomalies, such as inverter failure or component damage. At the same time, the system also extracts correlation features from the merged result set and analyzes the correlation strength values ​​between different parameters. For example, whether the correlation between light intensity and current output remains stable. If the correlation drops significantly, it may indicate that there is shading or contamination on the component surface. The system sets a correlation strength threshold. When the correlation is lower than this threshold, the data is placed in the correlation feature area. The identification of the correlation feature area helps to detect whether there are any abnormalities in the coordinated change relationship between the parameters within the system.

[0073] The results of the trend analysis, variation analysis, and correlation characteristic zones serve as important input for subsequent performance zoning. The trend analysis zone reflects the long-term trends in the PV unit's operating parameters, providing a macroscopic perspective for performance evaluation. The variation analysis zone captures short-term fluctuations in operating parameters, revealing potential unexpected issues. The correlation characteristic zone, based on the interrelationships between parameters, identifies any issues with system-wide synergy degradation. The combined analysis of these three zones provides a multi-dimensional description of the PV unit's operating status, providing data support for subsequent performance status classification and maintenance strategy development.

[0074] Throughout the data acquisition and preprocessing process, the system adopts a modular design, ensuring that each processing step operates independently and can be flexibly adjusted. The data acquisition module is responsible for acquiring and initially organizing sensor data. The data preprocessing module performs noise filtering and outlier removal. The trend identification module extracts changing trends in key parameters. The feature analysis module identifies correlations between parameters. The merging processing module integrates trends and correlation features. The trend matching module compares data with the historical operation database. The feature segmentation module completes the division of change analysis areas and correlation feature areas. Data is transferred between modules through standardized interfaces, ensuring the consistency of the processing flow and the integrity of the data.

[0075] Furthermore, the system incorporates a timestamp mechanism during data collection and processing, ensuring that all data is accurately time-stamped, facilitating subsequent time series analysis and trend tracking. This timestamp mechanism also enables horizontal comparison of data across different PV units, identifying any regional operational anomalies or environmental factors. The system also features data caching and breakpoint-resume functions, ensuring that critical data is retained in the event of network outages or equipment failures, preventing data loss.

[0076] Example 2: See Figure 3The zoning analysis step classifies the performance status of each PV unit based not only on real-time operating parameter values ​​but also incorporates operating time data and fault frequency data as auxiliary evaluation factors. Operating time data reflects the cumulative operating time of the PV unit and is an important indicator for assessing component degradation. Fault frequency data records the number of faults that occur per unit time, reflecting the stability of equipment operation. By incorporating this data into the performance zoning evaluation system, a more comprehensive assessment of the PV unit's operating status can be achieved.

[0077] During implementation, the system first prioritizes high-efficiency operating zones, performance degradation zones, and potential failure zones. High-efficiency operating zones typically consist of PV units with stable operating parameters, small fluctuations, and consistent with design expectations. PV units in performance degradation zones exhibit characteristics such as decreased output power, increased temperature, or delayed response. PV units in potential failure zones may not yet have obvious failures, but their operating parameter trends have deviated from normal ranges or have a high historical failure frequency. The system then fits these zones based on operating time and failure frequency to establish a mapping between performance zones and actual maintenance needs.

[0078] The construction of the mapping relationship adopts a multivariate fitting model, which takes the running time and the frequency of failure as input variables, and the output is the maintenance demand level The model expression is as follows:

[0079] ;

[0080] in, Indicates the maintenance requirement level, Indicates the cumulative operating time of the photovoltaic unit (in hours), Indicates the number of faults that occur per unit time (in times / month). These coefficients are obtained through historical data training and reflect the weight of the impact of operating time and failure frequency on maintenance demand. Maintenance demand level The value range of is set as a discrete interval to divide the maintenance priorities into different levels. A higher value indicates that the PV unit needs to be maintained or replaced first.

[0081] Through this model processing, the system can dynamically evaluate performance zones and adjust maintenance strategies accordingly. For example, a PV unit with high-efficiency parameters but a long operating life and a high historical failure rate may be classified as requiring high maintenance, triggering more frequent inspections or preventive maintenance. Conversely, a PV unit with large fluctuations in operating parameters but a short operating life and a low historical failure rate may be considered a temporary fluctuation and not be considered a priority maintenance target.

[0082] During the mapping construction step, the system accesses maintenance rule information and historical maintenance case data from the maintenance policy database to generate multiple unlabeled maintenance policy identification results. These identification results represent maintenance rule information and historical maintenance case data that have not yet been associated with the current performance partition parameter characteristics. Maintenance rule information includes descriptive information such as the policy name, applicable conditions, and execution method; historical maintenance case data records information such as the execution time, operation content, and result feedback of previous maintenance activities.

[0083] The system determines the validity of these unlabeled maintenance strategy identification results. This determination is based on multiple dimensions, including whether the strategy's scope of application matches the characteristics of the current performance partition, the number of successful applications of the strategy in historical maintenance cases, and the similarity between the strategy and the parameter characteristics of the current operating state. This similarity is calculated using a feature vector-based matching method, comparing the parameter characteristics of the current performance partition with those in historical cases to obtain a matching score.

[0084] If the judgment result is valid, the maintenance strategy identification result is considered the target maintenance strategy in the maintenance strategy database. The target maintenance strategy is then included in the candidate set for subsequent mapping relationship construction and associated with the corresponding performance partition. This process ensures that the rule information in the maintenance strategy database can be dynamically updated to adapt to new failure modes and maintenance experience, enhancing the system's adaptive capabilities.

[0085] In actual operation, the system continuously updates the maintenance policy database, adding newly generated valid maintenance policies to the database while archiving or marking historical policies that have not been used for a long time or have a low matching degree. The database update process includes steps such as data cleansing, rule optimization, and version control to ensure the accuracy and timeliness of database content. Data cleansing is used to remove redundant or invalid maintenance policies; rule optimization uses machine learning or expert systems to adjust parameters or optimize categories of existing policies; and version control records the content and time of each update for easy traceability and analysis.

[0086] The maintenance policy database update mechanism, combined with the dynamic evaluation of performance partitions, enables the system to continuously accumulate operational experience and automatically adjust maintenance policy matching relationships based on changes in equipment operating status. This adaptive mechanism not only improves the accuracy of operational decision-making but also enhances the system's adaptability to complex operating environments.

[0087] Throughout the partition analysis and mapping process, the system employs a modular architecture, with standardized interfaces enabling data exchange between functional modules. The partition assessment module conducts a comprehensive analysis of operating parameters, operating time, and failure frequency; the model fitting module performs multivariate fitting calculations to generate maintenance requirement levels; the maintenance strategy identification module extracts and evaluates maintenance strategies from the database; the effectiveness judgment module selects strategies based on multidimensional information; and the database update module is responsible for dynamic maintenance and version management of the maintenance strategy database. These modules work collaboratively to ensure efficient operation and data consistency throughout the entire process.

[0088] The system incorporates a timestamp mechanism during processing, ensuring that all partition assessment results, maintenance strategy identification results, and database update records are accurately timed, facilitating subsequent version backtracking and trend analysis. Furthermore, the system supports multi-level permission control to ensure the security and controllability of database update operations.

[0089] Example 3: The mapping construction step is based on the information of the trend analysis area, change analysis area and associated feature area in the performance partition, combined with the descriptive information of the maintenance strategy database and the maintenance strategy category, to construct a mapping relationship between the performance partition and the maintenance strategy. The trend analysis area reflects the long-term change trend of the operating parameters of the photovoltaic unit, the change analysis area captures the severity of the parameter change, and the associated feature area reflects the coordinated change relationship between different parameters. The maintenance strategy database contains various types of maintenance strategies, such as preventive maintenance, corrective maintenance, predictive maintenance, etc. Each type of strategy has different applicable conditions, execution methods and expected effects. The system establishes a mapping relationship between the performance partition characteristics and the maintenance strategy description information by analyzing the matching degree between the two, providing a basis for subsequent maintenance decisions.

[0090] During implementation, the system first extracts information from the trend analysis area, change resolution area, and associated feature area within the performance partition and converts it into feature vectors for matching. These feature vectors are composed of multiple dimensions, including trend change direction, change rate, fluctuation amplitude, and correlation strength. Simultaneously, the system extracts policy description information from the maintenance policy database, including policy name, applicable conditions, execution method, and historical application records, and converts this information into policy feature vectors. The policy feature vectors share the same dimensional structure as the performance partition feature vectors, facilitating similarity comparisons.

[0091] In order to measure the matching degree between performance partition characteristics and maintenance strategies, the system uses cosine similarity as the matching degree calculation method. The calculation formula is as follows:

[0092] ;

[0093] in, Represents the performance partition feature vector and maintenance strategy feature vector The similarity value between them; Represents the dot product of two vectors; and Represents the modulus of the two vectors. This formula calculates the cosine of the angle between two eigenvectors, with values ​​ranging from -1 to 1. Values ​​closer to 1 indicate greater similarity. This method allows the system to identify the maintenance strategy that best matches the current performance partition characteristics and incorporate it into the mapping.

[0094] During the combinatorial optimization step, the system clusters and evaluates the performance partitions and maintenance policy database based on maintenance type, policy priority, and policy function. This clustering evaluation is based on multidimensional characteristics, including the maintenance policy's execution frequency, scope of application, impact, and resource consumption. The system uses clustering algorithms, such as K-Means clustering or hierarchical clustering, to classify performance partitions and maintenance policies with similar characteristics. The core cluster centers after the clustering evaluation are designated as key partitions, representing the areas of the system requiring the most attention for maintenance.

[0095] The system further extracts the core feature items of the key partitions and evaluates the similarity matching between each core feature item. Core feature items include key parameter change patterns, fault types, impact scope, maintenance costs, etc. The similarity matching is also calculated using the cosine similarity method, comparing the feature vectors between different core feature items to obtain the degree of matching between them. Based on the similarity matching, the system sets common sequence information, which represents the maintenance requirement pattern shared by multiple core feature items. Common sequence information is composed of the intersection of multiple feature items and is used to identify common requirements between a set of maintenance strategies.

[0096] Based on the common sequence information, the system extracts the maintenance policy items present and determines the longest matching sequence between these maintenance policy items. The longest matching sequence is a sequence of feature items that can be continuously matched across multiple maintenance policy items. Its length represents the combined matching degree. The combined matching degree reflects the degree of synergy between maintenance policies. The higher the matching degree, the more coordinated the maintenance policies are in terms of function and execution. The system uses the length of the longest matching sequence as an indicator of the combined matching degree between maintenance policies and optimizes the combination of maintenance policies based on this indicator.

[0097] During the optimization process, the system considers the temporal distribution characteristics of each maintenance strategy item to determine a preferred combination of maintenance strategies. These characteristics include the optimal time period for strategy execution, its duration, and its temporal dependencies with other strategies. For example, some maintenance strategies are best executed during periods of low sunlight to minimize the impact on power generation efficiency, while others require low system loads to mitigate operational risks. Based on these temporal characteristics, the system categorizes maintenance strategies into different time groups and sets preferred combinations to ensure consistent and coordinated maintenance activities.

[0098] The system incorporates a weight adjustment mechanism into the combinatorial optimization process, dynamically adjusting the execution priority of each maintenance strategy based on the combination matching and time distribution characteristics. This weight adjustment mechanism includes strategy prioritization, resource allocation optimization, and execution order adjustment. Strategy prioritization prioritizes strategies with high matching and appropriate execution times based on combination matching and time distribution characteristics. Resource allocation optimization rationally arranges maintenance personnel and equipment based on the type and quantity of resources required by each strategy. Execution order adjustment ensures that time conflicts between strategies do not affect execution efficiency.

[0099] Throughout the mapping construction and combinatorial optimization process, the system adopts a modular design, with standardized interfaces used to exchange data between functional modules. The mapping construction module is responsible for calculating the matching between performance partitions and maintenance strategies; the feature extraction module extracts the core features of performance partitions and maintenance strategies; the similarity calculation module performs cosine similarity calculations; the clustering evaluation module classifies performance partitions and maintenance strategies; and the combinatorial optimization module is responsible for calculating the matching between maintenance strategies and setting biased combinations. These modules work together to ensure efficient operation and data consistency throughout the entire process.

[0100] The system introduces a timestamp mechanism during processing to ensure that all mapping relationships and optimization results are accurately timed, facilitating subsequent version backtracking and trend analysis. Furthermore, the system supports multi-level permission control to ensure the security and controllability of mapping construction and combined optimization operations.

[0101] Example 4: See Figure 4The state update step extracts the time distribution characteristics of each maintenance strategy item from the maintenance strategy bias combination obtained in the combination optimization step. These time distribution characteristics include the optimal time period for strategy execution, execution frequency, duration, and time dependency with other strategies. Based on these characteristics, the system sets the target maintenance path for the maintenance strategy bias combination. The target maintenance path represents a set of maintenance tasks that should be performed within a specific time period and their sequential arrangement. In order to further improve the adaptability of the maintenance plan, the system fits the target maintenance path of each maintenance strategy item and generates a fitted target maintenance path to reflect the possibility of executing the maintenance combination within different time periods.

[0102] The system analyzes the probability of occurrence of the fitted target maintenance path in each time period to determine the probability of occurrence of the maintenance strategy-biased combination. The probability of occurrence reflects the likelihood of executing the maintenance combination within a certain time period and is used to measure the degree of match between the maintenance plan and historical maintenance practices. The system compares the probability of occurrence of the combination with the historical maintenance case citations in the maintenance strategy database to identify discrepancies between the current maintenance plan and historical practices. Changes in the discrepancy value include combinations with a higher or lower frequency of occurrence than the historical citation frequency, as well as combinations with significant deviations from historical patterns in certain time periods.

[0103] Based on changes in variance values, the system categorizes and updates historical maintenance cases in the maintenance strategy database. This update process includes marking high-variance cases, adjusting the weights of low-matching strategies, and supplementing records of newly emerging maintenance requirements. Marked high-variance cases are further analyzed by operations and maintenance managers to determine whether adjustments to the maintenance strategy's execution methods or applicable conditions are necessary. Adjusting the weights of low-matching strategies helps optimize the generation of subsequent maintenance plans, making them more aligned with actual operating conditions. Newly emerging maintenance requirements are recorded in the database as a reference for future maintenance strategy optimization.

[0104] During implementation, the system updates the maintenance strategy bias combination for a photovoltaic power station. The power station contains multiple photovoltaic units, each with different operating conditions and maintenance requirements. Based on the maintenance strategy bias combination generated by the combinatorial optimization step, the system extracts the maintenance strategy items and their time distribution characteristics and sets a target maintenance path. For example, for a maintenance strategy for a specific inverter module, the system identifies that its optimal execution time is during the night when grid load is low, with a monthly execution frequency of approximately two hours. Based on this information, the system schedules the maintenance strategy for execution between 11:00 PM and 1:00 AM each night and incorporates it into the target maintenance path.

[0105] The system further fits the target maintenance paths for all maintenance strategy items, generates a fitted target maintenance path, and calculates its probability of occurrence in each time period. For example, the probability of a maintenance strategy combination occurring between 00:00 and 01:00 is 0.85, between 01:00 and 02:00 is 0.62, and between 02:00 and 03:00 is 0.41. These probabilities are used to assess the rationality of the maintenance combination in different time periods and compared with execution records in historical maintenance cases.

[0106] To more intuitively compare the execution of maintenance strategies during status updates with historical data, the system constructs the following data table, which shows the comparison between the combined occurrence probability of the five maintenance strategy items in different time periods and the historical reference frequency:

[0107] ;

[0108] As shown in the table above, the system compares the probability of occurrence of each maintenance strategy item in different time periods with its historical reference frequency, identifying changes in variance. For example, the probability of occurrence of the M003 strategy item between 00:00 and 01:00 is 0.62, while its historical reference frequency is only 0.45, indicating that the strategy's execution frequency during this time period has increased compared to historical practice. The probability of occurrence of the M004 strategy item between 01:00 and 02:00 is 0.41, significantly lower than its historical reference frequency of 0.58, suggesting that the maintenance schedule for this time period may not have fully considered historical experience.

[0109] Based on these variance changes, the system categorizes and updates the relevant maintenance cases in the maintenance strategy database. For example, for strategy M003, the system marks its execution records between 00:00 and 01:00 as high-variance cases and recommends that operations managers assess the feasibility of maintenance during that time period. For strategy M004, the system adjusts its execution weight between 01:00 and 02:00 to align it more closely with historical practice in subsequent maintenance plans. For strategy M005, since its combined probability of occurrence is close to its historical reference frequency, the system maintains its current execution schedule.

[0110] The system utilizes a modular design throughout the entire status update process, ensuring that each processing step operates independently and can be flexibly adjusted. The target path setting module extracts the time distribution characteristics of maintenance strategies and sets target maintenance paths; the path fitting module integrates the target paths for each maintenance strategy; the probability calculation module calculates the probability of occurrence of combinations and compares it with historical data; the variance analysis module identifies changes in variance values; and the database update module categorizes and updates cases in the maintenance strategy database. Data is transferred between modules via standardized interfaces, ensuring process consistency and data integrity.

[0111] The system introduces a timestamp mechanism during status updates to ensure that all update records have accurate time stamps, facilitating subsequent version backtracking and trend analysis. Furthermore, the system supports multi-level permission control to ensure the security and controllability of status update operations.

[0112] Example 5: In Example 5, the management output step extracts the maintenance target information in the distributed photovoltaic system based on the mapping relationship between the performance partitions and the maintenance strategies that are dynamically adjusted in the status update step. These maintenance target information include the number of the photovoltaic unit, the current performance partition affiliation, the abnormal characteristics of the operating parameters, the type of maintenance strategy recommended for execution, and the maintenance priority. Based on this information, the system comprehensively ranks the maintenance needs of all photovoltaic units. The ranking is based on the frequency characteristics of the maintenance target information, the severity of the performance partition, and the impact of the unit on the overall power generation efficiency. The frequency characteristics refer to the number of times a certain type of maintenance target is identified by the system as requiring processing within a certain period. This characteristic reflects the continuity and urgency of the maintenance needs.

[0113] During the sorting process, the system considers not only the operating status of individual PV units but also the environmental factors of the area in which they are located and the interdependencies between power generation units. For example, if a PV unit is operating in a performance degradation zone but is located in an area with poor lighting conditions, resulting in chronically low power generation efficiency, its maintenance priority may be appropriately increased. Conversely, if a unit's operating parameters experience short-term fluctuations, but the fluctuations are within normal thresholds and historical records show that these fluctuations are cyclical, its maintenance priority may be appropriately lowered. This dynamic sorting mechanism ensures the rationality and targeted nature of maintenance operations.

[0114] After determining the maintenance operation's objectives and execution sequence, the system combines these information into a pre-defined structured storage format to generate an operations and maintenance management library. This library uses standardized data formats, such as JSON or XML, and contains fields such as the maintenance task number, target unit number, recommended maintenance strategy, execution time window, required resource type, and expected execution duration. This library can be directly accessed by the operations and maintenance management system to generate work orders, allocate maintenance resources, schedule maintenance personnel, and interact with the scheduling system to ensure efficient execution of maintenance tasks.

[0115] The data interaction step establishes a closed-loop connection between the status update step and the management output step. The updated mapping relationship data generated by the status update step is passed to the management output step as input for generating the latest operation and maintenance management library. The management output step uses this data to generate a feedback signal containing information such as system status changes, maintenance plan adjustments, and policy matching updates. This feedback signal is then passed to the data collection step, triggering a new round of data collection. This closed-loop interaction mechanism ensures that the system can dynamically adjust maintenance plans based on the latest operating status, avoiding maintenance strategy deviations caused by data lag.

[0116] During the interaction between feedback signals and real-time operating parameters, the system continuously monitors the operating status of PV units, identifying changes in maintenance requirements due to environmental changes, equipment aging, or sudden failures. For example, if a PV unit was classified as operating in the high-efficiency zone in the previous cycle, but after a new round of data collection, its output voltage shows a continuous downward trend, the system will reclassify the unit as performing at a reduced performance and update its maintenance strategy matching relationship. The updated maintenance strategy will be incorporated into the operation and maintenance management library and transmitted to the data acquisition module via feedback signals, prompting the system to adjust the key parameters and monitoring frequency for subsequent data collection.

[0117] The generation and update process of the O&M management library supports a multi-tiered organization of maintenance tasks. The system categorizes maintenance tasks into different categories, such as routine inspections, preventive maintenance, and troubleshooting, based on factors such as urgency, execution difficulty, and resource requirements. Each task category has distinct execution processes and approval mechanisms. For example, routine inspection tasks can be automatically generated by the system and directly assigned to O&M personnel, while troubleshooting tasks require review by O&M management personnel before execution. This categorized management approach improves the organizational efficiency of O&M work and ensures that different types of maintenance tasks are properly handled according to their specific characteristics.

[0118] While generating the O&M management library, the system also supports a visual display of maintenance tasks. Through a graphical interface, O&M managers can view the status of all current maintenance tasks, including pending, ongoing, and completed. They can also view detailed information for each task, such as the target unit location, recommended execution strategy, and estimated completion time. This visual display enhances the transparency and controllability of O&M management, facilitating scheduling and decision-making.

[0119] In actual operation, the system incorporates a timestamp mechanism to ensure that all maintenance task generation, update, and execution records are accurately timed. This timestamp mechanism allows for the tracing of maintenance task history, facilitates analysis of the effectiveness of maintenance strategies, and provides data support for subsequent strategy optimization. Furthermore, the system features data caching and breakpoint-resume transfer capabilities, ensuring that critical data is retained even in the event of network interruptions or equipment failures, preventing the loss of maintenance task information.

[0120] The entire management output process runs through every aspect of operation and maintenance management, from extracting maintenance objectives to determining the execution sequence, from generating the operation and maintenance management library to transmitting feedback signals. This demonstrates the system's intelligent organization and management capabilities for operation and maintenance tasks. Through structured data processing and a closed-loop feedback mechanism, the system enables efficient and precise operation and maintenance management of distributed photovoltaic systems.

[0121] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0122] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A distributed photovoltaic intelligent operation and maintenance method, characterized in that: The steps include: Data collection step: real-time acquisition of operating parameters of multiple photovoltaic units in the distributed photovoltaic system, the operating parameters including voltage output data, current intensity data, temperature measurement data, and ambient light data; Partition analysis step: based on the operating parameters, classifying the performance status of each photovoltaic unit to identify performance partitions, wherein the performance partitions include a high-efficiency operation zone, a performance degradation zone, and a potential failure zone; Mapping construction step: extracting maintenance rule information from a maintenance policy database according to the performance partitions, and constructing a mapping relationship between the performance partitions and maintenance policies; Combination optimization step: extracting key partitions from the performance partitions, the key partitions being determined based on partition priority weights and system impact factors, and optimizing the combination matching degree of the maintenance strategies; Status update step: Based on the optimized maintenance strategy combination, the real-time operating status of the distributed photovoltaic system is verified, changes in maintenance requirements are identified, and the mapping relationship between the performance partitions and the maintenance strategies is updated; Management output step: Based on the updated mapping relationship, determine the objectives and execution sequence of the maintenance operations, and generate a structured operation and maintenance management library to guide subsequent maintenance activities.

2. The distributed photovoltaic intelligent operation and maintenance method according to claim 1, characterized in that: The data collection step is implemented as follows: For any photovoltaic unit in the distributed photovoltaic system, obtaining monitoring sensor data corresponding to the operating parameters; Using a data preprocessing model to perform noise filtering and outlier removal on the monitoring sensor data; Identifying key parameter change trends and related parameter association features in the monitoring sensor data to form a standardized data sample set; The key parameter change trends and related parameter association characteristics in the standardized data sample set are analyzed respectively, and the trend analysis area, change analysis area and association characteristic area corresponding to the standardized data sample set are obtained in sequence.

3. The distributed photovoltaic intelligent operation and maintenance method according to claim 2, characterized in that: The implementation method of obtaining the trend analysis area, the change analysis area, and the associated feature area corresponding to the standardized data sample set further includes: Merging the key parameter change trends and related parameter association features in the standardized data sample set according to the photovoltaic unit identification to obtain multiple merged result sets; extracting trend matching pairs from the merged result set, and comparing the trend matching pairs with a historical operation database to obtain a trend analysis area; Extracting the change rate threshold of the change analysis feature and the correlation strength value of the association feature in the merged result set, dividing the merged result set according to the change rate threshold and the correlation strength value to obtain the change analysis area and the association feature area.

4. The distributed photovoltaic intelligent operation and maintenance method according to claim 1, characterized in that: The implementation method of classifying the performance status of each photovoltaic unit in the partition analysis step further includes: Priority evaluation is performed on the performance partitions, the operating time data and fault frequency data of the photovoltaic units in the performance partitions are analyzed, the performance partitions are fitted according to the operating time data and fault frequency data, and a mapping relationship between the performance partitions and actual maintenance needs is constructed.

5. The distributed photovoltaic intelligent operation and maintenance method according to claim 1, characterized in that: The implementation of the mapping construction step includes: Retrieving maintenance rule information and historical maintenance case data in the maintenance policy database to generate a plurality of unlabeled maintenance policy identification results, wherein the unlabeled maintenance policy identification results represent maintenance rule information and historical maintenance case data that are not associated with parameter features in the performance partition; It is determined whether the plurality of unlabeled maintenance strategy identification results are valid maintenance strategy identification results. If they are valid maintenance strategy identification results, the valid maintenance strategy identification results are regarded as target maintenance strategies in the maintenance strategy database.

6. The distributed photovoltaic intelligent operation and maintenance method according to claim 3, characterized in that: The implementation of building the mapping relationship between the performance partition and the maintenance strategy includes: A mapping relationship is established between the performance partition and the maintenance strategy database using the information representation of the trend analysis area, the change parsing area and the associated feature area in the performance partition, the description information of the maintenance strategy database and the maintenance strategy category.

7. The distributed photovoltaic intelligent operation and maintenance method according to claim 1, characterized in that: The implementation of the combination optimization step includes: Performing cluster evaluation processing on the performance partition and the maintenance policy database according to the maintenance type, policy priority and policy function, and setting the core cluster center after the cluster evaluation processing as the key partition; Extracting core feature items of the key partitions, evaluating similarity matching between the core feature items, and setting common sequence information related to the similarity matching between the core feature items; Using the public sequence information related to the similarity matching between the core feature items, the maintenance strategy items present in the public sequence information are extracted, and the longest matching sequence between the maintenance strategy items is set, and the matching length value of the longest matching sequence is set as the combined matching degree between the maintenance strategy items; The maintenance strategy bias combination is set based on the combination matching degree between the maintenance strategy items and the time distribution characteristics of the maintenance strategy items.

8. The distributed photovoltaic intelligent operation and maintenance method according to claim 7, characterized in that: The implementation of the status update step includes: Extracting the time distribution characteristics of each maintenance policy item from the maintenance policy bias combination; setting the target maintenance path of the maintenance policy bias combination according to the time period information corresponding to the time distribution characteristics of each maintenance policy item; Fitting the target maintenance path of each maintenance strategy item in the maintenance strategy bias combination to obtain a fitted target maintenance path, and setting the occurrence probability value of the fitted target maintenance path in each time period information as the combined occurrence probability of the maintenance strategy bias combination; The occurrence probability of the maintenance strategy bias combination is compared with the historical maintenance case references in the maintenance strategy database, the difference value changes are identified, and the historical maintenance cases in the maintenance strategy database are classified and updated according to the difference value changes.

9. The distributed photovoltaic intelligent operation and maintenance method according to claim 1, characterized in that: The management output step is implemented as follows: Extracting maintenance target information from the distributed photovoltaic system based on the updated mapping relationship between the performance partitions and the maintenance strategy, sorting the maintenance target information according to the occurrence frequency characteristics of the maintenance target information, and obtaining the target and execution order of the maintenance operation; The maintenance target information and the execution sequence are combined and processed according to a structured storage format to obtain the operation and maintenance management library.

10. The distributed photovoltaic intelligent operation and maintenance method according to claim 1, characterized in that: Also includes: Data interaction step: in the state updating step, the updated mapping relationship data is transferred to the management output step; In the management output step, a feedback signal is generated using the updated mapping relationship data, and the feedback signal is transmitted to the data collection step to start a new round of data collection; Through the interaction of the feedback signal and the real-time operating parameters, closed-loop operation and maintenance management of the distributed photovoltaic system is achieved.

Images