Method and device for monitoring operation data of photovoltaic power station

Abstract

The invention belongs to the technical field of photovoltaic power station monitoring, and discloses a photovoltaic power station operation data monitoring method and device, and the method comprises the steps: collecting multi-source data of a photovoltaic power station in an operation process, and adding a cross-dimension identifier for each piece of data; the method comprises the steps of receiving multi-source data, performing cleaning, missing value completion and normalization on the multi-source data, performing causal chain type alignment based on a cross-dimension identifier, establishing a physical transmission link constraint for photovoltaic module side data in the multi-source data and power grid interaction data, and generating a data set with causal consistency; based on the data set with causal consistency, chain-type correlation features are generated layer by layer, and the chain-type correlation features comprise first-order features, second-order features and third-order features; and the stability, the reliability and the intelligent level of the operation of the photovoltaic power station are improved.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power plant monitoring technology, and more specifically, to a method and apparatus for monitoring photovoltaic power plant operation data. Background Technology

[0002] The main problems with the methods and devices for monitoring photovoltaic power plant operation data are as follows: With the large-scale construction and application of photovoltaic (PV) power plants, the complexity of their operation and maintenance is increasing. During operation, PV power plants generate a large amount of multi-source data, including PV modules, power inverters, and environmental parameters. Existing PV power plant monitoring methods mainly rely on single-dimensional operating indicators, such as the current, voltage, or power of a single PV module, the output status of the inverter, and environmental information such as temperature and sunlight. These methods typically cannot capture the dynamic interactions between multi-dimensional data, resulting in low accuracy in anomaly detection, especially in identifying potential high-order anomaly propagation paths within the data chain.

[0003] Existing technologies in data processing mostly employ statistical analysis methods or traditional machine learning models to process and analyze single-point data, neglecting causal chains and physical transmission relationships. Therefore, they cannot systematically quantify the interactions between chain nodes, the cumulative effects of anomalies, and higher-order dynamic relationships. Traditional methods struggle to fully describe the process of anomaly transmission, amplification, or coupling within the chain, leading to delayed anomaly warnings, frequent false alarms, and a lack of traceability and interpretability for operation and maintenance decisions. Current technologies in photovoltaic power plant operation monitoring still suffer from low monitoring accuracy, delayed anomaly identification, and a lack of multi-dimensional causal analysis capabilities.

[0004] In view of this, the present invention proposes a method and device for monitoring photovoltaic power plant operation data to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a method for monitoring photovoltaic power plant operation data, comprising: S1. Collect multi-source data during the operation of the photovoltaic power station and attach cross-dimensional identifiers to each data point; S2. Receive multi-source data, clean the multi-source data, fill in missing values ​​and normalize it, and perform causal chain alignment based on cross-dimensional identifiers. Establish physical transmission link constraints between photovoltaic module-side data and grid interaction data in the multi-source data to generate a dataset with causal consistency. S3. Based on a dataset with causal consistency, generate chain-related features layer by layer. The chain-related features include first-order features, second-order features and third-order features. S4. Based on the chain-like correlation characteristics, conduct a health assessment of the operating status of photovoltaic power plants. Use the link integrity index to determine whether the causal chain is broken; use the deviation accumulation index to measure the amplification effect of anomalies in the causal chain; use the robustness index to distinguish between transient disturbances and trend anomalies, forming a multi-dimensional dynamic health assessment result. S5. Based on the dynamic health assessment results, implement graded alarms and feedback adjustments. When the link integrity is partially damaged, a level 1 alarm is triggered to notify the operation and maintenance node. When the cumulative deviation exceeds the preset cumulative deviation threshold, a level 2 alarm is triggered and an automatic adjustment strategy is executed.

[0006] Furthermore, the method for collecting multi-source data during the operation of a photovoltaic power station includes: Multi-source data during the operation of a photovoltaic power station includes photovoltaic module operating status parameters, DC voltage, current, AC power, power quality parameters, interface acquisition efficiency, power factor, wind speed, wind direction, humidity, ambient temperature, degree of contamination on the surface of photovoltaic modules, power commands and power curtailment information issued by the power grid dispatch center; Voltage sensors, current sensors, temperature sensors, and illuminance sensors are deployed on the photovoltaic module side to collect photovoltaic module operating status parameters; on the inverter side, DC voltage and current at the inverter input terminal and AC power and power quality parameters at the output terminal are collected, and efficiency and power factor are obtained through the inverter monitoring interface; outside the power station, wind speed, wind direction, humidity, and ambient temperature are collected through a weather station, and the degree of pollution on the photovoltaic module surface is obtained through dust sensors or air particulate matter sensors; power commands and power curtailment information issued by the power grid dispatch center are received.

[0007] Furthermore, the method for attaching cross-dimensional identifiers to each data entry includes: For each piece of collected data, generate and attach a cross-dimensional identifier. The cross-dimensional identifier includes a source identifier and a functional role identifier. The source identifier is used to indicate that the multi-source data belongs to any of the following sources: the component side, the inverter side, the environment side, or the grid interaction side. Functional role identifiers are used to indicate the attributes of multi-source data as causal inputs, intermediate quantities, or outputs in a causal chain; cross-dimensional identifiers can be generated with an attached collection timestamp and collection device identifier to form a structured three-dimensional index; after data collection is completed, cross-dimensional identifiers are stored or transmitted synchronously with the data.

[0008] Furthermore, the method for cleaning, imputing missing values, and normalizing multi-source data includes: It receives multi-source data and performs data cleaning, missing value completion, and normalization on the multi-source data. Data cleaning includes outlier detection and processing. Outlier detection adopts statistical outlier rules, physical constraint rules, and time continuity rules. Processing methods include deleting, replacing, or marking outliers. Missing value completion determines the location of missing data in the causal chain based on cross-dimensional identifiers and functional roles, and completes the missing data through multiple strategies such as physical model constraints, time series extrapolation, and correlation feature extrapolation, while retaining the completion identifier; data normalization includes dimensional standardization of various data from the component side, inverter side, environment side, and grid interaction side to obtain standardized multi-source data.

[0009] Furthermore, the method for obtaining a dataset with causal consistency includes: Based on standardized multi-source data and cross-dimensional identifiers for each data point, causal chain alignment is performed, taking component-side data as the cause and inverter and grid interaction data as the effect, and aligning various types of data sequentially along the physical operation link of the photovoltaic system. During the alignment process, physical transmission link constraints are established based on the physical mapping relationship between the output voltage and current of the photovoltaic module and the power at the inverter end, the power transfer relationship between the inverter output power and the grid interaction data, and the response relationship of environmental data to the output power of the module. Standardized multi-source data that violates physical constraints are labeled or corrected, and the data that best fits the physical model is selected as the chain node value through constraint optimization methods. Through alignment and constraint operations, various types of data are combined along the chain direction to generate a dataset with causal consistency.

[0010] Furthermore, the method for generating chain-related features includes: Based on the dataset, chain-related features are generated layer by layer, including first-order features, second-order features and third-order features; the first-order features are the state features of each chain node, including the mean voltage and current variance of the component side, the inverter output power fluctuation rate and the environmental parameter change rate. Second-order features are generated based on adjacent nodes and their first-order features, reflecting the interaction relationships between nodes in the chain, including the cumulative deviation between component power and inverter power, the sensitivity of environmental parameters to power output, and the rate of coupling change. The third-order features are generated based on three consecutive nodes and their first-order and second-order features. They are used to capture high-order dynamic correlations and anomaly propagation patterns in the chain, including the power transfer consistency index of component output voltage, inverter power, and grid load, as well as the three-node coupling sensitivity of ambient temperature, component power, and grid frequency. The first-order, second-order, and third-order features are combined according to the chain order and hierarchical relationship to form a hierarchical chain-related feature matrix. Each feature retains the cross-dimensional identifier and timestamp of the corresponding node.

[0011] Furthermore, the method for generating multi-dimensional dynamic health assessment results includes: Based on the chain-like association characteristics, the link integrity index is calculated to determine whether there are breaks or discontinuities in each node and its connection in the causal chain; the deviation accumulation index is calculated to measure the propagation and amplification effect of anomalies between nodes in the chain. Calculate robustness indices to distinguish between transient disturbances and trend anomalies in the chain; integrate link integrity, deviation accumulation, and robustness indices to form a multi-dimensional dynamic health assessment result. Each chain node can obtain an independent health score, while generating the overall health status of the entire causal chain.

[0012] Furthermore, the method for triggering a level-one alarm notification to the operation and maintenance node includes: Receive the multi-dimensional health assessment results output from the dynamic health assessment step, and determine the thresholds for the link integrity index, deviation accumulation index, and robustness index; when the link integrity index is lower than the preset link integrity threshold, it is determined that some nodes in the causal chain are broken or discontinuous, and a first-level alarm is automatically triggered. Level 1 alarms include alarm level, cross-dimensional identifiers of affected link nodes, timestamps of the anomaly occurrence, and anomaly descriptions, alerting maintenance nodes to potential physical or data anomalies. After a Level 1 alarm is triggered, feedback adjustment strategies are executed, including adjusting data collection frequency, isolating or marking damaged node data, updating data processing parameters, or reconstructing the local chain structure.

[0013] Furthermore, the method for triggering a secondary alarm and executing an automatic adjustment strategy includes: When the deviation accumulation index output by the dynamic health assessment step exceeds the preset deviation accumulation threshold, it is determined that the anomaly in the chain may be amplified or spread between nodes, and a level 2 alarm is automatically triggered. The level 2 alarm includes the alarm level, the cross-dimensional identifiers of the affected chain nodes and chain segments, the time period of the anomaly, the anomaly type, and the anomaly spread trend information. After a level 2 alarm is triggered, an automatic adjustment strategy is executed based on the location and type of the abnormal node. This includes adjusting the power output control parameters of components or inverters, adjusting data acquisition or processing algorithm parameters, isolating or prioritizing the repair of affected nodes, updating the chain health calculation model, or reconstructing the local chain structure.

[0014] A monitoring device for photovoltaic power plant operation data includes a monitoring device, wherein the monitoring device is equipped with a data acquisition module, a cross-domain alignment module, a correlation feature module, a dynamic health monitoring module, and an adaptive alarm feedback module, forming a photovoltaic power plant operation data monitoring system.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention, through a hierarchical combination of first-order, second-order, and third-order features, fully describes the node status, inter-node interactions, and high-order dynamic correlations of photovoltaic modules, inverters, and environmental data, achieving a global and refined characterization of the operating status of photovoltaic power plants. Second-order features capture the interaction relationships between adjacent nodes, while third-order features capture the high-order dynamic correlations and anomaly propagation patterns of three consecutive nodes, solving the problem in existing technologies that cannot quantify the causal relationships of chain nodes and anomaly propagation paths.

[0016] The hierarchical chain-like feature matrix provides anomaly accumulation and coupling sensitivity indicators between nodes, enabling the system to identify the amplification effect and potential propagation paths of anomalies, thus improving the timeliness and accuracy of anomaly warnings. Each feature retains a cross-dimensional identifier and timestamp, allowing anomaly events to be traced back to specific chain nodes, supporting operations personnel in quickly locating problems and providing a reliable data foundation for subsequent dynamic health assessments and tiered alarms. Based on the chain-like associated feature matrix, the health assessment and alarm modules can distinguish between transient disturbances and trend anomalies, thereby enabling tiered alarms and automatic adjustment strategies. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a method for monitoring photovoltaic power plant operation data according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0019] Please see Figure 1 As shown in the figure, this embodiment further illustrates the method for monitoring photovoltaic power plant operation data proposed in this invention, including: S1. Collect multi-source data during the operation of the photovoltaic power station and attach cross-dimensional identifiers to each data point; S2. Receive multi-source data, clean the multi-source data, fill in missing values ​​and normalize it, and perform causal chain alignment based on cross-dimensional identifiers. Establish physical transmission link constraints between photovoltaic module-side data and grid interaction data in the multi-source data to generate a dataset with causal consistency. S3. Based on a dataset with causal consistency, generate chain-related features layer by layer. The chain-related features include first-order features, second-order features and third-order features. S4. Based on the chain-like correlation characteristics, conduct a health assessment of the operating status of photovoltaic power plants. Use the link integrity index to determine whether the causal chain is broken; use the deviation accumulation index to measure the amplification effect of anomalies in the causal chain; use the robustness index to distinguish between transient disturbances and trend anomalies, forming a multi-dimensional dynamic health assessment result. S5. Based on the dynamic health assessment results, implement graded alarms and feedback adjustments. When the link integrity is partially damaged, a level 1 alarm is triggered to notify the operation and maintenance node. When the cumulative deviation exceeds the preset cumulative deviation threshold, a level 2 alarm is triggered and an automatic adjustment strategy is executed.

[0020] Methods for collecting multi-source data during the operation of photovoltaic power plants include: Multi-source data during the operation of a photovoltaic power station includes photovoltaic module operating status parameters, DC voltage, current, AC power, power quality parameters, interface acquisition efficiency, power factor, wind speed, wind direction, humidity, ambient temperature, degree of contamination on the surface of photovoltaic modules, power commands and power curtailment information issued by the power grid dispatch center; Voltage sensors, current sensors, temperature sensors, and illuminance sensors are deployed on the photovoltaic module side to collect photovoltaic module operating status parameters; on the inverter side, DC voltage and current at the inverter input terminal and AC power and power quality parameters at the output terminal are collected, and efficiency and power factor are obtained through the inverter monitoring interface; outside the power station, wind speed, wind direction, humidity, and ambient temperature are collected through a weather station, and the degree of pollution on the photovoltaic module surface is obtained through dust sensors or air particulate matter sensors; power commands and power curtailment information issued by the power grid dispatch center are received.

[0021] Methods for attaching cross-dimensional identifiers to each data entry include: For each piece of collected data, generate and attach a cross-dimensional identifier. The cross-dimensional identifier includes a source identifier and a functional role identifier. The source identifier is used to indicate that the multi-source data belongs to any of the following sources: the component side, the inverter side, the environment side, or the grid interaction side. Functional role identifiers are used to indicate the attributes of multi-source data as causal inputs, intermediate quantities, or outputs in a causal chain; cross-dimensional identifiers can be generated with an attached collection timestamp and collection device identifier to form a structured three-dimensional index; after data collection is completed, cross-dimensional identifiers are stored or transmitted synchronously with the data.

[0022] Methods for cleaning, imputing missing values, and normalizing multi-source data include: It receives multi-source data and performs data cleaning, missing value completion, and normalization on the multi-source data. Data cleaning includes outlier detection and processing. Outlier detection adopts statistical outlier rules, physical constraint rules, and time continuity rules. Processing methods include deleting, replacing, or marking outliers. Missing value completion determines the location of missing data in the causal chain based on cross-dimensional identifiers and functional roles, and completes the missing data through multiple strategies such as physical model constraints, time series extrapolation, and correlation feature extrapolation, while retaining the completion identifier; data normalization includes dimensional standardization of various data from the component side, inverter side, environment side, and grid interaction side to obtain standardized multi-source data.

[0023] Methods for datasets that demonstrate causal consistency include: Based on standardized multi-source data and cross-dimensional identifiers for each data point, causal chain alignment is performed, taking component-side data as the cause and inverter and grid interaction data as the effect, and aligning various types of data sequentially along the physical operation link of the photovoltaic system. During the alignment process, physical transmission link constraints are established based on the physical mapping relationship between the output voltage and current of the photovoltaic module and the power at the inverter end, the power transfer relationship between the inverter output power and the grid interaction data, and the response relationship of environmental data to the output power of the module. Standardized multi-source data that violates physical constraints are labeled or corrected, and the data that best fits the physical model is selected as the chain node value through constraint optimization methods. Through alignment and constraint operations, various types of data are combined along the chain direction to generate a dataset with causal consistency.

[0024] For example, at a certain moment, the standardized multi-source data collected by a photovoltaic power station includes the output voltage and current of the photovoltaic modules, the output power of the inverter, the input power to the grid, and the ambient temperature. Causal chain alignment first takes the voltage and current of the photovoltaic modules as the cause, and the output power of the inverter and the input power to the grid as the effect, aligning various types of data sequentially along the physical operation link of the modules—inverter—grid, while simultaneously using ambient temperature as an input correction factor.

[0025] During the alignment process, the system checks against physical constraints, such as ensuring that the product of the component voltage and current is approximately equal to the inverter input power, that the inverter output power conforms to the power transfer relationship of the grid, and that ambient temperature will have a degrading effect on the component output power. If an inconsistency is detected between the inverter output power and the component power, the system corrects the data using constraint optimization methods and selects the value that best fits the physical model as the chain node value.

[0026] After the above causal chain alignment and physical constraint correction, the resulting multi-source data is combined along the photovoltaic system link direction to ensure that the value of each chain node is consistent in causality, thereby generating a dataset with causal consistency, which can be directly used for anomaly detection, causal analysis or high-order dynamic relationship modeling.

[0027] Methods for generating chain-related features include: Based on the dataset, chain-related features are generated layer by layer, including first-order features, second-order features and third-order features; the first-order features are the state features of each chain node, including the mean voltage and current variance of the component side, the inverter output power fluctuation rate and the environmental parameter change rate. Second-order features are generated based on adjacent nodes and their first-order features, reflecting the interaction relationships between nodes in the chain, including the cumulative deviation between component power and inverter power, the sensitivity of environmental parameters to power output, and the rate of coupling change. The third-order features are generated based on three consecutive nodes and their first-order and second-order features. They are used to capture high-order dynamic correlations and anomaly propagation patterns in the chain, including the power transfer consistency index of component output voltage, inverter power, and grid load, as well as the three-node coupling sensitivity of ambient temperature, component power, and grid frequency. The first-order, second-order, and third-order features are combined according to the chain order and hierarchical relationship to form a hierarchical chain-related feature matrix. Each feature retains the cross-dimensional identifier and timestamp of the corresponding node.

[0028] Methods for generating multi-dimensional dynamic health assessment results include: Based on the chain-like association characteristics, the link integrity index is calculated to determine whether there are breaks or discontinuities in each node and its connection in the causal chain; the deviation accumulation index is calculated to measure the propagation and amplification effect of anomalies between nodes in the chain. Calculate robustness indices to distinguish between transient disturbances and trend anomalies in the chain; integrate link integrity, deviation accumulation, and robustness indices to form a multi-dimensional dynamic health assessment result. Each chain node can obtain an independent health score, while generating the overall health status of the entire causal chain.

[0029] Methods for triggering a Level 1 alarm notification on the operations and maintenance node include: Receive the multi-dimensional health assessment results output from the dynamic health assessment step, and determine the thresholds for the link integrity index, deviation accumulation index, and robustness index; when the link integrity index is lower than the preset link integrity threshold, it is determined that some nodes in the causal chain are broken or discontinuous, and a first-level alarm is automatically triggered. Level 1 alarms include alarm level, cross-dimensional identifiers of affected link nodes, timestamps of the anomaly occurrence, and anomaly descriptions, alerting maintenance nodes to potential physical or data anomalies. After a Level 1 alarm is triggered, feedback adjustment strategies are executed, including adjusting data collection frequency, isolating or marking damaged node data, updating data processing parameters, or reconstructing the local chain structure.

[0030] Methods for triggering a level 2 alarm and executing an automatic adjustment strategy include: When the deviation accumulation index output by the dynamic health assessment step exceeds the preset deviation accumulation threshold, it is determined that the anomaly in the chain may be amplified or spread between nodes, and a level 2 alarm is automatically triggered. The level 2 alarm includes the alarm level, the cross-dimensional identifiers of the affected chain nodes and chain segments, the time period of the anomaly, the anomaly type, and the anomaly spread trend information. After a level 2 alarm is triggered, an automatic adjustment strategy is executed based on the location and type of the abnormal node. This includes adjusting the power output control parameters of components or inverters, adjusting data acquisition or processing algorithm parameters, isolating or prioritizing the repair of affected nodes, updating the chain health calculation model, or reconstructing the local chain structure.

[0031] A monitoring device for photovoltaic power plant operation data includes a monitoring device, wherein the monitoring device is equipped with a data acquisition module, a cross-domain alignment module, a correlation feature module, a dynamic health monitoring module, and an adaptive alarm feedback module, forming a photovoltaic power plant operation data monitoring system.

[0032] In this embodiment, a hierarchical combination of first-order, second-order, and third-order features is used to fully describe the node status, inter-node interactions, and high-order dynamic correlations of photovoltaic modules, inverters, and environmental data, thereby achieving a global and detailed characterization of the operating status of photovoltaic power plants. Second-order features capture the interaction relationships between adjacent nodes, while third-order features capture the high-order dynamic correlations and anomaly propagation patterns of three consecutive nodes, solving the problem in existing technologies that cannot quantify the causal relationships and anomaly propagation paths of chain nodes.

[0033] The hierarchical chain-like feature matrix provides anomaly accumulation and coupling sensitivity indicators between nodes, enabling the system to identify the amplification effect and potential propagation paths of anomalies, thus improving the timeliness and accuracy of anomaly warnings. Each feature retains a cross-dimensional identifier and timestamp, allowing anomaly events to be traced back to specific chain nodes, supporting operations personnel in quickly locating problems and providing a reliable data foundation for subsequent dynamic health assessments and tiered alarms. Based on the chain-like associated feature matrix, the health assessment and alarm modules can distinguish between transient disturbances and trend anomalies, thereby enabling tiered alarms and automatic adjustment strategies.

[0034] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0035] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method of monitoring operation data of a photovoltaic power plant, characterized in that, include: S1. Collect multi-source data during the operation of the photovoltaic power station and attach cross-dimensional identifiers to each data point; S2. Receive multi-source data, clean the multi-source data, fill in missing values ​​and normalize it, and perform causal chain alignment based on cross-dimensional identifiers. Establish physical transmission link constraints between photovoltaic module-side data and grid interaction data in the multi-source data to generate a dataset with causal consistency. S3. Based on a dataset with causal consistency, generate chain-related features layer by layer. The chain-related features include first-order features, second-order features and third-order features. S4. Based on the chain-like correlation characteristics, conduct a health assessment of the operating status of photovoltaic power plants. Use the link integrity index to determine whether the causal chain is broken; use the deviation accumulation index to measure the amplification effect of anomalies in the causal chain; use the robustness index to distinguish between transient disturbances and trend anomalies, forming a multi-dimensional dynamic health assessment result. S5. Based on the dynamic health assessment results, execute graded alarms and feedback adjustments. When the integrity of the link is partially damaged, trigger a level 1 alarm to notify the operation and maintenance node. When the cumulative deviation exceeds the preset cumulative deviation threshold, a level 2 alarm is triggered and an automatic adjustment strategy is executed.

2. The method for monitoring operation data of a photovoltaic power station according to claim 1, characterized in that, The method for collecting multi-source data during the operation of a photovoltaic power station includes: Multi-source data during the operation of a photovoltaic power station includes photovoltaic module operating status parameters, DC voltage, current, AC power, power quality parameters, interface acquisition efficiency, power factor, wind speed, wind direction, humidity, ambient temperature, degree of contamination on the surface of photovoltaic modules, power commands and power curtailment information issued by the power grid dispatch center; Voltage sensors, current sensors, temperature sensors, and illuminance sensors are deployed on the photovoltaic module side to collect photovoltaic module operating status parameters; on the inverter side, DC voltage and current at the inverter input terminal and AC power and power quality parameters at the output terminal are collected, and efficiency and power factor are obtained through the inverter monitoring interface; outside the power station, wind speed, wind direction, humidity, and ambient temperature are collected through a weather station, and the degree of pollution on the photovoltaic module surface is obtained through dust sensors or air particulate matter sensors; power commands and power curtailment information issued by the power grid dispatch center are received.

3. The method for monitoring photovoltaic power plant operation data according to claim 2, characterized in that, The method for attaching cross-dimensional identifiers to each data entry includes: For each piece of collected data, generate and attach a cross-dimensional identifier. The cross-dimensional identifier includes a source identifier and a functional role identifier. The source identifier is used to indicate that the multi-source data belongs to any of the following sources: the component side, the inverter side, the environment side, or the grid interaction side. Functional role identifiers are used to indicate the attributes of multi-source data as causal inputs, intermediate quantities, or outputs in a causal chain; cross-dimensional identifiers can be generated with an attached collection timestamp and collection device identifier to form a structured three-dimensional index; after data collection is completed, cross-dimensional identifiers are stored or transmitted synchronously with the data.

4. The method for monitoring photovoltaic power plant operation data according to claim 3, characterized in that, The methods for cleaning, imputing missing values, and normalizing multi-source data include: It receives multi-source data and performs data cleaning, missing value completion, and normalization on the multi-source data. Data cleaning includes outlier detection and processing. Outlier detection adopts statistical outlier rules, physical constraint rules, and time continuity rules. Processing methods include deleting, replacing, or marking outliers. Missing value completion determines the location of missing data in the causal chain based on cross-dimensional identifiers and functional roles, and completes the missing data through multiple strategies such as physical model constraints, time series extrapolation, and correlation feature extrapolation, while retaining the completion identifier; data normalization includes dimensional standardization of various data from the component side, inverter side, environment side, and grid interaction side to obtain standardized multi-source data.

5. The method for monitoring photovoltaic power plant operation data according to claim 4, characterized in that, The method for providing a dataset with causal consistency includes: Based on standardized multi-source data and cross-dimensional identifiers for each data point, causal chain alignment is performed, taking component-side data as the cause and inverter and grid interaction data as the effect, and aligning various types of data sequentially along the physical operation link of the photovoltaic system. During the alignment process, physical transmission link constraints are established based on the physical mapping relationship between the output voltage and current of the photovoltaic module and the power at the inverter end, the power transfer relationship between the inverter output power and the grid interaction data, and the response relationship of environmental data to the output power of the module. Standardized multi-source data that violates physical constraints are labeled or corrected, and the data that best fits the physical model is selected as the chain node value through constraint optimization methods. Through alignment and constraint operations, various types of data are combined along the chain direction to generate a dataset with causal consistency.

6. The method for monitoring photovoltaic power plant operation data according to claim 5, characterized in that, The method for generating chain-related features includes: Based on the dataset, chain-related features are generated layer by layer, including first-order features, second-order features and third-order features; the first-order features are the state features of each chain node, including the mean voltage and current variance of the component side, the inverter output power fluctuation rate and the environmental parameter change rate. Second-order features are generated based on adjacent nodes and their first-order features, reflecting the interaction relationships between nodes in the chain, including the cumulative deviation between component power and inverter power, the sensitivity of environmental parameters to power output, and the rate of coupling change. The third-order features are generated based on three consecutive nodes and their first-order and second-order features. They are used to capture high-order dynamic correlations and anomaly propagation patterns in the chain, including the power transfer consistency index of component output voltage, inverter power, and grid load, as well as the three-node coupling sensitivity of ambient temperature, component power, and grid frequency. The first-order, second-order, and third-order features are combined according to the chain order and hierarchical relationship to form a hierarchical chain-related feature matrix. Each feature retains the cross-dimensional identifier and timestamp of the corresponding node.

7. The method for monitoring photovoltaic power plant operation data according to claim 6, characterized in that, The method for generating multi-dimensional dynamic health assessment results includes: Based on the chain-like association characteristics, the link integrity index is calculated to determine whether there are breaks or discontinuities in each node and its connection in the causal chain; the deviation accumulation index is calculated to measure the propagation and amplification effect of anomalies between nodes in the chain. Calculate robustness indices to distinguish between transient disturbances and trend anomalies in the chain; integrate link integrity, deviation accumulation, and robustness indices to form a multi-dimensional dynamic health assessment result. Each chain node can obtain an independent health score, while generating the overall health status of the entire causal chain.

8. The method for monitoring photovoltaic power plant operation data according to claim 7, characterized in that, The method for triggering a Level 1 alarm notification to the operation and maintenance node includes: Receive the multi-dimensional health assessment results output from the dynamic health assessment step, and determine the thresholds for the link integrity index, deviation accumulation index, and robustness index; when the link integrity index is lower than the preset link integrity threshold, it is determined that some nodes in the causal chain are broken or discontinuous, and a first-level alarm is automatically triggered. Level 1 alarms include alarm level, cross-dimensional identifiers of affected link nodes, timestamps of the anomaly occurrence, and anomaly descriptions, alerting maintenance nodes to potential physical or data anomalies. After a Level 1 alarm is triggered, feedback adjustment strategies are executed, including adjusting data collection frequency, isolating or marking damaged node data, updating data processing parameters, or reconstructing the local chain structure.

9. A method for monitoring photovoltaic power plant operation data according to claim 8, characterized in that, The method for triggering a level-two alarm and executing an automatic adjustment strategy includes: When the deviation accumulation index output by the dynamic health assessment step exceeds the preset deviation accumulation threshold, it is determined that the anomaly in the chain may be amplified or spread between nodes, and a level 2 alarm is automatically triggered. The level 2 alarm includes the alarm level, the cross-dimensional identifiers of the affected chain nodes and chain segments, the time period of the anomaly, the anomaly type, and the anomaly spread trend information. After a level 2 alarm is triggered, an automatic adjustment strategy is executed based on the location and type of the abnormal node. This includes adjusting the power output control parameters of components or inverters, adjusting data acquisition or processing algorithm parameters, isolating or prioritizing the repair of affected nodes, updating the chain health calculation model, or reconstructing the local chain structure.

10. A monitoring device for photovoltaic power plant operation data, employing a monitoring method for photovoltaic power plant operation data as described in any one of claims 1-9, characterized in that, The system includes a monitoring device, which is equipped with a data acquisition module, a cross-domain alignment module, a correlation feature module, a dynamic health monitoring module, and an adaptive alarm feedback module to monitor the operation data of a photovoltaic power station.

Images