KR-20260066291-A - WEATHER FORECASTING METHOD, WEATHER FORECASTING SYSTEM AND OPERATION SUPPORT SYSTEM FOR POWER PLANT

KR20260066291AKR 20260066291 AKR20260066291 AKR 20260066291AKR-20260066291-A

Abstract

In order to provide reliable rainfall prediction information by merging the predicted rainfall generated by inputting actual rainfall into a deep learning model with the predicted rainfall collected from an external source, the weather prediction system according to the present invention includes a prediction processing unit that predicts rainfall based on rainfall-related data provided from an external source, wherein the prediction processing unit acquires actual rainfall data acquired from an observation station and predicted rainfall data generated from an external model, generates input data by converting the rainfall within the actual rainfall data according to the terrain of the observation point, generates predicted rainfall by inputting the input data into a weather prediction model, and generates a final predicted rainfall by applying different weights to the predicted rainfall data and the predicted rainfall according to a timeline set based on the observation time of the actual rainfall data.

Inventors

신홍준
윤상후
권태용
윤정현

Assignees

한국수력원자력 주식회사
대구대학교 산학협력단
(주) 솔텍시스템

Dates

Publication Date: 20260512
Application Date: 20241104

Claims (10)

It includes a prediction processing unit that predicts rainfall amount based on rainfall-related data provided from an external source, and The above prediction processing unit A weather forecasting system characterized by acquiring actual rainfall data obtained from an observation station and predicted rainfall data generated from an external model, converting the actual rainfall data according to the terrain of the observation point to generate input data, inputting the input data into a weather forecasting model to generate a predicted rainfall amount, and generating a final predicted rainfall amount by applying different weights to the predicted rainfall data and the predicted rainfall amount according to a timeline set based on the observation time of the actual rainfall data.
In Article 1, The above prediction processing unit A first weight is applied to generate the final predicted rainfall amount using the predicted rainfall amount in a first timeline from the observation point to the first point in time, and A second weight is applied to generate the final predicted rainfall amount using the predicted rainfall amount and the predicted rainfall data in a second timeline from the first time point to the second time point, and A weather forecasting system characterized by applying a third weight to generate the final predicted rainfall amount by utilizing the predicted rainfall data at a third timeline after the second timeline.
In Article 2, The above prediction processing unit When setting the second weight in the second timeline above, A weather forecasting system characterized by being set based on any one of the following: a linear operation method set through linear operation according to time point, a quantitative evaluation method set through quantitative evaluation of the actual rainfall data, a correlation analysis method set through correlation analysis of the actual rainfall data, and a reliability evaluation method set by evaluating the reliability of the prediction result according to time point.
In Article 1, The above weather forecasting model is Extracting temporal and spatial features of the above input data, and Enhancing the above extracted features, and A weather forecasting system characterized by generating the predicted rainfall amount by restoring the above-mentioned enhanced features to the size of the above-mentioned input data.
In Paragraph 4, The above weather forecasting model is A weather forecasting system characterized by including ConvLSTM-Unet.
In Article 1, The above measured rainfall data It includes rain gauge-based rainfall data and radar-based rainfall data, The above prediction processing unit Perform an analysis of the elevation rainfall of each of the above rain gauge-based rainfall data and the above radar-based rainfall data, and generate topographic rainfall data in which elevation rainfall amounts converted according to the topography of the region are input into each grid separating a predetermined region. A weather forecasting system characterized by generating input data for each of the above grids by utilizing at least one of the radar-based rainfall data and the topographic rainfall data according to preset conditions.
In Article 1, The above weather forecasting model is Predicting rainfall at preset time intervals, A weather forecasting system characterized by using a rainfall forecast value as an input value for rainfall forecasting after the above time interval.
In Article 1, The above time interval is A weather forecasting method characterized by being 10 minutes.
A step of acquiring actual rainfall data obtained from an observation station and predicted rainfall data generated from an external model; A step of generating input data by varying the rainfall amount within the above-mentioned actual rainfall data according to the topography of the observation point; A step of inputting the above input data into a weather prediction model to generate a predicted rainfall amount; and A weather forecasting method comprising the step of generating a final predicted rainfall amount by applying different weights to the predicted rainfall data and the predicted rainfall amount according to a timeline set based on the observation time of the actual rainfall data.
It includes a prediction processing unit that predicts rainfall amount based on rainfall-related data provided from an external source, and The above prediction processing unit A power plant operation support system characterized by acquiring actual rainfall data obtained from an observation station and predicted rainfall data generated from an external model, generating input data by converting the rainfall amount within the actual rainfall data according to the topography of the observation point, generating predicted rainfall amount by inputting the input data into a weather prediction model, and generating a final predicted rainfall amount by applying different weights to the predicted rainfall data and the predicted rainfall amount according to a timeline set based on the observation time of the actual rainfall data.

Description

Weather forecasting method, weather forecasting system and operation support system for power plant The present invention relates to a weather forecasting method, a weather forecasting system, and a power plant operation support system, and more specifically, to a weather forecasting method, a weather forecasting system, and a power plant operation support system that predicts rainfall based on a deep learning model. Generally, hydroelectric power plants utilize meteorological information to ensure stable dam operation during heavy rainfall. For example, hydroelectric power plants can perform flood control by calculating the discharge volume using hourly rainfall and determining whether to open the gates based on the calculated discharge volume. Recently, real-time rainfall is being predicted based on geographical information of the dam basin and historical meteorological data, and the predicted meteorological information is being used for dam operation. Conventional technology regarding a weather forecasting method for such dam operation is disclosed in Korean Registered Patent Publication No. 10-1285044 (Method for determining hydrological discharge volume during flood season of a power generation dam, July 4, 2013). The above-mentioned registered invention calculates an area-averaged rainfall based on short-term and long-term weather forecasting information calculated from radar weather forecasting information and actual weather measurement information, and generates a rainfall event based on the calculated area-averaged rainfall. In particular, in the case of narrow basins, water levels fluctuate rapidly during heavy rainfall, and ultra-short-term rainfall data in 10-minute intervals is required for the stable operation of power plants. However, conventional technology had a problem in that rainfall prediction models had low accuracy in short time periods of less than one hour, making actual application difficult. FIG. 1 is a conceptual diagram showing a power plant operation support system according to the present embodiment. Figure 2 is a conceptual diagram showing actual rainfall data. FIG. 3 is a flowchart illustrating a weather forecasting method of a power plant operation support system according to the present embodiment. Figure 4 is a flowchart showing the grid-specific conditional synthesis technique. Embodiments of the present invention will be described in detail below with reference to the attached drawings. However, the embodiments disclosed below are not limited to those disclosed below and may be implemented in various forms; the embodiments provided are merely intended to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention. The shapes of elements in the drawings may be exaggerated for clearer explanation, and elements indicated by the same reference numeral in the drawings represent the same element. FIG. 1 is a conceptual diagram showing a power plant operation support system according to the present embodiment, and FIG. 2 is a conceptual diagram showing actual rainfall data. As illustrated in FIGS. 1 and 2, the power plant operation support system (100, hereinafter referred to as the 'operation support system') according to the present embodiment can obtain actual rainfall data (10a) from an observation station (10) and obtain predicted rainfall data (30a) through an external server (30). The operation support system (100) can generate a predicted rainfall amount based on the actual rainfall data (10a) and the predicted rainfall data (30a). The operation support system (100) can provide the generated predicted rainfall amount to a user to support the user in operating the power plant. Here, the observation station (10) may be for observing rainfall around the target power plant. Multiple observation stations (10) may be provided, and each observation station (10) may be equipped with a rain gauge or radar equipment for measuring rainfall. The observation station (10) may measure rainfall using a rain gauge or radar equipment over a certain range of areas. Therefore, the actual rainfall data (10a) may include rain gauge-based rainfall data and radar-based rainfall data. For example, the measured rainfall data (10a) may be a converted form of rainfall measured from each observation station (10) (see FIG. 2). The measured rainfall data (10a) may include a grid (A) that separates a predetermined area and a rainfall distribution field (B) generated based on the measured rainfall amount. However, the observation station (10) may optionally be equipped with a rain gauge or radar equipment, and thus, the rainfall distribution field (B) represented by the rain gauge-based rainfall data and the radar-based rainfall data representing the rainfall distribution field at the same time may differ slightly from each other. An external server (30) generates predicted rainfall data (30a) based on an external model (31