Power station output multi-scale dynamic correlation analysis method and system

Abstract

The application provides a power station output multi-scale dynamic correlation analysis method and system, relates to the new energy power generation technical field, and determines the spatial interference relationship by respectively acquiring the output power, equipment parameters, three-dimensional space coordinates and meteorological data of wind and light power stations; the wind power station flicker characteristic vector and the light power station roughness characteristic vector are calculated based on this, and the resource coordination probability is obtained by analyzing the meteorological data; the initial correlation is calculated after the multi-scale decomposition of the output power, the resource coordination probability and the two characteristic vectors are used for correction to obtain the dynamic correlation; finally, the atlas is constructed according to the dynamic correlation, topological analysis is carried out, the key power station node is identified, and the fluctuation influence grade is evaluated. The application overcomes the defect of being separated from the micro physical environment, and realizes high-precision output correlation prediction and global control.

Description

Technical Field

[0001] This invention relates to the field of new energy power generation technology, specifically to a method and system for multi-scale dynamic correlation analysis of power plant output. Background Technology

[0002] With the accelerated transformation of the global energy structure, the penetration rate of renewable energy sources such as wind and solar power in the power grid is constantly increasing. However, wind and solar power generation has significant randomness, volatility, and intermittency, and the grid connection of large-scale wind-solar hybrid power plant clusters poses a huge challenge to the safe and stable operation and dispatch of the power system. At present, integrating software algorithms with grid dispatch has become an important approach, but existing systems often fail to deeply match the microscopic physical characteristics of the underlying space and meteorology of new energy power plants, and cannot meet the urgent need for high-precision output correlation prediction and microscopic coordinated control.

[0003] In existing power plant output correlation analysis techniques, models are typically built using purely data-driven time-domain statistics or multi-scale signal decomposition methods. For example, in existing technical solutions, the first step is to perform deep multi-scale decomposition of the output power time series of wind and solar power plants based on empirical mode decomposition and its improved algorithms to obtain internal mode functions; subsequently, time-domain correlation analysis is used to calculate the dynamic correlation coefficients of components at different time scales. The core of this approach is to attempt to establish a dynamic correlation model within a local time range solely from the perspective of the pure time-domain evolution of historical power.

[0004] However, the aforementioned technologies have significant limitations in practical industrial applications: First, they rely excessively on historical sequence decomposition, detaching themselves from the physical environment and ignoring the microscopic three-dimensional spatial interference problems arising when power plants are distributed in the same field or close to each other, leading to distortion of the extracted fluctuation features. Second, they only extract from the output signal end in reverse, failing to combine the synergistic laws of meteorological resources for forward analysis, and thus cannot truly reflect the deep probabilistic dependence of wind and solar resources under complex climate conditions. Finally, when dealing with power plant clusters, existing technologies are mostly limited to local comparisons between pairs of power plants, lacking the ability to evaluate the global topology network, making it difficult to measure fluctuation propagation paths or identify key fluctuation sources that have a decisive impact on the power grid. In summary, existing models are unable to reflect complex physical and electrical coupling mechanisms, and the accuracy and global control capabilities of multi-scale correlation analysis are insufficient.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for multi-scale dynamic correlation analysis of power plant output, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method and system for multi-scale dynamic correlation analysis of power plant output, the specific steps of which include: Step 1: Obtain the output power, equipment parameters, and three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, and at the same time obtain the meteorological data of the area where each power station is located. Based on the three-dimensional spatial coordinates, determine the spatial interference relationship between each wind power station and each photovoltaic power station. Step 2: Calculate the scintillation feature vector of the wind power station based on the spatial interference relationship, the equipment parameters of the wind power station, and the meteorological data; calculate the roughness feature vector of the photovoltaic power station based on the spatial interference relationship, the equipment parameters of the photovoltaic power station, and the meteorological data. Step 3: Construct wind resource characteristics and light resource characteristics based on the meteorological data, analyze the synergistic relationship between the wind resource characteristics and the light resource characteristics, and obtain the resource synergy probability at each sampling time; Step 4: Perform multi-scale decomposition on the output power to obtain power components at different time scales; calculate the initial correlation between power components of different power plants at the same time scale, and then correct the resource synergy of the initial correlation based on the resource synergy probability to obtain the resource-corrected correlation; and combine the scintillation feature vector and the roughness feature vector to correct the resource-corrected correlation to obtain the dynamic correlation. Step 5: Construct a dynamic correlation graph with power stations as nodes based on the dynamic correlation relationship, perform topological structure analysis on the dynamic correlation graph, extract the connectivity and eigenvector centrality of each node as topological feature parameters, identify key power station nodes based on the topological feature parameters and assess their fluctuation impact level, and finally generate power station output analysis results.

[0008] Furthermore, the equipment parameters of a wind power station include hub height, rotor diameter, and rotor azimuth angle; The equipment parameters of a photovoltaic power station include: installation area, installation tilt angle, height above ground, and spacing between installations. Meteorological data include wind speed, wind direction, solar irradiance, solar altitude angle, azimuth angle, and ambient temperature; Based on the three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, calculate the relative distance, horizontal relative azimuth angle and vertical relative height difference between each wind power station and each photovoltaic power station; The spatial interference relationship between each wind power station and each photovoltaic power station is determined based on the relative distance, horizontal relative azimuth angle and vertical relative height difference. The spatial interference relationship is used to characterize the spatial geometric conditions for the shadowing effect and wind field disturbance effect between the wind power station and the photovoltaic power station. The output power and the meteorological data are both collected synchronously at the same preset sampling frequency, and the preset sampling times collected continuously together constitute the analysis time window.

[0009] Furthermore, the direction of sunlight is determined based on the solar altitude angle and azimuth angle; Using the three-dimensional spatial coordinates of the wind power station as the origin, the direction from the wind power station to the photovoltaic power station as the first coordinate axis of the local coordinate system, and the direction perpendicular to the ground as the third coordinate axis, a local relative coordinate system between the wind power station and the photovoltaic power station is constructed. In the local relative coordinate system, the hub height is offset along the third coordinate axis to obtain the hub center coordinates. The impeller sweep profile is constructed based on the hub center coordinates and the impeller diameter. The spatial attitude of the impeller sweep profile in the local relative coordinate system is determined based on the impeller azimuth angle. The impeller sweep profile is projected onto the photovoltaic power station mounting plane along the direction of sunlight to obtain a shadow projection area; Perform a geometric intersection operation on the shadow projection area and the installation area, extract the closed boundary of the intersection area, and calculate the area inside the closed boundary as the overlapping area; The ratio of the overlapping area to the installation area is used as the instantaneous occlusion ratio, and the instantaneous occlusion ratios at each sampling moment within the analysis time window are combined to obtain the flicker feature vector.

[0010] Further, the sine function value of the installation tilt angle is calculated, and the installation area is multiplied by the sine function value to obtain the effective windward cross-sectional area; Based on the installation area and the installation tilt angle, the horizontal surface projection area of ​​the photovoltaic power station is obtained through cosine projection calculation. The total surface area of ​​the gaps is calculated based on the arrangement spacing. The horizontal surface projection area and the total surface area of ​​the gaps are added together to obtain the spatial footprint of the corresponding photovoltaic power station. The ratio of the effective windward cross-sectional area to the spatial footprint is used as the terrain resistance factor. Extract the preset aerodynamic empirical coefficient, multiply the terrain drag factor by the aerodynamic empirical coefficient, and obtain the equivalent roughness length increase caused by each photovoltaic power station on the ground surface; Extract the absolute angle of the wind direction, subtract the absolute angle from the horizontal relative azimuth angle to obtain the wind direction angle, calculate the cosine function value of the wind direction angle, and multiply the cosine function value by itself to obtain the wind direction disturbance parameter; The roughness feature value is obtained by multiplying the increase in the equivalent roughness length, the wind direction disturbance parameter, and the wind speed together. The roughness feature values ​​corresponding to each sampling time within the analysis time window are combined in chronological order to obtain the roughness feature vector.

[0011] Furthermore, the logic for obtaining the resource coordination probability is as follows: Wind resource characteristics are constructed based on the wind speed, wind direction, and ambient temperature; light resource characteristics are constructed based on the solar irradiance, solar altitude angle, and azimuth angle. The wind resource features and the solar resource features are input into a preset feature encoding network for encoding to obtain wind resource state feature vectors and solar resource state feature vectors, respectively. The distribution of the wind resource state feature vector and the solar resource state feature vector within the analysis time window is statistically analyzed to obtain their respective marginal probability distributions. The marginal probability distributions of wind resources and solar resources are input into a preset Copula function for fitting, thereby establishing a joint dependency model between wind resource status and solar resource status. The resource synergy probability of wind resource status and solar resource status at each sampling time within the analysis time window is calculated based on the joint dependency model.

[0012] Furthermore, the variational mode decomposition method is used to perform mode decomposition on the output power of each wind power station and each photovoltaic power station, decomposing the output power into multiple mode components with different center frequencies and a residual trend component. Calculate the center frequency corresponding to each modal component, and sort the modal components according to the magnitude of the center frequency; The modal components with a center frequency higher than a first preset threshold are identified as short-time-scale power components. The modal components whose center frequency is between the first preset threshold and the second preset threshold are defined as medium-time scale power components. The modal components with center frequencies below the second preset threshold and the residual trend components are identified as long-time scale power components. The first preset threshold is greater than the second preset threshold; The corresponding modal components at the same time scale are reconstructed to obtain short-time scale power sequences, medium-time scale power sequences and long-time scale power sequences. For the power sequences of different power plants at the same time scale, calculate the Pearson correlation coefficient between them, and use the Pearson correlation coefficient as the initial correlation at that time scale.

[0013] Further, the values ​​of the scintillation feature vector and the roughness feature vector are extracted at each sampling time within the analysis time window. The scintillation feature vector and the roughness feature vector are resampled according to the same time scale as the power component. The spatial interference intensity between the corresponding power station pairs is calculated based on the resampled scintillation feature value and roughness feature value. Extract the resource coordination probability corresponding to each sampling time, and use the resource coordination probability to correct the initial correlation at the corresponding time scale to obtain the resource correction correlation. The resource modification correlation is attenuated and corrected based on the spatial interference intensity to obtain the dynamic correlation corresponding to each time scale.

[0014] Furthermore, each power station is designated as a node; Based on the dynamic association relationship, connection edges are established between nodes, and the dynamic association relationship is used as the edge weight of the connection edge; Power plant association networks at different time scales were constructed, and the power plant association networks at different time scales were combined to obtain a dynamic association map; Traverse the dynamic association graph, for any node in the dynamic association graph, extract the edge weights of all connecting edges connected to that node, sum the edge weights to obtain the connectivity of the node; Construct a graph adjacency matrix with the edge weights of each node in the dynamic association graph as matrix elements; The maximum eigenvalue of the graph adjacency matrix is ​​solved using the power iteration algorithm, and the eigenvector corresponding to the maximum eigenvalue is calculated. Extract the element values ​​corresponding to the node from the feature vector as the feature vector centrality of the node; The nodes are sorted in descending order based on the magnitude of their topological feature parameters. Nodes with topological feature parameters greater than a preset key threshold are identified as key power plant nodes. The connectivity of the key power plant nodes is used as the correlation propagation range, and the feature vector centrality is used as the correlation strength. The fluctuation impact level of the corresponding power plant is determined by combining the correlation propagation range and the correlation strength. Power output analysis results are generated based on the fluctuation impact level of each power station.

[0015] A power plant output multi-scale dynamic correlation analysis system, wherein the power plant output multi-scale dynamic correlation analysis system is used to implement the above-mentioned power plant output multi-scale dynamic correlation analysis method, including: The data acquisition module is used to acquire the output power, equipment parameters and three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, and at the same time acquire the meteorological data of the area where each power station is located. Based on the three-dimensional spatial coordinates, the spatial interference relationship between each wind power station and each photovoltaic power station is determined. The feature calculation module is used to calculate the scintillation feature vector of the wind power station based on the spatial interference relationship, the equipment parameters of the wind power station and the meteorological data, and to calculate the roughness feature vector of the photovoltaic power station based on the spatial interference relationship, the equipment parameters of the photovoltaic power station and the meteorological data. The collaborative analysis module is used to construct wind resource characteristics and light resource characteristics based on the meteorological data, analyze the collaborative relationship between the wind resource characteristics and the light resource characteristics, and obtain the resource collaboration probability at each sampling time. The dynamic correlation module is used to perform multi-scale decomposition on the output power to obtain power components at different time scales; calculate the initial correlation relationship between power components of different power plants at the same time scale; then correct the resource synergy of the initial correlation relationship based on the resource synergy probability to obtain the resource-corrected correlation relationship; and combine the scintillation feature vector and the roughness feature vector to correct the resource-corrected correlation relationship to obtain the dynamic correlation relationship. The graph analysis module is used to construct a dynamic correlation graph with power stations as nodes based on the dynamic correlation relationship, perform topological structure analysis on the dynamic correlation graph, extract the connectivity and eigenvector centrality of each node as topological feature parameters, identify key power station nodes based on the topological feature parameters and assess their fluctuation impact level, and finally generate power station output analysis results.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention determines the spatial interference relationship between wind power stations and photovoltaic power stations based on the obtained output power, equipment parameters, three-dimensional spatial coordinates, and meteorological data, overcoming the shortcomings of existing analyses that often detach from the microscopic physical operating environment, leading to feature distortion. On this basis, by integrating spatial interference relationship, equipment parameters, and meteorological data, the invention calculates the scintillation feature vector of the wind power station and the roughness feature vector of the photovoltaic power station, closely combining pure data mining with the physical interference effect when power stations are distributed close together. Simultaneously, wind resource features and solar resource features are constructed based on meteorological data, and the resource coordination probability at each sampling time is obtained by analyzing the synergistic relationship. This scheme overcomes the limitations of reverse stripping from a single output power signal, truly reflecting the underlying probability dependency relationship, and significantly improving the objectivity and accuracy of feature extraction. This invention also decomposes output power into power components at different time scales through multi-scale decomposition and calculates the initial correlation between power components of different power plants at the same time scale, thus solving the problem that existing technologies are mostly limited to local comparisons and lack global evaluation capabilities. In this process, resource synergy is used to correct the initial correlation using resource synergy probability, resulting in a corrected correlation. This is further corrected by combining scintillation feature vectors and roughness feature vectors to obtain a dynamic correlation that accurately characterizes the evolution pattern. Furthermore, a dynamic correlation graph with power plants as nodes is constructed based on the dynamic correlation. Topological structure analysis is performed on this dynamic correlation graph, and the connectivity and feature vector centrality of each node are extracted as topological feature parameters. Based on these topological feature parameters, key power plant nodes can be accurately identified from a global network perspective, their fluctuation impact level can be assessed, and finally, power plant output analysis results can be generated. This design breaks through the superficiality and lag of local calculations, providing a global structural analysis perspective oriented towards clusters, and achieving efficient and accurate positioning of key nodes. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall method flow of the present invention.

[0018] Figure 2 This is a graph showing the changes in the scintillation feature vector and the roughness feature vector of the present invention.

[0019] Figure 3 This is a block diagram of the overall system modules of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0021] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0022] Example: Please see Figures 1-2The present invention provides a technical solution: The method for multi-scale dynamic correlation analysis of power plant output includes the following steps: Step 1: Obtain the output power, equipment parameters, and three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, and at the same time obtain the meteorological data of the area where each power station is located. Based on the three-dimensional spatial coordinates, determine the spatial interference relationship between each wind power station and each photovoltaic power station.

[0023] In this embodiment, the equipment parameters of the wind power station include hub height, rotor diameter, and rotor azimuth angle.

[0024] The equipment parameters of a photovoltaic power station include: installation area, installation tilt angle, height above ground, and spacing between installations.

[0025] Meteorological data include wind speed, wind direction, solar irradiance, solar altitude angle, azimuth angle, and ambient temperature.

[0026] Based on the three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, calculate the relative distance, horizontal relative azimuth angle, and vertical relative height difference between each wind power station and each photovoltaic power station.

[0027] In this embodiment, the calculation is not limited to two adjacent or single specified power plants, but rather includes... One wind power station and The renewable energy cluster of photovoltaic power stations performs many-to-many traversal and cross-calculation. Through nested loop logic, using any wind power station in the cluster as a reference, the three-dimensional spatial coordinates of each photovoltaic power station in the cluster are retrieved, and the relative distance, azimuth angle, and elevation difference are calculated sequentially. This is achieved by completing... This spatial geometric transformation can establish a spatial interference relationship matrix between all wind and solar power stations in the entire field, thereby ensuring that no potential physical interference link is missed when constructing the dynamic correlation map.

[0028] To determine the microscopic geometric interference relationship between the two power stations, the three-dimensional spatial coordinates of the wind power station and the photovoltaic power station are transformed into a unified geodetic rectangular coordinate system. The triaxial differences between the two points in the three-dimensional rectangular coordinate system are calculated, and geometric distance calculations are performed to obtain the relative distance between the two power stations. The horizontal relative azimuth angle between the two points is derived by performing arctangent trigonometric function calculations using the difference in the horizontal and vertical coordinates of the two points on the horizontal projection plane. Finally, the vertical relative elevation difference between the wind power station and the photovoltaic power station is obtained by directly subtracting their reference elevations.

[0029] The spatial interference relationship between each wind power station and each photovoltaic power station is determined based on the relative distance, horizontal relative azimuth angle and vertical relative height difference. The spatial interference relationship is used to characterize the spatial geometric conditions under which shading and wind field disturbance occur between the wind power station and the photovoltaic power station.

[0030] Among them, the relative distance represents the straight-line physical span between the wind power station and the photovoltaic power station, used to assess the spatial attenuation of interference intensity; the horizontal relative azimuth represents the absolute orientation relationship between the two on the ground plane, used for subsequent calculations of the solar radiation projection angle and the direction of the wind wake; the vertical relative height difference represents the difference in elevation between the two on the terrain, used to correct for the true relative height between the hub height and the windward side of the photovoltaic array. These three parameters are encapsulated and combined to establish the spatial interference relationship between the wind power station and the photovoltaic power station.

[0031] The output power and the meteorological data are both collected synchronously at the same preset sampling frequency, and the preset sampling times collected continuously together constitute the analysis time window.

[0032] The underlying data and physical parameters of the hybrid renewable energy complementary power generation cluster are synchronously collected. To ensure the accuracy of fluctuation capture, a fixed preset sampling frequency is set to perform high-frequency synchronous data acquisition on all equipment in the field. Multiple preset sampling times collected continuously constitute the analysis time window for subsequent data analysis.

[0033] In actual industrial scenarios, wind turbines and photovoltaics are not independent entities that do not interfere with each other. On the one hand, when the sun reaches a specific altitude and azimuth angle, the wind turbine tower and the huge rotating impeller will project a violently oscillating dynamic shadow on the photovoltaic panel downwind along the direction of sunlight, directly causing a momentary drop in the photovoltaic output signal. On the other hand, the large-area, tilted photovoltaic array changes the physical roughness of the natural surface, forming a new type of aerodynamic boundary layer, which generates frictional disturbances to the underlying flow field, thereby changing the inflow velocity and turbulence of the downwind wind turbine.

[0034] Constructing spatial interference relationships by extracting relative distance, azimuth, and elevation differences essentially embeds physical white-box constraints of optical projection geometry and fluid dynamics into the data model. This synchronous acquisition mechanism and spatial geometric dimensionality reduction effectively eliminate local environmental noise caused by terrain undulations and spatial distribution differences, establishing accurate boundary conditions for precisely locking the propagation path of power fluctuations and characterizing the underlying spatiotemporal coupling mechanism in subsequent multi-scale decomposition.

[0035] Step 2: Calculate the scintillation feature vector of the wind power station based on the spatial interference relationship, the equipment parameters of the wind power station, and the meteorological data; calculate the roughness feature vector of the photovoltaic power station based on the spatial interference relationship, the equipment parameters of the photovoltaic power station, and the meteorological data.

[0036] In this embodiment, the direction of sunlight is determined based on the solar altitude angle and azimuth angle.

[0037] First, real-time meteorological data at the current sampling moment is extracted, including the solar altitude angle, such as 45°, and the solar azimuth angle, such as 135°, i.e., southeast. These two angles directly determine the incident vector of sunlight, and the three-dimensional direction of sunlight at the current moment is established accordingly.

[0038] Using the three-dimensional spatial coordinates of the wind power station as the origin, the direction from the wind power station to the photovoltaic power station as the first coordinate axis of the local coordinate system, and the direction perpendicular to the ground as the third coordinate axis, a local relative coordinate system between the wind power station and the photovoltaic power station is constructed.

[0039] Using the center of the bottom of the wind power station as the origin Subsequently, based on the calculated horizontal relative azimuth angle, the line connecting the bottom of the wind turbine and the center of the photovoltaic power station array was set as... The axis, the first coordinate axis, means The axis always points towards the photovoltaic power station. Simultaneously, the upward direction perpendicular to the ground is set as... The third coordinate axis. Through this translation and rotation of relative coordinates, the complex absolute geodetic coordinates were successfully transformed into a microscopic coordinate environment that only focuses on the relative positions of the wind turbine and the photovoltaic panel.

[0040] In the local relative coordinate system, the hub height is offset along the third coordinate axis to obtain the hub center coordinates. The impeller sweep profile is constructed based on the hub center coordinates and the impeller diameter. The spatial attitude of the impeller sweep profile in the local relative coordinate system is determined based on the impeller azimuth angle.

[0041] Starting from the origin, along Extending the axis upwards to the height of the wheel hub, such as 100 meters, this point is the coordinate of the wheel hub center. Using the hub center as the center and the impeller diameter, such as 121 meters, as the diameter, a circular impeller sweep profile representing the area swept by the rotating blades is constructed. However, the wind turbine is not stationary; its impeller yaws according to the wind direction. Therefore, the real-time acquired impeller azimuth angle is introduced, such as the current turbine nacelle facing 90° due east. Using a three-dimensional rotation matrix, this initially constructed impeller sweep profile is rotated around... The shaft is rotated to its actual orientation relative to the wind. At this point, the three-dimensional obstruction surface of the wind turbine impeller in the local space has been accurately reproduced.

[0042] The impeller sweep profile is projected onto the photovoltaic power station mounting plane along the direction of sunlight to obtain a shadow projection area.

[0043] Using the direction of sunlight as the projection ray, the sweeping profile of the turbine rotor in its actual spatial orientation is projected parallel to these rays towards the ground. Because the photovoltaic panels have a fixed installation tilt angle and ground clearance, the projection surface is not a horizontal ground surface, but rather the actual installation plane of the photovoltaic power station. After projection, a two-dimensional shadow projection area formed by the turbine rotor's shading is obtained on this installation plane.

[0044] Perform a geometric intersection operation on the shadow projection area and the installation area, extract the closed boundary of the intersection area, and calculate the area inside the closed boundary as the overlapping area.

[0045] Since the installation area of ​​a photovoltaic power station is limited, the shadow of the rotor does not always fall entirely on the photovoltaic panel. Therefore, it is necessary to perform a geometric intersection calculation on the shadow projection area and the actual installation area outline of the photovoltaic power station. Through calculation algorithms, the closed polygon boundary of the overlapping part is extracted, and the precise area inside this boundary, i.e., the overlapping area, is obtained using the polygon area calculation formula.

[0046] The ratio of the overlapping area to the installation area is used as the instantaneous occlusion ratio, and the instantaneous occlusion ratios at each sampling moment within the analysis time window are combined to obtain the flicker feature vector.

[0047] Dividing the calculated overlap area at the current moment by the total installed area of ​​the photovoltaic power station yields the instantaneous shading ratio at that moment. This represents the effective shading effect of the wind turbine on the photovoltaic power station at the current moment. This calculation is repeated for each sampling moment throughout the entire analysis time window. Connecting the instantaneous shading ratios calculated at all moments in a time series yields a one-dimensional time series vector, which is the flicker feature vector used to characterize the dynamic shading effect of the wind turbine on the photovoltaic system.

[0048] The output of a photovoltaic (PV) power plant is directly and strongly dependent on solar irradiance. When wind farms and PV power plants are built together or distributed close to each other, the massive shadows cast by wind turbine towers reaching hundreds of meters in height and rotating blades tens of meters long will inevitably fall on the PV panels at specific solar altitude and azimuth angles. Because the wind turbine blades are constantly rotating, this shadow is not a static blockage, but rather causes high-frequency alternations of light and dark irradiance received by local panels, the so-called flicker effect. These high-frequency abrupt changes in irradiance can lead to current mismatch within the PV array strings, resulting in frequent activation of the maximum power point tracking controller and severe high-frequency fluctuations in the overall PV output power.

[0049] This embodiment innovatively abstracts the microscopic optical interference process into a quantifiable scintillation feature vector through rigorous 3D geometric modeling, local coordinate system construction, realistic pose reconstruction, and intersection of solar ray projections. This process not only realistically reproduces the physical scene but also accurately isolates the high-frequency power fluctuations caused by shadows, which were previously difficult to capture through macroscopic data. This provides crucial physical boundary inputs for subsequent correction of initial correlations, enabling the model to distinguish which power fluctuations are due to coordinated changes caused by macroscopic weather conditions and which are merely noise interference caused by local physical occlusion. This significantly improves the objectivity of low-level feature extraction and the final accuracy of multi-scale correlation analysis.

[0050] In this embodiment, the sine function value of the installation tilt angle is calculated, and the installation area is multiplied by the sine function value to obtain the effective windward cross-sectional area.

[0051] Extract the installation area and tilt angle of the photovoltaic power station panels. By converting the tilt angle using a sine trigonometric function and multiplying it by the total installation area, the vertical cross-sectional area that actually physically obstructs the photovoltaic array when airflow blows horizontally can be obtained.

[0052] In microfluidics, the obstruction effect of a tilted photovoltaic panel on airflow depends not on its absolute surface area, but on its projection onto the windward vertical plane. This step uses a sine function to extract the effective windward cross-sectional area, accurately reconstructing the actual obstruction section of the photovoltaic panel on the underlying wind field in terms of physical logic. This provides a geometric basis consistent with aerodynamic laws for subsequent quantification of wind resistance, avoiding serious deviations caused by directly using the total area for calculation.

[0053] Based on the installation area and the installation tilt angle, the horizontal surface projection area of ​​the photovoltaic power station is obtained through cosine projection calculation. The total surface area of ​​the gaps is calculated based on the arrangement spacing. The horizontal surface projection area is added to the total surface area of ​​the gaps to obtain the spatial footprint of the corresponding photovoltaic power station.

[0054] The cosine function value of the installation tilt angle is calculated and multiplied by the installation area to obtain the absolute shading area of ​​all photovoltaic panels laid flat on the ground. Then, based on the spacing between the photovoltaic arrays, the total area of ​​the ground surface not covered by the panels between the arrays is calculated. Adding the projected area to the gap area reconstructs the complete boundary range of the photovoltaic power station's alteration of the original landform, i.e., the spatial footprint.

[0055] A photovoltaic power station is not a monolithic entity; its unique array and gap arrangement constitute a discontinuous physical boundary layer. By adding the solid cosine projection to the gap area, the overall occupancy of the power station on the landform can be accurately reflected. This allows subsequent disturbance assessments to go beyond isolated photovoltaic panels, focusing on the overall disturbance scale of the macroscopic array, thus improving the objectivity of landform modeling.

[0056] The ratio of the effective windward cross-sectional area to the spatial footprint is used as the terrain resistance factor.

[0057] Extract the preset aerodynamic empirical coefficient, multiply the terrain drag factor by the aerodynamic empirical coefficient, and obtain the equivalent roughness length increase caused by each photovoltaic power station on the ground surface.

[0058] In this embodiment, the aerodynamic empirical coefficient is set to 0.5, a value that objectively reflects the degree of micro-damping of natural wind fields by conventional photovoltaic arrays. Typically, the frictional resistance characteristics of tilted photovoltaic panel arrays to surface airflow are similar to those of sparse windbreaks or low-rise building clusters.

[0059] The terrain drag factor directly reflects the density and three-dimensional obstruction of obstacles within a specified area. Introducing empirical aerodynamic coefficients for multiplication is a crucial step in transforming purely spatial geometric characteristics into microscopic hydrodynamic characteristics. This quantifies the substantial changes in the natural surface friction characteristics caused by large-scale photovoltaic cluster construction, solving the quantitative problem in traditional correlation analysis of the physical interference caused by power station infrastructure on local micro-meteorology.

[0060] Extract the absolute angle of the wind direction, subtract the absolute angle from the horizontal relative azimuth angle to obtain the wind direction angle, calculate the cosine function value of the wind direction angle, and multiply the cosine function value by itself to obtain the wind direction disturbance parameter.

[0061] The specific formula for calculating wind direction disturbance parameters is as follows: In the formula, For wind direction disturbance parameters, The absolute angle of the wind direction. The horizontal relative azimuth angle is given.

[0062] The cosine square function is used because the energy loss of wake disturbances in space is strongest along their central axis, exhibiting a rapid, non-linear decay as they shift to either side. When the wind direction is perfectly parallel to the line connecting the two power stations, the cosine value is 1, indicating the most severe disturbance effect. As the wind direction changes, the square operation ensures that this parameter remains a positive damping coefficient and allows for smooth decay. Its beneficial effect is that it perfectly matches the physical law of lateral wake decay in fluid mechanics, resulting in extremely high physical fidelity in the spatial directivity calculation of disturbance intensity.

[0063] The roughness characteristic value is obtained by multiplying the increase in the equivalent roughness length, the wind direction disturbance parameter, and the wind speed together.

[0064] The roughness feature values ​​corresponding to each sampling time within the analysis time window are combined in chronological order to obtain the roughness feature vector.

[0065] The formula for calculating the roughness characteristic value is: In the formula, These are roughness characteristic values. This is the increase in equivalent roughness length. For wind direction disturbance parameters, The wind speed is [value].

[0066] By deeply cross-multiplying static topographic changes, dynamic wind direction deflection weights, and real-time wind speed, a coupled feature quantity integrating spatiotemporal and meteorological variables is constructed. The isolated static physical arrangement of the photovoltaic array is transformed into a continuous dynamic feature vector that can participate in time-series correction. This enables the model to accurately determine whether the decline in wind turbine output is due to macroscopic windlessness or microscopic airflow obstruction caused by photovoltaic array interception under specific wind conditions, greatly improving the analytical accuracy of power plant output fluctuation correlation analysis.

[0067] Step 3: Construct wind resource characteristics and light resource characteristics based on the meteorological data, analyze the synergistic relationship between the wind resource characteristics and the light resource characteristics, and obtain the resource synergy probability at each sampling time.

[0068] In this embodiment, the logic for obtaining the resource collaboration probability is as follows: Wind resource characteristics are constructed based on the wind speed, wind direction, and ambient temperature; light resource characteristics are constructed based on the solar irradiance, solar altitude angle, and azimuth angle. The meteorological data were preprocessed using the minimum-maximum normalization method. The absolute angles of wind speed and direction and ambient temperature were extracted from the preprocessed data, and these three physical quantities were concatenated in sequence to construct a 3D initial feature vector of wind resources. Similarly, the normalized solar irradiance, solar altitude angle and azimuth angle were extracted and concatenated to construct a 3D initial feature vector of light resources.

[0069] The work potential of wind and solar energy cannot be determined by a single variable. For example, ambient temperature affects air density, which in turn affects the kinetic energy captured by wind turbine blades and the photoelectric conversion efficiency of photovoltaic modules. Elevation and azimuth angles determine the depth of light penetration. By aggregating multi-dimensional meteorological factors and constructing a unified resource feature vector, this study aims to comprehensively characterize the true work potential of raw meteorological resources from multiple dimensions, laying a high-quality data foundation for subsequent exploration of the deep complementary relationships between wind, solar, and meteorological energy.

[0070] The wind resource features and the solar resource features are input into a preset feature encoding network for encoding to obtain wind resource state feature vectors and solar resource state feature vectors, respectively. The preset feature encoding network adopts VQ-VAE. The input feature matrix, containing high-dimensional historical meteorological information and multi-channel satellite cloud images, is fed into the VQ-VAE encoder network, initially mapping it into continuous latent variables. Subsequently, a learnable discrete codebook is established in the latent space. The Euclidean distance between the continuous latent variables and each vector in the codebook is calculated, and the best-matching discrete vector is found for replacement. This embeds the high-dimensional features into the low-dimensional latent space, achieving vectorization and dimensionality reduction of the data features. The specific vector quantization calculation logic is as follows: In the formula, This represents the feature vector of wind or solar resource status. These are the high-dimensional continuous latent variables initially output by the encoder; For the first in the preset codebook A discrete feature vector, This is the traversal index.

[0071] Traditional autoencoders output continuous high-dimensional features that are highly susceptible to interference from extreme weather distortions or local noise in satellite cloud images, leading to blurred feature spaces. The VQ-VAE model addresses this by forcing complex, continuous weather changes to align with a finite number of typical weather patterns, i.e., discrete vectors in the codebook. This forced discretization dimensionality reduction is analogous to classifying ever-changing cloud trajectories and cyclone fluctuations into dozens of standard weather types. It filters out high-frequency redundant noise from the original data, avoiding posterior collapse, and the resulting vectorized dimensionality-reduced features exhibit strong robustness and representativeness, providing extremely clear state boundaries for subsequent probability calculations.

[0072] The distribution of the wind resource state feature vector and the solar resource state feature vector within the analysis time window is statistically analyzed to obtain their respective marginal probability distributions. Collect all wind resource state feature vectors and solar resource state feature vectors output within the entire analysis time window. For these discrete feature state data, use kernel density estimation to fit a continuous probability density. By integrating the probability density function, calculate the cumulative probability of each feature state occurring over a long historical period at each sampling time. This yields the probability values ​​of wind resources and solar resources in a certain state, both strictly constrained within the [0,1] interval, forming a complete marginal probability distribution sequence.

[0073] While the absolute values ​​of wind speed or irradiance are not comparable across regions due to different seasons or terrains, their transformation into marginal probability distributions reflects the current meteorological level's position within the entire annual cycle, such as whether it is at an extremely high or extremely low level. This eliminates the base differences in the underlying climate background, providing a standardized input format for establishing a unified joint dependency model.

[0074] The marginal probability distributions of wind resources and solar resources are input into a predefined Copula function for fitting, thereby establishing a joint dependency model between the states of wind resources and solar resources.

[0075] To address the complex nonlinear dependence between wind speed and solar power output, this scheme does not use a single fixed formula. Instead, it employs a multiple candidate set of Copula functions, including Gumbel, Clayton, and Frank functions, for correlation modeling. The marginal probability distributions of wind and solar resources are substituted into each candidate Copula function for calculation. Then, the goodness of fit is evaluated using the Bayesian information criterion, and the function with the best score is selected to construct the final binary joint probability distribution model. The optimal calculation logic is as follows: In the formula, The value is the Bayesian information criterion value. This represents the maximum likelihood estimate of the candidate Copula function with respect to the current marginal probability distribution. This represents the number of independent arguments for the corresponding Copula function. To analyze the total number of samples within the time window, For the optimal Copula function, The marginal probability distribution of wind resources. This represents the marginal probability distribution of light resources. For the Gumbel Copula function, For Clayton Copula functions, This is the Frank Copula function.

[0076] In complex and ever-changing real-world climate physics scenarios, the complementary nature of wind and solar resources is not static. For example, Gumbel Copula is highly sensitive to the upper-tail correlation of extreme weather conditions, Clayton Copula excels at capturing the lower-tail correlation under weak wind and low light conditions, while Frank Copula is better suited to describing symmetrical, normal correlations. If only one Copula function is preset, probability fitting under specific climatic conditions will inevitably fail. By introducing a goodness-of-fit evaluation for dynamic optimization, the model can adaptively select the joint probability function that best fits the underlying meteorological logic under different seasons and geographical wind belts. This endows the model with adaptive learning capabilities in heterogeneous climatic environments, avoids biases introduced by static models, and ensures that the final output binary joint probability distribution model possesses the highest physical fidelity and statistical accuracy.

[0077] The resource synergy probability of wind resource status and solar resource status at each sampling time within the analysis time window is calculated based on the joint dependency model.

[0078] The joint probability density perfectly eliminates the bias of the single Copula model under heterogeneous climates. This cooperative probability, as a weight parameter with extremely high physical confidence, directly determines the magnitude of subsequent multi-scale power correlation corrections. It can accurately distinguish whether the synchronous fluctuations in the power of the two power plants originate from genuine meteorological coordination of wind and solar power generation at the underlying level, or from spurious correlations caused by grid curtailment / equipment failure, thus solidifying the underlying physical logic of dynamic correlation analysis.

[0079] Step 4: Perform multi-scale decomposition on the output power to obtain power components at different time scales; calculate the initial correlation between power components of different power plants at the same time scale, and then correct the resource synergy of the initial correlation based on the resource synergy probability to obtain the resource-corrected correlation. Combine the scintillation feature vector and the roughness feature vector to correct the resource-corrected correlation to obtain the dynamic correlation.

[0080] In this embodiment, the variational mode decomposition method is used to perform mode decomposition on the output power of each wind power station and each photovoltaic power station, decomposing the output power into multiple mode components with different center frequencies and a residual trend component.

[0081] The actual output power time series of a wind power station and a photovoltaic power station within the analysis time window are extracted. A variational mode decomposition algorithm is then used to extract the signal from the one-dimensional non-stationary power signal. The core parameters of the variational mode decomposition algorithm are preset as follows: the number of mode decompositions is set to 8, and the penalty factor is 2000 to ensure the compactness of the mode bandwidth. By iteratively searching for the optimal solution of the variational model, the originally chaotic total output power signal is precisely decomposed into 8 mode components with specific bandwidths and 1 residual trend component reflecting the overall slow upward or downward trend throughout the year.

[0082] The actual output power of a wind-solar power station is the final result of the superposition and mixing of multiple different physical processes. If the original total power is analyzed directly, the fluctuations of various frequencies will mask each other, making it impossible to see the complementary relationship. By transforming it into a non-recursive variational optimization problem, the power fluctuations caused by different physical factors can be thoroughly separated, so that each modal component has a strict center frequency. This provides a high-fidelity, pure signal source for the subsequent accurate positioning of wind-solar complementary characteristics at different time scales.

[0083] Calculate the center frequency corresponding to each modal component, and sort the modal components according to the magnitude of the center frequency; The modal components with a center frequency higher than a first preset threshold are identified as short-time-scale power components. The modal components whose center frequency is between the first preset threshold and the second preset threshold are defined as medium-time scale power components. The modal components with center frequencies below a second preset threshold and the residual trend components are identified as long-time scale power components.

[0084] The first preset threshold is greater than the second preset threshold; For the eight modal components decomposed by VMD, their spectra are extracted using Fourier transform, their center frequencies are calculated, and they are arranged in descending order of frequency (i.e., period from shortest to longest). In this embodiment, a first preset threshold is set as... Hertz, corresponding to a physical cycle of 4 hours. Set the second preset threshold to... Hertz corresponds to a physical cycle of 3 days.

[0085] The system categorizes components based on conditional logic: modal components with a center frequency higher than the first preset threshold and a period of less than 4 hours are classified as short-timescale power components; modal components with a center frequency between the two and a period between 4 hours and 3 days are classified as medium-timescale power components; and modal components with a center frequency lower than the second preset threshold and a period of more than 3 days, along with the residual trend component representing extremely low frequencies, are collectively classified as long-timescale power components.

[0086] Setting a fixed frequency threshold for truncation and classification is intended to forcibly align the signal frequency bands of pure mathematics with the physical cycles of meteorology. Short timescales typically correspond to highly random high-frequency fluctuations caused by local cloud cover and micro-topographic gusts; medium timescales typically correspond to diurnal regular fluctuations caused by day-night alternation and local convective weather systems; and long timescales correspond to low-frequency trend fluctuations caused by alternating monsoon cyclones and gradual seasonal temperature changes.

[0087] The corresponding modal components at the same time scale are reconstructed to obtain short-time scale power sequences, medium-time scale power sequences and long-time scale power sequences. For the power sequences of different power plants at the same time scale, calculate the Pearson correlation coefficient between them, and use the Pearson correlation coefficient as the initial correlation at that time scale.

[0088] For wind power plant A and photovoltaic power plant B, the Pearson correlation coefficient is calculated between their short-timescale power series. This calculation logic yields a value between [-1, 1] by measuring the ratio of the product of the covariance and the standard deviation of the two series.

[0089] For example, the Pearson correlation coefficient between power plants A and B at the medium time scale is calculated to be -0.65, indicating a strong negative correlation between the two at this scale, i.e., good complementarity. The correlation coefficients for the sequence pairs at the short, medium, and long time scales are calculated separately and used as the initial association for the power plant pair.

[0090] This provides a baseline without any prior physical or meteorological knowledge. While this initial correlation has limitations such as superficiality and susceptibility to environmental noise, it objectively reflects the actual power fluctuation matching status received at the power grid end. It is precisely because of this purely data-level initial correlation as a raw target to be corrected that the scintillation feature vector, roughness feature vector, and resource coordination probability painstakingly obtained in the preceding steps can play their corrective role as physical and meteorological constraints, thus constructing an analytical closed loop driven by both surface and underlying principles, and by data and mechanisms.

[0091] In this embodiment, the values ​​of the scintillation feature vector and the roughness feature vector are extracted at each sampling time within the analysis time window. The scintillation feature vector and the roughness feature vector are resampled according to the same time scale as the power component. The spatial interference intensity between the corresponding power station pairs is calculated based on the resampled scintillation feature value and roughness feature value. The scintillation and roughness feature vectors are extracted. Since these two physical feature vectors are generated based on the original 15-minute high-frequency sampling time, and the power component has been divided into three specific time scales—short, medium, and long (e.g., the medium time scale corresponds to 4 hours to 3 days)—this scheme resamples the original feature vectors to achieve temporal alignment between the physical features and the power signal. For the medium time scale, a window averaging method is used to calculate the mean of all high-frequency feature values ​​within the corresponding time period, yielding the resampled scintillation and roughness feature values. Subsequently, these two features, belonging to the optical and fluid dynamics fields respectively, are fused and calculated. The formula for calculating the spatial interference intensity is as follows: In the formula, For spatial interference intensity, These are the flicker feature values ​​after resampling. These are the roughness feature values ​​after resampling. The preset theoretical maximum value for surface roughness is set to 0.8. and These are the preset penalty weights for the flicker effect and roughness disturbance on the overall interference, respectively. In this embodiment, they are set to 0.6 and 0.4, respectively, depending on the density of the power station layout.

[0092] flickering characteristics Essentially, it's about the area occlusion ratio, which is a dimensionless parameter of [0,1]; while roughness features... It is the equivalent roughness length change calculated based on fluid mechanics, with physical units and a large numerical fluctuation range. A preset theoretical maximum value is intentionally introduced into the formula. As the denominator, the fluid dynamics characteristics are forcibly transformed into a relative proportion of [0,1]. This logic solves the interdisciplinary problem of directly merging two completely different physical quantities, optical projection occlusion and wind field airflow disturbance, which cannot be directly calculated. Through and Linear combination allows for dynamic penalties based on the density of the power plant's micro-layout. For example, in areas with extremely dense photovoltaic panels, the penalty can be adjusted upwards. The weighting of this highlights the dominant role of optical occlusion.

[0093] Spatial interference intensity, used as a decay factor for subsequent correlation correction, mathematically requires its value to strictly converge within the interval [0,1). If a simple linear superposition is used, under extremely harsh conditions with severe shading and significant airflow disturbance, the calculated interference intensity can easily exceed 100%, leading to negative values ​​or numerical collapse in subsequent correlation corrections. Therefore, [the following approach is adopted]. The function, regardless of the degree of internal interference, will ultimately output... They will only infinitely approach 1 and never cross the boundary, ensuring the absolute robustness of the all-weather algorithm.

[0094] The two heterogeneous physical disturbances are transformed into a unified scalar of spatial interference intensity. The physical significance of the exponential decay structure is that when optical obstruction or flow field hindrance is extremely severe, the combined interference intensity smoothly approaches 1, avoiding numerical overflow that may be caused by linear addition. This allows the multidimensional physical environment boundary conditions to be successfully reduced to a standardized parameter, providing a reliable mathematical tool for subsequent correlation correction.

[0095] Extract the resource coordination probability corresponding to each sampling time, and use the resource coordination probability to correct the initial correlation at the corresponding time scale to obtain the corrected resource correlation.

[0096] The initial correlations within the corresponding time window are obtained. For example, the initial Pearson correlation coefficient of a certain wind and solar power station at a short time scale is -0.75, seemingly showing strong complementarity. At the same time, the resource collaboration probability at each sampling time is extracted from the Copula joint dependency model, and the average probability within that time scale window is calculated. For example, the calculated average meteorological collaboration probability for this period is only 0.25.

[0097] The initial correlation merely reflects changes in meter readings received at the end of the grid. However, in actual industrial operations, unnatural behaviors such as grid dispatch instructions, collective equipment maintenance, or inverter group control can cause the output curves of two power plants to exhibit extremely high temporal synchronization. Blindly believing this illusion can fatally mislead the grid's peak-shaving planning.

[0098] When the probability of meteorological correlation is extremely low, while the apparent power correlation is extremely high, it indicates that the current power fluctuations are dominated by human or non-natural factors. By multiplying by a low probability coefficient, this scheme can decisively eliminate spurious correlations caused by power grid curtailment or equipment failure, ensuring that the derived corrected correlation is entirely based on the objective foundation driven by natural meteorological factors, thus enhancing the anti-interference capability of the analysis model.

[0099] The resource modification correlation is attenuated and corrected based on the spatial interference intensity to obtain the dynamic correlation corresponding to each time scale.

[0100] To accurately reflect the destructive effect of microscopic physical obstruction on macroscopic meteorological coordination, an interference attenuation mathematical model is introduced, with the specific formula as follows: In the formula, It is a dynamic association. To correct the relationship between resources, The interference sensitivity coefficient is preset to 2.5 in this embodiment to amplify the destructive effect under strong interference scenarios. The intensity of spatial interference.

[0101] When spatial interference intensity =0, meaning the wind and solar power stations are far apart and there is no shade or wind disturbance, in which case the denominator is 1 + 0 = 1. = This is physically self-consistent; in the absence of microscopic physical interference, the correlation of power plant output is entirely equivalent to the pure macroscopic synergy of meteorological resources. In data processing, if linear subtraction is used, when the interference is extremely strong, it might reduce the original positive correlation to a negative number. However, in physical reality, physical obstruction can at most completely disrupt the original meteorological synergy, causing it to approach 0, i.e., become uncorrelated, but it can never turn two power plants with the same wind and solar radiation into negatively mutually exclusive ones. No matter how intense the interference, the attenuation factor will only approach 0 infinitely, thus allowing the dynamic correlation to smoothly converge to 0, and it will never overstep the boundary and change the polarity of the correlation.

[0102] Within the interval [0,1), the square of a very small number becomes even smaller. This means that for extremely slight spatial interference, such as a very brief sweep of the wind turbine blade edge, the squared term will significantly weaken the penalty. It is assumed that such slight interference is insufficient to disrupt the overall macro-level meteorological coordination, thus ensuring the model's robustness in the face of minor environmental noise.

[0103] As a constant coefficient placed in the penalty term of the denominator, this gives the formula strong engineering scalability. In actual power grids, the density of power plant clusters varies in different regions. For flat, densely packed wind-solar hybrid power bases, even small physical disturbances can have a significant impact. In such cases, adjusting the coefficient can help mitigate the impact. This amplifies the destructive power of interference; however, for mountainous power stations with varying elevations and complex terrain, the interference effect may be weakened by the terrain, in which case the interference can be reduced. This parameter transforms what were once rigid mathematical formulas into flexible tools that can be adapted to different situations.

[0104] When spatial interference is extremely weak When the dynamic correlation approaches zero, it is almost equal to the theoretical meteorological synergy. However, as the intensity of spatial interference increases, the denominator rapidly increases due to the amplification effect of the squared term, and the original complementary correlation will be drastically weakened or even disintegrated. This design perfectly reproduces the natural phenomenon that local microscopic shading directly blocks macroscopic meteorological complementarity, and completely opens up the entire chain of deduction from macroscopic meteorological background to microscopic spatial environment and finally to output power, so that the output analysis results delivered to the power grid have unprecedented physical realism and environmental adaptability.

[0105] Step 5: Construct a dynamic correlation graph with power stations as nodes based on the dynamic correlation relationship, perform topological structure analysis on the dynamic correlation graph, extract the connectivity and eigenvector centrality of each node as topological feature parameters, identify key power station nodes based on the topological feature parameters and assess their fluctuation impact level, and finally generate power station output analysis results.

[0106] In this embodiment, each power station is defined as a node; Based on the dynamic association relationship, connection edges are established between nodes, and the dynamic association relationship is used as the edge weight of the connection edge; Power plant association networks at different time scales were constructed, and the power plant association networks at different time scales were combined to obtain a dynamic association map.

[0107] First, each wind power station and photovoltaic power station within the cluster is abstracted as a graph node. If a non-zero dynamic correlation is found between two power stations in the previous steps, an undirected connection edge is established between these two nodes, and the absolute value of the dynamic correlation is used as the edge weight. Three independent single-layer power station correlation networks are constructed for short-term, medium-term, and long-term time scales, respectively. Subsequently, to form a globally unified evaluation view, these three layers of networks are combined using a weighted fusion method. Different fusion weight parameters are preset for different time scales; for example, the short-term time scale emphasizes sudden disturbances with a preset weight of 0.5; the medium-term time scale has a weight of 0.3; and the long-term time scale reflects seasonal trends with a weight of 0.2. The edge weights of the same pair of nodes in the three layers of the network are weighted and summed, ultimately forming a global power station dynamic correlation graph by combining all nodes and their fused connection edges.

[0108] Power fluctuations in a real power grid are not transmitted point-to-point, but rather spread in a network. An abnormal output from one power plant can affect multiple surrounding power plants through the shifts in power flow and weather systems.

[0109] Traverse the dynamic association graph, for any node in the dynamic association graph, extract the edge weights of all connecting edges connected to that node, sum the edge weights to obtain the connectivity of the node.

[0110] A traversal search is performed on the generated dynamic association graph of power plants. Taking a specific node in the graph, such as node A of a large wind farm, as an example, all the lines directly connected to it are extracted, representing the dynamic association relationships with other power plants, and the edge weights of these lines are read. By using basic arithmetic addition, all these edge weights are added together, and the sum is defined as the node connectivity of node A.

[0111] Node connectivity directly reflects the range of a power plant's direct influence within the entire cluster. A power plant with extremely high node connectivity means that when its output fluctuates drastically, it will directly affect the output status of a large number of surrounding power plants. Extracting this parameter allows for the rapid delineation of the range of power plants first impacted during a localized fluctuation outbreak.

[0112] Construct a graph adjacency matrix with the edge weights of each node in the dynamic association graph as matrix elements; The maximum eigenvalue of the graph adjacency matrix is ​​solved using the power iteration algorithm, and the eigenvector corresponding to the maximum eigenvalue is calculated. Extract the element values ​​corresponding to the node from the feature vector as the feature vector centrality of the node.

[0113] The nodes are sorted in descending order based on the magnitude of their topological feature parameters. Nodes with topological feature parameters greater than a preset key threshold are identified as key power plant nodes. The connectivity of the key power plant nodes is used as the correlation propagation range, and the feature vector centrality is used as the correlation strength. The fluctuation impact level of the corresponding power plant is determined by combining the correlation propagation range and the correlation strength. Power output analysis results are generated based on the fluctuation impact level of each power station.

[0114] The two extracted topological feature parameters, node connectivity and eigenvector centrality, are normalized and then sorted in descending order. A preset key threshold is imported; for example, values ​​in the top 15% are set as the threshold limit, and nodes exceeding this limit are marked as key power plant nodes. Subsequently, the physical meanings are reassigned: the calculated node connectivity is mapped to the propagation range of the power plant, representing the physical surface directly affected; the eigenvector centrality is mapped to the correlation strength, representing its penetrating and destructive power on the overall stability of the power grid.

[0115] In graph theory and complex network theory, the dynamic correlation graph of a power plant typically exhibits the characteristics of a scale-free network, meaning that a few key nodes control the vast majority of the connections. By plotting the cumulative distribution function curve of the topological characteristic parameters of all nodes in the graph, the inflection point of this curve can be identified. Typically, the number of nodes at the head of the long tail accounts for approximately 0% to 20% of the total number of nodes. In this embodiment, 15% is preferably selected as the preset key threshold dividing line. This threshold can minimize the dimensionality of control and calculation for power grid dispatchers without overlooking core fluctuation sources.

[0116] The impact level of fluctuations is assessed based on a pre-defined two-dimensional evaluation matrix. For example, if a key node possesses both a very large propagation range and a very strong propagation intensity, it is assessed as a Level 1 fluctuation impact source; if it only has high intensity but a limited propagation range, it is assessed as a Level 2 fluctuation impact source; the rest are marked as regular nodes. Finally, the identity attributes, level classifications, and propagation paths of all power plants are packaged to generate and output a complete results report.

[0117] By rigorously mapping abstract topological features to specific propagation ranges and correlation strengths, and conducting structured level assessments, when extreme weather conditions cause enormous pressure on the power grid's peak shaving, the dispatch center does not need to blindly impose power rationing on hundreds of power stations across the entire network. Instead, it can directly retrieve the generated analysis results, accurately identify and prioritize the control of key power station nodes that are assessed as first-level sources of fluctuation impact, thereby securing the most robust and stable foundation for the entire new power system with minimal cost of wind and solar curtailment.

[0118] Table 1: Comparison of the Relevance of the Method in this Application and the Methods in the Prior Art Based on the measured comparison data, the wind power nodes (01 to 03) and photovoltaic nodes (01 to 03) in the table show extremely low spatial interference intensity and extremely low resource coordination probability, indicating that they do not have strong coupling in physical space and natural weather. Existing technology, relying solely on the apparent power output signal, yields a value as high as 0.84. The strong correlation misjudgment of 0.91 is highly likely due to a false impression caused by unified power grid rationing or equipment group control. This application corrects this by introducing resource coordination probability, decisively reducing these false appearances to a high score of 0.12. The value of 0.18 eliminates interference from non-natural factors, ensuring that the analysis results are based on objective meteorological drivers.

[0119] Observing the wind power nodes 04 to 06 and the photovoltaic nodes 04 to 06, these power plants exhibit extremely high spatial interference intensity and extremely high resource synergy probability, indicating a deep mutual influence between them in terms of microscopic flow field shading and macroscopic meteorological conditions. Current technology cannot perceive this complex physical interference, and a value of 0.35 is given. The blind judgment showed a low correlation of 0.48. This application forward-embeds optical interferometry and hydrodynamic constraints, successfully identifying these hidden high-risk nodes and restoring the true correlation to 0.85. 0.94. This effectively overcomes the shortcomings of existing analyses that suffer from feature distortion due to being detached from the microscopic physical operating environment.

[0120] For wind power-07, 08 and photovoltaic-07, 08, as well as wind power-09, 10 and photovoltaic-09, 10, the scoring of this application maintains a high degree of consistency with the prior art in these groups of natural states that can truly reflect the laws of complementarity or non-complementarity. This proves that the method of this application does not completely overturn the existing system, but rather uses a mechanistic model to identify and fill gaps while retaining the objective background of the pure data layer.

[0121] Please see Figure 3 The present invention also provides a power plant output multi-scale dynamic correlation analysis system, which is used to implement the above-mentioned power plant output multi-scale dynamic correlation analysis method, including: The data acquisition module is used to acquire the output power, equipment parameters and three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, and at the same time acquire the meteorological data of the area where each power station is located. Based on the three-dimensional spatial coordinates, the spatial interference relationship between each wind power station and each photovoltaic power station is determined. The feature calculation module is used to calculate the scintillation feature vector of the wind power station based on the spatial interference relationship, the equipment parameters of the wind power station and the meteorological data, and to calculate the roughness feature vector of the photovoltaic power station based on the spatial interference relationship, the equipment parameters of the photovoltaic power station and the meteorological data. The collaborative analysis module is used to construct wind resource characteristics and light resource characteristics based on the meteorological data, analyze the collaborative relationship between the wind resource characteristics and the light resource characteristics, and obtain the resource collaboration probability at each sampling time. The dynamic correlation module is used to perform multi-scale decomposition on the output power to obtain power components at different time scales; calculate the initial correlation relationship between power components of different power plants at the same time scale; then correct the resource synergy of the initial correlation relationship based on the resource synergy probability to obtain the resource-corrected correlation relationship; and combine the scintillation feature vector and the roughness feature vector to correct the resource-corrected correlation relationship to obtain the dynamic correlation relationship. The graph analysis module is used to construct a dynamic correlation graph with power stations as nodes based on the dynamic correlation relationship, perform topological structure analysis on the dynamic correlation graph, extract the connectivity and eigenvector centrality of each node as topological feature parameters, identify key power station nodes based on the topological feature parameters and assess their fluctuation impact level, and finally generate power station output analysis results.

[0122] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0123] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0124] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0125] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for multi-scale dynamic correlation analysis of power plant output, characterized in that, The specific steps include: Step 1: Obtain the output power, equipment parameters, and three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, and at the same time obtain the meteorological data of the area where each power station is located. Based on the three-dimensional spatial coordinates, determine the spatial interference relationship between each wind power station and each photovoltaic power station. Step 2: Calculate the scintillation feature vector of the wind power station based on the spatial interference relationship, the equipment parameters of the wind power station, and the meteorological data; calculate the roughness feature vector of the photovoltaic power station based on the spatial interference relationship, the equipment parameters of the photovoltaic power station, and the meteorological data. Step 3: Construct wind resource characteristics and light resource characteristics based on the meteorological data, analyze the synergistic relationship between the wind resource characteristics and the light resource characteristics, and obtain the resource synergy probability at each sampling time; Step 4: Perform multi-scale decomposition on the output power to obtain power components at different time scales; calculate the initial correlation between power components of different power plants at the same time scale, and then correct the resource synergy of the initial correlation based on the resource synergy probability to obtain the resource-corrected correlation; and combine the scintillation feature vector and the roughness feature vector to correct the resource-corrected correlation to obtain the dynamic correlation. Step 5: Construct a dynamic correlation graph with power stations as nodes based on the dynamic correlation relationship, perform topological structure analysis on the dynamic correlation graph, extract the connectivity and eigenvector centrality of each node as topological feature parameters, identify key power station nodes based on the topological feature parameters and assess their fluctuation impact level, and finally generate power station output analysis results.

2. The method for multi-scale dynamic correlation analysis of power plant output according to claim 1, characterized in that: The equipment parameters of a wind power station include hub height, rotor diameter, and rotor azimuth angle; The equipment parameters of a photovoltaic power station include: installation area, installation tilt angle, height above ground, and spacing between installations. Meteorological data include wind speed, wind direction, solar irradiance, solar altitude angle, azimuth angle, and ambient temperature; Based on the three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, calculate the relative distance, horizontal relative azimuth angle and vertical relative height difference between each wind power station and each photovoltaic power station; The spatial interference relationship between each wind power station and each photovoltaic power station is determined based on the relative distance, horizontal relative azimuth angle and vertical relative height difference. The spatial interference relationship is used to characterize the spatial geometric conditions for the shadowing effect and wind field disturbance effect between the wind power station and the photovoltaic power station. The output power and the meteorological data are both collected synchronously at the same preset sampling frequency, and the preset sampling times collected continuously together constitute the analysis time window.

3. The method for multi-scale dynamic correlation analysis of power plant output according to claim 2, characterized in that: The direction of sunlight is determined based on the solar altitude angle and azimuth angle. Using the three-dimensional spatial coordinates of the wind power station as the origin, the direction from the wind power station to the photovoltaic power station as the first coordinate axis of the local coordinate system, and the direction perpendicular to the ground as the third coordinate axis, a local relative coordinate system between the wind power station and the photovoltaic power station is constructed. In the local relative coordinate system, the hub height is offset along the third coordinate axis to obtain the hub center coordinates. The impeller sweep profile is constructed based on the hub center coordinates and the impeller diameter. The spatial attitude of the impeller sweep profile in the local relative coordinate system is determined based on the impeller azimuth angle. The impeller sweep profile is projected onto the photovoltaic power station mounting plane along the direction of sunlight to obtain a shadow projection area; Perform a geometric intersection operation on the shadow projection area and the installation area, extract the closed boundary of the intersection area, and calculate the area inside the closed boundary as the overlapping area; The ratio of the overlapping area to the installation area is used as the instantaneous occlusion ratio, and the instantaneous occlusion ratios at each sampling moment within the analysis time window are combined to obtain the flicker feature vector.

4. The method for multi-scale dynamic correlation analysis of power plant output according to claim 2, characterized in that: Calculate the sine function value of the installation tilt angle, multiply the installation area by the sine function value to obtain the effective windward cross-sectional area; Based on the installation area and the installation tilt angle, the horizontal surface projection area of ​​the photovoltaic power station is obtained through cosine projection calculation. The total surface area of ​​the gaps is calculated based on the arrangement spacing. The horizontal surface projection area and the total surface area of ​​the gaps are added together to obtain the spatial footprint of the corresponding photovoltaic power station. The ratio of the effective windward cross-sectional area to the spatial footprint is used as the terrain resistance factor. Extract the preset aerodynamic empirical coefficient, multiply the terrain drag factor by the aerodynamic empirical coefficient, and obtain the equivalent roughness length increase caused by each photovoltaic power station on the ground surface; Extract the absolute angle of the wind direction, subtract the absolute angle from the horizontal relative azimuth angle to obtain the wind direction angle, calculate the cosine function value of the wind direction angle, and multiply the cosine function value by itself to obtain the wind direction disturbance parameter; The roughness feature value is obtained by multiplying the increase in the equivalent roughness length, the wind direction disturbance parameter, and the wind speed together. The roughness feature values ​​corresponding to each sampling time within the analysis time window are combined in chronological order to obtain the roughness feature vector.

5. The method for multi-scale dynamic correlation analysis of power plant output according to claim 2, characterized in that: The logic for obtaining the resource collaboration probability is as follows: Wind resource characteristics are constructed based on the wind speed, wind direction, and ambient temperature; light resource characteristics are constructed based on the solar irradiance, solar altitude angle, and azimuth angle. The wind resource features and the solar resource features are input into a preset feature encoding network for encoding to obtain wind resource state feature vectors and solar resource state feature vectors, respectively. The distribution of the wind resource state feature vector and the solar resource state feature vector within the analysis time window is statistically analyzed to obtain their respective marginal probability distributions. The marginal probability distributions of wind resources and solar resources are input into a preset Copula function for fitting, thereby establishing a joint dependency model between wind resource status and solar resource status. The resource synergy probability of wind resource status and solar resource status at each sampling time within the analysis time window is calculated based on the joint dependency model.

6. The method for multi-scale dynamic correlation analysis of power plant output according to claim 1, characterized in that: The variational mode decomposition method is used to perform mode decomposition on the output power of each wind power station and each photovoltaic power station, and the output power is decomposed into multiple mode components with different center frequencies and a residual trend component. Calculate the center frequency corresponding to each modal component, and sort the modal components according to the magnitude of the center frequency; The modal components with a center frequency higher than a first preset threshold are identified as short-time-scale power components. The modal components whose center frequency is between the first preset threshold and the second preset threshold are defined as medium-time scale power components. The modal components with center frequencies below the second preset threshold and the residual trend components are identified as long-time scale power components. The first preset threshold is greater than the second preset threshold; The corresponding modal components at the same time scale are reconstructed to obtain short-time scale power sequences, medium-time scale power sequences and long-time scale power sequences. For the power sequences of different power plants at the same time scale, calculate the Pearson correlation coefficient between them, and use the Pearson correlation coefficient as the initial correlation at that time scale.

7. The method for multi-scale dynamic correlation analysis of power plant output according to claim 5, characterized in that: The values ​​of the scintillation feature vector and the roughness feature vector at each sampling time within the analysis time window are extracted respectively. The scintillation feature vector and the roughness feature vector are resampled according to the same time scale as the power component. The spatial interference intensity between the corresponding power station pairs is calculated based on the resampled scintillation feature value and roughness feature value. Extract the resource coordination probability corresponding to each sampling time, and use the resource coordination probability to correct the initial correlation at the corresponding time scale to obtain the resource correction correlation. The resource modification correlation is attenuated and corrected based on the spatial interference intensity to obtain the dynamic correlation corresponding to each time scale.

8. The method for multi-scale dynamic correlation analysis of power plant output according to claim 1, characterized in that: Each power station was designated as a node; Based on the dynamic association relationship, connection edges are established between nodes, and the dynamic association relationship is used as the edge weight of the connection edge; Power plant association networks at different time scales were constructed, and the power plant association networks at different time scales were combined to obtain a dynamic association map; Traverse the dynamic association graph, for any node in the dynamic association graph, extract the edge weights of all connecting edges connected to that node, sum the edge weights to obtain the connectivity of the node; Construct a graph adjacency matrix with the edge weights of each node in the dynamic association graph as matrix elements; The maximum eigenvalue of the graph adjacency matrix is ​​solved using the power iteration algorithm, and the eigenvector corresponding to the maximum eigenvalue is calculated. Extract the element values ​​corresponding to the node from the feature vector as the feature vector centrality of the node; The nodes are sorted in descending order based on the magnitude of their topological feature parameters. Nodes with topological feature parameters greater than a preset key threshold are identified as key power plant nodes. The connectivity of the key power plant nodes is used as the correlation propagation range, and the feature vector centrality is used as the correlation strength. The fluctuation impact level of the corresponding power plant is determined by combining the correlation propagation range and the correlation strength. Power output analysis results are generated based on the fluctuation impact level of each power station.

9. A multi-scale dynamic correlation analysis system for power plant output, characterized in that: The power plant output multi-scale dynamic correlation analysis system is used to implement the power plant output multi-scale dynamic correlation analysis method according to any one of claims 1-8, including: The data acquisition module is used to acquire the output power, equipment parameters and three-dimensional spatial coordinates of each wind power station and each photovoltaic power station, and at the same time acquire the meteorological data of the area where each power station is located. Based on the three-dimensional spatial coordinates, the spatial interference relationship between each wind power station and each photovoltaic power station is determined. The feature calculation module is used to calculate the scintillation feature vector of the wind power station based on the spatial interference relationship, the equipment parameters of the wind power station and the meteorological data, and to calculate the roughness feature vector of the photovoltaic power station based on the spatial interference relationship, the equipment parameters of the photovoltaic power station and the meteorological data. The collaborative analysis module is used to construct wind resource characteristics and light resource characteristics based on the meteorological data, analyze the collaborative relationship between the wind resource characteristics and the light resource characteristics, and obtain the resource collaboration probability at each sampling time. The dynamic correlation module is used to perform multi-scale decomposition on the output power to obtain power components at different time scales; calculate the initial correlation relationship between power components of different power plants at the same time scale; then correct the resource synergy of the initial correlation relationship based on the resource synergy probability to obtain the resource-corrected correlation relationship; and combine the scintillation feature vector and the roughness feature vector to correct the resource-corrected correlation relationship to obtain the dynamic correlation relationship. The graph analysis module is used to construct a dynamic correlation graph with power stations as nodes based on the dynamic correlation relationship, perform topological structure analysis on the dynamic correlation graph, extract the connectivity and eigenvector centrality of each node as topological feature parameters, identify key power station nodes based on the topological feature parameters and assess their fluctuation impact level, and finally generate power station output analysis results.

Images