KR102527744B1 - Apparatus and system for predicting and monitoring water odor using national reference standard data - Google Patents

Abstract

According to one embodiment, provided is an odorant substance prediction and monitoring device which comprises: a data collection unit collecting linked water quality data provided by an environmental facility in real-time as water quality data measured in the past or currently measured, collecting measured water quality data in real-time, which is different from the linked water quality data and includes at least one of phycocyanin, chlorophyll a, water temperature, and pH (hydrogen ion concentration) as the currently measured water quality data, collecting national reference standard data on historical concentration of odorant substances including geosmin and 2-methylisoborneol (2-MIB), and having the national reference standard data converted into a database; a prediction unit receiving data from the data collection unit and predicting the concentration of odor substances including geosmin and 2-MIB for at least one of phycocyanin, chlorophyll a, water temperature, and pH (hydrogen ion concentration) by using a prediction model learned through machine learning, wherein the prediction unit includes the prediction model including an artificial neural network and predicting the concentration of odorant substances through the artificial neural network; and a learning unit training the prediction model by using the linked water quality data, the measured water quality data, and the national reference standard data. According to an embodiment of the present invention, the concentration of odorant substances can be indirectly predicted from the water quality data capable of being automatically measured.

Description

Odor substance prediction and monitoring device and system using national reference standard data

This embodiment relates to a technology for predicting an odorant.

Since most of the water sources in Korea use lakes and rivers, they are easily exposed to naturally occurring or artificial taste and odor substances. The taste and smell from tap water negatively affect the smooth supply of drinking water and the securing of stable water resources, causing not only national health but also enormous economic losses. Appropriate countermeasures are required. Odor substances are periodically generated every summer in multi-purpose dams such as Paldang and Daecheong, and in the case of Paldang Dam, the concentration has been detected up to about 1,200 ng/L. causes

On the other hand, the analysis of odorous substances requires a complicated pre-processing process and expensive analysis equipment (GC/MS, ELISA). Therefore, the development of online monitoring system technology that can provide feedback through on-site analysis results is required. In response to this, an attempt is being made to develop odorous substance analysis technology using some highly sensitive sensors. Accordingly, Korea Water Resources Corporation is developing a 'real-time automatic analysis system' that can quickly detect taste and odor substances in water purification plants. However, the Purge&Trap (pre-processing)-GC/MS- or SPME (pre-processing)-GC/MS- technology used by this system has problems in that the pre-processing is difficult and takes a long time.

Accordingly, the inventor of the present invention has completed the present invention after long research and trial and error.

Against this background, one purpose of this embodiment is to predict phycocyanin, chlorophyll-A (Chl-a; chlorophyll a), water temperature through a prediction model learning linked water quality data, measured water quality data, and national reference standard data. and pH (hydrogen ion concentration) to provide a technology for predicting and monitoring odorant concentrations.

Meanwhile, other unspecified objects of the present invention will be additionally considered within the scope that can be easily inferred from the following detailed description and effects thereof.

In order to achieve the above object, an embodiment collects in real time linked water quality data provided from environmental facilities as water quality data measured in the past or currently measured, and is different from the linked water quality data and currently measured water quality data. phycocyanin, chlorophyll-A (Chl-a; chlorophyll a), water temperature and pH (hydrogen ion concentration) are collected in real time, and geosmin and 2 -A data collection unit that collects national reference standard data on odorant concentrations in the past, including MIB (2-Methylisoborneol; 2-methylisoborneol), and stores the national reference standard data in a database; And receiving data from the data collection unit, using a prediction model learned through machine learning, phycocyanin, chlorophyll-A (Chl-a; chlorophyll a), water temperature and pH (hydrogen ion concentration) a prediction unit for predicting concentrations of odorants including geosmin and 2-MIB for at least one of the odorants, wherein the prediction unit includes: a prediction model including an artificial neural network and predicting concentrations of odorants through the artificial neural network; and a learning unit configured to learn the prediction model using the linked water quality data, the measured water quality data, and the national reference standard data.

In the device, the measured water quality data further includes a geosmin measurement value and a 2-MIB measurement value, and the learning unit includes a phycocyanin measurement value, a geosmin measurement value, and a 2-MIB measurement value included in the measured water quality data. value, and the geosmin standard value and 2-MIB standard value included in the national reference standard data can be used as test data for testing the learning performance of the prediction model.

In the device, the measured water quality data may further include a total nitrogen (TN) measurement value, a total phosphorus (TP) measurement value, and a total organic carbon (TOC) measurement value.

In the apparatus, an analysis unit extracts water quality factors from the linked water quality data, the measured water quality data, and the national reference standard data, and analyzes a correlation between the water quality factors, and the learning unit includes Geosmin and 2- The predictive model may be trained by using a water quality factor having a correlation with MIB as an input variable.

Another embodiment is an odorant predicting and monitoring device; a linked data providing device for providing the linked water quality data to the odorant prediction and monitoring device; a measurement data collection device including a measuring instrument for measuring the measured water quality data and a remote terminal unit (RTU) for transmitting the measured water quality data to the odorant predicting and monitoring device; and a national reference standard data providing device for providing the national reference standard data to the odorant predicting and monitoring device.

As described above, according to the present embodiment, the odorant concentration can be predicted indirectly from water quality data that can be automatically measured.

Further, according to the present embodiment, the utilization of the national reference standard data can be increased by using the national reference standard data to predict the odorant concentration.

On the other hand, even if the effects are not explicitly mentioned here, it is added that the effects described in the following specification expected by the technical features of the present invention and their provisional effects are treated as described in the specification of the present invention.

1 is a block diagram of an odorant prediction and monitoring system according to an embodiment.
2 is an exemplary diagram of national reference standard data including concentrations of odorous substances according to an embodiment.
3 is a configuration diagram of an apparatus for predicting and monitoring odorous substances according to an embodiment.
4 is a flowchart illustrating an operation of an apparatus for predicting and monitoring odorants according to an exemplary embodiment.
5 is a diagram for explaining characteristics of a substance that causes an odorous substance according to an exemplary embodiment.
6 is a flowchart illustrating an operation of learning a predictive model by analyzing correlations of water quality factors according to an embodiment.
7 is a first exemplary view of analysis results of correlations between water quality factors according to an embodiment.
8 is a second exemplary view of analysis results of correlations between water quality factors according to an embodiment.
It is revealed that the accompanying drawings are illustrated as references for understanding the technical idea of the present invention, and thereby the scope of the present invention is not limited.

In the description of the present invention, if it is determined that a related known function may unnecessarily obscure the subject matter of the present invention as an obvious matter to those skilled in the art, the detailed description thereof will be omitted.

Terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, the terms "include" or "have" are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded.

Hereinafter, an embodiment according to the present invention will be described in detail with reference to the accompanying drawings, and in the description with reference to the accompanying drawings, the same or corresponding components are assigned the same reference numerals and overlapping descriptions thereof are omitted. do it with

1 is a configuration diagram of an odorant prediction and monitoring system according to an embodiment, and FIG. 2 is an exemplary view of national reference standard data including odorant concentrations according to an embodiment.

Referring to FIG. 1, an odorant prediction and monitoring system (100, hereinafter referred to as 'system') includes a national reference standard data providing device 110, a linked data providing device 120, a measurement data collection device 130, and an odor A material prediction and monitoring device 140 may be included.

Water quality data below includes turbidity, pH (hydrogen ion concentration), DO (dissolved oxygen), BOD (biochemical oxygen demand), COD (chemical oxygen demand), SS (suspended solids) , suspended solids), TN (total nitorgen, total nitrogen), TP (total phosphorus, total phosphorus), TOC (total organic carbon), water temperature, phycocyanin, Chl-a (chlorophyll a, chlorophyll) It can be understood as values for -a) and flow rate, but is not limited thereto, and may further include values for odorants-geosmin and 2-MIB (2-methylisoborneol; 2-methylisoborneol). can Water quality data may be values already measured in the past or values currently measured.

The national reference standard data providing device 110 may provide the national reference standard data to the odorant prediction and monitoring device 140 . While the linked water quality data and the measured water quality data are collected in real time, the national reference standard data may be collected in advance and recorded in a database (not shown) linked to the odorant prediction and monitoring device 140 . National reference standard data include data on historical odorant concentrations, where odorants may include geosmin and 2-MIB. Therefore, national reference standard data may include concentration values for geosmin and 2-MIB.

Referring to FIG. 2, an example of national reference standard data may be shown. A (national) reference standard is a data prepared to be widely and continuously used or repeatedly used in all fields of the national society by scientifically analyzing, evaluating and accrediting the accuracy and reliability of measurement data and information (Article 3, Article 7 of the Framework Act on National Standards) ) - For example, semiconductor gas properties, high-speed properties, Korean body size, nuclear energy, environment, etc. -, national reference standard data may refer to data included in the reference standard. National reference standard data can be used by product developers at companies, researchers at schools and research institutes, and doctors at hospitals. As they can be referenced for similar tests and research and development, time and cost can be drastically reduced by preventing duplicate experiments. If national reference standard data are used in the early stage of research and product development, the leverage effect can be very large.

The national reference standard data shown in this figure may be about 'odorous substance concentration in raw water of a wide-area water purification plant' in the field of environmental water quality by the National Reference Standards Center. Specifically, this national reference standard data can indicate the concentration (μg/L) of odorous substances (geosmin, 2-MIB) in raw water from 13 water purification plants nationwide in August 2020. As an example, the result of analyzing a sample collected by analyst K-water at the Seongnam Water Purification Plant (San 88-5 Sasong-dong, Sujeong-gu, Seongnam-si, Gyeonggi-do) on 2020.08.27 is 'water temperature 23.1℃ pH 7.7, geosmin not detected, 2-MIB not detected' can include As another example, the result of analyzing a sample taken on August 27, 2020 by analyst K-water at the Goyang Water Purification Plant (300 Sanhwang-dong, Ilsandong-gu, Goyang-si, Gyeonggi-do) may include 'water temperature 22.2℃ pH 7.72, geosmin not detected, 2-MIB not detected'. can In this way, the values measured by a specific entity in each region can be recognized as national reference standard data after undergoing accuracy and reliability evaluation.

Returning to FIG. 1, the value included in the national reference standard data is initially a value measured at a specific point in the past by a specific subject, but becomes a standard value after being recognized for its accuracy and reliability through evaluation. The odorant prediction and monitoring device 140 may use the geosmin standard value and the 2-MIB standard value included in the national reference standard data to learn the prediction model.

The linked data providing device 120 may provide the odorous substance predicting and monitoring device 140 with water quality data measured in the past or currently measured by an institution. Such water quality data provided by the linked data providing device 120 may be referred to as linked water quality data. Alternatively, the linked water quality data may include past or present water quality data as water quality data possessed by an institution. Here, one institution may be an environmental facility such as a sewage treatment plant or water treatment plant. The environmental facility may provide linked water quality data to the odorant prediction and monitoring device 140 through the linked data providing device 120 .

The measurement data collection device 130 may provide the odorant prediction and monitoring device 140 with the water quality data collected through the measuring instrument 131 . The instrument 131 may include various sensors and meters. Such water quality data collected by the meter 131 and provided to the odorant prediction and monitoring device 140 may be referred to as measured water quality data. The meter 131 includes various sensors to measure turbidity, pH (hydrogen ion concentration), DO (dissolved oxygen), BOD (biochemical oxygen demand), COD (chemical oxygen demand) , SS (suspended solids), TN (total nitorgen), TP (total phosphorus), TOC (total organic carbon), water temperature, Chl-a (chlorophyll a, chlorophyll-a) ) and flow can be measured.

Here, the values of TN, TP, and TOC sent from the measurement data collection device 130 to the odorant predicting and monitoring device 140 are not only used for learning in the odorant predicting and monitoring device 140, but also used as national reference standard data. Values for TN, TP, and TOC designated as national reference standard data may be stored in the national reference standard data providing device 110.

In addition, the measurement data collection device 130 may further include a remote terminal unit (RTU) 132. The remote terminal 132 may collect data from a remote location, convert the data into a transmittable format, and transmit the data to an external device. Here, the remote terminal 132 may convert the measured water quality data collected by the meter 131 into a transmittable format and transmit the result to the odorant predicting and monitoring device 140 .

Meanwhile, the odorant predicting and monitoring device 140 may learn a predictive model through machine learning and predict odorant concentration based on the learned predictive model. Specifically, the odorant prediction and monitoring device 140 receives national reference standard data, linked water quality data, and measured water quality data, and predicts odorant concentrations for the input water quality data using a prediction model learned through machine learning. there is. For example, the odorant prediction and monitoring device 140 inputs at least one of phycocyanin, Chl-a (chlorophyll a, chlorophyll-a), water temperature, and pH (hydrogen ion concentration) to the learned prediction model. In response thereto, at least one of a geosmin prediction value and a 2-MIB prediction value may be output.

The odorant predicting and monitoring device 140 may provide the values of TN, TP, and TOC to the national reference standard data providing device 110 as national reference standard data after using the values of TN, TP, and TOC for learning the prediction model and predicting the odorant concentration. there is.

In addition, the user devices 150-1 and 150-2 may receive the analysis result from the smelly substance predicting and monitoring device 140 and output the result to the user. Here, the analysis result is an odorant concentration prediction value, and the odorant concentration prediction value may be a value output by a learned prediction model. The user devices 150-1 and 150-2 may output odorant concentration prediction values to the user in accordance with various user interfaces. The first user device 150-1 may output the odorant concentration predicted value through a web page, and the second user device 150-2 may output the odorant concentration predicted value through an application. Alternatively, the odorant concentration prediction value may be provided in the form of an open API, and the user devices 150-1 and 150-2 may output odorant concentration prediction values in conjunction therewith.

3 is a block diagram of an apparatus for predicting and monitoring odorous substances according to an embodiment.

Referring to FIG. 3 , the odorant prediction and monitoring device 140 (hereinafter referred to as 'device') includes a data collection unit 310, a data pre-processing unit 320, a prediction unit 330, a communication unit 340, and an analysis unit. (350). Here, the prediction unit 330 may further include a learning unit 331 and a prediction model 332 . The apparatus 140 may learn a predictive model through machine learning and predict odorant concentration based on the learned predictive model.

The data collection unit 310 may collect water quality data for learning of the prediction model 332 . Here, the water quality data may include national reference standard data, linked water quality data, and measured water quality data. Also, the collected water quality data may be current or past water quality data.

The collected water quality data can be divided into training data for training the predictive model 332 and verification data used to determine a learning stop point in the learning process. The predictive model 332 may learn training data and complete learning through validation data at a certain point in time—for example, before overfitting occurs. In addition, the performance of the predictive model 332 may be finally evaluated through test data different from training data and verification data.

National reference standard data may include odorant national reference standard data and water quality national reference standard data. The odorous substance national reference standard data may be data on a substance causing odor, and the water quality national reference standard data may be data on a scale representing the cleanliness of water.

Linked water quality data, measured water quality data, and national reference standard data for water quality include turbidity, pH (hydrogen ion concentration), DO (dissolved oxygen), BOD (biochemical oxygen demand), COD (chemical oxygen demand, chemical oxygen demand), SS (suspended solids), TN (total nitorgen), TP (total phosphorus), TOC (total organic carbon), water temperature, phycocyanin , Chl-a (chlorophyll a, chlorophyll-a) and values for flux.

In addition, linked water quality data, measured water quality data, and national reference standard data for odorants may also include values for geosmin and 2-MIB, which are odorants.

Here, phycocyanin measurement values, geosmin measurement values, and 2-MIB measurement values included in the measured water quality data, and geosmin standard values and 2-MIB standard values included in the national reference standard data can be used as training data and test data. there is.

For example, the learning unit 331 inputs the phycocyanin measurement value of the measured water quality data to the learned prediction model, and the geosmin prediction value output at this time is the measured water quality at the same time and place as the phycocyanin measurement value. It can be judged whether it matches the geosmin measurement value included in the data. If the results match, the learning unit 331 may determine that the performance of the predictive model is stable. Alternatively, the learning unit 331 may determine whether the geosmin predicted value matches a geosmin standard value included in national reference standard data at the same time and in the same place. If the results match, the learning unit 331 may determine that the performance of the predictive model is stable.

In addition, the data collection unit 310 may collect linked water quality data from the linked data providing device 120 operated by an environmental facility - for example, a water treatment plant or a sewage treatment plant. For example, the data collection unit 310 collects meteorological data from the meteorological data open portal (data.kma.go.kr) and water quality data from the real-time water quality information system, the automatic water quality measurement network (water.nier.go.kr). can be collected Alternatively, the data collection unit 310 may periodically collect linked water quality data in conjunction with an open API provided by the linked data providing device 120 .

The data pre-processing unit 320 may pre-process the data collected by the data collection unit 310 so that it can be used for learning and prediction of the prediction model 332 . Preprocessing may include a process of extracting a specific element or a transformation process such as vectorization or normalization.

The prediction unit 330 may train the prediction model 332 with respect to the odorant concentration. When learning is completed, the prediction unit 330 may predict the odorant concentration through the prediction model 332 . For example, the prediction unit 330 may receive at least one of phycocyanin, Chl-a, water temperature, and pH, and output at least one of a geosmin predicted value and a 2-MIB predicted value as a prediction result.

In the case of learning, the learning unit 331 of the prediction unit 330 may learn the prediction model 332 based on machine learning. The learning unit 331 may use national reference standard data, linked water quality data, and measured water quality data to learn the prediction model 332 .

The learning unit 331 may define the characteristics of the predictive model according to an algorithm used for machine learning. For example, the learning unit 331 may define the prediction model 332 as a classification model. To this end, the learning unit 331 may apply a decision tree (DT) algorithm, a support vector machine (SVM) algorithm, or a multi layer perceptron (MLP) algorithm to the prediction model 332 . Alternatively, the learning unit 331 may define the prediction model 332 as a regression model or a neural network model. To this end, the learning unit 331 may apply an artificial neural network (ANN) algorithm or a recurrent neural network (RNN) algorithm to the prediction model 332 .

In addition, the learning unit 331 may optimize the learning of the predictive model 332 by receiving feedback on the learning result. For example, the learning unit 331 may train the predictive model 332 by selecting an optimal algorithm from among a plurality of algorithms according to a learning result. Alternatively, the learning unit 331 may change variables of an algorithm applied to the predictive model 332 - for example, weights - according to a learning result.

In addition, the prediction unit 330 may transmit the odorant concentration value output by the prediction model 332 to the communication unit 340 . The odorant concentration value is converted so that the user can recognize it, and the communication unit 340 may transmit the odorant concentration value to the user devices 150-1 and 150-2. Alternatively, the communication unit 340 may transmit the measured water quality data collected by the measurement data collection device 130 - for example, the measured value of phycocyanin - to the user devices 150-1 and 150-2.

The analysis unit 350 may receive preprocessed linked water quality data, measured water quality data, and national reference standard data, and extract water quality factors from them. Water quality factors may include items included in linked water quality data, measured water quality data, or national reference standard data. For example, the water quality factor may include at least one of phycocyanin, chlorophyll-A, water temperature, pH, TN measurement value, TP measurement value, and TOC measurement value of the measured water quality data.

The analyzer 350 may analyze the correlation between water quality factors (step S605). The analyzer 350 may analyze the correlation between the water quality factor and the odorous substance - Geosmin or 2-MIB - using a statistical technique.

The learning unit 331 may learn the predictive model using the water quality factor. The learning unit 331 may set a water quality factor having a high correlation with the odorant as an input variable, and may train the predictive model with data for the input variable. For example, the learning unit 331 may set chlorophyll-A, algae population, and pH, which have a high correlation with 2-MIB, as input variables of the prediction model. The learning unit 331 may standardize learning data such as chlorophyll-A, algae population, and pH to unify numerical ranges, and apply the standardized learning data to a predictive model.

The learning unit 331 generates a plurality of candidate prediction models in various forms with a plurality of machine learning algorithms, such as long short-rerm memory (LSTM), bidirectional LSTM (LSTM), and gated recurrent units (GRU), for example, and , each can be trained in the same way, and a candidate prediction model with the highest accuracy can be determined as the prediction model.

Conventionally, the analysis of odorous substances not only requires complicated pre-processing and expensive analysis equipment (GC/MS, ELISA), but also makes it difficult to know measurement results in real time because measurement is discontinuous. However, the apparatus 140 for predicting and monitoring odorous substances according to an exemplary embodiment automatically measures water quality data-turbidity, pH, DO, BOD, COD, SS, TN, TP, TOC, water temperature, phycocyanin, and Chl-a. And the concentration of the odorant can be predicted through the flow rate.

In addition, according to an embodiment, by using the national reference standard data to predict the concentration of odorous substances in the future, the usability of the national reference standard data can be increased, and the collected measured water quality data can also be output in real time.

4 is a flowchart illustrating an operation of an apparatus for predicting and monitoring odorants according to an exemplary embodiment.

Referring to FIG. 4 , the device 140 may collect measured water quality data, linked water quality data, and national reference standard data (step S410).

The device 140 may learn the prediction model as an algorithm using the measured water quality data, the linked water quality data, and the national reference standard data (step S420).

The device 140 may determine whether learning of the predictive model is completed (step S430). If learning is not completed, the device 140 may continue learning (NO in step S430). When learning is complete, the device 140 may perform a prediction step (YES in step S430). The device 140 may predict the odorant concentration through the predictive model (step S440). The device 140 may output at least one of a geosmin prediction value and a 2-MIB prediction value as a result of the prediction.

5 is a diagram for explaining characteristics of a substance that causes an odorous substance according to an exemplary embodiment.

Referring to FIG. 5 , characteristics of a substance causing an odorous substance may be illustrated. According to an embodiment, the odorants predicted by the device are geosmin and 2-MIB, but other candidate odorants are included in the description in this specification.

Odor substances related to waterworks can be largely classified into natural and artificial, but naturally occurring substances due to biological action may be common. Odor substances caused by natural causes are mostly found to be caused by algae, and in particular, blue green algae can cause substances such as geosmin and 2-MIB.

Algae can give off odorants in the form of intracellular storage, extracorporeal secretion, and release of degradation products. When the algae die, the body of the algae decomposes and acts as a nutrient for actinomycetes, or substances contained in the algal cells can be separated and dissolved by other bacterial actions. Actinomycetes are filamentous unicellular bacteria that live mainly in soil and can contaminate raw water by leaking into rainwater.

Bacteria other than actinomycetes also release substances such as geosmin and 2-MIB, and fungi can also release geosmin. Major odorants generated by blue-green algae identified so far include geosmin, 2-MIB, IPMP, and IBMP. Among these, geosmin and 2-MIB are chemically stable and resistant to oxidation, so they are not oxidized well by conventional water treatment methods, and can be produced in large quantities in summer when algal activity is active. For Geosmin and 2-MIB, the monitoring standard for Geosmin in purified water is suggested as 20ng/L in the Ministry of Environment's Drinking Water Quality Standards (2017.06), but people with a normal sense of smell can detect it at 4-10ng/L.

6 is a flowchart illustrating an operation of learning a predictive model through analysis of correlations of water quality factors according to an embodiment, and FIG. 7 is a first exemplary diagram of analysis results of correlations of water quality factors according to an embodiment. 8 is a second exemplary diagram of analysis results of correlations between water quality factors according to an embodiment.

Referring to FIG. 6 , the device according to an embodiment may analyze the correlation of water quality factors and learn a predictive model using a water quality factor having the highest correlation with Geosmin or 2-MIB as an input variable.

The data pre-processing unit of the device may pre-process linked water quality data, measured water quality data, and national reference standard data (step S601).

The analysis unit of the apparatus may receive preprocessed linked water quality data, measured water quality data, and national reference standard data, and extract water quality factors from them (step S603). Water quality factors may include items included in linked water quality data, measured water quality data, or national reference standard data. For example, the water quality factor may include at least one of phycocyanin, chlorophyll-A, water temperature, pH, TN measurement value, TP measurement value, and TOC measurement value of the measured water quality data.

The analysis unit of the device measures water quality factors—for example, at least one of phycocyanin, chlorophyll-A, water temperature, pH, TN measurement value, TP measurement value, and TOC measurement value—from linked water quality data, measured water quality data, or national reference standard data. If is extracted, it is possible to analyze the correlation between water quality factors (step S605). The analysis unit of the device may analyze the correlation between the water quality factor and the odorous substance - Geosmin or 2-MIB - based on a statistical technique.

Referring to FIG. 7 , a first example of a correlation analysis result between algae population, chlorophyll-A, pH, water temperature, TOC, and 2-MIB may be shown. The analysis unit of the device can be derived by analyzing the correlation between the algae population, chlorophyll-A, pH, water temperature, and TOC, including 2-MIB, which is an odorous substance. In this figure, it can be seen that the correlation between chlorophyll-A and the algae population is high, and the correlation between the algae population and pH is high.

Referring to FIG. 8, a second example of a correlation analysis result between chlorophyll-A, pH, and geosmin may be shown. The analysis unit of the device may be derived by analyzing the correlation between chlorophyll-A, including geosmin, which is an odorous substance, and pH. In this figure, it can be seen that geosmin does not have a high correlation with other water quality factors.

Returning to FIG. 6 again, the learning unit of the device may learn the prediction model with the water quality factor (step S607). The learning unit may set a water quality factor having a high correlation with the odorant as an input variable, and may train data on the input variable to the predictive model. For example, the learning unit can set chlorophyll-A, algae population, and pH, which showed a high correlation with 2-MIB, as input variables of the prediction model. The learning unit may standardize learning data such as chlorophyll-A, algae population, and pH to unify numerical ranges, and apply the standardized learning data to a prediction model.

Here, the learning unit generates a plurality of candidate prediction models of various types with a plurality of machine learning algorithms, such as long short-rerm memory (LSTM), bidirectional LSTM (LSTM), and gated recurrent units (GRU), respectively. The candidate prediction model with the highest accuracy can be determined as the prediction model. The accuracy of the model can be evaluated by mean absolute error (MAE) and root mean squared error (RMSE). For example, the learning unit may select a candidate prediction model based on bidirectional LSTM as a final prediction model.

Aspects of the subject matter described herein may be described in the context of computer-executable instructions, such as program modules, being executed on a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., which perform particular tasks or particular abstract data types. Aspects of the subject matter described herein may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

Alternatively or additionally, the functionality described herein may be performed, at least in part, by one or more hardware logic components. By way of example and not limitation, example types of hardware logic components that may be used include field-programmable gate arrays (FPGAs), program-specific integrated circuits (ASICs), application Includes application-specific standard products (ASSPs), system-on-a-chip systems (SOCs), complex programmable logic devices (CPLDs), etc. do.

Meanwhile, the disclosed embodiments may be implemented in the form of a recording medium storing programs and/or commands executable by a computer. Instructions may be stored in the form of program codes, and when executed by a processor, create program modules to perform operations of the disclosed embodiments. The recording medium may be implemented as a computer-readable recording medium.

Computer-readable recording media include all types of recording media in which instructions that can be decoded by a computer are stored. For example, there may be read only memory (ROM), random access memory (RAM), a magnetic tape, a magnetic disk, a flash memory, an optical data storage device, and the like.

The protection scope of the present invention is not limited to the description and expression of the embodiments explicitly described above. In addition, it is added once again that the protection scope of the present invention cannot be limited due to obvious changes or substitutions in the technical field to which the present invention belongs.

Claims (4)

As water quality data measured in the past or currently, linked water quality data provided from environmental facilities is collected in real time, and is different from the linked water quality data and is currently measured. Phycocyanin, chlorophyll-A (Chl -a; chlorophyll a), water temperature and pH (hydrogen ion concentration) are collected in real time, and geosmin and 2-MIB (2-Methylisoborneol; 2-methylisoborneol) are collected in real time. a data collection unit that collects national reference standard data on odorant concentrations in the past, including all), and stores the national reference standard data in a database;
Receiving data from the data collection unit, using a prediction model learned through machine learning, at least one of phycocyanin, chlorophyll-A (Chl-a; chlorophyll a), water temperature and pH (hydrogen ion concentration) A prediction unit for predicting the concentration of odorous substances including geosmin and 2-MIB for one, and
An analysis unit extracting water quality factors from the linked water quality data, the measured water quality data, and the national reference standard data, and analyzing a correlation between the water quality factors,
The prediction unit may include a prediction model including an artificial neural network and predicting the concentration of the odorant through the artificial neural network; And a learning unit for learning the prediction model using the linked water quality data, the measured water quality data, and the national reference standard data,
The linked water quality data, the measured water quality data, and the national reference standard data include values for geosmin and 2-MIB,
The learning unit learns the prediction model using Geosmin measurement values and 2-MIB measurement values included in the measured water quality data and Geosmin standard values and 2-MIB standard values included in the national reference standard data, and Geosmin and a water quality factor having a correlation with 2-MIB set as an input variable to train the prediction model to output concentrations of odorants including geosmin and 2-MIB. delete delete A device for predicting and monitoring odorous substances according to claim 1;
a linked data providing device for providing the linked water quality data to the odorant prediction and monitoring device;
a measurement data collection device including a measuring instrument for measuring the measured water quality data and a remote terminal unit (RTU) for transmitting the measured water quality data to the odorant predicting and monitoring device; and
Odor substance prediction and monitoring system including a national reference standard data providing device for providing the national reference standard data to the odor substance prediction and monitoring device

Images