CN106203756A - A kind of trade effluent pollutant Source Tracing method of feature based data base - Google Patents

Abstract

A kind of trade effluent pollutant Source Tracing method of feature based data base, belongs to trade effluent pollutant regulation technique field.Its feature includes: set up the feature weight data base M of all types of factories pollutant parameter, gather the pollutant data that each factory in target area drains the sewage away, build the characteristic Sample Storehouse N of target area factory blowdown, described characteristic Sample Storehouse N is utilized to build the class variable of KD-tree, each pollutant measurement value vector to be identified z of input mixing industrial wastewater, use described kNN grader to carry out match cognization with the data in characteristic Sample Storehouse N, complete pollutant and trace to the source.The present invention can complete the pollutant of various region mixing industrial wastewater and trace to the source, provide the sequence of target discharge factory fast and effectively, the suitability and universality are strong, investigate pollution source factory in order provide technical support for the functional department that is correlated with of government, substantially increase investigation efficiency and success rate.

Description

A kind of trade effluent pollutant Source Tracing method of feature based data base

Technical field

The invention belongs to trade effluent pollutant regulation technique field, relate to a kind of trade effluent pollutant and trace to the source Analysis method, a kind of trade effluent pollutant Source Tracing method of feature based data base.

Background technology

The water pollution problems of China is quite serious at present, and this has remained incessant after repeated prohibition directly with the row's phenomenon steathily that illegally exceeds standard The relation connect.The supervision strengthening all types of industries sewage effluent has become as the important content of governments at all levels' work, Effectively supervision to illegal super row still lacks effective technological means with tracing to the source to call to account.Traditional skill of tracing to the source Art is typically all after contamination accident occurs, and is sampled the plant effuent near polluter upstream and downstream analyzing, Workload longer, poor in timeliness big, time-consuming, is often difficult to find polluter in time.

The most also emerge some new tracing technologies, but be typically all for a certain class industrial wastewater or Some anion and carry out, or in terms of regulatory management propose method, such as China Environmental Science The source tracing method of risk source " a kind of water environment pollution accident that happens suddenly be accused of " disclosed in institute 2012, publication number For CN102622670A, for another example disclosed in South China Institute of Environmental Sciences. MEP 2014 " one Plant the source tracing method of sudden water environment event polluter under limited conditions ", Publication No. CN104103005A, The method mainly gives a system, and its Matching Model is suitable for the most under the limited conditions.

Still lack a kind of source tracing method being generally suitable for various industry composite waste, meet variety classes industrial occupancy The demand of tracing to the source of territory combined sewage.

Summary of the invention

The invention aims to overcome the deficiency of above-mentioned background technology, it is provided that a kind of feature based data base Trade effluent pollutant Source Tracing method, it is possible to be applicable to various industry composite wastes trace to the source.

The trade effluent pollutant Source Tracing method such as Fig. 1 of a kind of feature based data base that the present invention provides Shown in, comprise the following steps that:

(1) collect all types of factory owner and want the pollutant parameter of sewage effluent, set up all types of factory and pollute The feature weight data base M of thing parameter；

(2) gather the pollutant data that each factory in target area drains the sewage away, build the blowdown of target area factory Characteristic Sample Storehouse N；

(3) utilize described characteristic Sample Storehouse N to build the class variable of KD-tree, classify as kNN The identification model of device；

(4) gather industrial computer by integrated form pollutant obtain or be directly manually entered mixing industrial wastewater Each pollutant measurement value, as to be identified vector z；

(5) described kNN grader is used to carry out match cognization with the data in characteristic Sample Storehouse N, In the matching process, described feature weight data base M is utilized according to factory types, target factory to be weighted, The sequence of target discharge factory is given according to matching degree；

(6) target discharge factory title is shown in order.

The core identification algorithm of described kNN grader is:

Input: the training sample set D of described characteristic Sample Storehouse N, described vector z to be identified, sample Class label collection L.

Output:c_zThe classification of ∈ L, i.e. sample vector.

For each sample y ∈ D do

Calculate z to y distance d (z, y)

end

K the training sample closest with z is selected to constitute set N,

$c_z=\underset{v\∈L}{\arg \max }{\mathrm{\Σ}}_{y\∈N}I(v=\mathrm{class}\left(c_y\right))$

Wherein, k is the number of the predetermined target discharge factory identified, I () is indicator function, if condition becomes Vertical value is 1, if condition is false, value is 0.

Treatment of details principle in kNN grader includes:

(z is y) sample to be identified and the measuring similarity of sample in training sample database, in representing feature space to A.d The distance of two points.Other mathematical distance alternative have multiple, such as Euler's distance, mahalanobis distance, Manhatton distance, Chebyshev's distance, Min Shi distance etc.；

B. use weighted voting algorithm carry out kind judging, take the principle that the minority is subordinate to the majority, in neighbour which Which class the point of classification is the most just divided into, and weight is chosen by factory types according to feature weight data base M；

C. characteristic vector is normalized, reduces the impact that variable codomain is adjusted the distance；

D. the method using principal component analysis, reduces the high-dimensional impact adjusted the distance and weigh.

The invention has the beneficial effects as follows, it is possible to the pollutant completing various region mixing industrial wastewater are traced to the source, soon Speed effectively provides the sequence of target discharge factory, and the suitability is strong with universality, is correlated with functional department for government Investigation pollution source factory provides technical support in order, substantially increases investigation efficiency and success rate.

Accompanying drawing explanation

Fig. 1 trade effluent pollutant Source Tracing method flow diagram.

Fig. 2 target area factory title diagram.

Detailed description of the invention:

The present invention is described in further detail with embodiment below in conjunction with the accompanying drawings, but this embodiment should not be managed Solve as limitation of the present invention.

In the present embodiment, industrial park, selection Daliang City, target area, in region, factory includes 12, As shown in Figure 2, each factory drains the sewage away pollutant kind 28 kinds specific name, including benzene, pH value, ammonia Nitrogen, colourity, toluene, total nitrogen, total cadmium, total chromium, total nickel, total arsenic, total copper, total zinc, dimethylbenzene, fluorine Compound, volatile phenol etc..

Trace to the source according to following operating procedure:

(1) collect all types of factory owner and want the pollutant parameter of sewage effluent, set up all types of factory and pollute The feature weight data base M1 of thing parameter；

(2) gather 28 kinds of pollutant data that each factory shown in Fig. 2 drains the sewage away, build target area work The characteristic Sample Storehouse N1 of factory's blowdown, and according to type, factory is classified；

(3) utilize described characteristic Sample Storehouse N1 to build the class variable of KD-tree, divide as kNN The identification model of class device；

(4) manually input mixes each pollutant value of industrial wastewater, as vector z1 to be identified；

(5) described kNN grader is used to carry out match cognization with the data in characteristic Sample Storehouse N1, In the matching process, described feature weight data base M1 is utilized according to factory types, target factory to be added Power, provides the sequence of target discharge factory according to matching degree；

(6) described target discharge factory title is shown in order.

In described kNN classifier algorithm, described distance d (z1, y) selects Euclidean distance, and formula is as follows:

$d_{\mathrm{euc}}(x,y)={\left[\underset{j=1}{\overset{d}{\mathrm{\Σ}}}{(x_j-y_j)}^2\right]}^{\frac{1}{2}}={\left[(x-y){(x-y)}^T\right]}^{\frac{1}{2}}$

In described kNN classifier algorithm, k value is chosen as 3, and franchise repeated root is according to feature weight data base M1 Choose by factory types, characteristic vector z1 has been carried out normalization, the method that have employed principal component analysis.

Claims (1)

1. the trade effluent pollutant Source Tracing method of a feature based data base, it is characterised in that

(1) collect all types of factory owner and want the pollutant parameter of sewage effluent, set up all types of factory and pollute The feature weight data base M of thing parameter；

(2) gather the pollutant data that each factory in target area drains the sewage away, build the blowdown of target area factory Characteristic Sample Storehouse N；

(3) utilize described characteristic Sample Storehouse N to build the class variable of KD-tree, classify as kNN The identification model of device；

(4) gather industrial computer by integrated form pollutant obtain or be directly manually entered mixing industrial wastewater Each pollutant measurement value, as to be identified vector z；

(5) described kNN grader is used to carry out match cognization with the data in characteristic Sample Storehouse N, In the matching process, described feature weight data base M is utilized according to factory types, target factory to be weighted, The sequence of target discharge factory is given according to matching degree；

Use weighted voting algorithm to carry out kind judging, take the principle that the minority is subordinate to the majority, which class in neighbour Which class other point is the most just divided into, and weight is chosen by factory types according to feature weight data base M；

Characteristic vector is normalized, reduces the impact that variable codomain is adjusted the distance；

The method using principal component analysis, reduces the high-dimensional impact adjusted the distance and weigh；

The core identification algorithm of described kNN grader is:

Input: the training sample set D of described characteristic Sample Storehouse N, described vector z to be identified, sample Class label collection L；

Output:c_zThe classification of ∈ L, i.e. sample vector；

For each sample y ∈ D do

Calculate z to y distance d (z, y)

end

K the training sample closest with z is selected to constitute set N,

$c_z=\underset{v\∈L}{\arg \max }{\mathrm{\Σ}}_{y\∈N}I(v=\mathrm{class}\left(c_y\right))$

Wherein, k is the number of the predetermined target discharge factory identified, I () is indicator function, if condition becomes Vertical value is 1, if condition is false, value is 0；

(z, is y) sample to be identified and the measuring similarity of sample in training sample database to d, represents in feature space two The distance of individual point；

(6) target discharge factory title is shown in order.