Monitoring methods, computer equipment and storage media for photovoltaic new energy systems

Abstract

This application discloses a method, computer equipment, and storage medium for monitoring a photovoltaic new energy system. The method includes: acquiring the current operating data of the target photovoltaic system and mining its spatiotemporal distribution characteristics, historical degradation characteristics, and micro-parameter characteristics; modeling an adjacency matrix based on the spatiotemporal distribution characteristics, historical degradation characteristics, and micro-parameter characteristics to obtain a string topology model of the target photovoltaic system; processing the current operating data using the string topology model to generate a dynamic mismatch feature set; using the dynamic mismatch feature set as input, calling a time-series prediction model to output the mismatch trend evolution result of the target photovoltaic system; generating a string-level control strategy based on the mismatch trend evolution result; and optimizing the control by combining the power allocation logic of the target inverter of the target photovoltaic system with the string-level control strategy during execution, and outputting the optimized control signal. This method improves the accuracy of photovoltaic system mismatch monitoring and the effectiveness of the control strategy.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and in particular to a method for monitoring a photovoltaic new energy system, a computer device, and a storage medium. Background Technology

[0002] With the accelerated global energy structure transformation, the installed capacity of photovoltaic (PV) new energy systems continues to grow, and their operational stability and power generation efficiency have become core issues of concern to the industry. Module mismatch is one of the key factors affecting the efficiency of PV systems. Module mismatch can lead to a decrease in system output power, and in severe cases, it can even cause hot spot effects and shorten the lifespan of the modules.

[0003] Existing technologies for monitoring photovoltaic (PV) system module mismatch mainly fall into three categories: monitoring methods based on operational data, image-based monitoring methods, and model-based monitoring methods. However, existing technologies cannot guarantee the accuracy of PV system mismatch monitoring and the effectiveness of control strategies in PV system module mismatch monitoring and control. Summary of the Invention

[0004] This application provides a photovoltaic new energy system monitoring method, computer equipment, and storage medium to improve the accuracy of photovoltaic system mismatch monitoring and the effectiveness of control strategies.

[0005] This application provides a method for monitoring a photovoltaic new energy system, applied to computer equipment, the method comprising: The current operating data of the target photovoltaic system is obtained, and the spatiotemporal distribution characteristics of the temperature gradient, shadow movement trajectory and component mismatch of the same array are mined to characterize the temporal coupling relationship between the component degradation rate curve and the degree of mismatch, as well as the microscopic parameter characteristics of the influence of the microcrack distribution in the EL image on the local mismatch. Based on the spatiotemporal distribution characteristics, the historical degradation characteristics, and the microscopic parameter characteristics, an adjacency matrix model is performed to obtain the string topology model of the target photovoltaic system. The current running data is processed using the string topology relationship model to generate a dynamic mismatch feature set. Using the dynamic mismatch feature set as input, the time series prediction model is called to output the mismatch trend evolution result of the target photovoltaic system. Based on the mismatch trend evolution results, a string-level control strategy is generated. During the execution of the string-level control strategy, the power distribution logic of the target inverter of the target photovoltaic system is combined to optimize the control, and the optimized control signal is output.

[0006] One embodiment of this application provides a computer device, including: A processor; a storage device having a computer program stored thereon; a network interface for providing network communication functions; when the computer program is executed by the processor, the processor enables the processor to implement any of the photovoltaic new energy system monitoring methods described above.

[0007] One embodiment of this application provides a readable storage medium on which a program or instruction is stored. When the program or instruction is executed by a processor, it implements the steps of the photovoltaic new energy system monitoring method.

[0008] Therefore, the embodiments of this application have the following beneficial effects: The embodiments of this application achieve a full-link innovation in photovoltaic system mismatch monitoring and control through the collaborative logic of multi-dimensional feature mining, adjacency matrix modeling, time-series prediction, and control optimization. First, by simultaneously mining spatiotemporal distribution features, historical degradation features, and micro-parameter features, it breaks through the limitations of traditional single-feature monitoring, constructing a multi-scale feature system from macro-array operation to micro-component defects, providing data support for mismatch analysis. Second, adjacency matrix modeling based on multi-dimensional features transforms component relationships into a quantifiable topological model, accurately restoring the actual connection logic and influence relationships within strings and between components, solving the problem of traditional topological modeling neglecting multi-feature coupling. Third, by using the string topological relationship model to generate a dynamic mismatch feature set and combining it with a time-series prediction model to output mismatch trend evolution results, it achieves accurate prediction of mismatch trends, overcoming the deficiency of traditional static analysis in not being able to provide early warnings. Finally, based on the trend results, a string-level control strategy is generated and optimized in conjunction with the inverter power allocation logic, achieving deep adaptation between the control strategy and hardware operation logic, improving the stability and efficiency of system power output.

[0009] The embodiments of this application form a logic of feature mining, topology modeling, trend prediction, and control optimization, which improves the accuracy of photovoltaic system mismatch monitoring, the foresight of trend prediction, and the effectiveness of control strategies, enabling the photovoltaic system to operate efficiently and stably. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart of a photovoltaic new energy system monitoring method provided in an embodiment of this application.

[0012] Figure 2This is a schematic diagram of the basic structure of a computer device provided in an embodiment of this application.

[0013] Figure 3 This is a functional block diagram of a photovoltaic new energy system monitoring device provided in an embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0015] Please see Figure 1 , Figure 1 This is a flowchart of a photovoltaic new energy system monitoring method provided in an embodiment of this application. The method can be executed by a computer device or by a computer device and a server. The method may include steps 110-140.

[0016] Step 110: Obtain the current operating data of the target photovoltaic system, mine the spatiotemporal distribution characteristics of the temperature gradient, shadow movement trajectory and component mismatch of the same array, the historical degradation characteristics of the temporal coupling relationship between the component attenuation rate curve and the degree of mismatch, and the microscopic parameter characteristics of the influence of the microcrack distribution in the EL image on the local mismatch.

[0017] In this embodiment, the target photovoltaic system is a distributed photovoltaic system installed on the roof of an industrial park, comprising three arrays: A, B, and C. Each array consists of several components, identified as Z1, Z2, Z3…Zn. The photovoltaic new energy system monitoring system acquires the current operating data of the target photovoltaic system through temperature sensors deployed on the surface of each component, shadow monitoring cameras installed at the edge of the array, string-level power acquisition devices, and EL image acquisition equipment. The current operating data includes real-time temperature data of each component, real-time shadow coverage data of each array, real-time output power data of each component, and EL image data of each component.

[0018] Then, the photovoltaic new energy system monitoring system performs feature mining on the current operating data: For spatiotemporal distribution features, it obtains the component temperature gradient by analyzing the differences in real-time temperature data of adjacent components in the same array; it obtains the shadow movement trajectory by analyzing the changes in the shadow coverage area in continuous frame images captured by the shadow monitoring camera; it obtains the component mismatch degree by analyzing the difference between the real-time output power data of each component in the same array and the average output power data of the array; and it then performs correlation analysis on the component temperature gradient, shadow movement trajectory, and component mismatch degree to obtain the spatiotemporal distribution features characterizing the relationship between the three. This spatiotemporal distribution feature is a multi-dimensional vector containing multiple values ​​of the temperature gradient dimension, shadow trajectory dimension, and mismatch degree dimension. For historical degradation features, the photovoltaic new energy system monitoring system retrieves the historical operating data of the target photovoltaic system. The historical operating data includes the output power data and degradation rate data of each component in the past six months. By analyzing the degradation rate data of each component over time, it obtains the component degradation rate curve. The system analyzes the time-series changes in the difference between the output power data of each component and the average output power data of the array to obtain the time-series change data of the component mismatch. Then, it couples the component attenuation rate curve with the time-series change data of the component mismatch to obtain historical degradation features characterizing the time-series coupling relationship between the two. These historical degradation features are multi-dimensional vectors containing multiple values ​​for both the attenuation rate curve dimension and the mismatch degree dimension. For micro-parameter features, the photovoltaic new energy system monitoring system processes the EL images of each component acquired by the EL image acquisition device. It identifies hidden crack regions in the EL images using image segmentation algorithms to obtain hidden crack distribution data. By analyzing the correlation between the hidden crack distribution data and the output power data of the component, it obtains the impact data of the hidden crack distribution on local mismatch. Finally, it integrates the hidden crack distribution data and the impact data to obtain micro-parameter features characterizing the impact of the hidden crack distribution in the EL image on local mismatch. These micro-parameter features are multi-dimensional vectors containing multiple values ​​for both the hidden crack distribution dimension and the local mismatch impact dimension.

[0019] Step 120: Based on the spatiotemporal distribution characteristics, historical degradation characteristics, and microscopic parameter characteristics, adjacency matrix modeling is performed to obtain the string topology model of the target photovoltaic system.

[0020] In this embodiment of the application, the photovoltaic new energy system monitoring system performs adjacency matrix modeling based on the spatiotemporal distribution characteristics, historical degradation characteristics and micro-parameter characteristics obtained in step 110, so as to construct the string topology relationship model of the target photovoltaic system.

[0021] Step 121: Extract the temperature gradient correlation vector, shadow trajectory correlation vector, and mismatch correlation vector from the spatiotemporal distribution features. The temperature gradient correlation vector represents the degree of correlation of temperature gradients between different components in the same array. The shadow trajectory correlation vector represents the degree of correlation of shadow movement trajectories between different components in the same array. The mismatch correlation vector represents the degree of correlation of mismatch between different components in the same array.

[0022] In this embodiment, the photovoltaic new energy system monitoring system analyzes the spatiotemporal distribution features obtained in step 110 and extracts the temperature gradient correlation vector, shadow trajectory correlation vector, and mismatch correlation vector. For the temperature gradient correlation vector, the correlation degree between the temperature gradients of the two components is obtained by calculating the Pearson correlation coefficient of the real-time temperature data of any two components in the same array, thus obtaining the temperature gradient correlation vector. Each element of this vector corresponds to the numerical value of the temperature gradient correlation degree between the two components in the same array. For the shadow trajectory correlation vector, the correlation degree between the shadow movement trajectories of the two components is obtained by calculating the dynamic time warping distance of the shadow coverage time series data of any two components in the same array, thus obtaining the shadow trajectory correlation vector. Each element of this vector corresponds to the numerical value of the shadow trajectory correlation degree between the two components in the same array. For the mismatch correlation vector, the correlation degree between the mismatch degree data of any two components in the same array is obtained by calculating the cosine similarity of the mismatch degree data of the two components in the same array, thus obtaining the mismatch correlation vector. Each element of this vector corresponds to the numerical value of the mismatch degree correlation degree between the two components in the same array.

[0023] Step 122: Extract the attenuation rate coupling vector and mismatch degree coupling vector from the historical degradation features. The attenuation rate coupling vector represents the degree of coupling between the attenuation rate curves of different components in the same string, and the mismatch degree coupling vector represents the degree of coupling between the degree of mismatch of different components in the same string.

[0024] In this embodiment, the photovoltaic new energy system monitoring system analyzes the historical degradation features obtained in step 110 and extracts the attenuation rate coupling vector and the mismatch degree coupling vector. For the attenuation rate coupling vector, the coupling degree between the attenuation rate curves of any two components in the same string is obtained by calculating the mean square error of the attenuation rate curves of the two components, and then the attenuation rate coupling vector is obtained. Each element of the vector corresponds to the coupling degree value of the attenuation rate curves of the two components in the same string. For the mismatch degree coupling vector, the coupling degree of the mismatch degree between the two components is obtained by calculating the mutual information value of the time-series change data of the mismatch degree of any two components in the same string, and then the mismatch degree coupling vector is obtained. Each element of the vector corresponds to the coupling degree value of the mismatch degree between the two components in the same string.

[0025] Step 123: Extract the hidden crack distribution correlation vector from the micro-parameter features. The hidden crack distribution correlation vector represents the degree of correlation between the hidden crack distribution in different regions of the same component and the degree of correlation with the degree of mismatch with adjacent components.

[0026] In this embodiment of the application, the photovoltaic new energy system monitoring system analyzes the microscopic parameter features obtained in step 110 and extracts the hidden crack distribution correlation vector: for the correlation degree of hidden crack distribution in different regions of the same component, the correlation degree of hidden crack distribution in different regions of the component is obtained by calculating the covariance of the hidden crack area in different regions of the component EL image; for the correlation degree of mismatch between the same component and adjacent components, the correlation degree of mismatch between the component and adjacent components is obtained by calculating the correlation coefficient between the mismatch data of the component and the mismatch data of adjacent components; then the correlation degree of hidden crack distribution in different regions of the same component and the correlation degree of mismatch with adjacent components are integrated to obtain the hidden crack distribution correlation vector, and each element of the vector corresponds to the correlation degree of hidden crack distribution in different regions of the same component or the correlation degree of mismatch with adjacent components.

[0027] Step 124: Using the component identifiers of the target photovoltaic system as nodes, and the temperature gradient association vector, shadow trajectory association vector, mismatch association vector, attenuation rate coupling vector, mismatch degree coupling vector, and hidden crack distribution association vector as edge weights, construct an initial adjacency matrix. The rows and columns of the initial adjacency matrix are all component identifiers, and the matrix elements are the edge weights between corresponding two component identifiers.

[0028] In this embodiment, the component identifiers of the target photovoltaic system are Z1, Z2, Z3, Z4, and Z5. The photovoltaic new energy system monitoring system uses these component identifiers as nodes to construct an initial adjacency matrix of 5 rows and 5 columns, with the rows and columns corresponding to component identifiers Z1 to Z5, respectively. Then, the photovoltaic new energy system monitoring system uses the temperature gradient correlation vector, shadow trajectory correlation vector, and mismatch correlation vector obtained in step 121, the attenuation rate coupling vector and mismatch degree coupling vector obtained in step 122, and the hidden crack distribution correlation vector obtained in step 123 as the weights of the edges to calculate the element values ​​of the initial adjacency matrix. For example, for component identifiers Z1 and Z2, the element values ​​at the corresponding positions in the initial adjacency matrix are weighted combinations of the temperature gradient correlation degree values, shadow trajectory correlation degree values, mismatch correlation degree values, attenuation rate coupling degree values, mismatch degree coupling degree values, and hidden crack distribution correlation degree values ​​between Z1 and Z2. The weights of the weighted combinations are preset according to the degree of influence of each vector on the string topology.

[0029] Step 125: Perform feature dimension alignment on the initial adjacency matrix to ensure that the weight dimensions of different edges in the initial adjacency matrix are consistent.

[0030] In this embodiment, the weight dimensions of different edges in the initial adjacency matrix differ. For example, the dimension of the temperature gradient correlation vector is 1, the dimension of the shadow trajectory correlation vector is 2, the dimension of the mismatch correlation vector is 1, the dimension of the attenuation rate coupling vector is 2, the dimension of the mismatch degree coupling vector is 1, and the dimension of the hidden crack distribution correlation vector is 2. The photovoltaic new energy system monitoring system performs feature dimension alignment processing on the initial adjacency matrix. The processing method is as follows: for vectors with a dimension of 1, their dimension is expanded to 2 by copying vector elements; for vectors with a dimension of 2, their dimension remains unchanged; then, the expanded vectors are normalized by mapping the vector element values ​​to the interval between 0 and 1, so that the weight dimension of different edges in the initial adjacency matrix is ​​2 and the element value range is consistent.

[0031] Step 126: Perform topological relationship constraint verification based on the aligned initial adjacency matrix to determine the set of valid edges in the initial adjacency matrix that conform to the string connection logic; wherein, the topological relationship constraint verification includes distance constraint verification of node connections, feature correlation constraint verification, and topological structure constraint verification.

[0032] In this embodiment, the photovoltaic new energy system monitoring system performs topological constraint verification on the aligned initial adjacency matrix: For distance constraint verification of node connections, the system retrieves the component installation layout data of the target photovoltaic system, which includes the physical location coordinates of each component. It calculates the distance between the physical location coordinates of the components corresponding to each node in the initial adjacency matrix and determines whether this distance is less than a preset distance threshold. If it is less, the node connection meets the distance constraint. For feature correlation constraint verification, the system calculates the magnitude of the feature correlation vector of each node connection in the initial adjacency matrix and determines whether this magnitude is greater than a preset correlation threshold. If it is greater, the node connection meets the feature correlation constraint. For topological structure constraint verification, the system analyzes the topological structure of the initial adjacency matrix and determines whether the topological structure conforms to a tree structure or chain structure of string connections. If it does, the node connection meets the topological structure constraint. Then, the system marks node connections that simultaneously meet the distance constraint, feature correlation constraint, and topological structure constraint as valid edges, obtaining a set of valid edges.

[0033] Step 1261: Extract the feature association vector set of each node in the aligned initial adjacency matrix. The feature association vector set includes the temperature gradient association vector, shadow trajectory association vector, and mismatch association vector of each node and other nodes.

[0034] In this embodiment of the application, the nodes of the aligned initial adjacency matrix are component identifiers Z1 to Z5. The photovoltaic new energy system monitoring system extracts the feature association vector set of each node. For example, for node Z1, its feature association vector set includes the temperature gradient association vector, shadow trajectory association vector and mismatch association vector between Z1 and Z2, the temperature gradient association vector, shadow trajectory association vector and mismatch association vector between Z1 and Z3, the temperature gradient association vector, shadow trajectory association vector and mismatch association vector between Z1 and Z4, and the temperature gradient association vector, shadow trajectory association vector and mismatch association vector between Z1 and Z5.

[0035] Step 1262: For each node, analyze the correlation distribution of each vector in the feature correlation vector set of that node, and determine the core correlation node set of that node. The core correlation node set contains nodes that have a strong correlation with the feature correlation vector of that node.

[0036] In this embodiment, the photovoltaic new energy system monitoring system calculates the correlation strength of each vector in the feature correlation vector set of each node. The correlation strength is calculated by weighted sum of vector element values, and the weights of the weighted sum are preset according to the degree of influence of each vector on the correlation. Then, the photovoltaic new energy system monitoring system determines the nodes corresponding to the vectors with correlation strength greater than a preset strong correlation threshold as the core correlation nodes of that node, thus obtaining the core correlation node set. For example, for node Z1, the correlation strength with Z2 in its feature correlation vector set is greater than the strong correlation threshold, the correlation strength with Z3 is less than the strong correlation threshold, the correlation strength with Z4 is greater than the strong correlation threshold, and the correlation strength with Z5 is less than the strong correlation threshold. Therefore, the core correlation node set of node Z1 is {Z2, Z4}.

[0037] Step 1263: Construct a topological subgraph of the core associated node set. Based on the connectivity and hierarchical structure of the topological subgraph, determine the subgraph structure that conforms to the string connection logic.

[0038] In this embodiment, the photovoltaic new energy system monitoring system constructs a topological subgraph using nodes in the core associated node set as vertices and connecting edges between nodes in the core associated node set as edges. Then, the photovoltaic new energy system monitoring system analyzes the connectivity of the topological subgraph to determine whether it is a connected graph. If it is a connected graph, it further analyzes the hierarchical structure of the topological subgraph to determine whether the hierarchical structure conforms to the sequential structure of string connections. For example, if the hierarchical structure of the topological subgraph is Z1→Z2→Z4, then the hierarchical structure conforms to the string connection logic. Finally, the photovoltaic new energy system monitoring system determines the topological subgraph that meets the connectivity and hierarchical structure requirements as a subgraph structure that conforms to the string connection logic.

[0039] Step 1264: Mark the connecting edges of each node in the topological subgraph as candidate valid edges, and extract the unmarked edges in the initial adjacency matrix.

[0040] In this embodiment of the application, the photovoltaic new energy system monitoring system marks the connecting edges in the subgraph structure that conforms to the string connection logic as candidate valid edges. For example, if the structure of the topological subgraph is Z1→Z2→Z4, then the connecting edges Z1-Z2 and Z2-Z4 are marked as candidate valid edges. Then, the photovoltaic new energy system monitoring system extracts the unmarked edges in the initial adjacency matrix, such as the connecting edges Z1-Z3, Z1-Z5, Z3-Z4, Z3-Z5, and Z4-Z5.

[0041] Step 1265: If an unmarked edge is synchronous with the feature association vector of the topological subgraph, then mark the edge as a candidate valid edge.

[0042] In this embodiment of the application, the photovoltaic new energy system monitoring system calculates the synchronicity between the feature correlation vector of the unmarked edge and the feature correlation vector of the topological subgraph. The synchronicity is calculated by calculating the correlation coefficient of the two vectors. If the correlation coefficient is greater than a preset synchronicity threshold, then the unmarked edge and the feature correlation vector of the topological subgraph are synchronized. For example, if the correlation coefficient between the feature correlation vector of the unmarked edge Z3-Z4 and the feature correlation vector of the topological subgraph is greater than the synchronicity threshold, then edge Z3-Z4 is marked as a candidate valid edge.

[0043] Step 1266: Perform collaborative verification of topological constraints on all candidate valid edges, and take the set of candidate valid edges that pass the collaborative verification as the set of valid edges; wherein, the collaborative verification includes consistency verification of subgraph structure, matching verification of feature association vectors, and compliance verification of string connection logic.

[0044] In this embodiment, the photovoltaic new energy system monitoring system performs collaborative verification of topological constraints on all candidate valid edges: For the consistency verification of the subgraph structure, the photovoltaic new energy system monitoring system determines whether the candidate valid edge belongs to the same topological subgraph. If it does, the candidate valid edge meets the consistency requirement of the subgraph structure. For the matching verification of the feature association vector, the photovoltaic new energy system monitoring system calculates the matching degree between the feature association vector of the candidate valid edge and the core feature association vector of the topological subgraph. The matching degree is calculated by calculating the cosine similarity of the two vectors. If the cosine similarity is greater than the preset matching degree threshold, the candidate valid edge meets the matching requirement of the feature association vector. For the compliance verification of the string connection logic, the photovoltaic new energy system monitoring system determines whether the connection of the candidate valid edge conforms to the sequential structure of the string connection. If it does, the candidate valid edge meets the compliance requirement of the string connection logic. Then, the photovoltaic new energy system monitoring system determines the candidate valid edges that simultaneously meet the consistency requirement of the subgraph structure, the matching requirement of the feature association vector, and the compliance requirement of the string connection logic as valid edges, thus obtaining a set of valid edges.

[0045] Step 12661: Extract the temperature gradient correlation vector, shadow trajectory correlation vector, and mismatch correlation vector of the node pairs corresponding to the candidate valid edges.

[0046] In this embodiment of the application, the node pairs corresponding to the candidate valid edges are Z1-Z2, Z2-Z4, and Z3-Z4. The photovoltaic new energy system monitoring system extracts the temperature gradient correlation vector, shadow trajectory correlation vector, and mismatch correlation vector of these node pairs. For example, the temperature gradient correlation vector of node pair Z1-Z2 is [T12], the shadow trajectory correlation vector is [S12a, S12b], and the mismatch correlation vector is [M12]; the temperature gradient correlation vector of node pair Z2-Z4 is [T24], the shadow trajectory correlation vector is [S24a, S24b], and the mismatch correlation vector is [M24]; the temperature gradient correlation vector of node pair Z3-Z4 is [T34], the shadow trajectory correlation vector is [S34a, S34b], and the mismatch correlation vector is [M34].

[0047] Step 12662: Determine the first association mode based on the association mode between the temperature gradient association vector and the shadow trajectory association vector; determine the second association mode based on the association mode between the shadow trajectory association vector and the mismatch association vector.

[0048] In this embodiment, the photovoltaic new energy system monitoring system calculates the correlation coefficient between the temperature gradient correlation vector and the shadow trajectory correlation vector. Based on the sign and magnitude of the correlation coefficient, a first correlation mode is determined. For example, if the correlation coefficient between the temperature gradient correlation vector of node Z1-Z2 and the shadow trajectory correlation vector is positive and large, then the first correlation mode is a positive correlation mode; if the correlation coefficient between the temperature gradient correlation vector of node Z2-Z4 and the shadow trajectory correlation vector is negative and large, then the first correlation mode is a negative correlation mode; if the correlation coefficient between the temperature gradient correlation vector of node Z3-Z4 and the shadow trajectory correlation vector is close to zero, then the first correlation mode is an uncorrelated mode. Similarly, the photovoltaic new energy system monitoring system calculates the correlation coefficient between the shadow trajectory correlation vector and the mismatch correlation vector. Based on the sign and magnitude of the correlation coefficient, the second correlation mode is determined. For example, if the correlation coefficient between the shadow trajectory correlation vector of node Z1-Z2 and the mismatch correlation vector is positive and large, then the second correlation mode is a positive correlation mode; if the correlation coefficient between the shadow trajectory correlation vector of node Z2-Z4 and the mismatch correlation vector is negative and large, then the second correlation mode is a negative correlation mode; if the correlation coefficient between the shadow trajectory correlation vector of node Z3-Z4 and the mismatch correlation vector is close to zero, then the second correlation mode is an uncorrelated mode.

[0049] Step 12663: Determine whether the first association pattern and the second association pattern conform to the preset string mismatch association pattern.

[0050] In this embodiment, the preset string mismatch association mode is either a positive correlation mode for both the first and second association modes, or a negative correlation mode for both the first and second association modes. The photovoltaic new energy system monitoring system determines whether the first and second association modes of each node pair conform to the preset mode. For example, if the first and second association modes of node pair Z1-Z2 are both positively correlated, they conform to the preset mode; if the first and second association modes of node pair Z2-Z4 are both negatively correlated, they conform to the preset mode; if the first and second association modes of node pair Z3-Z4 are both uncorrelated, they do not conform to the preset mode.

[0051] Step 12664: If the conditions are met, determine whether the node pairs corresponding to the candidate valid edges belong to the same topological subgraph.

[0052] In this embodiment of the application, node pairs Z1-Z2 and Z2-Z4 conform to a preset string mismatch association pattern. The photovoltaic new energy system monitoring system determines whether these node pairs belong to the same topological subgraph. For example, node pairs Z1-Z2 and Z2-Z4 belong to the same topological subgraph, while node pairs Z3-Z4 do not belong to the same topological subgraph.

[0053] Step 12665: If they belong to the same topological subgraph, determine whether the structure of the topological subgraph meets the hierarchical structure requirements of the group connection.

[0054] In this embodiment of the application, node pairs Z1-Z2 and Z2-Z4 belong to the same topological subgraph. The photovoltaic new energy system monitoring system determines whether the structure of the topological subgraph meets the hierarchical structure requirements of string connection. For example, the structure of the topological subgraph is Z1→Z2→Z4, which meets the sequential hierarchical structure requirements of string connection.

[0055] Step 12666: If the hierarchical structure requirements are met, determine whether the feature association vector of the node pair corresponding to the candidate valid edge matches the core feature association vector of the topological subgraph.

[0056] In this embodiment, the core feature association vector of the topological subgraph is the weighted average of the feature association vectors of Z1-Z2 and Z2-Z4. The photovoltaic new energy system monitoring system calculates the cosine similarity between the feature association vectors of Z1-Z2 and Z2-Z4 and the core feature association vector of the node. If the cosine similarity is greater than a preset matching threshold, it is judged as a match. For example, if the cosine similarity between the feature association vector of Z1-Z2 and the core feature association vector of the node is 0.85, which is greater than the matching threshold of 0.8, it is judged as a match; if the cosine similarity between the feature association vector of Z2-Z4 and the core feature association vector of the node is 0.90, which is greater than the matching threshold of 0.8, it is judged as a match.

[0057] Step 12667: If a match is found, the candidate valid edge is determined to have passed collaborative verification; otherwise, the candidate valid edge is determined to have failed collaborative verification.

[0058] In this embodiment, the feature association vectors of node pairs Z1-Z2 and Z2-Z4 match the core feature association vectors of the topological subgraph, thus determining that these candidate valid edges pass the collaborative verification; node pairs Z3-Z4 do not conform to the preset string mismatch association pattern, thus determining that the candidate valid edge does not pass the collaborative verification.

[0059] Step 127: Update the initial adjacency matrix based on the set of valid edges, set the weight of invalid edges to zero, and obtain the string topology model.

[0060] In this embodiment of the application, the set of valid edges is {Z1-Z2, Z2-Z4}. The photovoltaic new energy system monitoring system updates the initial adjacency matrix based on this set of valid edges, sets the weights of invalid edges (such as Z1-Z3, Z1-Z5, Z3-Z5, Z4-Z5, etc.) in the initial adjacency matrix to zero, and obtains a string topology model. This model is a 5x5 matrix, where the element values ​​corresponding to valid edges are non-zero, and the element values ​​corresponding to invalid edges are zero.

[0061] Step 130: Process the current running data using the string topology relationship model to generate a dynamic mismatch feature set. Using the dynamic mismatch feature set as input, call the time series prediction model to output the mismatch trend evolution result of the target photovoltaic system.

[0062] In this embodiment, the photovoltaic new energy system monitoring system inputs the current operating data into the string topology model. The string topology model integrates and processes the current operating data according to the correlation between each component to obtain the associated operating data of each component. Then, the photovoltaic new energy system monitoring system extracts features from the associated operating data to generate a dynamic mismatch feature set. This dynamic mismatch feature set includes data on the degree of mismatch change, temperature change, and shading change of each component. Next, the photovoltaic new energy system monitoring system uses the dynamic mismatch feature set as input and calls the time series prediction model. The time series prediction model analyzes and processes the dynamic mismatch feature set and outputs the mismatch trend evolution result of the target photovoltaic system. This result includes the trend of the degree of mismatch change of each component and the expansion trend of the mismatch area within the next week.

[0063] Step 131: Input the current running data into the string topology model to obtain the string association feature vector corresponding to the current running data. The string association feature vector represents the association relationship between different components in the same string and the association relationship with adjacent strings.

[0064] In this embodiment of the application, the photovoltaic new energy system monitoring system inputs the current operating data into the string topology relationship model. The string topology relationship model performs weighted integration on the current operating data according to the association relationship of each component to obtain the string association feature vector. This vector contains the association relationship values ​​between different components in the same string and the association relationship values ​​with adjacent strings. For example, the string association feature vector is [R12, R24, R14, R23, R34], where R12 represents the association relationship value between components Z1 and Z2, R24 represents the association relationship value between components Z2 and Z4, R14 represents the association relationship value between components Z1 and Z4, R23 represents the association relationship value between components Z2 and Z3, and R34 represents the association relationship value between components Z3 and Z4.

[0065] Step 132: Extract the mismatch change vector, temperature change vector, and shadow change vector from the string association feature vector. The mismatch change vector represents the degree of mismatch change of different components in the same string, the temperature change vector represents the temperature change of different components in the same string, and the shadow change vector represents the shadow change of different components in the same string.

[0066] In this embodiment, the photovoltaic new energy system monitoring system analyzes the string-related feature vectors and extracts the mismatch change vector, temperature change vector, and shading change vector: For the mismatch change vector, the difference between the current mismatch degree data and the historical mismatch degree data of different components in the same string is calculated to obtain the mismatch change vector, and each element of the vector corresponds to the mismatch degree change value of different components in the same string; For the temperature change vector, the difference between the current temperature data and the historical temperature data of different components in the same string is calculated to obtain the temperature change vector, and each element of the vector corresponds to the temperature change value of different components in the same string; For the shading change vector, the difference between the current shading coverage data and the historical shading coverage data of different components in the same string is calculated to obtain the shading change vector, and each element of the vector corresponds to the shading change value of different components in the same string.

[0067] Step 133: After standardizing the mismatch change vector, temperature change vector, and shadow change vector, feature concatenation is performed to obtain the initial dynamic mismatch feature vector.

[0068] In this embodiment, the photovoltaic new energy system monitoring system standardizes the mismatch change vector, temperature change vector, and shadow change vector by mapping the vector element values ​​to the range of 0 to 1. Then, the photovoltaic new energy system monitoring system performs feature concatenation on the standardized mismatch change vector, temperature change vector, and shadow change vector by connecting the three vectors in sequence into a longer vector to obtain the initial dynamic mismatch feature vector. For example, if the mismatch change vector is [ΔM1, ΔM2, ΔM3], the temperature change vector is [ΔT1, ΔT2, ΔT3], and the shadow change vector is [ΔS1, ΔS2, ΔS3], then the initial dynamic mismatch feature vector is [ΔM1, ΔM2, ΔM3, ΔT1, ΔT2, ΔT3, ΔS1, ΔS2, ΔS3].

[0069] Step 134: Using the initial dynamic mismatch feature vector as input, call the feature enhancement layer of the time series prediction model, and perform feature enhancement processing by combining the historical dynamic mismatch feature vectors within the preset time window to obtain the enhanced dynamic mismatch feature vector.

[0070] In this embodiment, the preset time window is the past 24 hours. The photovoltaic new energy system monitoring system retrieves the historical dynamic mismatch feature vector set of the past 24 hours. Then, the photovoltaic new energy system monitoring system takes the initial dynamic mismatch feature vector as input and calls the feature enhancement layer of the time series prediction model. The feature enhancement layer combines the historical dynamic mismatch feature vector set to enhance the initial dynamic mismatch feature vector to obtain the enhanced dynamic mismatch feature vector.

[0071] Step 1341: Extract the set of historical dynamic mismatch feature vectors within the preset time window. The set of historical dynamic mismatch feature vectors contains dynamic mismatch feature vectors at consecutive time points.

[0072] In this embodiment of the application, the preset time window is the past 24 hours. The photovoltaic new energy system monitoring system retrieves the dynamic mismatch feature vectors every hour in the past 24 hours to obtain a set of historical dynamic mismatch feature vectors. This set contains 24 dynamic mismatch feature vectors, and the dimension of each vector is consistent with the dimension of the initial dynamic mismatch feature vector.

[0073] Step 1342: Analyze the feature change patterns of the historical dynamic mismatch feature vector set. The feature change patterns include gradual change patterns or abrupt change patterns.

[0074] In this embodiment of the application, the photovoltaic new energy system monitoring system analyzes the historical dynamic mismatch feature vector set and analyzes the changing trend of each vector element value over time: if the change of the vector element value over time is relatively gradual and the change amplitude is less than the preset gradual threshold, then the feature change mode is determined to be a gradual mode; if the change of the vector element value over time is relatively drastic and the change amplitude is greater than the preset sudden threshold, then the feature change mode is determined to be a sudden mode; for example, if the element value of the mismatch degree change vector in the historical dynamic mismatch feature vector set gradually increases over time and the change amplitude is less than the gradual threshold, then the feature change mode is a gradual mode; if the element value of the mismatch degree change vector suddenly increases at a moment and the change amplitude is greater than the sudden threshold, then the feature change mode is a sudden mode.

[0075] Step 13421: Extract the feature vector differences between adjacent time points in the historical dynamic mismatch feature vector set. The feature vector differences include differences in temperature changes, differences in shading changes, and differences in mismatch changes.

[0076] In this embodiment of the application, the photovoltaic new energy system monitoring system calculates the difference in feature vectors between adjacent time points in the historical dynamic mismatch feature vector set. For example, for time points t1 and t2, the difference in feature vectors is the feature vector at time t2 minus the feature vector at time t1, resulting in the difference in temperature change, the difference in shadow change, and the difference in mismatch change.

[0077] Step 13422: Determine the temperature change pattern based on the changing trend of the temperature change difference; determine the shading change pattern based on the changing trend of the shading change difference; determine the mismatch change pattern based on the changing trend of the mismatch change difference.

[0078] In this embodiment, the photovoltaic new energy system monitoring system analyzes the trend of temperature variation differences over time: if the temperature variation difference gradually increases over time, the temperature variation pattern is determined to be an increasing pattern; if the temperature variation difference gradually decreases over time, the temperature variation pattern is determined to be a decreasing pattern; if the temperature variation difference remains stable over time, the temperature variation pattern is determined to be a stable pattern. Similarly, the photovoltaic new energy system monitoring system analyzes the trends of shading variation differences and mismatch variation differences over time to determine the shading variation pattern and mismatch variation pattern.

[0079] Step 13423: Determine whether the temperature change pattern, shadow change pattern, and mismatch change pattern are synchronous.

[0080] In this embodiment of the application, the photovoltaic new energy system monitoring system calculates the synchronization coefficients of temperature change mode, shadow change mode and mismatch change mode. The synchronization coefficient is calculated as the average of the correlation coefficients of the change trends of the three modes. If the synchronization coefficient is greater than the preset synchronization threshold, it is determined that the three modes are synchronized; otherwise, it is determined that the three modes are not synchronized.

[0081] Step 13424: If there is synchronization, then determine the feature change pattern of the historical dynamic mismatch feature vector set as a cooperative change pattern.

[0082] In this embodiment of the application, if the synchronization coefficients of the temperature change pattern, the shadow change pattern, and the mismatch change pattern are greater than the synchronization threshold, then the characteristic change pattern is determined to be a cooperative change pattern. For example, if the temperature change pattern is an upward pattern, the shadow change pattern is an upward pattern, the mismatch change pattern is an upward pattern, and the synchronization coefficients are greater than the synchronization threshold, then the characteristic change pattern is a cooperative upward pattern.

[0083] Step 13425: If there is no synchronicity, analyze the dominance of each mode and determine the characteristic change mode of the historical dynamic mismatch feature vector set as a non-cooperative change mode.

[0084] In this embodiment, if the synchronicity coefficients of the temperature change mode, the shadow change mode, and the mismatch change mode are less than the synchronicity threshold, the dominance of each mode is analyzed. The dominance analysis method is to calculate the proportion of the change amplitude of each mode to the total change amplitude, and the mode with the largest proportion is the dominant mode. Then, the characteristic change mode is determined to be a non-cooperative change mode. For example, if the temperature change mode is an increasing mode, the shadow change mode is a stable mode, and the mismatch change mode is a decreasing mode, and the change amplitude of the temperature change mode accounts for the largest proportion of the total change amplitude, then the characteristic change mode is a temperature-dominated non-cooperative change mode.

[0085] Step 13426: Based on the cooperative change mode or the non-cooperative change mode, determine the effective feature subset of the historical dynamic mismatch feature vector set.

[0086] In this embodiment of the application, if the feature change mode is a cooperative change mode, the photovoltaic new energy system monitoring system extracts a subset of historical dynamic mismatch feature vectors that are consistent with the cooperative change mode; if the feature change mode is a non-cooperative change mode, the system extracts a subset of historical dynamic mismatch feature vectors that are consistent with the dominant mode to obtain an effective feature subset.

[0087] Step 1343: For the initial dynamic mismatch feature vector, determine the matching degree with the feature change pattern of the historical dynamic mismatch feature vector set based on its feature change pattern.

[0088] In this embodiment of the application, the photovoltaic new energy system monitoring system analyzes the characteristic change pattern of the initial dynamic mismatch feature vector. For example, if the element value of the mismatch degree change vector of the initial dynamic mismatch feature vector gradually increases over time and the change amplitude is less than the gradual change threshold, then the characteristic change pattern is a gradual change pattern. Then, the photovoltaic new energy system monitoring system calculates the matching degree between the characteristic change pattern of the initial dynamic mismatch feature vector and the characteristic change pattern of the historical dynamic mismatch feature vector set. The matching degree is calculated by calculating the correlation coefficient between the two patterns. The larger the correlation coefficient, the higher the matching degree.

[0089] Step 1344: If the matching degree meets the requirements, extract the subset of the historical dynamic mismatch feature vector set that matches the pattern of the initial dynamic mismatch feature vector; perform cross-dimensional correlation analysis on the subset and the initial dynamic mismatch feature vector, calculate the correlation between the temperature change dimension and the mismatch change dimension, the correlation between the shadow change dimension and the mismatch change dimension, and perform feature weighted combination based on the correlation weight to obtain the enhanced dynamic mismatch feature vector.

[0090] In this embodiment, if the matching degree is greater than a preset matching degree threshold, the photovoltaic new energy system monitoring system extracts a subset from the historical dynamic mismatch feature vector set that matches the pattern of the initial dynamic mismatch feature vector. Then, the photovoltaic new energy system monitoring system performs cross-dimensional correlation analysis on the subset and the initial dynamic mismatch feature vector, calculating the correlation between the temperature change dimension and the mismatch change dimension, and the correlation between the shadow change dimension and the mismatch change dimension. The correlation is calculated using the Pearson correlation coefficient. Next, the photovoltaic new energy system monitoring system determines the weight based on the magnitude of the correlation; the greater the correlation, the greater the weight. Finally, the photovoltaic new energy system monitoring system performs a weighted combination of the initial dynamic mismatch feature vector and the vectors in the subset based on the weights to obtain the enhanced dynamic mismatch feature vector.

[0091] Step 1345: If the matching degree does not meet the requirements, extract the subset of the historical dynamic mismatch feature vector set that differs from the initial dynamic mismatch feature vector pattern; perform feature difference compensation fusion on the subset of differences and the initial dynamic mismatch feature vector to obtain the enhanced dynamic mismatch feature vector.

[0092] In this embodiment, if the matching degree is less than a preset matching degree threshold, the photovoltaic new energy system monitoring system extracts a subset of the historical dynamic mismatch feature vector set that differs from the initial dynamic mismatch feature vector pattern. Then, the photovoltaic new energy system monitoring system performs feature difference compensation fusion on the difference subset and the initial dynamic mismatch feature vector. The fusion method is to take a weighted average of the vector element values ​​in the difference subset and the element values ​​of the initial dynamic mismatch feature vector. The weight is determined according to the degree of difference; the greater the degree of difference, the smaller the weight. Finally, an enhanced dynamic mismatch feature vector is obtained.

[0093] Step 135: Input the enhanced dynamic mismatch feature vector into the trend prediction layer of the time series prediction model, and output the mismatch trend evolution result. The mismatch trend evolution result includes the changing trend information of the degree of mismatch and the expansion trend information of the mismatch region.

[0094] In this embodiment of the application, the photovoltaic new energy system monitoring system will input the enhanced dynamic mismatch feature vector into the trend prediction layer of the time series prediction model. The trend prediction layer will analyze and process the enhanced dynamic mismatch feature vector, predict the trend of mismatch degree change of each component and the expansion trend of mismatch area in the next week, and output the mismatch trend evolution result. The result is a multi-dimensional vector, which includes multiple values ​​of the mismatch degree change trend dimension and the mismatch area expansion trend dimension.

[0095] Step 140: Generate string-level control strategy based on the mismatch trend evolution result. During the execution of the string-level control strategy, optimize the control by combining the power distribution logic of the target inverter of the target photovoltaic system and output the optimized control signal.

[0096] In this embodiment, the photovoltaic new energy system monitoring system determines the strings that need to be adjusted and the adjustment method based on the mismatch trend evolution results, and generates a string-level control strategy. Then, during the execution of the string-level control strategy, the photovoltaic new energy system monitoring system, in conjunction with the power allocation logic of the target inverter, optimizes and adjusts the string-level control strategy and outputs an optimized control signal, which includes the power adjustment instructions of each component and the power allocation instructions of the inverter.

[0097] Step 141: Analyze the trend of mismatch expansion and the trend of mismatch degree change in the mismatch trend evolution results.

[0098] In this embodiment of the application, the photovoltaic new energy system monitoring system analyzes the mismatch trend evolution results and extracts the mismatch region expansion trend and the mismatch degree change trend: the mismatch region expansion trend includes the trend that the mismatch region will expand from module Z1 to modules Z2 and Z3 within the next week; the mismatch degree change trend includes the trend that the mismatch degree of module Z1 will increase from the current 10% to 15% within the next week, the mismatch degree of module Z2 will increase from the current 5% to 8%, and the mismatch degree of module Z3 will increase from the current 3% to 5%.

[0099] Step 142: Extract the current power output status and component association of the string corresponding to the mismatch region.

[0100] In this embodiment, the string corresponding to the mismatch region is string 1, which includes components Z1, Z2, and Z3. The photovoltaic new energy system monitoring system extracts the current power output status of string 1. The current power output status is that the real-time output power of string 1 is 5kW, the real-time output power of component Z1 is 1.5kW, the real-time output power of component Z2 is 1.8kW, and the real-time output power of component Z3 is 1.7kW. At the same time, the photovoltaic new energy system monitoring system extracts the component association relationship of string 1. The component association relationship is that components Z1 and Z2 are connected in series, and components Z2 and Z3 are connected in series.

[0101] Step 143: Utilize the matching relationship between the mismatch region expansion trend and the component association to determine the set of strings and components that need to be adjusted.

[0102] In this embodiment of the application, the photovoltaic new energy system monitoring system analyzes the matching between the mismatch region expansion trend and the component association relationship. For example, if the mismatch region expansion trend is from component Z1 to Z2 and Z3, and the component association relationship is Z1-Z2-Z3 in series, then the set of strings that need to be adjusted is determined to be {string 1}, and the set of components that need to be adjusted is {Z1, Z2, Z3}.

[0103] Step 144: For the set of strings that need to be adjusted, determine the coordination requirements for power adjustment based on the power correlation between the set of strings and adjacent strings.

[0104] In this embodiment, the set of strings to be adjusted is {string 1}, and the adjacent strings are string 2, which includes components Z4 and Z5. The photovoltaic new energy system monitoring system analyzes the power correlation between string 1 and string 2. The power correlation is that string 1 and string 2 are connected in parallel to the same inverter. Then, the photovoltaic new energy system monitoring system determines the coordination requirements for power adjustment. The coordination requirements are that the power adjustment of string 1 should not affect the normal power output of string 2, and the total output power of string 1 and string 2 should remain stable.

[0105] Step 145: Based on the trend of mismatch and coordination requirements, determine the power adjustment direction and method for each string.

[0106] In this embodiment, the photovoltaic new energy system monitoring system determines the power adjustment direction of string 1 to reduce the output power of component Z1 and increase the output power of components Z2 and Z3 based on the trend of mismatch degree changes. At the same time, according to the coordination requirements, the power adjustment method is determined to adjust the output power by adjusting the working voltage of the components. The adjustment range needs to be controlled within the range that does not affect the normal power output of string 2. For example, the working voltage of component Z1 is reduced from the current 30V to 28V, the working voltage of component Z2 is increased from the current 30V to 32V, and the working voltage of component Z3 is increased from the current 30V to 31V.

[0107] Step 146: Construct the initial group-string level control strategy, which includes group-string adjustment instructions, component adjustment instructions, and collaborative adjustment instructions.

[0108] In this embodiment, the photovoltaic new energy system monitoring system constructs an initial string-level control strategy based on the determined power adjustment direction and adjustment method: the string adjustment instruction is to reduce the output power of string 1 by 5%; the module adjustment instruction is to reduce the operating voltage of module Z1 to 28V, increase the operating voltage of module Z2 to 32V, and increase the operating voltage of module Z3 to 31V; the coordinated adjustment instruction is that during the power adjustment of string 1, the output power of string 2 should be maintained at approximately 6kW.

[0109] Step 147: Determine whether the initial string-level control strategy is compatible with the power conversion logic of the target photovoltaic system; if compatible, determine the initial string-level control strategy as a string-level control strategy; if incompatible, adjust the power adjustment direction or adjustment method, and reconstruct the initial string-level control strategy until it meets the compatibility requirements.

[0110] In this embodiment, the power conversion logic of the target photovoltaic system is such that the inverter's input voltage range is 200V to 300V, and the current input voltage of string 1 is 270V (the operating voltage of components Z1, Z2, and Z3 is 30V each, the voltage after series connection is 90V, and the voltage after three strings are connected in series is 270V). The photovoltaic new energy system monitoring system determines whether the initial string-level control strategy is compatible with the power conversion logic: the operating voltage of component Z1 decreases to 28V, the operating voltage of component Z2 increases to 32V, and the operating voltage of component Z3 increases to 31V. The input voltage of string 1 is 28+32+31=91V, and the voltage after three strings are connected in series is 273V, which is within the inverter's input voltage range. Therefore, the initial string-level control strategy is compatible with the power conversion logic, and the initial string-level control strategy is determined to be a string-level control strategy.

[0111] Step 148: Extract the string power adjustment feature set from the string-level control strategy. The string power adjustment feature set includes the adjustment correlation and adjustment constraint requirements of each string to be adjusted.

[0112] In this embodiment of the application, the string power adjustment feature set in the string-level control strategy includes the adjustment correlation relationship and adjustment constraint requirements of string 1: the adjustment correlation relationship is that the power adjustment of string 1 is related to the power output of string 2; the adjustment constraint requirements are that the input voltage of string 1 needs to be kept in the range of 200V to 300V, and the output power adjustment amplitude of string 1 needs to be less than 20% of the current output power.

[0113] Step 149: Obtain the power allocation logic of the target inverter, and extract the inverter power allocation constraint set from the power allocation logic. The inverter power allocation constraint set contains the priority rules and allocation range restrictions for power allocation.

[0114] In this embodiment, the power allocation logic of the target inverter prioritizes the power supply of important loads. The power requirement of important loads is 10kW, and the maximum output power of the inverter is 15kW. Therefore, the inverter power allocation constraint set includes the priority rule of power allocation, which prioritizes important loads. The allocation range is limited to the power supply of important loads being greater than or equal to 10kW, and the power supply of non-important loads being less than or equal to 5kW.

[0115] Step 1410: Generate a string-inverter association feature matrix with string identifiers as rows, inverter power allocation constraint identifiers as columns, and matrix elements representing the association strength features between the corresponding string and the power allocation constraint.

[0116] In this embodiment, string identifiers are string 1 and string 2, and inverter power allocation constraint identifiers are: priority for important loads, power supply of important loads ≥10kW, and power supply of non-important loads ≤5kW. The photovoltaic new energy system monitoring system generates a 3x2 string-inverter association feature matrix, where the matrix elements represent the association strength characteristics of the corresponding string and the power allocation constraint. For example, the association strength characteristic of string 1 with priority for important loads is 0.8, the association strength characteristic of string 1 with power supply of important loads ≥10kW is 0.9, and the association strength characteristic of string 1 with power supply of non-important loads ≤5kW is 0.7; the association strength characteristic of string 2 with priority for important loads is 0.7, the association strength characteristic of string 2 with power supply of important loads ≥10kW is 0.8, and the association strength characteristic of string 2 with power supply of non-important loads ≤5kW is 0.9.

[0117] Step 1411: Based on the string-inverter correlation feature matrix, identify the set of conflict points between the string power adjustment feature set and the inverter power allocation constraint set. The set of conflict points includes string adjustment terms that do not meet the power allocation constraints.

[0118] In this embodiment, the photovoltaic new energy system monitoring system analyzes the string-inverter correlation feature matrix to identify conflict points between the string power adjustment feature set and the inverter power allocation constraint set. For example, the power adjustment range of string 1 is 5%, the current output power is 5kW, and the adjusted output power is 4.75kW; the current output power of string 2 is 6kW, and the adjusted output power needs to be maintained at around 6kW; the power requirement of the important load is 10kW, and the total output power of string 1 and string 2 is 4.75 + 6 = 10.75kW, which is greater than the power requirement of the important load of 10kW. Therefore, there is no conflict point, and the conflict point set is empty.

[0119] Step 1412: Based on the real-time power status of the strings in the current running data and the degree of influence of the strings on the overall power output of the system, determine the priority order of each conflict item in the conflict point set.

[0120] In this embodiment, the set of conflict points is empty, so there is no need to sort them by priority. If the set of conflict points is not empty, then sorting is performed according to step 1412.

[0121] Step 1413: Based on priority sorting, perform constraint adaptation adjustment on the string power adjustment feature set to obtain the adapted string power adjustment feature set.

[0122] In this embodiment, the conflict point set is empty, therefore the adapted string power adjustment feature set is consistent with the string power adjustment feature set. If the conflict point set is not empty, step 1413 is executed according to priority.

[0123] Step 1414: Perform collaborative fusion processing between the adapted string power adjustment feature set and the inverter power allocation logic to obtain the fused power control feature set.

[0124] In this embodiment of the application, the photovoltaic new energy system monitoring system performs collaborative fusion processing on the adapted string power adjustment feature set and the inverter power allocation logic. The fusion method is to integrate the adjustment correlation and adjustment constraint requirements in the string power adjustment feature set with the priority rules and allocation range restrictions in the inverter power allocation logic to obtain the fused power control feature set. This set includes a power adjustment range of 5% for string 1, an output power of approximately 6kW for string 2, a power supply of 10kW for important loads, and a power supply of 0.75kW for non-important loads.

[0125] Step 1415: Generate a preliminary optimized control signal set based on the fused power control feature set.

[0126] In this embodiment, the photovoltaic new energy system monitoring system generates a preliminary optimized control signal set based on the fused power control feature set. This set includes power adjustment instructions for string 1, power hold instructions for string 2, and power allocation instructions for the inverter: the power adjustment instructions for string 1 are to reduce the operating voltage of component Z1 to 28V, increase the operating voltage of component Z2 to 32V, and increase the operating voltage of component Z3 to 31V; the power hold instructions for string 2 are to keep the operating voltages of components Z4 and Z5 unchanged; and the power allocation instructions for the inverter are to allocate 10kW of power to important loads and 0.75kW of power to non-important loads.

[0127] Step 1416: Determine whether the preliminary optimized control signal set matches the power stability constraints of the target photovoltaic system; if it matches, output the optimized control signal; if it does not match, return to the constraint adaptation adjustment step until an optimized control signal that meets the matching requirements is output.

[0128] In this embodiment, the power stability constraint of the target photovoltaic system is that the output power fluctuation of the system must be less than 5% of the current output power. The current output power of the system is 11kW (the output power of string 1 is 5kW and the output power of string 2 is 6kW). After the execution of the preliminary optimization control signal set, the output power of the system is 10.75kW, and the fluctuation range is (11-10.75) / 11≈2.27%, which is less than 5%. Therefore, the preliminary optimization control signal set matches the power stability constraint, and the optimized control signal is output.

[0129] In an extended embodiment, the method further includes: Step 210: Based on the array operation data of multiple arrays of the target photovoltaic system, mine the inter-array mismatch correlation characteristics of each array. The inter-array mismatch correlation characteristics characterize the mutual influence relationship of the degree of mismatch between different arrays.

[0130] In this embodiment, the target photovoltaic system includes three arrays: A, B, and C. The photovoltaic new energy system monitoring system acquires the array operation data of these three arrays, which includes the real-time output power data, mismatch degree data, and temperature data of each array. Then, the photovoltaic new energy system monitoring system analyzes the array operation data to mine the mismatch correlation characteristics between each array. For example, by calculating the correlation coefficient between the mismatch degree data of array A and the mismatch degree data of array B, the mutual influence relationship of the mismatch degree changes between array A and array B is obtained. Similarly, the correlation coefficients between array A and array C, and between array B and array C are calculated to obtain the mismatch correlation characteristics between arrays. This characteristic is a multi-dimensional vector that includes the numerical values ​​of the mutual influence relationship of the mismatch degree changes between arrays A and B, A and C, and B and C.

[0131] Step 220: Construct an inter-array correlation matrix based on the inter-array mismatch correlation features. The rows and columns of the inter-array correlation matrix are array identifiers, and the matrix elements are the degree of mismatch correlation between the corresponding two arrays.

[0132] In this embodiment of the application, the array identifiers are A, B, and C. The photovoltaic new energy system monitoring system uses these array identifiers as rows and columns to construct a 3-row, 3-column inter-array correlation matrix. The matrix elements are the mismatch correlation degree values ​​between corresponding two arrays. For example, the mismatch correlation degree value between array A and B is 0.8, the mismatch correlation degree value between array A and C is 0.6, and the mismatch correlation degree value between array B and C is 0.7.

[0133] Step 230: Use the inter-array correlation matrix to process the array operation data of multiple arrays and generate an inter-array dynamic mismatch feature set. The inter-array dynamic mismatch feature set characterizes the dynamic changes and mutual influences of the degree of mismatch between different arrays.

[0134] In this embodiment of the application, the photovoltaic new energy system monitoring system inputs the array operation data of multiple arrays into the inter-array correlation matrix. The inter-array correlation matrix integrates and processes the array operation data according to the correlation relationship of each array to obtain inter-array correlation operation data. Then, the photovoltaic new energy system monitoring system extracts features from the inter-array correlation operation data to generate an inter-array dynamic mismatch feature set. This set contains the dynamic change values ​​of the degree of mismatch between different arrays and the values ​​of mutual influence.

[0135] Step 240: Using the inter-array dynamic mismatch feature set as input, call the cross-array trend prediction model to output the inter-array mismatch trend evolution results of the target photovoltaic system. The inter-array mismatch trend evolution results include the changing trends of the degree of mismatch between different arrays and their mutual influence trends.

[0136] In this embodiment of the application, the photovoltaic new energy system monitoring system takes the dynamic mismatch feature set between arrays as input, calls the cross-array trend prediction model, and analyzes and processes the dynamic mismatch feature set between arrays to output the mismatch trend evolution result between arrays of the target photovoltaic system. The result includes the trend that the mismatch degree of array A will increase from the current 8% to 12%, the mismatch degree of array B will increase from the current 6% to 9%, and the mismatch degree of array C will increase from the current 4% to 6% within the next week. It also shows that the influence of the change in the mismatch degree of array A on array B will increase from the current 0.8 to 0.9, and the influence on array C will increase from the current 0.6 to 0.7.

[0137] Step 250: Generate a cross-array control strategy based on the inter-array mismatch trend evolution results. The cross-array control strategy includes power adjustment coordination rules between different arrays.

[0138] In this embodiment of the application, the photovoltaic new energy system monitoring system determines the power adjustment coordination rules between different arrays based on the mismatch trend evolution results between arrays. For example, the power adjustment range of array A must be less than 15% of the current output power, the power adjustment range of array B must be less than 10% of the current output power, and the power adjustment range of array C must be less than 5% of the current output power. Moreover, the power adjustment of array A must be coordinated with the power adjustment of arrays B and C to ensure the stability of the total output power of the system.

[0139] Step 260: During the execution of the cross-array control strategy, the power allocation logic of multiple target inverters of the target photovoltaic system is jointly optimized, and the optimized control signal is output. The optimized control signal contains power allocation coordination instructions between different inverters.

[0140] In this embodiment, the target photovoltaic system includes two target inverters: inverter 1 is connected to arrays A and B, and inverter 2 is connected to array C. During the execution of the cross-array control strategy, the photovoltaic new energy system monitoring system combines the power allocation logic of the two target inverters to perform collaborative optimization of the cross-array control strategy. For example, the power allocation logic of inverter 1 prioritizes the power supply to array A, while the power allocation logic of inverter 2 prioritizes the power supply to array C. Based on the inter-array mismatch trend evolution results, the photovoltaic new energy system monitoring system adjusts the power allocation logic of inverter 1 to ensure the power supply to array A while also considering the power supply to array B; it also adjusts the power allocation logic of inverter 2 to ensure the power supply to array C while coordinating with the power allocation of inverter 1. Finally, it outputs a collaboratively optimized control signal, which includes the power allocation instruction of inverter 1 to allocate 60% of the power to array A and 40% to array B, the power allocation instruction of inverter 2 to allocate 100% of the power to array C, and power adjustment instructions for arrays A, B, and C.

[0141] In an extended embodiment, the method further includes: Step 310: Obtain historical operating data of the target photovoltaic system for multiple cycles, and mine the historical degradation characteristics within each cycle and the historical degradation characteristics between cycles. The historical degradation characteristics within a cycle represent the coupling relationship between the module degradation rate curve and the degree of mismatch within a single cycle, while the historical degradation characteristics between cycles represent the changes in the coupling relationship between the module degradation rate curve and the degree of mismatch between different cycles.

[0142] In this embodiment of the application, the cycle of the target photovoltaic system is one month. The photovoltaic new energy system monitoring system acquires historical operating data from the past six months and mines the historical degradation characteristics within each month and the historical degradation characteristics during the week: for the historical degradation characteristics within the cycle, the coupling relationship between the component attenuation rate curve and the degree of mismatch within each month is analyzed to obtain the historical degradation characteristics within the cycle; for the historical degradation characteristics during the week, the changes in the coupling relationship between the component attenuation rate curve and the degree of mismatch between different months are analyzed to obtain the historical degradation characteristics during the week.

[0143] Step 320: Construct a degradation trend verification model based on historical degradation characteristics within the cycle and historical degradation characteristics during the cycle; use the degradation trend verification model to determine the state labels of historical degradation characteristics.

[0144] In this embodiment, the photovoltaic new energy system monitoring system uses historical degradation features within a cycle and historical degradation features during a cycle as input to construct a degradation trend verification model. This model is a machine learning model, which includes an input layer, a hidden layer, and an output layer. The dimension of the input layer is the total dimension of the historical degradation features within a cycle and historical degradation features during a cycle. The hidden layer contains two fully connected layers, and the output layer has a dimension of 2, corresponding to valid labels and invalid labels, respectively. Then, the photovoltaic new energy system monitoring system uses the degradation trend verification model to verify the historical degradation features obtained in step 110 and determine their status labels. For example, if the matching degree between the historical degradation features and the historical degradation features within a cycle and historical degradation features during a cycle is greater than a preset verification threshold, then the status label is a valid label; otherwise, it is an invalid label.

[0145] Step 330: If the state label is a valid label, then based on the historical degradation characteristics, perform confidence calibration on the mismatch trend evolution result to obtain the calibrated mismatch trend evolution result. The confidence calibration includes adjusting the confidence weight of the trend evolution result.

[0146] In this embodiment, the status label of the historical degradation feature is a valid label. The photovoltaic new energy system monitoring system uses the historical degradation feature as a basis to calibrate the confidence of the mismatch trend evolution result obtained in step 130. For example, if the coupling degree between the attenuation rate curve and the mismatch degree of component Z1 in the historical degradation feature is 0.8, which is greater than the preset coupling threshold of 0.7, then the confidence weight of the mismatch degree change trend of component Z1 is increased from the original 0.8 to 0.9; if the coupling degree between the attenuation rate curve and the mismatch degree of component Z2 in the historical degradation feature is 0.6, which is less than the preset coupling threshold of 0.7, then the confidence weight of the mismatch degree change trend of component Z2 is decreased from the original 0.7 to 0.6; finally, the calibrated mismatch trend evolution result is obtained.

[0147] Step 340: Update the string-level control strategy based on the calibrated mismatch trend evolution results to obtain the updated string-level control strategy.

[0148] In this embodiment, the photovoltaic new energy system monitoring system adjusts the power adjustment magnitude and adjustment method in the string-level control strategy based on the calibrated mismatch trend evolution results. For example, the confidence weight of the mismatch degree change trend of component Z1 is increased, so the power adjustment magnitude of component Z1 is increased from the original decrease of 5% to decrease of 8%; the confidence weight of the mismatch degree change trend of component Z2 is decreased, so the power adjustment magnitude of component Z2 is decreased from the original increase of 5% to increase of 3%; finally, the updated string-level control strategy is obtained.

[0149] Step 350: During the execution of the updated string-level control strategy, a secondary optimization is performed in conjunction with the power distribution logic of the target inverter, and the secondary optimized control signal is output. The secondary optimized control signal includes the updated power adjustment command and power distribution command.

[0150] In this embodiment, during the execution of the updated string-level control strategy, the photovoltaic new energy system monitoring system, in conjunction with the power allocation logic of the target inverter, performs secondary optimization of the updated string-level control strategy. For example, if the power adjustment range of component Z1 increases, resulting in a decrease in the input voltage of string 1, the photovoltaic new energy system monitoring system adjusts the power allocation logic of the inverter to adapt to the change in the input voltage of string 1. Finally, it outputs the secondary optimized control signal, which includes the updated power adjustment command for component Z1 to reduce the operating voltage to 26V, the power adjustment command for component Z2 to increase the operating voltage to 31V, the power adjustment command for component Z3 to increase the operating voltage to 30V, and the inverter's power allocation command to allocate 5.5kW of power to string 1 and 6kW of power to string 2.

[0151] This application's embodiments achieve end-to-end innovation in photovoltaic system mismatch monitoring and control through a collaborative logic of multi-dimensional feature mining, adjacency matrix modeling, time-series prediction, and control optimization. First, it simultaneously mines spatiotemporal distribution features, historical degradation features, and microscopic parameter features, overcoming the limitations of traditional single-feature monitoring and constructing a multi-scale feature system from macroscopic array operation to microscopic component defects, providing data support for mismatch analysis. Second, adjacency matrix modeling based on multi-dimensional features transforms component relationships into a quantifiable topological model, accurately reconstructing the actual connection logic and influence relationships within strings and between components, solving the problem of traditional topological modeling neglecting multi-feature coupling. Third, it uses the string topological relationship model to generate a dynamic mismatch feature set and combines it with a time-series prediction model to output mismatch trend evolution results, achieving accurate prediction of mismatch trends and overcoming the deficiency of traditional static analysis in not being able to provide early warnings. Finally, it generates string-level control strategies based on the trend results and optimizes them in conjunction with the inverter power allocation logic, achieving deep adaptation between the control strategy and hardware operating logic, improving the stability and efficiency of system power output.

[0152] The embodiments of this application form a logic of feature mining, topology modeling, trend prediction, and control optimization, which improves the accuracy of photovoltaic system mismatch monitoring, the foresight of trend prediction, and the effectiveness of control strategies, enabling the photovoltaic system to operate efficiently and stably.

[0153] Please see Figure 2The figure is a schematic diagram of the basic structure of a computer device 200 provided in an embodiment of this application. The computer device 200 includes: a processor 201; a storage device 202 on which a computer program 2020 is stored; and a network interface 203 for providing network communication functions. When the computer program 2020 is executed by the processor 201, the processor 201 implements any of the photovoltaic new energy system monitoring methods described above.

[0154] Please see Figure 3 This application provides a functional block diagram of a photovoltaic new energy system monitoring device, which includes: The data mining module is used to obtain the current operating data of the target photovoltaic system, mine the spatiotemporal distribution characteristics of the temperature gradient, shadow movement trajectory and component mismatch of the same array, the historical degradation characteristics of the temporal coupling relationship between the component degradation rate curve and the degree of mismatch, and the micro-parameter characteristics of the influence of the distribution of hidden cracks in the EL image on local mismatch. The adjacency matrix modeling module is used to perform adjacency matrix modeling based on the spatiotemporal distribution characteristics, the historical degradation characteristics, and the microscopic parameter characteristics to obtain the string topology model of the target photovoltaic system. The trend evolution prediction module is used to process the current running data using the string topology relationship model, generate a dynamic mismatch feature set, and use the dynamic mismatch feature set as input to call the time series prediction model to output the mismatch trend evolution result of the target photovoltaic system. The string control optimization module is used to generate a string-level control strategy based on the mismatch trend evolution result. During the execution of the string-level control strategy, it performs control optimization in conjunction with the power distribution logic of the target inverter of the target photovoltaic system and outputs the optimized control signal.

[0155] Based on the above, a readable storage medium is provided, on which a program or instructions are stored, and when the program or instructions are executed by a processor, the steps of the above method are implemented.

[0156] Furthermore, it should be noted that this application also provides a computer program product, which may include a computer program that can be stored in a computer-readable storage medium. The processor of a computer device reads the computer program from the computer-readable storage medium, and the processor can execute the computer program, causing the computer device to perform the aforementioned... Figure 1The methods described in the corresponding embodiments are already known, and therefore will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the computer program product embodiments related to this application, please refer to the description of the method embodiments of this application.

[0157] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

Claims

1. A monitoring method for a photovoltaic new energy system, characterized in that, The method includes: The current operating data of the target photovoltaic system is obtained, and the spatiotemporal distribution characteristics of the temperature gradient, shadow movement trajectory and component mismatch of the same array are mined to characterize the temporal coupling relationship between the component degradation rate curve and the degree of mismatch, as well as the microscopic parameter characteristics of the influence of the microcrack distribution in the EL image on the local mismatch. Based on the spatiotemporal distribution characteristics, the historical degradation characteristics, and the microscopic parameter characteristics, an adjacency matrix model is performed to obtain the string topology model of the target photovoltaic system. The current running data is processed using the string topology relationship model to generate a dynamic mismatch feature set. Using the dynamic mismatch feature set as input, the time series prediction model is called to output the mismatch trend evolution result of the target photovoltaic system. Based on the mismatch trend evolution results, a string-level control strategy is generated. During the execution of the string-level control strategy, the power distribution logic of the target inverter of the target photovoltaic system is combined to optimize the control, and the optimized control signal is output.

2. The photovoltaic new energy system monitoring method according to claim 1, characterized in that, The adjacency matrix modeling based on the spatiotemporal distribution characteristics, historical degradation characteristics, and microscopic parameter characteristics yields the string topology model of the target photovoltaic system, including: Extract the temperature gradient correlation vector, shadow trajectory correlation vector, and mismatch correlation vector from the spatiotemporal distribution features. The temperature gradient correlation vector represents the degree of correlation of temperature gradients between different components in the same array. The shadow trajectory correlation vector represents the degree of correlation of shadow movement trajectories between different components in the same array. The mismatch correlation vector represents the degree of correlation of mismatch between different components in the same array. Extract the attenuation rate coupling vector and the mismatch degree coupling vector from the historical degradation features. The attenuation rate coupling vector represents the degree of coupling of the attenuation rate curves between different components in the same string, and the mismatch degree coupling vector represents the degree of coupling of the mismatch degree between different components in the same string. Extract the hidden crack distribution correlation vector from the microscopic parameter features. The hidden crack distribution correlation vector represents the degree of correlation of hidden crack distribution in different regions of the same component and the degree of correlation with the degree of mismatch with adjacent components. Using the component identifiers of the target photovoltaic system as nodes, and the temperature gradient correlation vector, the shadow trajectory correlation vector, the mismatch correlation vector, the attenuation rate coupling vector, the mismatch degree coupling vector, and the hidden crack distribution correlation vector as edge weights, an initial adjacency matrix is ​​constructed. The rows and columns of the initial adjacency matrix are all component identifiers, and the matrix elements are the edge weights between corresponding two component identifiers. The initial adjacency matrix is ​​subjected to feature dimension alignment processing to ensure that the weight dimensions of different edges in the initial adjacency matrix are consistent. Based on the aligned initial adjacency matrix, a topological relationship constraint verification is performed to determine the set of valid edges in the initial adjacency matrix that conform to the string connection logic; wherein, the topological relationship constraint verification includes distance constraint verification of node connections, feature correlation constraint verification, and topological structure constraint verification; The initial adjacency matrix is ​​updated based on the set of valid edges, and the weights of invalid edges are set to zero to obtain the string topology model.

3. The photovoltaic new energy system monitoring method according to claim 2, characterized in that, The topological relationship constraint verification based on the aligned initial adjacency matrix, determining the set of valid edges in the initial adjacency matrix that conform to the string connection logic, includes: Extract the feature association vector set of each node in the aligned initial adjacency matrix. The feature association vector set includes the temperature gradient association vector, shadow trajectory association vector, and mismatch association vector of each node with other nodes. For each node, the correlation distribution of each vector in the feature correlation vector set of that node is analyzed to determine the core correlation node set of that node, which includes nodes that have a strong correlation with the feature correlation vector of that node; Construct a topological subgraph of the core associated node set, and determine a subgraph structure that conforms to the string connection logic based on the connectivity and hierarchical structure of the topological subgraph; The connecting edges of each node in the topological subgraph are marked as candidate valid edges, and the unmarked edges in the initial adjacency matrix are extracted. If an unmarked edge is synchronous with the feature association vector of the topological subgraph, then the edge is marked as a candidate valid edge. All candidate valid edges are subjected to collaborative verification of topological constraints, and the set of candidate valid edges that pass the collaborative verification is taken as the set of valid edges; wherein, the collaborative verification includes consistency verification of subgraph structure, matching verification of feature association vectors, and compliance verification of string connection logic.

4. The photovoltaic new energy system monitoring method according to claim 3, characterized in that, The collaborative verification of topological constraints on all candidate valid edges includes: Extract the temperature gradient correlation vector, shadow trajectory correlation vector, and mismatch correlation vector of the node pairs corresponding to the candidate valid edges; Based on the association pattern between the temperature gradient association vector and the shadow trajectory association vector, a first association pattern is determined; based on the association pattern between the shadow trajectory association vector and the mismatch association vector, a second association pattern is determined. Determine whether the first association pattern and the second association pattern conform to the preset string mismatch association pattern; If the conditions are met, then determine whether the node pairs corresponding to the candidate valid edges belong to the same topological subgraph; If they belong to the same topological subgraph, then determine whether the structure of the topological subgraph conforms to the hierarchical structure requirements of the group string connection; If the hierarchical structure requirements are met, then it is determined whether the feature association vector of the node pair corresponding to the candidate valid edge matches the core feature association vector of the topological subgraph. If a match is found, the candidate valid edge is determined to have passed collaborative verification; otherwise, the candidate valid edge is determined to have failed collaborative verification.

5. The photovoltaic new energy system monitoring method according to claim 1, characterized in that, The process of processing the current operating data using the string topology model to generate a dynamic mismatch feature set, and using the dynamic mismatch feature set as input to call the time series prediction model to output the mismatch trend evolution result of the target photovoltaic system, includes: The current running data is input into the string topology model to obtain the string association feature vector corresponding to the current running data. The string association feature vector represents the association relationship between different components in the same string and the association relationship with adjacent strings. Extract the mismatch change vector, temperature change vector, and shadow change vector from the string association feature vector. The mismatch change vector represents the degree of mismatch change of different components in the same string, the temperature change vector represents the temperature change of different components in the same string, and the shadow change vector represents the shadow change of different components in the same string. After standardizing the mismatch change vector, the temperature change vector, and the shadow change vector, feature concatenation is performed to obtain an initial dynamic mismatch feature vector. Using the initial dynamic mismatch feature vector as input, the feature enhancement layer of the time series prediction model is called, and the historical dynamic mismatch feature vectors within a preset time window are combined to perform feature enhancement processing to obtain the enhanced dynamic mismatch feature vector. The enhanced dynamic mismatch feature vector is input into the trend prediction layer of the time series prediction model, and the mismatch trend evolution result is output. The mismatch trend evolution result includes the changing trend information of the degree of mismatch and the expansion trend information of the mismatch region.

6. The photovoltaic new energy system monitoring method according to claim 5, characterized in that, The step of combining historical dynamic mismatch feature vectors within a preset time window for feature enhancement processing to obtain enhanced dynamic mismatch feature vectors includes: Extract the set of historical dynamic mismatch feature vectors within a preset time window, wherein the set of historical dynamic mismatch feature vectors contains dynamic mismatch feature vectors at consecutive time points; Analyze the feature change patterns of the historical dynamic mismatch feature vector set, where the feature change patterns include gradual or abrupt patterns. For the initial dynamic mismatch feature vector, the matching degree with the feature change pattern of the historical dynamic mismatch feature vector set is determined according to its feature change pattern. If the matching degree meets the requirements, then extract the subset of the historical dynamic mismatch feature vector set that matches the pattern of the initial dynamic mismatch feature vector; perform cross-dimensional correlation analysis on the subset and the initial dynamic mismatch feature vector, calculate the correlation between the temperature change dimension and the mismatch change dimension, the correlation between the shadow change dimension and the mismatch change dimension, and perform feature weighted combination based on the correlation weight to obtain the enhanced dynamic mismatch feature vector. If the matching degree does not meet the requirements, then extract the subset of the historical dynamic mismatch feature vector set that differs from the initial dynamic mismatch feature vector pattern; perform feature difference compensation fusion on the subset of differences and the initial dynamic mismatch feature vector to obtain the enhanced dynamic mismatch feature vector; The process of analyzing the feature change patterns of the historical dynamic mismatch feature vector set includes: Extract the feature vector differences between adjacent time points in the historical dynamic mismatch feature vector set. The feature vector differences include differences in temperature changes, differences in shading changes, and differences in mismatch changes. Based on the changing trends of the temperature change differences, determine the temperature change pattern; based on the changing trends of the shadow change differences, determine the shadow change pattern; based on the changing trends of the mismatch change differences, determine the mismatch change pattern. Determine whether the temperature change pattern, shadow change pattern, and mismatch change pattern are synchronous; If there is synchronization, then the feature change pattern of the historical dynamic mismatch feature vector set is determined to be a cooperative change pattern; If there is no synchronization, the dominance of each mode is analyzed to determine that the characteristic change mode of the historical dynamic mismatch feature vector set is a non-cooperative change mode. Based on the cooperative or non-cooperative change pattern, determine the effective feature subset of the historical dynamic mismatch feature vector set.

7. The photovoltaic new energy system monitoring method according to claim 1, characterized in that, The generation of a string-level control strategy based on the mismatch trend evolution result includes: Analyze the trends of mismatch region expansion and mismatch degree change in the mismatch trend evolution results; Extract the current power output status and component association relationship of the string corresponding to the mismatch region; By utilizing the matching relationship between the expansion trend of the mismatch region and the component association, the set of strings and the set of components that need to be adjusted are determined; For the set of strings that need adjustment, the coordination requirements for power adjustment are determined based on the power correlation between the set of strings and adjacent strings. Based on the trend of mismatch and coordination requirements, the power adjustment direction and method for each string are determined. Construct an initial string-level control strategy, which includes string adjustment instructions, component adjustment instructions, and collaborative adjustment instructions; Determine whether the initial string-level control strategy is compatible with the power conversion logic of the target photovoltaic system; if compatible, determine the initial string-level control strategy as the string-level control strategy; if incompatible, adjust the power adjustment direction or adjustment method, and reconstruct the initial string-level control strategy until it meets the compatibility requirements.

8. The photovoltaic new energy system monitoring method according to claim 1 or 7, characterized in that, The process of executing the string-level control strategy, combined with the power distribution logic of the target inverter of the target photovoltaic system, performs control optimization and outputs optimized control signals, including: Extract the string power adjustment feature set from the string-level control strategy. The string power adjustment feature set includes the adjustment correlation and adjustment constraint requirements of each string to be adjusted. Obtain the power allocation logic of the target inverter, and extract the inverter power allocation constraint set from the power allocation logic. The inverter power allocation constraint set includes the priority rules and allocation range restrictions for power allocation. Using string identifiers as rows and inverter power allocation constraint identifiers as columns, with matrix elements representing the correlation strength features between the corresponding strings and power allocation constraints, a string-inverter correlation feature matrix is ​​generated. Based on the string-inverter association feature matrix, a set of conflict points between the string power adjustment feature set and the inverter power allocation constraint set is identified, and the set of conflict points includes string adjustment terms that do not satisfy the power allocation constraints. Based on the real-time power status of the strings in the current operating data and the degree of influence of the strings on the overall power output of the system, the priority order of each conflict item in the conflict point set is determined. Based on the priority sorting, the string power adjustment feature set is constrained and adapted to obtain the adapted string power adjustment feature set. The adapted string power adjustment feature set is combined with the inverter power allocation logic to obtain the fused power control feature set. Based on the fused power control feature set, a preliminary optimized control signal set is generated; Determine whether the preliminary optimized control signal set matches the power stability constraints of the target photovoltaic system; if they match, output the optimized control signal; if they do not match, return to the constraint adaptation adjustment step until the optimized control signal that meets the matching requirements is output.

9. A computer device, characterized in that, include: A processor; a storage device having a computer program stored thereon; a network interface for providing network communication functions; when the computer program is executed by the processor, the processor enables the processor to implement the photovoltaic new energy system monitoring method as described in any one of claims 1-8.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions, which, when executed by a processor, implement the photovoltaic new energy system monitoring method as described in any one of claims 1-8.

Images