CN113627506A - Intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network - Google Patents

Intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network Download PDF

Info

Publication number
CN113627506A
CN113627506A CN202110883215.2A CN202110883215A CN113627506A CN 113627506 A CN113627506 A CN 113627506A CN 202110883215 A CN202110883215 A CN 202110883215A CN 113627506 A CN113627506 A CN 113627506A
Authority
CN
China
Prior art keywords
time
total phosphorus
interval
neural network
effluent
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110883215.2A
Other languages
Chinese (zh)
Inventor
韩红桂
孙晨暄
伍小龙
乔俊飞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Technology
Original Assignee
Beijing University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Technology filed Critical Beijing University of Technology
Priority to CN202110883215.2A priority Critical patent/CN113627506A/en
Publication of CN113627506A publication Critical patent/CN113627506A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/043—Architecture, e.g. interconnection topology based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Neurology (AREA)
  • Pure & Applied Mathematics (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Computational Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Automation & Control Theory (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention provides an intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network, which realizes intelligent detection of total phosphorus concentration in effluent in a sewage treatment process. Aiming at the characteristics of complexity and nonlinearity of a sewage treatment process, an accurate mathematical model is difficult to establish, the method for intelligently detecting the total phosphorus in the effluent establishes an effluent total phosphorus detection model based on an interval two-type fuzzy neural network by extracting characteristic variables related to the total phosphorus in the effluent, trains the detection model by using an information fusion method, realizes the automatic adjustment of the structure and parameters of the detection model, completes the design of the effluent total phosphorus intelligent detection model, ensures the real-time intelligent detection of the concentration of the total phosphorus in the effluent in the sewage treatment process and improves the detection precision on the premise of low-cost operation of a sewage treatment plant.

Description

Intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network
Technical Field
On the basis of analyzing the operating characteristics of the sewage treatment process, the intelligent detection method for the concentration of the total phosphorus in the effluent is realized by establishing an intelligent detection model for the total phosphorus in the effluent based on the interval two-type fuzzy neural network, designing a fuzzy rule by using an information fusion method and adjusting the parameters and the structure of the detection model. The method for intelligently detecting the total phosphorus in the effluent based on the information fusion-interval two-type fuzzy neural network can realize more accurate detection precision of the total phosphorus in the effluent on the premise of obtaining a compact model structure, and belongs to the field of water treatment.
Background
The premise of the sewage resource utilization is to ensure that the sewage is treated to obtain effluent which can reach the standard and be discharged. Other resources and energy sources are extracted from the sewage, and the method has important significance for optimizing water supply structure, increasing water resource supply, relieving contradiction between supply and demand, reducing water pollution and guaranteeing water ecological safety. Therefore, the method has wide application prospect in the research of the sewage treatment process.
Along with the research to sewage treatment process, the problem that partial organic matter of sewage discharges and exceeds standard is solved, but the problem that pollutant emissions such as nitrogen, phosphorus easily exceed standard still remains to be solved. The main standard exceeding indexes of the surface water source monitoring section are sulfate, permanganate index and total phosphorus. Therefore, the over-standard total phosphorus concentration is the main discharge problem of the current sewage treatment plant, and the content of the total phosphorus in the effluent of the sewage treatment plant is an important index for measuring the effluent quality of the sewage treatment plant. Meanwhile, the total phosphorus emission concentration is also used as an important measurement index of water quality in multiple industrial standards and national standards. Therefore, the method has great significance for preventing and treating water body pollution and recycling resources by accurately and quickly detecting the concentration of the total phosphorus in the effluent and timely treating and reducing the concentration of the total phosphorus in the effluent. Currently, sewage treatment plants can realize automatic detection of effluent water samples mainly by using an online total phosphorus analysis meter, and are simple to operate, but the instrument purchase and instrument maintenance costs are high. Therefore, the artificial neural network becomes an effective intelligent technology for detecting the total phosphorus in the water. The artificial neural network has nonlinear approximation capability and learning capability, is suitable for a sewage treatment process with nonlinear characteristics, and provides a new method for detecting the quality of sewage water through modeling. Therefore, the method has important significance in improving the detection precision of the effluent total phosphorus.
The invention designs an intelligent detection method of total phosphorus in effluent based on an information fusion-interval two-type fuzzy neural network, which establishes a detection model based on the interval two-type fuzzy neural network by extracting characteristic variables, adjusts the structure and parameters of the network by using the information fusion method to obtain the detection model with a compact structure, improves the detection precision of the total phosphorus in effluent, provides an effective method for realizing the accurate detection of the total phosphorus in effluent, and realizes the actual requirements of a sewage treatment plant.
Disclosure of Invention
The invention obtains an intelligent detection method for total phosphorus in effluent based on an information fusion-interval two-type fuzzy neural network, which determines main variables related to the total phosphorus in effluent by extracting characteristic variables so as to reduce the input dimension of a model, simultaneously optimizes the parameters and the structure of the interval two-type fuzzy neural network by using the information fusion method, constructs the model with a compact structure while ensuring the precision of the model, and improves the intelligent detection precision of the total phosphorus in effluent.
The invention adopts the following technical scheme and implementation steps:
an effluent total phosphorus intelligent detection method based on an information fusion-interval two-type fuzzy neural network is characterized by comprising the following steps of:
(1) selecting characteristic variables of effluent total phosphorus intelligent detection model
Selecting dissolved oxygen, oxidation-reduction potential, solid suspended matters, the pH value of a small water inlet chamber, ammonia nitrogen and temperature as characteristic variables influencing the total phosphorus in the effluent by taking a sewage treatment process as a research object; all variables are normalized according to the formula (1),
zi(t)＝(Di(t)-Di,min)/(Di,max-Di,min) (1)
wherein D is1(t) is the dissolved oxygen concentration at time t in mg/l, D2(t) oxidation at time tReduction potential in millivolts, D3(t) is the suspended solids concentration at time t in mg/l, D4(t) pH of the inlet cell at time t, D5(t) Ammonia nitrogen concentration at time t in mg/L, D6(t) is the temperature at time t, in degrees Celsius, D7(t) is the total phosphorus concentration in mg/l of the effluent at time t, Di,minIs the minimum value of the ith variable, Di,maxIs the maximum value of the ith variable; 1, …, 7; x is the number of1(t)＝z1(t) normalized dissolved oxygen, x, at time t2(t)＝z2(t) normalized Oxidation-reduction potential at time t, x3(t)＝z3(t) normalization of the solid suspension at time t, x4(t)＝z4(t) normalization of the inlet cell pH at time t, x5(t)＝z5(t) normalization of Ammonia Nitrogen, x at time t6(t)＝z6(t) is the normalized temperature at time t,
Figure BDA0003192980450000021
normalizing the total phosphorus concentration for time t, wherein Z is the total number of training samples, and t is 1, 2.
(2) Establishing an effluent total phosphorus intelligent detection model
The method comprises the following steps that an effluent total phosphorus intelligent detection model is established by utilizing a two-section type fuzzy neural network, wherein the two-section type fuzzy neural network comprises an input layer, a membership function layer, an activation layer, a back part layer and an output layer;
an input layer: consists of 6 neurons, with an input variable of x1(t),x2(t),x3(t),x4(t),x5(t),x6(t)；
Membership function layer: the neural network is composed of P neurons, wherein P is an integer in [2,20], and the output of each neuron of the membership function layer is an interval:
Figure BDA0003192980450000031
Figure BDA0003192980450000032
wherein, mu1ij(t) is the ith input, jth neuron upper bound output, μ at time t2ij(t) is the ith input, jth neuron lower bound output at time t, mij(t) is the upper center of the ith input jth neuron at time t, cij(t) is the lower center of the ith input jth neuron at time t, σij(t) is the width of the ith input jth neuron at time t, i ═ 1, …, 6; j is 1, …, P;
an active layer: consists of Q neurons, and Q ═ P, the activation layer upper and lower bound outputs are expressed as:
Figure BDA0003192980450000033
Figure BDA0003192980450000034
wherein f is1j(t) is the upper bound output of the jth neuron at time t, f2j(t) is the lower bound output of the jth neuron at time t;
a rear part layer: consisting of 2 neurons, the back-end layer output is expressed as:
Figure BDA0003192980450000035
Figure BDA0003192980450000036
Figure BDA0003192980450000037
wherein, y1(t) is the output value at time t, y2(t) is the output value at time t, wj(t) is the weight of the jth neuron at time t, aij(t) is the ith input of the jth neuron at time tWeight coefficient of input, bj(t) is the deviation of the jth neuron at time t;
an output layer: consists of 1 neuron, and the output layer output is expressed as:
y(t)＝q(t)y2(t)+(1-q(t))y1(t) (9)
wherein y (t) is the prediction output of the two-type fuzzy neural network in the t-time interval, and q (t) is a t-time proportionality coefficient;
(3) effluent total phosphorus intelligent detection model parameter adjustment based on information fusion
Initializing a detection model based on an interval two-type fuzzy neural network:
setting the current time t as 1, the initial upper center m of the interval type two fuzzy neural networkij(1) In [0,1 ]]Middle random value, initial lower center cij(1)＝0.5mij(1) Initial width σij(1) Initial weight a 1ij(1) In [ -1,1 [)]Middle random value, initial deviation bj(1) In [ -1,1 [)]Wherein, i is 1, …, 6; j is 1, …, P; the initial proportionality coefficient q (1) is 0.5; the threshold value is Ed，EdA positive number of 0.01 or less;
optimizing the parameters of the interval two-type fuzzy neural network:
the rule activation function of the interval type two fuzzy neural network is defined as follows:
Figure BDA0003192980450000041
wherein f isj(t)＝0.5f1j(t)+0.5f2j(t), g (t) is a rule activation function at time t; updating of the front-part parameters using a second order algorithm
Figure BDA0003192980450000042
Wherein, the rule activates the function parameter theta at the time tI(t)＝[m11(t),…,m6P(t),c11(t),…,c6P(t),σ11(t),…,σ6P(t)]Rule activation function parameter theta at time t +1I(t+1)＝[m11(t+1),…,m6P(t+1),c11(t+1),…,c6P(t+1),σ11(t+1),…,σ6P(t+1)]，
Figure BDA0003192980450000043
Figure BDA0003192980450000044
mij(t) is the upper center of the ith input jth neuron at time t, cij(t) is the lower center of the ith input jth neuron at time t, σij(t) is the width of the ith input jth neuron at time t, i ═ 1, …, 6; j is 1, …, P; i is an identity matrix, λI(t) is the learning rate of the rule activation function at time t, and the calculation method is as follows:
Figure BDA0003192980450000045
wherein, | | represents an absolute value, | represents a two-norm, and T represents a vector transposition;
defining a t-time error function e (t) of the interval type two fuzzy neural network as follows:
Figure BDA0003192980450000046
updating rule parameters using a second order algorithm
Figure BDA0003192980450000047
Wherein, the parameter vector theta of the error function at the time te(t)＝[m11(t),…,m6P(t),c11(t),…,c6P(t),σ11(t),…,σ6P(t),a11(t),…,a6P(t),b1(t),…,bP(t),q(t)]The parameter vector theta of the error function at time t +1e(t+1)＝[m11(t+1),…,m6P(t+1),c11(t+1),…,c6P(t+1),σ11(t+1),…,σ6P(t+1),a11(t+1),…,a6P(t+1),b1(t+1),…,bP(t+1),q(t+1)]，
Figure BDA0003192980450000048
Figure BDA0003192980450000049
aij(t) is a weight coefficient of the ith input of the jth neuron at time t, bj(t) is the deviation of the j-th neuron at time t, i ═ 1, …, 6; j is 1, …, P; lambda [ alpha ]e(t) is the learning rate of the error function at time t, and is calculated by:
Figure BDA0003192980450000051
adjusting the structure of the interval two-type fuzzy neural network:
the interval contribution degree of the fuzzy rule is as follows:
Figure BDA0003192980450000052
Figure BDA0003192980450000053
wherein M isj(t)＝[m1j(t),m2j(t),…,m6j(t)]T，Cj(t)＝[c1j(t),c2j(t),…,c6j(t)]T，Oj 1(t) is the input interval contribution at time t, Oj 2(t) is the contribution of the output interval at time t, fj(t)＝[q(t-L+1)f2j(t-L+1)+(1-q(t-L+1))f1j(t-L+1),…,q(t)f2j(t)+(1-q(t))f1j(t)]，Y(t)＝[y(t-L+1),y(t-L+2),…,y(t)]The number of samples L is a positive integer less than Z;
the interval correlation of the fuzzy rule is as follows:
Figure BDA0003192980450000054
Figure BDA0003192980450000055
wherein alpha isij(t) section-related information of the ith neuron and the jth neuron at time t;
when a neuron satisfies the following condition:
Figure BDA0003192980450000056
the interval type two fuzzy neural network adds a neuron, P is increased by 1, and Q is increased by 1; input contribution matrix O at time t1(t)＝[O1 1(t),O2 1(t)…,OP 1(t)]The matrix of contribution O is output at time t2(t)＝[O1 2(t),O2 2(t),…,OP 2(t)]The correlation contribution matrix α (t) at time t is [ α ]12(t),α13(t),…,α(P-1)P(t)]；
When a neuron satisfies the following condition:
Figure BDA0003192980450000061
the interval type two fuzzy neural network deletes one neuron, P is reduced by 1, and Q is reduced by 1;
when t is less than Z, increasing the time t by 1, and turning to the step II; otherwise, turning to the fifth step;
calculating the performance of the interval type II fuzzy neural network:
Figure BDA0003192980450000062
if E (t)>EdIf t is equal to 1, turning to the step II; otherwise, ending the circulation to obtain an effluent total phosphorus intelligent detection model;
(4) detection of total phosphorus concentration in effluent by using intelligent detection model
And (3) calculating the output of the water outlet total phosphorus intelligent detection model according to the formulas (2) to (9) by using the trained water outlet total phosphorus intelligent detection model and taking the dissolved oxygen concentration, the oxidation-reduction potential, the solid suspended matter concentration, the water inlet small chamber pH value, the ammonia nitrogen concentration and the temperature as the input of the water outlet total phosphorus intelligent detection model, wherein the output of the water outlet total phosphorus intelligent detection model is the detection value of the water outlet total phosphorus concentration.
The invention is mainly characterized in that:
(1) the invention provides an intelligent detection method of total phosphorus in effluent based on an interval type two fuzzy neural network aiming at longer detection period of total phosphorus in effluent in current sewage treatment, and solves the problem that the concentration of total phosphorus in effluent is difficult to measure in real time;
(2) the invention provides an information fusion method, which is characterized in that a model with a compact structure is obtained by automatically adjusting a two-section fuzzy neural network structure, so that the problem that the two-section fuzzy neural network structure is difficult to determine is solved, and the number of fuzzy rules is reduced; adopting an interval two-type fuzzy neural network based on an information fusion method to measure the total phosphorus in the effluent on line and ensure the intelligent detection precision of the model;
drawings
FIG. 1 is a structural diagram of an intelligent detection model of total phosphorus in effluent based on an interval two-type fuzzy neural network;
FIG. 2 is a graph of the training effect of total phosphorus in effluent according to the intelligent detection method of the present invention, wherein the solid line is the actual output value of total phosphorus in effluent, and the dotted line is the training value of the interval type two fuzzy neural network based on the information fusion method;
FIG. 3 is a graph of the total phosphorus training error in effluent from the intelligent detection method of the present invention;
FIG. 4 is a diagram of the predicted result of total phosphorus in effluent according to the intelligent detection method of the present invention, wherein the solid line is the actual output value of total phosphorus in effluent, and the dotted line is the predicted value of the interval type two fuzzy neural network based on the information fusion method;
FIG. 5 is a graph of the error of the prediction of total phosphorus in effluent according to the intelligent detection method of the present invention;
Detailed Description
The experimental data come from a water quality analysis table of 2021 years in a certain sewage treatment plant; taking actual detection data of inflow, anaerobic intermediate oxidation-reduction potential, anoxic pre-oxidation-reduction potential, anoxic final nitrate nitrogen concentration, aerobic pre-dissolved oxygen concentration, aerobic intermediate dissolved oxygen concentration, aerobic secondary dissolved oxygen concentration, pH value of a water inlet small chamber, water inlet small chamber solid suspended matters, water inlet small chamber chemical oxygen demand, water inlet small chamber ammonia nitrogen, primary sedimentation tank water outlet chemical oxygen demand, an external reflux flow meter, dosage, secondary fine final orthophosphate and residual sludge discharge amount as experimental sample data, and removing 1500 groups of available data after removing abnormal experimental samples, wherein 1000 groups are used as training samples, and the rest 500 groups are used as testing samples.
The invention adopts the following technical scheme and implementation steps:
1. an effluent total phosphorus intelligent detection method based on an information fusion-interval two-type fuzzy neural network is characterized by comprising the following steps of:
(1) selecting characteristic variables of effluent total phosphorus intelligent detection model
Selecting dissolved oxygen, oxidation-reduction potential, solid suspended matters, the pH value of a small water inlet chamber, ammonia nitrogen and temperature as characteristic variables influencing the total phosphorus in the effluent by taking a sewage treatment process as a research object; all variables are normalized according to equation (23),
zi(t)＝(Di(t)-Di,min)/(Di,max-Di,min) (23)
wherein D is1(t) is the dissolved oxygen concentration at time t in mg/l, D2(t) is the oxidation-reduction potential at time t in millivolts, D3(t) is the suspended solids concentration at time t in mg/l，D4(t) pH of the inlet cell at time t, D5(t) Ammonia nitrogen concentration at time t in mg/L, D6(t) is the temperature at time t, in degrees Celsius, D7(t) is the total phosphorus concentration in mg/l of the effluent at time t, Di,minIs the minimum value of the ith variable, Di,maxIs the maximum value of the ith variable; 1, …, 7; x is the number of1(t)＝z1(t) normalized dissolved oxygen, x, at time t2(t)＝z2(t) normalized Oxidation-reduction potential at time t, x3(t)＝z3(t) normalization of the solid suspension at time t, x4(t)＝z4(t) normalization of the inlet cell pH at time t, x5(t)＝z5(t) normalization of Ammonia Nitrogen, x at time t6(t)＝z6(t) is the normalized temperature at time t,
Figure BDA0003192980450000071
normalizing the total phosphorus concentration for time t, wherein t is 1, 2.., 1000;
(2) establishing an effluent total phosphorus intelligent detection model
The method comprises the following steps that an effluent total phosphorus intelligent detection model is established by utilizing a two-section type fuzzy neural network, wherein the two-section type fuzzy neural network comprises an input layer, a membership function layer, an activation layer, a back part layer and an output layer;
an input layer: consists of 6 neurons, with an input variable of x1(t),x2(t),x3(t),x4(t),x5(t),x6(t)；
Membership function layer: the neural network is composed of P neurons, wherein P is an integer in [2,20], and the output of each neuron of the membership function layer is an interval:
Figure BDA0003192980450000081
Figure BDA0003192980450000082
wherein, mu1ij(t) is the jth input of the ith input at time tUpper bound output, μ, of neurons2ij(t) is the ith input, jth neuron lower bound output at time t, mij(t) is the upper center of the ith input jth neuron at time t, cij(t) is the lower center of the ith input jth neuron at time t, σij(t) is the width of the ith input jth neuron at time t, i ═ 1, …, 6; j is 1, …, P;
an active layer: consists of Q neurons, and Q ═ P, the activation layer upper and lower bound outputs are expressed as:
Figure BDA0003192980450000083
Figure BDA0003192980450000084
wherein f is1j(t) is the upper bound output of the jth neuron at time t, f2j(t) is the lower bound output of the jth neuron at time t;
a rear part layer: consisting of 2 neurons, the back-end layer output is expressed as:
Figure BDA0003192980450000085
Figure BDA0003192980450000086
Figure BDA0003192980450000087
wherein, y1(t) is the output value at time t, y2(t) is the output value at time t, wj(t) is the weight of the jth neuron at time t, aij(t) is a weight coefficient of the ith input of the jth neuron at time t, bj(t) is the deviation of the jth neuron at time t;
an output layer: consists of 1 neuron, and the output layer output is expressed as:
y(t)＝q(t)y2(t)+(1-q(t))y1(t) (31)
wherein y (t) is the prediction output of the two-type fuzzy neural network in the t-time interval, and q (t) is a t-time proportionality coefficient;
(3) effluent total phosphorus intelligent detection model parameter adjustment based on information fusion
Initializing a detection model based on an interval two-type fuzzy neural network:
setting the current time t as 1, the initial upper center m of the interval type two fuzzy neural networkij(1) In [0,1 ]]Middle random value, initial lower center cij(1)＝0.5mij(1) Initial width σij(1) Initial weight a 1ij(1) In [ -1,1 [)]Middle random value, initial deviation bj(1) In [ -1,1 [)]Wherein, i is 1, …, 6; j is 1, …, P; the initial proportionality coefficient q (1) is 0.5; the threshold value is Ed，Ed＝0.01；
Optimizing the parameters of the interval two-type fuzzy neural network:
the rule activation function of the interval type two fuzzy neural network is defined as follows:
Figure BDA0003192980450000091
wherein f isj(t)＝0.5f1j(t)+0.5f2j(t), g (t) is a rule activation function at time t; updating of the front-part parameters using a second order algorithm
Figure BDA0003192980450000092
Wherein, the rule activates the function parameter theta at the time tI(t)＝[m11(t),…,m6P(t),c11(t),…,c6P(t),σ11(t),…,σ6P(t)]Rule activation function parameter theta at time t +1I(t+1)＝[m11(t+1),…,m6P(t+1),c11(t+1),…,c6P(t+1),σ11(t+1),…,σ6P(t+1)]，
Figure BDA0003192980450000093
Figure BDA0003192980450000094
mij(t) is the upper center of the ith input jth neuron at time t, cij(t) is the lower center of the ith input jth neuron at time t, σij(t) is the width of the ith input jth neuron at time t, i ═ 1, …, 6; j is 1, …, P; i is an identity matrix, λI(t) is the learning rate of the rule activation function at time t, and the calculation method is as follows:
Figure BDA0003192980450000095
wherein, | | represents an absolute value, | represents a two-norm, and T represents a vector transposition;
defining a t-time error function e (t) of the interval type two fuzzy neural network as follows:
Figure BDA0003192980450000096
updating rule parameters using a second order algorithm
Figure BDA0003192980450000097
Wherein, the parameter vector theta of the error function at the time te(t)＝[m11(t),…,m6P(t),c11(t),…,c6P(t),σ11(t),…,σ6P(t),a11(t),…,a6P(t),b1(t),…,bP(t),q(t)]The parameter vector theta of the error function at time t +1e(t+1)＝[m11(t+1),…,m6P(t+1),c11(t+1),…,c6P(t+1),σ11(t+1),…,σ6P(t+1),a11(t+1),…,a6P(t+1),b1(t+1),…,bP(t+1),q(t+1)]，
Figure BDA0003192980450000101
Figure BDA0003192980450000102
aij(t) is a weight coefficient of the ith input of the jth neuron at time t, bj(t) is the deviation of the j-th neuron at time t, i ═ 1, …, 6; j is 1, …, P; lambda [ alpha ]e(t) is the learning rate of the error function at time t, and is calculated by:
Figure BDA0003192980450000103
adjusting the structure of the interval two-type fuzzy neural network:
the interval contribution degree of the fuzzy rule is as follows:
Figure BDA0003192980450000104
Figure BDA0003192980450000105
wherein M isj(t)＝[m1j(t),m2j(t),…,m6j(t)]T，Cj(t)＝[c1j(t),c2j(t),…,c6j(t)]T，Oj 1(t) is the input interval contribution at time t, Oj 2(t) is the contribution of the output interval at time t, fj(t)＝[q(t-9)f2j(t-9)+(1-q(t-9))f1j(t-9),…,q(t)f2j(t)+(1-q(t))f1j(t)]，Y(t)＝[y(t-9),y(t-8),…,y(t)]；
The interval correlation of the fuzzy rule is as follows:
Figure BDA0003192980450000106
Figure BDA0003192980450000107
wherein alpha isij(t) section-related information of the ith neuron and the jth neuron at time t;
when a neuron satisfies the following condition:
Figure BDA0003192980450000108
the interval type two fuzzy neural network adds a neuron, P is increased by 1, and Q is increased by 1; input contribution matrix O at time t1(t)＝[O1 1(t),O2 1(t)…,OP 1(t)]The matrix of contribution O is output at time t2(t)＝[O1 2(t),O2 2(t),…,OP 2(t)]The correlation contribution matrix α (t) at time t is [ α ]12(t),α13(t),…,α(P-1)P(t)]；
When a neuron satisfies the following condition:
Figure BDA0003192980450000111
the interval type two fuzzy neural network deletes one neuron, P is reduced by 1, and Q is reduced by 1;
fourthly, when t is less than 1000, increasing 1 at the moment t, and turning to the second step; otherwise, turning to the fifth step;
calculating the performance of the interval type II fuzzy neural network:
Figure BDA0003192980450000112
if E (t) is greater than 0.01, if t is equal to 1, turning to the step II; otherwise, ending the circulation to obtain an effluent total phosphorus intelligent detection model;
(4) detection of total phosphorus concentration in effluent by using intelligent detection model
The training result of the intelligent detection method for the total phosphorus concentration of the effluent is shown in figure 2, and the X axis: number of training samples, unit is one, Y-axis: the unit of the output water total phosphorus training output is milligram/liter, the solid line is the output water total phosphorus actual output, and the dotted line is the output water total phosphorus prediction output; the error between the actual output of total phosphorus in the effluent and the training output is shown in fig. 3, X axis: number of training samples, unit is one, Y-axis: the total phosphorus training error of the effluent is milligram/liter;
and (3) calculating the output of the total phosphorus intelligent detection model of the effluent according to the formulas (24) to (31), namely the detection value of the total phosphorus concentration of the effluent, by using the trained total phosphorus intelligent detection model of the effluent and taking the concentration of dissolved oxygen, the oxidation-reduction potential, the concentration of suspended solids, the pH of a small water inlet chamber, the concentration of ammonia nitrogen and the temperature as the input of the total phosphorus intelligent detection model of the effluent.
The test result of the intelligent detection method for the total phosphorus concentration of the effluent is shown in figure 4, and the X axis: number of samples tested, in units of units, Y-axis: the unit of the output water total phosphorus prediction output is mg/L, the solid line is the output water total phosphorus actual output, and the dotted line is the output water total phosphorus prediction output; the error between the actual output and the test output of the total phosphorus in the effluent is shown in figure 5, and the X axis: number of samples tested, in units of units, Y-axis: the prediction error of the total phosphorus of the effluent is milligram/liter. The experimental result shows the effectiveness of the intelligent detection method for the total phosphorus in the effluent based on the information fusion-interval two-type fuzzy neural network.

Claims (1)

1. An effluent total phosphorus intelligent detection method based on information fusion-interval two-type fuzzy neural network is characterized by comprising the following steps:
(1) selecting characteristic variables of effluent total phosphorus intelligent detection model
Selecting dissolved oxygen, oxidation-reduction potential, solid suspended matters, the pH value of a small water inlet chamber, ammonia nitrogen and temperature as characteristic variables influencing the total phosphorus in the effluent by taking a sewage treatment process as a research object; all variables are normalized according to the formula (1),
zi(t)＝(Di(t)-Di,min)/(Di,max-Di,min) (1)
wherein D is1(t) is the dissolved oxygen concentration at time t in mg/l, D2(t) is the oxidation-reduction potential at time t in millivolts, D3(t) is the suspended solids concentration at time t in mg/l, D4(t) pH of the inlet cell at time t, D5(t) Ammonia nitrogen concentration at time t in mg/L, D6(t) is the temperature at time t, in degrees Celsius, D7(t) is the total phosphorus concentration in mg/l of the effluent at time t, Di,minIs the minimum value of the ith variable, Di,maxIs the maximum value of the ith variable; 1, …, 7; x is the number of1(t)＝z1(t) normalized dissolved oxygen, x, at time t2(t)＝z2(t) normalized Oxidation-reduction potential at time t, x3(t)＝z3(t) normalization of the solid suspension at time t, x4(t)＝z4(t) normalization of the inlet cell pH at time t, x5(t)＝z5(t) normalization of Ammonia Nitrogen, x at time t6(t)＝z6(t) is the normalized temperature at time t,
Figure FDA0003192980440000013
normalizing the total phosphorus concentration for time t, wherein Z is the total number of training samples, and t is 1, 2.
(2) Establishing an effluent total phosphorus intelligent detection model
The method comprises the following steps that an effluent total phosphorus intelligent detection model is established by utilizing a two-section type fuzzy neural network, wherein the two-section type fuzzy neural network comprises an input layer, a membership function layer, an activation layer, a back part layer and an output layer;
an input layer: consists of 6 neurons, with an input variable of x1(t),x2(t),x3(t),x4(t),x5(t),x6(t)；
Membership function layer: the neural network is composed of P neurons, wherein P is an integer in [2,20], and the output of each neuron of the membership function layer is an interval:
Figure FDA0003192980440000011
Figure FDA0003192980440000012
wherein, mu1ij(t) is the ith input, jth neuron upper bound output, μ at time t2ij(t) is the ith input, jth neuron lower bound output at time t, mij(t) is the upper center of the ith input jth neuron at time t, cij(t) is the lower center of the ith input jth neuron at time t, σij(t) is the width of the ith input jth neuron at time t, i ═ 1, …, 6; j is 1, …, P;
an active layer: consists of Q neurons, and Q ═ P, the activation layer upper and lower bound outputs are expressed as:
Figure FDA0003192980440000021
Figure FDA0003192980440000022
wherein f is1j(t) is the upper bound output of the jth neuron at time t, f2j(t) is the lower bound output of the jth neuron at time t;
a rear part layer: consisting of 2 neurons, the back-end layer output is expressed as:
Figure FDA0003192980440000023
Figure FDA0003192980440000024
Figure FDA0003192980440000025
wherein, y1(t) is the output value at time t, y2(t) is the output value at time t, wj(t) is the weight of the jth neuron at time t, aij(t) is a weight coefficient of the ith input of the jth neuron at time t, bj(t) is the deviation of the jth neuron at time t;
an output layer: consists of 1 neuron, and the output layer output is expressed as:
y(t)＝q(t)y2(t)+(1-q(t))y1(t) (9)
wherein y (t) is the prediction output of the two-type fuzzy neural network in the t-time interval, and q (t) is a t-time proportionality coefficient;
(3) effluent total phosphorus intelligent detection model parameter adjustment based on information fusion
Initializing a detection model based on an interval two-type fuzzy neural network:
setting the current time t as 1, the initial upper center m of the interval type two fuzzy neural networkij(1) In [0,1 ]]Middle random value, initial lower center cij(1)＝0.5mij(1) Initial width σij(1) Initial weight a 1ij(1) In [ -1,1 [)]Middle random value, initial deviation bj(1) In [ -1,1 [)]Wherein, i is 1, …, 6; j is 1, …, P; the initial proportionality coefficient q (1) is 0.5; the threshold value is Ed，EdA positive number less than 0.01;
optimizing the parameters of the interval two-type fuzzy neural network:
the rule activation function of the interval type two fuzzy neural network is defined as follows:
Figure FDA0003192980440000026
wherein f isj(t)＝0.5f1j(t)+0.5f2j(t), g (t) is a rule activation function at time t; updating of the front-part parameters using a second order algorithm
Figure FDA0003192980440000031
Wherein, the rule activates the function parameter theta at the time tI(t)＝[m11(t)，...，m6P(t)，c11(t)，...，c6P(t)，σ11(t)，...，σ6P(t)]Rule activation function parameter theta at time t +1I(t+1)＝[m11(t+1)，...，m6P(t+1)，c11(t+1)，...，c6P(t+1)，σ11(t+1)，...，σ6P(t+1)]，
Figure FDA0003192980440000037
Figure FDA0003192980440000038
mij(t) is the upper center of the ith input jth neuron at time t, cij(t) is the lower center of the ith input jth neuron at time t, σij(t) is the width of the ith input jth neuron at time t, i ═ 1.., 6; j 1.., P; i is an identity matrix, λI(t) is the learning rate of the rule activation function at time t, and the calculation method is as follows:
Figure FDA0003192980440000032
wherein, | | represents an absolute value, | represents a two-norm, and T represents a vector transposition;
defining a t-time error function e (t) of the interval type two fuzzy neural network as follows:
Figure FDA0003192980440000033
updating rule parameters using a second order algorithm
Figure FDA0003192980440000034
Wherein, the parameter vector theta of the error function at the time te(t)＝[m11(t)，...，m6P(t)，c11(t)，...，c6P(t)，σ11(t)，...，σ6P(t)，a11(t)，...，a6P(t)，b1(t)，...，bP(t)，q(t)]The parameter vector theta of the error function at time t +1e(t+1)＝[m11(t+1)，...，m6P(t+1)，c11(t+1)，...，c6P(t+1)，σ11(t+1)，...，σ6P(t+1)，a11(t+1)，...，a6P(t+1)，b1(t+1)，...，bP(t+1)，q(t+1)]，
Figure FDA0003192980440000039
Figure FDA00031929804400000310
aij(t) is a weight coefficient of the ith input of the jth neuron at time t, bj(t) is the deviation of the j-th neuron at time t, i ═ 1.., 6; j 1.., P; lambda [ alpha ]e(t) is the learning rate of the error function at time t, and is calculated by:
Figure FDA0003192980440000035
adjusting the structure of the interval two-type fuzzy neural network:
the interval contribution degree of the fuzzy rule is as follows:
Figure FDA0003192980440000036
Figure FDA0003192980440000041
wherein M isj(t)＝[m1j(t)，m2j(t)，...，m6j(t)]T，Cj(t)＝[c1j(t)，c2j(t)，...，c6j(t)]T，Oj 1(t) is the input interval contribution at time t,
Figure FDA0003192980440000042
is the contribution of the output interval at time t, fj(t)＝[q(t-L+1)f2j(t-+1)+(1-q(t-L+1))f1j(t-L+1)，...，q(t)f2j(t)+(1-q(t))f1j(t)]，Y(t)＝[y(t-L+1)，y(t-L+2)，...，y(t)]The number of samples L is a positive integer less than Z;
the interval correlation of the fuzzy rule is as follows:
Figure FDA0003192980440000043
Figure FDA0003192980440000044
wherein alpha isij(t) section-related information of the ith neuron and the jth neuron at time t;
when a neuron satisfies the following condition:
Figure FDA0003192980440000045
the interval type two fuzzy neural network adds a neuron, P is increased by 1, and Q is increased by 1; input contribution matrix O at time t1(t)＝[O1 1(t)，O2 1(t)...，OP 1(t)]The matrix of contribution O is output at time t2(t)＝[O1 2(t)，O2 2(t)，...，OP 2(t)]The correlation contribution matrix α (t) at time t is [ α ]12(t)，α13(t)，...，α(P-1)P(t)]；
When a neuron satisfies the following condition:
Figure FDA0003192980440000046
the interval type two fuzzy neural network deletes one neuron, P is reduced by 1, and Q is reduced by 1;
fourthly, when t is smaller than Z, increasing the time t by 1, and turning to the second step; otherwise, turning to the fifth step;
calculating the performance of the interval type II fuzzy neural network:
Figure FDA0003192980440000047
if E (t) > EdIf t is equal to 1, turning to the step II; otherwise, ending the circulation to obtain an effluent total phosphorus intelligent detection model;
(4) detection of total phosphorus concentration in effluent by using intelligent detection model
And (3) calculating the output of the water outlet total phosphorus intelligent detection model according to the formulas (2) to (9) by using the trained water outlet total phosphorus intelligent detection model and taking the dissolved oxygen concentration, the oxidation-reduction potential, the solid suspended matter concentration, the water inlet small chamber pH value, the ammonia nitrogen concentration and the temperature as the input of the water outlet total phosphorus intelligent detection model, wherein the output of the water outlet total phosphorus intelligent detection model is the detection value of the water outlet total phosphorus concentration.
CN202110883215.2A 2021-08-02 2021-08-02 Intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network Pending CN113627506A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110883215.2A CN113627506A (en) 2021-08-02 2021-08-02 Intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110883215.2A CN113627506A (en) 2021-08-02 2021-08-02 Intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network

Publications (1)

Publication Number Publication Date
CN113627506A true CN113627506A (en) 2021-11-09

Family

ID=78382298

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110883215.2A Pending CN113627506A (en) 2021-08-02 2021-08-02 Intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network

Country Status (1)

Country Link
CN (1) CN113627506A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114814130A (en) * 2022-03-04 2022-07-29 北京工业大学 Intelligent detection method for total nitrogen in effluent of interval type two fuzzy neural network based on nonsingular gradient descent algorithm
CN114912580A (en) * 2022-05-07 2022-08-16 北京工业大学 Intelligent detection method for total phosphorus in two-section fuzzy neural network effluent based on self-adaptive discrimination strategy

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114814130A (en) * 2022-03-04 2022-07-29 北京工业大学 Intelligent detection method for total nitrogen in effluent of interval type two fuzzy neural network based on nonsingular gradient descent algorithm
CN114814130B (en) * 2022-03-04 2024-04-26 北京工业大学 Intelligent detection method for total nitrogen in water outlet of interval two-type model neural network based on nonsingular gradient descent algorithm
CN114912580A (en) * 2022-05-07 2022-08-16 北京工业大学 Intelligent detection method for total phosphorus in two-section fuzzy neural network effluent based on self-adaptive discrimination strategy
CN114912580B (en) * 2022-05-07 2024-05-28 北京工业大学 Intelligent detection method for total phosphorus in water outlet of interval two-type fuzzy neural network based on self-adaptive discrimination strategy

Similar Documents

Publication Publication Date Title
CN105510546B (en) A kind of biochemical oxygen demand (BOD) BOD intelligent detecting methods based on self-organizing Recurrent RBF Neural Networks
CN106295800B (en) A kind of water outlet total nitrogen TN intelligent detecting method based on recurrence Self organizing RBF Neural Network
CN103197544B (en) Sewage disposal process multi-purpose control method based on nonlinear model prediction
CN107358021B (en) DO prediction model establishment method based on BP neural network optimization
CN102854296B (en) Sewage-disposal soft measurement method on basis of integrated neural network
CN111354423B (en) Method for predicting ammonia nitrogen concentration of effluent of self-organizing recursive fuzzy neural network based on multivariate time series analysis
Gontarski et al. Simulation of an industrial wastewater treatment plant using artificial neural networks
CN104360035B (en) A kind of sewage total phosphorus TP flexible measurement method based on self-organization population-radial base neural net
CN110378533B (en) Intelligent aeration management method based on big data analysis
CN112183719A (en) Intelligent detection method for total nitrogen in effluent based on multi-objective optimization-fuzzy neural network
CN113627506A (en) Intelligent detection method for total phosphorus in effluent based on information fusion-interval two-type fuzzy neural network
Pisa et al. LSTM-based wastewater treatment plants operation strategies for effluent quality improvement
CN109344971B (en) Effluent ammonia nitrogen concentration prediction method based on adaptive recursive fuzzy neural network
CN114275912B (en) Aeration system dissolved oxygen control method based on adaptive neural network model
CN108375534A (en) MBR fouling membrane intelligent early-warning methods
CN108562709A (en) A kind of sewage disposal system water quality monitoring method for early warning based on convolution self-encoding encoder extreme learning machine
CN112989704B (en) IRFM-CMNN effluent BOD concentration prediction method based on DE algorithm
CN114814130B (en) Intelligent detection method for total nitrogen in water outlet of interval two-type model neural network based on nonsingular gradient descent algorithm
CN103809436A (en) Method for intelligent modeling sewage disposal process using activated sludge process
CN111125907B (en) Sewage treatment ammonia nitrogen soft measurement method based on hybrid intelligent model
CN112417765A (en) Sewage treatment process fault detection method based on improved teacher-student network model
CN117389142A (en) Prediction deduction and safe real-time control method for biochemical reaction tank of sewage treatment plant
CN115356930A (en) Multi-objective optimization control system and method in sewage treatment process
CN113156074B (en) Effluent total nitrogen prediction method based on fuzzy migration
CN114861543A (en) Data-driven intelligent evaluation method for biodegradability of petrochemical sewage

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination