A Big Data-Based Load Forecasting Method for Photovoltaic Power Generation Systems

Abstract

This invention discloses a load forecasting method for photovoltaic power generation systems based on big data, relating to the field of photovoltaic power generation technology. The method includes: identifying multiple power generation units and their interconnection network; constructing constraints and multiple power generation databases; collecting first meteorological forecast information for a preset power supply area in the first forecast time zone, and inputting it into the multiple power generation databases for analysis to generate initial power generation data for multiple sub-units; analyzing the fusion loss after the power is processed by the inverter from the power generation units using the initial power generation data of the multiple sub-units, and constructing a forecast load for the photovoltaic power generation system that can be input into the user-side power grid in the first forecast time zone. This invention solves the technical problem of insufficient load forecasting accuracy caused by the lack of processing of equipment aging, interconnection relationships, and multi-source data fusion analysis in existing technologies, achieving the technical effect of improving the adaptability and accuracy of forecast results to the user-side power grid.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, and specifically to a method for load forecasting of photovoltaic power generation systems based on big data. Background Technology

[0002] Existing load forecasting methods for photovoltaic (PV) power generation systems primarily rely on static correlation modeling between meteorological parameters and power generation. They typically use single-dimensional meteorological data as input, neglecting the performance degradation of PV modules over long-term operation and the impact of the interconnection structure between various power generation units on energy transmission efficiency. Furthermore, traditional methods lack in-depth utilization of historical operating data, failing to construct multi-dimensional data models reflecting equipment type, service life, and meteorological cycle variations. This makes them ill-suited to the complex and ever-changing operating environments in real-world applications, resulting in low forecast accuracy and failing to meet the demands of refined scheduling and grid load management. Summary of the Invention

[0003] This application provides a big data-based load forecasting method for photovoltaic power generation systems, which addresses the technical problem of insufficient load forecasting accuracy caused by the lack of processing of equipment aging, connection relationships, and multi-source data fusion analysis in existing technologies.

[0004] In view of the above problems, this application provides a method for load forecasting of photovoltaic power generation systems based on big data.

[0005] This application provides a method for load forecasting of photovoltaic power generation systems based on big data, the method comprising:

[0006] Multiple power generation units within a photovoltaic power generation system within a preset power supply area are identified, along with their interconnection network. Constraints are constructed by collecting the photovoltaic panel models and service times of the multiple power generation units, as well as the meteorological cycle characteristics of the preset power supply area. Multiple power generation databases are built through historical data retrieval. First meteorological forecast information for the preset power supply area in the first forecast time zone is collected and input into the multiple power generation databases for analysis, generating initial power generation data for multiple sub-units. Based on the interconnection network, the initial power generation data of the multiple sub-units is analyzed to determine the fusion loss after the electrical energy is transmitted from the power generation units to the inverter for processing. The predicted load of the photovoltaic power generation system that can be input into the user-side power grid in the first forecast time zone is then constructed.

[0007] In a possible implementation, determining multiple power generation units within a photovoltaic power generation system in a preset power supply area, and the connection network of the multiple power generation units, includes: collecting data from each inverter processing node within the photovoltaic power generation system; using each inverter processing node as a partitioning node to partition the multiple power generation units and construct each power generation unit array; and constructing the connection network using each power generation unit array and each inverter processing node.

[0008] In a possible implementation, each power generation unit array includes several power generation units connected in series or in parallel, and the initial electrical energy of the several power generation units flows into the same inverter processing node for processing.

[0009] In a possible implementation, constraints are constructed by collecting the photovoltaic panel models and service times of the multiple power generation units and the meteorological cycle characteristics of the preset power supply area. Multiple power generation databases are then constructed through historical data retrieval, including: constructing a first data constraint based on the photovoltaic panel models and service times of the multiple power generation units; retrieving multiple historical power generation datasets based on big data technology, wherein the historical power generation data includes historical meteorological information and corresponding historical power generation data at any historical moment; constructing a second data constraint based on the photovoltaic panel models of the multiple power generation units and the meteorological cycle characteristics of the preset power supply area; retrieving multiple historical power generation data time series that change sequentially with service time; performing power generation quality degradation analysis on the multiple power generation units using the multiple historical power generation data time series to construct multiple power generation quality degradation curves; and establishing a mapping connection between the multiple power generation quality degradation curves and the multiple historical power generation datasets to obtain the multiple power generation databases.

[0010] In a possible implementation, the power generation quality degradation analysis of the multiple power generation units is performed using the time series of the multiple historical power generation data, and multiple power generation quality degradation curves are constructed. This includes: decomposing the multiple historical power generation data time series according to the power generation parameter type to generate multiple historical power generation parameter time series combinations; and based on the multiple historical power generation parameter time series combinations, fitting the relationship between the degree of degradation of power generation parameters and service time to generate the multiple power generation quality degradation curves.

[0011] In a possible implementation, based on the connection network, the initial power generation data of the multiple sub-units is analyzed for fusion losses after the power is transmitted from the power generation unit to the inverter for processing. This process constructs the predicted load that the photovoltaic power generation system can input to the user-side power grid in the first prediction time zone. This includes: based on the connection network, superimposing the initial power generation data of multiple sub-units in series and parallel states of each power generation unit array and analyzing power loss to generate each first output power generation data corresponding to each power generation unit array; performing power processing simulation on each first output power generation data at each inverter processing node based on the transmission parameter requirements of the user-side power grid to generate each second output power generation data; and performing power loss analysis on the access circuit from each inverter processing node to the user-side power grid, compensating for losses in each second output power generation data, and generating the predicted load.

[0012] In a possible implementation, based on the connection network, the initial power generation data of multiple sub-units in series and parallel states of each power generation unit array are superimposed and power loss analysis is performed to generate the first output power generation data corresponding to each power generation unit array. This includes: constructing a series-parallel power parameter superposition mechanism; collecting the connection circuit structure information of each power generation unit array; retrieving historical power transmission loss data based on the connection circuit structure information; analyzing the relationship between power loss and circuit block loss using the historical power transmission loss data; constructing a power loss compensation mechanism; and combining the series-parallel power parameter superposition mechanism and the power loss compensation mechanism to superimpose and analyze the initial power generation data of the multiple sub-units to generate the first output power generation data.

[0013] In a possible implementation, each inverter processing node performs power processing simulation on each first output power generation data based on the power transmission parameter requirements of the user-side power grid to generate each second output power generation data. This includes: performing twin simulation on each inverter processing node to establish each twin inverter, wherein the twin simulation includes at least twin simulation of the DC input terminal, current control circuit, filter circuit, and AC output terminal; using the power transmission parameter requirements as simulation output constraints, each twin inverter performs simulation operation with each first output power generation data as simulation input to generate each second output power generation data.

[0014] In possible implementations, the transmission parameter requirements include at least the input voltage, frequency, power factor, and harmonic constraints of the user-side power grid.

[0015] In a possible implementation, after constructing the photovoltaic power generation system to input the predicted load of the user-side power grid in the first predicted time zone, the method further includes: reading the power supply planning information of the preset power supply area in the first predicted time zone; determining whether the predicted load exceeds the power supply planning information; if so, obtaining the excess load and storing it through an energy storage device; if not, determining the power supply of the main distribution network connected to the user-side power grid based on the excess load.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0017] This application identifies multiple power generation units within a photovoltaic power generation system within a preset power supply area, and the connection network of these multiple power generation units. It collects the photovoltaic panel models and service times of the multiple power generation units, as well as the meteorological cycle characteristics of the preset power supply area, to construct constraints. Multiple power generation databases are built through historical data retrieval. First meteorological forecast information for the preset power supply area in a first forecast time zone is collected and input into the multiple power generation databases for analysis, generating initial power generation data for multiple sub-units. Based on the connection network, the initial power generation data of the multiple sub-units is analyzed to determine the fusion loss after the power is transmitted from the power generation units to the inverter for processing, thus constructing a predicted load for the photovoltaic power generation system that can be input into the user-side power grid in the first forecast time zone. This invention solves the technical problem of insufficient load forecast accuracy caused by the lack of processing of equipment aging, connection relationships, and multi-source data fusion analysis in existing technologies. By constructing a power generation database containing multi-dimensional constraints such as photovoltaic panel models, service times, and meteorological cycles, and combining this with the analysis of power fusion losses using the connection network, the technical effect of improving the adaptability and accuracy of the forecast results to the user-side power grid is achieved. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A schematic flowchart of a photovoltaic power generation system load forecasting method based on big data is provided for embodiments of this application;

[0020] Figure 2 This is a schematic diagram illustrating the process of constructing multiple power generation databases in a big data-based photovoltaic power generation system load forecasting method provided in this application embodiment. Detailed Implementation

[0021] This application provides a big data-based load forecasting method for photovoltaic power generation systems. This method addresses the technical problem of insufficient load forecasting accuracy caused by the lack of processing of equipment aging, connection relationships, and multi-source data fusion analysis in existing technologies. By constructing a power generation database containing multi-dimensional constraints such as photovoltaic panel model, service time, and weather cycle, and combining it with connection relationship network analysis of power fusion loss, the method achieves the technical effect of improving the adaptability and accuracy of the forecast results to the user-side power grid.

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0023] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0024] Examples, such as Figure 1 As shown, this application provides a method for load forecasting of photovoltaic power generation systems based on big data, the method comprising:

[0025] Step S100: Determine multiple power generation units within the photovoltaic power generation system in the preset power supply area, and the connection network of the multiple power generation units.

[0026] In this embodiment, a structural analysis of the photovoltaic power generation system within a preset power supply area is first performed to identify multiple power generation units. A power generation unit refers to a basic power generation module constituting a photovoltaic power generation array, typically composed of several photovoltaic panels, and possesses independent power generation capabilities.

[0027] Next, a network of connections between the various power generation units is constructed. Specifically, first, the data of each inverter processing node within the photovoltaic power generation system is collected; then, using the inverter processing node as a partitioning node, the multiple power generation units connected to it are divided into independent power generation unit arrays, with each array representing a group of power generation modules with the same electrical affiliation; finally, based on each power generation unit array and the inverter processing node, a network of connections between the multiple power generation units is constructed.

[0028] Furthermore, the method provided in the application embodiment, which determines multiple power generation units within a photovoltaic power generation system in a preset power supply area, and the connection network of the multiple power generation units, further includes:

[0029] Collect data from each inverter processing node within the photovoltaic power generation system; use each inverter processing node as a partitioning node to partition the multiple power generation units and construct each power generation unit array; construct the connection network using each power generation unit array and each inverter processing node.

[0030] Furthermore, the method provided in the application embodiments also includes:

[0031] Each power generation unit array includes several power generation units connected in series or in parallel, and the initial electrical energy of the several power generation units flows into the same inverter processing node for processing.

[0032] In this embodiment of the application, multiple inverter processing nodes in the photovoltaic power generation system within the preset power supply area are first determined. This process adopts the method of system structure identification and equipment information analysis. Combined with electrical wiring diagrams, deployment drawings and operation configuration data in the SCADA platform, the inverter locations, connection channels and their electrical relationships with the power generation unit in the photovoltaic system are identified, and each inverter processing node is determined.

[0033] After identifying the inverter processing nodes, each inverter processing node is further used as a partitioning node to divide the multiple power generation units into partitions, constructing various power generation unit arrays. Specifically, by reading wiring numbers, bus methods, and cable wiring relationships, multiple power generation units connected to the same inverter are identified and grouped into a single power generation unit array. Each array consists of several power generation units connected in series or parallel; series connection increases output voltage, while parallel connection enhances output current capability. Through structural identification methods and analysis of actual operating parameters, the internal connection methods of multiple power generation unit arrays are further clarified, and it is confirmed that the initial electrical energy of several power generation units flows uniformly into the same inverter processing node for centralized processing.

[0034] After completing the above structure identification and array division, the connection network is finally constructed using each generator unit array and each inverter processing node. Specifically, using a graph structure modeling method, the generator unit array and inverter processing nodes are treated as network nodes, with their physical connections and electrical paths as modeling edges. This establishes a complete structural model of the photovoltaic power generation system, showing how electrical energy originates from the generator units, converges through the array, and reaches the inverter processing nodes. The final result is the connection network of multiple generator units.

[0035] Step S200: Collect the photovoltaic panel models and service times of the multiple power generation units and the meteorological cycle characteristics of the preset power supply area to construct constraints, and construct multiple power generation databases through historical data retrieval.

[0036] In this embodiment, the equipment configuration files and installation records of the power station are first reviewed to determine the type of photovoltaic panel corresponding to each power generation unit and the installation and commissioning time, thereby calculating the actual service life of each photovoltaic panel. Subsequently, a periodic meteorological analysis method is used to collect the meteorological cycle characteristics of the preset power supply area. By accessing historical meteorological databases (such as the National Meteorological Center or regional environmental monitoring stations), key meteorological variables such as light intensity, ambient temperature, and wind speed over long periods are obtained, and periodic modeling tools are used to extract trends and analyze periodic patterns in the data. This process employs methods such as Fourier transform to identify typical cyclical structures such as daily, weekly, and yearly patterns to characterize the environmental cycle background in which the photovoltaic system operates, thereby obtaining the meteorological cycle characteristics of the preset power supply area.

[0037] After completing the multi-source information collection, the first data constraint is constructed based on the photovoltaic panel model and service time of multiple power generation units. Using big data retrieval methods, data records that meet the criteria of identical panel models and similar service times are selected from the system's historical database. This dataset contains meteorological information and corresponding power generation data at any historical point in time. Next, the second data constraint is constructed based on the photovoltaic panel model and the cyclical characteristics of meteorological periods. This selects data record sequences covering different meteorological stages and varying with service time, forming a historical power generation data time series with a time dimension.

[0038] After acquiring historical time-series data, a power generation degradation analysis method was employed. Based on multiple historical power generation data series, the changing trend of power generation capacity of each power generation unit over service time was modeled. Multiple power generation quality degradation curves were constructed using data fitting methods. Finally, these multiple power generation quality degradation curves were mapped and connected with multiple historical power generation datasets extracted in the first stage. This associates the time-series degradation model with actual power generation sample data, establishing a complete relationship between component characteristics, degradation model, environmental variables, and power generation performance, and constructing multiple power generation databases.

[0039] Further, such as Figure 2 As shown, the method provided in the application embodiment includes collecting the photovoltaic panel models and service times of the multiple power generation units and the meteorological cycle characteristics of the preset power supply area to construct constraints, and constructing multiple power generation databases through historical data retrieval. It also includes:

[0040] A first data constraint is constructed based on the photovoltaic panel model and service time of the multiple power generation units. Based on big data technology, multiple historical power generation datasets are retrieved, wherein the historical power generation data includes historical meteorological information and corresponding historical power generation data at any historical moment. A second data constraint is constructed based on the photovoltaic panel model of the multiple power generation units and the meteorological cycle characteristics of the preset power supply area. Multiple historical power generation data time series that change sequentially with service time are retrieved. Power generation quality degradation analysis is performed on the multiple power generation units based on the multiple historical power generation data time series to construct multiple power generation quality degradation curves. After establishing a mapping connection between the multiple power generation quality degradation curves and the multiple historical power generation datasets, the multiple power generation databases are obtained.

[0041] In this embodiment, an attribute-based filtering method is first employed to construct the first data constraint using the photovoltaic panel models and service times of multiple power generation units. This method, based on the component information table and operation record table in a big data platform, uses photovoltaic panel models (such as PERC) and service life ranges as filtering conditions to select data entries from the historical database that match the characteristics of the current power generation unit. Through this operation, multiple historical power generation datasets are retrieved. Each dataset contains photovoltaic module power generation output data at a specific historical moment and corresponding historical meteorological information, such as the light intensity, temperature, and wind speed at that moment.

[0042] Subsequently, a meteorological cycle matching method was employed to construct a second data constraint based on the photovoltaic panel models of multiple power generation units and the cyclical characteristics of the meteorological cycles in a preset power supply area. This method, based on time series analysis, utilizes STL decomposition (Seasonal-Trend-Residual Decomposition) to perform periodic modeling of historical meteorological data for the target area, extracting stable daily and annual cycle components. These components are then combined with the photovoltaic panel models to define the search criteria, ensuring that the data are consistent not only at the component level but also exhibit periodic similarity in meteorological patterns. Through this approach, multiple historical power generation data series that change sequentially with service time were retrieved.

[0043] The following case study employs a multi-time-series degradation analysis method, using multiple historical power generation data time series to analyze the power generation quality degradation of the multiple power generation units. This analysis first decomposes the multiple historical power generation data time series according to the type of power generation parameters, breaking down the power output into independent parameter change time series combinations based on key parameters (such as output power, voltage, current, and temperature coefficient). Then, based on multiple historical power generation parameter time series combinations, the relationship between the degradation degree of power generation parameters and service time is fitted. Common fitting methods such as linear regression are used to obtain degradation trend models for each parameter, ultimately generating multiple power generation quality degradation curves.

[0044] Finally, a data model mapping method is used to establish a mapping connection between multiple power generation quality degradation curves and multiple historical power generation datasets. Specifically, in each historical power generation data record, a degradation curve function corresponding to the component model and service time is embedded, forming a data structure with triple information binding of time dimension, environmental background and degradation model.

[0045] Furthermore, in the method provided in the application embodiments, the method further includes performing power generation quality degradation analysis on the multiple power generation units using the time sequence of the multiple historical power generation data to construct multiple power generation quality degradation curves, and also includes:

[0046] The time series of the multiple historical power generation data are decomposed according to the type of power generation parameter to generate multiple time series combinations of historical power generation parameters; based on the multiple time series combinations of historical power generation parameters, the relationship between the degree of degradation of power generation parameters and service time is fitted to generate multiple power generation quality degradation curves.

[0047] In this embodiment, a parameter classification and extraction method is first employed to decompose multiple historical power generation data time series according to the type of power generation parameter. This method extracts key power generation parameters such as output power, voltage, current, and temperature from the historical time series data by parsing the parameter fields in the historical data records, and reassembles them into independent time series data based on timestamps. Each parameter forms a continuous sequence of numerical changes, thus obtaining multiple combinations of historical power generation parameter time series.

[0048] Subsequently, a least squares regression fitting method was employed to fit the relationship between the degradation degree of power generation parameters and their service life, based on multiple historical power generation parameter time series combinations. Service life was obtained by subtracting the installation time of the module from the time of each data record, serving as the input independent variable; the time series values ​​corresponding to each parameter served as the input dependent variable, establishing a functional relationship between parameter values ​​and service life. The least squares method was used to fit the data points, and an appropriate mathematical model (such as a linear model or exponential model) was selected to express the relationship. The fitting effect was evaluated through residual testing, yielding a degradation trend function for each power generation parameter. This function accurately characterizes the change trend of the parameter during the module aging process.

[0049] Finally, a trend curve construction method is employed to generate multiple power generation quality degradation curves based on the fitted parameter degradation function. Specifically, according to the mathematical expression of the degradation function, the parameter values ​​corresponding to each time point are calculated within a set service time interval (e.g., 0 to 25 years), and a curve is plotted with time on the horizontal axis and the fitted value on the vertical axis. Each curve clearly shows the trajectory of a certain power generation parameter gradually degrading with the service time, possessing good visualization and quantitative characteristics. Through this process, multiple power generation quality degradation curves are ultimately generated.

[0050] Step S300: Collect the first meteorological forecast information of the preset power supply area in the first forecast time zone, and input it into the multiple power generation databases for analysis to generate initial power generation data for multiple sub-units.

[0051] In this embodiment, a numerical weather prediction interface is first used to collect the first meteorological forecast information for a preset power supply area in the first prediction time zone. By accessing a meteorological data service platform (such as ECMWF or GFS), the API interface is called to obtain hourly meteorological data covering the target area. The data includes key meteorological variables such as light intensity, ambient temperature, humidity, and wind speed, thus obtaining the first meteorological forecast information for the first prediction time zone.

[0052] Next, a similar meteorological retrieval method is used to input the first meteorological forecast information for the first forecast time zone into the power generation database for matching analysis. Using indicators such as Euclidean distance, the meteorological parameters of the forecast time zone are compared with typical meteorological scenarios recorded in the historical database to retrieve the historical dataset that is closest to the forecast conditions. Subsequently, a degradation model calibration method is used to correct the retrieved historical power generation data by combining it with the corresponding power generation quality degradation curve. The degradation curve is fitted based on the photovoltaic module model and service time, reflecting the performance decline trend of the module due to aging during actual operation. By substituting the historical power generation output data into the corresponding degradation function, the actual power generation capacity of the module at the current moment under the forecast meteorological conditions is calibrated to obtain key parameters such as power, voltage, and current.

[0053] Finally, a structured output generation method is used to generate initial power generation data for multiple sub-units based on the calibrated results. The initial power generation data for each sub-unit includes electrical parameters such as output power, voltage, and current.

[0054] Step S400: Based on the connection network, analyze the initial power generation data of the multiple sub-units to determine the fusion loss after the power is transmitted from the power generation unit to the inverter for processing, and construct the predicted load of the photovoltaic power generation system that can be input to the user-side power grid in the first prediction time zone.

[0055] In this embodiment, based on the connection relationship network, the initial power generation data of multiple sub-units are fused and analyzed. First, the series and parallel connection methods of each power generation unit array are identified, and the initial power generation data is superimposed with electrical energy and the loss during transmission is calculated to obtain the first output power generation data corresponding to each power generation unit array. Then, at the inverter processing node, according to the power transmission parameter requirements of the user-side power grid, the first output power generation data is simulated for power processing to obtain the second output power generation data. Finally, combined with the access circuit from the inverter processing node to the user-side power grid, the second output power generation data is compensated for power loss to construct the predicted load of the photovoltaic power generation system that can be input to the user-side power grid in the first predicted time zone.

[0056] Furthermore, in the method provided in the application embodiment, based on the connection relationship network, the analysis of the initial power generation data of the multiple sub-units to determine the fusion loss after the power is transmitted from the power generation unit to the inverter for processing, and the construction of the predicted load of the photovoltaic power generation system that can be input to the user-side power grid in the first prediction time zone, further includes:

[0057] Based on the aforementioned connection network, the initial power generation data of multiple sub-units in series and parallel states of each power generation unit array are superimposed and power loss analysis is performed to generate each first output power generation data corresponding to each power generation unit array. At each inverter processing node, based on the transmission parameter requirements of the user-side power grid, power processing simulation is performed on each first output power generation data to generate each second output power generation data. Power loss analysis is performed on the access circuit from each inverter processing node to the user-side power grid, and loss compensation is applied to each second output power generation data to generate the predicted load.

[0058] In this embodiment, a series-parallel power parameter superposition method is first employed to analyze the structure of each power generation unit array based on a connection relationship network. Specifically, the series or parallel connection status of each power generation unit in the array is identified, and the power, voltage, current, and other parameters in the initial power generation data of multiple sub-units are combined according to the principles of voltage superposition (series) and current superposition (parallel). To further correct energy attenuation during transmission, a circuit structure extraction and matching method is used to collect connection circuit information such as cable paths, resistance structures, and wiring methods of each array. Based on this information, historical power transmission loss data is retrieved, a power loss compensation mechanism is constructed, the superimposed parameters are corrected, and the first output power generation data corresponding to each power generation unit array is generated.

[0059] Subsequently, an inverter twin simulation method was employed to simulate the actual power conversion process at each inverter processing node. Specifically, a twin inverter model was established, including DC input, current control, filtering, and AC output structures, to simulate its response to received power. Using the first output power generation data as the simulation input, and combining the user-side grid transmission parameter requirements (such as voltage level, power factor, frequency range, etc.) as the simulation output constraints, the simulation process was run to calculate the actual output capacity of the inverter under the current conditions, obtaining the second output power generation data corresponding to each inverter processing node.

[0060] Next, a precise impedance modeling method for the access path is used to analyze the losses in the access circuit between the inverter output and the user's grid connection point. This method constructs a physical equivalent circuit model of the inverter output path, clarifies the equivalent series resistance and inductance of the cables, and establishes equivalent impedance models for single-phase and three-phase circuits based on the physical length, wiring method, and transformer parameters of each path. Combining the current amplitude information from the second output power generation data, an active power loss calculation method is used to perform node-level loss calculations for each path segment, and the power attenuation and voltage sag values ​​of the entire access path are accumulated. This step, through deterministic physical calculations, fully reflects the electrical performance degradation process from the inverter to the grid.

[0061] Finally, a path-correction-based power fusion method is used to compensate and integrate the output results. The path loss data calculated in the previous step is used to correct the power and voltage parameters in the second output power generation data. By directly subtracting the loss value from the AC output power, the net available power of each inverter node before grid connection is obtained. Subsequently, according to the multi-path power fusion rules of the system access point, the corrected output values ​​of multiple inverters are summed and aggregated. Combined with grid connection point power constraints, phase voltage balance conditions, and distribution network capacity constraints, the final predicted load of the photovoltaic power generation system that can be received by the user-side grid in the first prediction time zone is generated.

[0062] Furthermore, the method provided in the application embodiment, which involves superimposing initial power generation data of multiple sub-units in series and parallel states of each power generation unit array and analyzing power loss to generate each first output power generation data corresponding to each power generation unit array, further includes:

[0063] A series-parallel power parameter superposition mechanism is constructed; the connection circuit structure information of each of the generating unit arrays is collected; historical power transmission loss data is retrieved based on the connection circuit structure information, and the relationship between power loss and circuit block loss is analyzed using the historical power transmission loss data to construct a power loss compensation mechanism; the initial power generation data of the multiple sub-units are superimposed and loss analyzed by combining the series-parallel power parameter superposition mechanism and the power loss compensation mechanism to generate the first output power generation data of each sub-unit.

[0064] In this embodiment, a series-parallel power parameter superposition mechanism is first constructed. Based on the circuit characteristics of existing photovoltaic systems, this mechanism identifies the connection structure between each power generation unit, determining whether it is in series, parallel, or mixed connection, and integrates the initial power generation data of the sub-units according to electrical principles. Specifically, in a series structure, the current is consistent and the voltage is accumulated; in a parallel structure, the voltage is consistent and the current is accumulated. Through this mechanism, the voltage, current, and power parameters from the initial power generation data of multiple sub-units are logically superimposed according to the array topology, forming the overall power output parameters of each power generation unit array under ideal conditions.

[0065] Subsequently, by consulting electrical installation drawings and system wiring lists, the connection circuit structure information of each power generation unit array was collected, including specific physical parameters such as the wiring path between the array and the inverter, cable material (e.g., copper, aluminum), cable cross-sectional area, single-segment resistance of the cable, total length, connection method (e.g., overhead, underground), and number of connectors. Through this process, the structure information of each connection circuit was obtained.

[0066] Next, based on the collected connection circuit structure information, historical power transmission loss data was further retrieved. By accessing the operation logs in the photovoltaic power plant's historical operation and maintenance platform, historical records with cable path parameters similar to the current structure were selected, and the current, voltage, actual power, and theoretical output data contained therein were extracted. Combining the difference between the ideal output and the measured output, the line loss level under specific operating conditions was compared, and the relationship between power loss and loss between circuit blocks was analyzed accordingly. This analysis clarifies the correspondence between different wiring structures, current intensities, and energy attenuation, providing quantitative support for subsequent correction processes.

[0067] Based on the above analysis, a power loss compensation mechanism was constructed. This mechanism is based on commonly used engineering physical calculation methods. It calculates the active power loss of the entire cable path by multiplying the array output current by the path resistance. Simultaneously, it accumulates the voltage drop in the path segment by segment based on the voltage drop formula, forming a set of power and voltage correction rules for each path. This compensation mechanism feeds back the energy attenuation caused by the path to the array output parameters in real time, ensuring that subsequent calculation results are closer to the physical reality.

[0068] Finally, by combining the series and parallel power parameter superposition mechanism and the power loss compensation mechanism, the initial power generation data of each sub-unit is uniformly processed and loss analyzed. Through the integration of steps such as parameter superposition, structure identification, and path compensation, the comprehensive capability assessment of the power generation unit array under the real electrical configuration is completed, and finally, the first output power generation data of each unit is generated.

[0069] Furthermore, in the method provided in the application embodiment, in which each inverter processing node performs power processing simulation on each first output power generation data based on the power transmission parameter requirements of the user-side power grid to generate each second output power generation data, the method further includes:

[0070] A twin simulation is performed on each inverter processing node to establish each twin inverter. The twin simulation includes at least twin simulations of the DC input terminal, current control circuit, filter circuit, and AC output terminal. Using the power transmission parameter requirements as simulation output constraints, each twin inverter is used as the simulation input to perform simulation operation and generate each second output power generation data.

[0071] Furthermore, the method provided in the application embodiments also includes:

[0072] The required transmission parameters include at least the input voltage, frequency, power factor, and harmonic constraints of the user-side power grid.

[0073] In this embodiment, to recreate the power conversion process of the inverter in a photovoltaic power generation system, twin simulations are performed on each inverter processing node. This involves establishing a simulation entity with a functional structure consistent with the actual inverter through digital modeling. The simulation employs a structural equivalence restoration method to ensure that the behavior of each inverter processing node in the system can be fully mapped in the simulation environment, thereby supporting accurate prediction of subsequent power output capabilities. During the simulation, twin inverters are established, each containing key structural modules to achieve electrical consistency restoration of input and output behavior. This structure is divided into multiple functional sub-modules, covering the entire process from input acquisition, control calculation, filtering and shaping to output feedback, ensuring that the simulation data has response characteristics highly consistent with the actual inverter operating state. Specifically, the twin simulation includes at least twin simulations of the DC input terminal, current control circuit, filter circuit, and AC output terminal. The DC input terminal receives the first output power generation data from the photovoltaic array side, including parameters such as voltage, current, and power. The current control circuit uses a dual closed-loop PI control method to decouple voltage and current, achieving stable tracking of grid synchronization commands. The filter circuit is based on an LCL topology and is used to suppress high-frequency harmonics generated during the inverter process, improving the quality of output power. The AC output terminal simulates the electrical interface of the inverter's grid connection port, used to output modulated and shaped three-phase AC power to meet grid connection standards.

[0074] After establishing the twin inverter structure, the transmission parameter requirements are used as simulation output constraints to set the operating boundary conditions for each simulated inverter. These transmission parameter requirements include the input voltage, frequency, power factor, and harmonic constraints of the user-side power grid. The voltage and frequency are set to the national grid standard values, the power factor constrains the reactive power characteristics of the inverter output, and the harmonic constraints ensure that the output power quality does not affect grid stability. These parameters are incorporated into the control module to dynamically adjust the PWM modulation strategy and output waveform characteristics, ensuring that the simulation process strictly adheres to grid connection rules.

[0075] Subsequently, the dynamic simulation process is initiated by using the first output power generation data of each twin inverter as the simulation input. After receiving the output power, current, and voltage parameters of the corresponding array, each inverter simulation unit starts its internal control calculation and power conversion process, responding step by step from input to output. During the simulation, the controller dynamically adjusts the output amplitude and phase, the power devices simulate the actual switching process, and the filter circuit synchronously follows, gradually generating the AC signal at the output terminal.

[0076] Through the above process, various second output power generation data are generated, namely, the output AC power, voltage, current and other index data of each inverter processing node under the set constraints and input conditions.

[0077] Furthermore, in the method provided in the application embodiments, after constructing the photovoltaic power generation system to input the predicted load of the user-side power grid in the first predicted time zone, it further includes:

[0078] Read the power supply planning information of the preset power supply area in the first predicted time zone; determine whether the predicted load exceeds the power supply planning information. If yes, obtain the excess load and store it through the energy storage device. If no, determine the power supply of the main distribution network connected to the user-side power grid based on the excess load.

[0079] In this embodiment, a time-based supply and demand scheduling parsing method is first used to read the power supply planning information of the preset power supply area in the first prediction time zone. This information comes from the regional energy management system (EMS) or distribution automation platform, and its content includes the maximum power limit allowed to be connected in the area during the target prediction time period, the operating status of energy storage devices, the power supply strategy of the main distribution network, and load level boundaries, etc. The power supply planning table is retrieved and parsed through a standard communication protocol to extract the access capacity and load control strategy for the current time period, which serves as the basis for subsequent judgment and control.

[0080] Subsequently, a power boundary comparison method is used to determine whether the predicted load calculated by the photovoltaic system exceeds the allowable range in the power supply planning information. This method performs boundary condition judgment by comparing the predicted load obtained in the previous stage (i.e., the available power supply capacity of the photovoltaic system under specific environmental and conversion conditions) with the upper limit of the access power in the power supply plan. If the predicted load is greater than the upper limit, it is determined to be an overload; otherwise, it is considered that the current power supply capacity can be fully absorbed by the grid. This step provides control branch logic for subsequent energy storage call-up or main grid supplementary power strategy selection.

[0081] When the predicted load is determined to exceed the planned upper limit, a dynamic energy absorption regulation method is used to guide the excess electrical energy to local energy storage devices for storage. This method is based on a bidirectional converter control mechanism, converting excess electricity into the DC input required by the energy storage units (such as lithium batteries or supercapacitors). Dynamic control of the energy storage process is achieved by setting power limits, SOC (State of Charge) thresholds, and charging curve control methods. Before energy storage is executed, the real-time SOC and available capacity of the energy storage device are verified to ensure that redundant power can be fully received. Finally, the temporary storage of electrical energy is completed, avoiding power curtailment losses caused by photovoltaic power generation capacity exceeding grid-connected capacity.

[0082] When the predicted load is determined to be within the power supply planning limits, the main distribution network power allocation method is used to determine that the effective portion of the predicted load will be supplied by the interconnection channel between the user-side grid and the main distribution network. This method performs power allocation calculations based on the electrical parameters of the user access point (such as grid connection switch status, voltage level, distribution transformer load rate, etc.) and the main grid load margin information. Specific strategies may employ voltage adjustment or load transfer methods to identify the portion of the current photovoltaic output that can be supported or accepted by the main grid, and control its output as grid-connected power to ensure safe system operation.

[0083] In summary, the embodiments of this application have at least the following technical effects:

[0084] This application identifies multiple power generation units within a photovoltaic power generation system within a preset power supply area, and the connection network of these multiple power generation units. It collects the photovoltaic panel models and service times of the multiple power generation units, as well as the meteorological cycle characteristics of the preset power supply area, to construct constraints. Multiple power generation databases are built through historical data retrieval. First meteorological forecast information for the preset power supply area in a first forecast time zone is collected and input into the multiple power generation databases for analysis, generating initial power generation data for multiple sub-units. Based on the connection network, the initial power generation data of the multiple sub-units is analyzed to determine the fusion loss after the power is transmitted from the power generation units to the inverter for processing, thus constructing a predicted load for the photovoltaic power generation system that can be input into the user-side power grid in the first forecast time zone. This invention solves the technical problem of insufficient load forecast accuracy caused by the lack of processing of equipment aging, connection relationships, and multi-source data fusion analysis in existing technologies. By constructing a power generation database containing multi-dimensional constraints such as photovoltaic panel models, service times, and meteorological cycles, and combining this with the analysis of power fusion losses using the connection network, the technical effect of improving the adaptability and accuracy of the forecast results to the user-side power grid is achieved.

[0085] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0086] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0087] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for load forecasting of photovoltaic power generation systems based on big data, characterized in that, include: Determine multiple power generation units within a photovoltaic power generation system in a preset power supply area, as well as the connection network of the multiple power generation units; Constraints are constructed by collecting the photovoltaic panel models and service times of the multiple power generation units and the meteorological cycle characteristics of the preset power supply area, and multiple power generation databases are constructed by retrieving historical data. Collect the first meteorological forecast information of the preset power supply area in the first forecast time zone, and input it into the multiple power generation databases for analysis to generate initial power generation data for multiple sub-units; Based on the connection network, the initial power generation data of the multiple sub-units is analyzed to determine the fusion loss after the power is transmitted from the power generation unit to the inverter for processing, and the photovoltaic power generation system can be constructed to input the predicted load of the user-side power grid in the first prediction time zone; Constraints are constructed by collecting the photovoltaic panel models and service times of the multiple power generation units and the meteorological cycle characteristics of the preset power supply area. Multiple power generation databases are built through historical data retrieval, including: The first data constraint is constructed based on the photovoltaic panel model and service time of the multiple power generation units. Based on big data technology, multiple historical power generation datasets are retrieved. The historical power generation data includes historical meteorological information and corresponding historical power generation data at any historical moment. A second data constraint is constructed based on the photovoltaic panel model of the multiple power generation units and the meteorological cycle characteristics of the preset power supply area, and multiple historical power generation data time series that change sequentially with the service time are retrieved; The power generation quality degradation analysis of the multiple power generation units is performed using the time series of the multiple historical power generation data, and multiple power generation quality degradation curves are constructed. The multiple power generation quality degradation curves are mapped and connected to the multiple historical power generation datasets to obtain the multiple power generation databases.

2. The method for load forecasting of a photovoltaic power generation system based on big data as described in claim 1, characterized in that, Determine multiple power generation units within a photovoltaic power generation system in a preset power supply area, and the connection network of the multiple power generation units, including: Collect data from each inverter processing node within the photovoltaic power generation system; Each inverter processing node is used as a partitioning node to partition the multiple generator units and construct each generator unit array. The interconnection network is constructed using the various generator unit arrays and the various inverter processing nodes.

3. The method for load forecasting of a photovoltaic power generation system based on big data as described in claim 2, characterized in that, Each power generation unit array includes several power generation units connected in series or in parallel, and the initial electrical energy of the several power generation units flows into the same inverter processing node for processing.

4. The method for load forecasting of a photovoltaic power generation system based on big data as described in claim 1, characterized in that, Based on the time series of the aforementioned historical power generation data, a power generation quality degradation analysis is performed on the aforementioned power generation units to construct multiple power generation quality degradation curves, including: The time series of the multiple historical power generation data are decomposed according to the power generation parameter type to generate multiple combinations of historical power generation parameter time series. Based on the time series combination of the multiple historical power generation parameters, the relationship between the degradation degree of the power generation parameters and the service time is fitted to generate the multiple power generation quality degradation curves.

5. The method for load forecasting of a photovoltaic power generation system based on big data as described in claim 2, characterized in that, Based on the connection network, the initial power generation data of the multiple sub-units is analyzed to determine the fusion loss after the power is transmitted from the power generation unit to the inverter for processing. This analysis then constructs a forecast load that the photovoltaic power generation system can input into the user-side power grid in the first forecast time zone, including: Based on the connection network, the initial power generation data of multiple sub-units in series and parallel states of each power generation unit array are superimposed and the power loss is analyzed to generate the first output power generation data corresponding to each power generation unit array. Based on the power transmission parameter requirements of the user-side power grid, each inverter processing node performs power processing simulation on each first output power generation data to generate each second output power generation data. Power loss analysis is performed on the access circuit from each inverter processing node to the user-side power grid, and loss compensation is performed on each second output power generation data to generate the predicted load.

6. The method for load forecasting of a photovoltaic power generation system based on big data as described in claim 5, characterized in that, Based on the aforementioned connection network, the initial power generation data of multiple sub-units in series and parallel states of each power generation unit array are superimposed and power loss analysis is performed to generate the first output power generation data corresponding to each power generation unit array, including: Construct a mechanism for superimposing series and parallel power parameters; Collect the structural information of each connection circuit of each electron-generating unit array; Based on the structural information of each connection circuit, historical power transmission loss data is retrieved, and the relationship between power loss and circuit block loss is analyzed using the historical power transmission loss data to construct a power loss compensation mechanism. The initial power generation data of the multiple sub-units are superimposed and loss analyzed by combining the series-parallel power parameter superposition mechanism and the power loss compensation mechanism to generate the first output power generation data of each sub-unit.

7. The method for load forecasting of a photovoltaic power generation system based on big data as described in claim 5, characterized in that, Based on the power transmission parameter requirements of the user-side power grid, each inverter processing node performs power processing simulation on each first output power generation data to generate each second output power generation data, including: Perform twin simulations on each inverter processing node to establish each twin inverter. The twin simulations include at least twin simulations of the DC input terminal, current control circuit, filter circuit, and AC output terminal. Using the power transmission parameter requirements as simulation output constraints, each twin inverter performs simulation operation with each first output power generation data as simulation input to generate each second output power generation data.

8. The method for load forecasting of a photovoltaic power generation system based on big data as described in claim 7, characterized in that, The required transmission parameters include at least the input voltage, frequency, power factor, and harmonic constraints of the user-side power grid.

9. The method for load forecasting of a photovoltaic power generation system based on big data as described in claim 1, characterized in that, After constructing the photovoltaic power generation system to input the predicted load of the user-side power grid in the first prediction time zone, the system further includes: Read the power supply planning information of the preset power supply area in the first predicted time zone; Determine whether the predicted load exceeds the power supply planning information. If so, obtain the excess load and store it through an energy storage device. If not, determine the power supply of the main distribution network connected to the user-side power grid based on the excess load.

Images