A method and system for optimizing the inspection cycle of photovoltaic power plants based on fault analysis

Abstract

This invention relates to the field of photovoltaic power plant inspection technology, and discloses a method and system for optimizing the inspection cycle of photovoltaic power plants based on fault analysis. The method includes: collecting historical fault data of various equipment in the photovoltaic power plant, generating and analyzing several historical fault event sequences to obtain a basic inspection cycle and setting a first inspection cycle for each piece of equipment; setting several inspection tasks based on the first inspection cycle and basic information, and generating a second inspection cycle for the same inspection task; constructing a risk analysis model for each piece of equipment, and generating a predicted risk coefficient based on the risk analysis model and real-time operating data; determining whether to optimize the second inspection cycle based on the predicted risk coefficient, and if so, generating an optimization instruction for the second inspection cycle; dynamically adjusting the inspection cycle according to the actual fault situation and risk status of the equipment, thereby realizing intelligent adjustment of the inspection cycle and optimized allocation of inspection resources, and improving the accuracy and economy of photovoltaic power plant operation and maintenance.

Description

Technical Field

[0001] This application relates to the field of photovoltaic power plant inspection technology, and in particular to a method and system for optimizing the inspection cycle of photovoltaic power plants based on fault analysis. Background Technology

[0002] With the rapid growth of photovoltaic (PV) power generation capacity, the operation and maintenance (O&M) management of PV power plants faces significant challenges. Currently, PV power plants generally employ a fixed-cycle inspection model, such as a comprehensive inspection every month or quarter. However, this inspection method has obvious limitations: firstly, it lacks flexibility, failing to dynamically adjust inspection plans based on the actual operating status of different areas and equipment; secondly, it lacks in-depth mining and analysis of historical fault data, failing to fully utilize fault information to guide the scientific formulation of inspection cycles. Furthermore, the importance, operating environment, and fault modes of different equipment within a PV power plant vary, making a uniform inspection cycle insufficient to meet the needs of refined O&M. Summary of the Invention

[0003] To address the aforementioned technical issues, this application provides a method and system for optimizing the inspection cycle of photovoltaic power plants based on fault analysis. By collecting historical fault data from various devices and generating fault event sequences, the system analyzes these sequences to determine the basic inspection cycle and the first inspection cycle for each device. Based on this basic information, inspection tasks are grouped to generate a second inspection cycle at the task level. A risk analysis model is constructed and risk coefficients are predicted. The second inspection cycle is dynamically optimized based on these risk coefficients. Furthermore, the inspection cycle is dynamically adjusted according to the actual fault conditions and risk status of the equipment, achieving intelligent adjustment of the inspection cycle and optimized allocation of inspection resources, thereby improving the accuracy and economy of photovoltaic power plant operation and maintenance.

[0004] In some embodiments of this application, a method for optimizing the inspection cycle of a photovoltaic power plant based on fault analysis is provided, including: Collect historical fault data of each device in the photovoltaic power station and generate several historical fault event sequences for each device; Analyze the complete historical fault event sequence for each device to obtain the basic inspection cycle and set the first inspection cycle for each device; Based on the first inspection cycle of all equipment and basic information, several inspection tasks of the photovoltaic power station are set, and a second inspection cycle for each inspection task is generated. Construct risk analysis models for each device, and generate predicted risk coefficients for the corresponding devices based on the risk analysis models and real-time operational data; Based on the predicted risk coefficients of all equipment in the same inspection task, determine whether to optimize the second inspection cycle. If so, generate an optimization instruction for the second inspection cycle.

[0005] In some embodiments of this application, a sequence of historical fault events is generated for each device, including: Fault features are extracted from the historical fault data of each device to obtain several historical fault features, which are then clustered. Based on the processing results, several historical fault events are generated for each device. Identify several historical starting points for the same historical fault event, and generate corresponding historical fault coefficients based on the historical fault characteristics at each historical starting point. The historical fault coefficients corresponding to the historical fault features are sorted according to the time sequence of all historical start nodes of the same historical fault event to obtain the corresponding historical fault event sequence. Several historical fault event sequences are generated sequentially.

[0006] In some embodiments of this application, the basic inspection cycle is obtained, including: Generate the temporal characteristic change curves of the corresponding historical fault events based on the historical fault event sequence; Calculate the historical fault coefficient mean, historical fault frequency, and several slopes for each time-series characteristic change curve; The basic inspection cycle for corresponding historical fault events is set based on the historical average fault coefficient and historical fault frequency. Several compensation coefficients are generated based on several slopes to determine the basic inspection cycle of the corresponding event, and the average value of the compensation coefficients is calculated. The basic inspection cycle is corrected based on the average compensation coefficient, and the corrected basic inspection cycle is replaced with the basic inspection cycle corresponding to the historical fault events.

[0007] In some embodiments of this application, a first inspection cycle is set for each device, including: Obtain several historical impact data packets for each historical fault event, and each historical impact data packet is mapped to a corresponding historical fault coefficient. Generate the historical impact coefficient for each historical impact data packet, and construct a mapping table of historical failure coefficients and historical impact coefficients for the corresponding historical failure events; The historical impact level of the corresponding historical failure event is determined based on the mapping table; The weighting coefficients are set based on the historical impact levels of all historical fault events of the same equipment, and the weighted average is processed on the basic inspection cycle to obtain the first inspection cycle of each equipment.

[0008] In some embodiments of this application, several inspection tasks for the photovoltaic power station are set based on the first inspection cycle of all equipment and basic information, including: Construct a device topology distribution map of the photovoltaic power station based on the basic information of all equipment; Based on the equipment topology distribution map, calculate the physical distance between different devices. Combined with the first inspection cycle of each device, use a greedy algorithm to group and cluster the devices. Devices with physical distance differences less than a preset distance difference threshold and first inspection cycle differences less than a preset inspection cycle difference threshold are classified into the same inspection task. Several inspection tasks are generated sequentially.

[0009] In some embodiments of this application, generating a second inspection cycle for each inspection task includes: Based on the historical failure coefficient average, historical failure frequency, and historical impact level of each historical failure event for each device, the upper base radius, lower base radius, and height are respectively converted to obtain them. Based on the upper base radius, lower base radius, and height, a frustum model corresponding to the historical fault event is constructed. The volume of the frustum model is calculated and set as the fault event feature of the corresponding historical fault event. The comprehensive fault characteristic value of the equipment is obtained by summing the fault event characteristics of all historical fault events of the same equipment. The weight coefficient of each device is set according to the comprehensive fault characteristic value of each device, and the weight average is processed for the first inspection cycle of all devices in the same inspection task to obtain the second inspection cycle of the inspection task.

[0010] In some embodiments of this application, risk analysis models for various devices are constructed, including: Based on the comprehensive fault characteristic value of each device, a time interval is set, and several historical attention periods are generated according to the time interval and preset time window. Obtain historical operation data packets for each historical period of interest, compare and analyze all historical operation data packets for the same historical fault event, and determine the risk triggering characteristics of the corresponding historical fault event based on the analysis results; Each risk triggering feature includes several historical data intervals, and each historical data interval is mapped to the historical triggering probability and historical failure coefficient of the corresponding historical failure event. The first training dataset is constructed based on the risk triggering characteristics of all historical failure events of the same device. The risk identification model for the device is obtained by training the first training dataset; A second training dataset is constructed based on the historical failure coefficients and historical impact coefficients mapped from the historical impact data packets of all historical failure events of the same device. The risk association model for the device is obtained by training on the second training dataset; A risk analysis model for each device is generated based on the risk identification model and risk association model for each device.

[0011] In some embodiments of this application, a predicted risk coefficient for the corresponding device is generated based on a risk analysis model and real-time operational data, including: The system acquires real-time operating data for each device, inputs the real-time operating data into the risk identification model, performs feature extraction and matching on the real-time operating data, and outputs the corresponding predicted fault event, predicted trigger probability, and predicted fault coefficient if there are real-time data features that match the historical data range of risk triggering features. The output predicted failure events and predicted failure coefficients are input into the risk association model to obtain the predicted impact data package of the predicted failure events under the predicted failure coefficients and the corresponding predicted impact coefficients. The predicted risk coefficient of the equipment is obtained by weighting the predicted trigger probability, the predicted failure coefficient, and the predicted impact coefficient.

[0012] In some embodiments of this application, it is determined whether to optimize the second inspection cycle based on the predicted risk coefficients of all equipment in the same inspection task. If so, an optimization instruction for the second inspection cycle is generated, including: Set the real-time weighting coefficient for each device based on the predicted fault events of each device in the same inspection task. The comprehensive predicted risk coefficient for the corresponding inspection task is calculated based on the predicted risk coefficient and real-time weight coefficient of all equipment in the same inspection task. The comprehensive predicted risk coefficient is compared with the preset risk coefficient threshold of the corresponding inspection task. If the comprehensive predicted risk coefficient is greater than the preset risk coefficient threshold, it is determined that the second inspection cycle should be optimized and an optimization instruction to shorten the second inspection cycle is generated. If the overall predicted risk coefficient is not greater than the preset risk coefficient threshold, then it is determined that the second inspection cycle will not be optimized.

[0013] In some embodiments of this application, a photovoltaic power plant inspection cycle optimization system based on fault analysis is also included: The data acquisition module is used to collect historical fault data of various devices in the photovoltaic power station and generate several historical fault event sequences for each device. The analysis module is used to analyze all historical fault event sequences of each device to obtain the basic inspection cycle and set the first inspection cycle for each device. The setting module is used to set several inspection tasks for the photovoltaic power station based on the first inspection cycle of all equipment and basic information, and to generate the second inspection cycle for each inspection task. The prediction module is used to build risk analysis models for each device and generate the predicted risk coefficients for the corresponding devices based on the risk analysis models and real-time operating data. The optimization module is used to determine whether to optimize the second inspection cycle based on the predicted risk coefficients of all equipment in the same inspection task. If so, it generates optimization instructions for the second inspection cycle.

[0014] The photovoltaic power plant inspection cycle optimization method and system based on fault analysis according to the embodiments of this application have the following advantages compared with the prior art: By collecting historical fault data from various devices and generating fault event sequences, the analysis of these sequences determines the basic inspection cycle and the first inspection cycle for each device. Inspection tasks are grouped based on this basic information to generate a second inspection cycle at the task level. A risk analysis model is constructed and risk coefficients are predicted. The second inspection cycle is dynamically optimized based on these risk coefficients. The inspection cycle is also dynamically adjusted according to the actual fault conditions and risk status of the equipment, enabling intelligent adjustment of the inspection cycle and optimized allocation of inspection resources, thereby improving the accuracy and economy of photovoltaic power plant operation and maintenance. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a method for optimizing the inspection cycle of a photovoltaic power station based on fault analysis, as described in an embodiment of this application. Detailed Implementation

[0016] The specific embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but are not intended to limit the scope of this application.

[0017] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0018] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0019] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0020] like Figure 1 As shown in the figure, an embodiment of this application provides a method for optimizing the inspection cycle of a photovoltaic power plant based on fault analysis, comprising: S101: Collect historical fault data of each device in the photovoltaic power station and generate several historical fault event sequences for each device; S102: Analyze the entire historical fault event sequence of each device to obtain the basic inspection cycle and set the first inspection cycle for each device; S103: Based on the first inspection cycle of all equipment and basic information, set several inspection tasks for the photovoltaic power station and generate the second inspection cycle for each inspection task. S104: Construct risk analysis models for each device, and generate predicted risk coefficients for the corresponding devices based on the risk analysis models and real-time operating data; S105: Determine whether to optimize the second inspection cycle based on the predicted risk coefficients of all equipment in the same inspection task. If so, generate the optimization instruction for the second inspection cycle.

[0021] In some embodiments of this application, a sequence of historical fault events is generated for each device, including: Fault features are extracted from the historical fault data of each device to obtain several historical fault features, which are then clustered. Based on the processing results, several historical fault events are generated for each device. Identify several historical starting points for the same historical fault event, and generate corresponding historical fault coefficients based on the historical fault characteristics at each historical starting point. The historical fault coefficients corresponding to the historical fault features are sorted according to the time sequence of all historical start nodes of the same historical fault event to obtain the corresponding historical fault event sequence. Several historical fault event sequences are generated sequentially.

[0022] In this embodiment, historical fault events are a set of faults with similar fault manifestations and causes obtained by clustering after extracting fault features from historical fault data of equipment. Historical fault events include, but are not limited to, IGBT module damage faults of inverters and hot spot effect faults of photovoltaic modules.

[0023] In this embodiment, the historical fault coefficient is a comprehensive value obtained by quantifying the historical fault characteristics at the same historical starting point based on several preset fault assessment indicators corresponding to historical fault events. It is used to characterize the severity and scope of the fault at the historical starting point.

[0024] In this embodiment, the preset fault assessment indicators are set based on historical fault events. Different fault events have different fault assessment indicators. For example, for IGBT module damage faults in inverters, the preset fault assessment indicators may include the percentage of output power loss when the fault occurs, the module replacement cost coefficient, the impact level on grid stability, and the fault repair time index. For hot spot effect faults in photovoltaic modules, the preset fault assessment indicators may cover the degree by which the temperature of the hot spot area exceeds the normal range, the proportion of affected modules, the power generation attenuation rate, and the potential fire risk coefficient. Each assessment indicator is mapped to several preset feature intervals, and each preset feature interval corresponds to a preset fault coefficient. By matching historical fault characteristics with the preset feature intervals of these preset fault assessment indicators, selecting the corresponding preset fault coefficient and performing weight processing, the resulting historical fault coefficient can more accurately reflect the actual fault degree of different historical fault events at a specific starting point. When the historical fault coefficient is larger, it indicates a greater fault degree and a higher hazard, providing a reliable quantitative basis for subsequent inspection cycle optimization based on fault event sequences.

[0025] In some embodiments of this application, the basic inspection cycle is obtained, including: Generate the temporal characteristic change curves of the corresponding historical fault events based on the historical fault event sequence; Calculate the historical fault coefficient mean, historical fault frequency, and several slopes for each time-series characteristic change curve; The basic inspection cycle for corresponding historical fault events is set based on the historical average fault coefficient and historical fault frequency. Several compensation coefficients are generated based on several slopes to determine the basic inspection cycle of the corresponding event, and the average value of the compensation coefficients is calculated. The basic inspection cycle is corrected based on the average compensation coefficient, and the corrected basic inspection cycle is replaced with the basic inspection cycle corresponding to the historical fault events.

[0026] In this embodiment, the horizontal axis of the time-series characteristic change curve is the time axis. The time axis starts from the earliest historical starting node in the historical fault event sequence and extends sequentially to the latest historical starting node. Each historical starting node corresponds to a specific time point on the time axis. By connecting the historical fault coefficients of each historical starting node in the same historical fault event sequence in chronological order, a curve that can intuitively reflect the trend of the fault degree change at different time points of the historical fault event is formed.

[0027] In this embodiment, the slope = (follower historical fault coefficient - previous historical fault coefficient) / (follower historical start time - previous historical start time). A positive slope indicates that the fault severity is increasing, and the larger the absolute value of the slope, the faster the fault severity deteriorates; a negative slope indicates that the fault severity is decreasing, and the larger the absolute value, the faster the improvement; a slope of zero indicates that the fault severity remains stable during this period.

[0028] In this embodiment, the historical failure coefficient mean refers to the average value obtained by dividing the sum of all historical failure coefficients in the same historical failure event sequence by the number of historical failure coefficients. The historical failure coefficient mean reflects the average failure severity of the historical failure event throughout the entire recording period.

[0029] In this embodiment, the historical fault frequency is calculated by counting the total number of historical starting nodes in the historical fault event sequence and dividing it by the total time length corresponding to the sequence (i.e., the difference between the latest historical starting node time and the earliest historical starting node time), and is used to characterize the frequency of fault occurrence.

[0030] In this embodiment, a mapping relationship between the average fault coefficient, fault frequency, and basic inspection cycle is established through a preset mapping function. That is, the larger the average fault coefficient and the higher the fault frequency, the shorter the corresponding basic inspection cycle, and vice versa. The basic inspection cycle corresponding to the historical average fault coefficient and historical fault frequency is found in the mapping relationship and is initially determined as the basic inspection cycle of the historical fault event. The preset mapping function is a nonlinear function model trained by a large amount of actual fault case data, which can dynamically adjust the calculation weight of the basic inspection cycle according to the characteristics of different fault events.

[0031] In this embodiment, a pre-defined correspondence between slope intervals and compensation coefficients is established. The slope intervals include several positive slope intervals, several negative slope intervals, and zero, with each interval corresponding to a different compensation coefficient value. The positive slope intervals are divided into multiple sub-intervals in descending order: intervals with a positive slope greater than 0.8 correspond to a compensation coefficient of 0.75, the 0.5-0.8 interval to 0.85, and the 0.2-0.5 interval to 0.95; intervals with a negative slope absolute value greater than 0.8 correspond to a compensation coefficient of 1.25, the 0.5-0.8 interval to 1.15, and the 0.2-0.5 interval to 1.05. That is, when the slope is positive... When the slope is in the range of 0.75-1, the compensation coefficient is smaller as the absolute value of the slope increases. This is used to shorten the basic inspection cycle to cope with the upward trend of the fault severity. When the slope is in the negative slope range, the compensation coefficient is 1-1.25. The larger the absolute value, the larger the compensation coefficient. This is used to extend the basic inspection cycle to adapt to the downward trend of the fault severity. When the slope is zero, the compensation coefficient is 1 and the basic inspection cycle remains unchanged. This achieves refined compensation of the basic inspection cycle under different fault trends.

[0032] In this embodiment, the slope interval and the range of the compensation coefficient are determined by statistical analysis of historical fault data of a large number of different types of photovoltaic equipment (such as inverters, photovoltaic modules, combiner boxes, etc.) under different climatic conditions (high temperature, high humidity, windy sand, strong radiation, etc.) and different years of operation. The typical development trend characteristics of various equipment faults are identified, and the boundary values ​​of the slope interval are divided based on this. The range of the compensation coefficient is combined with the cost-benefit analysis of fault handling to ensure that the optimal allocation of inspection resources is achieved under the premise of ensuring the safe operation of the power station.

[0033] In this embodiment, after obtaining the compensation coefficients corresponding to several slopes, these compensation coefficients are weighted and averaged (the weights are set according to the length of the time interval corresponding to the slope; the longer the time interval, the smaller the weight, and vice versa), to obtain the average value of the compensation coefficients. Then, the average value is multiplied by the initially determined basic inspection cycle to complete the correction of the basic inspection cycle.

[0034] In this embodiment, the basic inspection cycle and the average compensation coefficient are determined by the average historical fault coefficient, the historical fault frequency, and the slope. This accurately reflects the severity, frequency, and development trend of historical fault events, thereby determining the basic inspection cycle after correction for each historical fault event. This provides a scientific basis for setting the first inspection cycle of the equipment, thereby realizing the quantitative transformation from a single fault event to the overall equipment inspection requirements, and laying the foundation for the planning of subsequent power plant-level inspection tasks.

[0035] In some embodiments of this application, a first inspection cycle is set for each device, including: Obtain several historical impact data packets for each historical fault event, and each historical impact data packet is mapped to a corresponding historical fault coefficient. Generate the historical impact coefficient for each historical impact data packet, and construct a mapping table of historical failure coefficients and historical impact coefficients for the corresponding historical failure events; The historical impact level of the corresponding historical failure event is determined based on the mapping table; The weighting coefficients are set based on the historical impact levels of all historical fault events of the same equipment, and the weighted average is processed on the basic inspection cycle to obtain the first inspection cycle of each equipment.

[0036] In this embodiment, the historical impact data package includes specific impact data of the historical fault event on the photovoltaic power plant from a historical start node to the fault repair completion node. The historical impact data includes, but is not limited to, power generation loss, equipment maintenance costs, impact on other equipment, fault handling time, and safety risk level. The historical impact data in the historical impact data package is all standardized quantitative data, which facilitates a unified assessment of the degree of impact.

[0037] In this embodiment, the historical impact coefficient is obtained by evaluating the historical impact data associated with the historical impact data package through several preset impact assessment indicators. Each preset impact assessment indicator is mapped to several preset impact data intervals, and each preset impact interval corresponds to a preset impact coefficient. By matching the historical impact data with the preset impact data intervals of these preset impact assessment indicators, the corresponding preset impact coefficient is selected and weighted summation is performed to obtain the historical impact coefficient, which is used to comprehensively characterize the actual impact of the historical fault event on the photovoltaic power station at a specific historical starting point. The value range of the historical impact coefficient is 0-1.

[0038] In this embodiment, the average historical influence coefficient is calculated based on all historical influence coefficients in the mapping table. The initial influence level is determined based on the average historical influence coefficient. The initial influence level includes four levels: Level 1, Level 2, Level 3, and Level 4. The average historical influence coefficient is less than 0.2, corresponding to Level 1; 0.2-0.5, corresponding to Level 2; 0.5-0.8, corresponding to Level 3; and 0.8-1, corresponding to Level 4.

[0039] In this embodiment, the initial impact level is dynamically adjusted according to the relationship between the historical impact coefficient and the historical failure coefficient in the mapping table. When the historical impact coefficient and the historical failure coefficient are positively correlated and the correlation coefficient is greater than 0.7, the initial impact level is increased by one level; when the historical impact coefficient and the historical failure coefficient are negatively correlated and the absolute value of the correlation coefficient is greater than 0.5, the initial impact level is decreased by one level; otherwise, the initial impact level remains unchanged, thus obtaining the final historical impact level.

[0040] In this embodiment, the weighting coefficients are set according to the historical impact level. The weighting coefficient for the fourth level of impact is 1.0, for the third level it is 0.8, for the second level it is 0.5, and for the first level it is 0.3. The basic inspection cycle of each historical fault event is multiplied by its corresponding weighting coefficient, summed, and then divided by the sum of all weighting coefficients to obtain the first inspection cycle of the equipment. This allows the first inspection cycle of the equipment to comprehensively consider the impact of different fault events on the power station and to achieve a scientific quantification of the overall inspection requirements of the equipment.

[0041] In this embodiment, by determining the historical impact level, the weight coefficient of the basic inspection cycle for each historical fault event is set, and the first inspection cycle for each device is calculated. This can fully reflect the comprehensive inspection needs of the device under the influence of different fault events, avoid the problem of unreasonable inspection cycle setting due to the one-sidedness of a single fault event, and ensure that the first inspection cycle considers both the severity and frequency of the fault, as well as the actual impact of the fault on the power station. This provides a precise basis for the planning of subsequent photovoltaic power station inspection tasks.

[0042] In some embodiments of this application, several inspection tasks for the photovoltaic power station are set based on the first inspection cycle of all equipment and basic information, including: Construct a device topology distribution map of the photovoltaic power station based on the basic information of all equipment; Based on the equipment topology distribution map, calculate the physical distance between different devices. Combined with the first inspection cycle of each device, use a greedy algorithm to group and cluster the devices. Devices with physical distance differences less than a preset distance difference threshold and first inspection cycle differences less than a preset inspection cycle difference threshold are classified into the same inspection task. Several inspection tasks are generated sequentially.

[0043] In this embodiment, the basic information includes device type, installation location coordinates, array number, and electrical connection relationship.

[0044] In this embodiment, the specific execution steps of the greedy algorithm are as follows: First, sort all devices according to their first inspection cycle from smallest to largest, and select the first unassigned device from the sorted device list as the starting device for the current inspection task; with the starting device as the center, determine whether the difference between its first inspection cycle and the first inspection cycle of the devices already included in the current inspection task is less than a preset inspection cycle difference threshold (5 days in this application). If the condition is met and the physical distance difference is less than the preset distance difference threshold (50 meters in this application), then add the device to the current inspection task; continue to use the geometric center of all devices in the current inspection task as the new search center, and repeat the above search and judgment process until no device that meets the conditions can be added to the current inspection task. Repeat the above steps until all devices are assigned to the corresponding inspection tasks.

[0045] In this embodiment, the above method can quickly group and cluster the equipment while ensuring that the physical locations of the equipment within the inspection task are concentrated and the inspection cycles are similar. This effectively reduces the travel path length of the inspection personnel, improves inspection efficiency, and avoids the problem of over- or under-inspection of some equipment due to excessive differences in inspection cycles.

[0046] In some embodiments of this application, generating a second inspection cycle for each inspection task includes: Based on the historical failure coefficient average, historical failure frequency, and historical impact level of each historical failure event for each device, the upper base radius, lower base radius, and height are respectively converted to obtain them. Based on the upper base radius, lower base radius, and height, a frustum model corresponding to the historical fault event is constructed. The volume of the frustum model is calculated and set as the fault event feature of the corresponding historical fault event. The comprehensive fault characteristic value of the equipment is obtained by summing the fault event characteristics of all historical fault events of the same equipment. The weight coefficient of each device is set according to the comprehensive fault characteristic value of each device, and the weight average is processed for the first inspection cycle of all devices in the same inspection task to obtain the second inspection cycle of the inspection task.

[0047] In this embodiment, the historical average failure coefficient, historical failure frequency, and historical impact level are mapped to preset size ranges. The allowable ranges for the upper and lower radii of the circular platform are set as [R1{min},R1{max}], the allowable ranges for the lower radii are [R2{min}, R2{max}], and the allowable ranges for the height are [H{min}, H{max}]. The conversion formula for the upper radii is = R1{min} + (R1{max} -R1{min})×μi μ{min} / μ{max} μ{min}, where μ{min} and μ{max} are the minimum and maximum values ​​of the historical fault coefficients. The conversion formulas for the lower radius and height are the same as above. Based on the preset size range, the average historical fault coefficient, historical fault frequency, and historical impact level are converted into the upper radius, lower radius, and height respectively through linear normalization.

[0048] In this embodiment, by constructing a frustum model for each historical fault event and calculating the volume of the frustum as a feature of the fault event, the information of the three key dimensions of historical fault coefficient mean, historical fault frequency, and historical impact level can be spatially integrated, so that the comprehensive characteristics of the fault event can be presented in an intuitive volume.

[0049] In this embodiment, the comprehensive fault feature value obtained by summing the fault event characteristics of all historical fault events of the same device comprehensively covers the overall impact of all historical fault events of the device. This provides a scientific and quantitative basis for setting the weight coefficient of the first inspection cycle of the device in the inspection task, thereby corresponding to the second inspection cycle of the task and improving the accuracy and adaptability of the inspection task.

[0050] In some embodiments of this application, risk analysis models for various devices are constructed, including: Based on the comprehensive fault characteristic value of each device, a time interval is set, and several historical attention periods are generated according to the time interval and preset time window. Obtain historical operation data packets for each historical period of interest, compare and analyze all historical operation data packets for the same historical fault event, and determine the risk triggering characteristics of the corresponding historical fault event based on the analysis results; Each risk triggering feature includes several historical data intervals, and each historical data interval is mapped to the historical triggering probability and historical failure coefficient of the corresponding historical failure event. The first training dataset is constructed based on the risk triggering characteristics of all historical failure events of the same device. The risk identification model for the device is obtained by training the first training dataset; A second training dataset is constructed based on the historical failure coefficients and historical impact coefficients mapped from the historical impact data packets of all historical failure events of the same device. The risk association model for the device is obtained by training on the second training dataset; A risk analysis model for each device is generated based on the risk identification model and risk association model for each device.

[0051] In this embodiment, the historical attention period refers to the time interval that is shifted forward from the historical starting point. The time point after the shift is taken as the end point of the historical attention period, and a preset time window length is extracted forward as the starting point of the historical attention period, thus forming the historical attention period.

[0052] In this embodiment, the larger the comprehensive fault feature value, the longer the time interval. That is, when the comprehensive fault feature value is larger, the interval between the historical attention period of the corresponding historical fault event and the corresponding historical starting node is longer. This allows for a more comprehensive capture of the long-term operational data change trend that may exist before the fault occurs, providing a more sufficient data foundation for the extraction of risk triggering features. The correspondence between the comprehensive fault feature value and the time interval is set based on historical data. The selected time interval refers to the length of time that can comprehensively capture the long-term operational data change trend that may exist before the fault occurs while providing timely warnings for subsequent fault events.

[0053] In this embodiment, the preset time window is set in advance. In this application, the preset time window is 30 days.

[0054] In this embodiment, risk triggering features refer to key operational data features that can predict the occurrence of specific historical fault events of photovoltaic equipment. These features are obtained by comparing and analyzing historical operational data packets within the historical period of interest. When the same historical operational data feature appears multiple times and the frequency of occurrence is significantly correlated with the occurrence of historical fault events, it is identified as a risk triggering feature. Each feature corresponds to multiple historical data intervals (set according to the historical operational data packets), and each historical data interval is associated with the probability of the corresponding historical fault event occurring within that interval (i.e., the historical triggering probability ranges from 25% to 100%) and the historical fault coefficient of the corresponding historical fault event.

[0055] In this embodiment, the historical association data packet of each historical fault event includes data such as the impact range, duration, degree of impact, and fault propagation path of the corresponding fault event of the current device on other devices. These data are obtained by analyzing information such as changes in the operating status of other devices, fault occurrence time sequence, and fault type correlation when the historical fault event occurs. They are used to characterize the potential impact relationship of historical fault events with different historical fault coefficients of the current device on other devices in the photovoltaic power station.

[0056] In this embodiment, the risk triggering features in the first training dataset are used as training input data, and the mapped historical fault events, historical trigger probabilities, and historical fault coefficients are used as training output data to train the neural network, so as to achieve accurate identification and probability prediction of equipment risk triggering features.

[0057] In this embodiment, historical fault events and their corresponding historical fault coefficients in the second training dataset are used as training input data, and the mapped historical impact data packets and their corresponding historical impact coefficients are used as training output data to train a neural network, so as to achieve a quantitative assessment of the degree of impact that may occur after a fault event occurs.

[0058] In this embodiment, the risk identification model can monitor the equipment's operating data in real time. When it detects operating data that matches the historical data range in the risk triggering characteristics, it can output the corresponding predicted fault event, the corresponding predicted trigger probability, and the predicted fault coefficient, thereby achieving early identification of potential fault risks. After the risk identification model outputs the predicted fault event and the predicted fault coefficient, the risk association model predicts the possible predicted impact data packets and the corresponding predicted impact coefficients, thereby assessing the comprehensive impact of the potential fault on the power plant.

[0059] In this embodiment, the risk identification model and the risk association model are integrated into a risk analysis model for the equipment. This model can not only warn of the possibility of failure, but also predict the potential impact of failure, providing a forward-looking basis for the dynamic adjustment of inspection strategies, and further improving the initiative and accuracy of photovoltaic power station inspection.

[0060] In some embodiments of this application, a predicted risk coefficient for the corresponding device is generated based on a risk analysis model and real-time operational data, including: The system acquires real-time operating data for each device, inputs the real-time operating data into the risk identification model, performs feature extraction and matching on the real-time operating data, and outputs the corresponding predicted fault event, predicted trigger probability, and predicted fault coefficient if there are real-time data features that match the historical data range of risk triggering features. The output predicted failure events and predicted failure coefficients are input into the risk association model to obtain the predicted impact data package of the predicted failure events under the predicted failure coefficients and the corresponding predicted impact coefficients. The predicted risk coefficient of the equipment is obtained by weighting the predicted trigger probability, the predicted failure coefficient, and the predicted impact coefficient.

[0061] In this embodiment, the weight of the predicted trigger probability is 0.4, the weight of the predicted fault coefficient is 0.3, the weight of the predicted impact coefficient is 0.3, and the value of the predicted risk coefficient ranges from 0 to 1. The larger the value, the higher the potential risk currently faced by the device.

[0062] In this embodiment, by introducing real-time operating data and risk analysis models, the current operating status and potential risks of the equipment can be dynamically captured, so that the predicted risk coefficient can reflect the risk level of the equipment in real time, providing timely and accurate risk basis for the dynamic adjustment of subsequent inspection cycles.

[0063] In some embodiments of this application, it is determined whether to optimize the second inspection cycle based on the predicted risk coefficients of all equipment in the same inspection task. If so, an optimization instruction for the second inspection cycle is generated, including: Set the real-time weighting coefficient for each device based on the predicted fault events of each device in the same inspection task. The comprehensive predicted risk coefficient for the corresponding inspection task is calculated based on the predicted risk coefficient and real-time weight coefficient of all equipment in the same inspection task. The comprehensive predicted risk coefficient is compared with the preset risk coefficient threshold of the corresponding inspection task. If the comprehensive predicted risk coefficient is greater than the preset risk coefficient threshold, it is determined that the second inspection cycle should be optimized and an optimization instruction to shorten the second inspection cycle is generated. If the overall predicted risk coefficient is not greater than the preset risk coefficient threshold, then it is determined that the second inspection cycle will not be optimized.

[0064] In this embodiment, the setting of the real-time weight coefficient is related to the severity of the predicted fault event of the device. The higher the historical impact level of the predicted fault event, the larger its corresponding real-time weight coefficient, and vice versa.

[0065] In this embodiment, the formula for calculating the comprehensive prediction risk coefficient is: Comprehensive risk value = Σ (predicted risk coefficient of device i × real-time weight coefficient of device i) / total number of devices in the inspection task. This formula can comprehensively reflect the current risk status of all devices in the inspection task.

[0066] In this embodiment, the preset risk coefficient threshold is pre-set based on the maximum risk coefficient corresponding to the second inspection cycle of the corresponding inspection task. Different second inspection cycles correspond to different preset risk coefficient thresholds. The shorter the second inspection cycle, the higher the preset risk coefficient threshold, so as to achieve refined control over high-risk inspection tasks.

[0067] In this embodiment, when generating the optimization instruction to shorten the second inspection cycle, the extent of the shortening is determined based on the degree to which the comprehensive predicted risk coefficient exceeds the preset risk coefficient threshold. The greater the exceedance, the greater the shortening. In this application, when the comprehensive predicted risk coefficient exceeds the preset threshold by 20%, the second inspection cycle can be shortened by 10%; when it exceeds 50%, it can be shortened by 30%. The specific shortening rules can be dynamically adjusted based on the actual operation and maintenance experience of the power plant and cost-benefit analysis.

[0068] In this embodiment, by calculating the comprehensive predicted risk coefficient of each inspection task and comparing it with the preset risk coefficient threshold, the risk level of the inspection task can be dynamically assessed, and optimization instructions to shorten the inspection cycle can be generated in a timely manner. This allows the inspection strategy to be flexibly adjusted according to the real-time risk status of the equipment, further improving the pertinence and timeliness of the inspection. This helps to ensure the safe and stable operation of the power plant while achieving the rational allocation and efficient utilization of inspection resources.

[0069] In some embodiments of this application, a photovoltaic power plant inspection cycle optimization system based on fault analysis is also included: The data acquisition module is used to collect historical fault data of various devices in the photovoltaic power station and generate several historical fault event sequences for each device. The analysis module is used to analyze all historical fault event sequences of each device to obtain the basic inspection cycle and set the first inspection cycle for each device. The setting module is used to set several inspection tasks for the photovoltaic power station based on the first inspection cycle of all equipment and basic information, and to generate the second inspection cycle for each inspection task. The prediction module is used to build risk analysis models for each device and generate the predicted risk coefficients for the corresponding devices based on the risk analysis models and real-time operating data. The optimization module is used to determine whether to optimize the second inspection cycle based on the predicted risk coefficients of all equipment in the same inspection task. If so, it generates optimization instructions for the second inspection cycle.

[0070] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of this application, and these improvements and substitutions should also be considered within the scope of protection of this application.

Claims

1. A method for optimizing the inspection cycle of a photovoltaic power station based on fault analysis, characterized in that, include: Collect historical fault data of each device in the photovoltaic power station and generate several historical fault event sequences for each device; Analyze the complete historical fault event sequence for each device to obtain the basic inspection cycle and set the first inspection cycle for each device; Based on the first inspection cycle of all equipment and basic information, several inspection tasks of the photovoltaic power station are set, and a second inspection cycle for each inspection task is generated. Construct risk analysis models for each device, and generate predicted risk coefficients for the corresponding devices based on the risk analysis models and real-time operational data; Based on the predicted risk coefficients of all equipment in the same inspection task, determine whether to optimize the second inspection cycle. If so, generate an optimization instruction for the second inspection cycle. 2.The method of claim 1, wherein, Generate a sequence of historical fault events for each device, including: Fault features are extracted from the historical fault data of each device to obtain several historical fault features, which are then clustered. Based on the processing results, several historical fault events are generated for each device. Identify several historical starting points for the same historical fault event, and generate corresponding historical fault coefficients based on the historical fault characteristics at each historical starting point. The historical fault coefficients corresponding to the historical fault features are sorted according to the time sequence of all historical start nodes of the same historical fault event to obtain the corresponding historical fault event sequence. Several historical fault event sequences are generated sequentially. 3.The method of claim 2, wherein, The basic inspection cycle is obtained, including: Generate the temporal characteristic change curves of the corresponding historical fault events based on the historical fault event sequence; Calculate the historical fault coefficient mean, historical fault frequency, and several slopes for each time-series characteristic change curve; The basic inspection cycle for corresponding historical fault events is set based on the historical average fault coefficient and historical fault frequency. Several compensation coefficients are generated based on several slopes to determine the basic inspection cycle of the corresponding event, and the average value of the compensation coefficients is calculated. The basic inspection cycle is corrected based on the average compensation coefficient, and the corrected basic inspection cycle is replaced with the basic inspection cycle corresponding to the historical fault events.

4. The photovoltaic power plant inspection cycle optimization method based on fault analysis as described in claim 1, characterized in that, Set the first inspection cycle for each device, including: Obtain several historical impact data packets for each historical fault event, and each historical impact data packet is mapped to a corresponding historical fault coefficient. Generate the historical impact coefficient for each historical impact data packet, and construct a mapping table of historical failure coefficients and historical impact coefficients for the corresponding historical failure events; The historical impact level of the corresponding historical failure event is determined based on the mapping table; The weighting coefficients are set based on the historical impact levels of all historical fault events of the same equipment, and the weighted average is processed on the basic inspection cycle to obtain the first inspection cycle of each equipment.

5. The photovoltaic power plant inspection cycle optimization method based on fault analysis as described in claim 1, characterized in that, Based on the first inspection cycle of all equipment and basic information, several inspection tasks are set for the photovoltaic power station, including: Construct a topology distribution map of the photovoltaic power station based on the basic information of all equipment; Based on the equipment topology distribution map, calculate the physical distance between different equipment. Combined with the first inspection cycle of each equipment, use a greedy algorithm to group and cluster the equipment. Equipment with physical distance differences less than a preset distance difference threshold and first inspection cycle differences less than a preset inspection cycle difference threshold are classified into the same inspection task. Several inspection tasks are generated in sequence.

6. The photovoltaic power plant inspection cycle optimization method based on fault analysis as described in claim 4, characterized in that, Generate the second inspection cycle for each inspection task, including: Based on the historical failure coefficient average, historical failure frequency, and historical impact level of each historical failure event for each device, the upper base radius, lower base radius, and height are respectively converted to obtain them. Based on the upper base radius, lower base radius, and height, a frustum model corresponding to the historical fault event is constructed. The volume of the frustum model is calculated and set as the fault event feature of the corresponding historical fault event. The comprehensive fault characteristic value of the equipment is obtained by summing the fault event characteristics of all historical fault events of the same equipment. The weight coefficient of each device is set according to the comprehensive fault characteristic value of each device, and the weighted average is processed for the first inspection cycle of all devices in the same inspection task to obtain the second inspection cycle of the inspection task.

7. The photovoltaic power plant inspection cycle optimization method based on fault analysis as described in claim 6, characterized in that, Construct risk analysis models for each device, including: Based on the comprehensive fault characteristic value of each device, a time interval is set, and several historical attention periods are generated according to the time interval and preset time window. Obtain historical operation data packets for each historical period of interest, compare and analyze all historical operation data packets for the same historical fault event, and determine the risk triggering characteristics of the corresponding historical fault event based on the analysis results; Each risk triggering feature includes several historical data intervals, and each historical data interval is mapped to the historical triggering probability and historical failure coefficient of the corresponding historical failure event. The first training dataset is constructed based on the risk triggering characteristics of all historical failure events of the same device. The risk identification model for the device is obtained by training the first training dataset; A second training dataset is constructed based on the historical failure coefficients and historical impact coefficients mapped from the historical impact data packets of all historical failure events of the same device. The risk association model for the device is obtained by training on the second training dataset; A risk analysis model for each device is generated based on the risk identification model and risk association model for each device.

8. The photovoltaic power plant inspection cycle optimization method based on fault analysis as described in claim 7, characterized in that, Based on the risk analysis model and real-time operational data, predictive risk coefficients for corresponding equipment are generated, including: The system acquires real-time operating data for each device, inputs the real-time operating data into the risk identification model, performs feature extraction and matching on the real-time operating data, and outputs the corresponding predicted fault event, predicted trigger probability, and predicted fault coefficient if there are real-time data features that match the historical data range of risk triggering features. The output predicted failure events and predicted failure coefficients are input into the risk association model to obtain the predicted impact data package of the predicted failure events under the predicted failure coefficients and the corresponding predicted impact coefficients. The predicted risk coefficient of the equipment is obtained by weighting the predicted trigger probability, the predicted failure coefficient, and the predicted impact coefficient.

9. The photovoltaic power plant inspection cycle optimization method based on fault analysis as described in claim 8, characterized in that, Based on the predicted risk coefficients of all equipment in the same inspection task, determine whether to optimize the second inspection cycle. If so, generate optimization instructions for the second inspection cycle, including: Set the real-time weighting coefficient for each device based on the predicted fault events of each device in the same inspection task. The comprehensive predicted risk coefficient for the corresponding inspection task is calculated based on the predicted risk coefficient and real-time weight coefficient of all equipment in the same inspection task. The comprehensive predicted risk coefficient is compared with the preset risk coefficient threshold of the corresponding inspection task. If the comprehensive predicted risk coefficient is greater than the preset risk coefficient threshold, it is determined that the second inspection cycle should be optimized and an optimization instruction to shorten the second inspection cycle is generated. If the overall predicted risk coefficient is not greater than the preset risk coefficient threshold, then it is determined that the second inspection cycle will not be optimized.

10. A photovoltaic power plant inspection cycle optimization system based on fault analysis, characterized in that, include: The data acquisition module is used to collect historical fault data of various devices in the photovoltaic power station and generate several historical fault event sequences for each device. The analysis module is used to analyze all historical fault event sequences of each device to obtain the basic inspection cycle and set the first inspection cycle for each device. The setting module is used to set several inspection tasks for the photovoltaic power station based on the first inspection cycle of all equipment and basic information, and to generate the second inspection cycle for each inspection task. The prediction module is used to build risk analysis models for each device and generate the predicted risk coefficients for the corresponding devices based on the risk analysis models and real-time operating data. The optimization module is used to determine whether to optimize the second inspection cycle based on the predicted risk coefficients of all equipment in the same inspection task. If so, it generates optimization instructions for the second inspection cycle.

Images