[water quality prediction] water quality prediction based on MATLAB Fuzzy Neural Network [including Matlab source code 1923]

One 、 Introduction to fuzzy neural network water quality prediction

1 Fuzzy neural network
From the perspective of the overall structural framework , Two inputs and one output , It is like a neural network ; In terms of function , It is a fuzzy system . The structure of fuzzy neural network is shown in the figure 1 Shown , This network structure is divided into 5 A hierarchical , It is a fuzzy inference system constructed by neural network , Based on the working process of fuzzy system . First , The input layer inputs two signals ; Next, blur the layer , Blur the input signal ; Thirdly, the fuzzy rule calculation layer ; Then there is the fuzzy decision-making level , Classify the quantities that meet the conditions and convert the fuzzy quantities into clear results ; Finally, the fifth layer is the output layer , Output the final operation result .

2 Water quality evaluation
So far, there are many methods of water quality monitoring and evaluation , But there are limitations . Our country has been using nutrients 、 Inorganic matter 、 Microorganisms and heavy metal ions are the main water quality parameters . Traditional biological or chemical water quality evaluation methods , We can only get whether it is polluted and the main components of sewage , It cannot reflect the real potential problems of water quality . Integrated water pollution system with online automatic continuous monitoring , Compared with the continuous automatic monitoring system of air pollution , Water quality monitoring is much more difficult . Of course , All these problems can be solved , But it is particularly important to find a suitable water quality evaluation method .

chart 2 General steps of fuzzy neural network water quality evaluation
2.1 Main parameters of water quality evaluation
There are six commonly used parameters in water quality evaluation : (1) Conventional water quality parameters ; (2) Oxygen balance parameters ; (3) Heavy metal parameters ; (4) Parameters of organic pollutants ; (5) Parameters of inorganic pollutants , ; (6) Biological parameters .

2.2 General steps of water quality evaluation
The main steps of the fuzzy neural network water quality evaluation algorithm experiment are shown in the figure 2 Shown , It is mainly divided into six steps .

3 be based on T-S Fuzzy neural network modeling
3.1 T-S Fuzzy model
T-S Fuzzy system can not only automatically update the membership function of fuzzy subset , And can constantly update the membership function . It belongs to a growing self-learning system . This model , T-S Fuzzy system is defined in the following “if-then” In the rules . According to rules Ri in , Fuzzy reasoning is as follows :

among , Aij Is the fuzzy set of fuzzy system ;pij (j=1, 2, …, k) Is the parameter of fuzzy system ;yi Is the output obtained according to fuzzy rules , Input part ( namely if part ) It's fuzzy , Output part ( namely then part ) Is to determine the , The fuzzy inference represents that the output is a linear combination of inputs .

Suppose the input quantity x=[x1, x2, …xk], First, calculate each input variable according to fuzzy rules , xj Membership of .

among , cji, bji They are the center and width of the membership function ;k For input parameters ;n Is the number of fuzzy subsets . Fuzzy calculation of each membership degree , Apply fuzzy operator as a continued multiplication operator .

Calculate the output value of the fuzzy model according to the fuzzy calculation results yi.

3.2 T-S Fuzzy neural network
T-S Fuzzy neural network consists of four layers : Input layer 、 Fuzzification layer 、 Fuzzy rule calculation layer and output layer . The learning algorithm of fuzzy neural network is as follows :

(1) Error calculation

In style , yd Is the expected output of the network ;yc Network actual output ;e Error between expected output and actual output .

(2) Coefficient correction

In style , pji Is the neural network coefficient , α For the efficiency of online learning , xj Enter parameters for the network ;ωi Is the product of the membership degree of the input parameter .

(3) Parameter correction

In style , cji, bji They are the center and width of the membership function .

3.3 Model building
Fuzzy neural network takes the dimension of training samples as the basic premise , The basic elements of constructing fuzzy neural network : Input 、 Number of output nodes 、 Number of fuzzy membership functions . Finally, the input data is selected as 6 individual ( Including ammonia nitrogen content 、 Dissolved oxygen content 、 Chemical oxygen demand 、 Permanganate index 、 Six indicators of total phosphorus and total nitrogen ) , The water quality grade is divided into five categories , Use numbers 1—5 Indicates the water quality grade I—V class , So the output node number is 1, To form the 6-12-1 Network structure .

Two 、 Partial source code

%---------------------------------------------%
%
%---------------------------------------------%

%%  Clear environment variables 
clc
clear

%%  Parameter initialization 
xite=0.001;
alfa=0.05;

% Network nodes 
I=6;   % Enter the number of nodes 
M=12;  % Number of hidden nodes 
O=1;   % Number of output nodes 

% Coefficient initialization 
p0=0.3*ones(M,1);p0_1=p0;p0_2=p0_1;
p1=0.3*ones(M,1);p1_1=p1;p1_2=p1_1;
p2=0.3*ones(M,1);p2_1=p2;p2_2=p2_1;
p3=0.3*ones(M,1);p3_1=p3;p3_2=p3_1;
p4=0.3*ones(M,1);p4_1=p4;p4_2=p4_1;
p5=0.3*ones(M,1);p5_1=p5;p5_2=p5_1;
p6=0.3*ones(M,1);p6_1=p6;p6_2=p6_1;

% Parameter initialization 
c=1+rands(M,I);c_1=c;c_2=c_1;
b=1+rands(M,I);b_1=b;b_2=b_1;

maxgen=100; % Number of evolutions 

% Network test data , And normalize the data 
load data1 input_train output_train input_test output_test

% Select and connect the sample input and output data 
[inputn,inputps]=mapminmax(input_train);
[outputn,outputps]=mapminmax(output_train);
[n,m]=size(input_train);

%%  Network training 
% Loop start , Evolutionary networks 
for iii=1:maxgen
    iii
    for k=1:m        
        x=inputn(:,k);
        
        % Output layer settlement 
        for i=1:I
            for j=1:M
                u(i,j)=exp(-(x(i)-c(j,i))^2/b(j,i));
            end
        end

3、 ... and 、 Running results

Four 、matlab Edition and references

1 matlab edition
2014a

2 reference
[1] Kang Caili . Application of fuzzy neural network in water quality evaluation [J]. Journal of Xinzhou Normal University . 2019,35(02)

3 remarks
This part of the introduction is taken from the Internet , For reference only , If infringement , Contact deletion