CN105320993A - Water source pollution detection method based on evidence theory - Google Patents

Abstract

The invention relates to a water source pollution detection method based on an evidence theory, and the method uses an information processing technology to detect the pollution degree of water. A fuzzy number model of the water resource water quality index is generated according to existing data, the similarity between a sampling value in a water resource distribution system and the fuzzy number model of the existing index is obtained according to a similarity measurement method of the fuzzy number, basic possibility appointment is generated on the basis of the similarity, a classic fusion method is used to fuse the basic possibility appointment, a main focal element of a fusion result is determined, the pollution degree of the system is obtained according to the correspondence between the pollution level and function elements of the basic possibility appointment, and water resource pollution is detected. The detection method has the advantages of higher versatility, low computational complexity and high instantaneity, and can be conveniently used in water resource pollution detection.

Description

Based on the water resource pollution detection method of evidence theory

Technical field

What the present invention relates to is a kind of method of technical field of information processing, specifically a kind of water resource pollution detection method based on evidence theory.

Background technology

In recent years, rivers and lakes water pollution issue and potential risk can not be ignored, deeply by social concerns.Support water quality detection is monitored, the modern means of science and technology such as computing machine, network communication and infotech, the water quality detection monitoring of Erecting and improving and information management system, timely grasp water quality condition, analyses and prediction Regular of Water Quality Variation and trend, to furnish a forecast Informational support for tackling water quality deterioration or alert in advance, having become one of current technical matters needing solution badly.Due to affect water qualitative factor and index numerous, and whole observation process relate to many uncertain factors, therefore adopt development in recent years to get up robotization comprehensive treatment technique---information fusion technology is very important to carry out process.This technology, by processing the information from different index or sensor, draws and comprehensively analyzes assessment result.It is a kind of science polymerization to observing and measuring.For monitoring and the inspection of a system or a process, often need polytype measurement or observe to describe its integrality.Data fusion is very useful for having repeatable objective polymerization.Many infrastructure project problems, such as, in the state estimation of assets, production run, quality monitoring and water quality monitoring all need multiple performance index to carry out definition status.In addition, for a prediction reliably, usually need to be polymerized in time or observation spatially (or multiple) performance index.

In some cases, different data sets (such as, dissimilar sensor and the measurement of detector, every water-quality guideline) provides each side information of system or process, plays complementary effect.Therefore, our object predicts accurately state to collect more information.In addition, if the information collected by multiple data set is used to process same problem, may redundancy be produced, but but turn improve reliability by certain confirmation each other observed/measure.In the state estimation and water quality monitoring of assets, the information of data set is supplemented with redundancy is the basis that data fusion is applied.

Drinking water dispenser system is owing to often suffering bad reaction and event and the tap water causing consumer to eat is become dangerous by high-quality, had the water of peculiar smell.Thus, for guaranteeing that consumer can obtain high-quality potable water, in distribution system, the periodic monitoring of raw water quality and the monitoring of processing procedure are that drinking water quality manages vital link.

In order to the water quality in monitor allocation system, the physics obtained in routine sampling, chemistry and Biological indicators can be analyzed through laboratory or portable set by some fields (American Public Health Association, water employer's organization of the U.S., Water Pollution Control Federation) usually.The existence of sensor technology make by on-line monitoring obtain performance index than by sample obtain more welcome.This technology constantly develops, and covers more eurypalynous water-quality guideline.Some common water-quality guideline in potable water distributes have turbidity, residual sanitizer, pH value, nitrate, phosphate, heterotrophic bacteria, organism etc.

The use of water creates a large amount of water quality datas by routine sampling, is used for controlling in water distribution system and maintaining the receptive phase of water quality.The technology that the information acquisition of different quality index uses is different (hand sampling, automatically sample and carry out lab analysis subsequently, on-line monitoring by automatic analyzer equipment) also.In actual applications, reasonable realization is on the comprehensive assessment of the various factors and index that affect water quality, and the overall treatment of the various uncertain factors that may occur in testing process also exists the problems such as versatility is not strong, computation complexity is high, poor real, uncertain information processing power are weak.

Summary of the invention

The present invention is directed to the current water quality pollution situation existed in water resource distribution system, provide a kind of water resource pollution detection method based on evidence theory, the roadmap that the method adopts Fuzzy Math Model to combine with evidence theory effectively can solve the uncertainty of Testing index and data, and in versatility, there is good improvement computation complexity and real-time aspect, can in the pollution detection conveniently for various water resource distribution system.

The present invention is achieved by the following technical solutions, the present invention includes following steps:

The first step, obtain the similarity of sampled value and existing water-quality guideline Fuzzy Math Model according to the method for measuring similarity of fuzzy number.

In this contamination detection method, first generate the Fuzzy Math Model of water-quality guideline according to the database in existing water resource distribution system, water-quality guideline comprises: residual chlorine (RC), haloform (THM), heterotrophic bacteria (HPC).Three principles should be followed: representativeness principle, comprehensive principle, operability principle when setting up data model.

Secondly, in water resource distribution system, the sampled value of water-quality guideline and water-quality guideline Fuzzy Math Model obtain the similarity between them according to the method for measuring similarity of fuzzy number.

Second step, generation basic probability assignment

2.1 classes of pollution and basic probability assignment function (BasicProbabilityAssignment, BPA) Jiao unit is corresponding: Jiao unit of basic probability assignment function is determined by the number of the class of pollution, and the class of pollution that this method adopts is five: very low (VL), low (L), medium (M), high (H), very high (VH).

2.2 obtain preliminary basic probability assignment according to the sampled value of water-quality guideline and the similarity of water-quality guideline Fuzzy Math Model, then based on the percentage contribution of each water-quality guideline (namely affecting water resource water qualitative factor), it is carried out to the division of weight, and weighted calculation obtains the basic probability assignment of each contamination index, obtain the BPA of this water resource distribution system thus.

The fusion method of the 3rd step basic probability assignment: when there being n index, combines the Dempster rule of combination of the basic probability assignment generated by following classics n-1 time.

$m\left(A\right)=\frac{\underset{B\∩C=A}{\mathrm{\Σ}}{m}_1\left(B\right)m_2\left(C\right)}{1-k}---\left(1\right)$

Wherein,

4th step, pollution level are evaluated: the BPA fusion results obtained in the 3rd step analyzed, find out the main burnt unit (Jiao Yuanzhong of basic probability assignment function occupies Jiao unit of clear superiority) of BPA after merging, by the corresponding relation of the burnt unit of the class of pollution and basic probability assignment function, draw the class of pollution in water resource distribution system, achieve the water resource pollution detection method based on evidence theory.

The invention has the beneficial effects as follows:

1) traditional based in the water quality detection method of evidence theory, usually need to adopt other technologies to carry out pre-service to revise evidence model to the sampled value of water quality parameter, to generate the evidence of available D-S rule of combination combination, and the evidence in some evidence theory being transformed by sampled value and obtain, its basic probability assignment is rule of thumb obtained by monitoring personnel or expert system, has very strong subjectivity; And the present invention is under the prerequisite providing certain water resource index sample data, by the corresponding relation of the class of pollution and Jiao unit, and the method for measuring similarity of fuzzy number, BPA function can be realized and automatically generate, greatly reduce workload, decrease the subjective impact of people.

2) application prerequisite of the present invention is the qualitative or quantitative database sample data of water pollution system index, and this is very easy to realize in water resource distribution system, has good versatility;

3) synthesize BPA by the class of pollution and Jiao unit corresponding relation, and merged it by weighting synthetic method, computation complexity is low, has good real-time, and increases the confidence level of court verdict.

4) between the fuzzy number proposed in the present invention, similarity calculating method has considered the factors that centroidal distance between fuzzy number, height and girth geometrically and area etc. affect fuzzy number similarity, have effectively achieved the valid metric of similarity between fuzzy number.

Accompanying drawing explanation

Fig. 1 is that water pollutions detection method runs schematic diagram.

Fig. 2 is the Fuzzy Math Model of water-quality guideline, wherein: the Fuzzy Math Model that (a) is residual chlorine; B Fuzzy Math Model that () is haloform; C Fuzzy Math Model that () is heterotrophic bacteria.

Fig. 3 is the pollution level overview that in different residual chlorine concentrations situation, water-quality guideline represents, wherein: (a) is situation when residual chlorine concentrations is 0mg/l; B () is situation when residual chlorine concentrations is 0.2mg/l; C () is situation when residual chlorine concentrations is 0.5mg/l; D () is situation when residual chlorine concentrations is 4mg/l.

Embodiment

Elaborate to embodiments of the invention below, the present embodiment is implemented under premised on technical solution of the present invention, give detailed embodiment and concrete operating process, but range of application of the present invention is not limited to following embodiment.The present embodiment is the water pollution testing process citing in water resource distribution system.The detailed process of this enforcement comprises following step:

(1) division of the class of pollution

The class of pollution adopted is five: very low (VL), low (L), medium (M), high (H), very high (VH).Jiao unit number of basic probability assignment function is corresponding with the class of pollution.

(2) Fuzzy Math Model of water-quality guideline

First generate the Fuzzy Math Model of water-quality guideline according to the database in existing water resource distribution system, water-quality guideline comprises: residual chlorine (RC), haloform (THM), heterotrophic bacteria (HPC).Its model as shown in Figure 2.

(3) similarity of sampled value and existing water-quality guideline Fuzzy Math Model generates basic probability assignment

Herein, sampled value (clear number) is regarded as a certain moduli paste number and is processed by us, and such as: the sampled value of residual chlorine concentrations is 0.09, then corresponding fuzzy number is (0.09,0.09,0.09,0.09; 1.0).Then obtain the similarity of sampled value and existing water-quality guideline Fuzzy Math Model according to the method for measuring similarity of fuzzy number, the sampled value being obtained water-quality guideline by the result detected water-quality constituents in water resource distribution system is as follows:

Residual chlorine=0.09mg/l; Heterotrophic bacteria=62/100ml; Haloform=118ppb.

The method for measuring similarity of the fuzzy number adopted herein is:

$S(A,B)=e^{-|x_A^{*}-x_B^{*}|}\×[1-|w_A-w_B\left|\right]\×\frac{\min (P\left(A\right),P\left(B\right))+\min (A\left(A\right),A\left(B\right))}{\max (P\left(A\right),P\left(B\right))+\max (A\left(A\right),A\left(B\right))}---\left(2\right)$

A and B is two Trapezoid Fuzzy Numbers, is designated as A=(a ₁, a ₂, a ₃, a ₄; w _a), B=(b ₁, b ₂, b ₃, b ₄; w _b), wherein 0≤a ₁≤ a ₂≤ a ₃≤ a ₄≤ 1,0≤b ₁≤ b ₂≤ b ₃≤ b ₄≤ 1,0≤w _a, w _b≤ 1.

The similarity that can be obtained sampled value and water-quality guideline Fuzzy Math Model by the method is as shown in table 1

The similarity of table 1 sampled value and water-quality guideline Fuzzy Math Model

According in D-S evidence theory, the basic probability assignment of all subsets and be 1 principle, be normalized the similarity in upper table, thus obtain basic probability assignment function as shown in table 2, wherein, normalization process adopts $x_i=x_i/{\mathrm{\Σ}}_{i=1}^n{x}_i,i=\mathrm{1,2},...,n$ Calculate.

The basic probability assignment of table 2 sampled value and water-quality guideline Fuzzy Math Model

According to the percentage contribution of three water-quality guideline (namely affecting water resource water qualitative factor), the division of its weight is as follows:

α _RC＝0.9；α _HPC＝0.5；α _THM＝0.8。

Weighted calculation formula is:

wherein α is weight coefficient, and n is Jiao unit number of basic probability assignment.It is as shown in table 3 that the basic probability assignment of three water-quality guideline calculates net result through weighted formula:

The result of basic probability assignment after weighting of table 3 sampled value and water-quality guideline Fuzzy Math Model

I.e. m (RC)=(0.1504,0.1433,0.1810,0.2579,0.2674);

m(HPC)＝(0.2445，0.2376，0.1971，0.1616，0.1592)；

M (THM)=(0.1482,0.1424,0.1737,0.2382,0.2975); Five states of five corresponding classes of pollution of burnt unit's difference.

(4) fusion method of BPA

Adopt classical Dempster rule of combination to merge the BPA generated, because water resource distribution system relates to three indexs, therefore merged 2 times, the result obtained:

m(WQ)＝(m(WQ) _VL，m(WQ) _L，m(WQ) _M，m(WQ) _H，m(WQ) _VH)

；

＝(0.1394，0.1240，0.1585，0.2540，0.3240)

(5) final assessment of pollution level

The BPA fusion results obtained in step (4) is analyzed, find out the main burnt unit (occupying Jiao unit of clear superiority in basic probability assignment function) of BPA after merging, by the corresponding relation of the burnt unit of the class of pollution and basic probability assignment function, the class of pollution of water resource distribution system can be obtained for very high.

This new regulation of permissible concentration about DBPs is being formulated in the U.S. and other drinking water supplies area.Sterilization decreases the pollution that infected by microbes brings, but but likely can bring cancer and other pollutions from DBPs (wherein, haloform is modal DBPs).In addition, many DBPs to be determined to public health the unknown are also had.Present society be faced with one known microbial contamination and have much more probabilistic DBPs pollute between difficult balance.Assessment between pollution this in potable water is accepted or rejected, must assess under same framework.

We are 0mg/l at residual chlorine concentrations respectively, 0.2mg/l, the balance between heterotroph number and haloform is analyzed when 0.5mg/l and 4mg/l, result as shown in Figure 3, wherein water-quality guideline (WQI) is used as the Substitute Indexes of pollution level, adopt RehanSadiq (RehanSadiq, ManuelJ.Rodriguez.Interpretingdrinkingwaterqualityinthed istributionsystemusingDempster-Shafertheoryofevidence.Ch emosphere59 (2005) 177-188.) linear function that proposes calculates, its computing formula is as follows:

$\mathrm{WQI}=\frac{u}{2^0}\left[\mathrm{bl}{\left(\mathrm{WQ}\right)}_{\mathrm{VH}}\right]+\frac{u}{2^1}\left[\mathrm{bl}{\left(\mathrm{WQ}\right)}_H\right]+\frac{u}{2^2}\left[\mathrm{bl}{\left(\mathrm{WQ}\right)}_M\right]+\frac{u}{2^3}\left[\mathrm{bl}{\left(\mathrm{WQ}\right)}_L\right]+\frac{u}{2^4}\left[\mathrm{bl}{\left(\mathrm{WQ}\right)}_{\mathrm{VL}}\right]---\left(3\right)$

Wherein, practical coefficient u ≈ 1.3.

As seen from Figure 3, when can't detect residual chlorine, when namely residual chlorine concentrations is 0, WQI approximately changes to 1 from 0.65.We can observe, even if also may have the higher class of pollution under heterotroph number and all low-down situation of haloform concentration, this is because for pre-preventing microbial contamination, a small amount of residual chlorine is necessary.But when residual chlorine concentrations is increased to 0.2mg/l, 0.5mg/l and 4mg/l, WQI approximately changes to 0.9 from 0.4, this is low much relative to the first situation.In addition, this three-dimensional character Risk profiles can be set up to predict whether arbitrary specific indexes reaches the accepted pollution level under specified criteria for various water-quality guideline.

Claims (5)

1. the water resource pollution detection method based on evidence theory, it is characterized in that, first the Fuzzy Math Model of water resource water-quality guideline is generated according to data with existing storehouse, the method for measuring similarity of fuzzy number is adopted to obtain the similarity of sampled value and existing index Fuzzy digital-to-analogue type in water resource distribution system, and basic probability assignment is generated based on this similarity, then according to classical fusion method, basic probability assignment is merged, finally determine the main burnt unit of fusion results, and the pollution level of this system is drawn by the corresponding relation of the burnt unit of the class of pollution and basic probability assignment function, realize the detection of water resource pollution.

2. the water resource pollution detection method based on evidence theory according to claim 1, is characterized in that, the Fuzzy Math Model of described water resource water-quality guideline refers to:

2.1 water resource water-quality guideline comprise: residual chlorine (RC), haloform (THM), heterotrophic bacteria (HPC);

2.2 according to the Fuzzy Math Model of the data genaration water resource water-quality guideline in existing Water resources data storehouse.

3. the water resource pollution detection method based on evidence theory according to claim 1, is characterized in that, in described water resource distribution system, the similarity of sampled value and existing index Fuzzy digital-to-analogue type obtains according to following steps:

The sampled value of 3.1 water-quality guideline: the sampled value obtaining residual chlorine (RC), haloform (THM) and heterotrophic bacteria (HPC) these three kinds of indexs by carrying out detection to water-quality constituents in water resource distribution system;

3.2 obtain the similarity of sampled value and existing water-quality guideline Fuzzy Math Model according to the method for measuring similarity of fuzzy number: at this, and the method for measuring similarity of the fuzzy number that we adopt is:

$S(A,B)=e^{-|x_A^{*}-x_B^{*}|}\×[1-|w_A-w_B\left|\right]\×\frac{\min (P\left(A\right),P\left(B\right))+\min (A\left(A\right),A\left(B\right))}{\max (P\left(A\right),P\left(B\right))+\max (A\left(A\right),A\left(B\right))}$

Wherein, A and B is two Trapezoid Fuzzy Numbers, is designated as A=(a ₁, a ₂, a ₃, a ₄; w _a), B=(b ₁, b ₂, b ₃, b ₄; w _b), 0≤a ₁≤ a ₂≤ a ₃≤ a ₄≤ 1,0≤b ₁≤ b ₂≤ b ₃≤ b ₄≤ 1,0≤w _a, w _b≤ 1.

4. the water resource pollution detection method based on evidence theory according to claim 1, it is characterized in that, described basic probability assignment obtains in the following manner:

Jiao unit number of 4.1 basic probability assignment functions is corresponding with the class of pollution, and the class of pollution that this method adopts is: very low (VL), low (L), medium (M), high (H), very high (VH) five;

4.2 weighting synthesis basic probability assignment functions: based on similarity, initial basic probability assignment function is generated by normalization, then according to the percentage contribution of each water-quality guideline to water resource pollution system, different weights is given respectively to it, obtain the basic probability assignment function after weighting, be specially: wherein α is weight coefficient, and n is Jiao unit number of basic probability assignment, obtains the basic probability assignment of this system thus.

5. the water resource pollution detection method based on evidence theory according to claim 1, it is characterized in that, described fusion refers to: when system has n index, and the basic probability assignment function after weighting being synthesized presses Dempster rule of combination combination n-1 time: $m\left(A\right)=\frac{\underset{B\∩C=A}{\mathrm{\Σ}}{m}_1\left(B\right)m_2\left(C\right)}{1-k},$ Wherein: