A remote fault diagnosis method for photovoltaic power generation equipment

Abstract

This application provides a remote fault diagnosis method for photovoltaic power generation equipment, relating to the field of data processing technology. The method includes: performing cluster analysis on distributed photovoltaic equipment based on environmental change feature sets, equipment model sets, and performance parameter sets to determine multiple equipment sets; identifying abnormal power generation based on multiple power generation sets to determine abnormal equipment sets; and predicting faults by matching the abnormal equipment model, performance parameters, and environmental monitoring data to determine an associated fault feature library and an associated fault prediction model, thereby determining the predicted fault type. This application addresses the technical problem in existing distributed photovoltaic power generation systems where the large number and dispersed nature of photovoltaic equipment makes it difficult to promptly and accurately detect abnormal operating states of individual devices, resulting in poor system efficiency, stability, and security. It improves the convenience and accuracy of identifying abnormal operating states, achieving intelligent and automated equipment fault diagnosis.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a remote fault diagnosis method for photovoltaic power generation equipment. Background Technology

[0002] A distributed photovoltaic (PV) power generation system is a power generation system that disperses PV power generation equipment at the user end, such as on or near the roofs of residential, commercial buildings, or industrial facilities, to achieve efficient utilization of solar energy. Unlike traditional centralized PV power plants, this system does not rely on large-scale solar power plants; instead, it installs PV modules near the user, directly providing electricity to the user or the power grid.

[0003] Currently, existing regional distributed photovoltaic power generation systems suffer from poor operating efficiency, stability, and safety due to the large number and dispersed nature of photovoltaic power generation equipment. This makes it difficult to detect abnormal operating conditions of individual devices in a timely and accurate manner, thus hindering timely measures for equipment fault diagnosis and maintenance. Summary of the Invention

[0004] The purpose of this application is to provide a remote fault diagnosis method for photovoltaic power generation equipment, in order to solve the technical problem that in existing regional distributed photovoltaic power generation systems, due to the large number and dispersion of photovoltaic power generation equipment, it is impossible to detect the abnormal working status of individual equipment in a timely and accurate manner, thus making it impossible to take timely measures for equipment fault diagnosis and maintenance, resulting in poor operating efficiency, stability and safety of photovoltaic power generation systems.

[0005] In view of the above problems, this application provides a remote fault diagnosis method for photovoltaic power generation equipment. The method includes: collecting equipment attribute information of distributed photovoltaic equipment within a target area to obtain an equipment attribute set, wherein the equipment attribute set includes an equipment model set, a performance parameter set, an installed capacity set, and a location coordinate set; based on predetermined power generation influencing factors and the location coordinate set, calling regional environmental monitoring logs to perform environmental change analysis; performing cluster analysis on the distributed photovoltaic equipment based on the environmental change feature set, the equipment model set, and the performance parameter set to determine multiple equipment sets; under a preset monitoring node, obtaining multiple power generation sets of the multiple equipment sets; and identifying abnormal power generation based on the multiple power generation sets and the installed capacity set to determine abnormal equipment. The system selects a first abnormal device from the set of abnormal devices, activates the associated monitoring array of the first abnormal device to perform device status monitoring and environmental monitoring, and obtains status monitoring data and environmental monitoring data. The first abnormal device is any abnormal device in the set of abnormal devices. Based on the device model, performance parameters and environmental monitoring data of the first abnormal device, an associated fault feature library and an associated fault prediction model are determined. The status monitoring data are input into the associated fault feature library and the associated fault prediction model respectively. The predicted fault type is determined by fusing the output results. A first fault diagnosis result is generated based on the predicted fault type and the location coordinates of the first abnormal device. A device fault diagnosis report for the target area is generated based on the first fault diagnosis result.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By collecting equipment attribute information of distributed photovoltaic (PV) devices within the target area, an equipment attribute set is obtained. This set includes a set of equipment models, performance parameters, installed capacity, and location coordinates. Based on predetermined power generation influencing factors and the location coordinate set, environmental change analysis is performed using regional environmental monitoring logs. Cluster analysis is then conducted on the distributed PV devices based on the environmental change feature set, the equipment model set, and the performance parameter set to identify multiple equipment sets. Under preset monitoring nodes, multiple power generation sets from these equipment sets are obtained. Based on these power generation sets and the installed capacity set, abnormal power generation is identified to determine abnormal equipment sets. Finally, a first abnormal device is selected from these abnormal equipment sets. Activate the associated monitoring array of the first abnormal device to perform device status monitoring and environmental monitoring, and obtain status monitoring data and environmental monitoring data, wherein the first abnormal device is any one of the abnormal devices in the abnormal device set; determine the associated fault feature library and associated fault prediction model by matching the device model, performance parameters and environmental monitoring data of the first abnormal device; input the status monitoring data into the associated fault feature library and associated fault prediction model respectively; determine the predicted fault type based on the fusion of the output results; generate a first fault diagnosis result based on the predicted fault type and the location coordinates of the first abnormal device; and generate a device fault diagnosis report for the target area based on the first fault diagnosis result. In other words, by clustering distributed photovoltaic (PV) devices based on environmental change feature sets, multiple initial device sets are obtained. Then, a second clustering is performed on these initial device sets based on device model sets and performance parameter sets, further defining multiple device sets. Next, abnormal power generation is identified based on multiple power generation sets across these device sets, identifying anomalous device sets. This allows for timely detection of abnormal operating states of individual devices, improving the convenience, accuracy, and efficiency of anomalous device identification. Furthermore, the associated monitoring array of anomalous devices is activated to acquire device status monitoring data and environmental monitoring data. Then, based on the device model, performance parameters, and environmental monitoring data of the anomalous devices, an associated fault feature library and associated fault prediction model are determined for fault prediction. The predicted fault type is determined by fusing the output results. Finally, a device fault diagnosis report for the target area is generated based on the predicted fault type and the location coordinates of the anomalous devices. This improves the convenience and accuracy of identifying abnormal device operating states and enables intelligent and automated remote fault diagnosis, enhancing the efficiency, accuracy, and reliability of fault diagnosis. Ultimately, this achieves the technical effect of ensuring the safe and stable operation of the PV power generation system and improving operational efficiency.

[0008] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating a remote fault diagnosis method for photovoltaic power generation equipment according to this application;

[0011] Figure 2 This is a schematic diagram illustrating the process of determining multiple device sets in a remote fault diagnosis method for photovoltaic power generation equipment according to this application. Detailed Implementation

[0012] This application provides a remote fault diagnosis method for photovoltaic (PV) power generation equipment. This method addresses the technical problem in existing regional distributed PV power generation systems where the large number and dispersed nature of PV devices makes it difficult to promptly and accurately detect abnormal operating states of individual devices. Consequently, timely fault diagnosis and maintenance measures cannot be taken, resulting in poor operational efficiency, stability, and safety of the PV power generation system. The method improves the convenience and accuracy of identifying abnormal operating states and enables intelligent and automated remote fault diagnosis, thereby enhancing the efficiency, accuracy, and reliability of fault diagnosis. Ultimately, this method ensures the safe and stable operation of the PV power generation system and improves its operational efficiency.

[0013] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0014] For examples, please refer to the appendix. Figure 1This application provides a remote fault diagnosis method for photovoltaic power generation equipment, the method specifically including the following steps:

[0015] Step 1: Collect equipment attribute information of distributed photovoltaic equipment in the target area and obtain the equipment attribute set, which includes equipment model set, performance parameter set, installed capacity set and location coordinate set.

[0016] Specifically, by accessing the database or operation and maintenance management system of the distributed photovoltaic (PV) system in the target area, equipment attribute information of all PV devices in the target area is collected. This equipment attribute information includes equipment model, performance parameters, installed capacity, and location coordinates. The equipment model includes equipment type and working principle, such as different working principles of solar panels. Performance parameters include output voltage, output current, output power, and conversion efficiency. Obtaining this set of equipment attributes provides data support for the next step of equipment clustering analysis.

[0017] Step 2: Based on the predetermined power generation influencing factors and the location coordinate set, call the regional environmental monitoring logs to perform environmental change analysis, and perform cluster analysis on the distributed photovoltaic equipment according to the environmental change feature set, the equipment model set, and the performance parameter set to determine multiple equipment sets.

[0018] Specifically, the predetermined factors affecting power generation are obtained, including sunlight intensity, ambient temperature, and ambient humidity. These factors directly affect the power generation efficiency and output power of photovoltaic (PV) equipment. For example, higher sunlight intensity results in greater current and voltage generated by the PV modules, leading to higher power generation efficiency; excessive humidity can cause condensation or corrosion on the PV panel surface, affecting power generation efficiency. Next, based on the predetermined factors affecting power generation and the set of location coordinates, regional environmental monitoring logs are retrieved to perform environmental change analysis, i.e., to determine the environmental change characteristics of individual PV devices and obtain an environmental change feature set, where each environmental change feature corresponds one-to-one with the device's location coordinates.

[0019] Then, based on the environmental change feature set, a first cluster analysis is performed on the distributed photovoltaic (PV) devices, grouping any two PV devices whose environmental feature deviations are within a preset deviation threshold into one class, resulting in multiple initial device sets. Further, a second cluster analysis is performed on these initial device sets based on the device model set and the performance parameter set, grouping initial devices with the same or similar device models and performance parameters into one class, resulting in multiple device sets. Within each device set, the power generation environment and device information of multiple devices can be considered identical. This clustering process, which identifies multiple device sets, provides a basis for subsequent abnormal device status identification and improves the convenience and accuracy of identifying abnormal devices within the same device set, thereby enhancing the accuracy, reliability, and efficiency of abnormal device detection.

[0020] Step 3: Under the preset monitoring node, acquire multiple power generation sets of multiple device sets, identify abnormal power generation based on the multiple power generation sets and the installed capacity set, and determine the abnormal device set.

[0021] Specifically, under preset monitoring nodes, power generation data is collected from photovoltaic devices in multiple device sets, resulting in multiple power generation sets. The preset monitoring nodes are the time points for data collection and can be set according to actual conditions, such as setting a monitoring node every 10 minutes. Next, based on the multiple power generation sets and the installed capacity set, the unit power generation of individual devices is calculated, and abnormal unit power generation within the same power generation set is identified. Abnormal unit power generation refers to unit power generation that deviates significantly from other unit power generation in the same power generation set. Then, the photovoltaic devices corresponding to the abnormal unit power generation are marked as abnormal, resulting in an abnormal device set. By constructing device sets for abnormal device status identification, this method is simple to operate and thus improves the convenience and efficiency of abnormal device status identification.

[0022] Step 4: Select the first abnormal device from the set of abnormal devices, activate the associated monitoring array of the first abnormal device to perform device status monitoring and environmental monitoring, and obtain status monitoring data and environmental monitoring data. The first abnormal device is any one of the abnormal devices in the set of abnormal devices.

[0023] Specifically, a first abnormal device is selected from the set of abnormal devices, where the first abnormal device is any one of the abnormal devices in the set. The associated monitoring array of the first abnormal device is then activated to perform device status monitoring and environmental monitoring. By activating the associated monitoring array after identifying the abnormal device, unnecessary monitoring resource consumption can be reduced, and the rational utilization rate of monitoring and computing resources can be improved. The associated monitoring array includes a device status monitoring array and an environmental parameter monitoring array. The device status monitoring indicators include at least the solar panel temperature, voltage, current, and output power. The environmental parameter monitoring indicator set includes at least weather type, temperature, humidity, wind speed, and wind direction, thereby obtaining the status monitoring data and environmental monitoring data of the first abnormal device.

[0024] Step 5: Based on the equipment model, performance parameters and environmental monitoring data of the first abnormal device, determine the associated fault feature library and associated fault prediction model. Input the status monitoring data into the associated fault feature library and associated fault prediction model respectively, and determine the predicted fault type based on the output results.

[0025] Specifically, the associated fault feature library and associated fault prediction model are further determined by matching the equipment model, performance parameters and environmental monitoring data of the first abnormal equipment. The equipment model, performance parameters, environmental monitoring data and associated fault feature library and associated fault prediction model of the first abnormal equipment are in a one-to-one correspondence. The associated fault feature library includes multiple fault feature matrix sets corresponding to multiple fault types. The associated fault prediction model is constructed based on a BP neural network and is used to predict faults based on equipment status characteristics.

[0026] The condition monitoring data is then input into a fault feature library and a fault prediction model. The condition monitoring data is compared with multiple fault feature matrix sets in the fault feature library, and fault types with a comprehensive similarity greater than a predetermined similarity threshold are selected as the first predicted fault type, outputting a first predicted fault type set. The fault prediction model then intelligently predicts the condition monitoring data, outputting a second predicted fault type set. Finally, the first and second predicted fault type sets are intersected, and the result is selected as the predicted fault type. That is, the same predicted fault type is selected from both sets, resulting in the predicted fault type. By performing an intersection operation on the first and second predicted fault type sets and selecting the result as the predicted fault type, the advantages of different prediction models can be combined, reducing the probability of false alarms and improving the accuracy and reliability of the predicted fault type.

[0027] Step 6: Generate a first fault diagnosis result based on the predicted fault type and the location coordinates of the first abnormal device, and generate a device fault diagnosis report for the target area based on the first fault diagnosis result.

[0028] Specifically, by integrating the predicted fault type and the location coordinates of the first abnormal device, a first fault diagnosis result is generated. Then, using the same method as the method used to obtain the first fault diagnosis result, intelligent fault analysis is performed on other abnormal devices in the abnormal device set to obtain multiple fault diagnosis results. Finally, based on the multiple fault diagnosis results, an equipment fault diagnosis report for the target area is generated. This enables intelligent and automated remote fault diagnosis of equipment, providing a basis for subsequent equipment maintenance.

[0029] This method provides a remote fault diagnosis approach for photovoltaic (PV) power generation equipment, addressing the technical problem of poor system efficiency, stability, and security in existing regional distributed PV power generation systems. This is because the large number and dispersed nature of PV equipment makes it difficult to promptly and accurately detect abnormal operating states of individual devices, hindering timely fault diagnosis and maintenance. First, equipment attribute information of distributed PV equipment within the target area is collected to obtain an equipment attribute set, which includes a set of equipment models, performance parameters, installed capacity, and location coordinates. Then, based on predetermined power generation influencing factors and the location coordinate set, regional environmental monitoring logs are used to analyze environmental changes. Cluster analysis is performed on the distributed PV equipment based on the environmental change feature set, the equipment model set, and the performance parameter set to identify multiple equipment sets. Next, under preset monitoring nodes, multiple power generation sets of these equipment sets are obtained. Abnormal power generation is identified based on these power generation sets and the installed capacity set to determine abnormal equipment sets. Finally, a first abnormal device is selected from the abnormal equipment sets. The system prepares to activate the associated monitoring array of the first abnormal device to perform device status monitoring and environmental monitoring, obtaining status monitoring data and environmental monitoring data. The first abnormal device is any one of the abnormal devices in the set of abnormal devices. Further, based on the device model, performance parameters, and environmental monitoring data of the first abnormal device, an associated fault feature library and an associated fault prediction model are determined. The status monitoring data are input into the associated fault feature library and the associated fault prediction model, respectively. The predicted fault type is determined by fusing the output results. Finally, a first fault diagnosis result is generated based on the predicted fault type and the location coordinates of the first abnormal device. Based on the first fault diagnosis result, a device fault diagnosis report for the target area is generated. By performing a primary clustering of distributed photovoltaic (PV) devices based on environmental change feature sets, multiple initial device sets are obtained. Then, a secondary clustering is performed on these initial device sets based on device model sets and performance parameter sets, further defining multiple device sets. Next, abnormal power generation is identified based on multiple power generation sets across these device sets, thus identifying abnormal device sets. This allows for timely detection of abnormal operating states of individual devices, improving the convenience, accuracy, and efficiency of abnormal device identification. Furthermore, the associated monitoring array of abnormal devices is activated to acquire device status monitoring data and environmental monitoring data. Then, based on the device model, performance parameters, and environmental monitoring data of the abnormal devices, an associated fault feature library and associated fault prediction model are determined for fault prediction. The predicted fault type is determined based on the fusion of the output results. Finally, a device fault diagnosis report for the target area is generated based on the predicted fault type and the location coordinates of the abnormal devices. This approach improves the convenience and accuracy of identifying abnormal device operating states and enables intelligent and automated remote fault diagnosis, improving the efficiency, accuracy, and reliability of fault diagnosis. Ultimately, this achieves the technical effect of ensuring the safe and stable operation of PV power generation systems and improving operational efficiency.

[0030] Furthermore, cluster analysis is performed on the distributed photovoltaic equipment based on the environmental change characteristic set, the equipment model set, and the performance parameter set, as shown in the appendix. Figure 2 As shown, step two of this application includes:

[0031] Information is extracted from regional environmental monitoring logs based on predetermined power generation influencing factors and the location coordinate set to obtain an environmental change feature set. The predetermined power generation influencing factors include light intensity, ambient temperature, and ambient humidity. Distributed photovoltaic equipment is clustered once based on predetermined environmental parameter thresholds and the environmental change feature set to obtain multiple initial equipment sets. Multiple initial equipment sets are then further divided based on the equipment model set and the performance parameter set to determine multiple equipment sets.

[0032] Specifically, the predetermined power generation influencing factors are obtained, including solar irradiance, ambient temperature, and ambient humidity. Next, the regional environmental monitoring log is accessed. This log is a document or system that records and stores environmental monitoring data within a specific area, typically used to track and record environmental factors related to photovoltaic power generation systems, such as solar irradiance, temperature, and humidity. These factors significantly impact the performance and power generation efficiency of photovoltaic equipment. Based on the predetermined power generation influencing factors and the set of location coordinates, information is extracted from the regional environmental monitoring log to obtain a set of environmental change features. These features characterize the comprehensive power generation environment of the photovoltaic equipment and correspond one-to-one with the location coordinates.

[0033] Further, based on a predetermined environmental parameter threshold and the environmental change feature set, the distributed photovoltaic (PV) devices are clustered. PV devices whose feature deviation values ​​between any two environmental change features in the environmental change feature set meet the predetermined environmental parameter threshold are grouped into one class, resulting in multiple initial device sets. This step groups PV devices with similar environmental conditions together, aiding in the identification of PV devices with similar environmental conditions. Next, the multiple initial device sets are further divided according to the device model set and the performance parameter set. Initial devices with the same or similar device models and performance parameters from the same initial device set are added to the same device set, resulting in multiple device sets. By performing cluster analysis on distributed PV devices based on environmental change characteristics, device models, and performance parameters, the granularity and accuracy of PV device clustering can be improved, thereby enhancing the precision and reliability of device set construction and providing a reliable basis for subsequent identification of abnormal devices.

[0034] Furthermore, to obtain the environmental change feature set, this application also includes the following steps:

[0035] A first location coordinate is selected from the location coordinate set. Information is extracted from the regional environmental monitoring log based on a predetermined time period, predetermined power generation influencing factors, and the first location coordinate to obtain a first environmental monitoring dataset. The first environmental monitoring dataset is preprocessed to obtain a first standard environmental monitoring dataset. The mean of the first standard environmental monitoring dataset is calculated to obtain a first environmental change feature. The first environmental change feature is added to the environmental change feature set.

[0036] Specifically, firstly, a first location coordinate is selected from the set of location coordinates, where the first location coordinate is any one of the coordinates in the set. Then, information is extracted from the regional environmental monitoring logs based on a predetermined time period, predetermined power generation influencing factors, and the first location coordinate. The predetermined time period is the time period for data extraction, which can be set according to actual conditions, such as the most recent month, to obtain a first environmental monitoring dataset. Next, the first environmental monitoring dataset undergoes data preprocessing, including data cleaning, missing data filling, data transformation, and data integration. For example, checking for duplicate, abnormal, or erroneous data points in the dataset and deleting or correcting them to ensure data integrity and accuracy; if missing values ​​exist in the dataset, interpolation methods (such as linear interpolation, spline interpolation, etc.) or deletion methods can be used to handle them, selecting the most suitable method to maintain data continuity and integrity; the numerical data in the dataset is converted into a format suitable for model analysis, such as standardization or normalization. Then, the mean of multiple first standard environmental monitoring data in the first standard environmental monitoring dataset is calculated to obtain the first environmental change feature. Finally, the first environmental change feature is added to the environmental change feature set.

[0037] Furthermore, based on predetermined environmental parameter thresholds and the environmental change feature set, this application also includes the following steps for clustering distributed photovoltaic devices:

[0038] The predetermined environmental parameter thresholds include a light intensity threshold, a temperature threshold, and a humidity threshold. A first environmental change feature and a second environmental change feature are randomly selected from the set of environmental change features, where the first and second environmental change features are any two environmental change features in the set. Feature deviation analysis is performed on the first and second environmental change features to obtain environmental feature deviation values, where the environmental feature deviation values ​​include light intensity deviation, temperature deviation, and humidity deviation. When the light intensity deviation meets the light intensity threshold, the temperature deviation meets the temperature threshold, and the humidity deviation meets the humidity threshold, the first photovoltaic device and the second photovoltaic device are added to the same initial device set, where the first photovoltaic device is the photovoltaic device corresponding to the first environmental change feature, and the second photovoltaic device is the photovoltaic device corresponding to the second environmental change feature. Iterative analysis is performed until all environmental change features are compared, resulting in multiple initial device sets.

[0039] Specifically, the predetermined environmental parameter thresholds characterize the range of variation of environmental parameters and can be determined based on historical environmental data analysis of the target area, including light intensity thresholds, temperature thresholds, and humidity thresholds. Next, a first environmental change feature and a second environmental change feature are randomly selected from the set of environmental change features, wherein the first and second environmental change features are any two environmental change features from the set. Furthermore, feature deviations are calculated between the first and second environmental change features to obtain environmental feature deviation values, which include light intensity deviation, temperature deviation, and humidity deviation.

[0040] The environmental feature deviation values ​​are judged based on the predetermined environmental parameter thresholds. When the light intensity deviation is less than or equal to the light intensity threshold, the temperature deviation is less than or equal to the temperature threshold, and the humidity deviation is less than or equal to the humidity threshold, the first photovoltaic device and the second photovoltaic device are added to the same initial device set. The first photovoltaic device corresponds to the first environmental change feature, and the second photovoltaic device corresponds to the second environmental change feature. Iterative analysis is performed using the same method until all environmental change features in the environmental change feature set have been compared, resulting in multiple initial device sets.

[0041] Furthermore, based on multiple power generation sets and the installed capacity set, abnormal power generation identification is performed. Step three of this application includes:

[0042] A first power generation set is randomly selected from multiple power generation sets. A first installed capacity set is obtained based on the first power generation set, wherein the first power generation and the first installed capacity correspond one-to-one. Based on the first power generation set and the first installed capacity set, the unit power generation is calculated to obtain a first unit power generation set. The average value of the first unit power generation set is calculated to obtain the first unit power generation average value. Based on the first unit power generation average value, the deviation of the first unit power generation set is calculated. Photovoltaic devices with a unit power generation deviation greater than a predetermined power generation deviation threshold are set as abnormal devices and added to the abnormal device set.

[0043] Specifically, firstly, a first power generation set is randomly selected from multiple power generation sets, where the first power generation set is any one of the multiple power generation sets. Next, based on the first power generation set, it is matched with the installed capacity set to obtain a first installed capacity set for the first power generation set, where there is a one-to-one correspondence between the first power generation and the first installed capacity. Then, based on the first power generation set and the first installed capacity set, unit power generation is calculated, where unit power generation is the ratio of power generation to installed capacity, to obtain a first unit power generation set. Finally, the average of multiple first unit power generation values ​​in the first unit power generation set is calculated to obtain the first unit power generation average.

[0044] Based on the average power generation of the first unit, the deviation of multiple first unit power generation in the first unit power generation set is calculated to obtain multiple unit power generation deviations. The unit power generation deviation is the difference between the average first unit power generation and the first unit power generation. Then, the photovoltaic equipment corresponding to the unit power generation deviation that is greater than the predetermined power generation deviation threshold is set as abnormal equipment. The predetermined power generation deviation threshold can be set according to the actual situation. Multiple abnormal equipment are obtained to construct an abnormal equipment set.

[0045] Furthermore, step four of this application includes:

[0046] A set of equipment status monitoring indicators and a set of environmental monitoring indicators are obtained. The set of equipment status monitoring indicators includes at least solar panel temperature, voltage, current, and output power, while the set of environmental monitoring indicators includes at least weather type, temperature, humidity, wind speed, and wind direction. Based on the set of equipment status monitoring indicators, correlation analysis is performed sequentially on the equipment status monitoring indicators and equipment faults. Equipment status monitoring indicators with correlation coefficients greater than a first predetermined threshold are selected as high-frequency status monitoring indicators, thus obtaining a high-frequency status monitoring indicator set. Based on the set of environmental monitoring indicators, correlation analysis is performed sequentially on the environmental monitoring indicators and equipment faults. Environmental monitoring indicators with correlation coefficients greater than a second predetermined threshold are selected as high-frequency environmental monitoring indicators, thus obtaining a high-frequency environmental monitoring indicator set. A correlation monitoring array is constructed based on the high-frequency status monitoring indicator set and the high-frequency environmental monitoring indicator set.

[0047] Specifically, a set of equipment condition monitoring indicators and a set of environmental monitoring indicators are obtained. The equipment condition monitoring indicator set includes at least solar panel temperature, voltage, current, and output power, while the environmental monitoring indicator set includes at least weather type, temperature, humidity, wind speed, and wind direction. Additional monitoring indicators can be added as needed. Next, based on the equipment condition monitoring indicator set, a correlation analysis is performed between the equipment condition monitoring indicators and equipment failures. For example, multiple sample equipment failure events can be obtained, and the frequency and impact of the equipment condition monitoring indicators in these events are statistically analyzed. Higher frequency and greater impact indicate a stronger correlation between the equipment condition monitoring indicators and equipment failures, resulting in a higher correlation coefficient. This yields multiple correlation coefficients for the various equipment condition monitoring indicators. Equipment condition monitoring indicators with correlation coefficients greater than a first predetermined threshold are then designated as high-frequency condition monitoring indicators. The first predetermined threshold is set based on actual conditions. Similarly, based on the environmental monitoring indicator set, a correlation analysis is performed between the environmental monitoring indicators and equipment failures. Environmental monitoring indicators with correlation coefficients greater than a second predetermined threshold are designated as high-frequency environmental monitoring indicators. This yields a high-frequency environmental monitoring indicator set.

[0048] Finally, based on the high-frequency condition monitoring index set and the high-frequency environmental monitoring index set, a correlation monitoring array is designed, wherein the correlation monitoring array should include all the sensors and monitoring equipment required for the high-frequency condition monitoring index and the environmental monitoring index; then, the sensors and monitoring equipment required for the correlation monitoring array are installed in the distributed photovoltaic system, while ensuring that the sensors and monitoring equipment can accurately collect the high-frequency condition monitoring index and the environmental monitoring index.

[0049] Furthermore, based on the equipment model, performance parameters, and environmental monitoring data of the first abnormal device, a related fault feature library and a related fault prediction model are determined. Step five of this application includes:

[0050] Multiple high-frequency environmental parameter thresholds are obtained, and these thresholds are divided according to multiple preset step sizes to determine multiple high-frequency environmental parameter interval sets. A parameter is randomly selected from each of these interval sets and combined to obtain multiple sample environmental parameter groups. A first sample environmental parameter group is selected from these groups. A first equipment model and a first performance parameter are obtained. Based on the first equipment model, the first performance parameter, and the first sample environmental parameter group, a first associated fault feature library and a first associated fault prediction model are constructed. Based on the classification decision principle, a fault prediction matching channel is constructed according to the mapping relationship between the first equipment model, the first performance parameter, the first sample environmental parameter group, the first associated fault feature library, and the first associated fault prediction model. The equipment model, performance parameters, and environmental monitoring data of the first abnormal equipment are input into the fault prediction matching channel for matching to determine the associated fault feature library and the associated fault prediction model.

[0051] Specifically, multiple high-frequency environmental parameter thresholds are obtained. These thresholds are then divided according to multiple preset step sizes, which can be set based on the type and fluctuation range of the environmental parameter. For example, a temperature step size of 0.1 degrees Celsius is used to determine multiple high-frequency environmental parameter interval sets. Next, a parameter interval is randomly selected from these interval sets and combined to obtain multiple sample environmental parameter groups. Finally, a first sample environmental parameter group is selected from these groups, where the first sample environmental parameter group is any one of the multiple sample environmental parameter groups.

[0052] Next, the first equipment model and first performance parameter are obtained, where the first equipment model is any one of the equipment models of multiple photovoltaic devices in the region, and the first performance parameter is any one of the equipment performance parameters of multiple photovoltaic devices in the region. Then, a first associated fault feature library and a first associated fault prediction model are constructed based on the first equipment model, the first performance parameter, and the first sample environmental parameter group. Further, a mapping relationship is established between the first equipment model, the first performance parameter, the first sample environmental parameter group, the first associated fault feature library, and the first associated fault prediction model. Based on the classification decision principle, a fault prediction matching channel is constructed according to the mapping relationship. That is, the first equipment model, the first performance parameter, and the first sample environmental parameter group are used as child nodes, and the first associated fault feature library and the first associated fault prediction model with the mapping relationship are used as leaf nodes to construct the fault prediction matching channel. Finally, the equipment model, performance parameter, and environmental monitoring data of the first abnormal equipment are input into the fault prediction matching channel for matching to determine the associated fault feature library and the associated fault prediction model.

[0053] Furthermore, based on the first equipment model, the first performance parameters, and the first sample environmental parameter group, a first associated fault feature library is constructed. This application also includes the following steps:

[0054] Constrained by a first equipment model, a first performance parameter, and a first sample environmental parameter group, data retrieval is performed via network connection to obtain a first sample dataset. This first sample dataset includes a first sample fault type set and multiple first sample fault feature sets, with a one-to-one relationship between the first sample fault feature sets and the first sample fault types. A predetermined similarity dimensionality reduction algorithm is used to perform similarity dimensionality reduction on the multiple first sample fault feature sets to obtain multiple first standard sample fault feature sets. The minimum similarity of the first standard sample fault features is less than a first similarity threshold, and the average similarity is less than a second similarity threshold, where the first similarity threshold is less than the second similarity threshold. Multiple first standard sample fault feature matrix sets are generated based on the multiple first standard sample fault feature sets. A first associated fault feature library is constructed based on the first sample fault type set and the multiple first standard sample fault feature matrix sets.

[0055] Specifically, the first equipment model, the first performance parameters, and the first sample environmental parameter group are then used as search constraints. Data retrieval is performed based on big data technology to obtain the first sample dataset. The first sample dataset includes a first sample fault type set and multiple first sample fault feature sets, and the first sample fault feature set and the first sample fault type have a one-to-one relationship.

[0056] Next, based on a predetermined similarity dimensionality reduction algorithm, a sample fault feature set is randomly selected from multiple first sample fault feature sets, and a sample fault feature is randomly selected from the sample fault feature set. Then, the sample fault feature and other sample fault features in the sample fault feature set are compared for similarity. For example, the existing cosine similarity comparison algorithm can be used to compare similarities, and multiple comparison similarities are obtained. The average of the multiple comparison similarities is calculated to obtain the average comparison similarity. Then, a first similarity threshold and a second similarity threshold are set, wherein the first similarity threshold is less than the second similarity threshold. When the minimum similarity among the multiple comparison similarities is less than the first similarity threshold and the average comparison similarity is less than the second similarity threshold, it indicates that the overall deviation between the sample fault feature and other sample fault features in the same sample fault feature set is large. Then, the sample fault feature is set as a standard sample fault feature, and multiple first standard sample fault feature sets are obtained.

[0057] Finally, multiple first standard sample fault feature matrix sets are generated based on multiple first standard sample fault feature sets, and a first associated fault feature library is constructed based on the mapping relationship between the first sample fault type and the first standard sample fault feature matrix set, according to the first sample fault type set and multiple first standard sample fault feature matrix sets.

[0058] Furthermore, based on the first equipment model, the first performance parameters, and the first sample environmental parameter group, a first associated fault prediction model is constructed. This application also includes the following steps:

[0059] Constrained by a first equipment model, a first performance parameter, and a first sample environmental parameter set, data retrieval is performed via network connection to obtain a second sample dataset. This second sample dataset includes multiple second sample fault type sets and multiple second sample fault feature sets, with a many-to-many relationship between the second sample fault features and the second sample fault types. The second sample dataset is divided into Q equal parts, and a first sample set is constructed by selecting Q times with replacement. This process is repeated Q times to generate Q sample sets. Using the second sample fault feature set as input and the second sample fault type set as supervision, a BP neural network is trained and cross-validated using the Q sample sets to obtain Q convergent fault prediction units. A first associated fault prediction model is then built based on these Q convergent fault prediction units, where the output of the first associated fault prediction model is the mode of the outputs of the Q convergent fault prediction units.

[0060] Specifically, using the first equipment model, first performance parameters, and first sample environmental parameter set as constraints, data retrieval is performed online based on big data technology to obtain a second sample dataset. This second sample dataset includes multiple sets of second sample fault types and multiple sets of second sample fault features, with a many-to-many relationship between the second sample fault features and the second sample fault types. The second sample dataset is then divided into Q equal parts, and a first sample set is constructed by selecting Q times with replacement from these Q parts. This selection process with replacement is repeated iteratively Q times, generating a new sample set after each iteration, resulting in Q sample sets.

[0061] Next, an initial fault prediction unit is constructed based on a backpropagation (BP) neural network. This initial fault prediction unit is a neural network model that can be iteratively optimized in machine learning. The input data of the initial fault prediction unit consists of multiple fault features, and the output data consists of multiple fault types. Then, using a second sample set of fault features as input and a second sample set of fault types as supervision, the BP neural network is trained and cross-validated using Q sample sets. For example, supervised training is performed using gradient descent and a loss function. After a preset number of training iterations, the training sets are swapped for cross-validation until the output accuracy meets the expected index, resulting in Q convergent fault prediction units. Finally, a first associated fault prediction model is built based on the Q convergent fault prediction units. The output of the first associated fault prediction model is the mode of the outputs of the Q convergent fault prediction units. By building the first associated fault prediction model based on Q convergent fault prediction units, prediction errors can be reduced, and the accuracy and reliability of fault prediction can be improved.

[0062] Furthermore, the condition monitoring data is input into the associated fault feature library and the associated fault prediction model respectively, and the predicted fault type is determined based on the fusion of the output results. Step five of this application includes:

[0063] The status monitoring data is compared with multiple sample fault feature matrix sets in the associated fault feature library to generate multiple similarity sets. The mean of the multiple similarity sets is calculated, and the fault types with a mean similarity greater than a predetermined abnormal similarity threshold are selected as predicted fault types, and a first predicted fault type set is output. The status monitoring data is input into the associated fault prediction model to generate a second predicted fault type set. The intersection of the first predicted fault type set and the second predicted fault type set is performed to obtain the predicted fault type.

[0064] Specifically, the condition monitoring data is compared with multiple sample fault feature matrix sets in the associated fault feature library using a similarity comparison algorithm. Commonly used similarity comparison algorithms include cosine similarity algorithm and Euclidean distance, and an appropriate similarity comparison algorithm can be selected according to the actual situation, outputting multiple similarity sets. Then, the mean of multiple similarity sets is calculated, and the fault types with a mean similarity greater than a predetermined abnormal similarity threshold are selected as predicted fault types, outputting a first predicted fault type set. On the other hand, the condition monitoring data is input into the associated fault prediction model, outputting a second predicted fault type set. Finally, the intersection of the first and second predicted fault type sets is performed, that is, the predicted fault types that are the same in the first and second predicted fault type sets are output to obtain the predicted fault type.

[0065] In summary, the remote fault diagnosis method for photovoltaic power generation equipment provided in this application has the following technical effects:

[0066] 1. This method involves clustering distributed photovoltaic (PV) devices based on environmental change characteristics to obtain multiple initial device sets. Then, a second clustering is performed on these initial sets based on device model and performance parameter sets to determine multiple device sets. Next, abnormal power generation is identified based on multiple power generation sets across these device sets to determine abnormal device sets. This allows for timely detection of abnormal operating states of individual devices, improving the convenience, accuracy, and efficiency of abnormal device identification. Furthermore, the associated monitoring array of the abnormal devices is activated to acquire device status monitoring data and environmental monitoring data. Then, based on the device model, performance parameters, and environmental monitoring data of the abnormal devices, an associated fault feature library and associated fault prediction model are determined for fault prediction. The predicted fault type is determined based on the fusion of the output results. Finally, a device fault diagnosis report for the target area is generated based on the predicted fault type and the location coordinates of the abnormal devices. This method improves the convenience and accuracy of identifying abnormal device operating states and enables intelligent and automated remote fault diagnosis, improving the efficiency, accuracy, and reliability of fault diagnosis. Ultimately, this achieves the technical effect of ensuring the safe and stable operation of the PV power generation system and improving its operational efficiency.

[0067] 2. By performing clustering to determine multiple device sets, a basis is provided for subsequent identification of abnormal device status. At the same time, it can improve the convenience and accuracy of identifying abnormal devices in the same device set, thereby improving the accuracy, reliability and efficiency of abnormal device detection.

[0068] 3. By activating the associated monitoring array after identifying abnormal devices, unnecessary monitoring resource consumption can be reduced, and the rational utilization rate of monitoring and computing resources can be improved.

[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0070] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A remote fault diagnosis method for photovoltaic power generation equipment, characterized in that, The method includes: Collect equipment attribute information of distributed photovoltaic equipment in the target area to obtain equipment attribute set, which includes equipment model set, performance parameter set, installed capacity set and location coordinate set; Based on the predetermined power generation influencing factors and the location coordinate set, the regional environmental monitoring logs are called to perform environmental change analysis. Based on the environmental change feature set, the equipment model set, and the performance parameter set, the distributed photovoltaic equipment is clustered to determine multiple equipment sets. Under a preset monitoring node, multiple power generation sets of multiple equipment sets are acquired, and abnormal power generation is identified based on the multiple power generation sets and the installed capacity set to determine the abnormal equipment set; Select a first abnormal device from the set of abnormal devices, activate the associated monitoring array of the first abnormal device to perform device status monitoring and environmental monitoring, and obtain status monitoring data and environmental monitoring data. The first abnormal device is any one of the abnormal devices in the set of abnormal devices. Based on the equipment model, performance parameters and environmental monitoring data of the first abnormal device, a related fault feature library and a related fault prediction model are determined. The status monitoring data are then input into the related fault feature library and the related fault prediction model respectively. The predicted fault type is determined by fusing the output results. A first fault diagnosis result is generated based on the predicted fault type and the location coordinates of the first abnormal device, and a device fault diagnosis report for the target area is generated based on the first fault diagnosis result.

2. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 1, characterized in that, Cluster analysis of distributed photovoltaic equipment is performed based on the environmental change characteristic set, the equipment model set, and the performance parameter set, including: Information is extracted from the regional environmental monitoring logs based on the predetermined power generation influencing factors and the location coordinate set to obtain an environmental change feature set, wherein the predetermined power generation influencing factors include light intensity, ambient temperature and ambient humidity; Based on predetermined environmental parameter thresholds and the environmental change feature set, the distributed photovoltaic devices are clustered once to obtain multiple initial device sets; Multiple initial device sets are further divided based on the device model set and the performance parameter set to determine multiple device sets.

3. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 2, characterized in that, The environmental change feature set is obtained, including: A first location coordinate is selected from the set of location coordinates. Information is extracted from the regional environmental monitoring log based on a predetermined time period, predetermined power generation influencing factors, and the first location coordinate to obtain a first environmental monitoring dataset. The first environmental monitoring dataset is preprocessed to obtain a first standard environmental monitoring dataset, and the mean of the first standard environmental monitoring dataset is calculated to obtain a first environmental change feature. The first environmental change feature is then added to the environmental change feature set.

4. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 2, characterized in that, Based on predetermined environmental parameter thresholds and the environmental change feature set, a clustering process is performed on distributed photovoltaic devices, including: The predetermined environmental parameter thresholds include light intensity threshold, temperature threshold, and humidity threshold; Randomly select a first environmental change feature and a second environmental change feature from the set of environmental change features, wherein the first environmental change feature and the second environmental change feature are any two environmental change features from the set of environmental change features; A feature deviation analysis is performed on the first environmental change feature and the second environmental change feature to obtain environmental feature deviation values, wherein the environmental feature deviation values ​​include light intensity deviation, temperature deviation and humidity deviation; When the light intensity deviation meets the light intensity threshold, the temperature deviation meets the temperature threshold, and the humidity deviation meets the humidity threshold, the first photovoltaic device and the second photovoltaic device are added to the same initial device set, wherein the first photovoltaic device is the photovoltaic device corresponding to the first environmental change characteristic, and the second photovoltaic device is the photovoltaic device corresponding to the second environmental change characteristic. Iterative analysis is performed until the environmental change characteristics are compared, resulting in multiple initial device sets.

5. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 1, characterized in that, Abnormal power generation identification is performed based on multiple power generation sets and the installed capacity set, including: A first power generation set is randomly selected from multiple power generation sets, and a first installed capacity set is obtained based on the first power generation set, wherein the first power generation and the first installed capacity correspond one-to-one. The unit power generation is calculated based on the first power generation set and the first installed capacity set to obtain the first unit power generation set, and the average value of the first unit power generation set is calculated to obtain the first unit power generation average value. Based on the average power generation per unit, the deviation of the first unit power generation set is calculated, and the photovoltaic equipment corresponding to the power generation deviation per unit that exceeds the predetermined power generation deviation threshold is set as abnormal equipment and added to the abnormal equipment set.

6. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 1, characterized in that, Activate the associated monitoring array of the first abnormal device to perform device status monitoring and environmental monitoring. Prior to this, the following steps were also taken: Acquire a set of equipment status monitoring indicators and an environmental monitoring indicator set. The set of equipment status monitoring indicators shall include at least the temperature, voltage, current and output power of the solar panel, and the set of environmental monitoring indicators shall include at least the weather type, temperature, humidity, wind speed and wind direction. Based on the equipment condition monitoring index set, the correlation analysis between the equipment condition monitoring index and the equipment fault is carried out in sequence. The equipment condition monitoring index with a correlation coefficient greater than the first predetermined coefficient threshold is selected as the high-frequency condition monitoring index, and the high-frequency condition monitoring index set is obtained. Based on the environmental monitoring index set, the correlation analysis between environmental monitoring indexes and equipment failures is carried out sequentially. Environmental monitoring indexes with a correlation coefficient greater than the second predetermined coefficient threshold are selected as high-frequency environmental monitoring indexes, thus obtaining the high-frequency environmental monitoring index set. A correlation monitoring array is constructed based on the high-frequency state monitoring index set and the high-frequency environmental monitoring index set.

7. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 6, characterized in that, Based on the equipment model, performance parameters, and environmental monitoring data of the first abnormal device, a related fault feature library and a related fault prediction model are determined, including: Obtain multiple high-frequency environmental parameter thresholds, divide the multiple high-frequency environmental parameter thresholds according to multiple preset step sizes, and determine multiple high-frequency environmental parameter interval sets; A parameter is randomly selected from multiple high-frequency environmental parameter ranges and combined to obtain multiple sample environmental parameter groups, and the first sample environmental parameter group is selected from the multiple sample environmental parameter groups. Obtain the first equipment model and the first performance parameters, and construct the first associated fault feature library and the first associated fault prediction model based on the first equipment model, the first performance parameters and the first sample environmental parameter group; Based on the principle of classification decision, a fault prediction matching channel is constructed according to the mapping relationship between the first equipment model, the first performance parameter, the first sample environmental parameter group, the first associated fault feature library, and the first associated fault prediction model. The device model, performance parameters, and environmental monitoring data of the first abnormal device are input into the fault prediction matching channel for matching to determine the associated fault feature library and associated fault prediction model.

8. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 7, characterized in that, A first associated fault feature library is constructed based on the first equipment model, the first performance parameters, and the first sample environmental parameter group, including: Using the first equipment model, first performance parameters, and first sample environmental parameter group as constraints, data retrieval is performed online to obtain the first sample dataset. The first sample dataset includes a first sample fault type set and multiple first sample fault feature sets, and the first sample fault feature set and the first sample fault type have a one-to-one relationship. Based on a predetermined similarity dimensionality reduction algorithm, multiple first sample fault feature sets are subjected to similarity dimensionality reduction to obtain multiple first standard sample fault feature sets. Among them, the minimum similarity of the first standard sample fault features is less than the first similarity threshold and the average similarity is less than the second similarity threshold. The first similarity threshold is less than the second similarity threshold. Multiple first standard sample fault feature matrix sets are generated based on multiple first standard sample fault feature sets, and a first associated fault feature library is constructed based on the first sample fault type set and multiple first standard sample fault feature matrix sets.

9. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 7, characterized in that, A first associated fault prediction model is constructed based on the first equipment model, the first performance parameters, and the first sample environmental parameter group, including: Using the first equipment model, first performance parameters, and first sample environmental parameter group as constraints, data retrieval is performed online to obtain the second sample dataset. The second sample dataset includes multiple second sample fault type sets and multiple second sample fault feature sets, and the second sample fault features and second sample fault types have a many-to-many relationship. Divide the second sample dataset into Q equal parts, and select Q times with replacement to construct the first sample set. Iterate Q times to generate Q sample sets. Using the second sample fault feature set as input and the second sample fault type set as supervision, the BP neural network is trained and cross-validated by Q sample sets to obtain Q convergent fault prediction units. A first associated fault prediction model is built based on the Q convergent fault prediction units, where the output of the first associated fault prediction model is the mode of the output results of the Q convergent fault prediction units.

10. The remote fault diagnosis method for photovoltaic power generation equipment according to claim 8, characterized in that, The condition monitoring data is input into the associated fault feature library and the associated fault prediction model, respectively. Based on the fusion of the output results, the predicted fault type is determined, including: The status monitoring data is compared with multiple sample fault feature matrix sets in the associated fault feature library to generate multiple similarity sets. The mean of the multiple similarity sets is calculated, and the fault type with the mean similarity greater than the predetermined abnormal similarity threshold is selected as the predicted fault type, and the first predicted fault type set is output. The condition monitoring data is input into the associated fault prediction model, and a second set of predicted fault types is output. The predicted fault type is obtained by performing an intersection operation on the first predicted fault type set and the second predicted fault type set.

Images