A photovoltaic power generation consumption prediction method and system based on time period division

Abstract

The present application relates to a kind of photovoltaic power generation consumption prediction method and system based on time period division, belong to photovoltaic consumption and data intelligent processing technical field.The technical scheme is: obtaining consumption statistical data;Determine multiple actual power generation;Build consumption prediction module;Build photovoltaic actual power generation prediction module;The output layer of photovoltaic actual power generation prediction module is merged with the input layer of consumption prediction module, and photovoltaic consumption prediction model is generated;Based on the actual power generation of photovoltaic actual power generation prediction module output, it is transmitted to consumption prediction module, and photovoltaic consumption prediction result is output;Determine surplus photovoltaic electric energy;Based on surplus photovoltaic electric energy and photovoltaic consumption area, determine photovoltaic consumption path, and surplus photovoltaic electric energy is surplus grid-connected consumption.The beneficial effects of the present application are: by optimizing consumption prediction and distribution mode, self-use consumption and surplus grid-connected consumption are reasonably combined, and the maximum utilization of resources is realized on the basis of guaranteeing supply and consumption.

Description

Technical Field

[0001] This invention relates to a method and system for predicting photovoltaic power generation consumption based on time period division, belonging to the field of photovoltaic consumption and intelligent data processing technology. Background Technology

[0002] To achieve sustainable energy development, energy conversion is carried out, transforming renewable energy into non-renewable energy to meet societal needs. Photovoltaic power plants convert solar energy into electricity to meet users' electricity demands. However, it is not convenient to store the electricity generated by photovoltaic power plants. To avoid energy waste, it is necessary to ensure the rational scheduling of electricity, maximize the consumption of electricity, and ensure the consistency of supply and consumption of photovoltaic power plants, i.e., energy supply and consumption. Currently, the distribution of photovoltaic electricity is mainly achieved through electricity market transactions. However, due to the limitations of existing technology, the distribution results are inconsistent and need further improvement.

[0003] In existing technologies, when photovoltaic power generation is integrated into the grid, the accuracy of the integration prediction results is insufficient and the intelligence of the allocation method is inadequate, which makes it impossible to achieve the expected results, resulting in a mismatch between supply and consumption and a waste of resources. Summary of the Invention

[0004] The purpose of this invention is to provide a photovoltaic power generation consumption prediction method and system based on time period division. By optimizing the consumption prediction and allocation method, it rationally combines self-consumption consumption with surplus grid-connected consumption, thereby maximizing resource utilization while ensuring supply and consumption consistency, and solving the problems existing in the background technology.

[0005] The technical solution of this invention is: A photovoltaic power generation consumption prediction method based on time period division includes the following steps: ① Set a preset time period to retrieve the regional power consumption data of the photovoltaic power station and obtain the power consumption statistics; ② Determine the installed capacity of the photovoltaic power station, and determine the actual power generation of various types based on the installed capacity and multi-dimensional environmental parameters; ③ Construct a consumption forecasting module based on consumption statistics; ④ Construct a photovoltaic actual power generation prediction module based on multiple types of actual power generation; ⑤ Merge the output layer of the photovoltaic actual power generation prediction module with the input layer of the grid connection prediction module to generate a photovoltaic grid connection prediction model; ⑥ Obtain real-time environmental parameters and input them into the photovoltaic power consumption prediction model. Based on the actual power generation prediction module, the actual power generation is output and transmitted to the power consumption prediction module to output the photovoltaic power consumption prediction result. ⑦ Determine surplus photovoltaic power based on actual power generation and photovoltaic consumption forecast results; ⑧ Determine the photovoltaic consumption path based on the surplus photovoltaic power and the photovoltaic consumption area, and carry out surplus photovoltaic power grid connection and consumption.

[0006] Step ① also includes the following steps: The system retrieves grid connection data for the photovoltaic power plant's management area based on a preset time period, and obtains the retrieval results. It then cleans the retrieved data to obtain standardized statistical data. Based on this standardized statistical data, it converts cumulative values ​​into instantaneous values ​​and constructs a trend chart. The system identifies missing data based on the trend chart and supplements it to obtain data preprocessing results. Based on the data preprocessing results, it calculates node increments. It sets increment thresholds and uses these thresholds to initially divide and label the trend chart, determining multi-period trend charts. Finally, it uses these multi-period trend charts as the grid connection statistics.

[0007] Step ③ further includes the following steps: The data on absorption statistics are divided into multiple sets of data based on seasonality, climate, and time zone. A set of data is randomly extracted from the multiple sets of data to construct the first prediction model. A second set of data is randomly extracted from the data on absorption statistics to construct the second prediction model. Sample data is extracted multiple times to construct N prediction models. The first prediction model, the second prediction model, and so on up to the Nth prediction model are integrated to generate the absorption prediction module.

[0008] The step of randomly extracting a set of data based on multiple sets of partitioned data to construct the first prediction model also includes the following steps: A set of data is randomly extracted from multiple sets of partitioned data as a sample dataset; the photovoltaic consumption ratio is determined based on the sample dataset as the consumption result; the sample dataset and the consumption result are labeled to generate a labeled sample dataset; based on the labeled sample dataset, a neural network is trained using the K-fold cross-validation method to generate the first prediction model.

[0009] The process also includes the following steps: dividing the labeled sample dataset into K groups of sample datasets in equal proportions; determining the training set and test set based on the K groups of sample datasets, wherein the training set includes K-1 groups of sample datasets and the test set includes one group of sample data; training a neural network based on the training set to determine the first adaptive model, testing the model based on the test set, and obtaining the first test result; adjusting the sample data of the training set and test set to obtain the second adaptive model and generating the second test result; repeating the sample data adjustment K times to obtain the Kth adaptive model and generating the Kth test result; ranking the first test result, the second test result, and so on up to the Kth test result to determine the optimal test result; and performing back-matching of the optimal test result to obtain the first prediction model.

[0010] Step ⑧ also includes the following steps: A 3D model of the photovoltaic (PV) absorption area is constructed to create a simulated scene space. Surplus PV power and PV absorption paths are used to generate absorption control parameters. The absorption control parameters are input into the simulated scene space to conduct a simulated experiment on surplus power absorption and obtain simulation training results. It is determined whether the simulation training results meet the absorption ratio threshold. When they do, the absorption control parameters are used as power dispatch parameters for real-time surplus power dispatch.

[0011] The process of generating consumption control parameters for surplus photovoltaic power and photovoltaic consumption paths further includes: collecting load node data for the photovoltaic consumption area and constructing a load node topology map; collecting historical load node consumption data and dividing the data to determine multiple sets of samples to construct data; constructing a surplus power consumption planning model based on the multiple sets of sample data and the load node topology map; and inputting the surplus photovoltaic power into the surplus power consumption planning model to obtain the photovoltaic consumption path.

[0012] A photovoltaic power generation consumption prediction system based on time period division includes a data acquisition module, a power generation determination module, a consumption prediction module construction module, a photovoltaic actual power generation prediction module construction module, a model generation module, a result output module, a surplus power determination module, and a power consumption module. The data acquisition module is used to set a preset time period to retrieve the photovoltaic power station's regional absorption data and obtain absorption statistics. The power generation determination module is used to determine the installed capacity of the photovoltaic power station and to determine various types of actual power generation based on the installed capacity and multi-dimensional environmental parameters. The absorption prediction module construction module is used to construct an absorption prediction module based on absorption statistics data; The photovoltaic actual power generation prediction module construction module is used to construct a photovoltaic actual power generation prediction module based on multiple types of actual power generation. The model generation module is used to merge the output layer of the photovoltaic actual power generation prediction module and the input layer of the grid absorption prediction module to generate a photovoltaic grid absorption prediction model. The result output module is used to obtain real-time environmental parameters, input them into the photovoltaic power consumption prediction model, output the actual power generation based on the photovoltaic power generation prediction module, transmit them to the power consumption prediction module, and output the photovoltaic power consumption prediction result. The surplus power determination module is used to determine surplus photovoltaic power based on actual power generation and photovoltaic consumption prediction results. The power consumption module is used to determine the photovoltaic consumption path based on the surplus photovoltaic power and the photovoltaic consumption area, and to carry out surplus photovoltaic power grid connection and consumption.

[0013] It also includes a data retrieval module, a data cleaning module, a statistical chart construction module, a data preprocessing module, an incremental calculation module, a multi-period trend statistical chart determination module, and a data absorption statistical data determination module; The data retrieval module is used to retrieve the absorption data of the photovoltaic power station management area based on a preset time period and obtain the data retrieval results. The data cleaning module is used to clean the data retrieval results and obtain standardized statistical data. The statistical chart construction module is used to convert cumulative values ​​into instantaneous values ​​based on standardized statistical data and construct trend statistical charts. The data preprocessing module is used to identify missing data based on trend statistics charts, supplement the missing data, and obtain data preprocessing results. The incremental calculation module is used to calculate the node increment based on the data preprocessing results; The multi-period trend chart determination module is used to set an incremental threshold, perform preliminary division and identification of the trend chart based on the incremental threshold, and determine the multi-period trend chart. The absorption statistics determination module is used to use multi-period trend statistics charts as absorption statistics.

[0014] The beneficial effects of this invention are: improving the accuracy of the consumption prediction results and the intelligence of the allocation method, so that the photovoltaic power generation consumption results meet the expected requirements; by optimizing the consumption prediction and allocation method, the self-consumption consumption and surplus grid-connected consumption are reasonably combined, so as to maximize the utilization of resources on the basis of ensuring the consistency of supply and consumption. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the prediction method of the present invention; Figure 2 This is a schematic diagram of the construction process of the absorption prediction module of the present invention; Figure 3 This is a schematic diagram of the surplus grid connection and consumption process of the present invention; Figure 4 This is a schematic diagram of the prediction system structure of the present invention; In the diagram: Data acquisition module 11, power generation determination module 12, power consumption prediction module construction module 13, photovoltaic actual power generation prediction module construction module 14, model generation module 15, result output module 16, surplus power determination module 17, power consumption module 18. Detailed Implementation

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] In this embodiment, refer to the appendix. Figure 1-3 A photovoltaic power generation consumption prediction method based on time period division includes the following steps: Step S100: Set a preset time period to retrieve the regional power consumption data of the photovoltaic power station and obtain the power consumption statistics. Specifically, it is not convenient to store electricity after photovoltaic power plants generate electricity. In order to avoid the waste of electricity, it is necessary to ensure the rational scheduling of electricity, maximize the consumption of electricity, and ensure the consistency of supply and consumption of photovoltaic power plants. This invention divides the actual power generation into multiple time periods based on the actual power generation situation, determines the actual power generation, and rationally plans and calls on surplus electricity on the basis of ensuring the needs of photovoltaic power plants, so as to maximize the utilization of electricity.

[0018] A preset time period is set, i.e., the time zone for historical data collection, to determine the control area of ​​the photovoltaic power station. For example, for a photovoltaic power station built for rooftop photovoltaic systems across an entire county, historical absorption data of the photovoltaic power station is retrieved based on the preset time period, including total absorption and absorption per unit load point. The acquired historical absorption data is preprocessed and standardized to generate the absorption statistics data. The absorption statistics data is used as the data source to be analyzed, providing a basis for subsequent absorption prediction of the control area.

[0019] Furthermore, step S100 also includes: Step S110: Based on the preset time period, retrieve the power consumption data of the photovoltaic power station management area and obtain the data retrieval results; Step S120: Perform data cleaning on the data retrieval results to obtain standardized statistical data; Step S130: Based on the standardized statistical data, convert the cumulative values ​​into instantaneous values ​​and construct a trend chart; Step S140: Based on the trend chart, determine the missing data, supplement the missing data, and obtain the data preprocessing results; Step S150: Calculate the node increment based on the data preprocessing results; Step S160: Set an incremental threshold, and perform preliminary division and identification of the trend statistics chart based on the incremental threshold to determine the multi-time period trend statistics chart; Step S170: Use the multi-period trend statistics chart as the absorption statistics.

[0020] Specifically, historical data on the photovoltaic power station's managed area are retrieved based on the preset time period. The retrieved data is then integrated based on the time series to obtain the data retrieval results. The data retrieval results contain some abnormal data, which is identified and removed to complete data cleaning and ensure the accuracy of the data retrieval results. Furthermore, dimensionless normalization processing is performed on the data to standardize the data format, ensuring the orderliness and regularity of the data, and obtaining the standardized statistical data.

[0021] Specifically, the standardized statistical data is quantified and integrated, converting cumulative values ​​into instantaneous values. Preferably, the statistical interval of the standardized statistical data is 5 minutes, with 0:00 each day as the daily statistical node, and the corresponding data for the daily statistical node is determined through data accumulation and conversion. The extreme values ​​of the standardized statistical data are determined, including peak values, valley values, and stable values, which are used as representative node data. A two-dimensional coordinate system is constructed using time and absorption data as coordinate axes. A trend chart is constructed in the two-dimensional coordinate system based on the corresponding data of the daily statistical node and the representative node data, respectively, which can ensure the data completeness of the trend chart.

[0022] Furthermore, missing data is identified and located based on the trend chart. For example, missing data can be supplemented using the mean interpolation method to obtain the data preprocessing results. The node data increment is then calculated for each of the two sets of statistical data to obtain the node increment. An increment threshold is set and used as the critical value for data division. Preferably, multiple levels of increment thresholds can be set to ensure the fineness of data division. The trend chart is divided into multiple time periods, and the division results are marked with node data to facilitate identification and differentiation. The absorption statistical data is then generated. By performing multi-dimensional processing on the collected data, the accuracy of subsequent data analysis is guaranteed.

[0023] Step S200: Determine the installed capacity of the photovoltaic power station, and determine the actual power generation of various types based on the installed capacity and multi-dimensional environmental parameters; Specifically, the installed capacity of the photovoltaic power station is determined, i.e., the total rated effective power of the units. The initial construction of the photovoltaic power station has been determined, and the theoretical power generation of the photovoltaic power station can be determined. Different external environmental conditions will affect the theoretical power generation of the photovoltaic power station, such as sunny days, rainy days, and sunlight angle. The external environment is classified into multiple types to determine the corresponding theoretical power generation under different external environmental conditions. At the same time, there are certain photovoltaic losses during the power generation process, such as dust and dirt covering, cable losses, etc. Power generation losses are excluded from the multiple types of theoretical power generation to determine the multiple types of actual power generation, which are the energy to be consumed.

[0024] Step S300: Construct a consumption prediction module based on the aforementioned consumption statistics; Furthermore, such as Figure 2 As shown, step S300 further includes: Step S310: Using seasonality, climate, and time zone as the basis for division, perform secondary division on the absorption statistics data to obtain multiple sets of division data; Step S320: Randomly extract a set of data based on the multiple sets of partitioned data to construct a first prediction model; Step S330: Based on the aforementioned absorption statistics, randomly extract a set of data again to construct a second prediction model; Step S340: Extract sample data multiple times to construct N prediction models; Step S350: Integrate the first prediction model, the second prediction model, and up to the Nth prediction model to generate the absorption prediction module.

[0025] Specifically, based on seasonality, climate, and time zone, multiple environmental parameters are determined, and the statistical data on absorption is divided into multiple sets of data. A set of data is randomly extracted from these multiple sets of data, evaluated and labeled for easy identification, and used as a labeling sample dataset. Based on this labeling sample dataset, a neural network is trained using K-fold cross-validation to generate the first prediction model. Again, a set of data is randomly extracted from the multiple sets of data, evaluated and labeled, and used as a labeling sample dataset. A neural network is then trained using K-fold cross-validation to generate the second prediction model. This modeling process is repeated N times, where N is the number of sets of data, to obtain N prediction models. The first prediction model, the second prediction model, and so on up to the Nth prediction model are integrated to form the absorption prediction module.

[0026] The modeling methods of the first prediction model, the second prediction model, and so on up to the Nth prediction model are the same. The extracted data sets and various environmental parameters correspond one-to-one with the multiple prediction models, which facilitates targeted model analysis for actual conditions and improves model calculation efficiency while ensuring the accuracy of prediction results.

[0027] Furthermore, step S320, which involves randomly extracting a set of data based on the multiple sets of segmented data to construct a first prediction model, further includes: Step S321: Randomly extract a set of data as a sample dataset based on the multiple sets of partitioned data; Step S322: Determine the photovoltaic power consumption ratio based on the sample dataset, as the consumption result; Step S323: Identify the corresponding data sets and the elimination results to generate an identified sample dataset; Step S324: Based on the identified sample dataset, train the neural network using the K-fold cross-validation method to generate the first prediction model.

[0028] Furthermore, step S324 also includes: Step S3241: Divide the identified sample dataset into K groups of sample datasets in equal proportions; Step S3242: Determine the training set and test set based on the K sets of sample datasets, wherein the training set includes K-1 sets of sample datasets and the test set includes 1 set of sample data; Step S3243: Train the neural network based on the training set to determine the first adaptive model, and test the model based on the test set to obtain the first test result; Step S3244: Adjust the sample data of the training set and the test set to obtain the second adaptive model and generate the second test result; Step S3245: Repeat the sample data adjustment K times to obtain the Kth adaptive model and generate the Kth test result; Step S3246: Sort the first test result, the second test result, and up to the Kth test result according to their merits and determine the optimal test result; Step S3247: Perform model back-matching on the optimal test result to obtain the first prediction model.

[0029] Specifically, a set of data is randomly extracted based on the multiple sets of segmented data and used as the sample dataset, i.e. the initial data for modeling. The actual power generation and the control area consumption data are extracted respectively, and the two are in one-to-one correspondence. The ratio of the control area consumption data to the actual power generation is used as the photovoltaic consumption ratio to obtain the consumption result. Then, the sample dataset and the consumption result are correspondingly labeled to obtain multiple sets of label data as the label sample dataset.

[0030] Based on the identified sample dataset, the first prediction model is constructed. The first prediction model corresponds to an environmental parameter. K is used as the sample data to divide the data into groups. The identified sample dataset is divided proportionally to obtain K groups of sample datasets. K-1 groups are randomly extracted from the K groups of sample datasets as training sets, and the remaining 1 group is used as a test set. The neural network is trained based on the training set. For example, the extracted absorption statistics are used as data identification nodes, and the corresponding absorption results are used as decision nodes to construct the first adaptive model. Then, the test set is input into the first adaptive model. The corresponding output result is determined by model analysis. The absorption result corresponding to the test set and the output result are analyzed for fit, and the fit information is used as the first test result.

[0031] Furthermore, the training set and the test set are adjusted by extracting any data that differs from the previous test set for test set iteration. The remaining sample datasets are used as the training set. Based on the above modeling and analysis steps, the second adaptive model is constructed, and the test set is input to determine the fit between the output result and the corresponding elimination result of the test set, which is used as the second test result. The above modeling and testing steps are repeated K times to ensure that each sample dataset is used as the test set for testing, thereby obtaining the Kth adaptive model and generating the Kth test result. The modeling mechanism of the adaptive model is the same as the model testing method.

[0032] The first test result, the second test result, and so on up to the Kth test result are used as evaluation data and ranked according to their merits. The one with the highest degree of fit is taken as the optimal test result. Then, the optimal test result is back-matched to determine the adaptive model corresponding to the optimal test result as the first prediction model. This can effectively improve the analysis accuracy of the first prediction model and ensure the fit between the model analysis results and the photovoltaic power station.

[0033] Step S400: Construct a photovoltaic actual power generation prediction module based on multiple types of actual power generation; Step S500: Merge the output layer of the photovoltaic actual power generation prediction module with the input layer of the grid absorption prediction module to generate the photovoltaic grid absorption prediction model; Step S600: Obtain real-time environmental parameters, input them into the photovoltaic power consumption prediction model, output the actual power generation based on the photovoltaic power generation prediction module, transmit it to the power consumption prediction module, and output the photovoltaic power consumption prediction result; Specifically, the various types of actual power generation are obtained, wherein each type of actual power generation corresponds one-to-one with the multi-dimensional environmental parameters. A set of environmental parameters includes season, weather, irradiance angle, irradiance, time, etc. The multi-dimensional environmental parameters are used as identification nodes, and the various types of actual power generation are used as decision nodes to generate a power generation decision tree. The photovoltaic actual power generation prediction module is constructed based on the power generation decision tree.

[0034] Specifically, the photovoltaic (PV) power generation prediction module and the PV absorption prediction module are embedded in the PV absorption prediction model. The output layer of the PV power generation prediction module and the input layer of the absorption prediction model are merged to generate the PV absorption prediction model. Real-time environmental parameters are collected, for example, based on sensor data identification using multiple types of sensors, with weather forecasts used as data references to ensure the accuracy of the collected data. The real-time environmental parameters are input into the PV absorption prediction model. First, the data is transmitted to the PV power generation prediction module for data identification and matching to obtain the actual power generation. Then, the actual power generation and the real-time environmental parameters are transmitted to the absorption prediction module. Based on the real-time environmental parameters, the corresponding prediction model is matched to predict the absorption of the actual power generation, obtaining the PV absorption prediction result. The actual power generation and the PV absorption prediction result are used as the output of the PV absorption prediction model. The PV absorption prediction result is the absorption result of the PV power plant's management area, and further analysis is performed based on the PV absorption prediction result.

[0035] Step S700: Determine the surplus photovoltaic power based on the actual power generation and the photovoltaic absorption prediction results; Step S800: Determine the photovoltaic consumption path based on the surplus photovoltaic power and the photovoltaic consumption area, and carry out surplus photovoltaic power grid connection and consumption.

[0036] Specifically, based on the photovoltaic power generation prediction model, the real-time environmental parameters are identified and analyzed to obtain the actual power generation and the photovoltaic power generation prediction results. The difference between the actual power generation and the photovoltaic power generation prediction results is calculated to obtain the surplus photovoltaic power. The surplus photovoltaic power is the power that cannot be consumed in the photovoltaic power station's management area. The surplus photovoltaic power is then consumed through grid-connected scheduling.

[0037] The surplus photovoltaic power is used as power to be connected to the grid for consumption. The photovoltaic consumption area, i.e. the grid-connected consumption area, is determined. The photovoltaic consumption path is obtained. Load nodes are allocated to the photovoltaic consumption area to realize the consumption of the surplus photovoltaic power. By linking the power generation of the photovoltaic power station with self-consumption and surplus grid-connected consumption, the consistency of power supply and consumption is achieved.

[0038] Furthermore, such as Figure 3 As shown, step S800 further includes: Step S810: Perform 3D modeling of the photovoltaic absorption area to construct a scene simulation space; Step S820: Generate absorption control parameters by combining the surplus photovoltaic power with the photovoltaic absorption path; Step S830: Input the absorption control parameters into the scenario simulation space, conduct a surplus power absorption simulation experiment, and obtain simulation training results; Step S840: Determine whether the simulated training result meets the elimination ratio threshold; Step S850: When the conditions are met, the absorption control parameters are used as power dispatch parameters to perform real-time surplus power dispatch.

[0039] Furthermore, step S820 also includes: Step S821: Collect load nodes in the photovoltaic consumption area and construct a load node topology map; Step S822: Collect historical load node absorption data, divide the data to determine multiple groups of samples to construct the data; Step S823: Based on the multiple sets of samples, construct the data and the load node topology map, and construct a surplus power consumption planning model; Step S824: Input the surplus photovoltaic power into the surplus power consumption planning model to obtain the photovoltaic consumption path.

[0040] Specifically, the surplus photovoltaic power is connected to the grid for consumption. The photovoltaic consumption area is determined, which is a consumption area different from the photovoltaic power station control area. There is a demand for electricity in this area. The photovoltaic consumption area is 3D modeled to construct a scene simulation space, wherein the scene simulation space is consistent with the photovoltaic consumption area.

[0041] Furthermore, a surplus power consumption planning model is constructed to process surplus power. Multiple load nodes in the photovoltaic consumption area are identified, and load nodes are linked based on grid flow to generate a load node topology map. Historical consumption data is collected for each node in the load node topology map, and data is segmented based on multi-dimensional environmental parameters to obtain multiple sets of sample construction data. Load nodes are identified in each set of sample construction data. For example, the surplus power consumption planning model architecture is constructed based on a machine learning algorithm. The load node topology map is used as the planning basis, and the multiple sets of sample construction data are used as matching reference data. The surplus power consumption planning model architecture is trained and optimized to obtain the surplus power consumption planning model.

[0042] The surplus photovoltaic power is input into the surplus power consumption planning model. The surplus photovoltaic power is allocated based on the load node absorption capacity. The flow direction of the surplus photovoltaic power is determined based on the load power topology map. The photovoltaic consumption path is generated and output. Further, the consumption control parameters are generated based on the surplus photovoltaic consumption control parameters and the photovoltaic consumption path, including power flow direction, load node absorption capacity, etc. The consumption control parameters are input into the scenario simulation space to conduct scenario simulation experiments and obtain simulation training results, that is, the photovoltaic consumption area's effect on the consumption of surplus photovoltaic power.

[0043] A threshold for the absorption ratio is set, which is the critical value for evaluating the qualification of absorbed electricity. For example, 95% of the surplus photovoltaic electricity is used as the absorption ratio threshold. When the simulation training results meet the absorption ratio threshold, it indicates that the absorption requirements can be met. The absorption control parameters are used as power dispatch parameters to control the absorption of surplus electricity. When the requirements are not met, it indicates that the absorption ratio of surplus photovoltaic electricity does not meet the predetermined requirements, and there is a certain amount of electricity waste. The surplus electricity absorption planning model is re-planned based on the surplus electricity absorption planning model, and then simulation verification is performed based on the scenario simulation space until the absorption requirements are met. This can effectively improve the photovoltaic absorption rate, maximize the utilization of photovoltaic power plant power generation, and avoid resource waste.

[0044] See attached document Figure 4 A photovoltaic power generation consumption prediction system based on time period division includes a data acquisition module 11, a power generation determination module 12, a consumption prediction module construction module 13, a photovoltaic actual power generation prediction module construction module 14, a model generation module 15, a result output module 16, a surplus power determination module 17, and a power consumption module 18. The data acquisition module 11 is used to set a preset time period to retrieve the photovoltaic power station's regional absorption data and obtain absorption statistics. The power generation determination module 12 is used to determine the installed capacity of the photovoltaic power station and to determine various types of actual power generation based on the installed capacity and multi-dimensional environmental parameters. The absorption prediction module construction module 13 is used to construct an absorption prediction module based on the absorption statistics data; The photovoltaic actual power generation prediction module construction module 14 is used to construct a photovoltaic actual power generation prediction module based on multiple types of actual power generation. The model generation module 15 is used to merge the output layer of the photovoltaic actual power generation prediction module and the input layer of the grid absorption prediction module to generate a photovoltaic grid absorption prediction model. The result output module 16 is used to acquire real-time environmental parameters, input them into the photovoltaic power consumption prediction model, output the actual power generation based on the photovoltaic power generation prediction module, transmit them to the power consumption prediction module, and output the photovoltaic power consumption prediction result. The surplus power determination module 17 is used to determine surplus photovoltaic power based on the actual power generation and the photovoltaic consumption prediction results. The power consumption module 18 is used to determine the photovoltaic consumption path based on the surplus photovoltaic power and the photovoltaic consumption area, and to carry out surplus photovoltaic power grid connection and consumption.

[0045] Furthermore, the prediction system also includes a data retrieval module, a data cleaning module, a statistical chart construction module, a data preprocessing module, an incremental calculation module, a multi-period trend statistical chart determination module, and a data absorption statistical data determination module; The data retrieval module is used to retrieve the absorption data of the photovoltaic power station management area based on the preset time period and obtain the data retrieval results. The data cleaning module is used to clean the data retrieval results and obtain standardized statistical data. The statistical chart construction module is used to convert cumulative values ​​into instantaneous values ​​based on the standardized statistical data and construct a trend statistical chart. The data preprocessing module is used to determine missing data based on the trend chart, supplement the missing data, and obtain data preprocessing results. The incremental calculation module is used to calculate the node increment based on the data preprocessing results; The multi-period trend chart determination module is used to set an incremental threshold, perform preliminary division and identification of the trend chart based on the incremental threshold, and determine the multi-period trend chart. The absorption statistics determination module is used to use the multi-period trend statistics chart as the absorption statistics.

[0046] Furthermore, the prediction system also includes a data partitioning module, a first prediction model construction module, a second prediction model construction module, N prediction model construction modules, and an elimination prediction module generation module; The data segmentation module is used to perform secondary segmentation of the absorption statistics data based on seasonality, climate, and time zone to obtain multiple sets of segmentation data. The first prediction model construction module is used to randomly extract a set of data based on the multiple sets of partitioned data to construct a first prediction model; The second prediction model building module is used to randomly extract a set of data again based on the absorption statistics data to build a second prediction model; The N prediction model building modules are used to extract sample data multiple times and build N prediction models; The absorption prediction module generation module is used to integrate the first prediction model, the second prediction model, and up to the Nth prediction model to generate the absorption prediction module.

[0047] Furthermore, the prediction system also includes a dataset extraction module, an elimination result determination module, a sample identification module, and a first prediction model generation module; The dataset extraction module is used to randomly extract a set of data as a sample dataset based on the multiple sets of partitioned data; The photovoltaic power consumption result determination module is used to determine the photovoltaic power consumption ratio based on the sample dataset, as the power consumption result; The sample identification module is used to identify the corresponding sample dataset and the elimination result, and generate an identified sample dataset; The first prediction model generation module is used to train a neural network based on the K-fold cross-validation method according to the identified sample dataset to generate the first prediction model.

[0048] Furthermore, the prediction system also includes a sample dataset partitioning module, a sample dataset determination module, a first test result acquisition module, a second test result acquisition module, a Kth test result acquisition module, an optimal test result acquisition module, and a model matching module; The sample dataset partitioning module is used to divide the identified sample dataset into K groups of sample datasets in an equal proportion. The sample dataset determination module is used to determine the training set and the test set based on the K sets of sample datasets, wherein the training set includes K-1 sets of sample datasets and the test set includes 1 set of sample data; The first test result acquisition module is used to train a neural network based on the training set to determine a first adaptive model, test the model based on the test set, and acquire a first test result. The second test result acquisition module is used to adjust the sample data of the training set and the test set, obtain the second adaptive model, and generate the second test result; The Kth test result acquisition module is used to repeat the sample data adjustment K times to obtain the Kth adaptive model and generate the Kth test result; The optimal test result acquisition module is used to sort the first test result, the second test result and up to the Kth test result to determine the optimal test result. The model matching module is used to perform reverse model matching on the optimal test result to obtain the first prediction model.

[0049] Furthermore, the prediction system also includes a space construction module, a control parameter generation module, a simulation training result acquisition module, a result judgment module, and a power dispatching module; The space construction module is used to create a 3D model of the photovoltaic consumption area and construct a scene simulation space; The control parameter generation module is used to generate absorption control parameters by combining the surplus photovoltaic power with the photovoltaic absorption path. The simulation training result acquisition module is used to input the absorption control parameters into the scene simulation space, conduct a surplus power absorption simulation experiment, and obtain the simulation training results. The result judgment module is used to determine whether the simulated training result meets the elimination ratio threshold. The power dispatch module is used to perform real-time surplus power dispatch by using the absorption control parameters as power dispatch parameters when the conditions are met.

[0050] Furthermore, the prediction system also includes a topology graph construction module, a data acquisition module, a planning model construction module, and a path planning module; The topology map construction module is used to collect load nodes in the photovoltaic consumption area and construct a load node topology map. The load absorption data acquisition module is used to collect historical load absorption data of load nodes, divide the data, determine multiple groups of samples to construct data; The planning model construction module is used to construct a surplus power consumption planning model based on the multiple sets of sample construction data and the load node topology map. The path planning module is used to input the surplus photovoltaic power into the surplus power consumption planning model to obtain the photovoltaic consumption path.

Claims

1. A photovoltaic power generation consumption prediction method based on time period division, characterized in that... It includes the following steps: ① Set a preset time period to retrieve the photovoltaic power station's regional grid connection data and obtain grid connection statistics. The grid connection data includes the total grid connection. The grid connection statistics are obtained by cleaning the retrieved data and obtaining standardized statistics. Based on the standardized statistics, the cumulative values ​​are converted into instantaneous values ​​to construct a trend chart. Based on the trend chart, the missing data is identified and supplemented to obtain data preprocessing results. Based on the data preprocessing results, the node increment is calculated. An increment threshold is set, and the trend chart is initially divided and labeled based on the increment threshold to determine the multi-time period trend chart. ② Determine the installed capacity of the photovoltaic power station, and determine multiple types of actual power generation based on the installed capacity and multi-dimensional environmental parameters; the installed capacity of the photovoltaic power station is determined, that is, the total rated effective power of the units. The initial construction of the photovoltaic power station has been determined, and the theoretical power generation of the photovoltaic power station can be determined. Different external environmental conditions will have different effects on the theoretical power generation of the photovoltaic power station. Therefore, the external environment is divided into multiple types to determine the corresponding theoretical power generation under different external environmental conditions. At the same time, photovoltaic losses exist during the power generation process. Power generation losses are excluded from multiple types of theoretical power generation to determine multiple types of actual power generation. The multiple types of actual power generation are the energy to be consumed. ③ Construct a photovoltaic (PV) grid connection prediction module based on grid connection statistics. The input of the grid connection prediction module is the actual power generation, and the output is the PV grid connection ratio. ④ Construct a photovoltaic actual power generation prediction module based on multiple types of actual power generation; The various types of actual power generation correspond one-to-one with the various environmental parameters. The various environmental parameters are used as identification nodes, and the various types of actual power generation are used as decision nodes to generate a power generation decision tree. The photovoltaic actual power generation prediction module is constructed based on the power generation decision tree. ⑤ Merge the output layer of the photovoltaic actual power generation prediction module with the input layer of the grid connection prediction module to generate a photovoltaic grid connection prediction model; ⑥ Obtain real-time environmental parameters and input them into the photovoltaic power consumption prediction model. Based on the actual power generation prediction module, the actual power generation is output and transmitted to the power consumption prediction module to output the photovoltaic power consumption prediction result. ⑦ Determine surplus photovoltaic power based on actual power generation and photovoltaic consumption forecast results; ⑧ Determine the photovoltaic consumption path based on the surplus photovoltaic power and the photovoltaic consumption area, and carry out surplus photovoltaic power grid connection and consumption; Load node data is collected for the photovoltaic (PV) grid connection area to construct a load node topology map; historical load node grid connection data is collected and the data is divided to determine multiple sets of samples for data construction; based on the multiple sets of sample data and the load node topology map, a surplus power grid connection planning model is constructed; surplus PV power is input into the surplus power grid connection planning model to obtain the PV grid connection path. A 3D model of the photovoltaic (PV) absorption area is constructed to create a simulated scene space. Surplus PV power and PV absorption paths are used to generate absorption control parameters. The absorption control parameters are input into the simulated scene space to conduct a simulated experiment on surplus power absorption and obtain simulation training results. It is determined whether the simulation training results meet the absorption ratio threshold. When they do, the absorption control parameters are used as power dispatch parameters for real-time surplus power dispatch. The power absorption control parameters include the direction of power flow and the absorption capacity of load nodes.

2. The photovoltaic power generation consumption prediction method based on time period division according to claim 1, characterized in that... Step ① also includes the following steps: The statistical interval for the standardized statistical data is 5 minutes. 00:00 is taken as the daily statistical node, and data accumulation and transformation are performed to determine the corresponding data for the daily statistical node. Extreme values, including peak values, trough values, and stable values, are determined for the standardized statistical data and used as representative node data. A two-dimensional coordinate system is constructed using time and absorption data as coordinate axes. Trend charts are constructed in the two-dimensional coordinate system based on the corresponding data for the daily statistical node and the representative node data, ensuring the data completeness of the trend charts.

3. A photovoltaic power generation consumption prediction method based on time period division according to claim 1 or 2, characterized in that... Step ③ further includes the following steps: The data on absorption statistics are divided into multiple sets of data based on seasonality, climate, and time zone. A set of data is randomly extracted from the multiple sets of data to construct the first prediction model. A second set of data is randomly extracted from the data on absorption statistics to construct the second prediction model. Sample data is extracted multiple times to construct N prediction models. The first prediction model, the second prediction model, and so on up to the Nth prediction model are integrated to generate the absorption prediction module.

4. The photovoltaic power generation consumption prediction method based on time period division according to claim 3, characterized in that: The step of randomly extracting a set of data based on multiple sets of partitioned data to construct the first prediction model further includes the following steps: randomly extracting a set of data as a sample dataset based on multiple sets of partitioned data; The photovoltaic consumption ratio is determined based on the sample dataset and used as the consumption result; the sample dataset and the consumption result are labeled to generate a labeled sample dataset; based on the labeled sample dataset, a neural network is trained using the K-fold cross-validation method to generate the first prediction model. Based on the multiple sets of segmented data, a set of data is randomly extracted and used as the sample dataset, i.e. the initial data for modeling. The actual power generation and the control area consumption data are extracted respectively, and the two are matched one-to-one. The ratio of the control area consumption data to the actual power generation is used as the photovoltaic consumption ratio to obtain the consumption result. Then, the sample dataset and the consumption result are correspondingly labeled to obtain multiple sets of label data as the label sample dataset.

5. The photovoltaic power generation consumption prediction method based on time period division according to claim 4, characterized in that... The process also includes the following steps: dividing the labeled sample dataset into K groups of sample datasets in equal proportions; determining the training set and test set based on the K groups of sample datasets, wherein the training set includes K-1 groups of sample datasets and the test set includes one group of sample data; training a neural network based on the training set to determine the first adaptive model, testing the model based on the test set, and obtaining the first test result; adjusting the sample data of the training set and test set to obtain the second adaptive model and generating the second test result; repeating the sample data adjustment K times to obtain the Kth adaptive model and generating the Kth test result; ranking the first test result, the second test result, and so on up to the Kth test result to determine the optimal test result; and performing back-matching of the optimal test result to obtain the first prediction model.

6. A photovoltaic power generation consumption prediction system based on time period division, characterized in that: It includes a data acquisition module (11), a power generation determination module (12), a power consumption prediction module construction module (13), a photovoltaic actual power generation prediction module construction module (14), a model generation module (15), a result output module (16), a surplus power determination module (17), and a power consumption module (18). The data acquisition module is used to set a preset time period to retrieve the photovoltaic power station's regional consumption data and obtain consumption statistics, wherein the consumption data includes the total consumption. It also includes a data cleaning module, a statistical chart construction module, a data preprocessing module, an incremental calculation module, a multi-period trend statistical chart determination module, and a power generation statistical data determination module. The data cleaning module cleans the retrieved data to obtain standardized statistical data. The statistical chart construction module converts cumulative values ​​into instantaneous values ​​based on the standardized statistical data to construct a trend statistical chart. The data preprocessing module identifies missing data based on the trend statistical chart and supplements the missing data to obtain data preprocessing results. The incremental calculation module calculates node increments based on the data preprocessing results. The multi-period trend statistical chart determination module sets an incremental threshold and performs preliminary segmentation and identification of the trend statistical chart based on the incremental threshold to determine the multi-period trend statistical chart. The power generation statistical data determination module uses the multi-period trend statistical chart as the power generation statistical data. The power generation determination module determines the installed capacity of the photovoltaic power station and determines various types of actual power generation based on the installed capacity and multi-dimensional environmental parameters. The installed capacity of the photovoltaic power station is determined, i.e., the total rated effective power of the units. The initial construction of the photovoltaic power station has been determined, and the theoretical power generation of the photovoltaic power station can be determined. Different external environmental conditions will result in differences in the theoretical power generation of the photovoltaic power station. The external environment is classified into multiple types to determine the corresponding theoretical power generation under different external environmental conditions. At the same time, photovoltaic losses exist during the power generation process. The power generation losses of multiple types of theoretical power generation are eliminated to determine multiple types of actual power generation. The multiple types of actual power generation are the energy to be consumed. The photovoltaic power generation prediction module is used to construct a photovoltaic power generation prediction module based on photovoltaic power generation statistics. The input of the photovoltaic power generation prediction module is the actual power generation, and the output is the photovoltaic power generation ratio. The photovoltaic actual power generation prediction module construction module is used to construct a photovoltaic actual power generation prediction module based on multiple types of actual power generation. The various types of actual power generation correspond one-to-one with the various environmental parameters. The various environmental parameters are used as identification nodes, and the various types of actual power generation are used as decision nodes to generate a power generation decision tree. The photovoltaic actual power generation prediction module is constructed based on the power generation decision tree. The model generation module is used to merge the output layer of the photovoltaic actual power generation prediction module and the input layer of the grid absorption prediction module to generate a photovoltaic grid absorption prediction model. The result output module is used to obtain real-time environmental parameters, input them into the photovoltaic power consumption prediction model, output the actual power generation based on the photovoltaic power generation prediction module, transmit them to the power consumption prediction module, and output the photovoltaic power consumption prediction result. The surplus power determination module is used to determine surplus photovoltaic power based on actual power generation and photovoltaic consumption prediction results. The power consumption module is used to determine the photovoltaic consumption path based on the surplus photovoltaic power and the photovoltaic consumption area, and to carry out surplus photovoltaic power grid connection and consumption. Load node data is collected for the photovoltaic (PV) grid connection area to construct a load node topology map; historical load node grid connection data is collected and the data is divided to determine multiple sets of samples for data construction; based on the multiple sets of sample data and the load node topology map, a surplus power grid connection planning model is constructed; surplus PV power is input into the surplus power grid connection planning model to obtain the PV grid connection path. A 3D model of the photovoltaic (PV) absorption area is constructed to create a simulated scene space. Surplus PV power and PV absorption paths are used to generate absorption control parameters. The absorption control parameters are input into the simulated scene space to conduct a simulated experiment on surplus power absorption and obtain simulation training results. It is determined whether the simulation training results meet the absorption ratio threshold. When they do, the absorption control parameters are used as power dispatch parameters for real-time surplus power dispatch. The power absorption control parameters include the direction of power flow and the absorption capacity of load nodes.

7. A photovoltaic power generation consumption prediction system based on time period division according to claim 6, characterized in that: The statistical interval for the standardized statistical data is 5 minutes. 00:00 is taken as the daily statistical node, and data accumulation and transformation are performed to determine the corresponding data for the daily statistical node. Extreme values, including peak values, trough values, and stable values, are determined for the standardized statistical data and used as representative node data. A two-dimensional coordinate system is constructed using time and absorption data as coordinate axes. Trend charts are constructed in the two-dimensional coordinate system based on the corresponding data for the daily statistical node and the representative node data, ensuring the data completeness of the trend charts.

Images