CN109408896A - A kind of anerobic sowage processing gas production multi-element intelligent method for real-time monitoring - Google Patents
A kind of anerobic sowage processing gas production multi-element intelligent method for real-time monitoring Download PDFInfo
- Publication number
- CN109408896A CN109408896A CN201811130613.1A CN201811130613A CN109408896A CN 109408896 A CN109408896 A CN 109408896A CN 201811130613 A CN201811130613 A CN 201811130613A CN 109408896 A CN109408896 A CN 109408896A
- Authority
- CN
- China
- Prior art keywords
- fuzzy
- network
- layer
- wavelet
- fwnn
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 34
- 238000012544 monitoring process Methods 0.000 title claims abstract description 28
- 238000012545 processing Methods 0.000 title abstract description 15
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 claims abstract description 30
- 239000010865 sewage Substances 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 6
- 238000005259 measurement Methods 0.000 claims abstract description 4
- 238000005070 sampling Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 48
- 238000013528 artificial neural network Methods 0.000 claims description 20
- 230000003044 adaptive effect Effects 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 9
- 230000002068 genetic effect Effects 0.000 claims description 8
- 238000011478 gradient descent method Methods 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000013519 translation Methods 0.000 claims description 6
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 6
- 210000000349 chromosome Anatomy 0.000 claims description 5
- 230000007423 decrease Effects 0.000 claims description 5
- 230000008901 benefit Effects 0.000 claims description 4
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 238000010353 genetic engineering Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000035772 mutation Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 239000007787 solid Substances 0.000 claims 1
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 description 2
- BVKZGUZCCUSVTD-UHFFFAOYSA-L Carbonate Chemical compound [O-]C([O-])=O BVKZGUZCCUSVTD-UHFFFAOYSA-L 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 239000002351 wastewater Substances 0.000 description 2
- 229910002092 carbon dioxide Inorganic materials 0.000 description 1
- 239000001569 carbon dioxide Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002986 genetic algorithm method Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000005416 organic matter Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- 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
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/048—Fuzzy inferencing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Genetics & Genomics (AREA)
- Physiology (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Automation & Control Theory (AREA)
- Fuzzy Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of anerobic sowages to handle gas production multi-element intelligent method for real-time monitoring, comprising: (1) constructs heredity-Fuzzy Wavelet Network;Determine the number of fuzzy rules of network;Network is modified and is improved;(2) it regard heredity-Fuzzy Wavelet Network establishment as program and by burning program in embedded Anaerobic Treatment monitoring system;(3) it is input to sewage data as training data in embedded Anaerobic Treatment monitoring system, trained network structure state is kept in systems;(4) trained embedded Anaerobic Treatment monitoring system access sewage treatment scene is subjected to on-line measurement, system is based on improved heredity-Fuzzy Wavelet Network and quickly obtains gas production and methane content;(5) primary every sampling in 30 minutes, it repeats step (4).The present invention intelligently monitors anerobic sowage processing system by Accurate Prediction gas production and methane content, promotes the efficient stable operation of anerobic sowage processing system.
Description
Technical field
The present invention relates to sewage treatment field more particularly to a kind of anerobic sowage processing gas production multi-element intelligent real time monitorings
Method.
Background technique
Anaerobic Methods in Treating is that under anaerobic, organic matter is converted methane by amphimicrobian and anaerobe group
With the process of carbon dioxide, it is a kind of not only energy conservation but also sewage treatment process of production capacity, not only can handle high concentrated organic wastewater,
And can handle the organic wastewater of middle low concentration, it is widely used in worldwide.
In order to improve anaerobic technique treatment effect, make gas production maximum while guaranteeing water quality, establishes Anaerobic Treatment prison
Control system is necessary.However, anerobic sowage processing system, which is one, is related to the multiple subjects such as chemistry, physics, biology
Complication system, there is complexity, time variation, non-linear and uncertain, so that showing between gas production and many parameters
Nonlinear relationship brings certain difficulty to the foundation for producing gas prediction modeling technique.Intelligent method has very strong study
Ability and adaptive ability, itself be it is nonlinear, can be simulated by learning reach required non-linear shape pair
As model, it is suitble to multi-input multi-output system, and intelligent method cannot use rule or public affairs to a large amount of initial data are handled
The problem of formula describes shows great flexibility and adaptivity, can be used for the mould of the changeable system for anaerobic treatment of dynamic
It is quasi-.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of anerobic sowages to handle gas production multi-element intelligent
Method for real-time monitoring.The present invention is based between the gas production of anerobic sowage processing system, influent quality and system control parameters
Correlation, by wavelet analysis, fuzzy theory, genetic algorithm and neural network organically combine, building GA-FWNN gas production it is pre-
Model is surveyed, and processing analysis is carried out to monitoring data by cloud computing storage platform, predicts that gas production (Biogas) and methane contain
Measure (CH4), anerobic sowage processing system is intelligently monitored, the operation of anerobic sowage processing system efficient stable is promoted.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of anerobic sowage processing gas production multi-element intelligent method for real-time monitoring, specific steps include:
(1) heredity-Fuzzy Wavelet Network is constructed;Determine that heredity-is fuzzy using adaptive fuzzy C means clustering algorithm
The number of fuzzy rules of wavelet neural network;Using heredity and gradient decline hybrid algorithm to heredity-Fuzzy Wavelet Network into
Row amendment and improvement;
(2) improved heredity-Fuzzy Wavelet Network is regard as system program by computer language establishment, and will
Burning program is in embedded Anaerobic Treatment monitoring system;
(3) it is input to sewage data as training data in embedded Anaerobic Treatment monitoring system, makes to improve in system
Heredity afterwards-Fuzzy Wavelet Network structure reaches neural network accuracy target, trained network structure state is maintained at is
In system;
(4) trained embedded Anaerobic Treatment monitoring system access sewage treatment scene is subjected to on-line measurement, passed through
The real-time collection site water sample testing data of sensor simultaneously enters data into system, and it is fuzzy that system is based on improved heredity-
Wavelet neural network quickly obtains gas production and methane content;
(5) primary every sampling in 30 minutes, it repeats step (4).
Specifically, in the present invention, when choosing water inlet chemical requirement (COD), organic loading rate (OLR), hydraulic retention
Between (HRT), pH value (pH) and carbonate total alkalinity (ALK) be used as input quantity, with gas production (Biogas) and methane content
(CH4) it is used as output quantity.
Specifically, in step (1), the heredity-Fuzzy Wavelet Network (GA-FWNN) constructed specifically includes five layers
Network structure, network inputs domain output between mapping relations be expressed as:
Wherein, woj(t) connection weight function of the wavelet network node layer to output layer, c are expressed asijAnd σijIt is expressed as
The center of subordinating degree function (Gaussian function) and width, xi(t) expression input node parameter, j=1,2,3...18, i=1,2,
3,4,5。
Further, the first layer of GA-FWNN structure is input layer, is responsible for receiving all input factors and is distributed to net
Next layer of network, there are five node, respectively COD (t), HRT (t), OLR (t), pH (t), ALK (t) for this layer.
Further, the second layer of GA-FWNN structure is blurring layer, this layer introduces Fuzzy Set Theory, makes network
Semantic variant can be handled.This layer is by being blurred each input, so that each node respectively corresponds a Vague language
Speech.In the present invention, Gaussian function is chosen as excitation function, and the output of each node indicates are as follows:
Wherein, cijAnd σijIt is expressed as in i-th of input variable Gaussian function relevant to j-th of fuzzy rule
The heart and width.Fuzzy division is carried out to network using adaptive fuzzy C means clustering algorithm, determines number of fuzzy rules.
Further, the third layer of GA-FWNN structure is fuzzy rule layer, and the number of nodes in this layer indicates fuzzy rule
With fuzzy segments, the corresponding fuzzy rule of each node.The operation of and is indicated using meeting, and realizes fuzzy reasoning function
Can, the output result of the fuzzy rule base generated by given data set, this layer indicates are as follows:
Wherein, n expression number of fuzzy rules, specially 18.
Further, the 4th layer of GA-FWNN structure is wavelet network layer (WNN layers), by three layers of wavelet neural network
Activation letter as the consequent part of fuzzy rule, using wavelet transformation local characteristics, using wavelet basis function as neuron
Number is embodied as j-th of wavelet neural member:
Wherein,aij、bijAnd wjRespectively indicate the translation of wavelet function, expansion factor and weight.
Further, the layer 5 of GA-FWNN structure is output layer, for calculating the output result of whole network.This
Layer is main to be considered to carry out the output result of each wavelet network the analysis of defuzzification as a result, output result is embodied as:
Wherein, ykThe output of FWNN, the i.e. gas production (Biogas) and methane content of forecasting system are indicated, for supervising in real time
Control the operating status of anaerobic system.
Specifically, in step (1), the fuzzy rule of GA-FWNN is determined using adaptive fuzzy C means clustering algorithm
Number, method particularly includes:
Fuzzy C-means clustering (FCM) is to obtain each sample point by optimization object function to be subordinate to all class centers
Degree, to determine the generic of sample point to achieve the purpose that automatically to classify to sample data.Assuming that sample set is X=
{x1,x2,…,xq, it is divided into c ambiguity group, makes the subordinated-degree matrix U=[u of objective function J (U, V)ij]c×qReach most
It is small, and obtain cluster centre V={ v1,v2,…,vc}。
dij=| | xj-vi|| (9)
Wherein, p expression parameter number, specially 5;Q indicates that sample is total, specially 120 groups;C indicates that cluster belongs to, h table
Show FUZZY WEIGHTED index, dijFor data xjTo cluster centre viEuclidean distance.
In order to improve segmentation quality, Validity Function B (c) is added in FCM, adaptive fuzzy C mean cluster is formed and calculates
Method, specific steps are as follows:
(1-2-1) gives iteration precision ε=0.001, k=0, c=2, and FUZZY WEIGHTED index h=2, B (1)=0 choose
[0,1] uniform random number on is as initial cluster center V(0);
(1-2-2) calculates the subordinated-degree matrix U of kth step(k), calculation formula are as follows:
(1-2-3) corrects cluster centre V(k+1), representation formula are as follows:
(1-2-4) if | | V(k+1)-V(k)| |≤ε, then iteration stopping, otherwise k=k+1, goes to step (1-2-2);
(1-2-5) calculates Validity Function B (c), in the case where c>2 and c<n, if B (c-1)>B (c-2) and B (c-1)>
B (c), then cluster process terminates, and otherwise sets c=c+1, goes to step (1-2-1).
WhereinIndicate the center vector of conceptual data sample.
Specifically, in step (1), FWNN is improved using heredity and gradient decline hybrid algorithm, specific steps
Are as follows:
The initialization procedure of (1-3-1) genetic algorithm realization network structure;
During initialization, parameter all in network is all optimized, former piece parameter (the degree of membership letter including rule
Several center cijAnd width csij) and consequent parameter (the translation a of wavelet functionij, expansion factor bijAnd weight wj)。
The present invention be using the standard error between network desired value and network reality output as the model fitness function,
Representation method are as follows:
Wherein, ydkIndicate desired output, ykIndicate the reality output of network, q indicates population at individual number, s in network
The output of a chromosome can be obtained by following formula:
Wherein,
Therefore s-th of chromosome is defined as:
Wherein
The initial parameter of FWNN in this way after three kinds of genetic manipulations (selection, intersect and variation) using can be obtained.Initially
Population number N pop is 100, crossover probability PcIt is 0.7, mutation probability PmIt is 200 for 0.01 and maximum number of iterations.
(1-3-2) gradient descent method realizes parameters revision process;
By the FWNN that genetic algorithm initializes, network parameter reaches near global optimum or approximate global optimum,
Gradient descent algorithm is recycled to adjust network parameter in real time, to obtain the parameter c of FWNNij、σij、aij、bijAnd wj.Ladder
The objective function for spending descent method indicates are as follows:
Wherein, yd(t) indicate that desired output, y (t) indicate current output.
Pass through objective function E and gradient descent method, the parameter c of FWNNij、σij、aij、bijAnd wjFollowing formula can be passed through
It obtains:
Wherein, learning rate η is 0.02, and factor of momentum ξ is 0.5.Formula (18)-(22) can be calculated by formula (10)~formula (13):
Wherein,
Specifically, above-mentioned GA-FWNN network structure, adaptive is worked out using VC and Matlab language joint in the present invention
Fuzzy C-Means Cluster Algorithm and the combination process for improving Multiple algorithm.
The present invention compared to the prior art, have it is below the utility model has the advantages that
1, the present invention combines the advantages of intelligent Theories such as wavelet analysis, fuzzy theory, genetic algorithm and neural network, building
GA-FWNN gas production prediction model, can accurately and rapidly realize gas production (Biogas) and methane content (CH4)
Line monitoring.
2, the present invention solves nerve using by genetic algorithm and gradient descent method hybrid algorithm Optimal Neural Network Architectures
Network falls into the problem of local optimum too early, has many advantages, such as that result lag time is short, error is small, strong environmental adaptability, energy
Enough so that system for anaerobic treatment is efficient, stablizes, economically runs.
Detailed description of the invention
Fig. 1 is heredity-Fuzzy Wavelet Network structure chart;
The anerobic sowage processing system schematic diagram that Fig. 2 is monitored in real time for the present invention;
Fig. 3 is heredity-Fuzzy Wavelet Network model training result figure;
Fig. 4 is heredity-Fuzzy Wavelet Network gas production prediction result figure.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
Embodiment
A kind of anerobic sowage processing gas production multi-element intelligent method for real-time monitoring, specific steps include:
(1) heredity-Fuzzy Wavelet Network is constructed;Determine that heredity-is fuzzy using adaptive fuzzy C means clustering algorithm
The number of fuzzy rules of wavelet neural network;Using heredity and gradient decline hybrid algorithm to heredity-Fuzzy Wavelet Network into
Row amendment and improvement.In the present invention, water inlet chemical requirement (COD), organic loading rate (OLR), hydraulic detention time are chosen
(HRT), pH value (pH) and carbonate total alkalinity (ALK) are used as input quantity, with gas production (Biogas) and methane content
(CH4) it is used as output quantity.
Specifically, in step (1), structure such as Fig. 1 institute of heredity-Fuzzy Wavelet Network (GA-FWNN) of construction
Show, specifically include five layer network structures, the mapping relations between the output of network inputs domain are expressed as:
Wherein, woj(t) connection weight function of the wavelet network node layer to output layer, c are expressed asijAnd σijIt is expressed as
The center of subordinating degree function (Gaussian function) and width, xi(t) expression input node parameter, j=1,2,3...18, i=1,2,
3,4,5。
Further, the first layer of GA-FWNN structure is input layer, is responsible for receiving all input factors and is distributed to net
Next layer of network, there are five node, respectively COD (t), HRT (t), OLR (t), pH (t), ALK (t) for this layer.
Further, the second layer of GA-FWNN structure is blurring layer, this layer introduces Fuzzy Set Theory, makes network
Semantic variant can be handled.This layer is by being blurred each input, so that each node respectively corresponds a Vague language
Speech.In the present invention, Gaussian function is chosen as excitation function, and the output of each node indicates are as follows:
Wherein, cijAnd σijIt is expressed as in i-th of the input variable Gaussian function first closed with j-th of fuzzy rule
The heart and width.Fuzzy division is carried out to network using adaptive fuzzy C means clustering algorithm, determines number of fuzzy rules.
Further, the third layer of GA-FWNN structure is fuzzy rule layer, and the number of nodes in this layer indicates fuzzy rule
With fuzzy segments, the corresponding fuzzy rule of each node.The operation of and is indicated using meeting, and realizes fuzzy reasoning function
Can, the output result of the fuzzy rule base generated by given data set, this layer indicates are as follows:
Wherein, n expression number of fuzzy rules, specially 18.
Further, the 4th layer of GA-FWNN structure is wavelet network layer (WNN layers), by three layers of wavelet neural network
Activation letter as the consequent part of fuzzy rule, using wavelet transformation local characteristics, using wavelet basis function as neuron
Number is embodied as j-th of wavelet neural member:
Wherein,aij、bijAnd wjRespectively indicate the translation of wavelet function, expansion factor and weight.
Further, the layer 5 of GA-FWNN structure is output layer, for calculating the output result of whole network.This
Layer is main to be considered to carry out the output result of each wavelet network the analysis of defuzzification as a result, output result is embodied as:
Wherein, ykIndicate the output of FWNN, the gas production (Biogas) and methane content of forecasting system, for monitoring in real time
The operating status of anaerobic system.
Specifically, in step (1), the fuzzy rule of GA-FWNN is determined using adaptive fuzzy C means clustering algorithm
Number, method particularly includes:
Fuzzy C-means clustering (FCM) is to obtain each sample point by optimization object function to be subordinate to all class centers
Degree, to determine the generic of sample point to achieve the purpose that automatically to classify to sample data.Assuming that sample set is X=
{x1,x2,…,xq, it is divided into c ambiguity group, makes the subordinated-degree matrix U=[u of objective function J (U, V)ij]c×qReach most
It is small, and obtain cluster centre V={ v1,v2,…,vc}。
dij=| | xj-vi|| (9)
Wherein, p expression parameter number, specially 5;Q indicates that sample is total, specially 120 groups;C indicates that cluster belongs to, h table
Show FUZZY WEIGHTED index, dijFor data xjTo cluster centre viEuclidean distance.
In order to improve segmentation quality, Validity Function B (c) is added in FCM, adaptive fuzzy C mean cluster is formed and calculates
Method, specific steps are as follows:
(1-2-1) gives iteration precision ε=0.001, k=0, c=2, and FUZZY WEIGHTED index h=2, B (1)=0 choose
[0,1] uniform random number on is as initial cluster center V(0);
(1-2-2) calculates the subordinated-degree matrix U of kth step(k), calculation formula are as follows:
(1-2-3) corrects cluster centre V(k+1), representation formula are as follows:
(1-2-4) if | | V(k+1)-V(k)| |≤ε, then iteration stopping, otherwise k=k+1, goes to step (1-2-2);
(1-2-5) calculates Validity Function B (c), in the case where c>2 and c<n, if B (c-1)>B (c-2) and B (c-1)>
B (c), then cluster process terminates, and otherwise sets c=c+1, goes to step (1-2-1).
WhereinIndicate the center vector of conceptual data sample.
Specifically, in step (1), FWNN is improved using heredity and gradient decline hybrid algorithm, specific steps
Are as follows:
The initialization procedure of (1-3-1) genetic algorithm realization network structure;
During initialization, parameter all in network is all optimized, former piece parameter (the degree of membership letter including rule
Several center cijAnd width csij) and consequent parameter (the translation a of wavelet functionij, expansion factor bijAnd weight wj).In this implementation
In example, occurrence of the former piece parameter and consequent parameter of GA-FWNN model when number of fuzzy rules is 18 is respectively such as Tables 1 and 2
It is shown.
The former piece parameter of table 1.GA-FWNN model
The consequent parameter of table 2.GA-FWNN model
The present invention be using the standard error between network desired value and network reality output as the model fitness function,
Representation method are as follows:
Wherein, ydkIndicate desired output, ykIndicate the reality output of network, q indicates population at individual number, s in network
The output of a chromosome can be obtained by following formula:
Wherein,
Therefore s-th of chromosome is defined as:
Wherein
The initial parameter of FWNN in this way after three kinds of genetic manipulations (selection, intersect and variation) using can be obtained.Initially
Population number N pop is 100, crossover probability PcIt is 0.7, mutation probability PmIt is 200 for 0.01 and maximum number of iterations.
(1-3-2) gradient descent method realizes parameters revision process;
By the FWNN that genetic algorithm initializes, network parameter reaches near global optimum or approximate global optimum,
Gradient descent algorithm is recycled to adjust network parameter in real time, to obtain the parameter c of FWNNij、σij、aij、bijAnd wj.Ladder
The objective function for spending descent method indicates are as follows:
Wherein, yd(t) indicate that desired output, y (t) indicate current output.
Pass through objective function E and gradient descent method, the parameter c of FWNNij、σij、aij、bijAnd wjFollowing formula can be passed through
It obtains:
Wherein, learning rate η is 0.02, and factor of momentum ξ is 0.5.
Formula (18)-(22) can be calculated by formula (10)~formula (13):
Wherein,
(2) improved heredity-Fuzzy Wavelet Network is regard as system program by the establishment of VC and Matlab language,
And by burning program in embedded Anaerobic Treatment monitoring system;
(3) the sewage data in anerobic sowage processing system embedded Anaerobic Treatment is input to as training data to monitor
In system, improved heredity-Fuzzy Wavelet Network structure in system is set to reach neural network accuracy target, by trained net
Network configuration state is kept in systems;Anerobic sowage processing system is as shown in Figure 2;Heredity-Fuzzy Wavelet Network model
Training result is as shown in Figure 3.
(4) trained embedded Anaerobic Treatment monitoring system access sewage treatment scene is subjected to on-line measurement, passed through
The real-time collection site water sample testing data of sensor simultaneously enters data into system, and it is fuzzy that system is based on improved heredity-
Wavelet neural network quickly obtains gas production and methane content;
(5) primary every sampling in 30 minutes, it repeats step (4).Heredity-fuzzy neural network model gas production prediction knot
Fruit figure is as shown in Figure 4.
From fig. 4, it can be seen that model curve of output tracks reality output curve well, sample pattern output and reality are defeated
Error out is very small, coefficient R and R2Greatly, illustrate GA-FWNN method can be realized to Anaerobic Treatment gas production it is quick,
Accurate monitoring.
The present invention is compared with fuzzy neural network (FNN), wavelet neural network (WNN) and neural network (NN), knot
Fruit is as shown in table 3.It is analyzed by calculating, estimated performance of the invention better than than other three kinds of models, compares FNN, WNN and NN
Model, the square error (RMSE) of GA-FWNN model prediction, mean absolute percentage error (MAPE) and mean square error MSE are
Minimum, while coefficient R and R2It is all maximum.Therefore using GA-FWNN as Anaerobic Treatment gas production is simulated is one good
Method.
The various model performances of table 3 compare
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (5)
1. a kind of anerobic sowage handles gas production multi-element intelligent method for real-time monitoring, which is characterized in that specific steps include:
(1) heredity-Fuzzy Wavelet Network is constructed;Heredity-fuzzy wavelet is determined using adaptive fuzzy C means clustering algorithm
The number of fuzzy rules of neural network;Heredity-Fuzzy Wavelet Network is repaired using heredity and gradient decline hybrid algorithm
Just and improve;
(2) improved heredity-Fuzzy Wavelet Network is regard as system program by computer language establishment, and by program
Burning is in embedded Anaerobic Treatment monitoring system;
(3) it is input in embedded Anaerobic Treatment monitoring system, makes improved in system using sewage data as training data
Heredity-Fuzzy Wavelet Network structure reaches neural network accuracy target, and trained network structure state is kept in systems;
(4) trained embedded Anaerobic Treatment monitoring system access sewage treatment scene is subjected to on-line measurement, passes through sensing
The real-time collection site water sample testing data of device simultaneously enters data into system, and system is based on improved heredity-fuzzy wavelet
Neural network quickly obtains gas production and methane content;
(5) primary every sampling in 30 minutes, it repeats step (4).
2. a kind of anerobic sowage according to claim 1 handles gas production multi-element intelligent method for real-time monitoring, feature exists
In in step (1), the heredity-Fuzzy Wavelet Network constructed specifically includes five layer network structures, and network inputs domain is defeated
Mapping relations between out are expressed as:
Wherein, woj(t) connection weight function of the wavelet network node layer to output layer, c are expressed asijAnd σijIt is expressed as being subordinate to
Spend center and the width of function, xi(t) expression input node parameter, j=1,2,3...18, i=1,2,3,4,5;N indicates fuzzy
Regular number.
3. a kind of anerobic sowage according to claim 2 handles gas production multi-element intelligent method for real-time monitoring, feature exists
In the first layer of GA-FWNN structure is input layer, is responsible for receiving all input factors and is distributed to next layer of network, this layer
There are five node, respectively COD (t), HRT (t), OLR (t), pH (t), ALK (t);
The second layer of GA-FWNN structure is blurring layer, this layer introduces Fuzzy Set Theory, network is enable to handle semantic variant;
This layer is by being blurred each input, so that each node respectively corresponds a fuzzy language;In the present invention, it selects
Take Gaussian function as excitation function, the output of each node indicates are as follows:
Wherein, cijAnd σijIt is expressed as center and the width of i-th of the input variable Gaussian function first closed with j-th of fuzzy rule
Degree, xiIt indicates input node parameter, fuzzy division is carried out to network using adaptive fuzzy C means clustering algorithm, determines network
Number of fuzzy rules;
The third layer of GA-FWNN structure is fuzzy rule layer, and the number of nodes in this layer indicates fuzzy rule and fuzzy segments, often
The corresponding fuzzy rule of a node;The operation of and is indicated using meeting, realizes fuzzy reasoning function, by given data
Collect the fuzzy rule base generated, the output result of this layer indicates are as follows:
Wherein, n expression number of fuzzy rules, specially 18;M expression input number of nodes, specially 5;
The 4th layer of GA-FWNN structure be wavelet network layer, the consequent part by three layers of wavelet neural network as fuzzy rule,
Using the good characteristic of wavelet transformation local characteristics, activation primitive using wavelet function as neuron, for j-th of small echo mind
Through member, it is embodied as:
Wherein,aij、bijAnd wjRespectively indicate the translation of wavelet function, expansion factor and weight;
The layer 5 of GA-FWNN structure is output layer, for calculating the output result of whole network;This layer mainly considers to each
The output result of wavelet network carries out the analysis of defuzzification as a result, output result is embodied as:
Wherein, ykIndicate the output of FWNN, the gas production and methane content of forecasting system, for monitoring the fortune of anaerobic system in real time
Row state.
4. a kind of anerobic sowage according to claim 1 handles gas production multi-element intelligent method for real-time monitoring, feature exists
In determining the number of fuzzy rules of GA-FWNN, specific method using adaptive fuzzy C means clustering algorithm in step (1)
Are as follows:
Fuzzy C-means clustering is that each sample point is obtained by optimization object function to the degree of membership at all class centers, thus certainly
The generic of sample point is determined to achieve the purpose that automatically to classify to sample data;Assuming that sample set is X={ x1,x2,…,
xq, it is divided into c ambiguity group, makes the subordinated-degree matrix U=[u of objective function J (U, V)ij]c×qReach minimum, and is gathered
Class center V={ v1,v2,…,vc};
dij=| | xj-vi|| (9)
Wherein, p expression parameter number, specially 5;Q indicates that sample is total, specially 120 groups;C indicates that cluster belongs to, and h indicates mould
Paste Weighted Index, dijFor data xjTo cluster centre viEuclidean distance;
In order to improve segmentation quality, Validity Function B (c) is added in FCM, forms adaptive fuzzy C means clustering algorithm, tool
Body step are as follows:
(1-2-1) gives iteration precision ε=0.001, k=0, c=2, and FUZZY WEIGHTED index h=2, B (1)=0 choose [0,1]
On uniform random number as initial cluster center V(0);
(1-2-2) calculate kth step subordinated-degree matrix U (k), calculation formula are as follows:
(1-2-3) amendment cluster centre V (k+1), representation formula are as follows:
(1-2-4) if | | V(k+1)-V(k)| |≤ε, then iteration stopping, otherwise k=k+1, goes to step (1-2-2);
(1-2-5) calculates Validity Function B (c), in the case where c>2 and c<n, if B (c-1)>B (c-2) and B (c-1)>B
(c), then cluster process terminates, and otherwise sets c=c+1, goes to step (1-2-1);
WhereinIndicate the center vector of conceptual data sample.
5. a kind of anerobic sowage according to claim 1 handles gas production multi-element intelligent method for real-time monitoring, feature exists
In, in step (1), hybrid algorithm is declined using heredity and gradient, FWNN is improved, specific steps are as follows:
The initialization procedure of (1-3-1) genetic algorithm realization network structure;
During initialization, parameter all in network is all optimized, the former piece parameter including rule, i.e. subordinating degree function
Center cijAnd width csijAnd consequent parameter, i.e. the translation a of wavelet functionij, expansion factor bijAnd weight wj;
The present invention is indicated using the standard error between network desired value and network reality output as the model fitness function
Method are as follows:
Wherein, ydkIndicate desired output, ykIndicate the reality output of network, q indicates population at individual number, s-th of dye in network
The output of colour solid can be obtained by following formula:
Wherein,
Therefore s-th of chromosome is defined as:
Wherein
The initial parameter of FWNN in this way after three kinds of genetic manipulations using can be obtained;Initial population number N pop is 100, is intersected
Probability PcIt is 0.7, mutation probability PmIt is 200 for 0.01 and maximum number of iterations;
(1-3-2) gradient descent method realizes parameters revision process;
By the FWNN that genetic algorithm initializes, network parameter reaches near global optimum or approximate global optimum, then benefit
Network parameter is adjusted in real time with gradient descent algorithm, to obtain the parameter c of FWNNij、σij、aij、bijAnd wj;Under gradient
The objective function of drop method indicates are as follows:
Wherein, yd(t) indicate that desired output, y (t) indicate current output;
Pass through objective function E and gradient descent method, the parameter c of FWNNij、σij、aij、bijAnd wjIt can be obtained by following formula:
Wherein, learning rate η is 0.02, and factor of momentum ξ is 0.5;
Formula (18)-(22) can be calculated by formula (10)~formula (13):
Wherein,
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811130613.1A CN109408896B (en) | 2018-09-27 | 2018-09-27 | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811130613.1A CN109408896B (en) | 2018-09-27 | 2018-09-27 | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109408896A true CN109408896A (en) | 2019-03-01 |
CN109408896B CN109408896B (en) | 2024-01-05 |
Family
ID=65466467
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811130613.1A Active CN109408896B (en) | 2018-09-27 | 2018-09-27 | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109408896B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110824915A (en) * | 2019-09-30 | 2020-02-21 | 华南师范大学 | GA-DBN network-based intelligent monitoring method and system for wastewater treatment |
CN111798134A (en) * | 2020-07-06 | 2020-10-20 | 青岛洪锦智慧能源技术有限公司 | Method for improving methane yield of sewage treatment plant based on data-driven model |
CN116084892A (en) * | 2023-02-14 | 2023-05-09 | 电子科技大学 | Automatic perforation system based on fuzzy neural network |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102262147A (en) * | 2011-07-15 | 2011-11-30 | 华南理工大学 | Soft measurement method and system for effluent chemical oxygen demand (COD) of waste water treatment system |
CN102663493A (en) * | 2012-03-23 | 2012-09-12 | 天津工业大学 | Delaying nerve network used for time sequence prediction |
CN106529818A (en) * | 2016-11-16 | 2017-03-22 | 河南工程学院 | Water quality evaluation prediction method based on fuzzy wavelet neural network |
CN107132325A (en) * | 2017-04-14 | 2017-09-05 | 华南理工大学 | A kind of flexible measurement method of the Anaerobic Waste Treatment System water outlet volatile fatty acid based on particle cluster algorithm and SVMs |
CN108088974A (en) * | 2017-11-30 | 2018-05-29 | 华南理工大学 | A kind of flexible measurement method of anaerobism while denitrification methane phase process water outlet nitrate nitrogen |
-
2018
- 2018-09-27 CN CN201811130613.1A patent/CN109408896B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102262147A (en) * | 2011-07-15 | 2011-11-30 | 华南理工大学 | Soft measurement method and system for effluent chemical oxygen demand (COD) of waste water treatment system |
CN102663493A (en) * | 2012-03-23 | 2012-09-12 | 天津工业大学 | Delaying nerve network used for time sequence prediction |
CN106529818A (en) * | 2016-11-16 | 2017-03-22 | 河南工程学院 | Water quality evaluation prediction method based on fuzzy wavelet neural network |
CN107132325A (en) * | 2017-04-14 | 2017-09-05 | 华南理工大学 | A kind of flexible measurement method of the Anaerobic Waste Treatment System water outlet volatile fatty acid based on particle cluster algorithm and SVMs |
CN108088974A (en) * | 2017-11-30 | 2018-05-29 | 华南理工大学 | A kind of flexible measurement method of anaerobism while denitrification methane phase process water outlet nitrate nitrogen |
Non-Patent Citations (1)
Title |
---|
RAHIB HIDAYAT ABIYEV: "Fuzzy Wavelet Neural Networks for Identification and Control of Dynamic Plants—A Novel Structure and a Comparative Study", 《IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110824915A (en) * | 2019-09-30 | 2020-02-21 | 华南师范大学 | GA-DBN network-based intelligent monitoring method and system for wastewater treatment |
CN110824915B (en) * | 2019-09-30 | 2022-06-07 | 华南师范大学 | GA-DBN network-based intelligent monitoring method and system for wastewater treatment |
CN111798134A (en) * | 2020-07-06 | 2020-10-20 | 青岛洪锦智慧能源技术有限公司 | Method for improving methane yield of sewage treatment plant based on data-driven model |
CN111798134B (en) * | 2020-07-06 | 2022-04-05 | 青岛洪锦智慧能源技术有限公司 | Method for improving methane yield of sewage treatment plant based on data-driven model |
CN116084892A (en) * | 2023-02-14 | 2023-05-09 | 电子科技大学 | Automatic perforation system based on fuzzy neural network |
CN116084892B (en) * | 2023-02-14 | 2024-04-23 | 电子科技大学 | Automatic perforation system based on fuzzy neural network |
Also Published As
Publication number | Publication date |
---|---|
CN109408896B (en) | 2024-01-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109492822B (en) | Air pollutant concentration time-space domain correlation prediction method | |
CN111832814B (en) | Air pollutant concentration prediction method based on graph attention mechanism | |
Asadi et al. | A new hybrid artificial neural networks for rainfall–runoff process modeling | |
Firat et al. | River flow estimation using adaptive neuro fuzzy inference system | |
Kalogirou et al. | Modeling of solar domestic water heating systems using artificial neural networks | |
Nikoo et al. | Flood-routing modeling with neural network optimized by social-based algorithm | |
Chen et al. | Deformation prediction of landslide based on improved back-propagation neural network | |
CN106529818B (en) | Water quality assessment Forecasting Methodology based on Fuzzy Wavelet Network | |
CN109146162B (en) | A kind of probability wind speed forecasting method based on integrated Recognition with Recurrent Neural Network | |
Gill et al. | Training back propagation neural networks with genetic algorithm for weather forecasting | |
CN108009674A (en) | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks | |
Solaimani | Rainfall-runoff prediction based on artificial neural network (a case study: Jarahi watershed) | |
CN111160520A (en) | BP neural network wind speed prediction method based on genetic algorithm optimization | |
Jiang et al. | Research and application of an innovative combined model based on a modified optimization algorithm for wind speed forecasting | |
CN107506590A (en) | A kind of angiocardiopathy forecast model based on improvement depth belief network | |
CN106022954B (en) | Multiple BP neural network load prediction method based on grey correlation degree | |
CN109408896A (en) | A kind of anerobic sowage processing gas production multi-element intelligent method for real-time monitoring | |
CN111242380A (en) | Lake (reservoir) eutrophication prediction method based on artificial intelligence algorithm | |
CN109934422A (en) | Neural network wind speed prediction method based on time series data analysis | |
Chidthong et al. | Developing a hybrid multi‐model for peak flood forecasting | |
CN107818340A (en) | Two-stage Air-conditioning Load Prediction method based on K value wavelet neural networks | |
Li et al. | A method of rainfall runoff forecasting based on deep convolution neural networks | |
CN116542382A (en) | Sewage treatment dissolved oxygen concentration prediction method based on mixed optimization algorithm | |
CN114580762A (en) | Hydrological forecast error correction method based on XGboost | |
CN110929809B (en) | Sewage key water quality index soft measurement method of characteristic self-enhanced cyclic neural network |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |