Study on water quality analysis and early-warning technology based on rough set and evidence theory

Abstract

With the development and progress of society,the problem of water environment pollution has become one of the topics which are cared and taken seriously by people.Higher requirements are brought forward to come up with the frequent and various water pollution events.The current used monitoring systems are difficult to meet the higher requirements referred before.How to monitor water quality effectively and to analyze the concentration of contaminants in water qualitatively during or after water pollution events have become arduous tasks for a water quality monitoring system.As it is known to all,there are lots of monitoring parameters in water quality detection,monitoring and information management system,and the evolution principles and changing trends of water quality are difficult to be obtained.The accuracy of water quality analysis and early-warning is hard to guarantee.Aiming to treat these problems,a new method for a water monitoring system is needed to study on.In this study,the rough set and Dempster-Shafer(D-S) evidence theories were researched,and the water quality analysis and early-warning algorithm based on rough set and D-S evidence theories were presented.The original monitoring water quality data was given(Table 1).The discretization in this paper was based on GB3838-2002,and the result of discretization was also given(Table 3).After discretization,the incomplete data and the conflicting data were deleted(Table 3).Moreover,in order to solve the problem of redundancy parameters,the rough set was used.After these steps,a minimal data list was acquired.Then,the data in the list was recovered to take the place of discrete value.Based on the function of basic probability assignment and the D-S rule of combination,the probability was assigned and the data was fused(Table 4),and the result of data fusion could reflect the water quality or the concentration of contaminants in the water qualitatively.Refer to the national standard,the value of CODMn and the water quality in different time points were acquired(Fig.4,5).The analysis results which were based on multistage function fitting and rough set theory were also given so as to compare with the result based on rough set and D-S evidence theories(Fig.2-3).Through the results of analysis,the method based on rough set and D-S evidence theories was proved to be feasible and well.Comparing to the result based on rough set and D-S evidence theories,the result based on multistage function fitting had larger deviation with the passage of time,and the result based on rough set theory had the case which could not be inferred by historical data.In conclusion,the method based on rough set and D-S evidence theories can reduce the influence of historical water data losses,conflicts and redundancy,and overcome the data shortage in practical systems.The reduction of water monitoring parameters which is based on rough set theory also reduces the calculation cost dramatically.