KR102283487B1 - A Forecasting System of Photovoltaic Generation Based on Machine-learning Using Realtime Satellite Data and Numerical Modeling Data - Google Patents

Abstract

The present invention relates to a solar power generation amount prediction system, and more particularly, to correct weather information predicted by a numerical model using weather information observed by a satellite, and predict the amount of power generation of a solar power generation device using this It is related to a machine learning-based solar power generation forecasting system using real-time satellite data and numerical model data that can increase the accuracy of forecasting.

Description

A Forecasting System of Photovoltaic Generation Based on Machine-learning Using Realtime Satellite Data and Numerical Modeling Data

The present invention relates to a solar power generation amount prediction system, and more particularly, to correct weather information predicted by a numerical model using weather information observed by a satellite, and predict the amount of power generation of a solar power generation device using this It is related to a machine learning-based solar power generation forecasting system using real-time satellite data and numerical model data that can increase the accuracy of forecasting.

In general, solar power generation is in the spotlight as a new and renewable energy power generation technology because it uses non-polluting, unlimited solar power and has little adverse effect on the environment. The spread of new and renewable energy power generation facilities is increasing.

However, the amount of solar power generation is highly uncertain due to temporal and spatial variables, and it is very difficult to predict because it varies according to regional characteristics such as latitude and longitude, and the solar altitude and atmospheric condition according to time.

In this way, if the amount of power generation cannot be accurately predicted, the power generation complex cannot be operated efficiently, and a plan for the distribution of the produced electricity cannot be properly established, so that economic effects through solar power generation cannot be achieved and an imbalance in power supply and demand occurs. .

Therefore, as in the following patent documents, the power to be generated is predicted based on the existing renewable energy produced or the amount of solar power generation is predicted based on the weather forecast. There was a problem with falling.

(Patent Literature)

Registered Patent Publication No. 10-1020639 (Registered on Mar. 02, 2011) "Renewable Energy Production Prediction Method"

The present invention has been devised to solve the above problems,

The present invention corrects the weather information predicted by the numerical model using the weather information observed by the satellite, and predicts the amount of power generation of the photovoltaic device using this to increase the accuracy of the prediction. It aims to provide a prediction system.

An object of the present invention is to provide a solar power generation prediction system that enables accurate correction of predicted weather information by analyzing the correlation between predicted weather information and satellite observation weather information by linear regression analysis to correct the predicted weather information. there is.

An object of the present invention is to provide a solar power generation amount prediction system that enables accurate correlation analysis by matching spatial domains and temporal resolutions of predicted weather information and phase observation weather information.

The present invention is implemented by an embodiment having the following configuration in order to achieve the above object.

According to an embodiment of the present invention, the solar power generation amount prediction system according to the present invention comprises: a photovoltaic device installed at a certain point to generate power by sunlight; a monitoring server for predicting the weather at the point where the photovoltaic device is installed, and predicting the amount of power generation of the photovoltaic device by machine learning using the predicted weather information; Including; a display device for displaying the weather information and power generation predicted by the monitoring server, wherein the monitoring server corrects the weather forecast information according to the correlation between the weather forecast information for a specific point and the meteorological observation information by the satellite It is characterized in that it is used to predict the amount of power generation of the solar power generation device.

According to another embodiment of the present invention, in the solar power generation amount prediction system according to the present invention, the monitoring server includes a weather forecasting and collecting unit for collecting weather information predicted by a numerical model, and weather observation information by satellite. A satellite data collection unit that collects, an error correction unit that corrects the forecasted weather information by analyzing the correlation between the collected weather forecast information and satellite weather observation information, and an algorithm between the corrected forecasted weather information and the amount of electricity generated by machine learning It characterized in that it comprises a machine learning performing unit to derive, and a power generation forecasting unit for predicting the amount of power generation of the solar power generation device by inputting weather prediction information to the algorithm derived by the machine learning performing unit.

According to another embodiment of the present invention, in the solar power generation amount prediction system according to the present invention, the error correction unit includes a weather forecast information loading module for loading the collected weather forecast information, and satellite data for loading the satellite observation weather information. A loading module, a linear regression analysis module that analyzes the correlation between the weather forecast information and the phase observation weather information for the same location at the same time through a linear regression equation, and the weather that corrects the weather forecast information according to the analyzed linear regression equation It is characterized in that it includes a prediction correction module.

According to another embodiment of the present invention, in the solar power generation forecasting system according to the present invention, the monitoring server calculates the spatial area and temporal resolution of weather forecast information and satellite observation weather information to correct the collected weather forecast information. It is characterized in that it includes a preprocessing unit for matching.

According to another embodiment of the present invention, in the solar power generation amount prediction system according to the present invention, the machine learning execution unit includes a prediction factor input module for inputting a weather prediction factor for predicting the amount of power generation, and according to the input weather prediction factor A power generation input module for inputting power generation information, an algorithm generation module for deriving a prediction algorithm for weather prediction factors and power generation using machine learning, and an optimization to modify the algorithm by comparing the power generation predicted by the prediction algorithm with the actual power generation It is characterized in that it includes a module.

The present invention can obtain the following effects by the configuration, combination, and use relationship described below with the present embodiment.

The present invention has the effect of improving the accuracy of prediction by correcting the weather information predicted by the numerical model using the weather information observed by the satellite, and predicting the amount of power generation of the photovoltaic device using this. .

The present invention has the effect of enabling accurate correction of predicted weather information by analyzing the correlation between predicted weather information and satellite-observed weather information by linear regression analysis to correct the predicted weather information.

The present invention has the effect of enabling accurate correlation analysis by matching the spatial domain and temporal resolution of predicted weather information and phase observation weather information.

1 is a configuration diagram of a solar power generation amount prediction system according to an embodiment of the present invention;
Figure 2 is a block diagram showing the configuration of the monitoring server of Figure 1;
3 is a block diagram showing the configuration of the preprocessing unit of FIG.
4 is a block diagram showing the configuration of the error correction unit of FIG.
5 is a block diagram showing the configuration of the machine learning performing unit of FIG.

Hereinafter, preferred embodiments of a machine learning-based solar power generation prediction system using real-time satellite data and numerical model data according to the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, if it is determined that a detailed description of a well-known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. Throughout the specification, when a part "includes" a certain component, it means that other components may be further included, rather than excluding other components, unless otherwise stated. Terms such as “…unit” and “…module” mean a unit that processes at least one function or operation, and may be implemented as hardware or software or a combination of hardware and software.

When the machine learning-based solar power generation prediction system using real-time satellite data and numerical model data according to an embodiment of the present invention is described with reference to FIGS. 1 to 5, the solar power generation prediction system is installed at a certain point and the A photovoltaic device (1) for generating electric power by light; a monitoring server 3 for predicting the weather at the point where the photovoltaic device 1 is installed, and predicting the amount of power generation of the photovoltaic device 1 by machine learning using the predicted weather information; and a display device (5) for displaying weather information and power generation predicted by the monitoring server (3).

The prediction of the amount of power generation by the conventional photovoltaic device 1 was made by performing a learning model such as machine learning using weather prediction information predicted by a WRF numerical model or the like. However, there is a problem in that the accuracy of the forecasting of the meteorological information is poor, and the prediction of the amount of power generation using it is also poor. Accordingly, in the present invention, the accuracy of weather forecast information and the accuracy of power generation prediction can be improved by correcting the weather forecast data by the WRF numerical module or the like using the meteorological data observed by the satellite.

The photovoltaic device 1 is installed at a certain point to generate power by sunlight, and may be formed such that a plurality of photovoltaic modules 11 are gathered to form a single power generation complex, and other homes and buildings It includes a solar module 11 installed on the back comprehensively. The power produced by the photovoltaic device 1 may be stored and supplied in a certain storage device or may be the subject of power trade, and the amount of power generated is accurately predicted by the system so that supply, distribution and sales plans are established. can do.

The monitoring server 3 predicts the weather at the point where the photovoltaic device 1 is located, and predicts the amount of power generation of the photovoltaic device 1 using the predicted weather information, and the predicted information is the It can be displayed through the display device (5). In particular, the monitoring server 3 corrects the weather forecast information using the weather information observed by the satellite for more accurate prediction of the amount of power generation, and analyzes the algorithm between the corrected weather forecast information and the amount of power generation information by machine learning. let it do To this end, the monitoring server 3 includes a weather forecasting measurement and collection unit 31 , a satellite data collection unit 32 , a preprocessing unit 33 , an error correction unit 34 , a machine learning execution unit 35 , and a power generation forecasting unit. part 36 .

The weather forecasting measurement and collection unit 31 is configured to collect weather forecast information for predicting power generation, and may collect weather information predicted by, for example, a Weather Research and Forecasting (WRF) numerical model. The weather forecast measurement and collection unit 31 may collect weather information such as temperature, humidity, wind speed, insolation, precipitation, cloudiness, etc. to predict the amount of power generation, and the collected information is compared with the weather information observed by the satellite and correlation is analyzed.

The satellite data collection unit 32 is configured to collect weather information observed by satellites, and collects real-time weather information from satellites installed for meteorological observation. The meteorological information collected by the satellite data collection unit 32 enables correlation analysis with the weather information collected by the weather forecasting measurement collection unit 31 for the same location and time, for example, insolation and cloudiness. , precipitation, etc. can be collected.

The pre-processing unit 33 is configured to analyze the space and time of the weather information for correlation analysis between the predicted weather information collected by the weather forecasting measurement and collection unit 31 and the satellite observation weather information collected by the satellite data collection unit 32 . As a configuration for matching the criteria, a spatial domain matching module 331 for matching spatial domains and a temporal resolution matching module 332 for matching temporal resolution may be included.

The spatial domain matching module 331 is configured to match the spatial domains of the predicted weather information and the satellite observation weather information, and allows the correlation analysis to be performed by matching the different grid domains with the same domain. The spatial region matching module 331 can match a high-resolution region of a 1km × 1km grid, and can be matched with weather information for a specific region by using various interpolation techniques such as linear interpolation, distance-weighted interpolation, and nearest interpolation. make it possible

The temporal resolution matching module 332 is configured to match the temporal resolution of the predicted weather information collected by the weather forecasting measurement collection unit 31 and the satellite observation weather information collected by the satellite data collection unit 32, and the same It is possible to analyze the correlation between predicted weather information and satellite observation weather information with respect to time.

The error correction unit 34 is configured to correct the predicted weather information according to the correlation between the weather prediction information and the satellite observation weather information, so that the accuracy of the predicted weather information can be increased. Since the weather information observed by the satellite has higher accuracy than the weather information predicted by the numerical model when compared with the actual weather information, the error correction unit 34 corrects the weather forecast information using the satellite observation information. By doing so, it is possible to increase the accuracy of forecasting weather information and the amount of electricity generated through it. To this end, the error correction unit 34 may include a weather forecast information loading module 341 , a satellite data loading module 342 , a linear regression analysis module 343 , and a weather forecast correction module 344 .

The weather forecast information loading module 341 loads the weather forecast information collected by the weather forecast and measurement collection unit 31, and loads the weather forecast information for a certain period of time. In addition, the weather forecast information loaded by the weather forecast information loading module 341 matches the satellite observation weather information with the spatial domain and temporal resolution by the pre-processing unit 33, and Correction is made through correlation analysis.

The satellite data loading module 342 loads the satellite observation weather information collected by the satellite data collection unit 32, makes the weather prediction information match the spatial domain and the temporal resolution, and compares the weather prediction information with the weather prediction information. correlation is analyzed through

The linear regression analysis module 343 is configured to analyze the correlation between the weather forecast information and the satellite observation weather information, so that the correlation between the weather forecast information and the satellite observation weather information is analyzed for a certain period using a linear regression equation. . The linear regression analysis module 343 analyzes the correlation of the same variables of weather forecast information and satellite observation weather information.

The weather forecast correction module 344 is configured to modify the weather forecast information by reflecting the linear regression equation analyzed by the linear regression analysis module 343 on the weather forecast information, so that the modified forecast weather information is used for prediction of power generation. do. Therefore, the weather information corrected by the satellite observation meteorological information is input as a predictor factor of the machine learning performing unit 35, the amount of power generated by the photovoltaic device 1 and machine learning are learned by machine learning, and an algorithm for predicting the amount of power through this This is derived so that the accuracy of power generation forecasting can be improved.

The machine learning performing unit 35 is configured to derive an algorithm for predicting the amount of power generation of the photovoltaic device 1 using weather prediction information, and a non-linear relationship between weather prediction information for a specific location and time and the amount of power generation to derive The machine learning performing unit 35 may use the XGBoost technique as an example, and may make a predictive model with high accuracy by repeating the steps of creating a classification rule by calculating weights between variables in a non-linear relationship. To this end, the machine learning performing unit 35 may include a predictive factor input module 351 , a power generation input module 352 , an algorithm generating module 353 , and an optimization module 354 .

The prediction factor input module 351 is configured to input weather prediction information for predicting the amount of power generation of the photovoltaic device 1, and the predicted weather information such as insolation, cloudiness, rainfall, temperature, humidity, wind speed, etc. is a prediction model to be entered in In particular, the prediction factor input module 351 may allow the predicted weather information corrected by the error correction unit 34 to be input, thereby increasing the accuracy of the generation amount prediction.

The power generation input module 352 is configured to input actual power generation information for a specific point and time into the prediction model, and updates the weights of the prediction model.

The algorithm generating module 353 is configured to derive an algorithm for predicting power generation while calculating weights by inputting predicted weather information and power generation data accumulated for a certain period to a prediction model, and the accuracy of the algorithm increases as data accumulates. .

The optimization module 354 is configured to verify the prediction model by comparing the generation amount information predicted according to the algorithm derived by the algorithm generation module 353 and the actual generation amount information, and the weights for each variable are updated This allows the prediction model to be optimized.

The generation amount prediction unit 36 is configured to predict the generation amount by the prediction model generated by the machine learning performing unit 35, and input the predicted weather information into the prediction model so that the predicted generation amount can be calculated, at this time Even in the predicted weather information, the information corrected by the error correction unit 34 may be used. In addition, the generation amount information predicted by the generation amount prediction unit 36 is compared with the actual generation amount information so that the prediction model can be verified.

The display device 5 is configured to display weather information predicted by the monitoring server 3, generation amount information, and the like, and various devices such as a computer and a mobile terminal may be applied.

In the above, the applicant has described various embodiments of the present invention, but these embodiments are only one embodiment that implements the technical idea of the present invention, and any changes or modifications as long as the technical idea of the present invention is implemented in the present invention should be construed as falling within the scope of

1: Solar power generation device 11: Solar module
3: Monitoring Server
31: weather forecasting and collecting unit 32: satellite data collecting unit
33: preprocessor 331: spatial domain matching module
332: temporal resolution matching module 34: error correction unit
341: weather forecast information loading module 342: satellite data loading module
343: linear regression analysis module 344: weather forecast correction module
35: machine learning performing unit 351: predictor input module
352: power generation input module 353: algorithm generation module
354: optimization module 36: power generation forecasting unit
5: Display device

Claims (5)

a photovoltaic device installed at a certain point to generate electric power by sunlight; a monitoring server that predicts the weather at the point where the photovoltaic device is installed, and predicts the amount of power generation of the photovoltaic device by machine learning using the predicted weather information; A display device for displaying weather information and power generation predicted by the monitoring server;
The monitoring server,
Correct the weather forecast information according to the correlation between the weather forecast information for a specific point and the meteorological observation information by the satellite and use it to predict the amount of electricity generated by the photovoltaic device;
The monitoring server,
A weather forecasting and collecting unit that collects weather information predicted by the WRF numerical model, a satellite data collection unit that collects weather observation information by satellite, and weather forecasting information and satellite observation meteorology to correct the collected weather forecast information A preprocessing unit that matches the spatial domain and temporal resolution of information, an error correction unit that corrects the forecasted weather information by analyzing the correlation between the collected weather forecast information and satellite weather observation information, and A machine learning performing unit for deriving an algorithm by machine learning, and a power generation forecasting unit for predicting the amount of power generation of the solar power generation device by inputting weather prediction information to the algorithm derived by the machine learning performing unit,
The error correction unit,
A weather prediction information loading module for loading the collected weather forecast information, a satellite data loading module for loading satellite observation weather information, and a linear regression equation for the correlation between the weather forecast information and the phase observation weather information for the same location at the same time It includes a linear regression analysis module that analyzes through
The preprocessor is
It includes a spatial domain matching module for matching spatial domains of weather forecast information and satellite data with a grid area of 1km × 1km, and a temporal resolution matching module for matching temporal resolutions of weather forecasting information and satellite data,
The machine learning performing unit,
A prediction factor input module for inputting weather prediction factors for predicting the amount of power generation, a generation amount input module for inputting generation information according to the input weather prediction factors, and a prediction algorithm for weather prediction factors and power generation using machine learning A solar power generation prediction system, comprising: an algorithm generation module to, and an optimization module for correcting the algorithm by comparing the generation amount predicted by the prediction algorithm with the actual generation amount. delete delete delete delete

Images