A Development of the Multi-item Water Quality Monitoring System Using Markov Process

Transcription

1 , pp A Development of the Multi-item Water Quality Monitoring System Using Markov Process ByungMun, Lee, Un-Gu Kang, 1 Dept. of Computer Science, Gachon University, Gyeonggi-do, Korea {bmlee, ugkang}@gachon.ac.kr Abstract. The conventional water quality testing method measured the pollution level after pollutants are discharged. If the water quality deteriorates on a constant cycle by nature, such cyclic pollution can be prevented by monitoring the water quality through the prediction of such cycle. To build a water quality prediction system using cyclic patterns, the author applied Markov Model in this study. In other words, since the probability values are dependent of the immediate past, an analysis of them will enable it to prevent water pollution. This approach was applied to both the pollution prediction by the manipulated pollution discharge and the pollution prediction on a natural cycle, which is expected to increase the accuracy in pollution prediction using Markov Process. Keywords: water pollution, water pollution patterns, water pollution prediction, Markov Process 1 Introduction With the increasing importance of water, it is essential water quality is monitored constantly and real-time information is provided, and the water quality prediction system [1] is developed to prevent the water pollution. It will be too late to take effective measures after the water pollution occurs, or the water pollution might have progressed considerably. Moreover, it is practically impossible to monitor the quality of every river of many rivers in Korea. However if the water pollution is predictable, it will prevent the water pollution more readily. However monitoring every river is easily said and done. It is because the time of malicious pollutant discharge is not pre-determined, or it is difficult to predict the time of discharge accurately and take timely measures even if the pollution is cyclic due to the natural phenomenon. Looking at the water pollution patterns over the last years, there are some patterns. In particular, the pollutant discharge got worse during the rainy season. More specifically, it occurs on specific summer days. Table 1 demonstrates the cycle of river pollution and the annual patterns of river pollution in major rivers of Korea [2]. 1 This work was supported by Business for Cooperative R&D between Industry, Academy, and Research Institute funded Korea Small and Medium Business Administration in ISSN: ASTL Copyright 2015 SERSC

2 As shown in the table 1, intensive pollution control is required during summer, yet some items (i.e. hydrogen ion concentration, dissolved oxygen, suspended solids and so on) require constant water quality control throughout the year, indicating continued control is required for domestic rivers [3]. Table 1. Monthly pollution level of Nak-dong River Month Temperature ( ) Hydrogen Ion Concentratio n (Ph) Dissolved Biological Demand Chemical Demand Suspended Matters Colon Bacillus Group (MPL/100) Total Jan Feb Mar Apr May Jun Jul ,300.0 Aug ,600.0 Sep Act Nov Dec Under such circumstances, it is almost impossible to mobilize many staff for monitoring throughout the year. Thus there still is a need for a system which predicts the probability of water pollution using an automated algorithm applied with patterns, the probability theory and statistical techniques [4]. To solve the problems set forth above, in this study the author, therefore, developed a water quality prediction algorithm using Markov Process with the aim to build a Smart Prediction System based on the forgoing algorithm. 2 Related Research The development of the suggested water pollution prediction algorithm is based on Markov Process Model. Markov Process Model explains the water pollution process using Markov Process, assuming that the future water pollution is always dependent of the present state of water. Fig. 1. Markov Process Model 184 Copyright 2015 SERSC

3 To predict the water quality, the constraints of Markov Process need to be satisfied. As shown in Fig 1, the probabilistic constraints can be expressed as shown in the formulas below because the state transition probability is determined by the present state only. P [ q t = j q t 1 = i, q t 2 = k, ] = P [ q t = j q t 1 = i ] (1) Second, the state transition probability is independent of time. a ij = P [ q t = j q t 1 = i ] (2) Finally, the event corresponds to each state. The listed in Table 2 were obtained by applying the properties of Markov Process to actual water pollution prediction. It was assumed there are only 3-state water pollution levels for the convenience s sake. Table 2. Examples of the water quality prediction using Markov Process Model Pollution Level Pollution Level Pattern If the algorithm is applied State 1: Severe Pollution, State 2 : Moderate Pollution, State 3 : Normal Calculate the probability of pollution having the pattern: Normal- Normal-Normal-Severe Pollution Severe Pollution- Normal Moderate Pollution P(O M)=P [ 3, 3, 3, 1, 1, 3, 2, 3 M ] =P[3] * P[3 3] 2 * P[1 3] * P[1 1] * P[3 1] * P[2 3] * P[3 2] =1.536 * 10-4 If the examples in the above Table 2 are implemented as Markov Model, the state transition probabilities can be calculated as shown in Fig 2. In this case, the state transition probability can be determined, which enables it to predict the following state and take measures on the basis of the present water pollution state. Fig. 2. Pollution level probability and Probability of having an effect on each state 3 System Design and Implementation The system was configured in such a way that the suggested algorithm of the monitoring system predicts the water pollution level on the basis of the information acquired through the sensors installed on the sites. Fig 3(a) is Entire System Copyright 2015 SERSC 185

4 Configuration and Fig 3(b) is Monitoring System implemented by applying the suggested algorithm. The proposed system was implemented to support one-stop monitoring, enabling it to monitor on a real-time basis and respond to disastrous situations systematically and immediately, which has overcome the limitations of the conventional system. (a) Entire System Configuration (b) Monitoring System Fig. 3. Suggested system based on Markov Process 4 Evaluation and Result The suggested algorithm of the suggested system was developed using Markov Process Model. Thus it is required to compare before the application of the suggested algorithm and after the application of the suggested algorithm. The item to be compared was the accuracy in the water quality prediction, which is expected to contribute to improving the accuracy of the suggested system compared to the conventional system such as a real-time water quality abnormality determination system using multivariate statistical techniques [4] or a water quality monitoring network using Genetic Algorithms [5]. The benefits of the suggested system include: 1) System based automated real-time monitoring, enabling it to save time required for testing water quality, and 2) Integrative management of target water quality/discharge measurement project for total pollutant management [6][7]. In addition to the above benefits, the suggested system enables it to use the basic data of key water quality policies (e.g. an analysis of water quality and the establishment of base flow) [8] and analyze the quality of rivers though an analysis of point-specific spatiotemporal properties, using the reference numerical/hydrological and water quality data though an analysis of point-specific spatiotemporal properties. References 1. A Study on the Development of Internet-based Remote Water Quality Monitoring and Prediction System, Pohang University of Science and Technology, Yonsei University, the Ministry of Environment Report, (2004) 2. The National Statics Portal, Copyright 2015 SERSC

5 3. Installation of groundwater quality monitoring network and water contamination measurement plan, the Ministry of Environment Report, (2014) 4. Tae-Young Heo.Hang-Bae Jeon.Sang-Min Park.Young-Joo Lee.: Development of Real-Time Water Quality Abnormality Warning System for Using Multivariate Statistical Method. J. Korean Soc. Environ. Eng. Chungbuk National University, vol 37 (3), pp , (2015) 5. Su Young Park.: A Design of a Water Quality Monitoring Network in the Nakdong River Using a Genetic Algorithm. Graduate School of Ewha Womans University, Master's thesis (2003) 6. Water Pollution-Measurement_ KBI Index, 7. The level of water quality management measures, 8. Water Quality Analysis Methodology (Seawater, Fresh Water), Water Quality Lab of the Department of Environmental Engineering, Pukyong National University. Copyright 2015 SERSC 187