Photovoltaic device operation monitoring method and system based on big data

Abstract

This application relates to the field of photovoltaic power generation technology, and in particular to a method and system for monitoring the operation of photovoltaic equipment based on big data. The method includes: obtaining anomaly candidate points based on an isolated forest; calculating the symmetry recovery index of monitoring data for each equipment type at the anomaly candidate point; calculating the anomaly contribution of each monitoring data point at the anomaly candidate point, and obtaining the absolute value of the difference between the total contribution of environmental monitoring data and equipment monitoring data; when the absolute value of the difference is greater than a judgment threshold and the symmetry recovery index of each equipment type monitoring data is lower than a preset index, the anomaly candidate point is determined to be a fault anomaly of the photovoltaic equipment; otherwise, it is determined to be a non-fault anomaly caused by environmental disturbance. Through the technical solution of this application, equipment anomalies caused by environmental interference and actual equipment failures can be effectively distinguished, improving the accuracy of photovoltaic equipment operation monitoring results.

Description

Technical Field

[0001] This application relates to the field of photovoltaic power generation technology, and in particular to a method and system for monitoring the operation of photovoltaic equipment based on big data. Background Technology

[0002] As an important source of clean energy, the stability and efficiency of photovoltaic power plants are of paramount importance. Real-time monitoring and analysis of photovoltaic equipment operating data, enabling timely detection and diagnosis of equipment anomalies, is crucial for ensuring the safe and efficient operation of power plants. Therefore, intelligent operation and maintenance of photovoltaic power plants has become an industry trend.

[0003] During the operation of photovoltaic (PV) equipment, its power generation performance is affected not only by the equipment's own health condition but also by environmental factors such as sunlight intensity and ambient temperature. Therefore, when monitoring data shows anomalies, it is difficult to distinguish whether the anomaly is caused by temporary environmental disturbances such as cloud cover or shadows, or by genuine faults such as aging or damage to equipment components. This confusion often leads to a large number of false alarms, increasing the burden on maintenance personnel and reducing the accuracy and efficiency of fault diagnosis.

[0004] Existing technologies typically employ anomaly detection algorithms such as isolated forests or local outliers to detect anomalies in multidimensional monitoring data. For instance, the isolated forest algorithm constructs multiple isolated trees to segment data points, enabling rapid filtering of potential anomalies from massive datasets. However, existing anomaly detection algorithms cannot effectively distinguish the nature of anomalies; that is, they cannot differentiate between recoverable environmental disturbances and irrecoverable equipment failures, leading to inaccurate monitoring results for photovoltaic equipment operation. Summary of the Invention

[0005] To address the technical problem of inaccurate photovoltaic equipment operation monitoring results, this application provides a photovoltaic equipment operation monitoring method and system based on big data, which can effectively distinguish between equipment anomalies caused by environmental interference and actual equipment failures, thereby improving the accuracy of photovoltaic equipment operation monitoring results.

[0006] In a first aspect, this application provides a method for monitoring the operation of photovoltaic equipment based on big data. The monitoring method includes: obtaining anomaly scores of photovoltaic equipment operation data at any given time based on isolated forest methods; identifying operation data with anomaly scores greater than anomaly thresholds as candidate anomaly points; the operation data includes multi-dimensional monitoring data, including environmental monitoring data and equipment monitoring data; within a monitoring window centered on the candidate anomaly point, calculating the symmetry recovery index of each equipment monitoring data point; calculating the anomaly contribution of each monitoring data point within the candidate anomaly point, and obtaining the absolute value of the difference between the total contribution of environmental monitoring data and equipment monitoring data; when the absolute value of the difference is greater than a judgment threshold and the symmetry recovery index of each equipment monitoring data is lower than a preset index, the candidate anomaly point is determined to be a fault anomaly of the photovoltaic equipment; otherwise, it is determined to be a non-fault anomaly caused by environmental disturbance.

[0007] Preferably, the environmental monitoring data includes irradiance and solar panel backsheet temperature; the equipment monitoring data includes the output power, output voltage, and output current of the photovoltaic equipment.

[0008] The monitoring data was divided into environmental monitoring data and equipment monitoring data, which provided a data foundation for subsequent calculation of the symmetry recovery index and anomaly contribution, and ensured the effectiveness of the diagnostic results.

[0009] Preferably, the calculation of the symmetric recovery index of the monitoring data of each equipment type at the candidate anomaly point includes: within the monitoring window, before the timestamp of the candidate anomaly point, taking the integral of the difference between the actual measured value and the reference theoretical value in the monitoring data of any equipment type as the power loss; after the timestamp of the candidate anomaly point, taking the integral of the difference between the actual measured value and the reference theoretical value in the monitoring data of the equipment type as the power recovery; the symmetric recovery index is the ratio of the power recovery to the power loss.

[0010] By comparing the integral quantification symmetric recovery index of the difference between the actual value and the theoretical value within two time windows before and after the anomaly point, the degree of recovery of equipment monitoring data is accurately measured.

[0011] Preferably, the method for obtaining the reference theoretical value includes: inputting the environmental monitoring data corresponding to the timestamp of the anomaly candidate point into a pre-trained machine learning model to obtain the reference theoretical value of the equipment monitoring data.

[0012] Preferably, calculating the anomaly contribution of each monitoring data point among the anomaly candidate points includes: tracing and recording the isolated paths from the root node to the leaf node of each isolated tree in the isolated forest, and the monitoring data used by each split node on the isolated path; the anomaly contribution of any monitoring data is obtained by summing the weighted contributions of the monitoring data in all isolated trees; wherein, in a single isolated tree, the weighted contribution of the monitoring data is positively correlated with the number of times the monitoring data appears in the split nodes of the isolated path, and negatively correlated with the length of the isolated path.

[0013] By tracing isolated paths and weighting the frequency of occurrence of each feature along the path according to the path length, the anomaly contribution of each monitoring data can be obtained, thereby accurately locating whether the root cause of the anomaly belongs to environmental monitoring data or equipment monitoring data.

[0014] Preferably, the calculated abnormal contribution of each monitoring data is normalized to obtain the contribution percentage of each monitoring data.

[0015] Preferably, the step of calculating the absolute value of the difference between the total contribution of environmental monitoring data and equipment monitoring data includes: summing the abnormal contribution values ​​of all environmental monitoring data to obtain the total environmental contribution value; summing the abnormal contribution values ​​of all equipment monitoring data to obtain the total equipment contribution value; and calculating the absolute value of the difference between the total environmental contribution value and the total equipment contribution value.

[0016] The difference directly quantifies whether the abnormal event is dominated by environmental factors or equipment factors, providing strong evidence to distinguish whether the candidate abnormal point belongs to environmental disturbance or actual equipment failure.

[0017] Preferably, the abnormal score obtained from the isolated forest for the photovoltaic equipment's operation data at any time includes: calculating the average path length of the operation data in each isolated tree of the isolated forest; the abnormal score is negatively correlated with the average path length.

[0018] Preferably, the monitoring window is centered on the timestamp of the abnormal candidate point and takes a time range of multiple moments before and after it.

[0019] In a second aspect, this application also provides a photovoltaic equipment operation monitoring system based on big data, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the photovoltaic equipment operation monitoring method based on big data according to the first aspect of this application is implemented.

[0020] The technical solution of this application has the following beneficial technical effects:

[0021] The isolated forest algorithm efficiently filters the collected operational data, quickly identifying potential anomaly candidates. Then, these candidates are precisely differentiated from environmental disturbances to genuine equipment failures using two dimensions. First, within a time window centered on the anomaly, the symmetry of loss and recovery in equipment monitoring data before and after the anomaly is quantified to determine process reversibility, thus distinguishing between environmental disturbances and equipment failures. Second, the contribution of each monitoring data point to the anomaly candidate identification is quantified, and the difference between the total contribution of environmental and equipment data is calculated. When an anomaly candidate simultaneously meets the conditions of a low symmetry recovery index and a large difference between environmental and equipment contributions, it is determined to be a genuine photovoltaic equipment failure. This achieves accurate attribution of anomalies and significantly improves the accuracy of monitoring and early warning. Attached Figure Description

[0022] Figure 1 This is a flowchart of a photovoltaic equipment operation monitoring method based on big data according to an embodiment of this application.

[0023] Figure 2 This is a schematic diagram of the screening results of isolated forest anomaly candidate points according to an embodiment of this application.

[0024] Figure 3 This is a schematic diagram illustrating the calculation process of the symmetric recovery index of the abnormal candidate point corresponding to 11:07 according to an embodiment of this application.

[0025] Figure 4 This is a schematic diagram of the fault diagnosis results according to an embodiment of this application.

[0026] Figure 5 This is a structural block diagram of a photovoltaic equipment operation monitoring system based on big data, according to an embodiment of this application. Detailed Implementation

[0027] According to the first aspect of this application, this application provides a big data-based photovoltaic equipment operation monitoring method for intelligent operation and maintenance of large-scale centralized photovoltaic power plants. The power generation of photovoltaic equipment is greatly affected by environmental factors. When the power generation of photovoltaic equipment is abnormal, it may be caused by environmental disturbances or actual equipment failures. Therefore, this application aims to solve the problem of inaccurate operation monitoring results caused by the inability to accurately distinguish between benign environmental disturbances, such as cloud cover or tree shadows, and actual equipment failures.

[0028] Figure 1 This is a flowchart of a photovoltaic equipment operation monitoring method based on big data, according to an embodiment of this application. Figure 1As shown, the photovoltaic equipment operation monitoring method based on big data includes steps S101 to S104, which are described in detail below.

[0029] S101, based on the anomaly score of the photovoltaic equipment's operating data at any time obtained from the isolated forest, the operating data with anomaly scores greater than the anomaly threshold are taken as anomaly candidate points. The operating data includes multi-dimensional monitoring data, which includes environmental monitoring data and equipment monitoring data.

[0030] In one embodiment, pre-deployed sensors continuously collect operational data characterizing the operating status of the photovoltaic (PV) equipment. Specifically, the operational data includes multi-dimensional monitoring data, which is divided into two categories: environmental monitoring data and equipment monitoring data. The environmental monitoring data includes irradiance and the backsheet temperature of the solar panels; the equipment monitoring data includes the output power, output voltage, and output current of the PV equipment. These multi-source data collectively constitute a high-dimensional time-series feature vector, serving as the basis for all subsequent analyses. Irradiance can be collected using illuminance sensors deployed on the surface of the PV equipment.

[0031] Understandably, normalization can be performed before inputting these monitoring data into the model to eliminate the influence of different physical dimensions and improve the stability and convergence speed of the algorithm.

[0032] The operational data at any given time is input into a pre-trained isolated forest for anomaly scoring, and the isolated forest consists of multiple isolated trees. The core idea of ​​the isolated forest is that, due to its sparse and discrete characteristics, anomalous data is more easily isolated during random data partitioning, meaning that a leaf node can be reached from the root node via a shorter path. Therefore, obtaining the anomaly score of the photovoltaic device's operational data at any given time based on the isolated forest includes: calculating the average path length of the operational data in each isolated tree of the isolated forest; the anomaly score is negatively correlated with the average path length.

[0033] Abnormal scores Satisfying the relation:

[0034] ;

[0035] in, For photovoltaic equipment at all times Operational data; For running data The length of an isolated path in any isolated tree; For running data The average length of isolated paths for all isolated trees in an isolated forest; For a sample size of The expected average path length at time is used to... Normalization is performed. The sample size is... Expected average path length satisfy: And sample size .

[0036] When the average path length of the running data is much lower than the average level When the exponent of the exponent approaches 0, the outlier score approaches 1, indicating a high degree of suspicion for the data anomalies; conversely, when the average path length is close to or greater than the average level... When the anomaly score approaches 0.5 or less, it indicates that the running data is normal. The anomaly threshold can be set to 0.6; when the anomaly score is greater than this threshold, the running data at that moment is considered an anomaly candidate. Please refer to [link to relevant documentation]. Figure 2 This is a schematic diagram of the screening results of isolated forest anomaly candidate points according to an embodiment of this application.

[0037] In this way, the initial screening of anomalies can be completed efficiently and quickly from massive continuous data streams, anomaly candidate points can be located, and anomaly candidate points can be accurately identified, thereby improving monitoring efficiency.

[0038] S102, within the monitoring window centered on the candidate anomaly point, calculate the symmetry recovery index of the monitoring data of each equipment type at the candidate anomaly point.

[0039] In one embodiment, after screening out the abnormal candidate points, it is necessary to distinguish their root causes, that is, to distinguish whether the abnormal candidate point belongs to environmental disturbance or to a real equipment failure.

[0040] For environmental disturbances such as cloud cover, the changes in equipment monitoring data caused by them are temporary and reversible. Once the disturbance disappears, the power generation capacity of the photovoltaic equipment should recover, and the equipment monitoring data should return to the level before the disturbance. However, actual equipment failure is an irreversible process, and the lost power generation capacity cannot be recovered on its own. Therefore, a monitoring window can be defined, and the symmetry recovery index of the equipment monitoring data within the monitoring window can be calculated to distinguish whether anomaly candidate points belong to environmental disturbances or actual equipment failures.

[0041] The monitoring window is centered on the timestamp of the candidate anomaly point and includes a time range of multiple moments before and after it. For example, if the data collection frequency is 1 minute / time, the half-width of the monitoring window can be set to 10 minutes, that is, the number of multiple moments is 10, which means that the monitoring window covers 10 minutes of data before and after the anomaly point occurs.

[0042] Within this monitoring window, the symmetry recovery index of monitoring data for each equipment type is calculated. This calculation process includes: within the monitoring window, before the timestamp of the anomaly candidate point, the integral of the difference between the actual measured value and the reference theoretical value in the monitoring data of any equipment type is used as the power loss; after the timestamp of the anomaly candidate point, the integral of the difference between the actual measured value and the reference theoretical value in the monitoring data of the equipment type is used as the power recovery; finally, the symmetry recovery index is the ratio of the power recovery to the power loss.

[0043] The formula for calculating the Symmetric Recovery Index (ISR) is as follows:

[0044] ;

[0045] in, The timestamp of the abnormal candidate point. To monitor the half width of the window, For photovoltaic equipment, timestamp Equipment monitoring data The actual measured value; For reference to theoretical values, this refers to values ​​relative to the timestamp. Under the same environmental conditions, the equipment monitoring data that a healthy photovoltaic device should have. Theoretical value; For a preset positive integer, such as Its function is to prevent the denominator from being zero. Please see Figure 3 This is a schematic diagram illustrating the calculation process of the symmetric recovery index of the abnormal candidate point corresponding to 11:07 according to an embodiment of this application.

[0046] Understandably, when equipment monitoring data is output power, the denominator in the ISR calculation formula represents the total power loss relative to theoretical power within the time window before the anomaly occurs; the numerator represents the total power recovery relative to theoretical power within the time window after the anomaly occurs. Therefore, when the ISR value approaches 1, it indicates that the lost power has been almost completely recovered, strongly characterizing the reversibility of the process, and is highly likely to be an environmental disturbance event. When the ISR value is much less than 1, it indicates that the power has not been effectively recovered, characterizing the irreversibility of the process, and is highly likely to be a real equipment failure.

[0047] The method for obtaining the reference theoretical value includes: inputting the environmental monitoring data corresponding to the anomaly candidate point timestamp into a pre-trained machine learning model to obtain the reference theoretical value of the equipment monitoring data. The environmental monitoring data characterizes the environmental conditions of the anomaly candidate point timestamp, and the machine learning model can be used to obtain the theoretical value of the equipment monitoring data corresponding to the environmental conditions of the anomaly candidate point timestamp.

[0048] It is understandable that machine learning models, such as neural networks and support vector regression, are obtained by training on a large amount of data during the historical healthy operation of the device; and one machine learning model corresponds to one device monitoring data.

[0049] Thus, a symmetric recovery index was constructed for the candidate points of anomalies, which can provide the first key basis for distinguishing between faults and disturbances from the perspective of the reversibility of the recovery process of equipment monitoring data.

[0050] S103, calculate the abnormal contribution of each monitoring data in the candidate abnormal point, and obtain the absolute value of the difference between the total contribution of environmental monitoring data and equipment monitoring data.

[0051] To analyze why a photovoltaic device is identified as an abnormal candidate point at a given time, it is necessary to calculate the abnormal contribution of each monitoring data point within that abnormal candidate point.

[0052] Specifically, calculating the anomalous contribution of each monitoring data point among the anomalous candidate points includes: tracing and recording the isolated paths from the root node to the leaf node of each isolated tree in the isolated forest, and the monitoring data used by each split node on the isolated path; the anomalous contribution of any monitoring data point is obtained by summing the weighted contributions of the monitoring data point in all isolated trees; wherein, in a single isolated tree, the weighted contribution of the monitoring data point is positively correlated with the number of times the monitoring data point appears in the split nodes of the isolated path, and negatively correlated with the length of the isolated path.

[0053] Understandably, in an isolated tree, starting from the root node, passing through multiple split nodes, one will eventually reach a leaf node, thus obtaining the isolated path of the isolated tree; and each split node in the isolated path will correspond to a type of monitoring data.

[0054] Monitoring data Abnormal contribution Used to measure monitoring data The contribution of the isolated anomaly candidate point. Specifically, at time... runtime data China's monitoring data Abnormal contribution Satisfying the relation:

[0055] ;

[0056] in, For a moment The running data, This represents the total number of isolated trees in an isolated forest. For running data In the Path length in an isolated tree For the first An isolated path of an isolated tree. isolated path One of the split nodes, For split nodes The monitoring data used in the process; As an indicator function, when a node splits... The monitoring data used in this study is monitoring data. hour, ,on the contrary, .

[0057] Statistical monitoring data The number of times it appears on a single isolated path; the more times it appears, the more important it is in determining the running data. When it belongs to an abnormal candidate point, the monitoring data The greater the effect, the better; furthermore, by using the reciprocal of the path length... Assign a weight to the path; the shorter the path, the greater the weight. That is, monitoring data In the The weighted contribution of each isolated tree is calculated; finally, the weighted contributions of all isolated trees are summed to obtain the monitoring data. Abnormal contribution.

[0058] In one embodiment, the calculated anomaly contribution of each monitoring data point is normalized to obtain the contribution percentage of each monitoring data point, and the contribution percentage of each monitoring data point constitutes a contribution vector. For example, if the normalized contribution vector is {output power: 45%, irradiance: 40%, others: 15%}, it indicates that the anomalies in the real-time operating data are mainly caused by power and irradiance, which highly matches the cloud cover scenario; if the contribution vector is {output power: 30%, output voltage: 35%, output current: 30%, irradiance: 5%}, it indicates that the anomalies in the real-time operating data are caused by a fault in the electrical characteristics of the equipment itself.

[0059] After obtaining the contribution of each monitoring data point, the absolute value of the difference between the total contribution of environmental monitoring data and equipment monitoring data is further calculated. Specifically, this includes: summing the abnormal contribution of all environmental monitoring data to obtain the total environmental contribution; summing the abnormal contribution of all equipment monitoring data to obtain the total equipment contribution; and finally, calculating the absolute value of the difference between the total environmental contribution and the total equipment contribution. This absolute value quantitatively describes whether the anomalies in the operational data at any given time are dominated by environmental factors or by factors inherent to the equipment itself. The environmental monitoring data includes irradiance and backplane temperature, while the equipment monitoring data includes output power, output voltage, and output current.

[0060] S104. When the absolute value of the difference is greater than the judgment threshold and the symmetry recovery index of each type of equipment monitoring data is lower than the preset index, the abnormal candidate point is judged to be a fault abnormality of photovoltaic equipment; otherwise, it is judged to be a non-fault abnormality caused by environmental disturbance.

[0061] In one embodiment, a threshold for distinguishing abnormality types and a preset index for assessing the degree of recovery are set; wherein, the threshold is 0.2 and the preset index is 0.9.

[0062] On the one hand, when the symmetry recovery index of each equipment monitoring data of the abnormal candidate point is lower than the preset index, it indicates that the loss of photovoltaic equipment is irreversible. On the other hand, when the absolute value of the difference between the total contribution of environmental monitoring data and equipment monitoring data of the abnormal candidate point is greater than the judgment threshold, it indicates that in the isolated forest, the total contribution of equipment is greater than the total contribution of environment, indicating that the monitoring data that caused the operation data to become an abnormal candidate point is mainly concentrated in the equipment monitoring data.

[0063] Therefore, when the absolute value of the difference is greater than the judgment threshold and the symmetry recovery index of each equipment monitoring data is lower than the preset index, the candidate anomaly is ultimately determined to be a photovoltaic equipment failure. Conversely, if the symmetry recovery index is not lower than the preset index, meaning the photovoltaic equipment loss is reversible, or if the absolute value of the difference is not greater than the judgment threshold, indicating that the anomaly contribution of environmental monitoring parameters and equipment monitoring parameters is comparable, the system determines the anomaly to be a non-fault anomaly caused by environmental disturbance and can automatically filter the alarm to avoid false alarms. Please refer to [link to relevant documentation]. Figure 4 This is a schematic diagram of the fault diagnosis results according to the embodiments of this application. When a real fault is diagnosed, an early warning message is issued in a timely manner, and the photovoltaic equipment is repaired to ensure the stable operation of the photovoltaic equipment in the future.

[0064] In this way, by making joint decisions based on dual conditions, it is possible to accurately determine the nature of abnormal events, and ultimately achieve the beneficial effect of reducing false alarm rates and improving monitoring accuracy.

[0065] According to a second aspect of this application, this application also provides a photovoltaic equipment operation monitoring system based on big data. Figure 5 This is a structural block diagram of a photovoltaic equipment operation monitoring system based on big data, according to an embodiment of this application. Figure 5 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the photovoltaic equipment operation monitoring method based on big data according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described further here.

[0066] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. A method for monitoring the operation of photovoltaic equipment based on big data, characterized in that, The monitoring methods include: obtaining anomaly scores of photovoltaic equipment operation data at any time based on isolated forests, and using operation data with anomaly scores greater than the anomaly threshold as anomaly candidate points. The operation data includes multi-dimensional monitoring data, and the monitoring data includes environmental monitoring data and equipment monitoring data. Within the monitoring window centered on the candidate anomaly point, the symmetric recovery index of the monitoring data for each equipment type at the candidate anomaly point is calculated. This includes: within the monitoring window, before the timestamp of the candidate anomaly point, the integral of the difference between the actual measured value and the reference theoretical value in the monitoring data of any equipment type is used as the power loss; after the timestamp of the candidate anomaly point, the integral of the difference between the actual measured value and the reference theoretical value in the monitoring data of the equipment type is used as the power recovery; the symmetric recovery index is the ratio of the power recovery to the power loss. The calculation of the anomaly contribution of each monitoring data point in the anomaly candidate point includes: tracing and recording the isolated paths from the root node to the leaf node of each isolated tree in the isolated forest, and the monitoring data used by each split node on the isolated path; the anomaly contribution of any monitoring data is obtained by summing the weighted contributions of the monitoring data in all isolated trees; wherein, in a single isolated tree, the weighted contribution of the monitoring data is positively correlated with the number of times the monitoring data appears in the split nodes of the isolated path, and negatively correlated with the length of the isolated path, and the absolute value of the difference between the total contribution of environmental monitoring data and the total contribution of equipment monitoring data is calculated; When the absolute value of the difference is greater than the judgment threshold and the symmetry recovery index of the monitoring data of each equipment type is lower than the preset index, the abnormal candidate point is judged to be a fault abnormality of photovoltaic equipment; otherwise, it is judged to be a non-fault abnormality caused by environmental disturbance.

2. The photovoltaic equipment operation monitoring method based on big data according to claim 1, characterized in that, The environmental monitoring data includes irradiance and solar panel backsheet temperature; the equipment monitoring data includes the output power, output voltage, and output current of the photovoltaic equipment.

3. The photovoltaic equipment operation monitoring method based on big data according to claim 1, characterized in that, The method for obtaining the reference theoretical value includes: inputting the environmental monitoring data corresponding to the timestamp of the anomaly candidate point into a pre-trained machine learning model to obtain the reference theoretical value of the equipment monitoring data.

4. The photovoltaic equipment operation monitoring method based on big data according to claim 1, characterized in that, The calculated abnormal contribution of each monitoring data point is normalized to obtain the percentage contribution of each monitoring data point.

5. The photovoltaic equipment operation monitoring method based on big data according to claim 1, characterized in that, The process of calculating the absolute value of the difference between the total contribution of environmental monitoring data and the total contribution of equipment monitoring data includes: summing the abnormal contribution values ​​of all environmental monitoring data to obtain the total environmental contribution value; summing the abnormal contribution values ​​of all equipment monitoring data to obtain the total equipment contribution value; and calculating the absolute value of the difference between the total environmental contribution value and the total equipment contribution value.

6. The photovoltaic equipment operation monitoring method based on big data according to claim 1, characterized in that, The anomaly score obtained from the isolated forest for photovoltaic equipment operation data at any time includes: calculating the average path length of the operation data in each isolated tree of the isolated forest; the anomaly score is negatively correlated with the average path length.

7. The photovoltaic equipment operation monitoring method based on big data according to claim 1, characterized in that, The monitoring window is centered on the timestamp of the abnormal candidate point and takes a time range of multiple moments before and after it.

8. A photovoltaic equipment operation monitoring system based on big data, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the photovoltaic equipment operation monitoring method based on big data according to any one of claims 1 to 7.

Images