CN110232256A - A kind of sewage disposal process monitoring method based on KPLS and RWFCM - Google Patents

A kind of sewage disposal process monitoring method based on KPLS and RWFCM Download PDF

Info

Publication number
CN110232256A
CN110232256A CN201910573311.XA CN201910573311A CN110232256A CN 110232256 A CN110232256 A CN 110232256A CN 201910573311 A CN201910573311 A CN 201910573311A CN 110232256 A CN110232256 A CN 110232256A
Authority
CN
China
Prior art keywords
matrix
sample
variable
sewage disposal
disposal process
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
Application number
CN201910573311.XA
Other languages
Chinese (zh)
Other versions
CN110232256B (en
Inventor
周平
张瑞垚
王宏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northeastern University China
Original Assignee
Northeastern University China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Northeastern University China filed Critical Northeastern University China
Priority to CN201910573311.XA priority Critical patent/CN110232256B/en
Publication of CN110232256A publication Critical patent/CN110232256A/en
Application granted granted Critical
Publication of CN110232256B publication Critical patent/CN110232256B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F9/00—Multistage treatment of water, waste water or sewage
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F1/00—Treatment of water, waste water, or sewage
    • C02F1/52—Treatment of water, waste water, or sewage by flocculation or precipitation of suspended impurities
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F1/00—Treatment of water, waste water, or sewage
    • C02F2001/007—Processes including a sedimentation step
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F2101/00—Nature of the contaminant
    • C02F2101/10—Inorganic compounds
    • C02F2101/16—Nitrogen compounds, e.g. ammonia
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F2101/00—Nature of the contaminant
    • C02F2101/30—Organic compounds
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F2209/00—Controlling or monitoring parameters in water treatment
    • C02F2209/08—Chemical Oxygen Demand [COD]; Biological Oxygen Demand [BOD]
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F2209/00—Controlling or monitoring parameters in water treatment
    • C02F2209/10—Solids, e.g. total solids [TS], total suspended solids [TSS] or volatile solids [VS]
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F2209/00—Controlling or monitoring parameters in water treatment
    • C02F2209/14—NH3-N
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F2209/00—Controlling or monitoring parameters in water treatment
    • C02F2209/22—O2
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F3/00—Biological treatment of water, waste water, or sewage
    • C02F3/30—Aerobic and anaerobic processes
    • C02F3/302—Nitrification and denitrification treatment
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A20/00—Water conservation; Efficient water supply; Efficient water use
    • Y02A20/152—Water filtration
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02W—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO WASTEWATER TREATMENT OR WASTE MANAGEMENT
    • Y02W10/00—Technologies for wastewater treatment
    • Y02W10/10—Biological treatment of water, waste water, or sewage

Abstract

The present invention relates to quality of sewage disposal monitoring technical fields, provide a kind of sewage disposal process monitoring method based on KPLS and RWFCM.The present invention acquires nominal situation and the sewage disposal process data sample comprising unusual service condition first, using sewage treatment run variable, effluent characteristics variable data as input, output data matrix, and standardize two matrixes；Then KPLS model is constructed, and solves score matrix；Then score matrix is clustered based on RWFCM algorithm, obtains subordinated-degree matrix, unusual service condition monitoring is carried out to sewage disposal process according to subordinated-degree matrix；Finally, establishing the linear regression model (LRM) of subordinated-degree matrix and sample variable, solves variable and contribute matrix, and contribute matrix to carry out unusual service condition identification to sewage disposal process according to variable.The present invention can to high dimensional data carry out dimensionality reduction and handle nonlinear data and to outlier it is insensitive, can be improved sewage disposal process monitoring and identification timeliness and accuracy.

Description

A kind of sewage disposal process monitoring method based on KPLS and RWFCM
Technical field
The present invention relates to quality of sewage disposal monitoring technical fields, more particularly to a kind of dirt based on KPLS and RWFCM Water treatment procedure monitoring method.
Background technique
With the quickening of Chinese Urbanization, process of industrialization, demand of the society to freshwater resources increasingly increases, and needs to accelerate City domestic sewage handles disposal facility construction, improves city domestic sewage processing capacity.Sewage disposal process by activated sludge process It is the main method of currently processed municipal sewage.Activated sludge purification sewage mainly has initial stage absorption, the metabolism of microorganism, is formed 3 processes of floccule body and precipitating will its essence is series of biochemical reactions is passed through using the micropopulation in activated sludge Biodegradable organic matter in sewage is adsorbed, decomposed and is aoxidized, to it be separated from sewage, to reach To the purpose of purification sewage.
Currently, generalling use biochemical oxygen demand (BOD) ([BOD]), COD ([COD]), suspended matter ([SS]), ammonia nitrogen ([NH]), total phosphorus ([TP]) are used as sewage discharge index.In sewage disposal process, flow of inlet water, water inlet composition, pollutant are dense The parameters such as degree, Changes in weather are all passively to receive, microbial life active receiving dissolved oxygen concentration, microbial population, sewage pH The many factors such as value influence, therefore it is very difficult for keeping municipal sewage plant to run steadily in the long term.Sewage treatment plant's hair Raw failure easily causes that effluent quality is not up to standard, increase operating cost and causes environmental pollution.So if cannot detect in time Sewage disposal process unusual service condition out leads to make correctly judgement and does not take effective measures to be adjusted in time and entangles Just, it will cause the irreversible loss of sewage disposal process.Therefore, operator is by detection sewage disposal process, to unusual service condition Accurate judgement is made, and is timely and accurately taken measures, to the direct motion of sewage treatment safety and stability is guaranteed, to guarantee effluent characteristics It is particularly important.
Process monitoring method based on data-driven is widely applied, wherein multivariate statistical process monitoring method One of the research hotspot for having become process monitoring field, the research for having pushed sewage treatment effluent characteristics to monitor.Sewage treatment Process is a complicated non-linear process, and historical failure data and not perfect, and therefore, sewage treatment effluent characteristics monitored Cheng Zhong, the fault identification of failure variable are still a relatively difficult problem.
Existing sewage disposal process monitoring method mostly uses data digging method in recent years, the main reason is that in the presence of Mass data can be widely used, and there is an urgent need to convert the data into useful information and knowledge.At sewage Reason process data does not have class indication, and the generation of sewage treatment failure does not have great association with the time, so being not suitable for using Classification or sequential mode mining are excavated.And the clustering in data mining technology is a kind of unsupervised segmentation technology, It can be used to carry out the analysis of the few data of priori knowledge well, therefore Clustering Analysis Technology obtains in sewage process monitoring It is widely applied.
FuzzycMeans Clustering (fuzzy c-means clustering, FCM) algorithm is classical one of clustering algorithm. FCM gives sample for the degree of uncertainty of classification, establishes sample for the uncertainty description of classification, more meets pair The description of objective world.However, sewage disposal process data with it is high-dimensional and non-linear and have outlier, traditional FCM Algorithm can not handle higher-dimension and nonlinear data and very sensitive to outlier, and this adds increased the difficulty of process monitoring, drops The low reliability of fault detection, has a huge impact sewage effluent characteristics, causes certain economic loss even thing Therefore occur.Meanwhile the clusters number needs of FCM algorithm are artificially preset, and have great limitation in practical applications.
Summary of the invention
In view of the problems of the existing technology, the present invention provides a kind of sewage disposal process prison based on KPLS and RWFCM Survey method, can to high dimensional data carry out dimensionality reduction and handle nonlinear data and to outlier it is insensitive, can be accurately and easily It determines clusters number, improves the timeliness and accuracy of sewage disposal process monitoring and identification.
The technical solution of the present invention is as follows:
A kind of sewage disposal process monitoring method based on KPLS and RWFCM, which is characterized in that include the following steps:
Step 1: nominal situation and the sewage disposal process data sample comprising unusual service condition are acquired respectively, at the sewage Managing process data sample includes m1A sewage treatment runs variable, m2A effluent characteristics variable；From time angle by nominal situation Sewage disposal process data sample is added in front of the sewage disposal process data sample comprising unusual service condition, forms blended data sample This collection；By m in blended data sample set1The data of a sewage treatment operation variable are as input data matrix X, by blended data M in sample set2The data of a effluent characteristics variable are as output data matrix Y；
Step 2: input data matrix X and output data matrix Y are pre-processed；The pretreatment includes calculating input The mean value and standard deviation of each variable in data matrix X and output data matrix Y, and by input data matrix X and output data square Battle array Y is standardized into the data of zero-mean and unity standard deviation；
Step 3: the KPLS model of building sewage disposal process monitoring maps the input sample x in input data matrix X It to high-dimensional feature space F:x → Φ (x) ∈ F, introduces gaussian kernel function and obtains the Gram matrix K of input data matrix X, and is right Gram matrix K carries out centralization processing；
Step 4: pivot number being determined using cross-validation method, and solves score matrix T；
Step 5: score matrix T being clustered based on RWFCM algorithm, subordinated-degree matrix U is obtained, according to subordinated-degree matrix U carries out unusual service condition monitoring to sewage disposal process: if a certain moment sample being subordinate to the cluster centre of nominal situation sample Degree is less than μ, then exception has occurred at the sample in sewage disposal process；
Step 6: establishing the linear regression model (LRM) of variable in subordinated-degree matrix U and input data matrix X, and bright using glug Day multiplier method solution variable contributes matrix N, contributes matrix N to carry out unusual service condition identification to sewage disposal process according to variable: if Contribution { η of a-th of variable to all clustersa1,...,ηacIn maximum value be ηag, then a-th of variable is abnormal with g kind The relevant failure variable of operating condition；Wherein, c is clusters number, and g ∈ { 1 ..., c-1 }, c-th of cluster is nominal situation sample Cluster.
The sewage disposal process uses activated sludge process, and raw sewage, into biochemistry pool part, carries out after coagulation After biological denitrificaion, a part is entered secondary settling tank and is precipitated by inner circulating reflux denitrogenation again, another part；Biochemistry pool part Including biochemistry pool l ∈ { 1,2,3,4,5 }, wherein biochemistry pool l1∈ { 1,2 } is the main anoxic for completing anti-nitration reaction process Area, biochemistry pool l2∈ { 3,4,5 } is the main aerobic zone for completing nitration reaction process；In the step 1, the m1At a sewage Reason operation variable includes flow of inlet water, influent ammonia nitrogen amount, the active heterotrophism bacteria biomass in biochemistry pool l ∈ { 1,2,3,4,5 }, life Change easily biological-degradable organic substrates amount, the basicity in biochemistry pool l ∈ { 1,2,3,4,5 }, biochemistry in pond l ∈ { 1,2,3,4,5 } Pond l1Nitrate nitrogen amount, biochemistry pool l in ∈ { 1,2 }2Active autotrophy bacteria biomass, biochemistry pool l in ∈ { 3,4,5 }2∈{3,4,5} In ammonia nitrogen amount, biochemistry pool l2Dissolved oxygen content in ∈ { 3,4,5 }；The m2A effluent characteristics variable includes biochemical oxygen demand (BOD), changes Learn oxygen demand, suspended matter, water outlet ammonia nitrogen amount.
The step 3 includes the following steps:
Step 3.1: the KPLS model of building sewage disposal process monitoring is
Φ=TP1'+Φr
Y=TQ'+Yr
Step 3.2: the input sample x in input data matrix X is mapped to high-dimensional feature space F:x → Φ (x) ∈ F, It introduces gaussian kernel function and obtains the Gram matrix K of input data matrix X, and centralization processing is carried out to Gram matrix K, it will KPLS model conversion is
K=TP2'+E
Y=TQ'+Yr
Wherein, the element of the i-th row jth column of Gram matrix K is Kij=k (xi,xj)=< Φ (xi),Φ(xj) >, xi、xjPoint It Wei not i-th of input sample x in input data matrix Xi, j-th of input sample xj, k (xi,xj) it is gaussian kernel function, i, j ∈ { 1,2 ..., n }, n are the number of samples in input data matrix X；T is high dimensional data Φ={ Φ (xi),i∈{1, 2 ..., n } } score matrix, T=[t1,...,tA], A is pivot number, P1=[p11,...,p1A]、P2=[p21,..., p2A], Q=[q1,...,qA] be respectively matrix Φ, Gram matrix K, output data matrix Y loading matrix, Φr、E、YrRespectively For matrix Φ, Gram matrix K, the modeling residual error of output data matrix Y.
In the step 4, pivot number A is determined using cross-validation method, and solve score matrix T, included the following steps:
Step 4.1: enabling u is the either rank of output data matrix Y；
Step 4.2: calculating score vector: t=Ku；
Step 4.3: by score vector t normalized: | | t | | → 1；
Step 4.4: being respectively listed on score vector t in output data matrix Y being returned: q=Y't；
Step 4.5: calculating the new score of output data matrix Y: u=Yq；
Step 4.6: by u vector normalized: | | u | | → 1；
Step 4.7: judging whether u restrains: if it is, jumping to step 4.8；If it is not, then jumping to step 4.2；
Step 4.8: matrix: K=(I-tt') K (I-tt'), Y=Y-tq' is updated, repeats step 4.2 to step 4.7, into The calculating of the next score vector of row, until A score vector is extracted；Wherein, I is unit matrix.
In the step 3, centralization treated Gram matrixWherein, EnFor The unit matrix of n × n, 1nComplete 1 column vector is tieed up for n, 1 'nIt is 1nTransposed matrix.
The step 5 includes the following steps:
Step 5.1: score matrix T being clustered based on RWFCM algorithm, building RWFCM objective function is
Wherein,For i-th of row vector of score matrix T,For the input sample x of m1 dimensioniAfter corresponding dimensionality reduction A reform sample, uijFor sampleTo j-th of cluster centre vjDegree of membership, sijFor sampleBelong to j-th of cluster can Energy property, subordinated-degree matrix U=(uij)n×c, cluster centre matrix V=(vj)c×A, c is clusters number；M ∈ [1 ,+∞] is fuzzy Index；For sampleWith j-th of cluster centre vjBetween mahalanobis distance, SjTo be blurred covariance square Battle array, SjFor positive definite matrix；For penalty term, ηjFor penalty factor, p is Possibility index；To score matrix T The c cluster centre clustered includes the cluster centre of nominal situation sample and the cluster of c-1 kind unusual service condition sample Center；
Step 5.2: solve subordinated-degree matrix U:
Step 5.2.1: initialization RWFCM algorithm parameter: determining clusters number c, sets Fuzzy Exponential m and Possibility index P, setting algorithm terminate limit ε, maximum number of iterations count, initialize the number of iterations k=1, random initializtion subordinated-degree matrix U(k)=(uij (k))n×c, random initializtion cluster centre matrix V(k)=(vj (k))c×A, random initializtion blurring covariance matrix Collect S(k)=(Sj (k))n×n×c；
Step 5.2.2: by uij (k)、vj (k)、Sj (k)Substitute into formulaCalculate the A possibility that k+1 iteration matrix B(k+1)=(sij (k+1))n×c；
Step 5.2.3: by sij (k+1)、vj (k)、Sj (k)Substitute into formulaMeter Calculate the subordinated-degree matrix U of+1 iteration of kth(k+1)=(uij (k+1))n×c；
Step 5.2.4: by uij (k+1)、sij (k+1)Substitute into formulaCalculate+1 iteration of kth Cluster centre matrix be V(k+1)=(vj (k+1))c×A；
Step 5.2.5: by uij (k+1)、sij (k+1)、vj (k+1)Substitute into formula Calculate the blurring covariance matrix collection S of+1 iteration of kth(k+1)=(Sj (k+1))n×n×c；Wherein, γjFor Lagrange multiplier；
Step 5.2.6: if | | U(k+1)-U(k)| | < ε or the number of iterations k > count then stops iteration, obtains final person in servitude Category degree matrix U, enters step 5.3；Otherwise, k=k+1, return step 5.2.2 are enabled；
Step 5.3: unusual service condition monitoring being carried out to sewage disposal process according to subordinated-degree matrix U: if i-th of sample μ is less than to the degree of membership of the cluster centre of nominal situation sample, then exception has occurred at i-th of sample in sewage disposal process； If exception has occurred in sewage disposal process, 6 are entered step；If there is no exceptions for sewage disposal process, terminate the base In the sewage disposal process monitoring method of KPLS and RWFCM.
In the step 5.2.1, determine that clusters number c includes:
Calculate the input sample x in input data matrix XiDot density value DiFor
Wherein,rdFor neighborhood density effective radius,
Calculating constructed fuction S (j) is
The image for drawing constructed fuction S (j), using the slope number of the image of constructed fuction S (j) as clusters number c.
The step 6 includes the following steps:
Step 6.1: the linear regression model (LRM) for establishing variable in subordinated-degree matrix U and input data matrix X is
Wherein, N0=(η0j)1×cFor constant term, εijFor error term, it is assumed that meet: E (εij)=0, Var (εij)=δ2, δ is Constant；xiaFor i-th of input sample x in input data matrix XiA-th of variable-value；Square is contributed for variable Battle array, ηajFor regression coefficient, ηajThe contribution that j-th is clustered for a-th of variable；
Step 6.2: variable is solved using method of Lagrange multipliers and contributes matrix N:
Step 6.2.1: initialize each parameter: setting algorithm terminates limit τ, maximum number of iterations T, initializes the number of iterations k =1, random initializtion variable contributes matrix
Step 6.2.2: by ηaj (k)Substitute into formulaCalculate the N of+1 iteration of kth0 (k+1) =(η0j (k+1))1×c；
Step 6.2.3: by η0j (k+1)Substitute into formula The variable for calculating+1 iteration of kth contributes matrix
Step 6.2.4: if | | N(k+1)-N(k)| | < τ or the number of iterations k > T then stops iteration, enters step 6.3；It is no Then, k=k+1, return step 6.2.2 are enabled；
Step 6.3: contributing matrix N to carry out unusual service condition identification to sewage disposal process according to variable: if a-th of variable pair Contribution { the η of all clustersa1,...,ηacIn maximum value be ηag, then a-th of variable is event relevant to g kind unusual service condition Hinder variable；Wherein, g ∈ { 1 ..., c-1 }, c-th of cluster are the cluster of nominal situation sample.
The invention has the benefit that
(1) present invention combines KPLS algorithm with RWFCM algorithm, constructs KPLS model and RWFCM model to describe just Normal production process only needs nominal situation data as flag data without the priori knowledge of sewage disposal process unusual service condition. It is primarily based on data-driven method, using gaussian kernel function, the process variable after standardization is projected into high-dimensional feature space, High-dimensional feature space establishes the KPLS model of sewage disposal process monitoring, after determining pivot number using cross-validation method, to height It ties up input data and carries out dimensionality reduction, obtain input data of the score matrix T as clustering in RWFCM algorithm, reaching dimensionality reduction Nonlinear data can not be handled and to outlier sensitive issue by solving FCM while purpose again.
(2) the present invention is based on Density functional calculations constructed fuctions, and solve clusters number according to constructed fuction, can be accurate Easily determine clusters number, the clusters number for solving RWFCM algorithm needs artificial preset confinement problems.
(3) the present invention is based on RWFCM algorithms clusters score matrix T, subordinated-degree matrix U is obtained, according to degree of membership Matrix U carries out unusual service condition monitoring to sewage disposal process, can monitor that the moment occurs for unusual service condition by sample degree of membership And identify the number of unusual service condition.The present invention is by establishing the linear regression mould of subordinated-degree matrix U Yu sewage disposal process variable Type solves linear regression model (LRM) using method of Lagrange multipliers and obtains variable contribution matrix N, and contributes matrix N pair according to variable Sewage disposal process carries out unusual service condition identification, can identify that every kind of unusual service condition is relevant to the contribution of cluster by variable Failure variable.The present invention is high to the monitoring of sewage disposal process unusual service condition and identification timeliness, accuracy, can facilitate operation Monitored by personnel's sewage disposal process, to sewage treatment effluent characteristics fluctuation make accurate judgement, and take timely measure processing and It corrects, and then guarantees the stabilization of sewage plant, efficient, safe operation, guarantees effluent characteristics.
Detailed description of the invention
Fig. 1 is the flow chart of the sewage disposal process monitoring method of the invention based on KPLS and RWFCM；
Fig. 2 is that monitor sample illustrates the degree of membership of the cluster centre of nominal situation sample in the specific embodiment of the invention Figure；
Fig. 3 is monitor sample being subordinate to the cluster centre of the 1st class unusual service condition sample in the specific embodiment of the invention Spend schematic diagram；
Fig. 4 is monitor sample being subordinate to the cluster centre of the 2nd class unusual service condition sample in the specific embodiment of the invention Spend schematic diagram；
Fig. 5 is the Clustering Effect schematic diagram of score matrix T in the specific embodiment of the invention；
Fig. 6 is each sample variable in the specific embodiment of the invention to the contribution schematic diagram of each cluster.
Specific embodiment
Below in conjunction with the drawings and specific embodiments, the invention will be further described.
As shown in Figure 1, being the flow chart of the sewage disposal process monitoring method of the invention based on KPLS and RWFCM.This The sewage disposal process monitoring method based on KPLS and RWFCM of invention, which is characterized in that include the following steps:
Step 1: nominal situation and the sewage disposal process data sample comprising unusual service condition are acquired respectively, at the sewage Managing process data sample includes m1A sewage treatment runs variable, m2A effluent characteristics variable；From time angle by nominal situation Sewage disposal process data sample is added in front of the sewage disposal process data sample comprising unusual service condition, forms blended data sample This collection；By m in blended data sample set1The data of a sewage treatment operation variable are as input data matrix X, by blended data M in sample set2The data of a effluent characteristics variable are as output data matrix Y.
In the present embodiment, the sewage disposal process uses activated sludge process.Activated Sludge Process process usually according to Degree for the treatment of is divided into coagulation, two stage treatment and tertiary treatment.Raw sewage is after coagulation, into biochemistry pool part, into After row biological denitrificaion, a part is entered secondary settling tank and is precipitated by inner circulating reflux denitrogenation again, another part.Biochemistry pool is Complete biochemical reaction process, the purification most important place of sewage.Biochemistry pool part includes biochemistry pool l ∈ { 1,2,3,4,5 }, In, biochemistry pool l1∈ { 1,2 } is the main anoxic zone for completing anti-nitration reaction process, biochemistry pool l2∈ { 3,4,5 } is main complete At the aerobic zone of nitration reaction process；In the step 1, as shown in table 1 below, the m1A sewage treatment operation variable include into Water flow, influent ammonia nitrogen amount, the active heterotrophism bacteria biomass in biochemistry pool l ∈ { 1,2,3,4,5 }, biochemistry pool l ∈ 1,2,3,4, 5 } the easily biological-degradable organic substrates amount in, the basicity in biochemistry pool l ∈ { 1,2,3,4,5 }, biochemistry pool l1Nitre in ∈ { 1,2 } Nitrogen quantity, biochemistry pool l2Active autotrophy bacteria biomass, biochemistry pool l in ∈ { 3,4,5 }2Ammonia nitrogen amount, biochemistry pool in ∈ { 3,4,5 } l2Dissolved oxygen content in ∈ { 3,4,5 }；The m2A effluent characteristics variable include biochemical oxygen demand (BOD), COD, suspended matter, It is discharged ammonia nitrogen amount.
The unusual service condition is sludge bulking well known to those skilled in the art, foam, dross, toxicity impact, heavy rain day Gas etc..In the present embodiment, acquire 100 nominal situations sewage disposal process data sample and 1300 comprising rainstorm weather, Toxicity impacts the sewage disposal process data sample of both unusual service conditions, and composition includes the blended data sample of 1400 samples Collection.By m in blended data sample set1The data of=28 sewage treatment operation variables are as input data matrix X ∈ R1400×28, By m in blended data sample set2The data of=4 effluent characteristics variables are as output data matrix Y ∈ R1400×4。
Table 1
Step 2: input data matrix X and output data matrix Y are pre-processed；The pretreatment includes calculating input The mean value and standard deviation of each variable in data matrix X and output data matrix Y, and by input data matrix X and output data square Battle array Y is standardized into the data of zero-mean and unity standard deviation.
Step 3: the KPLS model of building sewage disposal process monitoring maps the input sample x in input data matrix X It to high-dimensional feature space F:x → Φ (x) ∈ F, introduces gaussian kernel function and obtains the Gram matrix K of input data matrix X, and is right Gram matrix K carries out centralization processing.
The step 3 includes the following steps:
Step 3.1: the KPLS model of building sewage disposal process monitoring is
Φ=TP1'+Φr
Y=TQ'+Yr
Step 3.2: the input sample x in input data matrix X is mapped to high-dimensional feature space F:x → Φ (x) ∈ F, It introduces gaussian kernel function and obtains the Gram matrix K of input data matrix X, and centralization processing is carried out to Gram matrix K, it will KPLS model conversion is
K=TP2'+E
Y=TQ'+Yr
Wherein, the element of the i-th row jth column of Gram matrix K is Kij=k (xi,xj)=< Φ (xi),Φ(xj) >, xi、xjPoint It Wei not i-th of input sample x in input data matrix Xi, j-th of input sample xj, k (xi,xj) it is gaussian kernel function, i, j ∈ { 1,2 ..., n }, n are the number of samples in input data matrix X；T is high dimensional data Φ={ Φ (xi),i∈{1, 2 ..., n } } score matrix, T=[t1,...,tA], A is pivot number, P1=[p11,...,p1A]、P2=[p21,..., p2A], Q=[q1,...,qA] be respectively matrix Φ, Gram matrix K, output data matrix Y loading matrix, Φr、E、YrRespectively For matrix Φ, Gram matrix K, the modeling residual error of output data matrix Y.
In the step 3, centralization treated Gram matrixWherein, En is The unit matrix of n × n, 1nComplete 1 column vector is tieed up for n, 1 'nIt is 1nTransposed matrix.
Wherein, KPLS model is constructed using non-linear least square iterative algorithm, KPLS, that is, core offset minimum binary constructs (kernel projection to latent structures).In the present embodiment, gaussian kernel function isIts In, c1For gaussian kernel function width parameter, c1Value by 5m1Principle of experience determines, that is, determines c1=5*m1=140.
Step 4: pivot number being determined using cross-validation method, and solves score matrix T.
In the step 4, pivot number A is determined using cross-validation method, and solve score matrix T, included the following steps:
Step 4.1: enabling u is the either rank of output data matrix Y；
Step 4.2: calculating score vector: t=Ku；
Step 4.3: by score vector t normalized: | | t | | → 1；
Step 4.4: being respectively listed on score vector t in output data matrix Y being returned: q=Y't；
Step 4.5: calculating the new score of output data matrix Y: u=Yq；
Step 4.6: by u vector normalized: | | u | | → 1；
Step 4.7: judging whether u restrains: if it is, jumping to step 4.8；If it is not, then jumping to step 4.2；
Step 4.8: matrix: K=(I-tt') K (I-tt'), Y=Y-tq' is updated, repeats step 4.2 to step 4.7, into The calculating of the next score vector of row, until A score vector is extracted；Wherein, I is unit matrix.
In the present embodiment, pivot number A=3 is determined using cross-validation method.
Step 5: score matrix T being clustered based on RWFCM algorithm, subordinated-degree matrix U is obtained, according to subordinated-degree matrix U carries out unusual service condition monitoring to sewage disposal process: if a certain moment sample being subordinate to the cluster centre of nominal situation sample Degree is less than μ, then exception has occurred at the sample in sewage disposal process.
The step 5 includes the following steps:
Step 5.1: score matrix T being clustered based on RWFCM algorithm, building RWFCM objective function is
Wherein,For i-th of row vector of score matrix T,After the corresponding dimensionality reduction of input sample xi for m1 dimension A reform sample, uijFor sampleTo j-th of cluster centre vjDegree of membership, sijFor sampleBelong to j-th of cluster can Energy property, subordinated-degree matrix U=(uij)n×c, cluster centre matrix V=(vj)c×A, c is clusters number；M ∈ [1 ,+∞] is fuzzy Index；For sampleWith j-th of cluster centre vjBetween mahalanobis distance, SjTo be blurred covariance square Battle array, SjFor positive definite matrix；For penalty term, ηjFor penalty factor, p is Possibility index；To score matrix T The c cluster centre clustered includes the cluster centre of nominal situation sample and the cluster of c-1 kind unusual service condition sample Center；Step 5.2: solve subordinated-degree matrix U:
By introducing Lagrange multiplier λ and γ, construction such as minor function:
Function L is asked respectively about sij、uij、vj、SjPartial derivative, obtain
By?
It willAbove formula is brought into obtain
Sj -1+(Sj -1)TIt is reversible, it solves
Step 5.2.1: initialization RWFCM algorithm parameter: determining clusters number c, sets Fuzzy Exponential m and Possibility index P, setting algorithm terminate limit ε, maximum number of iterations count, initialize the number of iterations k=1, random initializtion subordinated-degree matrix U(k)=(uij (k))n×c, random initializtion cluster centre matrix V(k)=(vj (k))c×A, random initializtion blurring covariance matrix Collect S(k)=(Sj (k))n×n×c；
Step 5.2.2: by uij (k)、vj (k)、Sj (k)Substitute into formulaCalculate kth A possibility that+1 iteration matrix B(k+1)=(sij (k+1))n×c；
Step 5.2.3: by sij (k+1)、vj (k)、Sj (k)Substitute into formulaIt calculates The subordinated-degree matrix U of+1 iteration of kth(k+1)=(uij (k+1))n×c；
Step 5.2.4: by uij (k+1)、sij (k+1)Substitute into formulaCalculate+1 iteration of kth Cluster centre matrix is V(k+1)=(vj (k+1))c×A；
Step 5.2.5: by uij (k+1)、sij (k+1)、vj (k+1)Substitute into formula Calculate the blurring covariance matrix collection S of+1 iteration of kth(k+1)=(Sj (k+1))n×n×c；Wherein, γjFor Lagrange multiplier；
Step 5.2.6: if | | U(k+1)-U(k)| | < ε or the number of iterations k > count then stops iteration, obtains final person in servitude Category degree matrix U, enters step 5.3；Otherwise, k=k+1, return step 5.2.2 are enabled；
Step 5.3: unusual service condition monitoring being carried out to sewage disposal process according to subordinated-degree matrix U: if i-th of sample μ is less than to the degree of membership of the cluster centre of nominal situation sample, then exception has occurred at i-th of sample in sewage disposal process； If exception has occurred in sewage disposal process, 6 are entered step；If there is no exceptions for sewage disposal process, terminate the base In the sewage disposal process monitoring method of KPLS and RWFCM.
In the step 5.2.1, determine that clusters number c includes:
Calculate the input sample x in input data matrix XiDot density value DiFor
Wherein,rdFor neighborhood density effective radius,
Calculating constructed fuction S (j) is
The image for drawing constructed fuction S (j), using the slope number of the image of constructed fuction S (j) as clusters number c.
Wherein, the slope of constructed fuction S (j) has reacted the dot density value of sample data, and practical significance is to construct letter Slope of number S (j) at homogeneous data is identical.In the present embodiment, the slope number of the image of constructed fuction S (j) is 3, According to analysis, can determine whether that blended data sample set is divided into 3 classes: the 1st class class1 is the 1st class unusual service condition sample class, the 2nd class Class2 is the 2nd class unusual service condition sample class, and the 3rd class class3 is nominal situation sample class, so that it is determined that clusters number c=3.
Wherein, Fuzzy Exponential m influences the fog-level of subordinated-degree matrix.In the present embodiment, Fuzzy Exponential m=2.4 is set, Enable to the effect of algorithm optimal.RWFCM is robust Weighted Fuzzy c mean cluster (robust weight fuzzy c- means clustering)。
In the present embodiment, monitor sampleTo the cluster centre v of nominal situation sample3, the 1st class it is different The cluster centre v of normal operating condition sample1, the 2nd class unusual service condition sample cluster centre v2Degree of membership ui3、ui1、ui2Respectively as schemed 2, shown in Fig. 3, Fig. 4, the Clustering Effect of score matrix T is as shown in Figure 5.Set μ=0.5.It can be seen from Fig. 2, Fig. 3, Fig. 4 At the 200th and the 800th sample, sampleThe degree of membership of the cluster centre of nominal situation sample is started less than 0.5, To judge that exception takes place at the 200th and the 800th sample in sewage disposal process.As it can be seen that monitoring method of the invention The generation of sewage disposal process unusual service condition can be monitored in time.
Step 6: establishing the linear regression model (LRM) of variable in subordinated-degree matrix U and input data matrix X, and bright using glug Day multiplier method solution variable contributes matrix N, contributes matrix N to carry out unusual service condition identification to sewage disposal process according to variable: if Contribution { η of a-th of variable to all clustersa1,...,ηacIn maximum value be ηag, then a-th of variable is abnormal with g kind The relevant failure variable of operating condition；Wherein, c is clusters number, and g ∈ { 1 ..., c-1 }, c-th of cluster is nominal situation sample Cluster.
The step 6 includes the following steps:
Step 6.1: the linear regression model (LRM) for establishing variable in subordinated-degree matrix U and input data matrix X is
Wherein, N0=(η0j)1×cFor constant term, εijFor error term, it is assumed that meet: E (εij)=0, Var (εij)=δ2, δ is Constant；xiaFor i-th of input sample x in input data matrix XiA-th of variable-value；Square is contributed for variable Battle array, ηajFor regression coefficient, ηajThe contribution that j-th is clustered for a-th of variable.
Step 6.2: variable is solved using method of Lagrange multipliers and contributes matrix N:
It wherein, is solution ηaj、η0j, introducing loss function is
ηajIndicate that j-th of cluster by the explanation degree of a-th of variable, is constrained to the introducing of above-mentioned loss function
Variable is solved using method of Lagrange multipliers and contributes matrix N, introduces Lagrange multiplier ζ, construction objective function is
Objective function L is asked respectively about η0j、ηajPartial derivative, obtain
ByBy formulaAbove formula is brought into obtain
Step 6.2.1: initialize each parameter: setting algorithm terminates limit τ, maximum number of iterations T, initializes the number of iterations k =1, random initializtion variable contributes matrix
Step 6.2.2: by ηaj (k)Substitute into formulaCalculate the N of+1 iteration of kth0 (k+1) =(η0j (k+1))1×c；
Step 6.2.3: by η0j (k+1)Substitute into formula The variable for calculating+1 iteration of kth contributes matrix
Step 6.2.4: if | | N(k+1)-N(k)| | < τ or the number of iterations k > T then stops iteration, enters step 6.3；It is no Then, k=k+1, return step 6.2.2 are enabled；
Step 6.3: contributing matrix N to carry out unusual service condition identification to sewage disposal process according to variable: if a-th of variable pair Contribution { the η of all clustersa1,...,ηacIn maximum value be ηag, then a-th of variable is event relevant to g kind unusual service condition Hinder variable；Wherein, g ∈ { 1 ..., c-1 }, c-th of cluster are the cluster of nominal situation sample.
In the present embodiment, the contribution margin that 28 sample variables cluster 3 is as shown in Figure 6.It can be identified by Fig. 6: with The relevant failure variable of 1st class unusual service condition is active heterotrophism bacteria biomass, easily biological-degradable organic substrates in biochemistry pool 1-5 It measures, the active autotrophy bacteria biomass in biochemistry pool 3-5, the nitrate nitrogen amount in biochemistry pool 1；Failure relevant to the 2nd class unusual service condition Variable is flow of inlet water, influent ammonia nitrogen amount, the basicity in biochemistry pool 1-5, the nitrate nitrogen amount in biochemistry pool 2, molten in biochemistry pool 3-5 Ammonia nitrogen amount in solution oxygen amount, biochemistry pool 3-5.As it can be seen that method of the invention can identify that the failure of every kind of unusual service condition becomes in time Amount.
Obviously, above-described embodiment is only a part of the embodiments of the present invention, instead of all the embodiments.Above-mentioned implementation Example for explaining only the invention, is not intended to limit the scope of the present invention..Based on the above embodiment, those skilled in the art Member's every other embodiment obtained namely all in spirit herein and original without making creative work Made all modifications, equivalent replacement and improvement etc., are all fallen within the protection domain of application claims within reason.

Claims (8)

1. a kind of sewage disposal process monitoring method based on KPLS and RWFCM, which is characterized in that include the following steps:
Step 1: acquiring nominal situation and the sewage disposal process data sample comprising unusual service condition, the sewage treatment respectively Journey data sample includes m1A sewage treatment runs variable, m2A effluent characteristics variable；From time angle by the sewage of nominal situation Process data sample is added in front of the sewage disposal process data sample comprising unusual service condition, forms blended data sample Collection；By m in blended data sample set1The data of a sewage treatment operation variable are as input data matrix X, by blended data sample This concentration m2The data of a effluent characteristics variable are as output data matrix Y；
Step 2: input data matrix X and output data matrix Y are pre-processed；The pretreatment includes calculating input data The mean value and standard deviation of each variable in matrix X and output data matrix Y, and input data matrix X and output data matrix Y is equal It is standardized into the data of zero-mean and unity standard deviation；
Step 3: the input sample x in input data matrix X is mapped to height by the KPLS model of building sewage disposal process monitoring Dimensional feature space F:x → Φ (x) ∈ F introduces gaussian kernel function and obtains the Gram matrix K of input data matrix X, and to Gram square Battle array K carries out centralization processing；
Step 4: pivot number being determined using cross-validation method, and solves score matrix T；
Step 5: score matrix T being clustered based on RWFCM algorithm, subordinated-degree matrix U is obtained, according to U pairs of subordinated-degree matrix Sewage disposal process carries out unusual service condition monitoring: if a certain moment sample is small to the degree of membership of the cluster centre of nominal situation sample In μ, then exception has occurred at the sample in sewage disposal process；
Step 6: establishing the linear regression model (LRM) of variable in subordinated-degree matrix U and input data matrix X, and multiplied using Lagrange Sub- method solves variable and contributes matrix N, contributes matrix N to carry out unusual service condition identification to sewage disposal process according to variable: if a-th Contribution { η of the variable to all clustersa1,...,ηacIn maximum value be ηag, then a-th variable be and g kind unusual service condition phase The failure variable of pass；Wherein, c is clusters number, and g ∈ { 1 ..., c-1 }, c-th of cluster is the cluster of nominal situation sample.
2. the sewage disposal process monitoring method according to claim 1 based on KPLS and RWFCM, which is characterized in that institute Sewage disposal process is stated using activated sludge process, raw sewage, into biochemistry pool part, carries out biological denitrificaion after coagulation Afterwards, a part is entered secondary settling tank and is precipitated by inner circulating reflux denitrogenation again, another part；Biochemistry pool part includes biochemistry Pond l ∈ { 1,2,3,4,5 }, wherein biochemistry pool l1∈ { 1,2 } is the main anoxic zone for completing anti-nitration reaction process, biochemistry pool l2 ∈ { 3,4,5 } is the main aerobic zone for completing nitration reaction process；In the step 1, the m1A sewage treatment runs variable Including in flow of inlet water, influent ammonia nitrogen amount, biochemistry pool l ∈ { 1,2,3,4,5 } active heterotrophism bacteria biomass, biochemistry pool l ∈ 1, 2,3,4,5 } the easily biological-degradable organic substrates amount in, the basicity in biochemistry pool l ∈ { 1,2,3,4,5 }, biochemistry pool l1∈{1,2} In nitrate nitrogen amount, biochemistry pool l2Active autotrophy bacteria biomass, biochemistry pool l in ∈ { 3,4,5 }2Ammonia nitrogen amount in ∈ { 3,4,5 }, Biochemistry pool l2Dissolved oxygen content in ∈ { 3,4,5 }；The m2A effluent characteristics variable includes biochemical oxygen demand (BOD), COD, hangs Floating object, water outlet ammonia nitrogen amount.
3. the sewage disposal process monitoring method according to claim 1 or 2 based on KPLS and RWFCM, which is characterized in that The step 3 includes the following steps:
Step 3.1: the KPLS model of building sewage disposal process monitoring is
Φ=TP1'+Φr
Y=TQ'+Yr
Step 3.2: the input sample x in input data matrix X being mapped to high-dimensional feature space F:x → Φ (x) ∈ F, is introduced Gaussian kernel function obtains the Gram matrix K of input data matrix X, and carries out centralization processing to Gram matrix K, by KPLS model It is converted to
K=TP2'+E
Y=TQ'+Yr
Wherein, the element of the i-th row jth column of Gram matrix K is Kij=k (xi,xj)=< Φ (xi),Φ(xj) >, xi、xjRespectively I-th of input sample x in input data matrix Xi, j-th of input sample xj, k (xi,xj) be gaussian kernel function, i, j ∈ 1, 2 ..., n }, n is the number of samples in input data matrix X；T is high dimensional data Φ={ Φ (xi),i∈{1,2,...,n}} Score matrix, T=[t1,...,tA], A is pivot number, P1=[p11,...,p1A]、P2=[p21,...,p2A], Q= [q1,...,qA] be respectively matrix Φ, Gram matrix K, output data matrix Y loading matrix, Φr、E、YrRespectively matrix The modeling residual error of Φ, Gram matrix K, output data matrix Y.
4. the sewage disposal process monitoring method according to claim 3 based on KPLS and RWFCM, which is characterized in that institute It states in step 4, pivot number A is determined using cross-validation method, and solve score matrix T, included the following steps:
Step 4.1: enabling u is the either rank of output data matrix Y；
Step 4.2: calculating score vector: t=Ku；
Step 4.3: by score vector t normalized: | | t | | → 1；
Step 4.4: being respectively listed on score vector t in output data matrix Y being returned: q=Y't；
Step 4.5: calculating the new score of output data matrix Y: u=Yq；
Step 4.6: by u vector normalized: | | u | | → 1；
Step 4.7: judging whether u restrains: if it is, jumping to step 4.8；If it is not, then jumping to step 4.2；
Step 4.8: update matrix: K=(I-tt') K (I-tt'), Y=Y-tq' repeat step 4.2 to step 4.7, carry out down The calculating of one score vector, until A score vector is extracted；Wherein, I is unit matrix.
5. the sewage disposal process monitoring method according to claim 3 based on KPLS and RWFCM, which is characterized in that institute It states in step 3, centralization treated Gram matrixWherein, EnFor the unit of n × n Matrix, 1nComplete 1 column vector, 1' are tieed up for nnIt is 1nTransposed matrix.
6. the sewage disposal process monitoring method according to claim 4 based on KPLS and RWFCM, which is characterized in that institute Step 5 is stated to include the following steps:
Step 5.1: score matrix T being clustered based on RWFCM algorithm, building RWFCM objective function is
Wherein,For i-th of row vector of score matrix T,For m1The input sample x of dimensioniA after corresponding dimensionality reduction reforms Sample, uijFor sampleTo j-th of cluster centre vjDegree of membership, sijFor sampleA possibility that belonging to j-th of cluster, Subordinated-degree matrix U=(uij)n×c, cluster centre matrix V=(vj)c×A, c is clusters number；M ∈ [1 ,+∞] is Fuzzy Exponential；For sampleWith j-th of cluster centre vjBetween mahalanobis distance, SjTo be blurred covariance matrix, Sj For positive definite matrix；For penalty term, ηjFor penalty factor, p is Possibility index；Score matrix T is carried out Clustering c obtained cluster centre includes the cluster centre of nominal situation sample and the cluster centre of c-1 kind unusual service condition sample；
Step 5.2: solve subordinated-degree matrix U:
Step 5.2.1: initialization RWFCM algorithm parameter: determining clusters number c, sets Fuzzy Exponential m and Possibility index p, if It sets algorithm and terminates limit ε, maximum number of iterations count, initialize the number of iterations k=1, random initializtion subordinated-degree matrix U(k)= (uij (k))n×c, random initializtion cluster centre matrix V(k)=(vj (k))c×A, random initializtion blurring covariance matrix collection S(k) =(Sj (k))n×n×c；
Step 5.2.2: by uij (k)、vj (k)、Sj (k)Substitute into formulaIt calculates kth+1 time A possibility that iteration matrix B(k+1)=(sij (k+1))n×c；
Step 5.2.3: by sij (k+1)、vj (k)、Sj (k)Substitute into formulaCalculating kth+ The subordinated-degree matrix U of 1 iteration(k+1)=(uij (k+1))n×c；
Step 5.2.4: by uij (k+1)、sij (k+1)Substitute into formulaCalculate the cluster of+1 iteration of kth Center matrix is V(k+1)=(vj (k+1))c×A；
Step 5.2.5: by uij (k+1)、sij (k+1)、vj (k+1)Substitute into formula Calculate the blurring covariance matrix collection S of+1 iteration of kth(k+1)=(Sj (k+1))n×n×c；Wherein, γjFor Lagrange multiplier；
Step 5.2.6: if | | U(k+1)-U(k)| | < ε or the number of iterations k > count then stops iteration, obtains final degree of membership Matrix U enters step 5.3；Otherwise, k=k+1, return step 5.2.2 are enabled；
Step 5.3: unusual service condition monitoring being carried out to sewage disposal process according to subordinated-degree matrix U: if i-th of sampleTo just The degree of membership of the cluster centre of normal operating condition sample is less than μ, then exception has occurred at i-th of sample in sewage disposal process；If dirty Exception has occurred in water treatment procedure, then enters step 6；If sewage disposal process there is no exception, terminates described be based on The sewage disposal process monitoring method of KPLS and RWFCM.
7. the sewage disposal process monitoring method according to claim 6 based on KPLS and RWFCM, which is characterized in that institute It states in step 5.2.1, determines that clusters number c includes:
Calculate the input sample x in input data matrix XiDot density value DiFor
Wherein,rdFor neighborhood density effective radius,
Calculating constructed fuction S (j) is
The image for drawing constructed fuction S (j), using the slope number of the image of constructed fuction S (j) as clusters number c.
8. the sewage disposal process monitoring method according to claim 6 based on KPLS and RWFCM, which is characterized in that institute Step 6 is stated to include the following steps:
Step 6.1: the linear regression model (LRM) for establishing variable in subordinated-degree matrix U and input data matrix X is
Wherein, N0=(η0j)1×cFor constant term, εijFor error term, it is assumed that meet: E (εij)=0, Var (εij)=δ2, δ is normal Number；xiaFor i-th of input sample x in input data matrix XiA-th of variable-value；Square is contributed for variable Battle array, ηajFor regression coefficient, ηajThe contribution that j-th is clustered for a-th of variable；
Step 6.2: variable is solved using method of Lagrange multipliers and contributes matrix N:
Step 6.2.1: initialize each parameter: setting algorithm terminates limit τ, maximum number of iterations T, initializes the number of iterations k=1, Random initializtion variable contributes matrix
Step 6.2.2: by ηaj (k)Substitute into formulaCalculate the N of+1 iteration of kth0 (k+1)= (η0j (k+1))1×c；
Step 6.2.3: by η0j (k+1)Substitute into formula The variable for calculating+1 iteration of kth contributes matrix
Step 6.2.4: if | | N(k+1)-N(k)| | < τ or the number of iterations k > T then stops iteration, enters step 6.3；Otherwise, Enable k=k+1, return step 6.2.2；
Step 6.3: contributing matrix N to carry out unusual service condition identification to sewage disposal process according to variable: if a-th of variable is to all Contribution { the η of clustera1,...,ηacIn maximum value be ηag, then a-th of variable is failure relevant to g kind unusual service condition change Amount；Wherein, g ∈ { 1 ..., c-1 }, c-th of cluster are the cluster of nominal situation sample.
CN201910573311.XA 2019-06-28 2019-06-28 KPLS (kernel principal component system) and RWFCM (wireless remote control unit) -based sewage treatment process monitoring method Active CN110232256B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910573311.XA CN110232256B (en) 2019-06-28 2019-06-28 KPLS (kernel principal component system) and RWFCM (wireless remote control unit) -based sewage treatment process monitoring method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910573311.XA CN110232256B (en) 2019-06-28 2019-06-28 KPLS (kernel principal component system) and RWFCM (wireless remote control unit) -based sewage treatment process monitoring method

Publications (2)

Publication Number Publication Date
CN110232256A true CN110232256A (en) 2019-09-13
CN110232256B CN110232256B (en) 2022-11-15

Family

ID=67856526

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910573311.XA Active CN110232256B (en) 2019-06-28 2019-06-28 KPLS (kernel principal component system) and RWFCM (wireless remote control unit) -based sewage treatment process monitoring method

Country Status (1)

Country Link
CN (1) CN110232256B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021109180A1 (en) * 2019-12-03 2021-06-10 北京工业大学 Sewage treatment process fault monitoring method based on fuzzy width adaptive learning model
CN113222324A (en) * 2021-03-13 2021-08-06 宁波大学科学技术学院 Sewage quality monitoring method based on PLS-PSO-RBF neural network model
CN113223633A (en) * 2021-03-13 2021-08-06 宁波大学科学技术学院 Width GRNN model-based papermaking process drain water quality prediction method
CN113468479A (en) * 2021-06-16 2021-10-01 北京科技大学 Data-driven cold continuous rolling industrial process monitoring and anomaly detection method
CN116092078A (en) * 2023-04-11 2023-05-09 深圳市信远环保水务有限公司 Intelligent control system for sewage treatment
CN116110516A (en) * 2023-04-14 2023-05-12 青岛山青华通环境科技有限公司 Method and device for identifying abnormal working conditions in sewage treatment process

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0676070A1 (en) * 1992-12-23 1995-10-11 Baleanu Michael Alin Method and device for signal analysis, process identification and monitoring of a technical process.
CN106600605A (en) * 2016-12-14 2017-04-26 陕西科技大学 Unsupervised fast image segmentation algorithm
CN107463093A (en) * 2017-07-13 2017-12-12 东北大学 A kind of blast-melted quality monitoring method based on KPLS robust reconstructed errors
US20180276531A1 (en) * 2017-03-27 2018-09-27 Beijing University Of Technology Fault Identifying Method for Sludge Bulking Based on a Recurrent RBF Neural Network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0676070A1 (en) * 1992-12-23 1995-10-11 Baleanu Michael Alin Method and device for signal analysis, process identification and monitoring of a technical process.
CN106600605A (en) * 2016-12-14 2017-04-26 陕西科技大学 Unsupervised fast image segmentation algorithm
US20180276531A1 (en) * 2017-03-27 2018-09-27 Beijing University Of Technology Fault Identifying Method for Sludge Bulking Based on a Recurrent RBF Neural Network
CN107463093A (en) * 2017-07-13 2017-12-12 东北大学 A kind of blast-melted quality monitoring method based on KPLS robust reconstructed errors

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021109180A1 (en) * 2019-12-03 2021-06-10 北京工业大学 Sewage treatment process fault monitoring method based on fuzzy width adaptive learning model
US11403546B2 (en) 2019-12-03 2022-08-02 Beijing University Of Technology Fault monitoring method for sewage treatment process based on fuzzy width adaptive learning model
CN113222324A (en) * 2021-03-13 2021-08-06 宁波大学科学技术学院 Sewage quality monitoring method based on PLS-PSO-RBF neural network model
CN113223633A (en) * 2021-03-13 2021-08-06 宁波大学科学技术学院 Width GRNN model-based papermaking process drain water quality prediction method
CN113222324B (en) * 2021-03-13 2023-04-07 宁波大学科学技术学院 Sewage quality monitoring method based on PLS-PSO-RBF neural network model
CN113223633B (en) * 2021-03-13 2024-04-05 宁波大学科学技术学院 Width GRNN model-based water quality prediction method for sewage discharge outlet in papermaking process
CN113468479A (en) * 2021-06-16 2021-10-01 北京科技大学 Data-driven cold continuous rolling industrial process monitoring and anomaly detection method
CN113468479B (en) * 2021-06-16 2023-08-08 北京科技大学 Cold continuous rolling industrial process monitoring and abnormality detection method based on data driving
CN116092078A (en) * 2023-04-11 2023-05-09 深圳市信远环保水务有限公司 Intelligent control system for sewage treatment
CN116092078B (en) * 2023-04-11 2023-06-09 深圳市信远环保水务有限公司 Intelligent control system for sewage treatment
CN116110516A (en) * 2023-04-14 2023-05-12 青岛山青华通环境科技有限公司 Method and device for identifying abnormal working conditions in sewage treatment process

Also Published As

Publication number Publication date
CN110232256B (en) 2022-11-15

Similar Documents

Publication Publication Date Title
CN110232256A (en) A kind of sewage disposal process monitoring method based on KPLS and RWFCM
Farhi et al. Prediction of wastewater treatment quality using LSTM neural network
Gibert et al. Knowledge discovery with clustering based on rules by states: A water treatment application
Han et al. Hierarchical neural network modeling approach to predict sludge volume index of wastewater treatment process
CN109492265B (en) Wastewater effluent index prediction method based on dynamic nonlinear PLS soft measurement method
CN110232062A (en) A kind of sewage disposal process monitoring method based on KPLS and FCM
Sin et al. Improving the Morris method for sensitivity analysis by scaling the elementary effects
Banadda et al. A review of modeling approaches in activated sludge systems
Chen et al. Research progress on integrated treatment technologies of rural domestic sewage: a review
Qambar et al. Optimizing dissolved oxygen requirement and energy consumption in wastewater treatment plant aeration tanks using machine learning
CN108088974A (en) A kind of flexible measurement method of anaerobism while denitrification methane phase process water outlet nitrate nitrogen
Ba-Alawi et al. Missing data imputation and sensor self-validation towards a sustainable operation of wastewater treatment plants via deep variational residual autoencoders
CN103809436A (en) Method for intelligent modeling sewage disposal process using activated sludge process
Torregrossa et al. A data analysis approach to evaluate the impact of the capacity utilization on the energy consumption of wastewater treatment plants
Kocijan et al. Sequencing batch-reactor control using Gaussian-process models
Zhong et al. Water quality prediction of MBR based on machine learning: A novel dataset contribution analysis method
Xu et al. Prediction and evaluation of indirect carbon emission from electrical consumption in multiple full-scale wastewater treatment plants via automated machine learning-based analysis
Yang et al. Cloud-model-based feature engineering to analyze the energy–water nexus of a full-scale wastewater treatment plant
Qambar et al. Development of local and global wastewater biochemical oxygen demand real-time prediction models using supervised machine learning algorithms
Singh et al. Performance prediction of horizontal flow constructed wetlands by employing machine learning
Xu et al. A novel long short-term memory artificial neural network (LSTM)-based soft-sensor to monitor and forecast wastewater treatment performance
CN111204867B (en) Membrane bioreactor-MBR membrane pollution intelligent decision-making method
Comas Matas Development, implementation and evaluation of an activated sludge supervisory system for the granollers WWTP
Beraud et al. Multiobjective genetic algorithms for the optimisation of wastewater treatment processes
CN116040892A (en) Domestic garbage sewage treatment method

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