Distributed photovoltaic characteristic clustering analysis method based on multi-source data

Abstract

The present invention discloses a distributed photovoltaic characteristic clustering analysis method based on multi-source data, which relates to the field of photovoltaic power generation operation and maintenance management technology. The method includes the following steps: for distributed photovoltaic equipment under the jurisdiction of a city, collecting the power generation curves of different photovoltaic equipment on N sunny days each day, and performing cluster analysis on the distributed photovoltaic equipment under the jurisdiction in combination with the characteristics of the photovoltaic equipment; according to the characteristics of the distributed photovoltaic equipment after the cluster analysis, respectively establishing databases, collecting the equipment power generation data of different photovoltaic equipment in different databases each day, and setting humidity sensors based on the geographical location of the equipment to collect humidity data. The present invention comprehensively applies data science and energy management technology to achieve refined and intelligent management of distributed photovoltaic equipment in the city. By unifying variables, it is easy to discover abnormalities in distributed photovoltaic equipment, thereby improving the management and maintenance effects of distributed photovoltaic equipment in the city.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation operation and maintenance management technology, and more specifically, to a distributed photovoltaic characteristic clustering analysis method based on multi-source data. Background Technology

[0002] Distributed photovoltaic (PV) systems refer to PV power generation systems deployed at or near user sites. These systems are typically used for self-consumption and can meet some or all of the electricity demand. They usually consist of solar panels, inverters, brackets, cables, etc., and are often installed on the roofs, walls, and ground of urban buildings.

[0003] However, environmental factors in cities are more complex. Dust and pollution conditions vary in different areas of the city, which can easily lead to dirt on the surface of exposed photovoltaic panels, thereby reducing the light absorption capacity of the photovoltaic panels and affecting power generation efficiency. Furthermore, since the photovoltaic devices in a distributed photovoltaic system are located in different and dispersed locations, managers cannot promptly determine the abnormal conditions of the photovoltaic panels and cannot carry out targeted maintenance on abnormal photovoltaic devices in a timely manner, resulting in problems of low practicality and functionality.

[0004] There are currently no effective solutions to the problems in the relevant technologies. Summary of the Invention

[0005] To address the problems in related technologies, this invention proposes a distributed photovoltaic characteristic clustering analysis method based on multi-source data to overcome the aforementioned technical problems in existing related technologies.

[0006] Therefore, the specific technical solution adopted by the present invention is as follows:

[0007] A distributed photovoltaic characteristic clustering analysis method based on multi-source data, the method includes the following steps:

[0008] S1. For distributed photovoltaic (PV) equipment under the city's jurisdiction, collect the power generation curves of different PV equipment on N sunny days each day, and perform cluster analysis on the distributed PV equipment under the jurisdiction based on the characteristics of the PV equipment.

[0009] S2. Based on the characteristics of distributed photovoltaic equipment after cluster analysis, establish databases respectively, collect the power generation data of different photovoltaic equipment in different databases every day, and at the same time, set up humidity sensors based on the geographical location of the equipment to collect humidity data.

[0010] S3. Based on the power generation of different photovoltaic devices within n periods in different databases, determine and analyze the causes of abnormality of abnormal photovoltaic devices in the same database, including dust accumulation and humidity effects.

[0011] S4. Mark the abnormal photovoltaic equipment in the database, and at the same time transmit the abnormal information to the relevant photovoltaic equipment manager's email address.

[0012] As a preferred implementation, S11, for distributed photovoltaic equipment under the jurisdiction of the city, collect the power generation data of different photovoltaic equipment for each day and each hour in N sunny weather conditions;

[0013] S12. Take the average power generation data of N photovoltaic devices in the same hour, and plot the representative power generation curves of different photovoltaic devices based on the average hourly power generation. The specific steps are as follows:

[0014] ;

[0015] in, This represents the power generation capacity in the h-th hour of day d. This represents the average power generation of the current photovoltaic equipment in the h-th hour over all N days, where h ranges from 1 to 24 and d ranges from 1 to N.

[0016] The current photovoltaic equipment every hour Connect them sequentially to form the daily power generation curve:

[0017] ;

[0018] in, This is the representative power generation curve;

[0019] S13. Based on the representative power generation curves of each photovoltaic device and the model of the photovoltaic device, perform cluster analysis and grouping of the subordinate distributed photovoltaic devices.

[0020] In a preferred embodiment, S13 includes the following sub-steps:

[0021] S131. Collect relevant data for each photovoltaic device, including representative power generation curves and device models. Use the representative power generation curves as time series features to extract the daily maximum power generation, minimum power generation, average power generation, and power generation curve slope.

[0022] S132. The photovoltaic equipment under its jurisdiction is initially grouped according to its model number, with photovoltaic equipment of the same model classified into the same group. Then, the extracted photovoltaic equipment features are combined into a feature vector, which serves as the comprehensive feature representation of each photovoltaic equipment. The algorithm formula is as follows:

[0023] ;

[0024] in, This represents the feature vector of the i-th photovoltaic device. These represent average power, minimum power, maximum power, and slope, respectively.

[0025] S133. Group all subordinate photovoltaic devices using K-means clustering, combined with the feature vectors of the photovoltaic devices. The specific steps are as follows:

[0026] For photovoltaic devices grouped into different types, K devices are randomly selected as initial cluster centers. For each device Calculate the Euclidean distance to the K cluster centers, and assign the user to the nearest cluster center based on the Euclidean distance. The algorithm formula is as follows:

[0027] ;

[0028] in, Represents specific data points The process of assigning to the k-th cluster, Representative data points To the cluster center The distance;

[0029] S134. Based on the allocation results, recalculate the centers of the K clusters. The algorithm formula is as follows:

[0030] ;

[0031] Repeat the assignment and update steps until the cluster centers no longer change significantly or the maximum number of iterations is reached, based on the obtained updates. For each device under different initial groups Assign the newest cluster center to the nearest cluster center based on the Euclidean distance, and establish secondary groupings based on the number of cluster centers.

[0032] In a preferred embodiment, S2 includes the following steps:

[0033] S21. For the primary grouping obtained by equipment type, establish a primary classification database through MySQL. Based on the secondary grouping under the primary grouping, establish a secondary classification database in the primary classification database. Then, classify the photovoltaic equipment under the primary grouping into the secondary classification database based on the secondary grouping and establish a separate photovoltaic equipment file, which includes the photovoltaic equipment number and the person in charge's email address.

[0034] S22. Collect hourly power generation data of photovoltaic equipment from different secondary classification databases every day, and record the power generation data with timestamps in the relevant photovoltaic equipment files;

[0035] S23. Based on the photovoltaic equipment under the secondary classification, obtain the latitude and longitude information of the photovoltaic equipment through GPS, use the clustering algorithm to cluster the location information of the photovoltaic equipment, select the center point of each cluster as the representative location, establish a tertiary classification database in the secondary classification database, and classify the photovoltaic equipment files that meet the clustering into the tertiary classification database.

[0036] S24. Set humidity sensors at the center points of the clusters to collect humidity data in real time, every day and every hour, and record it in the three-level classification database.

[0037] In a preferred embodiment, step S3 includes the following steps:

[0038] S31. Statistically analyze the power generation data of photovoltaic equipment in the three-level classification database within n periods, and make a horizontal comparison of the power generation data of the same hour in n periods to identify undetermined abnormal photovoltaic power generation equipment.

[0039] S32. For photovoltaic power generation equipment with pending abnormalities, conduct anomaly judgment and analysis by combining humidity data in the three-level classification database to determine the cause of the photovoltaic equipment abnormality.

[0040] In a preferred embodiment, step S31 includes the following sub-steps:

[0041] S311. Collect the photovoltaic power generation data of photovoltaic equipment in the three-level classification database within n periods, collect the photovoltaic power generation of each day and each hour within the three-level classification within n periods, and calculate the mean and standard deviation of the power generation of all equipment in each hour. The algorithm formula is as follows:

[0042] ;

[0043] in, Represents the mean. Let m be the standard deviation, and m be the total number of photovoltaic devices in the current three-level classification database. This represents the power generation of the j-th device in the current hour;

[0044] S312. Calculate the Z-score for each device within the same number of days and hours over n periods. Based on the Z-score, make an initial judgment on anomalies. The algorithm formula is as follows:

[0045] ;

[0046] Initialize the anomaly detection coefficient based on the period n. , The initial value is 0, when Greater than the threshold When this occurs, it indicates that the photovoltaic power generation of the relevant equipment is abnormal for the current hour. The abnormality judgment coefficient is incremented by 1, and the abnormal value is recorded. ,right An initial anomaly determination is performed hourly to obtain the periodic anomaly determination coefficient. ,when When the time is specified, it indicates that the relevant photovoltaic equipment is abnormal and is marked as photovoltaic power generation equipment with pending abnormality. h is the judgment coefficient with a value of 8n.

[0047] In a preferred embodiment, step S32 includes the following sub-steps:

[0048] S321. Collect the abnormal hourly humidity values ​​S from the relevant abnormal photovoltaic equipment cycle n in the three-level classification database, based on the humidity threshold. Make a judgment;

[0049] Set initial exception detection coefficient , The initial value is 0, when S is greater than the threshold. When the humidity of the relevant equipment is also abnormal during the current abnormal hour, the abnormality judgment coefficient is incremented by 1, and the abnormal value is recorded as follows. For each abnormal hour of the photovoltaic equipment, an anomaly determination is performed to obtain the periodic anomaly determination coefficient. ;

[0050] when When h is the humidity of the photovoltaic power generation equipment that is pending abnormality, it indicates that there is an abnormality in humidity. h is the judgment coefficient, and its value is two-thirds of the total number of humidity judgments.

[0051] S322, when If this occurs, it indicates that the relevant photovoltaic power generation equipment has an abnormal accumulation of dust.

[0052] In a preferred embodiment, step S4 includes the following sub-steps:

[0053] S41. Mark the devices with abnormal humidity in the Level 3 database in blue, and mark the data with abnormal dust accumulation in the Level 3 database in red. Extract the relevant abnormal device numbers and the email addresses of the responsible persons.

[0054] S42. Use the email sending library to send the default email for the anomaly report to the email address of the person in charge of the relevant abnormal photovoltaic equipment.

[0055] The beneficial effects of the present invention are:

[0056] This invention uses cluster analysis to group distributed photovoltaic (PV) equipment under the jurisdiction of a city, grouping equipment of the same model with similar PV power generation curves into the same secondary group. This enables more centralized and efficient management, making it easier for managers to plan and maintain distributed PV equipment under the city's jurisdiction.

[0057] This invention groups distributed photovoltaic devices within a secondary group into a tertiary group based on humidity. Simultaneously, it collects the power generation of each photovoltaic device within the tertiary group in the same hour according to a periodic data. Photovoltaic devices with abnormal power generation are identified, and abnormal photovoltaic devices can be determined under the same variable, thus predicting potential failure risks.

[0058] In this invention, by collecting humidity data and grouping it into three levels, it is possible to determine whether the photovoltaic equipment has abnormal humidity and dust accumulation when abnormal power generation occurs. By using a unified variable, it is possible to quickly identify whether the abnormal power generation is caused by humidity or dust accumulation, thus enhancing the functionality and practicality of the method.

[0059] Through long-term data accumulation and analysis, this invention can predict potential risks of abnormal humidity and dust accumulation, take preventive maintenance measures in advance, reduce downtime and power generation efficiency loss caused by abnormalities, and send the information to the relevant photovoltaic equipment manager via email, thereby facilitating targeted maintenance of abnormal photovoltaic equipment.

[0060] This invention presents a clustering analysis method for distributed photovoltaic characteristics based on multi-source data. It comprehensively applies data science and energy management technologies to achieve refined and intelligent management of distributed photovoltaic equipment in cities. By using unified variables, it facilitates the detection of anomalies in distributed photovoltaic equipment, thereby improving the management and maintenance effectiveness of urban distributed photovoltaic equipment. Attached Figure Description

[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0062] Figure 1 This is a flowchart of a distributed photovoltaic characteristic clustering analysis method based on multi-source data according to an embodiment of the present invention. Detailed Implementation

[0063] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0064] According to an embodiment of the present invention, a distributed photovoltaic characteristic clustering analysis method based on multi-source data is provided.

[0065] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments:

[0066] Example 1: Figure 1 As shown, according to an embodiment of the present invention, a distributed photovoltaic characteristic clustering analysis method based on multi-source data includes the following steps:

[0067] S1. For distributed photovoltaic (PV) equipment under the city's jurisdiction, collect the power generation curves of different PV equipment on N sunny days each day, and perform cluster analysis on the distributed PV equipment under the jurisdiction based on the characteristics of the PV equipment.

[0068] S11. For distributed photovoltaic equipment under the city's jurisdiction, collect power generation data of different photovoltaic equipment for each hour of each day during N sunny weather days.

[0069] S12. Take the average power generation data of N photovoltaic devices in the same hour, and plot the representative power generation curves of different photovoltaic devices based on the average hourly power generation. The specific steps are as follows:

[0070] ;

[0071] in, This represents the power generation capacity in the h-th hour of day d. This represents the average power generation of the current photovoltaic equipment in the h-th hour over all N days, where h ranges from 1 to 24 and d ranges from 1 to N.

[0072] The current photovoltaic equipment every hour Connect them sequentially to form the daily power generation curve:

[0073] ;

[0074] in, This is the representative power generation curve;

[0075] It should be noted that the value of N is usually set to 15. N can be increased or decreased according to the actual situation. Sunny weather can be determined based on the weather forecast results in the meteorological forecast. By averaging, the impact of random fluctuations can be effectively reduced when the amount of data is large, the overall trend of the curve can be extracted, and the influence of a single anomaly on the final curve fitting can be avoided.

[0076] S13. Based on the representative power generation curves of each photovoltaic device and the model of the photovoltaic device, perform cluster analysis and grouping of the subordinate distributed photovoltaic devices;

[0077] S131. Collect relevant data for each photovoltaic device, including representative power generation curves and device models. Use the representative power generation curves as time series features to extract the daily maximum power generation, minimum power generation, average power generation, and power generation curve slope.

[0078] S132. The photovoltaic equipment under its jurisdiction is initially grouped according to its model number, with photovoltaic equipment of the same model classified into the same group. Then, the extracted photovoltaic equipment features are combined into a feature vector, which serves as the comprehensive feature representation of each photovoltaic equipment. The algorithm formula is as follows:

[0079] ;

[0080] in, This represents the feature vector of the i-th photovoltaic device. These represent average power, minimum power, maximum power, and slope, respectively.

[0081] S133. Group all subordinate photovoltaic devices using K-means clustering, combined with the feature vectors of the photovoltaic devices. The specific steps are as follows:

[0082] For photovoltaic devices grouped into different types, K devices are randomly selected as initial cluster centers. For each device Calculate the Euclidean distance to the K cluster centers, and assign the user to the nearest cluster center based on the Euclidean distance. The algorithm formula is as follows:

[0083] ;

[0084] in, Represents specific data points The process of assigning to the k-th cluster, Representative data points To the cluster center The distance;

[0085] It should be noted that, This is used to find a value of k that minimizes the following distance, where k ranges from 1 to K. The calculation is performed using Euclidean distance, and the algorithm formula is as follows:

[0086] ;

[0087] in, For data points The j-th feature, Cluster center The j-th feature.

[0088] S134. Based on the allocation results, recalculate the centers of the K clusters. The algorithm formula is as follows:

[0089] ;

[0090] Repeat the assignment and update steps until the cluster centers no longer change significantly or the maximum number of iterations is reached, based on the obtained updates. For each device under different initial groups Assign the newest cluster center to the nearest cluster center based on the Euclidean distance, and establish secondary groupings based on the number of cluster centers.

[0091] It should be noted that after the K-means clustering algorithm converges, the subordinate devices are grouped according to the existing clusters. That is, based on the distance between the device and the cluster center, the device is assigned to the nearest cluster. The final grouping results are used for further data analysis or decision support. The photovoltaic devices after secondary classification can be understood as having similar working environments, and their power generation values ​​are similar in the same hour under sufficient sunlight.

[0092] S2. Based on the characteristics of distributed photovoltaic equipment after cluster analysis, establish databases respectively, collect the power generation data of different photovoltaic equipment in different databases every day, and at the same time, set up humidity sensors based on the geographical location of the equipment to collect humidity data.

[0093] S21. For the primary grouping obtained by equipment type, establish a primary classification database through MySQL. Based on the secondary grouping under the primary grouping, establish a secondary classification database in the primary classification database. Then, classify the photovoltaic equipment under the primary grouping into the secondary classification database based on the secondary grouping and establish a separate photovoltaic equipment file, which includes the photovoltaic equipment number and the person in charge's email address.

[0094] S22. Collect hourly power generation data of photovoltaic equipment from different secondary classification databases every day, and record the power generation data with timestamps in the relevant photovoltaic equipment files;

[0095] S23. Based on the photovoltaic equipment under the secondary classification, obtain the latitude and longitude information of the photovoltaic equipment through GPS, use the clustering algorithm to cluster the location information of the photovoltaic equipment, select the center point of each cluster as the representative location, establish a tertiary classification database in the secondary classification database, and classify the photovoltaic equipment files that meet the clustering into the tertiary classification database.

[0096] S24. Set humidity sensors at the center points of the clusters to collect humidity data in real time, every day and every hour, and record it in the three-level classification database.

[0097] It should be noted that by performing geographic information clustering on photovoltaic equipment and selecting representative centers to set up humidity sensors, real-time humidity information of the photovoltaic equipment at the cluster locations can be obtained. This information can then be used to determine and analyze faulty photovoltaic equipment in subsequent fault diagnosis.

[0098] Example 2: S3. Based on the power generation of different photovoltaic devices within n periods in different databases, determine and analyze the causes of abnormalities in the same database, including dust accumulation and humidity effects.

[0099] S31. Statistically analyze the power generation data of photovoltaic equipment in the three-level classification database within n periods, and make a horizontal comparison of the power generation data of the same hour in n periods to identify undetermined abnormal photovoltaic power generation equipment.

[0100] S311. Collect the photovoltaic power generation data of photovoltaic equipment in the three-level classification database within n periods, collect the photovoltaic power generation of each day and each hour within the three-level classification within n periods, and calculate the mean and standard deviation of the power generation of all equipment in each hour. The algorithm formula is as follows:

[0101] ;

[0102] in, Represents the mean. Let m be the standard deviation, and m be the total number of photovoltaic devices in the current three-level classification database. This represents the power generation of the j-th device in the current hour;

[0103] S312. Calculate the Z-score for each device within the same number of days and hours over n periods. Based on the Z-score, make an initial judgment on anomalies. The algorithm formula is as follows:

[0104] ;

[0105] Initialize the anomaly detection coefficient based on the period n. , The initial value is 0, when Greater than the threshold When this occurs, it indicates that the photovoltaic power generation of the relevant equipment is abnormal for the current hour. The abnormality judgment coefficient is incremented by 1, and the abnormal value is recorded. ,right An initial anomaly determination is performed hourly to obtain the periodic anomaly determination coefficient. ,when When the time is specified, it indicates that the relevant photovoltaic equipment is abnormal and is marked as photovoltaic power generation equipment with pending abnormality. h is the judgment coefficient with a value of 8n.

[0106] It should be noted that, The values ​​can be set by consulting experts in the relevant field, and are usually set to 3. n is usually set to 7, but can also be adjusted according to the actual situation. h is the determination coefficient, with a value of 8n. This is achieved by taking the average daily direct sunlight duration as 12 hours. minus and multiply The data is obtained because summer has longer daylight hours, with approximately 14 hours of sunshine per day, while winter has shorter daylight hours, with approximately 10 hours of sunshine per day. The h data can be flexibly adjusted according to the actual conditions in different regions to meet the usage needs of different areas.

[0107] S32. For photovoltaic power generation equipment with pending abnormalities, conduct anomaly judgment and analysis by combining humidity data in the three-level classification database to determine the cause of the photovoltaic equipment abnormality.

[0108] S321. Collect the abnormal hourly humidity values ​​S from the relevant abnormal photovoltaic equipment cycle n in the three-level classification database, based on the humidity threshold. Make a judgment;

[0109] Set initial exception detection coefficient , The initial value is 0, when S is greater than the threshold. When the humidity of the relevant equipment is also abnormal during the current abnormal hour, the abnormality judgment coefficient is incremented by 1, and the abnormal value is recorded as follows. For each abnormal hour of the photovoltaic equipment, an anomaly determination is performed to obtain the periodic anomaly determination coefficient. ;

[0110] when When h is the humidity of the photovoltaic power generation equipment that is pending abnormality, it indicates that there is an abnormality in humidity. h is the judgment coefficient, and its value is two-thirds of the total number of humidity judgments.

[0111] S322, when If this occurs, it indicates that the relevant photovoltaic power generation equipment has an abnormal accumulation of dust.

[0112] S4. Mark the abnormal photovoltaic equipment in the database, and at the same time transmit the abnormal information to the relevant photovoltaic equipment manager's email address.

[0113] S41. Mark the devices with abnormal humidity in the Level 3 database in blue, and mark the data with abnormal dust accumulation in the Level 3 database in red. Extract the relevant abnormal device numbers and the email addresses of the responsible persons.

[0114] S42. Use the email sending library to send the default email for the anomaly report to the email address of the person in charge of the relevant abnormal photovoltaic equipment.

[0115] It should be noted that, based on the different marker colors, two preset emails are set up, and the corresponding preset emails are sent to the person in charge's email address using the email module in Python.

[0116] In summary, this invention uses cluster analysis to group distributed photovoltaic (PV) devices under a city's jurisdiction. Devices with similar PV power generation curves of the same model are grouped into the same secondary group, enabling more centralized and efficient management. This facilitates planning and maintenance of distributed PV devices under the city's jurisdiction by managers. By grouping the distributed PV devices within the secondary group into tertiary groups based on humidity, and collecting the power generation of each PV device in the tertiary group in the same hour according to a cycle, abnormal power generation devices can be identified. This allows for the identification of abnormal PV devices under the same variable and the prediction of potential failure risks.

[0117] By collecting humidity data and grouping it into three levels, abnormal humidity and dust accumulation in photovoltaic (PV) equipment can be identified when PV power generation is abnormal. By using unified variables, it is possible to quickly identify whether the abnormal power generation is caused by humidity or dust accumulation, thus enhancing the functionality and practicality of the method. By comprehensively applying data science and energy management technologies, it enables refined and intelligent management of distributed PV equipment in cities. The unified variables facilitate the detection of anomalies in distributed PV equipment, improving the management and maintenance effectiveness of urban distributed PV equipment.

[0118] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A distributed photovoltaic characteristic clustering analysis method based on multi-source data, characterized in that, The method includes the following steps: S1. For distributed photovoltaic (PV) devices under the city's jurisdiction, collect the daily power generation curves of different PV devices on N sunny days. Combined with the characteristics of the PV devices, perform cluster analysis on the distributed PV devices under the jurisdiction. Based on the representative power generation curves of each PV device and the model of the PV device, through... Clustering: The distributed photovoltaic equipment under its jurisdiction is clustered and grouped, and secondary groups are established based on the number of cluster centers. S2. Based on the characteristics of distributed photovoltaic equipment after cluster analysis, establish databases respectively, use clustering algorithms to cluster the location information of photovoltaic equipment, select the center point of each cluster as the representative location, collect the power generation data of different photovoltaic equipment in different databases every day, and at the same time, based on the geographical location of the equipment, set up humidity sensors at the center point of the cluster to collect humidity data and record it in the three-level classification database. S3. Based on the power generation of different photovoltaic devices within n periods in different databases, determine and analyze the causes of abnormality of abnormal photovoltaic devices in the same database, including dust accumulation and humidity effects. S31. Statistically analyze the power generation data of photovoltaic equipment in the three-level classification database within n periods, perform a horizontal comparison of the power generation data for the same hour within n periods, and calculate the power generation data for each device within the same number of days and the same hour within n periods. ,based on An initial assessment of the anomalies was conducted to identify the photovoltaic power generation equipment that was suspected of being abnormal. S32. For photovoltaic power generation equipment with pending anomalies, anomaly determination and analysis are performed by combining humidity data from the three-level classification database. Abnormal hourly humidity values ​​S for relevant abnormal photovoltaic equipment in the three-level classification database within period n are collected, and based on humidity thresholds... Make a judgment to determine the cause of the photovoltaic equipment malfunction; S4. Mark the abnormal photovoltaic equipment in the database, and at the same time transmit the abnormal information to the relevant photovoltaic equipment manager's email address.

2. The distributed photovoltaic characteristic clustering analysis method based on multi-source data according to claim 1, characterized in that, S1 includes the following sub-steps: S11. For distributed photovoltaic equipment under the city's jurisdiction, collect power generation data of different photovoltaic equipment for each hour of each day during N sunny weather days. S12. Take the average power generation data of N photovoltaic devices in the same hour, and plot the representative power generation curves of different photovoltaic devices based on the average hourly power generation. The specific steps are as follows: ； in, This represents the power generation capacity in the h-th hour of day d. This represents the average power generation of the current photovoltaic equipment in the h-th hour over all N days, where h ranges from 1 to 24 and d ranges from 1 to N. The current photovoltaic equipment every hour Connect them sequentially to form the daily power generation curve: ； in, This is the representative power generation curve; S13. Based on the representative power generation curves of each photovoltaic device and the model of the photovoltaic device, perform cluster analysis and grouping of the subordinate distributed photovoltaic devices.

3. The distributed photovoltaic characteristic clustering analysis method based on multi-source data according to claim 2, characterized in that, S13 includes the following sub-steps: S131. Collect relevant data for each photovoltaic device, including representative power generation curves and device models. Use the representative power generation curves as time series features to extract the daily maximum power generation, minimum power generation, average power generation, and power generation curve slope. S132. The photovoltaic equipment under its jurisdiction is initially grouped according to its model number, with photovoltaic equipment of the same model classified into the same group. Then, the extracted photovoltaic equipment features are combined into a feature vector, which serves as the comprehensive feature representation of each photovoltaic equipment. The algorithm formula is as follows: ； in, This represents the feature vector of the i-th photovoltaic device. These represent average power, minimum power, maximum power, and slope, respectively. S133, Through Clustering, which combines the feature vectors of photovoltaic devices, groups all subordinate photovoltaic devices. The specific steps are as follows: For photovoltaic devices grouped into different types, randomly select 10 devices as the initial cluster centers. For each device Calculate the Euclidean distance to the K cluster centers, and assign the user to the nearest cluster center based on the Euclidean distance. The algorithm formula is as follows: ； in, Represents specific data points The process of assigning to the k-th cluster, Representative data points To the cluster center The distance; S134. Based on the allocation results, recalculate the centers of the K clusters. The algorithm formula is as follows: ； Repeat the assignment and update steps until the cluster centers no longer change significantly or the maximum number of iterations is reached, based on the obtained updates. For each device under different initial groups Assign the newest cluster center to the nearest cluster center based on the Euclidean distance, and establish secondary groupings based on the number of cluster centers.

4. The distributed photovoltaic characteristic clustering analysis method based on multi-source data according to claim 1, characterized in that, S2 includes the following steps: S21. For the primary grouping obtained by equipment type, establish a primary classification database through MySQL. Based on the secondary grouping under the primary grouping, establish a secondary classification database in the primary classification database. Then, classify the photovoltaic equipment under the primary grouping into the secondary classification database based on the secondary grouping and establish a separate photovoltaic equipment file, which includes the photovoltaic equipment number and the person in charge's email address. S22. Collect hourly power generation data of photovoltaic equipment from different secondary classification databases every day, and record the power generation data with timestamps in the relevant photovoltaic equipment files; S23. Based on the photovoltaic equipment under the secondary classification, obtain the latitude and longitude information of the photovoltaic equipment through GPS, use the clustering algorithm to cluster the location information of the photovoltaic equipment, select the center point of each cluster as the representative location, establish a tertiary classification database in the secondary classification database, and classify the photovoltaic equipment files that meet the clustering into the tertiary classification database. S24. Set humidity sensors at the center points of the clusters to collect humidity data in real time, every day and every hour, and record it in the three-level classification database.

5. The distributed photovoltaic characteristic clustering analysis method based on multi-source data according to claim 1, characterized in that, S31 includes the following sub-steps: S311. Collect the photovoltaic power generation data of photovoltaic equipment in the three-level classification database within n periods, collect the photovoltaic power generation of each day and each hour within the three-level classification within n periods, and calculate the mean and standard deviation of the power generation of all equipment in each hour. The algorithm formula is as follows: ； in, Represents the mean. Let m be the standard deviation, and m be the total number of photovoltaic devices in the current three-level classification database. This represents the power generation of the j-th device in the current hour; S312. For each device, calculate the same number of days and the same number of hours within the n-period. ,based on For the initial determination of anomalies, the algorithm formula is as follows: ； Initialize the anomaly detection coefficient based on the period n. , The initial value is 0, when Greater than the threshold When this occurs, it indicates that the photovoltaic power generation of the relevant equipment is abnormal for the current hour. The abnormality judgment coefficient is incremented by 1, and the abnormal value is recorded. ,right An initial anomaly determination is performed hourly to obtain the periodic anomaly determination coefficient. ,when When the time is specified, it indicates that the relevant photovoltaic equipment is abnormal and is marked as photovoltaic power generation equipment with pending abnormality. h is the judgment coefficient with a value of 8n.

6. The distributed photovoltaic characteristic clustering analysis method based on multi-source data according to claim 5, characterized in that, S32 includes the following sub-steps: S321. Collect the abnormal hourly humidity values ​​S from the relevant abnormal photovoltaic equipment cycle n in the three-level classification database, based on the humidity threshold. Make a judgment; Set initial exception detection coefficient , The initial value is 0, when S is greater than the threshold. When the humidity of the relevant equipment is also abnormal during the current abnormal hour, the abnormality judgment coefficient is incremented by 1, and the abnormal value is recorded as follows. For each abnormal hour of the photovoltaic equipment, an anomaly determination is performed to obtain the periodic anomaly determination coefficient. ; when When h is the humidity of the photovoltaic power generation equipment that is pending abnormality, it indicates that there is an abnormality in humidity. h is the judgment coefficient, and its value is two-thirds of the total number of humidity judgments. S322, when If this occurs, it indicates that the relevant photovoltaic power generation equipment has an abnormal accumulation of dust.

7. The distributed photovoltaic characteristic clustering analysis method based on multi-source data according to claim 1, characterized in that, S4 includes the following sub-steps: S41. Mark the devices with abnormal humidity in the Level 3 database in blue, and mark the data with abnormal dust accumulation in the Level 3 database in red. Extract the relevant abnormal device numbers and the email addresses of the responsible persons. S42. Use the email sending library to send the default email for the anomaly report to the email address of the person in charge of the relevant abnormal photovoltaic equipment.

Images