KR102073768B1 - Drought information supply system based on portal - Google Patents

Abstract

The present invention relates to a portal-based drought information providing system which performs integrated production and management upon drought data, which have been individually produced by water-related organizations to be utilized corresponding to a field controlled by the organizations, according to a standard management system. Therefore, a foundation is provided such that a pre-emptive response can be taken corresponding to the drought data which are consistently and precisely analyzed by a standardized criterion, thereby minimizing the damage of a drought. In addition, the function of a drought forecast, which citizens can easily access like a weather forecast, is provided in order to display actions corresponding to a drought policy. To this end, in regard to a groundwater-based drought step, a floodgate-based drought step, and an underground water-based drought step, current state data and prediction data for each administrative district are generated such that anyone can inquire into and utilize the same through a portal web server (30). Furthermore, a service specialized for a manager of each administrative district and each water-related organization is provided in order to provide a foundation for guaranteeing an administrative drought response.

Description

Drought information supply system based on portal}

The present invention is integrated production management according to the standard management system of the drought data produced by each of the water-related institutions and used according to the area under the jurisdiction of the institution, according to preemptive response according to the drought data analyzed consistently and accurately according to the standardization standards. It provides a framework for minimizing drought damage, and provides a portal-based drought information provision system that functions as a drought forecaster that allows people to easily approach and respond to drought policy as weather forecasts.

As the frequency and intensity of droughts in Korea and abroad are increasing, it is necessary to shift the national drought response system from the recovery of post-damage recovery to the drought response system. To this end, there is an urgent need to establish a system that can proactively establish countermeasures by forecasting or warning short and long term droughts.

However, drought data is distributed and produced by relevant departments or agencies in each field such as meteorology, agriculture, life and industry, so that they can be used on their own or shared by local governments. It was unable to respond effectively.

Accordingly, the applicant of the present invention is the data of each institution including the observation data of the water source divided into dams, reservoirs, rivers and groundwater, the water supply system and facility data, the weather station (or rainfall observation station) status and observation / forecast data Data collection, establish a standardized database management system, establish standardization standards for drought judgment, drought index data for each district, forecast data of drought stage, drought status data, and big data of drought-related news. A drought information portal (http: //drought.kwater.or.) That generates a variety of data such as analysis data, drought stage response measures, and national behavioral guidelines, and can easily select and utilize the data not only for water related organizations but also for the general public. kr or http://www.drought.go.kr) (see Non-Patent Document 2).

However, the drought information portal in operation provides various data that can be easily searched and utilized by various institutions and local governments related to water, but specialized personnel can manage the drought data of the administrative districts in charge. It lacks functionality. In other words, while providing various and accurate drought data for national administrative districts, in practice, the drought data should be utilized in connection with the water supply system of the jurisdiction and its operation, and the drought data in the jurisdiction can be easily recognized. In addition, various specialized functions are required, such as registering the measures taken in response to drought, managing them in aggregate, and using them as statistical data to use as basic data for future drought response.

In addition, the Bayes-ESP technique (see Non-Patent Literature 1), which is an improvement of the Ensemble Streamflow Prediction (ESP) technique using the flow ensemble, is applied to predict the hydrologics of the water sources necessary for the determination of the drought stage. Although it could be increased, it was found that the distribution of the flow ensembles was not considered, and it was necessary to upgrade Bayes-ESP.

The standardized groundwater level index (SGI) should be calculated according to the monthly monthly groundwater level distribution, and there are underground observation stations where historical observation data reflecting the groundwater level distribution are still insufficiently accumulated. In addition, the reliability of the calculated SGI was difficult to secure.

In addition, SGI was predicted after learning artificial neural networks with 1-12 months duration data of Standardized Precipitation Index (SPI), where a large number of neurals were placed in the silvering layer for deep learning. The operation of the system was burdened by the administrative districts.

KR 10-1718294 B1 2017.03.14. KR 10-2017-0005553 A 2017.01.16. KR 10-1954570 B1 2019.02.26.

 Development and Evaluation of Bayesian Based Dam Estimation Inflow Estimation, v.50 no.7, 2017, pp.489-502  2017 Annual Report on Drought Information Analysis, http://www.drought.go.kr, Professional Information-Resources-'2017 Annual Report on National Drought Information Analysis Center ', 2018-01-29.

Accordingly, an object of the present invention is to quickly identify and manage drought data in a jurisdiction under specialized connection environments by water-related organizations or local governments, to ensure drought response, and to provide drought data by providing more accurate drought data. It is to provide an advanced portal-based drought information providing system with improved accuracy.

In order to achieve the above object, the present invention provides a portal-based drought information providing system, the past hydrological observation data, past meteorological data, and water supply source of groundwater which is a source of water supply source and non-water supply area divided into dams, reservoirs and rivers Database of historical basic data including past water supply data, past drought index data, water supply system data by administrative districts, drought judgment criteria by administrative districts, and log-in information for each administrative district and water related organizations. DB server 10; A data collection module 21 for collecting and basicizing current basic data and weather forecast data; An index-based drought analysis module 22 for calculating and databaseing an index-based drought stage by calculating current and future drought indices for each administrative district according to basic data and weather forecast data; In the drought index, the SGI (Standardized Groundwater level Index) for each administrative district is calculated according to the basic data, and the future value is estimated according to the SPI based on the correlation with the SPI (Standardized Precipitation Index). And groundwater drought analysis module 23 for determining the current and future groundwater drought stages and database; A monitoring module 24 for determining and databaseting the current hydrological-based drought stage according to the basic data for each administrative district according to the drought criteria in relation to the drought by the water supply source; Regarding the drought caused by water supply sources, the future hydrology according to the source and the weather forecast data is predicted by the rainfall-flow model by the source, and the future hydrologic drought stage is determined by administrative district according to the drought criteria. A drought outlook module 25 for database; General service that provides database data in non-login status, monitoring / provisional status board which collects database data of the administrator's jurisdiction in the login status of the personnel, and the database data of the administrative region selected by the staff for the purpose of inquiring. A decision support service that provides a comprehensive drought situation board including a status board that collects the data, provides a report that records the data of the monitoring / forecast situation board, and registers and records the emergency water supply status data of the administrative area of the person in charge. A portal web server 30 that provides an emergency water supply management service for making a query and database the emergency water supply status data; It includes.

According to one embodiment of the present invention, the drought comprehensive situation board is a drought including the alternative source information retrieved according to the water supply system data, the source information of the emergency jurisdiction surrounding jurisdiction administrative districts and drought response plan specialized to the administrative districts or related institutions It shows a response strategy and includes a drought analysis status board that allows you to simulate a drought response strategy that corresponds to a randomly chosen drought stage.

According to an embodiment of the present invention, the drought prospect module 25 learns the rainfall-flow model for each source according to the past basic data, and applies the weather ensembles obtained from past weather data to the rainfall-flow model to apply the flow ensemble. After that, the weighted average value of the flow ensemble is estimated by applying the probability of weather forecast data, and the hydrologic drought stage is determined by applying the stochastic hydrologic according to the predicted flow rate.

와

의 유량

로 획득한 확률적 수문을 사용하며, 사후 분포 관계식에서,

및

는 과거 수문관측자료의 유량 평균 및 분산이고,

및

는 미래 예측 시점의 유량 앙상블 맴버의 평균 및 분산이며,

,

및

는 과거 기초자료로 얻는 유량 앙상블 평균

및 수문관측 유량

사이의 상관을 나타내는 선형 회귀 모델

의 절편

, 기울기

및 잔차

의 분산

으로 한다.According to an embodiment of the present invention, the drought prospect module 25 has a posterior distribution relationship.

Wow

Flow rate

Using the stochastic hydrograph obtained by

And

Is the mean and variance of the flows of past hydrologic observations,

And

Is the mean and variance of the flow ensemble members at the future forecast,

,

And

Is the mean of the flow ensemble obtained from the past basic data.

And hydrologic flow

Linear regression model representing the correlation between

Intercept

, inclination

And residuals

Dispersion

It is done.

According to one embodiment of the invention, the flow ensemble is converted to have a normal distribution and then used.

According to an embodiment of the present invention, the drought prospect module 25 further obtains the quantitative hydrologic obtained by predicting the flow rate obtained by applying the current weather forecast data to the rainfall-flow model, and then the probabilistic hydrological and quantitative hydrologic Among them, one hydrograph is selected according to a predetermined selection rule, and a future hydrologic drought stage is determined according to the selected hydrology.

According to an embodiment of the present disclosure, the preset selection rule is a rule for selecting a low value among quantitative and probabilistic sluices.

According to an embodiment of the present invention, the drought prospect module 25 predicts the weekly and monthly hydrology using the ABCD model and the TANK model as the rainfall-flow model, but determines the monthly hydrologic drought stage. A relatively low flow rate is applied after the ABCD and TANK models are run simultaneously.

According to an embodiment of the present invention, the groundwater drought analysis module 23 calculates the monthly probability density function (PDF: probability density function) according to the probability distribution of the groundwater level observed from the past hydrological observation data on a daily basis. It is obtained by the KDE (kernel density estimation) technique. The current SGI is determined by determining the current groundwater drought stage by obtaining a qunatile value of the monthly mean groundwater level from the probability density function of the current month.

According to an embodiment of the present disclosure, the groundwater drought analysis module 23 may include an input layer for receiving a 1 to 12-month sustained unit SPI obtained until the prediction period has elapsed and an SGI obtained before the prediction period has elapsed; The artificial neural network having a hidden layer composed of neurons exceeding a predetermined number and an output layer outputting an SGI when a prediction period has elapsed is trained as basic data, but preset by pruning the neurons of the hidden layer. After reducing and learning by count, the future SGI is predicted according to the data that has been databased so far.

According to an embodiment of the present invention, the groundwater level at the present time to be applied to select the current SGI according to the quantile in the groundwater drought analysis module 23 is the observed groundwater level at the present time instead of the monthly average groundwater level. .

The present invention constituted as described above is determined by providing an index-based drought stage, a hydrology-based drought stage and a groundwater drought stage for each administrative region according to a standardized method, which is consistently unified and drought according to the business area of the administrative region and water related institutions. In addition to providing a foundation to cope with the problem, anyone can respond to droughts by using the portal web server to look up and utilize them, and use services specialized for administrative districts and water-related institutions to increase the value of drought data and Make effective use.

According to an embodiment of the present invention, by providing a drought analysis situation board, it is possible to induce drought response cooperation between personnel in each administrative district and water-related organizations, as well as to make full use of the drought response by using a virtual simulation.

According to an embodiment of the present invention, since the weather forecasting probability is reflected in the prediction of the hydrology using the weather ensemble and the flow ensemble according to the ESP technique, the prediction accuracy may be improved by adapting to the current weather forecast and predicting the hydrology.

According to an embodiment of the present invention, by using the improved Bayes-ESP method reflecting the probability distribution of the flow ensemble, it is possible to further increase the accuracy of hydrologic prediction.

According to an embodiment of the present invention, by obtaining both quantitative and probabilistic sluice, and by using both the ABCD model and the TANK model for the rainfall-runoff model, even if the water circulation process appears variously by watershed basin and administrative district, This situation can be predicted without omission.

According to an embodiment of the present invention, a standardized groundwater level index (SGI) is calculated according to a probability density function (PDF: probability density function) obtained as a daily observation groundwater, and predicted by an artificial neural network optimized for hidden layers. Therefore, the groundwater drought stage can be obtained more accurately.

1 is a block diagram of a portal-based drought information providing system according to an embodiment of the present invention.
2 is a detailed block diagram of the groundwater drought analysis module 23 and the drought prospect module 25 provided in the drought analysis server 20.
3 is a view showing a process of calculating the Standardized Groundwater Level Index (SGI).
4 is a view showing the difference between the monthly groundwater level distribution (a) and the daily groundwater level distribution (b).
5 shows an artificial neural network (a) and an input dataset.
6 is a schematic diagram of a rainfall-flow model.
7 shows exemplary data of a weather ensemble and flow forecasting process by the Bayes-ESP company.
8 is a screen configuration diagram of a drought general situation board provided by the portal web server 30.
9 is an emergency water supply performance registration screen configuration provided by the portal web server 30.
10 is a block diagram of the water supply support results provided by the portal web server 30.

Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention.

In addition, in describing the embodiments of the present invention, when it is determined that a detailed description of a related known function or known configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted and for the purpose of understanding. Reference is made to known documents as deemed necessary.

Referring to the block diagram shown in Figure 1, the portal-based drought information providing system according to an embodiment of the present invention is a DB server 10 of the database prepared in advance, the water management information distribution system linked through a communication network ( The drought analysis server 20 collects and basicizes the current basic data from WINS, 1) and generates and analyzes the drought analysis data, and the portal web server providing the databased data to clients connected through a communication network. It is configured to include).

The drought analysis server 20 generates the index-based drought stage determined according to the family index, and the hydrology-based drought stage determined according to the hydrology of the water source as daily data as well as weather forecast data and improved methods. The forecast data of the index-based drought stage and the hydrologic-based drought stage, which are predicted according to the predicted hydrologic water source, are generated as monthly data up to 3 months on the designated date every month and weekly data up to 1 month on the designated day of each week. That is, 1/2/3 month forecast data is generated once a month, so that you can forecast / alarm up to 3 months per month, 1/2/3/4 week forecast data is generated once a week, Make it a view.

Here, the water gate of Suwon is predicted according to a technique using at least two rainfall-runoff models (or hydrologic models), an improved Bayes-ESP technique, and a combination of quantitative and probabilistic predictions, thereby increasing prediction accuracy. . Since rainfall is observed at stations located at designated locations all over the country, the rainfall observed at the point is converted to area precipitation of watershed unit or administrative unit according to thissen method.

The drought level determined by the source is applied to determine the drought level of the administrative area supplied from the source, so that the drought level is obtained by the administrative area.

In addition, the drought analysis server 20 separately calculates and predicts the Standardized Groundwater Level Index (SGI) during the drought land to generate the status data and prospect data of the groundwater drought stage.

Here, SGI collects the monthly groundwater level observed for many years, obtains the monthly groundwater level distribution, and calculates it based on it, so it is to be generated on the basis of the end of each month. According to an embodiment of the present invention, the groundwater level observed daily is used as it is, which is different from the conventional method using the monthly mean groundwater level probability distribution, thereby obtaining a more accurate probability distribution from data of a limited observation period, and using KDE ( By using the probability density function that generalizes the probability distribution with the kernel density estimation technique, it is possible to estimate the spread density that faithfully reproduces the histogram of the original data even if the probability distribution varies with each groundwater station. In addition, Thiessen network is constructed according to the groundwater observation network, and the area-weighted average SGI is obtained.

In addition, the SGI predicts the value of 1 month / 2 months / 3 months after the end of each month to use for example / alarm, for this purpose, SGI (Standardized Precipitation Index, SPI of 1 to 12 months duration) It is used to learn artificial neural network that receives standard precipitation index) and outputs predicted SGI after 1 month. Here, the artificial neural network is provided for each administrative district and monthly, and by simplifying and optimizing the number of neurons in the hidden layer by initially pruning method to reduce the number to a predetermined number.

As such, the family analysis data including the drought index, the index-based drought stage, the predicted hydrology, the hydrologic drought stage, the SGI and the groundwater drought stage are databased in the DB server 10, and the historical data for the generation of the sentence analysis data, Drought criteria, etc. are databased in advance in the DB server 10, and continuously updated during the system operation process.

The portal web server 30 is a portal-based server for providing database-based data through a communication network. In addition to the data provided in a non-login state, a service specialized in a jurisdiction in a log-in state of a representative of each administrative district and water-related organizations is provided. To provide. Here, specialized services include decision support services, emergency water supply management services, and water supply support management services.

Hereinafter, with reference to the drawings will be described in more detail.

The preliminary data databased on the DB server 10 includes historical basic data, historical drought index data obtained according to the historical basic data, drought judgment criteria for each administrative district, water supply system data for each administrative district, and administration. Including the past emergency water supply data for each zone, and the login information of the person in charge per administrative district or water-related organizations are also stored in the DB server (10).

The basic data includes hydrologic observation data obtained by observing the hydrologic water of groundwater, which is a source of water supply and non-water supply area divided into dams, reservoirs and rivers, meteorological data, and water supply data of the water supply source. Databases are used to learn rainfall-runoff models (or hydrologic models) and neural networks as described below, and to obtain groundwater level probability distributions for SGI (Standardized Groundwater Level Index). Of course, the locations of water sources and stations in various parts of the country are databased by linking them with hydrologic observation data, and the locations of stations that obtain weather data are also databased by connecting them with meteorological data.

As is known in the art to which the present invention belongs, the water gate of the water source may be water level, flow rate, low water amount, etc. according to the type of water source, the weather data may include rainfall, temperature, sunshine time, etc. The data may be related to runoff, including water consumption, discharge, and drainage of feedwater sources (dams, reservoirs, and streams).

Here, since the weather data are observation data for each weather station, it is necessary to include the data by the watershed basin converted using the Thiessen polygon, and also the data by administrative district.

The drought index data include Standardized Groundwater Level Index (SGI), Standardized Precipitation Index (SPI), Palmer Drought Severity Index (PDSI), Modified Surface Water Supply Index (MSWSI) Drought index data including feed index) and SMI (Soil Moisture Index) were calculated based on past basic data and then databased.

However, the SGI is preferably databased after obtaining the past value by applying the past basic data according to the method of calculating the current SGI in the groundwater drought analysis module 23 to be described later.

SPI database all indexes with a duration of 1 to 12 months to be used in the artificial neural network as described below, but in the portal web server 30, for example, SPI1, 3 months duration of 1 month Only SPI3 and 6 months SPI6 may be selected and provided.

The water supply system data is data related to the water supply system from the source to the demand source. For example, the water supply system data may be a path and facility data from the water source, the intake station, the water purification plant, the drainage basin, and the water demand source. This water supply system data is used as water supply system data for each district.

The drought criteria are selected for each source, as well as for determining the drought stage for each administrative area according to the water supply system data. Dams and reservoirs can be based on drought criteria as well as reservoirs, inflows and outflows (e.g. runoff based on water supply data), and streams can be characterized by water levels, flow rates, and sulfur curves. The level of drought can be used as the criteria for drought, or the 10-year frequency. The drought criteria can be collected and determined based on the source used by the sources in different parts of the country, but they can be generated and standardized according to the collected basic data. You may also do it. However, in the embodiment of the present invention, a separate criterion for determining the drought stage according to the size of the SGI related to the groundwater is provided for the non-water supply area.

On the other hand, according to the present invention, since the emergency water supply data is collected and databased as described below, by collecting the past emergency water supply data to the database, it is good to enable the inquiry of the data before the construction of the system.

In addition, a database of water quality measurement data related to the water quality measurement network and the measured water quality may be provided by the portal web server 30.

The log-in information is the log-in information of the person in charge of each administrative area divided into a regional government, a basic self-government, an eupmun-dong, or a person in charge of water-related organizations such as the Water Resources Corporation, the Rural and Fishery Corporation, the Ministry of Environment, the Ministry of Public Administration and Security, and the Flood Control Center. Contains information related to

The database data described above is updated by the drought analysis server 20 and the portal web server 30, as will be described later, the drought analysis server 20 is utilized for the drought status, drought prediction and drought warning.

The drought analysis server 20 grasps the current drought status for each administrative region and database the drought analysis data for each administrative region, which is obtained by predicting the future drought, in the DB server 10 to recall the drought analysis data for each administrative region. As a server that can be provided through the portal well server 30, the data collection module 21, index-based drought analysis module 22, groundwater drought analysis module 23, monitoring module 24 and drought prospects Module 25.

The data collection module 21 collects current basic data and current weather forecast data periodically or in real time to make a database. The basic data includes, as mentioned above, hydrologic observation data obtained by observing the hydrologics of dams, reservoirs, rivers and groundwater, meteorological data observed at meteorological stations, and water supply data measuring the water supply of water sources. In addition to the weekly forecast data, the Korea Meteorological Administration can use the 1- to 3-month forecast data of the National Meteorological Observatory branch. The basic data and the weather forecast data can be collected in conjunction with the Water Management Information Distribution System (WINS). Although it is possible to do so, it may be collected directly through the relevant institutions.

Based on the collected basic data, the data for each basin and administrative district are generated and further databased by meteorological data and weather forecast data. When the flow rate is calculated according to the watershed, when the drought index is calculated for each administrative district, It can be used for the purpose of estimating the amount of rainfall.

The radix-based drought analysis module 22 includes the drought index SPI (Standardized Precipitation Index), PDSI (Palmer Drought Severity Index), MSWSI (Modified Surface Water Supply Index) SMI (Soil Moisture Index) is calculated on a daily basis according to basic data, monthly and weekly based on weather forecast data, and the calculated and predicted drought index is calculated in administrative districts according to interpolation or Thyssen network. And then determine the current and future index-based drought levels for each district. When calculating the drought index, the drought index may be obtained for each administrative region by using weather data converted into administrative districts. The drought index is calculated by a known technique and changed in units of administrative regions, and thus detailed description thereof will be omitted.

The drought index calculation and index-based drought stage determination by administrative districts are conducted once daily, and the forecast for 1/2/3 months on a designated date every month, and the data including the drought index and the index-based drought stage are generated. The database is added to the existing database-based drought index data, and the index-based drought stages by administrative districts continue to accumulate after the system is newly created and created.

2 (a) is a detailed block diagram of the groundwater drought analysis module 23.

The groundwater drought analysis module 23 is a PDF acquisition unit 23a, SGI calculation unit 23b and drought stage determination unit for obtaining a standardized groundwater level index (SGI) once a month according to the basic data ( 23c) and artificial neural network learning unit (23d) and SGI for predicting future SGI using the standardized Precipitation Index (SPI) for 1 month, 2 months and 3 months. The prediction unit 23e and the drought stage prediction unit 23f are included.

The SGI calculation process and the groundwater drought stage determination process by the PDF acquisition unit 23a, the SGI calculation unit 23b and the drought stage determination unit 23c will be described with reference to FIGS. 3 and 4.

The PDF acquisition unit 23a estimates a PDF (probability density function) of the groundwater level observed in the past. For this purpose, as shown in Fig. 3 (a), the Gamma distribution is obtained from the histogram of the groundwater level data, and the PDF is estimated. However, since the groundwater level distribution is very diverse, it is difficult to treat it as a generalized distribution formula. Therefore, the applicant obtained a KDE (kernel density estimation) and used it as a PDF. As is known, it may be defined by the following equation.

는 지하수위 X에 대한 커널밀도를 나타낸다. In Equation 1, K is a kernel function, x _i is the groundwater level observed, n is the number of observation data, and h is a bandwidth for adjusting the smoothness of the kernel. , The number of bands is 100, KDE

Denotes the kernel density for groundwater level X.

As such, the PDF used for KDE is obtained by the groundwater level station and the month and applies to the month corresponding to the station location and the SGI calculation point.

The SGI calculation unit 23b calculates the SGI according to the current groundwater level at the time of calculating the SGI. That is, as shown in FIG. 3 (b), a CDF (cumulative distribution function) is obtained according to the probability distribution obtained by kernel density, and the cumulative probability of the current groundwater level in the CDF is shown in FIG. 3 (c). As described above, the SGI is calculated by projecting the CDF onto the CDF of the standard normal distribution to obtain the SGI.

The SGI obtained here is a value for the groundwater level station. Therefore, according to the position of the groundwater level station, a Tiessen polygon is constructed, and the weighted ratio of the area occupied by the station in each administrative area is applied to the SGI in the administrative area. Calculate

The drought stage determination unit 23c determines the current groundwater drought stage for each administrative region according to the size of the SGI calculated for each administrative region. SGI and groundwater drought stages obtained by the administrative region is databased in the DB server (30).

The SGI described above is calculated using the monthly probability density function PDF as a high and low level of the current groundwater level compared to an average year, and thus can reflect seasonal periodicity. Therefore, sufficient past observations are needed.

However, conventionally, by using the average monthly groundwater level, it was difficult to utilize realistic historical observations. For example, for a station having a 20-year observation period, each of the 20 monthly average groundwater levels can be used to obtain a probability density function of January to December, so it is difficult to reflect the monthly probability density of the actual groundwater level.

Thus, the embodiment of the present invention uses the past groundwater level observed daily as it is instead of the monthly average groundwater level. That is, after grouping the groundwater level observed for many years by month, the monthly probability distribution is obtained by applying the observed data instead of the average value. For example, for a station with a 20-year observation period, a monthly probability density function PDF using approximately 600 monthly groundwater level observations can be obtained. As a result, many observations use the probability density function PDF, which more accurately reflects the monthly distribution of the groundwater level, thereby making it possible to estimate SGI more accurately.

Referring to FIG. 4, since the groundwater level distribution (b) obtained according to the daily groundwater level is more complicated than the groundwater level distribution (a) obtained according to the monthly average groundwater level, the reality that the monthly distribution of the groundwater level is complicated It can be seen that it reflects faithfully.

On the other hand, the daily groundwater level may contain missing data, so missing data is used. For example, a method may be used to exclude measurement data when groundwater levels are actually measured below the measurable limits.

In addition, the current groundwater level for calculating the current SGI at the time of calculation was the monthly average groundwater level of the month to which the current time point belongs, but as an alternative embodiment of the present invention, the quartile of the groundwater level observed at the current time point instead of the monthly average groundwater level Depending on the quantile, the current SGI can be estimated. That is, as a specific example, since the SGI is calculated at the end of each month and provided in the next month, the SGI is calculated according to the groundwater level observed at the end of the month. As another alternative embodiment, the SGI may be calculated for an average value of the groundwater level obtained until the end of each month and a predetermined number of days before the end of each month.

The future SGI prediction process and the groundwater level drought stage determination process by the artificial neural network learning unit 23d, the SGI predicting unit 23e, and the drought stage prediction unit 23f will be described with reference to FIG. 5.

The artificial neural network learning unit 23d has one month, two months and three months based on the correlation between the SGI and the Standardized Precipitation Index (SPI) and the prediction of the SPI according to the weather forecast data. As a configuration for estimating the SGI at the elapsed time, an artificial neural network is prepared for each administrative region to reflect the correlation between the SGI and the SPI that are different for each administrative region.

SPI is a drought index divided into SPI1, SPI2, SPI3, SPI4, SPI5, SPI6, SPI7, SPI8, SPI9, SPI10, SPI11, and SPI12 according to the duration of 1 to 12 months. It is used in the field.

However, the SPI calculated by the duration was found to be different from SGI to the administrative districts. Therefore, all SPIs of 1 to 12 months were used as input data of the artificial neural network.

In the artificial neural network used in the embodiment of the present invention, as shown in FIG. 5 (a), an input for receiving a 1 to 12 month sustained unit SPI obtained until the prediction period has elapsed and an SGI obtained before the prediction period has elapsed Nonlinear Autoregressive model process with exogenous input having an input layer, a hidden layer composed of more than a predetermined number of neurons, and an output layer for outputting an SGI when a prediction period has elapsed. ) Model is used, and the output of the neural network can be expressed by Equation 2 below.

Where y _i is the predicted SGI, {x} is the vector of SPI1 to SPI12, {y} is the previous SGI vector, [U] and [V] are the weighted parameter matrix of the hidden layer, and {w} is Is the weighting parameter vector of the output layer, s and {b} are the value and vector of the bias. tanh () denotes an activity function of each neuron provided in the hidden layer.

Here, SPI1 to SPI12 and SGI are used as values obtained according to the monthly precipitation amount and the monthly average groundwater level, and the SGI at the time when the forecast period elapses is affected by the previous SGI and SPI, and in the embodiment of the present invention, the influence is The period of giving was set at 3 months. Of course, SPI and SGI are drought indices converted to administrative units.

Referring to Figure 5 (b) and look at in detail as follows.

In estimating the SGI on a monthly basis, to calculate the SGI at t + 1 at the end of one month, the SPI at t + 1 obtained according to the weather forecast data, the SPI and SGI at the current t, and one month ago SPI and SGI at t-2 and t-2 months ago as the input dataset, and the input dataset should be the same due to the characteristics of the NARX model. To complete. Any SGI at this time does not affect the output.

When predicting the SGI at t + 2 after 2 months, the SPI and SGI at t-1, t, t + 1 and t + 2 are input datasets, and the SGI at t + 1 is the artificial neural network. The predicted value is autoregressed to be entered. Of course, at time t + 2, the SPI is obtained as weather forecast data, and the SGI uses a random value as well. In this way, the SGI at t + 3 after 3 months can be predicted.

The artificial neural network learning unit 23d learns the artificial neural network based on past basic data. That is, the past SPI can be calculated according to the basic data obtained by past measurement as known, and the past SGI can be obtained by the past groundwater level in the SGI calculation unit 23b, and thus obtained according to the observation data. The data set is composed of SPI and SGI, inputted into the artificial neural network, and then reverse propagated according to the difference between the SGI predicted and output from the artificial neural network and the SGI at the prediction time according to the observation data.

However, in order to deepen the artificial neural network, it is better to increase the number of neurons in the hidden layer sufficiently, but since the correlation between the SGI and the SPI appears to be different for each administrative area, the number of neurons in the hidden layer of the artificial neural network to be used for prediction in the learning process It is desirable to simplify the model by reducing it to an appropriate number.

Accordingly, the Applicant sets the number of neurons in the hidden layer to 30, which is more than twice the number of the inputters at the beginning of learning in consideration of the number of input data, and pruning during the learning process to have 12 neurons. Here, pruning is a method of removing neurons one at a time according to the predictive performance of the artificial neural network, maintaining the predictive performance and reducing the number of neurons in the hidden layer.

Probability density functions obtained at daily groundwater levels and pruning techniques to reduce neurons in hidden layers were evaluated before and after performance. For the evaluation method, SGI obtained from 167 basic municipalities according to the basic data was evaluated as correlation coefficient between SGI predicted by cited neural network and actual observed SGI. This is to compare the number and ratio of zones before and after upgrading.

Correlation coefficient 0.9 or more 0.9-0.8 0.8-0.7 0.7 ~ 0.6 Less than 0.6 Before upgrading 43
(26%) 51
(31%) 53
(31%) 15
(9%) 6
(4%) After advancement 47
(28%) 58
(35%) 45
(27%) 13
(8%) 4
(2%)

Referring to Table 1, the number of administrative districts by correlation coefficient section decreased from 0.8 to 0.7 section, but the section with higher correlation coefficient increased, and the lower section tended to decrease. In addition, as a result of examining the increase and decrease of correlation coefficient for each administrative district, 167 (77%) administrative districts increased and 39 (23%) administrative districts decreased. Even if the correlation coefficient is lowered, the correlation coefficient is mostly lowered slightly when the correlation coefficient is higher than 0.8. Exceptionally, the correlation coefficient in one basic self-government was lowered to less than 0.7, but this may be due to the location of groundwater stations and the application of Thyssen network. Here, it is appropriate to use monthly average data rather than daily data in consideration of groundwater level fluctuating by various causes.

The SGI predicting unit 23e predicts SGI at a future time point by constructing an input data set using the SPI and SGI obtained according to the basic data databased to date, and inputting it into the artificial neural network as shown in FIG. Since SGI is calculated according to the monthly average groundwater level, it is preferable to calculate the SGI at the end of each month and to predict the SGI after 1 month, 2 months and 3 months by applying the calculated SGI.

The drought stage prediction unit 23f determines a 1/2/3 month groundwater drought stage according to the predicted 1/2/3 month SGI.

As described above, the groundwater drought stage estimated by the current SGI, the groundwater drought stage determined according to the current SGI, and the SGI predicted by the 1/2/3 month period, and the predicted SGI as of the end of each month obtained from the groundwater drought analysis module 23. Database.

In addition, the PDF acquisition unit 23a upgrades the probability density function according to the increase of the groundwater level data, and the artificial neural network learning unit 23d re-learns and upgrades the artificial neural network according to the increase of the SPI and SGI data.

On the other hand, the past SPI and SGI to be used when learning the first artificial neural network may be generated in advance and made into a database, but since the SGI may vary depending on the upgrade of the probability density function, it is recalculated according to the upgrade of the probability density function. Also good.

The monitoring module 24 compares the current basic data of the water supply source collected in relation to the dams, reservoirs and rivers with the drought judgment criteria, and determines the current hydrologic drought stage by the source. Therefore, the hydrologic drought stage of the administrative districts supplied from the water source is also determined, and the hydrologic drought stages of the determined source and administrative districts are databased. In some cases, drought criteria that reflect water supply data are used, as well as the hydrology of water sources.

The drought prospect module 25 is configured to predict the hydrologic drought stage of each district by the water supply source divided into dams, reservoirs and rivers, and predicts the future of 1/2/3 months on a designated date every month, and also every week. Predict the future of 1/2/3/4 weeks on a given day, and update the 1/2/3/4 and 1/2/3 month hydrological-based drought-level database drought analysis data weekly and monthly, respectively. The detailed block diagram shown in FIG. 2 (b) and FIGS. 6 and 7 will be described.

The drought prospect module 25 includes a rainfall-flow model learning unit 25a, a Bayes-ESP fitting unit 25b, a hydrologic prediction unit 25c, and a drought stage prediction unit 25d.

The rainfall-flow model learning unit 25a operates a rainfall-flow model for each water supply source to predict an inflow amount (or a water level, a flow rate) of the water supply source divided into a dam, a reservoir, and a river. Learn from past weather and hydrologic observations.

Runoff refers to the flow of rainfall down to the ground, and the rainfall-runoff model is a simplified model of the relationship between rainfall and runoff for watersheds that contribute to the inflow of water sources. This rainfall-runoff model has been used to predict the floodgates of water sources by predicting runoff for rainfall due to weather forecasts.

In an embodiment of the present invention, one ABCD model suitable for monthly simulation and one TANK model suitable for daily simulation are operated in parallel for each water source.

The ABCD model is a rainfall-flow model in which rainfall flows out through the soil layer and the groundwater layer, as shown in the schematic diagram shown in FIG. 6 (a), and includes four A, B, C, D parameters related to the runoff. In other words, the initial parameters of soil moisture and groundwater are expressed in simple equations as the main factors of watershed characteristics and runoff processes. Applicant added five parameters including T _rain , T _snow , meltmax, α and initial value Snostor _{o related} to the snow melting, and improved the model to reflect the delay effect of snow and snow by inputting precipitation, evapotranspiration, and average temperature. There is a bar. In addition, the parameter should obtain the optimal solution value so that the output value of the model obtained according to the input of the weather data converges to the hydrologic observation data as the target value, and for this purpose, the Shuffled Complex Evolution method developed at The University of Arizona).

The TANK model is a rainfall-flow model in which the rainfall flows out through the storage tanks from the 1st stage to the 4th stage, as shown in the conceptual diagram shown in FIG. It is expressed as a parameter using the outlet coefficient and the initial value as parameters. The parameters of the TANK model were also estimated as global optimal solutions based on historical weather and hydrologic observations.

The above-described ABCD model and TANK model are well-known models used as rainfall-runoff models, and further description thereof is omitted.

However, according to an embodiment of the present invention, the ABCD model suitable for the monthly prediction and the TANK model suitable for the daily prediction are operated according to the characteristics of the model, so that the monthly 1/2/3 month prediction and the daily prediction are repeated. In addition to making unit 1/2/3/4 week predictions, we also perform 1/2/3 month predictions with the TANK model, compare the results with the ABCD model, and adopt relatively low hydrologic flow rates. It was made. This is to predict drought without omission.

The Bayes-ESP fitting part 25b fits a Bayes-ESP algorithm for predicting drought caused by a water supply source divided into a dam, a reservoir, and a river.

Applicant predicted the flow rate or the hydrologic inflow rate by Bayes-ESP technique as described in Non-Patent Document 1 mentioned as background of the present invention.

The flow forecasting technique using the Ensemble Streamflow Prediction (ESP) combines the hydrologic state at the time of the forecast with the past meteorological data through the rainfall-flow model, assuming that past meteorological phenomena can be reproduced in the future. An empirical long-term hydrographing technique that predicts.

According to this, all weather scenarios that are likely to occur in the future, namely weather ensembles, are obtained from past weather data according to the assumption that weather phenomena are reproduced, and are classified by month, and the rainfall-runoff model applying the hydrologic state of the forecast time is determined by The meteorological ensemble is input to calculate the hydrologic inflow for each scenario, and each member of the flow ensemble is obtained. And the average value of the member of the flow volume ensemble obtained at this time, or the weighted average value of the member of the weather ensemble is estimated by flow volume.

Bayes-ESP utilizes Bayes' theorem to improve the accuracy of the ESP method. It is a pre-distribution of the inflow and the flow ensemble predicted by the ESP method when the inflow is observed. A posterior distribution of the flow rate is derived according to the Likelihood function representing the conditional probability distribution. The flow rate prediction value is obtained from the post-distribution at this time.

However, the Bayes-ESP technique used previously did not reflect the weather forecast data, so it was not possible to obtain the forecasted results by adapting to the current weather forecast. In the present invention, the existing Bayes-ESP technique is improved.

The Bayes-ESP fitting part 25b fits internal parameter values using historical data in order to apply the Bayes-ESP method considering the distribution of the flow ensemble members.

First, the theoretical basis is explained.

는 아래의 수학식 3으로 표현된다.Applying Bayes' theory, the posterior probability of the flow rate when the flow ensemble y occurs

Is expressed by Equation 3 below.

Here, p (θ) is a probability indicating a prior distribution, and can be obtained according to past hydrological observation data as a prior probability of the observed flood (flow or inflow), and can be converted into a normal distribution as follows.

P (y) is the marginal probability of the flow ensemble y.

는 우도 함수(Likelihood function)로서, 관측수문 θ가 주어졌을 때의 유량 앙상블의 조건부확률로 표현되며, 유량 앙상블 평균

, 즉 ESP 기법으로 전망하는 유량의 확률에 의해서 아래와 같이 표현할 수 있다.

Is the likelihood function, expressed as the conditional probability of the flow ensemble given the observed hydrograph θ, and the mean of the flow ensemble

That is, it can be expressed as follows by the probability of the flow rate predicted by the ESP method.

In addition, the flow ensemble mean and the observation hydrograph can be expressed by the following linear regression model having the intercept α, the slope β and the residual ε.

의 분포를 갖게 되고, 유량 앙상블의 멤버도 정규분포를 갖는다고 가정하면

의 분포를 갖게 된다.Here, suppose the residual ε has a normal distribution.

Suppose we have a distribution of, and the members of the flow ensemble also have a normal distribution.

Has a distribution of.

가

에 독립적(independent)이면 수학식 5의 우도함수는 다음과 같은 확률분포를 갖는다고 할 수 있다.And,

end

If independent, the likelihood function of Equation 5 has a probability distribution as follows.

은 평균

및 분산

을 갖는 정규분포로 표현할 수 있다.As shown in Equations 4 and 7, if the pre-distribution and the likelihood function has a normal distribution, the posterior distribution

Is average

And distributed

It can be expressed as a normal distribution with

를 Bayes-ESP 기법으로 예측할 수문(유량)으로 할 수 있다. After all, average

Can be used as the predicted hydrograph (flow rate) by the Bayes-ESP technique.

및

를 과거 관측수문의 분포에 따라 얻는다.The Bayes-ESP fitting portion 25b is of a pre-distribution so that the hydrologic prediction according to Equation (8)

And

Is obtained according to the distribution of past observations.

를 얻어야 하며, 이를 위해서 과거 기상자료를 입력으로 한 ESP 기법에 따라 과거 유량 앙상블을 얻어서, 유량 앙상블 평균 및 분산을 얻고, 과거 수문관측자료의 따른 관측수문(유량 또는 유입량)을 상기 수학식 6의 관측수문의 값으로 적용함으로써, 수학식 6 및 수학식 8로부터 α,β, 및

를 얻는다.In addition, the Bayes-ESP fitting portion 25b includes α, β, and the likelihood function.

For this purpose, the historical flow ensemble is obtained according to the ESP technique with the input of past meteorological data, and the flow ensemble average and variance are obtained. By applying the value of the observation hydrograph, α, β, and

Get

,

, α,β, 및

는 수원별 및 월별로 얻는다.sure,

,

, α, β, and

Is obtained by source and month.

The hydrologic prediction unit 25c predicts the hydrology by water source at 1/2/3 month elapsed by month or the hydrology by water source at 1/2/3/4 week elapsed by week. When predicting the hydrology for each source, either the quantitatively predicted flow rate or the stochastic predicted flow rate is selected. Here, the selection criterion was to select a lower value between quantitative and probabilistic hydrological. In addition, when forecasting hydrology for one week, TANK model suitable for daily prediction is used in rainfall-flow model, and relatively low flow rate is applied by using both ABCD model and TANK model for prediction of monthly hydrology.

The above quantitative hydrologic is obtained by applying the current weather forecast data to be predicted to the rainfall-runoff model, and may be water level or flow rate depending on the type of water source.

The probabilistic sluice is a sluice obtained by applying the weather ensembles obtained from past meteorological data to the rainfall-flow model to obtain the flow ensembles as predicted flow rates averaged by the Bayes-ESP fitting unit 25b. Predicted according to the fitted Bayes-ESP algorithm, described with reference to FIG.

First, the weather ensemble is obtained by dividing the historical precipitation, temperature, wind speed, and sunshine data by watershed into monthly data. Of course, the Bayes-ESP fitting portion 25b has not been described in detail, but the gaseous ensemble required for operating the Bayes-ESP fitting portion 25b is also used in the same manner.

Referring to the weather ensembles illustrated in FIG. 7 (a), the classifications are classified into 'less', 'similar' and 'many' by using similar ranges of average annual rainfall by administrative district and monthly, and representative of 'similar' members. Representative values of 'less' and 'large' were determined as weather ensemble members by referring to average values and similar ranges. As an example, the number of members of the weather ensemble is three, but is not limited thereto, and may be divided into four or more. Of course, when the flow ensemble is a basin unit, the weather ensemble uses a weather ensemble in the watershed basin unit.

Next, a flow ensemble is obtained by inputting a weather ensemble member to the rainfall-flow model to which the hydrologic state at the time of prediction is applied. Of course, the monthly weather ensemble corresponding to the forecasted time is input to obtain the flow ensemble for the month.

과 분산

을 얻는다.Average of members from the flow ensemble

And dispersion

Get

은 예측하려는 현재의 기상예보자료에서 멤버별 기상전망의 확률 가중치를 부여하여 얻는 가중 평균치로 한다. Average of flow ensemble members

Is the weighted average obtained by assigning the probability weights of each member's weather forecast from the current weather forecast data to be predicted.

Referring to the schematic diagram illustrated in FIG. 7 (b), the weather forecast is provided as a probability for each precipitation section, and thus, the probability value for each section of the weather ensemble is weighted and averaged by the members of the flow ensemble. FIG. 7 (b) is illustrated as forecasting flow rates for April, May and June, for example, April flow forecast Q _April is flow ensemble member Q _{low, April} , Q _{med, April} , Q _high, the probability of the forecast P _{low, April,} P _med, by summing weighted values _April, P _{high, April} corresponding to each member in _April.

Of course, three ensemble members were used, divided into three sections, but the number of members may be increased by further subdividing.

과 분산

를 상기 수학식 8에 대입하여 얻는

를 확률적 예측 유량으로 한다. 앞서 언급하였듯이, 상기 수학식 8에서,

,

,

,

및

의 값은 상기 Bayes-ESP 피팅부(25b)에서 의해서 정해져 있으므로,

를 산정할 수 있다.Next, the average of the members from the flow ensemble obtained here

And dispersion

Obtained by substituting Equation 8 into

Is a probabilistic predicted flow rate. As mentioned above, in Equation 8,

,

,

,

And

Since the value of is determined by the Bayes-ESP fitting portion 25b,

Can be calculated.

On the other hand, since Equation 8 is obtained by assuming a normal distribution, it is preferable to convert the flow rate data having skewness to have a normalized distribution. For example, the Box-Cox transform removes the distortion of the distribution and then applies it to the Bayes-ESP technique. As a specific example, the flow ensemble member is converted to have a normal distribution and then used. Similarly, prior distributions obtained from observation hydrographs can be transformed to have normal distributions.

The drought stage prediction unit 25d determines the drought stage for each water source according to the hydrology predicted for each water source by the hydrologic prediction unit 25c, and also determines the hydrologic drought stage of the administrative area supplied from the water source, and the database. Make up. Of course, the drought judgment criteria are applied to determine the hydrologic drought level by source, and the hydrologic drought level by administrative district. According to an embodiment of the present invention, a 1/2/3/4 weekly forecast, a 1/2/3 month monthly forecast, and a database.

The portal web server 30 is a basic data (hydrology observation data, weather data, water supply data), water supply system data and the drought analysis data added by the drought analysis server 20 databased in the DB server 30 It provides not only server access to the users, but also measures related to prevention, preparedness, response to drought stages, recovery and measures related to droughts, and emergency contact networks for drought response according to drought stages, and national behavioral tips by drought stage and water use purpose.

In the embodiment of the present invention, in addition to the general service unit 31 that provides the above-mentioned various database data in a non-login connection state, in order to provide a specialized service to the person in charge of the person in charge by administrative districts and water-related organizations, Decision support service unit 32, emergency water supply management service unit 33, and water supply support management service unit 34.

The general service unit 31 provides a water supply facility, a water supply system including a water source location, a specification, a water supply population, a water supply status, a current status, and a weather / water level / water quality / groundwater station location in association with a map, every month. Services that enable you to map, view, and compare historical and drought stages and groundwater drought stages based on drought and drought indexes, including updated SGI and daily updated SPI, PDSI, MSWSI and SMI, Daily, 1/2/3/4 week, or 1/2/3 month forecasts, depending on when the drought analysis data were generated, including current and forecasting index-based drought stages, hydrologic-based drought stages, and groundwater drought stages. Drought forecasting service displayed on the drought map in the administrative district, database corresponding to the administrative district corresponding to the location of the access terminal or the administrative district selected on the drought map It is a service that collects and displays drought analysis data in our neighborhood, compares the historical records of drought stages, recognizes trends, responds to drought stages, and guides people's actions. Access through the service. Of course, such a service is provided by constructing a web page that can be easily recognized by using the database materialized in the DB server 30.

The following components provide services that can be obtained by logging in to representatives of administrative districts and water-related organizations.

The decision support service unit 32 monitors the status of droughts and warnings for administrative districts, water sources, or watersheds, and provides a comprehensive status board for droughts to support decision-making in response to droughts.

Referring to the screen configuration shown in FIG. 8, the drought composite situation board is configured to display a monitoring / view situation board, a gathering situation board, and a drought analysis situation board, respectively.

As illustrated in FIG. 8 (a), the monitoring / prospect situation board automatically selects the jurisdiction and the date of change and enables changes (①). It can display the drought status (current drought stage), drought forecast (predicted drought stage), water supply system, emergency water supply status and drought news of the jurisdiction, and facilities (water source, observation, water supply facilities, etc.). Drought Map (②), Rainfall Status Window showing Rainfall-related Data (③), Hydrologic Status Window Showing Hydrologic Data (④), Drought Status and Historical Records are linked to the county and water population by drought stage. The drought status window (⑤), the drought prospect window (⑥) showing the drought forecast data, and the emergency water supply / drought news window (⑦) showing drought news related to emergency water supply data and jurisdiction. Ball, and is collected by the data showing the screen features a menu (⑧), which can be downloaded as a report format.

Collecting status panel is a status panel that enables the editing of items by collecting the data of selected items for the purpose of inquiring about by the person in charge among the many data displayed in the monitoring / viewing situation panel. The date (for example, the day) is set as the default and the area and date to be inquired can be changed (1), the check box (2) to select from the items provided in the monitoring / view board and the data for each item selected. A display window (③) to show and a setting menu (④) for storing the selected item for later use.

The drought analysis board shows drought response strategies, including alternative source information retrieved according to the water supply system data, source information for emergency water supply areas surrounding the emergency jurisdiction, and drought response plans specific to the administrative district or related agencies. When selected, it is possible to simulate a drought response strategy corresponding to the selected drought stage. Here, the source information around the alternative water source may be determined in advance for each administrative district, but may be determined according to the water supply system or may be determined according to the past water source replacement record or emergency water supply record.

The emergency water supply management service unit 33 provides a service to register, database and query the emergency water supply records of the jurisdiction.

Referring to the emergency water supply performance registration screen illustrated in FIG. 9, the person in charge provides an input window for registering the emergency water supply performance carried out in the competent administrative area. According to this, after selecting the area where emergency water supply was carried out, the water supply division (village water supply, small water supply facility, supply area, etc.) of the area is selected, and the water supply generation, water supply population and water source of the area are It will be automatically selected according to the data registered as the supply system data, and select the emergency water supply status divided into restricted water, transport water and restricted + transport water.

The emergency water supply records inputted by the input window are databased, and the emergency water supply history and statistics of the jurisdiction can be inquired. And, it is possible to inquire in the drought comprehensive situation board.

The water supply support management service unit 34 provides a water supply support management service that allows a person in charge to register a water supply support-related data to database and query the results. For example, K-water supports emergency waters using dams, beams, and water supplies to drought areas using various resources such as bottles and water trucks. In addition, each administrative district can use its support to respond to droughts.

Referring to the water supply support results inquiry screen illustrated in FIG. 10, the water supply support performance is divided into watersheds according to watersheds, water supply organizations, support water, industrial water, agricultural water, support uses, support areas, support sources, and Includes support supplies and allows you to select and register such items. In addition, statistical data may be generated and provided according to a support purpose, a support region, and the like.

As such, the emergency water supply results and the water supply support results were utilized in the drought analysis situation edition of the drought comprehensive situation edition.

Specific embodiments have been shown and described in order to illustrate the technical idea of the present invention, but the present invention is not limited to the same configuration and operation as the specific embodiments as described above, and various modifications are made without departing from the scope of the present invention. It can be carried out in. Accordingly, such modifications should also be regarded as falling within the scope of the present invention, and the scope of the present invention should be determined by the claims that follow.

10: DB server
20: drought analysis server
21: data collection module
22: Index based drought analysis module
23: groundwater drought analysis module
23a: PDF Acquisition Unit 23b: SGI Mountain Government
23c: drought stage determination unit 23d: artificial neural network learning unit
23e: SGI predictor 23f: Drought stage predictor
24: monitoring module
25: Drought Outlook Module
25a: Rainfall-leak model learning section 25b: Bayes-ESP fitting section
25c: hydrologic prediction unit 25d: drought stage prediction unit
30: portal web server
31: General Service Department 32: Decision Support Service Department
33: Emergency water supply management service department 34: Water supply support management service department

Claims (11)

Historical basic data, including past hydrologic observations of historical sources, historical weather data, and historical water supply of water sources, including water sources in dams, reservoirs and streams, and non-water supply areas, historical drought index data, and administrative districts. A DB server 10 having a database of water supply system data and drought judgment criteria for each administrative district, and storing login information for a representative for each administrative district and water-related organizations;
A data collection module 21 for collecting and basicizing current basic data and weather forecast data;
An index-based drought analysis module 22 for determining and databaseing an index-based drought stage by calculating current and future drought indices for each administrative district according to basic data and weather forecast data;
In the drought index, the SGI (Standardized Groundwater level Index) for each administrative district is calculated according to the basic data, and the future value is estimated according to the SPI based on the correlation with the SPI (Standardized Precipitation Index). And groundwater drought analysis module 23 for determining the current and future groundwater drought stages and database;
A monitoring module 24 for determining and databaseting the current hydrological-based drought stage according to the basic data for each administrative district according to the drought criteria in relation to the drought by the water supply source;
Regarding the drought caused by water supply sources, the future hydrology according to the source and the weather forecast data is predicted by the rainfall-flow model by the source, and the future hydrologic drought stage is determined by administrative district according to the drought criteria. A drought outlook module 25 for database;
General service that provides database data in non-login status, monitoring / provisional status board which collects database data of the administrator's jurisdiction in the login status of the personnel, and the database data of the administrative region selected by the staff for the purpose of inquiring. A decision support service that provides a comprehensive drought situation board including a status board that collects data, provides a report that records the data of the monitoring / forecast situation board, and registers and records the emergency water supply status data of the administrative area under the authority of the person in charge. A portal web server 30 that provides an emergency water supply management service for making a query and database the emergency water supply status data;
Including,
The drought prospect module 25
Predicting weekly and monthly hydrographs using the ABCD and TANK models as rainfall-runoff models, and applying relatively low flow rates after concurrently operating ABCD and TANK models in determining monthly hydrologic drought stages. doing
Portal based drought information system. The method of claim 1,
The drought comprehensive situation board
It shows drought response strategies including alternative source information retrieved according to the water supply system data, source information of the emergency jurisdiction's surrounding jurisdiction, and drought response plans specific to the administrative district or related institutions. Includes a drought analysis status panel that allows you to simulate response strategies
Portal based drought information system. The method of claim 1,
The drought prospect module 25
Based on past basic data, the rainfall-flow model for each source is studied, and the flow ensemble is obtained by applying the weather ensembles obtained from past weather data to the rainfall-flow model. To predict the hydrologic drought stage by applying the stochastic hydrograph according to the predicted flowrate.
Portal based drought information system. The method of claim 3, wherein
The drought prospect module 25
Posterior distributon relation

Wow

Flow rate

Using stochastic hydrographs obtained by
In the posterior distribution equation,

And

Is the mean and variance of the flows of past hydrologic observations,

And

Is the mean and variance of the flow ensemble members at the future forecast,

,

And

Is the mean of the flow ensemble obtained from the past basic data.

And hydrologic flow

Linear regression model representing the correlation between

Intercept

, inclination

And residuals

Dispersion

By
Portal based drought information system. The method of claim 4, wherein
After converting the flow ensemble to have a normal distribution,
Portal based drought information system. The method according to claim 3 or 4,
The drought prospect module 25
After obtaining the quantitative hydrologic data obtained by applying the current weather forecast data to the rainfall-flow model, and selecting one of the probabilistic hydrologic and quantitative hydrologic according to the preset selection rule, To determine future stages of hydrologic drought
Portal based drought information system. The method of claim 6,
The predetermined selection rule is a rule for selecting a low value between quantitative and probabilistic hydrologic
Portal based drought information system. delete The method of claim 1,
The groundwater drought analysis module 23
Monthly probability density function (PDF) according to the probability distribution of groundwater level observed from the previous hydrologic observation data is obtained by kernel density estimation (KDE) method. The probability density function is used to determine the current SGI and determine the current groundwater drought stage by obtaining a qunatile value of the monthly mean groundwater level.
Portal based drought information system. The method of claim 9,
The groundwater drought analysis module 23
1 to 12-month sustained unit SPI obtained until the forecast period has elapsed, an input layer for receiving SGI obtained before the forecast period has elapsed, a hidden layer composed of more than a predetermined number of neurons, and The artificial neural network having the output layer that outputs the SGI of the viewpoint is trained as basic data in the past, but it is reduced to a predetermined number by learning by pruning the neurons of the hidden layer, and the future SGI according to the database data so far. To predict
Portal based drought information system. The method of claim 9,
In the groundwater drought analysis module 23, the groundwater level at the present time to be applied to select the current SGI according to the quantile is
Instead of the monthly average groundwater level,
Portal based drought information system.

Images