Photovoltaic system power generation prediction method and device based on edge cluster

Abstract

The present application provides a photovoltaic system power generation prediction method and device based on edge clusters, applicable to distributed photovoltaic power generation systems. The method comprises: obtaining a first power generation, a first location information, a second power generation, a second location information, a first maintenance plan, and a first environmental data at intervals of a first preset duration, and performing preprocessing to obtain a first training data set; performing feature screening on the first training data set using a recursive feature elimination method to obtain a second training data set; performing feature extraction on the second training data set using a factor analysis method to obtain a third training data set; training a prediction model based on the third training data set to obtain a power generation prediction model; obtaining target prediction parameters, and inputting the target prediction parameters into the power generation prediction model to obtain a target power generation. This method solves the problem in the prior art of the lack of a distributed photovoltaic power generation power prediction method based on multidimensional feature data.

Description

Technical Field

[0001] The present invention relates to the field of power monitoring technology, and in particular to a method, device, computer-readable storage medium and monitoring system for predicting the power generation of a photovoltaic system based on an edge cluster. Background Art

[0002] At present, the construction of the edge cluster system of the cloud-edge integrated intelligent scheduling and operation platform (CEP) requires the realization of distributed photovoltaic power generation prediction.

[0003] The core architecture is a cloud-edge integrated intelligent scheduling and operation platform. The bottom layer comprises model data services, operational data services, and real-time data services, providing the data foundation for load forecasting. The second layer comprises the computing engine and the artificial intelligence engine, respectively providing basic services such as data cleaning, a deep learning environment, and power forecasting for applications. The top layer, the application layer, facilitates the push of power forecast models and results, visualizing data, and managing forecast results.

[0004] In summary, the existing technology lacks a distributed photovoltaic power prediction method based on an edge cluster system, and a method for predicting short-term power generation within different ranges based on multi-dimensional input data. Summary of the Invention

[0005] The main purpose of this application is to provide a photovoltaic system power generation prediction method, device, computer-readable storage medium and monitoring system based on edge clusters, so as to at least solve the problem in the prior art of the lack of a distributed photovoltaic power generation prediction method based on multi-dimensional feature data.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for predicting the power generation of a photovoltaic system based on an edge cluster is provided, and the method is applicable to a distributed photovoltaic power generation system, and the distributed photovoltaic power generation system includes a metering system, a scheduling management system and a meteorological system. The method includes: at each first preset time interval, respectively obtaining the power generation power and location information of the distributed photovoltaic from the metering system to obtain the first power generation power and the first location information, obtaining the power generation power of the distributed photovoltaic directly connected to the grid on the user side from the metering system to obtain the second power generation power, obtaining the feeder corresponding to the distributed photovoltaic directly connected to the grid on the user side from the scheduling management system to obtain the second location information, obtaining the maintenance record from the scheduling management system to obtain the first maintenance plan, obtaining the environmental data from the meteorological system to obtain the first environmental data, and performing the first power generation power, the first location information, the second power generation power, the second location information, the first maintenance plan and the first Environmental data is preprocessed to obtain a first training data set, and the preprocessing includes at least data cleaning, data clipping and data completion; the first training data set is feature screened according to a recursive feature elimination method to obtain a second training data set; the second training data set is feature extracted based on a factor analysis method to obtain a third training data set; a prediction model is trained based on the third training data set to obtain a power generation prediction model, and the power generation prediction model is used to predict the power generation power of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data; target prediction parameters are obtained, and the target prediction parameters are input into the power generation prediction model to obtain target power generation power, and the target prediction parameters include at least one of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data within the past first preset time period.

[0007] Optionally, the first power generation power, the first location information, the second power generation power, the second location information and the first environmental data are preprocessed to obtain a first training data set, including: determining the substation corresponding to the first power generation power based on the first location information, and summing the first power generation powers corresponding to each substation to obtain a corresponding third power generation power; determining the substation to which the feeder corresponding to the second power generation power belongs based on the second location information, and accumulating the power generation powers corresponding to each feeder according to the relationship between the feeder and the substation, the second power generation power and the third power generation power to obtain a corresponding fourth power generation power; constructing a fourth training data set based on the third power generation power, the fourth power generation power, the first maintenance plan and the first environmental data; and performing missing value processing, continuous constant value processing, mutation outlier processing and trend outlier processing on the fourth training data set in sequence to obtain the first training data set.

[0008] Optionally, the fourth training data set is sequentially subjected to missing value processing and continuous constant value processing, including: traversing the fourth training data set to determine the null value or NAN value of the sample data therein, and determining it as the missing value; replacing the missing value based on a first preset value to complete the missing value processing; calculating the difference between the sample data at each adjacent moment to obtain a first power increment, summing the adjacent first power increments, and when the second power increment is less than a first threshold, determining the sample data corresponding to the second power increment as the continuous constant value; when the duration corresponding to the continuous constant value is less than a second threshold, processing is performed based on the linear filling method, and when the duration corresponding to the continuous constant value is greater than or equal to the second threshold, processing is performed based on the similarity day method.

[0009] Optionally, performing mutation outlier processing on the fourth training data set includes: based on a preset formula The sample data in the fourth training data set are screened to obtain the mutation outlier value, where prob() is a probability function, ΔP t is the change in sample data corresponding to adjacent moments, and E is ΔP t The corresponding expectation, k represents the degree to which the sample deviates from the expectation; the mutation outliers are deleted and filled based on linear interpolation.

[0010] Optionally, trend outlier processing is performed on the fourth training data set, including: determining the changing trend of irradiance based on the irradiance in the first environmental data in the fourth training data set, and repairing the changing trend of the third power generation power or the fourth power generation power based on the linear complement method when the changing trend of the third power generation power or the fourth power generation power in the fourth training data set is opposite to the changing trend of the irradiance.

[0011] Optionally, feature screening is performed on the first training data set according to a recursive feature elimination method to obtain a second training data set, including: taking the first training data set as an initial feature set, testing the performance of the preset learner based on the initial feature set to obtain a performance score; when the performance score of the preset learner is less than a third threshold, performing a feature score according to the weight coefficient of each feature in the initial feature set to obtain a target feature score; arranging each feature from small to large based on the target feature score and deleting a preset number of features; re-testing the performance of the preset learner based on the updated initial feature set and updating the initial feature set until the performance score is greater than or equal to the third threshold, and determining the initial feature set as the second training data set.

[0012] Optionally, feature extraction is performed on the second training data set based on a factor analysis method to obtain a third training data set, including: constructing a sample sequence based on the second training data set, wherein each row of the sample sequence represents a sample, and each column of the sample sequence represents a characteristic variable; calculating a correlation coefficient matrix between each of the characteristic variables, performing factor analysis based on the correlation coefficient matrix, and determining a first target factor; expressing the characteristic variables as a linear combination of the common first target factors, and obtaining factor loadings of each of the characteristic variables on each common factor; screening the main factors corresponding to each of the characteristic variables based on the factor loadings, and determining the characteristic variables whose factor loadings corresponding to the corresponding main factors are greater than a fourth threshold as target variables; and constructing the third training data set based on the target variables.

[0013] According to another aspect of the present application, a method for predicting the power generation of a photovoltaic system based on an edge cluster is provided, and the device is applicable to a distributed photovoltaic power generation system, and the distributed photovoltaic power generation system includes a metering system, a scheduling management system and a meteorological system, and the device includes: a first acquisition unit, for acquiring the power generation power and location information of the distributed photovoltaic from the metering system at intervals of a first preset time, to obtain the first power generation power and the first location information, acquiring the power generation power of the distributed photovoltaic directly connected to the grid on the user side from the metering system, to obtain the second power generation power, acquiring the feeder corresponding to the distributed photovoltaic directly connected to the grid on the user side from the scheduling management system, to obtain the second location information, acquiring the maintenance record from the scheduling management system, to obtain the first maintenance plan, acquiring the environmental data from the meteorological system, to obtain the first environmental data, pre-processing the first power generation power, the first location information, the second power generation power, the second location information, the first maintenance plan and the first environmental data, to obtain A first training data set, wherein the preprocessing includes at least data cleaning, data clipping and data completion; a first calculation unit, used to perform feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set; a second calculation unit, used to perform feature extraction on the second training data set based on a factor analysis method to obtain a third training data set; a prediction unit, used to perform prediction model training based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is used to predict the power generation power of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data; obtain target prediction parameters, input the target prediction parameters into the power generation prediction model to obtain target power generation power, wherein the target prediction parameters include at least one of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data within the past first preset time period.

[0014] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the methods described.

[0015] According to another aspect of the present application, a monitoring system is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of the methods described.

[0016] Applying the technical solution of the present application, in the above-mentioned photovoltaic system power generation prediction method based on edge cluster, first, at each first preset time interval, the power generation power and location information of the distributed photovoltaic are obtained from the above-mentioned metering system respectively to obtain the first power generation power and the first location information, the power generation power of the distributed photovoltaic directly connected to the grid on the user side is obtained from the above-mentioned metering system to obtain the second power generation power, the feeder corresponding to the above-mentioned distributed photovoltaic directly connected to the grid on the user side is obtained from the above-mentioned dispatching and management system to obtain the second location information, the maintenance record is obtained from the above-mentioned dispatching and management system to obtain the first maintenance plan, the environmental data is obtained from the above-mentioned meteorological system to obtain the first environmental data, the first power generation power, the first location information, the second power generation power, the second location information, the first maintenance plan and the first environmental data are preprocessed to obtain a first training data set, and the above-mentioned preprocessing at least includes data cleaning and data trimming. Cut and complete the data; then, perform feature screening on the first training data set according to the recursive feature elimination method to obtain a second training data set; then, perform feature extraction on the second training data set based on the factor analysis method to obtain a third training data set; then, perform prediction model training based on the third training data set to obtain a power generation prediction model, which is used to predict the power generation power of the distributed photovoltaic power generation system within the second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data; finally, obtain the target prediction parameters, input the target prediction parameters into the power generation prediction model to obtain the target power generation power, which includes at least one of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data within the past first preset time period. This application proposes a distributed photovoltaic power generation power prediction method based on edge groups, which uses power generation power, meteorological data and maintenance plan data as input to predict the distributed photovoltaic power generation power converged to different substations and feeders, thereby achieving accurate prediction of distributed photovoltaic power generation power, and solving the problem in the prior art of lacking a distributed photovoltaic power generation power prediction method based on multidimensional feature data. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The following is a hardware structure block diagram of a mobile terminal according to a photovoltaic system power generation prediction method based on edge clusters provided in an embodiment of the present application;

[0018] Figure 2 A schematic diagram of a process for predicting photovoltaic system power generation based on an edge cluster according to an embodiment of the present application is shown;

[0019] Figure 3A structural block diagram of a photovoltaic system power generation prediction device based on edge clusters provided according to an embodiment of the present application is shown.

[0020] The above drawings include the following reference numerals:

[0021] 102. Processor; 104. Memory; 106. Transmission device; 108. Input / output device. DETAILED DESCRIPTION

[0022] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0023] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] As introduced in the background technology, the existing technology lacks a distributed photovoltaic power generation prediction method based on an edge cluster system, and a method for predicting short-term power generation within different ranges based on multi-dimensional input data. In order to solve the problem that the existing technology lacks a distributed photovoltaic power generation prediction method based on multi-dimensional feature data, the embodiments of the present application provide a photovoltaic system power generation prediction method, device, computer-readable storage medium and monitoring system based on an edge cluster.

[0026] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0027] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal for a photovoltaic system power generation prediction method based on edge clusters according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown) a processor 102 (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA and other processing devices) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0028] The memory 104 can be used to store computer programs, such as software programs and modules of application software, such as the computer program corresponding to the device information display method in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implementing the above-mentioned method. The memory 104 may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned networks include but are not limited to the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include a wireless network provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0029] In this embodiment, a method for predicting photovoltaic system power generation based on an edge cluster that runs on a mobile terminal, a computer terminal, or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0030] Figure 2 FIG is a flow chart of a photovoltaic system power generation prediction method based on edge cluster according to an embodiment of the present application. Figure 2 As shown, the method includes the following steps:

[0031] Step S201: at intervals of a first preset time, respectively obtain the generated power and location information of the distributed photovoltaic system from the metering system to obtain the first generated power and the first location information; obtain the generated power of the distributed photovoltaic system directly connected to the grid on the user side from the metering system to obtain the second generated power; obtain the feeder corresponding to the distributed photovoltaic system directly connected to the grid on the user side from the dispatching and management system to obtain the second location information; obtain the maintenance record from the dispatching and management system to obtain the first maintenance plan; obtain the environmental data from the meteorological system to obtain the first environmental data; pre-process the first generated power, the first location information, the second generated power, the second location information, the first maintenance plan, and the first environmental data to obtain a first training data set, wherein the pre-processing at least includes data cleaning, data clipping, and data completion;

[0032] Specifically, distributed power generation data and the distribution of distributed photovoltaics are obtained from the metering system on a daily basis to obtain the first power generation and the first location information. The power generation of distributed photovoltaics directly connected to the grid on the user side is obtained from the metering system on a daily basis to obtain the second power generation. The feeders corresponding to the distributed photovoltaics are obtained from the dispatching and management system to obtain the second location information. Furthermore, a maintenance plan is obtained from the dispatching and management system to obtain the first maintenance plan. Environmental data is simultaneously obtained from the meteorological system to obtain the first environmental data. Preprocessing is then performed using the distributed photovoltaic power generation prediction data processing algorithm configured in this application. The processed distributed photovoltaic power generation data and meteorological data are merged according to time to obtain a multidimensional dataset, thereby obtaining the first training dataset.

[0033] Step S202, performing feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set;

[0034] Specifically, feature selection is based on the recursive feature elimination method to select important features of distributed photovoltaic historical power data, gridded meteorological data and power grid maintenance planning, and new features are generated based on the important features to enrich the data dimension of the training data set to obtain the above-mentioned second training data set.

[0035] Step S203, performing feature extraction on the second training data set based on factor analysis to obtain a third training data set;

[0036] Specifically, factor analysis is used to fully explore the intrinsic structure of the feature data from the selected features, further derive new features, and perform further changes based on mathematical principles to provide more features for subsequent model training, thereby obtaining the third data set mentioned above.

[0037] Step S204: training a prediction model based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is used to predict the power generation of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan, and the first environmental data;

[0038] Specifically, model training is performed based on the data in the third training data set to obtain a model for predicting power generation based on multi-dimensional data.

[0039] Step S205, obtain target prediction parameters, input the above target prediction parameters into the above power generation prediction model to obtain target power generation power, the above target prediction parameters include at least one of the above first power generation power, the above second power generation power, the above first maintenance plan and the above first environmental data within the above first preset time period in the past.

[0040] Specifically, relevant data is obtained from the above-mentioned metering system, scheduling management system and meteorological system at preset intervals to obtain the above-mentioned target prediction parameters, and then the above-mentioned target prediction parameters are input into the above-mentioned power generation prediction model to predict the power generation power and obtain the above-mentioned target power generation power.

[0041] Through this embodiment, first, at every first preset time interval, the power generation power and location information of the distributed photovoltaic are respectively obtained from the above-mentioned metering system to obtain the first power generation power and the first location information, the power generation power of the distributed photovoltaic directly connected to the grid on the user side is obtained from the above-mentioned metering system to obtain the second power generation power, the feeder corresponding to the above-mentioned distributed photovoltaic directly connected to the grid on the user side is obtained from the above-mentioned dispatching and management system to obtain the second location information, the maintenance record is obtained from the above-mentioned dispatching and management system to obtain the first maintenance plan, the environmental data is obtained from the above-mentioned meteorological system to obtain the first environmental data, the first power generation power, the first location information, the second power generation power, the second location information, the first maintenance plan and the first environmental data are preprocessed to obtain a first training data set, and the preprocessing at least includes data cleaning, data clipping and data completion; then, according to the iteration The feature elimination method is used to perform feature screening on the first training data set to obtain a second training data set; then, the feature extraction is performed on the second training data set based on the factor analysis method to obtain a third training data set; then, a prediction model is trained based on the third training data set to obtain a power generation prediction model, which is used to predict the power generation power of the distributed photovoltaic power generation system within the second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data; finally, the target prediction parameters are obtained, and the target prediction parameters are input into the power generation prediction model to obtain the target power generation power, which includes at least one of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data within the past first preset time period. This application proposes a distributed photovoltaic power generation power prediction method based on edge groups, which uses power generation power, meteorological data and maintenance plan data as input to predict the distributed photovoltaic power generation power converged to different substations and feeders, thereby achieving accurate prediction of distributed photovoltaic power generation power, and solving the problem in the prior art of lacking a distributed photovoltaic power generation power prediction method based on multi-dimensional feature data.

[0042] In order to obtain the first training data set, in an optional implementation, step S201 includes:

[0043] Step S2011, determining the substation corresponding to the first power generation power based on the first location information, and summing the first power generation powers corresponding to the substations to obtain the corresponding third power generation power;

[0044] Specifically, based on the first generated power and the first location information, the first generated powers of all distributed photovoltaics under the substation are accumulated to obtain the third generated power.

[0045] Step S2012: determining the substation to which the feeder corresponding to the second generated power belongs based on the second location information, and accumulating the generated power corresponding to each feeder according to the relationship between the feeder and the substation, the second generated power, and the third generated power to obtain a corresponding fourth generated power;

[0046] Specifically, based on the correspondence between distributed photovoltaics and substations, and the subordinate relationship between feeders and substations, the distributed power generation gathered in the substations is decomposed, and then the power generation gathered in the feeders is accumulated to obtain the fourth power generation mentioned above.

[0047] Step S2013: constructing a fourth training data set based on the third generated power, the fourth generated power, the first maintenance plan, and the first environmental data;

[0048] Specifically, the third generated power, the fourth generated power, the first maintenance plan and the first environmental data are stored in a database for maintenance to obtain the fourth training data set.

[0049] Step S2014 , sequentially performing missing value processing, continuous constant value processing, sudden change outlier processing, and trend outlier processing on the fourth training dataset to obtain the first training dataset.

[0050] Specifically, the fourth training data set is subjected to identification and filling of abnormal data such as missing values, sudden change values, and continuous constant values ​​to obtain a training data set from which redundant data is removed, namely, the first training data set.

[0051] In order to obtain the first training data set, in an optional implementation, step S2014 includes:

[0052] Step S20141, traversing the fourth training data set to determine null values ​​or NAN values ​​of sample data therein, and determining them as the missing values;

[0053] Specifically, the null value or NAN value of the fourth training data and the sample data is detected, and the detection result is determined as the missing value.

[0054] Step S20142: replacing the missing value based on a first preset value to complete the missing value processing;

[0055] Specifically, all the missing values ​​are filled with 0.

[0056] It is understandable that the sample data filled with 0 values ​​will become other types of abnormal data, and then further processed by other means.

[0057] Step S20143: Calculate the difference between sample data at adjacent moments to obtain a first power increment. If the sum of adjacent first power increments is less than a first threshold, determine the sample data corresponding to the second power increment as the continuous constant value.

[0058] Specifically, for continuous constant values, the load difference between two adjacent moments is calculated to obtain the load increment, and then the adjacent load increments are summed up in pairs. When the summation result is less than or equal to a certain minimum value (the first threshold mentioned above), the sample data of the three moments corresponding to the summation result are determined to be continuous constant values.

[0059] Step S20144: When the duration corresponding to the above-mentioned continuous constant value is less than the second threshold, processing is performed based on the linear filling method; when the duration corresponding to the above-mentioned continuous constant value is greater than or equal to the above-mentioned second threshold, processing is performed based on the similar day method.

[0060] Specifically, when the time points corresponding to the continuous constant values ​​are less than five, correction is performed based on linear filling; when the time points corresponding to the continuous constant values ​​are greater than or equal to five, filling is performed based on the similar day method.

[0061] In order to obtain the first training data set, in an optional implementation, step S2014 further includes:

[0062] Step S20145, based on the preset formula The sample data in the fourth training data set are screened to obtain the above-mentioned mutation outlier value, where prob() is the probability function, ΔP t is the change in sample data corresponding to adjacent moments, and E is ΔP t Corresponding to the expectation, k represents the degree to which the sample deviates from the expectation;

[0063] Specifically, for mutation outliers, Chebyshev's inequality is used for detection. Chebyshev's inequality is applicable to data of any distribution. Based on the above formula, the probability that the sample deviates from the expected k times the standard deviation is less than

[0064] In the specific implementation, if the value of k is 5, at least 96% of the samples should be distributed within the ±5D interval around the mean, and those points that deviate from the incremental mean outside the 5D interval are regarded as mutation outliers.

[0065] Step S20146: Delete the above-mentioned mutation outliers and fill them based on linear interpolation.

[0066] Specifically, the above-mentioned mutation outliers are deleted and filled based on the linear interpolation method.

[0067] In order to obtain the first training data set, in an optional implementation, step S2014 further includes:

[0068] Step S20147, based on the irradiance in the above-mentioned first environmental data in the above-mentioned fourth training data set, determine the changing trend of the irradiance. When the changing trend of the above-mentioned third power generation or the above-mentioned fourth power generation in the above-mentioned fourth training data set is opposite to the changing trend of the irradiance, the changing trend of the above-mentioned third power generation or the above-mentioned fourth power generation is repaired based on the linear complement method.

[0069] Specifically, taking the irradiance data as a benchmark, the upward or downward trends of the power generation data and the meteorological irradiance data are compared, and the power generation data that are opposite to the irradiance trend are repaired. For example, at a certain moment, the irradiance data is on an upward trend, but the power generation data is on a downward trend. At this time, the power generation data can be regarded as a data anomaly and repaired using linear filling.

[0070] In a specific embodiment, gridded meteorological information is obtained from the meteorological system on a daily basis. The obtained meteorological information includes temperature, humidity, rainfall, air pressure, etc., and the data time interval is 15 minutes. A distributed photovoltaic power generation prediction data processing algorithm is used to detect and fill in abnormal points in the meteorological data to obtain meteorological data.

[0071] Processing of meteorological data such as irradiance, temperature, and rainfall, including detection and filling of missing value anomalies, continuous constant values, and sudden change anomalies.

[0072] In order to obtain the second training data set, in an optional implementation, step S202 includes:

[0073] Step S2021, using the first training data set as an initial feature set, testing the performance of the preset learner based on the initial feature set to obtain a performance score;

[0074] Specifically, the recursive feature elimination method is a model-based feature selection method. First, a preset learner is constructed based on the initial features of the first training dataset. In one embodiment, the preset learner can be constructed based on a logistic regression algorithm. The initial performance of the preset learner is then determined based on the initial feature set to obtain the aforementioned performance score.

[0075] Step S2022: If the performance score of the preset learner is less than a third threshold, perform feature scoring based on the weight coefficients of each feature in the initial feature set to obtain a target feature score;

[0076] Specifically, when the performance of the preset learner does not meet the standard (the performance score of the preset learner is less than the third threshold), a feature score is performed according to the weight coefficient of each feature in the initial feature set to obtain the target feature score.

[0077] Step S2023, arranging the features in ascending order based on the target feature scores and deleting a preset number of the features;

[0078] Specifically, feature screening is performed based on the target feature score by deleting several features corresponding to smaller performance scores. First, the features are arranged from small to large based on the target feature score and a preset number of features are deleted in sequence, that is, a preset number of features are eliminated.

[0079] Step S2024: re-perform the performance test of the preset learner based on the updated initial feature set and update the initial feature set until the performance score is greater than or equal to the third threshold, and determine the initial feature set as the second training data set.

[0080] Specifically, the screening process is repeated until the number of features reaches a threshold or the model performance index meets the preset requirements.

[0081] In order to obtain the third training data set, in an optional implementation, step S203 includes:

[0082] Step S2031: construct a sample sequence based on the second training data set, where each row of the sample sequence represents a sample, and each column of the sample sequence represents a feature variable;

[0083] Specifically, the correlation between features can be eliminated, the intrinsic structure of the dataset can be extracted, and feature compression can be achieved while maintaining information integrity. First, a sample sequence is constructed based on the second training dataset, where each row represents a sample, and each column of the sample sequence represents a feature variable.

[0084] Step S2032: Calculate the correlation coefficient matrix between the above-mentioned characteristic variables, perform factor analysis based on the above-mentioned correlation coefficient matrix, and determine the first target factor;

[0085] Specifically, the correlation coefficient matrix between the characteristic variables is calculated, and factor analysis is performed based on the correlation coefficient matrix to determine the common factors, thereby obtaining the above-mentioned first target factor.

[0086] It can be understood that the common factors reflect the inherent correlation structure of the original characteristic variables.

[0087] Step S2033: Express the characteristic variables as a linear combination of the common first target factor, and obtain the factor loading of each characteristic variable on each common factor;

[0088] Specifically, each original characteristic variable is expressed as a linear combination of common factors, and the factor loading of each characteristic on each common factor is obtained.

[0089] Step S2034, screening the main factors corresponding to the characteristic variables based on the factor loadings, and determining the characteristic variables whose factor loadings corresponding to the main factors are greater than a fourth threshold as target variables;

[0090] Step S2035: construct the third training data set based on the target variable.

[0091] Specifically, based on the size of the factor loading, we determine which common factors each feature is mainly related to, and select the original features with larger factor loadings on the main factors as the new feature set.

[0092] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0093] The embodiment of the present application also provides a photovoltaic system power generation prediction device based on an edge cluster. It should be noted that the photovoltaic system power generation prediction device based on an edge cluster in the embodiment of the present application can be used to execute the photovoltaic system power generation prediction method based on an edge cluster provided in the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation methods, and those that have been explained will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and conceived.

[0094] The following introduces the photovoltaic system power generation prediction device based on edge cluster provided in an embodiment of the present application.

[0095] Figure 3 This is a structural block diagram of a photovoltaic system power generation prediction device based on edge clusters according to an embodiment of the present application. Figure 3 As shown, the device includes:

[0096] A first acquisition unit 10 is configured to obtain, at intervals of a first preset time duration, the generated power and location information of the distributed photovoltaic system from the metering system to obtain the first generated power and the first location information, obtain the generated power of the distributed photovoltaic system directly connected to the grid on the user side from the metering system to obtain the second generated power, obtain the feeder corresponding to the distributed photovoltaic system directly connected to the grid on the user side from the dispatching and management system to obtain the second location information, obtain the maintenance record from the dispatching and management system to obtain the first maintenance plan, obtain the environmental data from the meteorological system to obtain the first environmental data, and pre-process the first generated power, the first location information, the second generated power, the second location information, the first maintenance plan, and the first environmental data to obtain a first training data set, wherein the pre-processing at least includes data cleaning, data clipping, and data completion;

[0097] Specifically, distributed power generation data and the distribution of distributed photovoltaics are obtained from the metering system on a daily basis to obtain the first power generation and the first location information. The power generation of distributed photovoltaics directly connected to the grid on the user side is obtained from the metering system on a daily basis to obtain the second power generation. The feeders corresponding to the distributed photovoltaics are obtained from the dispatching and management system to obtain the second location information. Furthermore, a maintenance plan is obtained from the dispatching and management system to obtain the first maintenance plan. Environmental data is simultaneously obtained from the meteorological system to obtain the first environmental data. Preprocessing is then performed using the distributed photovoltaic power generation prediction data processing algorithm configured in this application. The processed distributed photovoltaic power generation data and meteorological data are merged according to time to obtain a multidimensional dataset, thereby obtaining the first training dataset.

[0098] A first computing unit 20 is configured to perform feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set;

[0099] Specifically, feature selection is based on the recursive feature elimination method to select important features of distributed photovoltaic historical power data, gridded meteorological data and power grid maintenance planning, and new features are generated based on the important features to enrich the data dimension of the training data set to obtain the above-mentioned second training data set.

[0100] A second computing unit 30 is configured to perform feature extraction on the second training data set based on a factor analysis method to obtain a third training data set;

[0101] Specifically, factor analysis is used to fully explore the intrinsic structure of the feature data from the selected features, further derive new features, and perform further changes based on mathematical principles to provide more features for subsequent model training, thereby obtaining the third data set mentioned above.

[0102] a training unit 40 configured to perform prediction model training based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is configured to predict the power generation of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation, the second power generation, the first maintenance plan, and the first environmental data;

[0103] Specifically, model training is performed based on the data in the third training data set to obtain a model for predicting power generation based on multi-dimensional data.

[0104] The prediction unit 50 is used to obtain target prediction parameters, input the above target prediction parameters into the above power generation prediction model, and obtain the target power generation power. The above target prediction parameters include at least one of the above first power generation power, the above second power generation power, the above first maintenance plan and the above first environmental data within the above first preset time period in the past.

[0105] Specifically, relevant data is obtained from the above-mentioned metering system, scheduling management system and meteorological system at preset intervals to obtain the above-mentioned target prediction parameters, and then the above-mentioned target prediction parameters are input into the above-mentioned power generation prediction model to predict the power generation power and obtain the above-mentioned target power generation power.

[0106] Through this embodiment, the first acquisition unit obtains the generated power and location information of the distributed photovoltaic from the above-mentioned metering system at intervals of a first preset time length to obtain the first generated power and the first location information, obtains the generated power of the distributed photovoltaic directly connected to the grid on the user side from the above-mentioned metering system to obtain the second generated power, obtains the feeder corresponding to the above-mentioned distributed photovoltaic directly connected to the grid on the user side from the above-mentioned dispatching and management system to obtain the second location information, obtains the maintenance record from the above-mentioned dispatching and management system to obtain the first maintenance plan, obtains the environmental data from the above-mentioned meteorological system to obtain the first environmental data, pre-processes the above-mentioned first generated power, the above-mentioned first location information, the above-mentioned second generated power, the above-mentioned second location information, the above-mentioned first maintenance plan and the above-mentioned first environmental data to obtain a first training data set, and the above-mentioned pre-processing at least includes data cleaning, data clipping and data completion; the first calculation unit calculates the generated power according to the iteration. The feature elimination method performs feature screening on the first training data set to obtain a second training data set; the second calculation unit performs feature extraction on the second training data set based on the factor analysis method to obtain a third training data set; the training unit performs prediction model training based on the third training data set to obtain a power generation prediction model, and the power generation prediction model is used to predict the power generation of the distributed photovoltaic power generation system within the second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data; the prediction unit obtains the target prediction parameter, inputs the target prediction parameter into the power generation prediction model to obtain the target power generation, and the target prediction parameter includes at least one of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data within the past first preset time period. The present application proposes a distributed photovoltaic power generation power prediction method based on edge groups, which uses power generation power, meteorological data and maintenance plan data as input to predict the distributed photovoltaic power generation power converged to different substations and feeders, thereby achieving accurate prediction of distributed photovoltaic power generation, and solving the problem in the prior art of lacking a distributed photovoltaic power generation power prediction method based on multi-dimensional feature data.

[0107] In order to obtain the first training data set, in an optional embodiment, the first obtaining unit includes:

[0108] A first determining module is configured to determine a substation corresponding to the first generated power based on the first location information, and to sum the first generated powers corresponding to the substations to obtain a corresponding third generated power;

[0109] Specifically, based on the first generated power and the first location information, the first generated powers of all distributed photovoltaics under the substation are accumulated to obtain the third generated power.

[0110] a second determining module, configured to determine, based on the second location information, the substation to which the feeder corresponding to the second generated power belongs, and to accumulate the generated power corresponding to each of the feeders according to the relationship between the feeder and the substation, the second generated power, and the third generated power, to obtain a corresponding fourth generated power;

[0111] Specifically, based on the correspondence between distributed photovoltaics and substations, and the subordinate relationship between feeders and substations, the distributed power generation gathered in the substations is decomposed, and then the power generation gathered in the feeders is accumulated to obtain the fourth power generation mentioned above.

[0112] A first construction module is configured to construct a fourth training data set based on the third generated power, the fourth generated power, the first maintenance plan, and the first environmental data;

[0113] Specifically, the third generated power, the fourth generated power, the first maintenance plan and the first environmental data are stored in a database for maintenance to obtain the fourth training data set.

[0114] The first processing module is used to sequentially perform missing value processing, continuous constant value processing, mutation outlier processing, and trend outlier processing on the fourth training data set to obtain the first training data set.

[0115] Specifically, the fourth training data set is subjected to identification and filling of abnormal data such as missing values, sudden change values, and continuous constant values ​​to obtain a training data set from which redundant data is removed, namely, the first training data set.

[0116] In order to obtain the first training data set, in an optional embodiment, the first processing module includes:

[0117] A first determining submodule is configured to traverse the fourth training data set to determine null values ​​or NAN values ​​of sample data therein as the missing values;

[0118] Specifically, the null value or NAN value of the fourth training data and the sample data is detected, and the detection result is determined as the missing value.

[0119] A first processing submodule is configured to replace the missing value based on a first preset value to complete the missing value processing;

[0120] Specifically, all the missing values ​​are filled with 0.

[0121] It is understandable that the sample data filled with 0 values ​​will become other types of abnormal data, and then further processed by other means.

[0122] a second determining submodule, configured to calculate the difference between sample data at adjacent moments to obtain a first power increment, sum the adjacent first power increments, and if the second power increment is less than a first threshold, determine the sample data corresponding to the second power increment as the continuous constant value;

[0123] Specifically, for continuous constant values, the load difference between two adjacent moments is calculated to obtain the load increment, and then the adjacent load increments are summed up in pairs. When the summation result is less than or equal to a certain minimum value (the first threshold mentioned above), the sample data of the three moments corresponding to the summation result are determined to be continuous constant values.

[0124] The second processing submodule is used to process based on the linear filling method when the duration corresponding to the above-mentioned continuous constant value is less than the second threshold, and to process based on the similar day method when the duration corresponding to the above-mentioned continuous constant value is greater than or equal to the above-mentioned second threshold.

[0125] Specifically, when the time points corresponding to the continuous constant values ​​are less than five, correction is performed based on linear filling; when the time points corresponding to the continuous constant values ​​are greater than or equal to five, filling is performed based on the similar day method.

[0126] In order to obtain the first training data set, in an optional embodiment, the first processing module further includes:

[0127] The third processing submodule is used to process the The sample data in the fourth training data set are screened to obtain the above-mentioned mutation outlier value, where prob() is the probability function, ΔP t is the change in sample data corresponding to adjacent moments, and E is ΔP t Corresponding to the expectation, k represents the degree to which the sample deviates from the expectation;

[0128] Specifically, for mutation outliers, Chebyshev's inequality is used for detection. Chebyshev's inequality is applicable to data of any distribution. Based on the above formula, the probability that the sample deviates from the expected k times the standard deviation is less than

[0129] In the specific implementation, if the value of k is 5, at least 96% of the samples should be distributed within the ±5D interval around the mean, and those points that deviate from the incremental mean outside the 5D interval are regarded as mutation outliers.

[0130] The fourth processing submodule is used to delete the above-mentioned mutation outliers and fill them based on linear interpolation.

[0131] Specifically, the above-mentioned mutation outliers are deleted and filled based on the linear interpolation method.

[0132] In order to obtain the first training data set, in an optional embodiment, the first processing module further includes:

[0133] The fifth processing submodule is used to determine the changing trend of irradiance based on the irradiance in the above-mentioned first environmental data in the above-mentioned fourth training data set, and when the changing trend of the above-mentioned third power generation or the above-mentioned fourth power generation in the above-mentioned fourth training data set is opposite to the changing trend of irradiance, the changing trend of the above-mentioned third power generation or the above-mentioned fourth power generation is repaired based on the linear complement method.

[0134] Specifically, taking the irradiance data as a benchmark, the upward or downward trends of the power generation data and the meteorological irradiance data are compared, and the power generation data that are opposite to the irradiance trend are repaired. For example, at a certain moment, the irradiance data is on an upward trend, but the power generation data is on a downward trend. At this time, the power generation data can be regarded as a data anomaly and repaired using linear filling.

[0135] In a specific embodiment, gridded meteorological information is obtained from the meteorological system on a daily basis. The obtained meteorological information includes temperature, humidity, rainfall, air pressure, etc., and the data time interval is 15 minutes. A distributed photovoltaic power generation prediction data processing algorithm is used to detect and fill in abnormal points in the meteorological data to obtain meteorological data.

[0136] Processing of meteorological data such as irradiance, temperature, and rainfall, including detection and filling of missing value anomalies, continuous constant values, and sudden change anomalies.

[0137] In order to obtain the second training data set, in an optional embodiment, the first calculation unit includes:

[0138] A testing module, configured to use the first training data set as an initial feature set, and to test the performance of the preset learner based on the initial feature set to obtain a performance score;

[0139] Specifically, the recursive feature elimination method is a model-based feature selection method. First, a preset learner is constructed based on the initial features of the first training dataset. In one embodiment, the preset learner can be constructed based on a logistic regression algorithm. The initial performance of the preset learner is then determined based on the initial feature set to obtain the aforementioned performance score.

[0140] A second processing module is configured to, when the performance score of the preset learner is less than a third threshold, perform feature scoring based on the weight coefficients of each feature in the initial feature set to obtain a target feature score;

[0141] Specifically, when the performance of the preset learner does not meet the standard (the performance score of the preset learner is less than the third threshold), a feature score is performed according to the weight coefficient of each feature in the initial feature set to obtain the target feature score.

[0142] A third processing module is used to arrange the above features in ascending order based on the above target feature scores and delete a preset number of the above features;

[0143] Specifically, feature screening is performed based on the target feature score by deleting several features corresponding to smaller performance scores. First, the features are arranged from small to large based on the target feature score and a preset number of features are deleted in sequence, that is, a preset number of features are eliminated.

[0144] The third determination module is used to re-perform the performance test of the preset learner based on the updated initial feature set and update the initial feature set until the performance score is greater than or equal to the third threshold, and determine the initial feature set as the second training data set.

[0145] Specifically, the screening process is repeated until the number of features reaches a threshold or the model performance index meets the preset requirements.

[0146] In order to obtain the third training data set, in an optional embodiment, the second calculation unit includes:

[0147] A second construction module is configured to construct a sample sequence based on the second training data set, wherein each row of the sample sequence represents a sample, and each column of the sample sequence represents a feature variable;

[0148] Specifically, the correlation between features can be eliminated, the intrinsic structure of the dataset can be extracted, and feature compression can be achieved while maintaining information integrity. First, a sample sequence is constructed based on the second training dataset, where each row represents a sample, and each column of the sample sequence represents a feature variable.

[0149] A first calculation module is used to calculate the correlation coefficient matrix between the above-mentioned characteristic variables, perform factor analysis based on the above-mentioned correlation coefficient matrix, and determine the first target factor;

[0150] Specifically, the correlation coefficient matrix between the characteristic variables is calculated, and factor analysis is performed based on the correlation coefficient matrix to determine the common factors, thereby obtaining the above-mentioned first target factor.

[0151] It can be understood that the common factors reflect the inherent correlation structure of the original characteristic variables.

[0152] A second calculation module is used to express the characteristic variables as a linear combination of the common first target factor to obtain the factor loading of each characteristic variable on each common factor;

[0153] Specifically, each original characteristic variable is expressed as a linear combination of common factors, and the factor loading of each characteristic on each common factor is obtained.

[0154] a third calculation module, configured to screen the main factors corresponding to the characteristic variables based on the factor loadings, and determine the characteristic variables whose factor loadings corresponding to the main factors are greater than a fourth threshold as target variables;

[0155] The third construction module is used to construct the third training data set based on the target variable.

[0156] Specifically, based on the size of the factor loading, we determine which common factors each feature is mainly related to, and select the original features with larger factor loadings on the main factors as the new feature set.

[0157] The aforementioned edge cluster-based photovoltaic system power generation prediction device includes a processor and memory. The aforementioned first acquisition unit, first calculation unit, second calculation unit, training unit, and prediction unit are all stored as program units in the memory. The processor executes these program units stored in the memory to implement the corresponding functions. All of the aforementioned modules are located in the same processor; alternatively, the aforementioned modules can be located in different processors in any combination.

[0158] The processor contains a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the prediction accuracy can be improved by adjusting the kernel parameters.

[0159] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0160] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program. When the program is running, the device where the computer-readable storage medium is located is controlled to execute the photovoltaic system power generation power prediction method based on the edge cluster.

[0161] Specifically, the photovoltaic system power generation prediction method based on edge cluster includes:

[0162] Step S201: at intervals of a first preset time, respectively obtain the generated power and location information of the distributed photovoltaic system from the metering system to obtain the first generated power and the first location information; obtain the generated power of the distributed photovoltaic system directly connected to the grid on the user side from the metering system to obtain the second generated power; obtain the feeder corresponding to the distributed photovoltaic system directly connected to the grid on the user side from the dispatching and management system to obtain the second location information; obtain the maintenance record from the dispatching and management system to obtain the first maintenance plan; obtain the environmental data from the meteorological system to obtain the first environmental data; pre-process the first generated power, the first location information, the second generated power, the second location information, the first maintenance plan, and the first environmental data to obtain a first training data set, wherein the pre-processing at least includes data cleaning, data clipping, and data completion;

[0163] Step S202, performing feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set;

[0164] Step S203, performing feature extraction on the second training data set based on factor analysis to obtain a third training data set;

[0165] Step S204: training a prediction model based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is used to predict the power generation of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan, and the first environmental data;

[0166] Step S205: Obtain target prediction parameters and input them into the power generation prediction model to obtain a target power generation. The target prediction parameters include at least one of the first power generation, the second power generation, the first maintenance plan, and the first environmental data within the first preset period of time. An embodiment of the present invention provides a processor configured to execute a program, wherein the program, when executed, executes the method for predicting photovoltaic system power generation based on an edge cluster.

[0167] Specifically, the photovoltaic system power generation prediction method based on edge cluster includes:

[0168] Step S201: at intervals of a first preset time, respectively obtain the generated power and location information of the distributed photovoltaic system from the metering system to obtain the first generated power and the first location information; obtain the generated power of the distributed photovoltaic system directly connected to the grid on the user side from the metering system to obtain the second generated power; obtain the feeder corresponding to the distributed photovoltaic system directly connected to the grid on the user side from the dispatching and management system to obtain the second location information; obtain the maintenance record from the dispatching and management system to obtain the first maintenance plan; obtain the environmental data from the meteorological system to obtain the first environmental data; pre-process the first generated power, the first location information, the second generated power, the second location information, the first maintenance plan, and the first environmental data to obtain a first training data set, wherein the pre-processing at least includes data cleaning, data clipping, and data completion;

[0169] Step S202, performing feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set;

[0170] Step S203, performing feature extraction on the second training data set based on factor analysis to obtain a third training data set;

[0171] Step S204: training a prediction model based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is used to predict the power generation of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan, and the first environmental data;

[0172] Step S205: Obtain target prediction parameters and input the target prediction parameters into the power generation prediction model to obtain target power generation. The target prediction parameters include at least one of the first power generation, the second power generation, the first maintenance plan, and the first environmental data within the past first preset period of time. An embodiment of the present invention provides a monitoring system comprising a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs at least the following steps:

[0173] Step S201: at intervals of a first preset time, respectively obtain the generated power and location information of the distributed photovoltaic system from the metering system to obtain the first generated power and the first location information; obtain the generated power of the distributed photovoltaic system directly connected to the grid on the user side from the metering system to obtain the second generated power; obtain the feeder corresponding to the distributed photovoltaic system directly connected to the grid on the user side from the dispatching and management system to obtain the second location information; obtain the maintenance record from the dispatching and management system to obtain the first maintenance plan; obtain the environmental data from the meteorological system to obtain the first environmental data; pre-process the first generated power, the first location information, the second generated power, the second location information, the first maintenance plan, and the first environmental data to obtain a first training data set, wherein the pre-processing at least includes data cleaning, data clipping, and data completion;

[0174] Step S202, performing feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set;

[0175] Step S203, performing feature extraction on the second training data set based on factor analysis to obtain a third training data set;

[0176] Step S204: training a prediction model based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is used to predict the power generation of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan, and the first environmental data;

[0177] Step S205: Obtain target prediction parameters and input the target prediction parameters into the power generation prediction model to obtain target power generation. The target prediction parameters include at least one of the first power generation, the second power generation, the first maintenance plan, and the first environmental data within the past first preset period of time. The present application also provides a computer program product that, when executed on a data processing device, is adapted to execute a program that initializes at least the following method steps:

[0178] Step S201: at intervals of a first preset time, respectively obtain the generated power and location information of the distributed photovoltaic system from the metering system to obtain the first generated power and the first location information; obtain the generated power of the distributed photovoltaic system directly connected to the grid on the user side from the metering system to obtain the second generated power; obtain the feeder corresponding to the distributed photovoltaic system directly connected to the grid on the user side from the dispatching and management system to obtain the second location information; obtain the maintenance record from the dispatching and management system to obtain the first maintenance plan; obtain the environmental data from the meteorological system to obtain the first environmental data; pre-process the first generated power, the first location information, the second generated power, the second location information, the first maintenance plan, and the first environmental data to obtain a first training data set, wherein the pre-processing at least includes data cleaning, data clipping, and data completion;

[0179] Step S202, performing feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set;

[0180] Step S203, performing feature extraction on the second training data set based on factor analysis to obtain a third training data set;

[0181] Step S204: training a prediction model based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is used to predict the power generation of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan, and the first environmental data;

[0182] Step S205, obtain target prediction parameters, input the above target prediction parameters into the above power generation prediction model to obtain target power generation power, the above target prediction parameters include at least one of the above first power generation power, the above second power generation power, the above first maintenance plan and the above first environmental data within the above first preset time period in the past.

[0183] Obviously, those skilled in the art will appreciate that the various modules or steps of the present invention described above can be implemented using a general-purpose computing device, can be centralized on a single computing device, or can be distributed across a network of multiple computing devices. They can be implemented using program code executable by the computing device, and thus, can be stored in a storage device and executed by the computing device. In some cases, the steps shown or described herein can be performed in a different order than that shown, or can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0184] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0185] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0186] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0187] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0188] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0189] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0190] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0191] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0192] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0193] 1) The photovoltaic system power generation prediction method based on edge cluster of the present application, first, at each first preset time interval, the power generation power and location information of the distributed photovoltaic are obtained from the above-mentioned metering system respectively to obtain the first power generation power and the first location information, the power generation power of the distributed photovoltaic directly connected to the grid on the user side is obtained from the above-mentioned metering system to obtain the second power generation power, the feeder corresponding to the above-mentioned distributed photovoltaic directly connected to the grid on the user side is obtained from the above-mentioned dispatching and management system to obtain the second location information, the maintenance record is obtained from the above-mentioned dispatching and management system to obtain the first maintenance plan, the environmental data is obtained from the above-mentioned meteorological system to obtain the first environmental data, the first power generation power, the first location information, the second power generation power, the second location information, the first maintenance plan and the first environmental data are preprocessed to obtain a first training data set, and the preprocessing at least includes data cleaning, data clipping and data Data completion; then, feature screening is performed on the above-mentioned first training data set according to the recursive feature elimination method to obtain a second training data set; then, feature extraction is performed on the above-mentioned second training data set based on the factor analysis method to obtain a third training data set; then, a prediction model is trained based on the above-mentioned third training data set to obtain a power generation prediction model, and the above-mentioned power generation prediction model is used to predict the power generation power of the above-mentioned distributed photovoltaic power generation system within the second preset time period in the future based on one or more of the above-mentioned first power generation power, the above-mentioned second power generation power, the above-mentioned first maintenance plan and the above-mentioned first environmental data; finally, the target prediction parameters are obtained, and the above-mentioned target prediction parameters are input into the above-mentioned power generation prediction model to obtain the target power generation power, and the above-mentioned target prediction parameters include at least one of the above-mentioned first power generation power, the above-mentioned second power generation power, the above-mentioned first maintenance plan and the above-mentioned first environmental data within the past first preset time period. This application proposes a distributed photovoltaic power generation power prediction method based on edge groups, which uses power generation power, meteorological data and maintenance plan data as input to predict the distributed photovoltaic power generation power converged to different substations and feeders, and realizes accurate prediction of distributed photovoltaic power generation power, solving the problem in the prior art of lacking a distributed photovoltaic power generation power prediction method based on multi-dimensional feature data.

[0194] 2) The photovoltaic system power generation prediction device based on edge cluster of the present application, the first acquisition unit obtains the power generation and location information of the distributed photovoltaic from the above-mentioned metering system at intervals of a first preset time length, obtains the first power generation and the first location information, obtains the power generation of the distributed photovoltaic directly connected to the grid on the user side from the above-mentioned metering system, obtains the second power generation, obtains the feeder corresponding to the above-mentioned distributed photovoltaic directly connected to the grid on the user side from the above-mentioned dispatching and management system, obtains the second location information, obtains the maintenance record from the above-mentioned dispatching and management system, obtains the first maintenance plan, obtains the environmental data from the above-mentioned meteorological system, obtains the first environmental data, pre-processes the above-mentioned first power generation, the above-mentioned first location information, the above-mentioned second power generation, the above-mentioned second location information, the above-mentioned first maintenance plan and the above-mentioned first environmental data, and obtains the first training data set. The above-mentioned pre-processing at least includes data cleaning, data clipping and data completion. ; The first calculation unit performs feature screening on the first training data set according to the recursive feature elimination method to obtain a second training data set; the second calculation unit performs feature extraction on the second training data set based on the factor analysis method to obtain a third training data set; the training unit performs prediction model training based on the third training data set to obtain a power generation prediction model, and the power generation prediction model is used to predict the power generation of the distributed photovoltaic power generation system within the second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data; the prediction unit obtains the target prediction parameter, inputs the target prediction parameter into the power generation prediction model to obtain the target power generation, and the target prediction parameter includes at least one of the first power generation power, the second power generation power, the first maintenance plan and the first environmental data within the past first preset time period. This application proposes a distributed photovoltaic power generation power prediction method based on edge groups, which uses power generation power, meteorological data and maintenance plan data as input to predict the distributed photovoltaic power generation power converged to different substations and feeders, thereby achieving accurate prediction of distributed photovoltaic power generation, and solving the problem in the prior art of lacking a distributed photovoltaic power generation power prediction method based on multi-dimensional feature data.

[0195] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A photovoltaic system power generation prediction method based on edge cluster, characterized in that: The method is applicable to a distributed photovoltaic power generation system, which includes a metering system, a dispatching management system, and a meteorological system. The method includes: At intervals of a first preset time length, the generated power and location information of the distributed photovoltaic system are obtained from the metering system to obtain the first generated power and the first location information; the generated power of the distributed photovoltaic system directly connected to the grid on the user side is obtained from the metering system to obtain the second generated power; the feeder corresponding to the distributed photovoltaic system directly connected to the grid on the user side is obtained from the dispatching and management system to obtain the second location information; the maintenance record is obtained from the dispatching and management system to obtain the first maintenance plan; the environmental data is obtained from the meteorological system to obtain the first environmental data; the first generated power, the first location information, the second generated power, the second location information, the first maintenance plan and the first environmental data are preprocessed to obtain a first training data set, wherein the preprocessing at least includes data cleaning, data clipping and data completion; Performing feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set; Performing feature extraction on the second training data set based on factor analysis to obtain a third training data set; Performing prediction model training based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is used to predict the power generation power of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan, and the first environmental data; Obtaining target prediction parameters, inputting the target prediction parameters into the power generation prediction model to obtain target power generation, wherein the target prediction parameters include at least one of the first power generation within the first preset time period, the second power generation, the first maintenance plan, and the first environmental data; Preprocessing the first generated power, the first location information, the second generated power, the second location information, and the first environmental data to obtain a first training data set includes: determining a substation corresponding to the first generated power based on the first location information, and summing the first generated powers corresponding to the substations to obtain a corresponding third generated power; Determine the substation to which the feeder corresponding to the second generated power belongs based on the second location information, and accumulate the generated power corresponding to each feeder according to the relationship between the feeder and the substation, the second generated power, and the third generated power to obtain a corresponding fourth generated power; constructing a fourth training data set according to the third generated power, the fourth generated power, the first maintenance plan, and the first environmental data; performing missing value processing, continuous constant value processing, sudden change outlier processing, and trend outlier processing on the fourth training data set in sequence to obtain the first training data set; Performing trend outlier processing on the fourth training data set includes: Taking the irradiance in the first environmental data in the fourth training data set as a benchmark, determine the changing trend of the irradiance. When the changing trend of the third power generation power or the fourth power generation power in the fourth training data set is opposite to the changing trend of the irradiance, repair the changing trend of the third power generation power or the fourth power generation power based on the linear complement method.

2. The method according to claim 1, characterized in that The fourth training data set is sequentially subjected to missing value processing and continuous constant value processing, including: Traversing the fourth training data set to determine null values ​​or NAN values ​​of sample data therein, and determining them as the missing values; Replacing the missing value based on a first preset value to complete the missing value processing; Calculating the difference between sample data at adjacent moments to obtain a first power increment, summing adjacent first power increments, and if a second power increment is obtained that is less than a first threshold, determining the sample data corresponding to the second power increment as the continuous constant value; When the duration corresponding to the continuous constant value is less than the second threshold, processing is performed based on the linear filling method; when the duration corresponding to the continuous constant value is greater than or equal to the second threshold, processing is performed based on the similar day method.

3. The method according to claim 1, characterized in that Performing mutation outlier processing on the fourth training data set includes: Based on preset formula The sample data in the fourth training data set is screened to obtain the mutation outlier value, wherein: is the probability function, is the change in sample data corresponding to adjacent moments, for The corresponding expectations, Indicates the degree to which the sample deviates from the expectation; The mutation outliers are deleted and filled based on linear interpolation.

4. The method according to claim 1, wherein Performing feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set, including: Using the first training data set as an initial feature set, testing the performance of the preset learner based on the initial feature set to obtain a performance score; When the performance score of the preset learner is less than a third threshold, performing feature scoring according to the weight coefficient of each feature in the initial feature set to obtain a target feature score; Arrange the features in ascending order based on the target feature score and delete a preset number of features; The performance test of the preset learner is re-performed based on the updated initial feature set and the initial feature set is updated until the performance score is greater than or equal to the third threshold, and the initial feature set is determined as the second training data set.

5. The method according to claim 4, characterized in that Performing feature extraction on the second training data set based on factor analysis to obtain a third training data set, including: Constructing a sample sequence according to the second training data set, wherein each row of the sample sequence represents a sample, and each column of the sample sequence represents a feature variable; Calculating a correlation coefficient matrix between each of the characteristic variables, performing factor analysis based on the correlation coefficient matrix, and determining a first target factor; Expressing the characteristic variables as linear combinations of the common first target factors, and obtaining factor loadings of the characteristic variables on the common factors; Screening the main factors corresponding to each of the characteristic variables based on the factor loads, and determining the characteristic variables whose factor loads corresponding to the main factors are greater than a fourth threshold as target variables; The third training dataset is constructed based on the target variable.

6. A photovoltaic system power generation prediction device based on edge cluster, characterized in that: The device is applicable to a distributed photovoltaic power generation system, which includes a metering system, a dispatching management system, and a meteorological system. The device includes: a first acquisition unit, configured to acquire, at intervals of a first preset time duration, the generated power and location information of the distributed photovoltaic system from the metering system to obtain the first generated power and the first location information, acquire the generated power of the distributed photovoltaic system directly connected to the grid on the user side from the metering system to obtain the second generated power, acquire the feeder corresponding to the distributed photovoltaic system directly connected to the grid on the user side from the dispatching and management system to obtain the second location information, acquire the maintenance record from the dispatching and management system to obtain the first maintenance plan, acquire the environmental data from the meteorological system to obtain the first environmental data, and pre-process the first generated power, the first location information, the second generated power, the second location information, the first maintenance plan, and the first environmental data to obtain a first training data set, wherein the pre-processing at least includes data cleaning, data clipping, and data completion; a first computing unit, configured to perform feature screening on the first training data set according to a recursive feature elimination method to obtain a second training data set; a second computing unit, configured to perform feature extraction on the second training data set based on a factor analysis method to obtain a third training data set; a training unit, configured to perform prediction model training based on the third training data set to obtain a power generation prediction model, wherein the power generation prediction model is configured to predict the power generation power of the distributed photovoltaic power generation system within a second preset time period in the future based on one or more of the first power generation power, the second power generation power, the first maintenance plan, and the first environmental data; a prediction unit, configured to obtain a target prediction parameter, input the target prediction parameter into the power generation prediction model, and obtain a target power generation, wherein the target prediction parameter includes at least one of the first power generation, the second power generation, the first maintenance plan, and the first environmental data within the first preset time period; The device is also used to perform the following steps: determining a substation corresponding to the first generated power based on the first location information, and summing the first generated powers corresponding to the substations to obtain a corresponding third generated power; Determine the substation to which the feeder corresponding to the second generated power belongs based on the second location information, and accumulate the generated power corresponding to each feeder according to the relationship between the feeder and the substation, the second generated power, and the third generated power to obtain a corresponding fourth generated power; constructing a fourth training data set according to the third generated power, the fourth generated power, the first maintenance plan, and the first environmental data; performing missing value processing, continuous constant value processing, sudden change outlier processing, and trend outlier processing on the fourth training data set in sequence to obtain the first training data set; Taking the irradiance in the first environmental data in the fourth training data set as a benchmark, determine the changing trend of the irradiance. When the changing trend of the third power generation power or the fourth power generation power in the fourth training data set is opposite to the changing trend of the irradiance, repair the changing trend of the third power generation power or the fourth power generation power based on the linear complement method.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 5.

8. A monitoring system, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing the method of any one of claims 1 to 5.

Images