CN107132325B - A kind of flexible measurement method based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid - Google Patents
A kind of flexible measurement method based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid Download PDFInfo
- Publication number
- CN107132325B CN107132325B CN201710250802.1A CN201710250802A CN107132325B CN 107132325 B CN107132325 B CN 107132325B CN 201710250802 A CN201710250802 A CN 201710250802A CN 107132325 B CN107132325 B CN 107132325B
- Authority
- CN
- China
- Prior art keywords
- data
- water outlet
- fatty acid
- model
- volatile fatty
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/18—Water
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Chemical & Material Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Analytical Chemistry (AREA)
- Medicinal Chemistry (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Immunology (AREA)
- Biochemistry (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Pathology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Food Science & Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Purification Treatments By Anaerobic Or Anaerobic And Aerobic Bacteria Or Animals (AREA)
Abstract
The invention discloses a kind of flexible measurement methods based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid.In order to improve the accuracy and robustness of model, which combines particle swarm algorithm and support vector machines.Flexible measurement method of the invention solves the problems, such as that measuring device lag time long, it really realizes and wastewater effluent quality is monitored in real time comprehensively, prevent pop-up threat, it can be used for the optimal control that feedback control realizes anaerobic wastewater treatment system simultaneously, for monitoring, optimization and understand that anaerobic digestion process provides guidance, it can be used for replacing the expensive measuring device in part, maintenance cost saving, reduce cost for wastewater treatment, it is easy to promote and apply in waste water treatment engineering, there is good Social benefit and economic benefit.
Description
Technical field
The present invention relates to technical field of waste water processing, and in particular to a kind of anaerobic wastewater treatment system water outlet volatile fat
The flexible measurement method of acid.
Background technique
With the quickening of China's economic development and process of industrialization, the wastewater discharge in China increases year by year, water pollution control
System shoulders heavy responsibilities.Anaerobic treatment technique have power consumption is small, sludge yield is few, low to the requirement of nitrogen and phosphorus, can
The advantages that reduction supplement nitrogen and the expense of phosphorus nutrition, anaerobic digestion can produce biological energy source, so that anaerobic bio-treated is in Industry Waste
More and more important in water process, utilization is also more and more extensive.However, anaerobic digestion process is an extremely complex, non-linear, pole
Has uncertain biochemical process.On the one hand, anaerobe to including trichlorophenol, 2,4,6,-T etc. toxic organic compounds it is more sensitive, if for
Toxic wastewater property understands insufficient or misoperation, may cause the deterioration of reactor service condition in severe case；It is another
The effect of aspect, anaerobic bio-treated is easy the disturbance changed by operating condition, anaerobe (especially methanogen)
Environmental condition is required harsher.When influent load increases suddenly or water temperature, basicity are too low, Hydrolysis Acidification and methane
Change process is easy unbalance, largely accumulates so as to cause organic acid, anaerobic methane is suppressed even anaerobic system collapse.
The characteristics of complexity due to anaerobic System and the disturbance vulnerable to service condition, the monitoring of anaerobic digestion process
Analysis with water quality parameter is particularly important for the stability, the high efficiency that guarantee Anaerobic treatment technique.However, big portion, China
Manually to determine these parameters, artificial chemical examination is not only time-consuming to take for chemical examination mostly during handling waste water for point waste water treatment plant
Power also consumes a large amount of chemicals, increases the discharge amount of pollutant, has seriously affected the control effect of waste water treatment process.
Although occurring can be with the analysis meter of these waste water qualities of on-line measurement, these are specially due to the development of instrumental technique in recent years
Developed in China than later with instrument, usually selection external product；Meanwhile the generally existing expensive, equipment of meter specially is thrown
The problems such as money and operation expense are high, detection lag time is long, stability is bad and poor repeatability, to reduce important ginseng
Several Control platforms, this makes its application receive further limitation.
In recent years, soft-measuring technique the answering in wastewater treatment process based on artificial neural network, mathematical statistics etc.
With more and more extensively.Soft-sensing model is suitable for nonlinear system modeling, can mention significantly using fast parallel Processing Algorithm
Height identification speed.Method for system modelling is very simple, only need to output and input data by system, make system for water
The frequent biological wastewater treatment of qualitative changeization still has relatively good precision.In recent years, researcher is to soft sensor modeling
Technology and its use research and practice, especially nearest one or two years intelligent algorithm and each field utilization gradually at
For research hotspot, the hard measurement system constructed by soft-measuring technique can gradually be used to replace conventional hardware instrument, can also be with
It is used to ensure that the accuracy of measurement simultaneously with hardware instrument.
Support vector machines (Support Vector Machine, SVM) is to attract attention in machine learning field in recent years
A kind of more soft sensor modeling new technology, SVM is based on Principle of Statistics, compared to the discovery learning mechanism of neural network, SVM
Experience ingredient it is very few, have more stringent mathematic(al) argument.Meanwhile SVM for provided sample data dependence compared with
It is few, and generalization ability is stronger, locally optimal solution must be globally optimal solution, generation dimension disaster be avoided, especially suitable for solution
Certainly the problems such as small sample, non-linear, local minimum point.Currently, being directed to the detection of volatile fatty acid, researcher is proposed very
Multi-method, the flexible measurement method of the anaerobic wastewater treatment system water outlet volatile fatty acid based on support vector machines also have relevant
Document and patent report.Such as Liu Bo proposes a kind of anaerobic wastewater treatment system water outlet VFA on-line prediction based on PCA-LSSVM
Model, simulation result show that the model has good simulation capacity (Liu Bo, Wan Jinquan, the base such as yellow wisdom under Stable State Environment
VFA on-line prediction model [J] ACTA Scientiae Circumstantiae, 35 (6): 1768- are discharged in the anaerobic wastewater treatment system of PCA-LSSVM
1778.).However, from these documents it is found that model is imitative under unsteady condition after unstable state data are added in metadata set
True performance will receive interference.
Summary of the invention
The object of the present invention is to provide a kind of hard measurement sides of Anaerobic Waste Treatment System water outlet volatile fatty acid (VFA)
Method, in order to improve the accuracy and robustness of model, which combines particle swarm algorithm and support vector machines.
Particle swarm algorithm (Particle Swarm Optimization, PSO) is for excellent to selecting for model parameter.
The problems such as support vector machines (SVM) is especially suitable for solving small sample, non-linear, local minimum point, and training speed
Degree is fast, and has effective classification policy to the metadata set of addition；SVM model be Statistical Learning Theory VC dimension it is theoretical and
The new intelligent algorithm of the one kind developed on Structural risk minization principle, is by Corinna Cortes and Vapnik et al.
(Cortes C, Vapnik V.1995.Support-Vector networks [J] .Machine Learning, 20 (3):
273-297) put forward first, basic thought is exactly the concept for introducing kernel function, and the inseparable problem of lower dimensional space is passed through
Kernel function is mapped to higher dimensional space linear separability, does classification or recurrence processing in higher dimensional space.
SVM is initially that academia puts forward for the classification problem of two data categories, and classification problem can mathematical expression
Are as follows: for sample set (xi, yi), i=1,2 ..., n, xn∈Rn, yn∈ { -1,1 }, structural classification face WX+b=0, the classifying face
Can be faultless separated by two class samples, and make the distance between two classes maximum；W, X is n-dimensional vector, linear discriminant function
General type is g (x)=Wx+b, it is normalized with times scaling W, a b, meets the sample nearest from classifying face | g
(x) |=1, then all samples of two classes all meet | and g (x) | >=1, wherein the class interval of two class samples is 2/ | | W | |.
Then the SVM mathematical problem to be solved is: under conditions of meeting formula (3), seek the minimum value of formula (4):
yi(wTxi+ b) -1 >=0, i=1,2,3 ... n (3)；
Define lagrange function, wherein aiFor lagrange coefficient, ai≥0；Problem, which is converted into, asks w and b
Lagrange functional minimum value.
Respectively to w, b, aiPartial differential is sought, when being equal to zero, is obtained:
Formula (6) and formula (3) constitute the constraint condition for former problem being converted into the dual problem of following convex quadratic programming:
Former problem forms the quadratic function mechanism problem under inequality constraints, existence anduniquess optimal solution；IfFor optimal solution,
ThenWhereinThe sample being not zero is supporting vector；Therefore, the weight coefficient of optimal classification surface to
Amount is the linear combination of supporting vector；
B* is solved when being equal to zero by constraint equation (3), and the optimal classification function thus acquired is:
Finally, SVM, which solves regression problem, can be described as following optimization problem:
In formula (9),For the training error of i-th of sample point,For empiric risk,Measure machine
The complexity of study；γ > 0 is penalty factor, also known as regularization parameter, to balancing machine learns in training complexity and
Empiric risk；Formula (9) should meet constraint condition:
yi=wTΦ(xi)+b+ξi, i=1,2,3 ..., n (10)；
Lagrange function is introduced in formula (10):
In formula (11), aiIt is Lagrange multiplier, utilizes Karush-Kuhn-Tucker ' s (KKT) optimal condition to formula
(11) it optimizes, to w, b, ξ, aiAsk local derviation that can obtain:
The problem of elimination w, ξ, formula (11) optimizes, can be converted into following linear equation Solve problems:
In formula (13), by k (xi,xj) it is defined as kernel function, it solves equation (13) and acquires ξ, aiValue, the regression model of acquisition,
Expression are as follows:
The present invention is used as kernel function using Radial basis kernel function (RBF, Radial Basis Functions), it may be assumed that
The classifier obtained by formula (15) and the important difference of tradition RBF method are that each Basis Function Center is one corresponding
Supporting vector, output weight are generally automatically determined by algorithm, and σ is core width in formula.
Both particle swarm algorithm (PSO) and support vector machines (SVM) are implemented in combination with anaerobic wastewater treatment system volatility rouge
The hard measurement of fat acid, on-line monitoring and subsequent technique optimal control to anaerobic wastewater treatment system are of great significance, simultaneously
To anaerobic wastewater treatment system, other are difficult to the index monitored on-line with directive significance.
In order to achieve the object of the present invention, technical scheme is as follows.
It is a kind of that the soft of volatile fatty acid is discharged based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines
Measurement method includes the following steps:
(1) anaerobic wastewater treatment system is built, obtains and collects and corresponding under the conditions of input quantity and different input quantity detest
The metadata of oxygen reactor output quantity；
(2) obtained metadata is divided into training set pTraining={ x；Y } and test set and pTest={ x '；Y ' }, and by data
Collect p={ x x '；Y y ' } do normalized；Training set data is used for SVM-regression model modeling, is trained
Training set is divided into two training according to the relative error of model output Y and training set actual value y by the model output Y of collection
Collection, i.e. training set 1. with training set 2. namely pTraining 1={ x1；y1And pTraining 2={ x2；y2, and mark training set 1. and training set
2. metadata tag be respectively 1 and -1；
(3) according to the training set of classification 1. with training set 2., using SVM-classification model by test set point
At two corresponding test sets, i.e., test set 1. with test set 2. namely pTest 1={ x '1；y'1And pTest 2={ x '2；y'2, from
And metadata is divided into two group data sets, i.e., data set 1. with data set 2. namely p1={ x1 x’1；y1 y’1And p2={ x2
x’2；y2 y’2}；
(4) 1. 2. obtained training set SVM-regression model is utilized respectively with training set to handle, and point
The validity of model is not 1. 2. verified with test set with corresponding test set；When the precision of model reaches sets requirement, wrapped
The water outlet volatile fatty acid hard measurement of the PSO-SVM of model containing SVM-classification and SVM-regression model
Model；
(5) the water outlet volatile fatty acid soft-sensing model of the PSO-SVM after training and test is embedded in industrial personal computer, and
Human-computer interaction interface is built using the configuration software in industrial personal computer；
(6) the water outlet volatility rouge based on PSO-SVM is realized using OPC (OLE for Process Control) technology
The data of collection are delivered to going out for PSO-SVM by the data exchange between the aching and limp measurement model of fat and the configuration software of industrial personal computer
Water volatile fatty acid soft-sensing model, calculates the predicted value of water outlet Vfa Concentration, then the predicted value is back to
Industrial personal computer human-computer interaction interface is shown；
(7) step (6) constantly are repeated, to realize the online real-time of anaerobic wastewater treatment system water outlet volatile fatty acid
Monitoring；
(8) system is exported and is showed with report form, and provide the characteristic index index of characterization system.
Further, the particle swarm algorithm is the algorithm of iteration pattern, mathematical description are as follows:
It is located in a n dimension search space, population x={ x1,x2,……,xNIt is to be made of N number of particle, wherein i-th
Current location locating for particle is x1={ xi1,xi2,……,xinT, speed v1={ vi1,vi2,……,vinT, the particle
Individual extrema representation is P1={ Pi1,Pi2,……,PinT, the global extremum of entire population is expressed as Pg={ Pg1,Pg2,……,
PgnT, according to the principle of the continuous optimizing of particle, the particle x of particle swarm algorithmjSpeed and location update formula it is as follows:
In formula, Xj is particle vector position, and Vj is the speed of particle, and w is weighted value, c1、c2For aceleration pulse；rand1、
rand2It is random function, effect is the random number in order to generate (0,1)；Pbest is individual extreme value, and gbest is global extremum；k
Indicate the number of iterations；J indicates vector dimension.
Further, in step (1), the anaerobic wastewater treatment system includes that anaerobic reactor and anaerobic effluent volatilize
Property fatty acid it is relevant monitoring and detection instrument, industrial personal computer, D/A and A/D module.
Further, in step (1), the input quantity includes water inlet organic loading (COD), water outlet pH, leaving water temperature
(T), the amount of gas and the oxidation-reduction potential (ORP) of component, influent alkalinity and anaerobic reactor are produced.
Further, in step (1), the component and yield of the production gas include the yield of methane and carbon dioxide.
Further, in step (1), the output quantity includes being discharged the concentration of general volatile fatty acid.
Further, in step (2), the normalized are as follows:
In formula, S (i) is any group of data in data set；Min (S) is to be worth the smallest one group of data in data set；max
It (S) is to be worth maximum one group of data in data set.
Further, in step (2), the relative error of model output Y and training set actual value y are instruction lower than 50%
Practice collection 1.；The relative error of model output Y and training set actual value y be greater than or equal to 50% be training set 2..
Further, in step (3), the SVM-classification model are as follows:
In formula, x is auxiliary variable sample data,For Lagrange multiplier, b*For offset, w*For optimal weights value, T
Represent transposition, xi *For optimal sample data, the natural number that n is >=1；(xi,yj) it is training set sample, xn∈Rn, yn∈ -1,
1}。
Further, step (2), in (4), the SVM-regression model are as follows:
In formula, x is auxiliary variable sample data, aiFor Lagrange multiplier, b is offset, the natural number that n is >=1；k
(xi,xj) be kernel function and meet Mercer condition.
Further, in step (4), the required precision of the model is by user's self-setting as needed.
Further, in step (8), the characteristic index index includes related coefficient (R), mean absolute percentage error
(MAPE), relative error (RE) and root-mean-square error (MSRE).
The mathematical expression of characteristic index index is as follows:
Related coefficient (correlation coefficient, R) reflects the strong of predicted value and actual value linear relationship
Weak, R is closer to 1, then predicted value and actual value are closer.
Mean absolute percentage error (Mean Absolute Percent Error, MAPE) is all relative errors
The average value of absolute value summation, can preferably reflect the actual conditions of predicted value on the whole；
Relative error (Relative Error, RE) indicates the ratio between the true value of absolute error value and tested magnitude, relatively
Error can more reflect the degree of reliability of prediction；
Root-mean-square error (Root Mean Square Error, RMSE), observation and true value deviation square with observation
The square root of frequency n ratio, RMSE is primarily to illustrate the dispersion degree of sample.
The value of RMSE is smaller, illustrates that prediction model describes experimental data with better levels of precision, conversely, model prediction
Precision is poor；
In above R, MAPE, RE, RMSE formula, y is actual value, y*For actual value mean value, ypFor predicted value,For actual value
Mean value, m are sample size.
Further, by transmission control protocol, internet protocol negotiation serial data standard, pass through computer and two-way
Letter monitors measurement output data in real time, and is shown by the monitoring window of configuration software, so that system is in time, accurately
Ground understands the index situation of change of sewage disposal system, promotes the operation of sewage treatment plant's efficient stable.
Compared with prior art, the invention has the advantages that and the utility model has the advantages that
(1) the present invention provides new modeling method, introduce classification policy handle data volume it is big caused by model
Degradation problem has preferable simulation capacity according to experimental applications in reality, can be used for during Anaerobic wastewater treatment going out
The real-time detection of water VFA concentration；
(2) modeling method of the invention meets preferably to anaerobic effluent volatile fatty acid predicted value and truthful data,
Error is relatively also smaller, fully meets practical application；
(3) flexible measurement method of the invention solves the problems, such as that measuring device lag time long, really realizes waste water
Water water quality monitors in real time comprehensively, prevents pop-up threat, while can be used for feedback control and realizing anaerobic wastewater treatment system
Optimal control, for monitoring, optimization and understand that anaerobic digestion process provides guidance；
(4) present invention can be used for replace the expensive measuring device in part, maintenance cost saving, reduce wastewater treatment at
This, is easy to promote and apply in waste water treatment engineering, has good Social benefit and economic benefit.
Detailed description of the invention
Fig. 1 is the Anaerobic wastewater treatment real-time monitoring system schematic diagram built in embodiment 1；
Fig. 2 a, Fig. 2 b and Fig. 2 c are the searching process schematic diagram of PSO algorithm in embodiment 1；
The modeling procedure for the water outlet volatile fatty acid soft-sensing model that Fig. 3 a and Fig. 3 b are the PSO-SVM in embodiment 1
Schematic diagram；
Fig. 4 is the data exchange flow process schematic diagram in embodiment 1 between soft-sensing model MATLAB and configuration software MCGS；
Simulation result diagram before the water outlet volatile fatty acid soft-sensing model classification that Fig. 5 a is PSO-SVM in embodiment 1；
Fig. 5 b and Fig. 5 c are the sorted emulation of water outlet volatile fatty acid soft-sensing model of PSO-SVM in embodiment 1
Result figure.
Specific embodiment
Below in conjunction with specific embodiment and attached drawing, the present invention is further elaborated.Following embodiment is only used for this
Invention explains rather than limiting the invention.If it is noted that below have not especially detailed description process or symbol,
It is that those skilled in the art can refer to prior art understanding or realize such as ellipsis.
Embodiment 1
Hard measurement based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid
Method uses MCGS to collect the information data come from sensor as configuration software, uses MATLAB as soft sensor modeling
The step of platform, measurement method, is as follows:
(1) IC reactor is built as Anaerobic Waste Treatment System, and whole system builds schematic diagram as shown in Figure 1, empty
Line indicates that the transmission route of data information, solid line indicate that the circulation path of waste water, pecked line indicate the circulation path of lye；Experiment
Anaerobic reactor used is the IC anaerobic reactor of autonomous Design, which is organic glass production, high 1272mm, internal diameter
The volume ratio of 200mm, dischargeable capacity 25.1L, the first reaction zone and second reaction zone is 4: 1；
Anaerobic Waste Treatment System include alkali liquid tank 1, water inlet storage tank 2, on-line temperature monitoring instrument 3, ORP on-line computing model 4,
PH on-line computing model 5, wet gas flow meter 6, data processing centre 7, anaerobic reactor 9, lye peristaltic pump 10 and water inlet are compacted
Dynamic pump 11；Storage tank 2 of intaking is interior to be equipped with agitating device, and to mix well waste water quality, the lye prepared is placed in alkali liquid tank 1；Waste water
Pass through two BT600-2J type peristaltic pumps respectively with lye, i.e., water inlet peristaltic pump 11 and lye peristaltic pump 10 are delivered to anaerobic reaction
In device 9；Water inlet peristaltic pump 11 and lye peristaltic pump 10 adjust revolving speed by adjusting input voltage, control lye and waste water with this
Flow；Conduit is equipped at the top of device to export the biogas of reactor generation, for later period gas collection and analysis；For convenience of anaerobism
The detection of the parameters such as digestion system water quality indicator, system are equipped with pH on-line computing model 5, ORP on-line computing model 4, temperature online prison
Survey the on-line monitoring instruments such as instrument 3 and wet gas flow meter 6；
By variation water inlet organic loading in reactor operational process, the mode for changing influent alkalinity changes reactor water inlet
Condition and treatment conditions obtain the variation of Anaerobic Treatment water outlet VFA concentration under different condition, and according to national standard monitoring influent COD,
It is discharged the input quantity of pH, water outlet T, gas production and gas generating component, reactor oxidation-reduction potential as system, reactor operation
60d collects totally 159 groups of data；
(2) obtained metadata is divided into training set pTraining={ x；Y } and test set pTest={ x '；Y ' }, data set p at this time
={ x x '；Y y ' } data set (amounting to 159 groups of data) is divided into training set (amounting to 118 groups of data) and test set is (41 groups total
Data), and do normalized:
In formula, S (i) is any group of data in data set；Min (S) is to be worth the smallest one group of data in data set；max
It (S) is to be worth maximum one group of data in data set；
(3) training set data is used for SVM-regression model modeling, obtains the model output of training set at this time
Y, according to the relative error of model output Y and training set actual value y, training set is divided into two training sets, and (relative error is low
Be one kind in 50%, be greater than or equal to 50% to be another kind of), i.e., training set 1. with training set 2. namely pTraining 1={ x1；Y1 } and
pTraining 2={ x2；Y2 }, and 1. marking training set is respectively 1 and -1 with the metadata tag of training set 2.；
(4) according to the training set of classification 1. with training set 2., using SVM-classification model by test set point
At two corresponding test sets, i.e., test set 1. with test set 2. namely pTest 1={ x '1；y'1And pTest 2={ x '2；y'2, from
And metadata is divided into two group data sets, i.e., data set 1. with data set 2. namely p1={ x1 x’1；y1 y’1And p2={ x2
x’2；y2 y’2}；Wherein data set is 1.: 20 groups of 51 groups of data of training set, test set data；Data set is 2.: 67 groups of numbers of training set
According to, 21 groups of data of test set；
(5) 1. 2. obtained training set is utilized respectively SVM-regression model with training set and carries out modeling processing,
And 1. 2. verify the validity of model with test set with corresponding test set respectively, PSO parameter optimization process is shown in Fig. 2 a to Fig. 2 c,
From Fig. 2 a to Fig. 2 c it is found that model reaches optimal adaptation degree (Fig. 2 a) by 127 iteration before classification, after classification, two
Model passes through 8 times respectively and 10 iteration reach optimal adaptation degree (see Fig. 2 b and 2c), it can be seen that have metadata set classification
Accelerate to effect the training speed of model；
The modeling procedure of entire soft-sensing model is as shown in Figure 3a and Figure 3b shows, obtains comprising SVM-classification mould
The water outlet volatile fatty acid soft-sensing model of the PSO-SVM of type and SVM-regression model；
(6) the water outlet volatile fatty acid soft-sensing model of the PSO-SVM after training and test is embedded in industrial personal computer, and
Human-computer interaction interface is built using the configuration software in industrial personal computer；
(7) the water outlet volatility rouge based on PSO-SVM is realized using OPC (OLE for Process Control) technology
Data exchange between the aching and limp measurement model of fat and the configuration software of industrial personal computer, the data exchange process between MCGS and MATLAB
See Fig. 4.MCGS configuration software and its advanced development kit are installed to correct path, opens the opc server of MSGS, is wanted according to engineering
Seek the good human-computer interaction interface of configuration.In configuration software, defines including influent COD, water outlet pH, water outlet T, gas production and produce gas
Component, the parameter of reactor oxidation-reduction potential and water outlet Vfa Concentration, is written and read for MCGS.Starting
MATLAB, auxiliary variable order opcregister (' install ') in command window, for installing by OPCFoundation
The a set of core component that can be browsed other computers on network and can communicate provided.Auxiliary variable order again: a=
Opcda (' localhost ', ' MCGS.OPC.Server '): connect (da), then MCGS and MATLAB establish connection.It will instruction
After the PSO-SVM soft-sensing model insertion industrial personal computer perfected, during system is run, computer the waterpower influent COD being collected into,
It is discharged pH, T, gas production and gas generating component, reactor oxidation-reduction potential and the data for being discharged volatile fatty acid pass to soft
Part, operation PSO-SVM soft-sensing model carry out prediction water outlet volatile fatty acid number, hard measurement value are returned to human-computer interaction circle
Face；
(8) step (7) constantly are repeated, to realize the online real-time of anaerobic wastewater treatment system water outlet volatile fatty acid
Monitoring；
(9) system is exported and is showed with report form, and provide the characteristic index index of characterization system.
After network analysis operation, experiment actual value and system operations assay value difference are shown in Fig. 5 a (SVM-regression
Before category of model), Fig. 5 b and Fig. 5 c (after SVM-regression category of model), the characteristic index index of obtained simulation result
As shown in table 1.
1 system emulation result of table
By Fig. 5 a~Fig. 5 c and table 1 it is found that before classification, each performance indicator (test set) RMSE of model is 59.75,
MAPE is 42.97%, and correlation R is 85.25%；After classification, data set is 1. and data set collection performance indicator 2. is respectively as follows:
RMSE is 20.45,9.64, MAPE 12.11%, 7.16%, and correlation R is 99.14%, 99.59%.
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 in protection scope of the present invention.
Claims (7)
1. a kind of soft survey based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid
Amount method, which comprises the steps of:
(1) anaerobic wastewater treatment system is built, corresponding anaerobism is anti-under the conditions of obtaining and collecting input quantity and different input quantities
Answer the metadata of device output quantity；
(2) obtained metadata is divided into training set pTraining={ x；Y } and test set pTest={ x '；Y ' }, and by data set p={ x
x'；Y y ' } do normalized；Training set data is used for SVM-regression model modeling, obtains the model of training set
Training set is divided into two training sets, that is, instructed by output quantity Y according to the relative error of model output Y and training set actual value y
Practice collection 1. with training set 2. namely pTraining 1={ x1；y1And pTraining 2={ x2；y2, and mark member of the training set 1. with training set 2.
Data label is respectively 1 and -1；The relative error of model output Y and training set actual value y lower than 50% be training set 1.；
The relative error of model output Y and training set actual value y be greater than or equal to 50% be training set 2.；The SVM-
Regression model are as follows:
In formula, x is auxiliary variable sample data, aiFor Lagrange multiplier, b is offset, the natural number that n is >=1；k(xi,
xj) be kernel function and meet Mercer condition；
(3) according to the training set of classification 1. with training set 2., test set is divided into two using SVM-classification model
A corresponding test set, i.e. test set 1. with test set 2. namely pTest 1={ x '1；y'1And pTest 2={ x '2；y'2, thus will
Metadata is divided into two group data sets, i.e., data set 1. with data set 2. namely p1={ x1x’1；y1 y’1And p2={ x2 x’2；
y2 y’2}；The SVM-classification model are as follows:
In formula, x is auxiliary variable sample data,For Lagrange multiplier, b*For offset, w*For optimal weights value, T is represented
Transposition, xi *For optimal sample data, the natural number that n is >=1；(xi,yj) it is training set sample, xn∈Rn, yn∈ { -1,1 }；
(4) 1. 2. obtained training set is utilized respectively SVM-regression model with training set to handle, and used respectively
1. 2. corresponding test set verifies the validity of model with test set；When the precision of model reaches sets requirement, included
The water outlet volatile fatty acid hard measurement mould of the PSO-SVM of SVM-classification model and SVM-regression model
Type；
(5) the water outlet volatile fatty acid soft-sensing model of the PSO-SVM after training and test is embedded in industrial personal computer, and utilized
Configuration software in industrial personal computer builds human-computer interaction interface；
(6) realize that the configuration of water outlet volatile fatty acid soft-sensing model and industrial personal computer based on PSO-SVM is soft using OPC technology
The data of collection are delivered to the water outlet volatile fatty acid soft-sensing model of PSO-SVM, calculated by the data exchange between part
The predicted value of water Vfa Concentration, then the predicted value is back to industrial personal computer human-computer interaction interface and is shown；
(7) step (6) constantly are repeated, to realize the online real-time prison of anaerobic wastewater treatment system water outlet volatile fatty acid
It surveys；
(8) system is exported and is showed with report form, and provide the characteristic index index of characterization system.
2. according to claim 1 a kind of based on the water outlet of the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines
The flexible measurement method of volatile fatty acid, which is characterized in that in step (1), the anaerobic wastewater treatment system includes that anaerobism is anti-
Answer device, monitoring relevant to anaerobic effluent volatile fatty acid and detection instrument, industrial personal computer, D/A and A/D module；The input
Amount include water inlet organic loading, water outlet pH, leaving water temperature, produce gas component and yield, influent alkalinity and anaerobic reactor oxygen
Change reduction potential；The component for producing gas and yield include the yield of methane and carbon dioxide.
3. according to claim 1 a kind of based on the water outlet of the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines
The flexible measurement method of volatile fatty acid, which is characterized in that in step (1), the output quantity includes water outlet general volatile fat
The concentration of acid.
4. according to claim 1 a kind of based on the water outlet of the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines
The flexible measurement method of volatile fatty acid, which is characterized in that the particle x of the particle swarm algorithmjSpeed and location updating it is public
Formula is as follows:
In formula, Xj is particle vector position, and Vj is the speed of particle, and w is weighted value, c1、c2For aceleration pulse；rand1、rand2
It is random function, effect is the random number in order to generate (0,1)；Pbest is individual extreme value, and gbest is global extremum；K is indicated
The number of iterations；J indicates vector dimension.
5. according to claim 1 a kind of based on the water outlet of the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines
The flexible measurement method of volatile fatty acid, which is characterized in that in step (2), the normalized are as follows:
In formula, S (i) is any group of data in data set；Min (S) is to be worth the smallest one group of data in data set；max(S)
To be worth maximum one group of data in data set.
6. according to claim 1 a kind of based on the water outlet of the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines
The flexible measurement method of volatile fatty acid, which is characterized in that in step (4), the required precision of the model is by user according to need
Want self-setting；In step (8), the characteristic index index includes related coefficient, mean absolute percentage error, relative error
And root-mean-square error.
7. according to claim 1 a kind of based on the water outlet of the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines
The flexible measurement method of volatile fatty acid, which is characterized in that by transmission control protocol, internet protocol negotiation serial data mark
Standard monitors measurement output data in real time by computer and two-way communication, and aobvious by the monitoring window of configuration software
Show, so that system understands the index situation of change of sewage disposal system accurately and in time, promotes sewage treatment plant's efficient stable fortune
Row.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710250802.1A CN107132325B (en) | 2017-04-14 | 2017-04-14 | A kind of flexible measurement method based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710250802.1A CN107132325B (en) | 2017-04-14 | 2017-04-14 | A kind of flexible measurement method based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107132325A CN107132325A (en) | 2017-09-05 |
| CN107132325B true CN107132325B (en) | 2019-07-16 |
Family
ID=59716665
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710250802.1A Active CN107132325B (en) | 2017-04-14 | 2017-04-14 | A kind of flexible measurement method based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107132325B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109669017B (en) * | 2017-10-17 | 2021-04-27 | 中国石油化工股份有限公司 | Refinery distillation tower top cut water ion concentration prediction method based on deep learning |
| CN108549740A (en) * | 2018-03-06 | 2018-09-18 | 华南理工大学 | A kind of anaerobic system water outlet ammonia nitrogen flexible measurement method based on integrated intelligent algorithm |
| CN109273058A (en) * | 2018-09-20 | 2019-01-25 | 中轻国环（北京）环保科技有限公司 | A kind of composite algorism for the exceeded early warning of anaerobic processes volatile fatty acid |
| CN109408896B (en) * | 2018-09-27 | 2024-01-05 | 华南师范大学 | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production |
| CN111160518B (en) * | 2018-11-08 | 2023-08-15 | 中国石油化工股份有限公司 | Soft measurement method and device for COD of sewage refining and Chemical Oxygen Demand (COD) and machine-readable storage medium |
| CN109975366B (en) * | 2019-03-25 | 2020-08-14 | 浙江大学 | Rural domestic sewage A2Soft measurement method and device for COD concentration of effluent from O treatment terminal |
| CN112456639B (en) * | 2020-11-10 | 2022-11-22 | 盐城工学院 | Intelligent prediction method for hydrogen sulfide generation amount in anaerobic digestion |
| CN114093434A (en) * | 2021-11-30 | 2022-02-25 | 华能国际电力股份有限公司 | Soft measurement method for components and concentrations of organic waste hydrolysis acidification volatile fatty acid |
| CN114354867A (en) * | 2021-12-27 | 2022-04-15 | 江苏海峡环保科技发展有限公司 | Method for predicting volatile fatty acid in sewage |
| CN114460255B (en) * | 2022-02-08 | 2024-06-11 | 北京工商大学 | LSTM-based method and system for measuring volatile fatty acid in wastewater |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101763084A (en) * | 2009-12-29 | 2010-06-30 | 浙江大学 | System and method for minimizing chemical oxygen demand (COD) discharge of pesticide production waste liquid incinerator |
| CN102073797A (en) * | 2011-02-23 | 2011-05-25 | 江南大学 | Analysis method for identifying main influence factors on water qualities of rivers flowing into Tai Lake on basis of PSO (Particle Swarm Optimization) and SVM (Support Vector Machine) hybrid algorithm |
| CN102902985A (en) * | 2012-10-10 | 2013-01-30 | 常州大学 | Coastal water quality evaluation method based on two-classification support vector machines and particle swarm algorithm |
| CN103020434A (en) * | 2012-11-30 | 2013-04-03 | 南京航空航天大学 | Particle swarm optimization-based least square support vector machine combined predicting method |
| CN103488090A (en) * | 2013-09-22 | 2014-01-01 | 浙江大学 | Crowd-sourced and machine learning incinerator noxious substance discharging standard reaching control system and method |
| CN103617147A (en) * | 2013-11-27 | 2014-03-05 | 中国地质大学（武汉） | Method for identifying mine water-inrush source |
| CN105045951A (en) * | 2015-05-27 | 2015-11-11 | 华南理工大学 | Soft-measurement method for volatile fatty acid in effluent of anaerobic wastewater treatment system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8158377B2 (en) * | 2005-08-31 | 2012-04-17 | Ut-Battelle, Llc | Biosensor method and system based on feature vector extraction |
-
2017
- 2017-04-14 CN CN201710250802.1A patent/CN107132325B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101763084A (en) * | 2009-12-29 | 2010-06-30 | 浙江大学 | System and method for minimizing chemical oxygen demand (COD) discharge of pesticide production waste liquid incinerator |
| CN102073797A (en) * | 2011-02-23 | 2011-05-25 | 江南大学 | Analysis method for identifying main influence factors on water qualities of rivers flowing into Tai Lake on basis of PSO (Particle Swarm Optimization) and SVM (Support Vector Machine) hybrid algorithm |
| CN102902985A (en) * | 2012-10-10 | 2013-01-30 | 常州大学 | Coastal water quality evaluation method based on two-classification support vector machines and particle swarm algorithm |
| CN103020434A (en) * | 2012-11-30 | 2013-04-03 | 南京航空航天大学 | Particle swarm optimization-based least square support vector machine combined predicting method |
| CN103488090A (en) * | 2013-09-22 | 2014-01-01 | 浙江大学 | Crowd-sourced and machine learning incinerator noxious substance discharging standard reaching control system and method |
| CN103617147A (en) * | 2013-11-27 | 2014-03-05 | 中国地质大学（武汉） | Method for identifying mine water-inrush source |
| CN105045951A (en) * | 2015-05-27 | 2015-11-11 | 华南理工大学 | Soft-measurement method for volatile fatty acid in effluent of anaerobic wastewater treatment system |
Non-Patent Citations (4)
| Title |
|---|
| 基于PSO-LSSVM污水处理***出水数据的软测量;徐方舟 等;《江南大学学报（自然科学版）》;20100630;第9卷(第3期);第253-256页 |
| 基于PSO-SVM的废水厌氧处理过程软测量模型;刘林 等;《环境科学学报》;20160711;第37卷(第6期);第2124页左栏-2125页左栏，图1、3 |
| 基于PSO-SVM算法的环境监测数据异常监测和确实补全;魏晶茹 等;《环境监测管理与技术》;20160831;第28卷(第4期);第53-56页 |
| 基于动力学和PSO-SVM的废水厌氧处理产气量的混合软测量模型;刘林 等;《中国造纸》;20170315;第36卷(第3期);第31-36页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107132325A (en) | 2017-09-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107132325B (en) | A kind of flexible measurement method based on the Anaerobic Waste Treatment System of particle swarm algorithm and support vector machines water outlet volatile fatty acid | |
| Bagherzadeh et al. | Prediction of energy consumption and evaluation of affecting factors in a full-scale WWTP using a machine learning approach | |
| Niu et al. | A novel effluent quality predicting model based on genetic-deep belief network algorithm for cleaner production in a full-scale paper-making wastewater treatment | |
| Li et al. | Analysis and forecasting of the energy consumption in wastewater treatment plant | |
| Mehrani et al. | Application of a hybrid mechanistic/machine learning model for prediction of nitrous oxide (N2O) production in a nitrifying sequencing batch reactor | |
| CN110320335B (en) | Multi-element robust soft measurement method for effluent quality index of sewage treatment | |
| CN101630376A (en) | Soft-sensing modeling method and soft meter of multi-model neural network in biological fermentation process | |
| Liu et al. | Online prediction of effluent COD in the anaerobic wastewater treatment system based on PCA-LSSVM algorithm | |
| Li et al. | Assessment for surface water quality in Lake Taihu Tiaoxi River Basin China based on support vector machine | |
| Han et al. | Prediction of sludge bulking using the knowledge-leverage-based fuzzy neural network | |
| Zhang et al. | Plant-scale biogas production prediction based on multiple hybrid machine learning technique | |
| CN105045951A (en) | Soft-measurement method for volatile fatty acid in effluent of anaerobic wastewater treatment system | |
| Hong et al. | Concurrent monitoring strategy for static and dynamic deviations based on selective ensemble learning using slow feature analysis | |
| Wang et al. | A full-view management method based on artificial neural networks for energy and material-savings in wastewater treatment plants | |
| Ruan et al. | Application of fuzzy neural networks for modeling of biodegradation and biogas production in a full-scale internal circulation anaerobic reactor | |
| Xu et al. | Integrated data-driven strategy to optimize the processes configuration for full-scale wastewater treatment plant predesign | |
| Cheng et al. | Knowledge and data-driven hybrid system for modeling fuzzy wastewater treatment process | |
| Yasmin et al. | Improved support vector machine using optimization techniques for an aerobic granular sludge | |
| Fawzy et al. | Trio-V wind analyzer: a generic integral system for wind farm suitability design and power prediction using big data analytics | |
| Yang et al. | Cloud-model-based feature engineering to analyze the energy–water nexus of a full-scale wastewater treatment plant | |
| CN104345636A (en) | Dissolved-oxygen control method based on improved differential algorithm | |
| EP0446036A2 (en) | Control apparatus | |
| Dong et al. | Virtual sample generation empowers machine learning-based effluent prediction in constructed wetlands | |
| Huang et al. | Enhancing dissolved oxygen control using an on-line hybrid fuzzy-neural soft-sensing model-based control system in an anaerobic/anoxic/oxic process | |
| Mathur et al. | Modeling of two-stage anaerobic onsite wastewater sanitation system to predict effluent soluble chemical oxygen demand through machine learning |
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 |