Photovoltaic power generation scheme prediction method and system

By sorting and transmitting data at the relay end of the photovoltaic power generation area and training the neural network model on the main control end, the problem of low prediction efficiency of photovoltaic power generation schemes in the existing technology is solved, and a more efficient prediction process is achieved.

CN118734140BActive Publication Date: 2024-11-08ANHUI UNIVERSITY OF ARCHITECTURE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411223949.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-11-08
Estimated Expiration
2044-09-03

AI Technical Summary

Technical Problem

The prediction process of existing intelligent photovoltaic power generation solutions is inefficient and cannot effectively improve the prediction workload and efficiency.

Method used

By setting up multiple relay terminals of built-in classifiers in the photovoltaic power generation area, data classification and transmission are performed, and the neural network model of type tag-prediction scores is synchronized at the overall control end to optimize the data processing and identification process.

Benefits of technology

This greatly improves the prediction efficiency, reduces the identification frequency of the overall control terminal, and improves the prediction speed and accuracy of photovoltaic power generation data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118734140B_ABST
    Figure CN118734140B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of intelligent prediction technology, and specifically discloses a photovoltaic power generation scheme prediction method and system, the method comprising: performing type calibration on each relay terminal based on the type of the relay terminal; collecting photovoltaic power generation data by the relay terminal, and inserting a type label into the photovoltaic power generation data according to the type of the current relay terminal; classifying the photovoltaic power generation data by a classifier built into the relay terminal, and determining the next relay terminal; and sending the photovoltaic power generation data containing the type label to a master control terminal by the relay terminal; the present invention sets a plurality of relay terminals with built-in classifiers, classifies the acquired photovoltaic power generation data, and transmits the data between the relay terminals, and finally obtains photovoltaic power generation data with classification results, and when identifying the photovoltaic power generation data, synchronously trains the mapping relationship between the classification result and the recognition result, and when the error is small enough, it can replace the existing recognition process, thereby greatly improving the prediction efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent prediction technology, and in particular to a photovoltaic power generation scheme prediction method and system. Background Art

[0002] "Photovoltaic power generation scheme prediction" refers to the process of using intelligent technology and data analysis methods to predict the results of photovoltaic power generation schemes. This prediction can involve many aspects, including energy utilization efficiency, cost-effectiveness, environmental impact, etc.

[0003] The existing intelligent prediction process usually uses various sensors, data acquisition systems and intelligent analysis algorithms to collect and analyze data related to power supply, such as indoor and outdoor temperature, power supply equipment operating status, energy consumption, etc. By analyzing these data, problems or potential improvement spaces in the power supply system can be identified, and corresponding optimization plans can be proposed.

[0004] However, most of the existing intelligent prediction solutions perform global prediction on the acquired data based on preset algorithms. The prediction object is all the acquired data. The prediction workload is large and the efficiency is low. How to improve the efficiency of the prediction process is the technical problem that the technical solution of the present invention wants to solve. Summary of the invention

[0005] The purpose of the present invention is to provide a photovoltaic power generation scheme prediction method and system to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A photovoltaic power generation scheme prediction method and system, the method comprising:

[0008] Step S1: receiving the prediction index of photovoltaic power generation data input by the staff, and determining the type of the relay terminal according to the prediction index; wherein the value space of the type is the same as the index space of the prediction index; and the prediction index contains a sequential feature;

[0009] Step S2: obtaining the positions of all predicted objects in the photovoltaic power generation area, determining the position of the relay terminal according to the positions, and calibrating the type of each relay terminal based on the type of the relay terminal; wherein, during the type calibration process, a classifier is synchronously inserted into the relay terminal;

[0010] Step S3: The relay terminal collects photovoltaic power generation data, and inserts a type tag into the photovoltaic power generation data according to the type of the current relay terminal; wherein the order of inserting the type tag is recorded;

[0011] Step S4: the classifier built into the relay terminal classifies the photovoltaic power generation data and determines the next relay terminal;

[0012] Step S5: looping through steps S3 and S4, when the difference between the tag set constructed by all type tags of photovoltaic power generation data acquired by any relay terminal and the value space is empty, the photovoltaic power generation data containing the type tag is sent to the master control terminal;

[0013] Step S6: The master control terminal identifies the photovoltaic power generation data, calculates the prediction score, and simultaneously trains a neural network model containing the ordered type label-prediction score.

[0014] As a further solution of the present invention: the steps of obtaining the positions of all predicted objects in the photovoltaic power generation area, determining the position of the relay terminal according to the positions, and calibrating the type of each relay terminal based on the type of the relay terminal include:

[0015] Query the location and power demand of all predicted objects in the photovoltaic power generation area;

[0016] Determine the influence radius of each location based on the power demand and construct an influence circle;

[0017] Count all influence circles, calculate the intersection, and take the number of influence circles corresponding to the intersection as the coreness of the intersection;

[0018] Select the intersection according to the descending order of coreness, randomly select a position in the selected intersection as the relay end position, and set the relay end;

[0019] Calculate the transmission time of each predicted object and each relay end, and for each predicted object, select the relay end corresponding to the minimum transmission time to establish a connection channel;

[0020] Each relay end is type-labeled based on the type of the relay end.

[0021] As a further solution of the present invention: the step of classifying each relay terminal based on the type of the relay terminal includes:

[0022] Acquire the number of types of relay terminals, and align the total number of relay terminals according to the number of types;

[0023] The total number after registration is distributed based on the number of types to obtain the number of relay terminals of each type;

[0024] Select relay terminals based on the number of relay terminals of each type and perform type calibration;

[0025] The selected relays of each type meet the following conditions: ; ; In the formula, is the total number of all relays of each type, is the preset threshold, It is The coordinates of the relay terminals, It is the first relay of the same type. The coordinates of the relay terminals, is a non-zero integer, For point and Point distance.

[0026] As a further solution of the present invention: the step of classifying the photovoltaic power generation data by the classifier built into the relay end and determining the next relay end includes:

[0027] Obtain photovoltaic power generation data samples and positively transform each indicator in each sample;

[0028] Determine the arithmetic progression based on the maximum and mean values ​​after forward transformation;

[0029] For the photovoltaic power generation data obtained by the relay end, locate the index of the photovoltaic power generation data according to the type of the relay end, and after positively converting the index, determine its matching sequence according to the arithmetic progression;

[0030] Query the next indicator according to the matching sequence number, query the relay end corresponding to the next indicator and randomly select one as the next relay end.

[0031] As a further solution of the present invention: the rule of positive transformation is:

[0032] When the indicator is a reverse indicator: ;

[0033] When the indicator is a moderate indicator: ;

[0034] The common difference of an arithmetic sequence is:

[0035] ;

[0036] In the above content, is the first numerical values, is the first numerical values, is the preset standard value; is the tolerance, is the maximum value after positive transformation, is the minimum value after positive transformation, is the average value after normalization, is the preset correction factor.

[0037] As a further solution of the present invention: the steps of identifying photovoltaic power generation data by the master control terminal, calculating the prediction score, and synchronously training the neural network model containing the ordered type label-prediction score include:

[0038] The master control terminal identifies the photovoltaic power generation data based on the entropy weight method and calculates the prediction score;

[0039] Get the type label with order as input, take the prediction score as output, and construct the sample set;

[0040] Train a neural network model containing ordered type labels-prediction scores from the sample set;

[0041] Calculate the accuracy of the neural network model in real time, and adjust the entropy weight method and the usage frequency of the neural network model based on the accuracy.

[0042] The technical solution of the present invention also provides a photovoltaic power generation scheme prediction system, the system comprising:

[0043] An initialization module is used to receive the prediction index of photovoltaic power generation data input by the staff, and determine the type of the relay end according to the prediction index; wherein the value space of the type is the same as the index space of the prediction index; and the prediction index contains a sequential feature;

[0044] A type calibration module is used to obtain the positions of all predicted objects in the photovoltaic power generation area, determine the position of the relay terminal according to the position, and perform type calibration on each relay terminal based on the type of the relay terminal; wherein, during the type calibration process, a classifier is synchronously inserted into the relay terminal;

[0045] The tag insertion module is used to collect photovoltaic power generation data from the relay end, and insert a type tag into the photovoltaic power generation data according to the type of the current relay end; wherein the order of inserting the type tag is recorded;

[0046] A data classification module is used to classify the photovoltaic power generation data by a classifier built into the relay end and determine the next relay end;

[0047] The data upload module is used to cyclically execute the label insertion module and the data classification module. When the difference between the label set constructed by all types of labels of the photovoltaic power generation data obtained by any relay end and the value space is empty, the photovoltaic power generation data containing the type label is sent to the master control end;

[0048] The model training module is used to identify photovoltaic power generation data from the master control end, calculate the prediction score, and simultaneously train the neural network model containing the ordered type label-prediction score.

[0049] As a further solution of the present invention: the type calibration module includes:

[0050] An information acquisition unit, used to query the location and power supply demand of all predicted objects in the photovoltaic power generation area;

[0051] An influence circle construction unit, used to determine the influence radius of each location based on the power demand and construct an influence circle;

[0052] The core degree calculation unit is used to count all influence circles, calculate the intersection, and take the number of influence circles corresponding to the intersection as the core degree of the intersection;

[0053] A position selection unit is used to select an intersection according to the descending order of coreness, randomly select a position in the selected intersection as the relay end position, and set the relay end;

[0054] A channel establishment unit is used to calculate the transmission time of each prediction object and each relay end, and for each prediction object, select the relay end corresponding to the minimum transmission time to establish a connection channel;

[0055] The calibration execution unit is used to perform type calibration on each relay terminal based on the type of the relay terminal.

[0056] As a further solution of the present invention: the label insertion module includes:

[0057] A positive conversion unit is used to obtain photovoltaic power generation data samples and positively convert each indicator in each sample;

[0058] A sequence construction unit is used to determine an arithmetic sequence according to the maximum and mean values ​​after forward transformation;

[0059] The data matching unit is used to locate the index of the photovoltaic power generation data acquired by the relay end according to the type of the relay end, and after positively converting the index, determine its matching sequence according to the arithmetic progression;

[0060] The query selection unit is used to query the next indicator according to the matching sequence number, query the relay end corresponding to the next indicator and randomly select one as the next relay end.

[0061] As a further solution of the present invention: the model training module includes:

[0062] The data identification unit is used for the master control end to identify the photovoltaic power generation data based on the entropy weight method and calculate the prediction score;

[0063] The sample set construction unit is used to obtain the type label containing the order as input, take the prediction score as output, and construct the sample set;

[0064] A training unit, used for training a neural network model containing ordered type labels and prediction scores from a sample set;

[0065] The application unit is used to calculate the accuracy of the neural network model in real time, and adjust the entropy weight method based on the accuracy and the usage frequency of the neural network model.

[0066] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention sets up multiple relay terminals with built-in classifiers, classifies the acquired photovoltaic power generation data, and transmits it between the relay terminals, and finally obtains photovoltaic power generation data with classification results. When the photovoltaic power generation data is identified, the mapping relationship between the classification result and the identification result is synchronously trained. When the error is small enough, it can replace the existing identification process, greatly improving the prediction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0068] Figure 1 The overall flow chart of the photovoltaic power generation scheme prediction method is shown.

[0069] Figure 2 The structural diagram of the photovoltaic power generation scheme prediction system is shown. DETAILED DESCRIPTION

[0070] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0071] Figure 1 The overall flow chart of the photovoltaic power generation scheme prediction method and system is as follows. In an embodiment of the present invention, a photovoltaic power generation scheme prediction method is provided, and the method includes:

[0072] Step S1: receiving the prediction index of photovoltaic power generation data input by the staff, and determining the type of the relay terminal according to the prediction index; wherein the value space of the type is the same as the index space of the prediction index; and the prediction index contains a sequential feature;

[0073] The staff inputs the prediction standards of photovoltaic power generation data in advance, which are called prediction indicators. For example, the prediction indicators include average power supply power, maximum power, power supply frequency, active adjustment vector, user information and user satisfaction; among them, the active adjustment vector is the adjustment of the power supply process by the power supply recipient. The adjustment process includes adjustment direction and adjustment amount, and is therefore represented by a vector; user information is the main information of the power supply recipient, and user satisfaction is the satisfaction of the power supply recipient with the power supply process.

[0074] After the prediction indicators are determined, the type of relay terminal is determined according to the type of prediction indicators, wherein the number of types of relay terminals is the same as the number of types of prediction indicators; in addition, the determined prediction indicators themselves contain an order, and the same prediction indicators with different arrangements are considered to be different.

[0075] Step S2: obtaining the positions of all predicted objects in the photovoltaic power generation area, determining the position of the relay terminal according to the positions, and calibrating the type of each relay terminal based on the type of the relay terminal; wherein, during the type calibration process, a classifier is synchronously inserted into the relay terminal;

[0076] The photovoltaic power generation area is the area where photovoltaic power generation data prediction is required. The prediction object is the power supply recipient within the photovoltaic power generation area. The positions of all prediction objects are obtained, a relay terminal is established based on the position, and a connection channel between the relay terminal and the prediction object is established to obtain photovoltaic power generation data in real time. At the same time, each relay terminal is type-calibrated. The type calibration process is very simple, which is to insert a type-indicator into each relay terminal. Each type corresponds to each indicator. This allows the relay terminal to analyze only one indicator when obtaining the data.

[0077] Step S3: The relay terminal collects photovoltaic power generation data, and inserts a type tag into the photovoltaic power generation data according to the type of the current relay terminal; wherein the order of inserting the type tag is recorded;

[0078] For each relay end, the relay end collects photovoltaic power generation data. Each relay end has a unique type. The type query indicator is used to insert the type tag into the photovoltaic power generation data. The purpose of the insertion process is to indicate that the indicator of the photovoltaic power generation data has been processed. It should be noted that, similar to the indicator, when inserting the type tag, the order must be retained, and type tags of different orders are considered different.

[0079] Step S4: the classifier built into the relay terminal classifies the photovoltaic power generation data and determines the next relay terminal;

[0080] Whenever a relay terminal receives photovoltaic power generation data, the classifier built into the relay terminal will classify the photovoltaic power generation data and then determine the next indicator to be processed. The indicator is matched with the relay terminal of the corresponding type and the photovoltaic power generation data is sent to it.

[0081] Step S5: looping through steps S3 and S4, when the difference between the tag set constructed by all type tags of photovoltaic power generation data acquired by any relay terminal and the value space is empty, the photovoltaic power generation data containing the type tag is sent to the master control terminal;

[0082] Steps S3 and S4 are executed in a loop to continuously classify the various indicators in the photovoltaic power generation data. When the last type of relay terminal is reached (meaning that the difference is empty), the photovoltaic power generation data and type tags are sent to the main control terminal; among them, the type tag contains an order, and the order is a very important parameter because all type tags are arrangements of existing types.

[0083] Step S6: The master control terminal identifies the photovoltaic power generation data, calculates the prediction score, and simultaneously trains a neural network model containing the ordered type label-prediction score;

[0084] Finally, the main control end identifies the photovoltaic power generation data and calculates the prediction score. The purpose of this application is to transfer the recognition process to the relay end. The recognition process of the relay end is reflected by the type label containing the order, and the type label-prediction score sample is constructed to train the neural network model. When the accuracy of the neural network model is high, the prediction score can be quickly output based on the type label. The frequency of the main control end to identify the photovoltaic power generation data can be greatly reduced, which improves the prediction speed.

[0085] The prediction score of the present application is used to characterize the quality of the photovoltaic power generation process.

[0086] In some embodiments of the present invention, the steps of obtaining the positions of all predicted objects in the photovoltaic power generation area, determining the position of the relay terminal according to the positions, and calibrating the type of each relay terminal based on the type of the relay terminal include:

[0087] Query the location and power demand of all predicted objects in the photovoltaic power generation area;

[0088] Determine the influence radius of each location based on the power demand and construct an influence circle;

[0089] Count all influence circles, calculate the intersection, and take the number of influence circles corresponding to the intersection as the coreness of the intersection;

[0090] Select the intersection according to the descending order of coreness, randomly select a position in the selected intersection as the relay end position, and set the relay end;

[0091] Calculate the transmission time of each predicted object and each relay end, and for each predicted object, select the relay end corresponding to the minimum transmission time to establish a connection channel;

[0092] Each relay end is type-labeled based on the type of the relay end.

[0093] The above content limits the type calibration process of the relay end, and queries the location and power supply demand of the predicted object in the photovoltaic power generation area. The power supply demand is the data uploaded by the predicted object. In the technical solution of the present invention, it is regarded as known data, and the radius is determined by the power supply demand. The influence circle is constructed with the position as the center. Generally, the larger the power supply demand, the larger the influence circle; after constructing an influence circle for each predicted object, the intersection of all influence circles is calculated. The intersection represents that the area can collect data for multiple predicted objects. The more influence circles corresponding to the intersection, the more prediction objects that can collect data. Therefore, the number of influence circles corresponding to the intersection is used as the coreness of the intersection.

[0094] By selecting the intersection in descending order of coreness and continuously setting up relay terminals, data can be collected on more prediction objects under the premise of a certain number, thus reducing costs.

[0095] Then, the transmission time of each predicted object and each relay end is tested, the minimum time is selected, the predicted objects and relay ends are paired, and a connection channel is established.

[0096] Finally, assign the type attribute to each relay end.

[0097] The process of assigning the type attribute to each relay end is as follows:

[0098] The step of classifying each relay terminal based on the type of the relay terminal comprises:

[0099] Acquire the number of types of relay terminals, and align the total number of relay terminals according to the number of types;

[0100] The total number after registration is distributed based on the number of types to obtain the number of relay terminals of each type;

[0101] Relay terminals are selected based on the number of each type of relay terminals and type calibration is performed.

[0102] In order to make the processing structure more balanced, the present application sets the same number of relay terminals for each type. At this time, the total number of relay terminals can be evenly divided by type. Since the total number of relay terminals may not be evenly divisible by type, it is necessary to align the total number of relay terminals. The alignment process is to make it a multiple of the number of types.

[0103] The registered relay ends are evenly distributed, and the same number of relay ends of each type is set. Then, for each type, a relay end is selected from the existing relay ends and set to the corresponding type.

[0104] Specifically, regarding the process of selecting relay terminals based on the number of relay terminals of each type and performing type calibration, the specific requirements are as follows:

[0105] The selected relays of each type meet the following conditions: ; ; In the formula, is the total number of all relays of each type, is the preset threshold, It is The coordinates of the relay terminals, It is the first relay of the same type. The coordinates of the relay terminals, is a non-zero integer, For point and Point distance.

[0106] Since the relay end in this application is used to collect photovoltaic power generation data, photovoltaic power generation data is generated by the prediction object, and the prediction objects in the photovoltaic power generation area are evenly distributed, this application hopes that each relay end is also evenly distributed in the photovoltaic power generation area, and try not to concentrate in a certain area. The purpose of the above conditions is to prevent a large number of concentrations of each interrupt end. The specific meaning of the conditions is:

[0107] For any type of relay, The distance between a relay and the nearest relay of the same type is called the nearest distance. The sum of all the nearest distances among the relays of the same type is calculated to make it greater than the threshold. (If it is less than the threshold , reselect), this can prevent large-scale concentration (the more concentrated, the smaller the sum of the closest distances).

[0108] In some embodiments of the present invention, the step of classifying the photovoltaic power generation data by a classifier built into the relay terminal and determining the next relay terminal includes:

[0109] Obtain photovoltaic power generation data samples and positively transform each indicator in each sample;

[0110] Determine the arithmetic progression based on the maximum and mean values ​​after forward transformation;

[0111] For the photovoltaic power generation data obtained by the relay end, locate the index of the photovoltaic power generation data according to the type of the relay end, and after positively converting the index, determine its matching sequence according to the arithmetic progression;

[0112] Query the next indicator according to the matching sequence number, query the relay end corresponding to the next indicator and randomly select one as the next relay end.

[0113] Whenever a relay terminal reads photovoltaic power generation data, it analyzes the photovoltaic power generation data according to the type positioning indicator of the current relay terminal, obtains the classification result, and then determines the next relay terminal.

[0114] The analysis and classification process is as follows:

[0115] First, the execution subject of this method counts multiple photovoltaic power generation data, and positively converts each indicator in the photovoltaic power generation data to obtain data with the same evaluation standard, and the larger the evaluation standard, the better; then, for each indicator, the maximum value and the mean are determined according to the positively converted data, the maximum value is used as the first and last term of the arithmetic sequence, and the tolerance is determined by the difference between the maximum value and the mean, thereby obtaining an arithmetic sequence with a very high matching degree, and the received photovoltaic power generation data can be analyzed and classified by the arithmetic sequence.

[0116] The specific, positive rules are:

[0117] When the indicator is a reverse indicator: ;

[0118] When the indicator is a moderate indicator: ;

[0119] The common difference of an arithmetic sequence is:

[0120] ;

[0121] In the above content, is the first numerical values, is the first numerical values, is the preset standard value; is the tolerance, is the maximum value after positive transformation, is the minimum value after positive transformation, is the average value after normalization, is the preset correction factor.

[0122] Specifically, the positive rule is mainly to convert the reverse indicator and the moderate indicator into the positive indicator. The meaning of the reverse indicator is that the smaller the value, the better. The meaning of the moderate indicator is that the closer the value is to a certain standard value, the better. The meaning of the positive indicator is that the larger the value, the better.

[0123] In addition, the tolerance of an arithmetic sequence is affected by two factors: one is the difference between the mean and the maximum value, and the other is the difference between the mean and the minimum value. It is proportional to the smaller value and also proportional to the larger value. The difference is that the larger value has less influence on the tolerance, and a logarithmic function is used for compounding.

[0124] In some embodiments of the present invention, the steps of identifying photovoltaic power generation data by the master control terminal, calculating prediction scores, and synchronously training a neural network model containing a sequential type label-prediction score include:

[0125] The master control terminal identifies the photovoltaic power generation data based on the entropy weight method and calculates the prediction score;

[0126] Get the type label with order as input, take the prediction score as output, and construct the sample set;

[0127] Train a neural network model containing ordered type labels-prediction scores from the sample set;

[0128] Calculate the accuracy of the neural network model in real time, and adjust the entropy weight method and the usage frequency of the neural network model based on the accuracy.

[0129] Calculating the prediction score of each photovoltaic power generation data is the power supply plan prediction process. There are two calculation schemes for the prediction score. One is an algorithm-based prediction process, which needs to predict all indicators of photovoltaic power generation data and consumes a long time. The other is a neural network model-based prediction process, which only needs to predict the type label unique to this application (which relay terminals the photovoltaic power generation data passes through), which consumes a short time. Considering the accuracy problem, at the beginning, it is necessary to use as many algorithm-based prediction processes as possible. When the number of samples is large and the error of the neural network model is small enough, the prediction process based on the neural network model can be used as much as possible. That is, the frequency of use of the neural network model is directly proportional to the accuracy, and the frequency of use of the algorithm-based prediction process is inversely proportional to the accuracy.

[0130] It is worth mentioning that the present application predicts multiple indicators of photovoltaic power generation data. On the basis of positive transformation of each indicator, the entropy weight method can be used for algorithm-based prediction process, that is, the algorithm is limited to the entropy weight method.

[0131] Figure 2 The structure diagram of the photovoltaic power generation scheme prediction system is shown. In a preferred embodiment of the technical solution of the present invention, a photovoltaic power generation scheme prediction system is also provided. The system 10 includes:

[0132] Initialization module 11 is used to receive the prediction index of photovoltaic power generation data input by the staff, and determine the type of the relay terminal according to the prediction index; wherein the value space of the type is the same as the index space of the prediction index; and the prediction index contains a sequential feature;

[0133] The type calibration module 12 is used to obtain the positions of all predicted objects in the photovoltaic power generation area, determine the position of the relay terminal according to the position, and perform type calibration on each relay terminal based on the type of the relay terminal; wherein, during the type calibration process, a classifier is synchronously inserted into the relay terminal;

[0134] The tag insertion module 13 is used to collect photovoltaic power generation data from the relay end, and insert a type tag into the photovoltaic power generation data according to the type of the current relay end; wherein the order of inserting the type tag is recorded;

[0135] The data classification module 14 is used to classify the photovoltaic power generation data by a classifier built into the relay end, and determine the next relay end;

[0136] The data uploading module 15 is used to cyclically execute the label inserting module 13 and the data classification module 14. When the difference between the label set constructed by all types of labels of the photovoltaic power generation data acquired by any relay end and the value space is empty, the photovoltaic power generation data containing the type label is sent to the master control end;

[0137] The model training module 16 is used for the master control terminal to identify the photovoltaic power generation data, calculate the prediction score, and simultaneously train the neural network model containing the ordered type label-prediction score.

[0138] Furthermore, the type identification module 12 includes:

[0139] An information acquisition unit, used to query the location and power supply demand of all predicted objects in the photovoltaic power generation area;

[0140] An influence circle construction unit, used to determine the influence radius of each location based on the power demand and construct an influence circle;

[0141] The core degree calculation unit is used to count all influence circles, calculate the intersection, and take the number of influence circles corresponding to the intersection as the core degree of the intersection;

[0142] A position selection unit is used to select an intersection according to the descending order of coreness, randomly select a position in the selected intersection as the relay end position, and set the relay end;

[0143] A channel establishment unit is used to calculate the transmission time of each prediction object and each relay end, and for each prediction object, select the relay end corresponding to the minimum transmission time to establish a connection channel;

[0144] The calibration execution unit is used to perform type calibration on each relay terminal based on the type of the relay terminal.

[0145] Specifically, the label insertion module 13 includes:

[0146] A positive conversion unit is used to obtain photovoltaic power generation data samples and positively convert each indicator in each sample;

[0147] A sequence construction unit is used to determine an arithmetic sequence according to the maximum and mean values ​​after forward transformation;

[0148] The data matching unit is used to locate the index of the photovoltaic power generation data acquired by the relay end according to the type of the relay end, and after positively converting the index, determine its matching sequence according to the arithmetic progression;

[0149] The query selection unit is used to query the next indicator according to the matching sequence number, query the relay end corresponding to the next indicator and randomly select one as the next relay end.

[0150] Furthermore, the model training module 16 includes:

[0151] The data identification unit is used for the master control end to identify the photovoltaic power generation data based on the entropy weight method and calculate the prediction score;

[0152] The sample set construction unit is used to obtain the type label containing the order as input, take the prediction score as output, and construct the sample set;

[0153] A training unit, used for training a neural network model containing ordered type labels and prediction scores from a sample set;

[0154] The application unit is used to calculate the accuracy of the neural network model in real time, and adjust the entropy weight method based on the accuracy and the usage frequency of the neural network model.

[0155] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A photovoltaic power generation scheme prediction method, characterized in that: The method comprises: Step S1: receiving the prediction index of photovoltaic power generation data input by the staff, and determining the type of the relay terminal according to the prediction index; wherein the value space of the type is the same as the index space of the prediction index; and the prediction index contains a sequential feature; Step S2: obtaining the positions of all predicted objects in the photovoltaic power generation area, determining the position of the relay terminal according to the positions, and calibrating the type of each relay terminal based on the type of the relay terminal; wherein, during the type calibration process, a classifier is synchronously inserted into the relay terminal; Step S3: The relay terminal collects photovoltaic power generation data, and inserts a type tag into the photovoltaic power generation data according to the type of the current relay terminal; wherein the order of inserting the type tag is recorded; Step S4: the classifier built into the relay terminal classifies the photovoltaic power generation data and determines the next relay terminal; Step S5: looping through steps S3 and S4, when the difference between the tag set constructed by all type tags of photovoltaic power generation data acquired by any relay terminal and the value space is empty, the photovoltaic power generation data containing the type tag is sent to the master control terminal; Step S6: The master control terminal identifies the photovoltaic power generation data, calculates the prediction score, and simultaneously trains a neural network model containing the ordered type label-prediction score; The steps of obtaining the positions of all predicted objects in the photovoltaic power generation area, determining the position of the relay terminal according to the positions, and calibrating the type of each relay terminal based on the type of the relay terminal include: Query the location and power demand of all predicted objects in the photovoltaic power generation area; Determine the influence radius of each location based on the power demand and construct an influence circle; Count all influence circles, calculate the intersection, and take the number of influence circles corresponding to the intersection as the coreness of the intersection; Select the intersection according to the descending order of coreness, randomly select a position in the selected intersection as the relay end position, and set the relay end; Calculate the transmission time of each predicted object and each relay end, and for each predicted object, select the relay end corresponding to the minimum transmission time to establish a connection channel; Perform type calibration on each relay end based on the type of the relay end; The step of classifying each relay terminal based on the type of the relay terminal comprises: Acquire the number of types of relay terminals, and align the total number of relay terminals according to the number of types; The total number after registration is distributed based on the number of types to obtain the number of relay terminals of each type; Select relay terminals based on the number of relay terminals of each type and perform type calibration; The selected relays of each type meet the following conditions: ; ; In the formula, is the total number of all relays of each type, is the preset threshold, is the coordinate of the ith relay terminal, It is the first relay of the same type. The coordinates of the relay terminals, is a non-zero integer, For point and Point distance.

2. The photovoltaic power generation scheme prediction method according to claim 1, characterized in that: The step of classifying the photovoltaic power generation data by the classifier built into the relay end and determining the next relay end includes: Obtain photovoltaic power generation data samples and positively transform each indicator in each sample; Determine the arithmetic progression based on the maximum and mean values ​​after forward transformation; For the photovoltaic power generation data obtained by the relay end, locate the index of the photovoltaic power generation data according to the type of the relay end, and after the index is positively converted, determine its matching sequence according to the arithmetic progression; Query the next indicator according to the matching sequence number, query the relay end corresponding to the next indicator and randomly select one as the next relay end.

3. The photovoltaic power generation scheme prediction method according to claim 2, characterized in that: The rules for forwarding are: When the indicator is a reverse indicator: ; When the indicator is a moderate indicator: ; The common difference of an arithmetic sequence is: ; In the above content, is the jth value of a certain indicator before it is positively transformed, is the jth value of a certain indicator after positive transformation, k is the preset standard value; A is the tolerance, is the maximum value after positive transformation, is the maximum value after positive transformation, is the average value after normalization, is the preset correction factor.

4. The photovoltaic power generation scheme prediction method according to claim 3, characterized in that: The steps of identifying photovoltaic power generation data, calculating prediction scores, and synchronously training a neural network model containing sequential type labels-prediction scores by the master control terminal include: The master control terminal identifies the photovoltaic power generation data based on the entropy weight method and calculates the prediction score; Get the type label with order as input, take the prediction score as output, and construct the sample set; Train a neural network model containing ordered type labels-prediction scores from the sample set; Calculate the accuracy of the neural network model in real time, and adjust the entropy weight method and the usage frequency of the neural network model based on the accuracy.

5. A photovoltaic power generation scheme prediction system, characterized in that: The system comprises: An initialization module is used to receive the prediction index of photovoltaic power generation data input by the staff, and determine the type of the relay end according to the prediction index; wherein the value space of the type is the same as the index space of the prediction index; and the prediction index contains a sequential feature; A type calibration module is used to obtain the positions of all predicted objects in the photovoltaic power generation area, determine the position of the relay terminal according to the position, and perform type calibration on each relay terminal based on the type of the relay terminal; wherein, during the type calibration process, a classifier is synchronously inserted into the relay terminal; The tag insertion module is used to collect photovoltaic power generation data from the relay end, and insert a type tag into the photovoltaic power generation data according to the type of the current relay end; wherein the order of inserting the type tag is recorded; A data classification module is used to classify the photovoltaic power generation data by a classifier built into the relay end to determine the next relay end; The data upload module is used to cyclically execute the label insertion module and the data classification module. When the difference between the label set constructed by all types of labels of the photovoltaic power generation data obtained by any relay end and the value space is empty, the photovoltaic power generation data containing the type label is sent to the master control end; Model training module, used for the master control terminal to identify photovoltaic power generation data, calculate prediction scores, and simultaneously train a neural network model containing sequential type labels-prediction scores; The type calibration module includes: An information acquisition unit, used to query the location and power supply demand of all predicted objects in the photovoltaic power generation area; An influence circle construction unit, used to determine the influence radius of each location based on the power demand and construct an influence circle; The core degree calculation unit is used to count all influence circles, calculate the intersection, and take the number of influence circles corresponding to the intersection as the core degree of the intersection; A position selection unit is used to select an intersection according to the descending order of coreness, randomly select a position in the selected intersection as the relay end position, and set the relay end; A channel establishment unit is used to calculate the transmission time of each prediction object and each relay end, and for each prediction object, select the relay end corresponding to the minimum transmission time to establish a connection channel; A calibration execution unit, used for calibrating the type of each relay terminal based on the type of the relay terminal; The content of classifying each relay terminal based on the type of the relay terminal includes: Acquire the number of types of relay terminals, and align the total number of relay terminals according to the number of types; The total number after registration is distributed based on the number of types to obtain the number of relay terminals of each type; Select relay terminals based on the number of relay terminals of each type and perform type calibration; The selected relays of each type meet the following conditions: ; ; In the formula, is the total number of all relays of each type, is the preset threshold, is the coordinate of the ith relay terminal, It is the first relay of the same type. The coordinates of the relay terminals, is a non-zero integer, For point and Point distance.

6. The photovoltaic power generation scheme prediction system according to claim 5, characterized in that: The label insertion module comprises: A positive conversion unit is used to obtain photovoltaic power generation data samples and positively convert each indicator in each sample; A sequence construction unit is used to determine an arithmetic sequence according to the maximum and mean values ​​after forward transformation; The data matching unit is used to locate the index of the photovoltaic power generation data acquired by the relay end according to the type of the relay end, and after positively converting the index, determine its matching sequence according to the arithmetic progression; The query selection unit is used to query the next indicator according to the matching sequence number, query the relay end corresponding to the next indicator and randomly select one as the next relay end.

7. The photovoltaic power generation scheme prediction system according to claim 6, characterized in that: The model training module includes: The data identification unit is used for the master control end to identify the photovoltaic power generation data based on the entropy weight method and calculate the prediction score; The sample set construction unit is used to obtain the type label containing the order as input, take the prediction score as output, and construct the sample set; A training unit, used for training a neural network model containing ordered type labels and prediction scores from a sample set; The application unit is used to calculate the accuracy of the neural network model in real time, and adjust the entropy weight method based on the accuracy and the usage frequency of the neural network model.