CN112102891B - Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering - Google Patents

Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering Download PDF

Info

Publication number
CN112102891B
CN112102891B CN202010771841.8A CN202010771841A CN112102891B CN 112102891 B CN112102891 B CN 112102891B CN 202010771841 A CN202010771841 A CN 202010771841A CN 112102891 B CN112102891 B CN 112102891B
Authority
CN
China
Prior art keywords
energy consumption
energy
abnormal
generalization
dissimilarity
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
Application number
CN202010771841.8A
Other languages
Chinese (zh)
Other versions
CN112102891A (en
Inventor
杨海东
印四华
徐康康
朱成就
曾超湛
胡罗克
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.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN202010771841.8A priority Critical patent/CN112102891B/en
Publication of CN112102891A publication Critical patent/CN112102891A/en
Application granted granted Critical
Publication of CN112102891B publication Critical patent/CN112102891B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/10—Analysis or design of chemical reactions, syntheses or processes
    • C—CHEMISTRY; METALLURGY
    • C03—GLASS; MINERAL OR SLAG WOOL
    • C03B—MANUFACTURE, SHAPING, OR SUPPLEMENTARY PROCESSES
    • C03B5/00—Melting in furnaces; Furnaces so far as specially adapted for glass manufacture
    • C03B5/16—Special features of the melting process; Auxiliary means specially adapted for glass-melting furnaces
    • C—CHEMISTRY; METALLURGY
    • C03—GLASS; MINERAL OR SLAG WOOL
    • C03B—MANUFACTURE, SHAPING, OR SUPPLEMENTARY PROCESSES
    • C03B5/00—Melting in furnaces; Furnaces so far as specially adapted for glass manufacture
    • C03B5/16—Special features of the melting process; Auxiliary means specially adapted for glass-melting furnaces
    • C03B5/24—Automatically regulating the melting process
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/231—Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/70—Machine learning, data mining or chemometrics

Landscapes

  • Engineering & Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Materials Engineering (AREA)
  • Computing Systems (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Evolutionary Computation (AREA)
  • Organic Chemistry (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Analytical Chemistry (AREA)
  • Chemical Kinetics & Catalysis (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to the technical field of glass processing, in particular to a horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering, which comprises the following steps: s10, constructing an energy generalization hierarchical structure by using a root cause analysis method by taking a layered energy model of the horseshoe flame glass melting furnace as a prototype; s20, calculating dissimilarity of energy composition item attributes based on the energy generalization hierarchical structure of the step S10, and accumulating the dissimilarity of all the energy composition item attributes to represent the dissimilarity d (C) of the abnormal energy consumption samplei,Cj) (ii) a Obtaining a generalization abnormal energy consumption sample; and S30, establishing an energy consumption abnormity positioning model based on a hierarchical clustering algorithm, inputting the generalized abnormal energy consumption samples obtained in the step S20, and outputting an energy composition item set represented in a generalized manner. The invention obtains the generalized energy composition item from the clustering result and rapidly positions the affiliated parts, thereby effectively solving the problem that the abnormal energy consumption positioning precision is influenced by the data acquisition delay.

Description

Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering
Technical Field
The invention relates to the technical field of glass processing, in particular to a horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering.
Background
The glass melting furnace is used as core thermal equipment for glass production, the energy consumption of the glass melting furnace accounts for more than 80% of the energy of the whole plant, and the energy cost accounts for more than 50% of the total production cost. When glass is produced, huge energy consumption of the glass is always a difficult problem which puzzles energy-saving production of enterprises. In particular, in the periodic production process, the yield of glass products is reduced due to factors such as unstable fuel supply and unstable batch supply, which directly results in energy loss. The abnormal energy consumption loss is an important incentive for causing the energy waste in kiln production, and the timely discovery of the abnormal kiln is an important link for the safety production and the energy saving of glass enterprises. At present, the energy consumption information is often combined to perform abnormal positioning, the abnormal positioning of the energy consumption information is to find an energy composition item with abnormal energy consumption value fluctuation by analyzing various energy composition items of kiln equipment, and then to position related equipment components to solve the abnormality in time, so that the accidental loss of energy is avoided. However, since the characteristic dimension of the energy consumption information is too large, a worker cannot locate the position of the abnormal part and the cause of the abnormality at the first time, thereby delaying the best time to solve the problem.
Chinese patent CN110378371A discloses an energy consumption anomaly detection method based on average neighbor distance anomaly factor, which includes: acquiring energy consumption data and converting the energy consumption data into alternating data; defining a time sequence characteristic value of the energy consumption data, and dividing the time sequence into subsequences which are respectively mapped to a four-dimensional characteristic space; respectively calculating average neighbor distance abnormal factors of the time subsequences in a four-dimensional feature space; processing the average neighbor distance abnormal factor of the subsequence to obtain an average neighbor distance abnormal factor of the time sequence; and calculating an average neighbor distance abnormal threshold according to the average neighbor distance abnormal factor, and judging whether the mode is abnormal. Although the scheme can eliminate the interference of the mode abnormity detection, effectively improve the precision of the mode abnormity detection and accurately position the abnormal position, the energy consumption data is often acquired by measuring and acquiring working condition information through the sensing test instrument, and the sensing test instrument can cause the problems of data acquisition delay, deviation and the like when being applied to the glass melting furnace due to factors such as installation position, network environment, service life and the like.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides a horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering, and can quickly and accurately find abnormity causes and improve troubleshooting efficiency.
In order to solve the technical problems, the invention adopts the technical scheme that:
the method for positioning the abnormal energy consumption of the horseshoe flame glass melting furnace based on root cause analysis hierarchical clustering comprises the following steps:
s10, constructing an energy generalization hierarchical structure by using a root cause analysis method by taking a layered energy model of the horseshoe flame glass melting furnace as a prototype; the energy generalization hierarchical structure is a tree generalization hierarchical structure with the energy flow characteristic of the horseshoe flame glass melting furnace, an input energy system flows from top to bottom in a layered energy model, and leaf nodes form a final energy output system;
s20, calculating dissimilarity of energy composition item attributes based on the energy generalization hierarchical structure of the step S10, and accumulating the dissimilarity of all the energy composition item attributes to represent the dissimilarity d (C) of the abnormal energy consumption samplei,Cj) (ii) a Obtaining a generalization abnormal energy consumption sample;
and S30, establishing an energy consumption abnormity positioning model based on a hierarchical clustering algorithm, inputting the generalized abnormal energy consumption samples obtained in the step S20, and outputting an energy composition item set represented in a generalized manner.
According to the horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering, the generalized energy composition item is obtained from the clustering result, the affiliated part is quickly positioned, and the problem that the abnormal energy consumption positioning precision is influenced by data acquisition delay can be effectively solved.
Preferably, in the layered energy model of step S10, the input energy system flows from top to bottom, energy is input by the small furnace, passes through the flame space and is output by the melting tank and the regenerator, and energy components of the small furnace, the flame space, the regenerator and the melting tank respectively form an energy hierarchy.
Preferably, the energy generalization hierarchy in step S10 is performed according to the following steps:
s11, determining characteristic variables and value domains Dom (c)x): heat Q generated by combustion of fuel in small furnaceqAs root point, heat Q absorbed by the glass surface in the flame spacelevelHeat Q dissipated by arch top and kiln wallcomb,wallHigh-temperature flue gas takes away heat Qreg,flueAs branch nodes, melting bathsThe inner glass liquid takes away heat QglassHeat of reaction heat consumption Q in melting glassreactionThe generated gas of the glass melting reaction takes away heat Qbatch,airHeat quantity Q dissipated from bottom of melting tank and kiln wallmelt,wallAs a leaf node, the combustion air in the heat storage tank brings physical sensible heat Qreg,airThe waste gas takes away heat QwgasHeat Q of heat dissipation of wall of heat storage chamberreg,wallHeat loss heat Q of grid body furnace dustbrickAs leaf nodes;
s12, determining a hierarchical relationship: qreg,flueInput energy system as regenerator and output energy system Qreg,air、Qwgas、Qreg,wallAnd QbrickSum of the junction values of (1), Qreg,flueIs Qreg,air、Qwgas、Qreg,wallAnd QbrickA parent node of (a); qlevelInput energy system as melting tank with output energy system Qglass、Qreaction、Qbatch,airAnd Qmelt,wallSum of the junction values of (1), QlevelIs Qglass、Qreaction、Qbatch,airAnd Qmelt,wallA parent node of (a); qqAs input energy system of small furnace, its value is output energy system Qlevel、Qcomvb,wall、Qreg,flueThe sum of the junction values of (a); qx、QbatchRespectively representing the physical sensible heat of the fuel and the mixture, and being a generalized hierarchical structure only comprising root nodes;
s13, generating an energy generalization hierarchical structure set G ═ G1,G2,G3}: for attribute cxThe method comprises the steps of establishing a corresponding generalization hierarchical structure Gx by utilizing a root cause analysis method, wherein the Gx structure is a data structure tree, and a root node is the topmost generalization result which can be obtained by each sample and has uniqueness.
Preferably, in step S12, each node stores the information of the parent node or the ancestor node, and Q isx、QbatchAre respectively denoted as c2、c3，QqIs denoted as c1，QlevelIs denoted as c1、c4，Qcomvb,wallIs denoted as c1、c5，Qreg,flueIs denoted as c1、c6，QglassIs denoted as c1、c4、c7，QreactionIs denoted as c1、c4、c8，Qbatch,airIs denoted as c1、c4、c9，Qmelt,wallIs denoted as c1、c4、c10，Qreg,airIs denoted as c1、c4、c11，QwgasIs denoted as c1、c4、c12，Qreg,wallIs denoted as c1、c4、c13，QbrickIs denoted as c1、c4、c14(ii) a The node information is a characteristic variable.
Preferably, step S20 is performed according to the following steps:
s21, acquiring an abnormal energy consumption sample set A and an energy consumption sample C of the energy management systemiEnergy consumption sample CiEnergy composition item attribute of cx，Ci＝{c1,c2,…,cm}：
Figure GDA0003529534720000031
Figure GDA0003529534720000032
Respectively is a row vector of the abnormal energy consumption sample set A;
transposing the matrix A:
Figure GDA0003529534720000033
calculating the mathematical expectation μ for each feature attribute cx in the ATxSum variance σx：
Figure GDA0003529534720000034
Figure GDA0003529534720000035
ci,xA mathematical expectation representing the ith energy contribution;
s22, giving a significance level a and setting a random variable XxEnergy composition item attribute c representing a samplexSequentially calculating abnormal energy consumption samples CiEach energy composition item attribute cxThe chebyshev inequality of (c):
Figure GDA0003529534720000041
s23, using the reasonable interval principle of Chebyshev inequality, the method is suitable for the application
Figure GDA0003529534720000042
When the energy composition item exceeds the value range of the normal category, the attributes have no similarity; the attribute c is generalized according to the energy generalization hierarchy GxIs replaced by GxIn (c)xTo obtain a generalized representation of the abnormal energy consumption sample Ci(ii) a On the contrary, no generalization operation is performed and the attribute c is calculatedxDegree of dissimilarity δ ofx(x1)；
S24, repeating the steps S22 and S23 until the dissimilarity degree delta of all the energy composition item attributes is calculated, accumulating the dissimilarity degrees delta of all the attributes and representing an abnormal energy consumption sample CiDissimilarity of (a):
Figure GDA0003529534720000043
d(Ci[cx]) The dissimilarity of the samples of abnormal energy consumption represented for generalization.
Preferably, the hierarchical clustering algorithm described in step S30 includes the following steps:
a. obtaining an original abnormal energy consumption sample set A ═ C1,C2,…,CkH, selecting a certain attribute cxAll abnormal energy consumption sample attributes c in AxValue is replaced by GxIn (c)xA parent value of;
b. repeating the step a, continuously generalizing the abnormal energy consumption until a subset L of A is found, wherein the number of elements in the subset L is | L | ≧ minisize, and H (L) is minimum, and terminating the generalization operation;
c. the generalized representation of the energy composition item set L is output.
Preferably, step S30 is performed as follows:
s31, importing an abnormal energy consumption sample set A, and calculating the attribute c of each energy composition itemxMathematical expectation of (a)xSum variance σx；
S32, selecting an abnormal energy consumption sample C from the abnormal energy consumption sample set Ai；
S33, selecting attributes c from bottom to top in sequence according to the energy generalization hierarchical structurex(ii) a Judging the attribute c by the Chebyshev inequalityxValue of (d) Dom (c)x) Whether the reasonable interval requirement of the Chebyshev inequality is met: if yes, let cxReplacement to Gx in cxTo obtain a generalized representation of the abnormal energy consumption sample CiCumulative equivalence and CiAnd is stored in A and the original C is deletediCalculating the attribute dissimilarity δx；
S34, repeatedly executing the step S33 to determine the dissimilarity delta of all the attributesxCumulatively represent energy consumption samples CiDegree of dissimilarity d ofx；
S35, repeatedly executing the steps S32-S34 until the generalization process of all the abnormal energy consumption samples and the dissimilarity calculation are completed; searching L, merging all samples with the same generalization representation abnormal energy consumption, and generalizing the tableAverage dissimilarity of anomalously powered clusters
Figure GDA0003529534720000051
Representing the current generalization effect;
s36, searching whether the element number of the abnormal energy consumption cluster represented in the generalization mode is larger than the minsize parameter or not, and the average dissimilarity degree of the current abnormal energy consumption cluster
Figure GDA0003529534720000052
Is the minimum value; if the condition is not met, the steps S32-S36 are repeatedly executed;
and S37, terminating the hierarchical clustering iteration condition, and outputting the energy represented in a generalization manner to form an item set A.
Preferably, in step S36, the minsize parameter is adapted as follows: setting the initial value minszie to ms0Then, a smaller value w is defined, w ∈ [0,1 ]](ii) a When ms is0、ms0·(1-w)、ms0When the clustering effect of (1+ w) is the same, ms is considered0The method is the best value of the current data scale.
Preferably, the initial value size of minsize is 1/5 for the abnormal energy consumption sample set size.
Preferably, in step S23, a reasonable interval of each feature attribute is determined according to the range value e of the deviation mean, and if e is set to be 2 times of the standard deviation, then there is
Figure GDA0003529534720000053
Indicating that the probability of values other than (m-2s, m +2s) is only 0.5%.
Compared with the prior art, the invention has the beneficial effects that:
according to the horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering, the generalized energy composition item is obtained from the clustering result, the affiliated part is quickly positioned, and the problem that the abnormal energy consumption positioning precision is influenced by data acquisition delay can be effectively solved.
Drawings
FIG. 1 is a flowchart of an algorithm for generalizing hierarchical clustering for abnormal energy consumption;
FIG. 2 is an energy generalization hierarchy for a horseshoe flame glass melter;
FIG. 3 is a flow chart of a horseshoe flame glass furnace energy consumption anomaly location model.
Detailed Description
The present invention will be further described with reference to the following embodiments.
Examples
Fig. 1 to 2 show an embodiment of the method for positioning abnormal energy consumption of horseshoe flame glass melting furnace based on root cause analysis hierarchical clustering, which comprises the following steps:
s10, constructing an energy generalization hierarchical structure by using a root cause analysis method by taking a layered energy model of the horseshoe flame glass melting furnace as a prototype; the energy generalization hierarchical structure is a tree generalization hierarchical structure with the horseshoe flame glass melting furnace energy flow characteristic, an input energy system flows from top to bottom in a hierarchical energy model, leaf nodes form a final energy output system, and each abnormal energy consumption sample has a unique generalization result at the topmost layer due to the characteristics of the tree structure; wherein the energy generalization hierarchy can be defined as:
G＝(V,E)
wherein G is a directed tree with the related properties and definitions of the tree; v is a tree node set and represents a heat balance energy composition item of each subsystem of the horseshoe flame glass melting furnace; and E is an edge set of the tree, nodes in the tree are connected by edges and represent the energy flow relationship among the kiln subsystems and the dependency relationship of upper and lower levels of energy composition items.
S20, calculating dissimilarity of energy composition item attributes based on the energy generalization hierarchical structure of the step S10, and accumulating the dissimilarity of all the energy composition item attributes to represent the dissimilarity d (C) of the abnormal energy consumption samplei,Cj) (ii) a Obtaining a generalization abnormal energy consumption sample;
and S30, establishing an energy consumption abnormity positioning model based on a hierarchical clustering algorithm, inputting the generalized abnormal energy consumption sample obtained in the step S20, and outputting an energy composition item set represented in a generalized manner.
In the layered energy model described in step S10, the input energy system flows from top to bottom, energy is input by the port, passes through the flame space, and is output by the melting tank and the regenerator, and energy components of the port, the flame space, the regenerator, and the melting tank respectively form an energy hierarchy.
The energy generalization hierarchy in step S10 is performed according to the following steps:
s11, determining characteristic variables and value domains Dom (c)x): heat Q generated by combustion of fuel in small furnaceqAs root point, heat Q absorbed by the glass surface in the flame spacelevelHeat Q dissipated by arch top and kiln wallcomb,wallHigh-temperature flue gas takes away heat Qreg,flueAs a branch node, the molten glass in the melting tank takes away heat QglassHeat of reaction heat consumption Q in melting glassreactionThe generated gas of the glass melting reaction takes away heat Qbatch,airHeat quantity Q dissipated from bottom of melting tank and kiln wallmelt,wallAs a leaf node, the combustion air in the heat storage tank brings physical sensible heat Qreg,airThe waste gas takes away heat QwgasAnd heat quantity Q of heat dissipation of wall of heat storage chamberreg,wallHeat loss heat Q of grid body furnace dustbrickAs leaf nodes;
s12, determining a hierarchical relationship: qreg,flueInput energy system as regenerator and output energy system Qreg,air、Qwgas、Qreg,wallAnd QbrickSum of the junction values of (c), Qreg,flueIs Qreg,air、Qwgas、Qreg,wallAnd QbrickA parent node of (2); qlevelInput energy system as melting tank with output energy system Qglass、Qreaction、Qbatch,airAnd Qmelt,wallSum of the junction values of (1), QlevelIs Qglass、Qreaction、Qbatch,airAnd Qmelt,wallA parent node of (a); qqAs input energy system of small furnace, its value is output energy system Qlevel、Qcomvb,wall、Qreg,flueThe sum of the junction values of (a); qx、QbatchRespectively represents the physical sensible heat of the fuel and the mixture and only comprises root knotsA generalized hierarchical structure of points; wherein Q isqAs energy constituting terms of the small furnace, Qlevel、Qcomvb,wall、Qreg,flueBeing an energy component of the flame space, Qglass、Qreaction、Qbatch,airAnd Qmelt,wallFor the energy composition of the melting bath, Qreg,air、Qwgas、Qreg,wallAnd QbrickIs an energy component of the regenerator.
S13, generating an energy generalization hierarchical structure set G ═ G1,G2,G3}，G1,G2,G3Respectively, the substructures of the energy generalization hierarchy: for attribute cxThe method comprises the steps of establishing a corresponding generalization hierarchical structure Gx by utilizing a root cause analysis method, wherein the Gx structure is a data structure tree, and a root node is the topmost generalization result which can be obtained by each sample and has uniqueness.
In order to avoid repeatedly traversing the hierarchical tree to find a father node in the generalization process, each node in the tree stores information of the father node or an ancestor node, specifically: qx、QbatchIs respectively represented as c2、c3，QqIs denoted as c1，QlevelIs denoted as c1、c4，Qcomvb,wallIs denoted as c1、c5，Qreg,flueIs denoted as c1、c6，QglassIs denoted as c1、c4、c7，QreactionIs denoted as c1、c4、c8，Qbatch,airIs denoted as c1、c4、c9，Qmelt,wallIs denoted as c1、c4、c10，Qreg,airIs denoted as c1、c4、c11，QwgasIs denoted as c1、c4、c12，Qreg,wallIs denoted as c1、c4、c13，QbrickIs denoted as c1、c4、c14The resulting energy generalization hierarchy is shown in fig. 2. The node information corresponds to the energy composition item attribute, and the node information also corresponds to the characteristic variable in step S11.
In step S20, the dissimilarity calculation of the energy components is derived from the energy generalization hierarchy, each attribute cxBy generalizing the hierarchy GxFind a common parent node p to connect x1、x2And accumulating the length of the shortest path connecting the two nodes to represent the dissimilarity between the attributes, and simultaneously, the father node is the maximum generalized representation obtained under the condition of minimizing the dissimilarity of the attributes. Step S20 is performed according to the following steps:
s21, acquiring an abnormal energy consumption sample set A and an energy consumption sample C of the energy management systemiEnergy consumption sample CiEnergy composition attribute of (2) is noted as cx，Ci＝{c1,c2,…,cm}：
Figure GDA0003529534720000071
Figure GDA0003529534720000072
Respectively is a row vector of the abnormal energy consumption sample set A;
transposing the matrix A:
Figure GDA0003529534720000073
calculating each feature attribute c in the ATxMathematical expectation ofxSum variance σxWherein the characteristic attribute is an energy composition item:
Figure GDA0003529534720000074
Figure GDA0003529534720000075
ci,xa mathematical expectation representing the ith energy contribution;
s22, giving a significance level a and setting a random variable XxEnergy composition item attribute c representing a samplexSequentially calculating abnormal energy consumption samples CiEach energy composition item attribute cxExpressed as:
Figure GDA0003529534720000081
wherein P represents probability, and ε represents any positive number;
s23, using the reasonable interval principle of Chebyshev inequality, the method is suitable for the application
Figure GDA0003529534720000082
If the value of the energy composition item exceeds the value range of the normal category, the attributes have no similarity; the attribute c is generalized according to the energy generalization hierarchy GxIs replaced by GxIn (c)xTo obtain a generalized representation of the abnormal energy consumption sample Ci(ii) a On the contrary, no generalization operation is performed and the attribute c is calculatedxDissimilarity δ of;
s24, repeatedly executing the steps S22 and S23 until the dissimilarity degree delta of all the energy composition item attributes is calculated, accumulating the dissimilarity degrees delta of all the attributes, and representing an abnormal energy consumption sample CiDissimilarity of (a):
Figure GDA0003529534720000083
d(Ci[cx]) The dissimilarity of the samples of abnormal energy consumption represented for generalization.
The abnormal energy consumption generalization hierarchical clustering algorithm in the step S30 comprises the following steps:
a. getObtaining an original abnormal energy consumption sample set A ═ C1,C2,…,CkH, selecting a certain attribute cxAll the abnormal energy consumption sample attributes c in AxValue replacement is GxIn (c)xA parent value of;
b. repeating the step a, continuously generalizing the abnormal energy consumption until a subset L of A is found, wherein the number of elements in the subset L is | L | ≧ minisize, and H (L) is minimum, and terminating the generalization operation;
c. the generalized representation of the energy composition item set L is output.
The algorithm pseudo code of the abnormal energy consumption generalization hierarchical clustering algorithm is as follows:
inputting: forming a cluster minimum statistic mini and a generalized hierarchical structure G by using an abnormal energy consumption sample set A
And (3) outputting: generalized energy constituent set L
The process is as follows:
Figure GDA0003529534720000091
step S30 is performed as follows:
s31, importing an abnormal energy consumption sample set A, copying the sample set to be recorded as L, loading an energy generalization hierarchical structure G, and calculating the attribute c of each energy composition itemxMathematical expectation ofxSum variance σx；
S32, selecting an abnormal energy consumption sample C from the abnormal energy consumption sample set Ai；
S33, selecting attributes c from bottom to top in sequence according to the energy generalization hierarchical structurex(ii) a Judging the attribute c by the Chebyshev inequalityxValue of (d) Dom (c)x) Whether the reasonable interval requirement of the Chebyshev inequality is met: if yes, let cxReplacement to Gx in cxTo obtain a generalized representation of the abnormal energy consumption sample CiCumulative equivalence and CiAnd is stored in A and the original C is deletediCalculating the attribute dissimilarity δx；
S34, repeatedly executing the step S33 to belong to all the usersDegree of dissimilarity δ of propertiesxCumulatively represent energy consumption samples CiDegree of dissimilarity of dx；
S35, repeatedly executing the steps S32-S34 until the generalization process of all the abnormal energy consumption samples and the dissimilarity calculation are completed; and L is searched, all the same samples of the abnormal energy consumption represented by the generalization are merged, and the average dissimilarity degree of the clusters of the abnormal energy consumption represented by the generalization
Figure GDA0003529534720000101
Representing the current generalization effect; mean degree of dissimilarity
Figure GDA0003529534720000102
Calculated as follows:
Figure GDA0003529534720000103
s36, searching whether the element number of the abnormal energy consumption cluster represented in the generalization mode is larger than the minsize parameter or not, and the average dissimilarity degree of the current abnormal energy consumption cluster
Figure GDA0003529534720000104
Is the minimum value; if the condition is not met, the steps S32-S36 are repeatedly executed;
and S37, terminating the hierarchical clustering iteration condition, and outputting the energy represented in a generalization manner to form an item set A.
In step S36, the minisize parameter indicates the element number statistic in the generalized cluster, and is a boundary condition for terminating the generalized clustering process. The mini parameter setting is not suitable to be too large or too small, and when the set value is too small, the algorithm is forced to combine energy consumption samples with different sources; similarly, when the set value is too small, generalized clustering is ended in advance, so that energy consumption samples with the same source are mixed in different clusters.
The mini parameter of the present embodiment is adaptive according to the following method: setting the initial value minszie to ms0Then define a smaller value w, w ∈ [0,1 ]](ii) a When ms is0、ms0·(1-w)、ms0Poly of (1+ w)If the class effect is the same, then ms is considered0The method is the best value of the current data scale. In this embodiment, the initial value of the minisize is 1/5 of the size of the abnormal energy consumption sample set, and when the initial value of the minisize takes this value, the clustering result has better robustness.
In step S23, in the method for measuring similarity of energy consumption by chebyshev inequalities, the key step is to calculate the chebyshev reasonable interval of each characteristic attribute, and the range value e of the deviation mean value needs to be determined by calculating the chebyshev inequalities. The characteristic attribute of the energy composition item in the known energy consumption sample is obtained by calculating an energy consumption value from working condition data, and the value of the original working condition data has continuity, such as temperature, fuel consumption, smoke emission and the like, i.e. the characteristic attribute of the energy composition item is a continuous random variable and obeys normal distribution. From the characteristics of normal distribution, when e is set to 2 times the standard deviation, there is a case where
Figure GDA0003529534720000105
The probability of values except (m-2s, m +2s) is only 0.5%, and the method has a good screening effect on abnormal data.
TABLE 1 partial abnormal energy consumption notes
Figure GDA0003529534720000111
In this embodiment, an ECD data set is used as an experimental data set, a detection period is set to be 2 minutes, an energy consumption abnormality detection process with a sample size of 574 is input each time, an energy consumption abnormality detection result is recorded as an input space, and an abnormal energy consumption sample is shown in table 1. From the energy calculation, Q is knownbatch、Qreaction、Qbatch,air、Qreg,wallThe attribute value of the energy composition item is stable, so that the attribute of a stable factor can be ignored to reduce the calculation.
Then, the mathematical expectation m and variance s of the characteristic attribute of each energy component are found, as shown in table 2. And constructing Chebyshev inequalities corresponding to the characteristic attributes according to the mathematical expectation and the variance of each energy composition item so as to initialize the energy consumption dissimilarity measuring method.
TABLE 2 mathematical expectation and variance of energy composition item attributes
Figure GDA0003529534720000121
And finally, initializing a mini parameter and an e parameter of a root cause analysis hierarchical clustering algorithm (RCA-HCA algorithm), importing an energy generalization hierarchical structure G, inputting the abnormal energy consumption samples into the RCA-HCA algorithm for clustering generalization, and obtaining clustering generalization results shown in table 3.
TABLE 3 generalized sample set of abnormal energy consumption
Figure GDA0003529534720000122
The abnormal energy consumption samples represented by generalization show that each generalization result is obtained by minimizing the dissimilarity of the average energy consumption of the clusters, and the generalization results are composed of nodes of an energy generalization hierarchy to represent a class of abnormal conditions. As shown in Table 3, the present embodiment obtains 9 abnormal energy consumption sample records, wherein the generalized result { Q ] is obtained through the clustering processq6 pieces in total, and as can be seen from the characteristics of the clustering algorithm, each generalized result represents a cluster group, { Q }qThe generalization result can represent a large cluster group in the clustering result of the time, namely, the main factor for causing the energy consumption abnormity at this time is a small furnace.
When the sample size to be detected reaches 574, the FMI index reaches 0.834, namely, the energy consumption sample with the actual abnormality is detected with the accuracy of 83.4%.
In the detailed description of the embodiments, various technical features may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. This need not be, nor should it be exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims (8)

1. A horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering is characterized by comprising the following steps:
s10, constructing an energy generalization hierarchical structure by using a root cause analysis method by taking a layered energy model of the horseshoe flame glass melting furnace as a prototype;
s20, calculating the dissimilarity of the attributes of the energy composition items based on the energy generalization hierarchical structure of the step S10, and accumulating the dissimilarity of the attributes of all the energy composition items to represent the dissimilarity d (C) of the abnormal energy consumption samplei,Cj) Obtaining a generalization abnormal energy consumption sample;
s30, establishing an energy consumption abnormity positioning model based on a hierarchical clustering algorithm, inputting the generalized abnormal energy consumption sample obtained in the step S20, and outputting an energy composition item set represented in a generalized manner;
in the layered energy model described in step S10, an input energy system flows from top to bottom, energy is input by a small furnace, passes through a flame space, and is output by a melting tank and a regenerator, and energy constituting items of the small furnace, the flame space, the regenerator, and the melting tank respectively constitute an energy hierarchy;
the energy generalization hierarchy in step S10 is performed according to the following steps:
s11, determining characteristic variables and value domains Dom (c)x): heat Q generated by combustion of fuel in small furnaceqAs root point, heat Q absorbed by the glass surface in the flame spacelevelHeat Q dissipated by arch top and kiln wallcomb,wallHigh-temperature flue gas takes away heat Qreg,flueAs a branch node, the molten glass in the melting tank takes away heat QglassGlass meltingHeat of reaction consumption QreactionThe generated gas of the glass melting reaction takes away heat Qbatch,airHeat quantity Q dissipated from bottom of melting tank and kiln wallmelt,wallAs a leaf node, the combustion-supporting air in the heat storage tank carries physical sensible heat Qreg,airThe waste gas takes away heat QwgasHeat Q of heat dissipation of wall of heat storage chamberreg,wallHeat loss Q of grid body furnace dustbrickAs leaf nodes;
s12, determining a hierarchical relationship: qreg,flueInput energy system as regenerator and output energy system Qreg,air、Qwgas、Qreg,wallAnd QbrickSum of the junction values of (1), Qreg,flueIs Qreg,air、Qwgas、Qreg,wallAnd QbrickA parent node of (a); qlevelAs input energy system of the melting tank, the value of which is the output energy system Qglass、Qreaction、Qbatch,airAnd Qmelt,wallSum of the junction values of (1), QlevelIs Qglass、Qreaction、Qbatch,airAnd Qmelt,wallA parent node of (a); qqAs input energy system of small furnace, its value is output energy system Qlevel、Qcomvb,wall、Qreg,flueThe sum of the junction values of (a); qx、QbatchRespectively representing the physical sensible heat of the fuel and the mixture, and being a generalized hierarchical structure only comprising root nodes;
s13, generating an energy generalization hierarchical structure set G ═ G1,G2,G3}: for attribute cxThe method comprises the steps of establishing a corresponding generalization hierarchical structure Gx by utilizing a root cause analysis method, wherein the Gx structure is a data structure tree, and a root node is the topmost generalization result which can be obtained by each sample and has uniqueness.
2. The method of claim 1, wherein in step S11, each node stores information about a parent node or an ancestor node, and Q is a factorx、QbatchIs respectively represented as c2、c3，QqIs denoted as c1，QlevelIs denoted as c1、c4，Qcomvb,wallIs denoted as c1、c5，Qreg,flueIs denoted as c1、c6，QglassIs denoted as c1、c4、c7，QreactionIs denoted as c1、c4、c8，Qbatch,airIs denoted as c1、c4、c9，Qmelt,wallIs denoted as c1、c4、c10，Qreg,airIs denoted as c1、c4、c11，QwgasIs denoted as c1、c4、c12，Qreg,wallIs denoted as c1、c4、c13，QbrickIs denoted as c1、c4、c14(ii) a The node information is a characteristic variable.
3. The method for positioning abnormal energy consumption of the horseshoe flame glass melting furnace based on the root cause analysis hierarchical clustering as claimed in claim 1 or 2, wherein the step S20 is performed according to the following steps:
s21, acquiring an abnormal energy consumption sample set A and an energy consumption sample C of the energy management systemiEnergy consumption sample CiEnergy composition item attribute of cx，Ci＝{c1,c2,…,cm}：
Figure FDA0003602459330000021
Figure FDA0003602459330000022
Respectively, abnormal energy consumptionThe row vectors of this set A;
transposing the matrix A:
Figure FDA0003602459330000023
calculating each feature attribute c in the ATxMathematical expectation ofxSum variance σx：
Figure FDA0003602459330000024
Figure FDA0003602459330000025
ci,xA mathematical expectation representing the ith energy contribution;
s22, giving a significance level a and setting a random variable XxEnergy composition item attribute c representing a samplexSequentially calculating abnormal energy consumption samples CiEach energy composition item attribute cxThe chebyshev inequality of (c):
Figure FDA0003602459330000026
wherein P represents probability, and ε represents any positive number;
s23, using the reasonable interval principle of Chebyshev inequality, the method is suitable for the application
Figure FDA0003602459330000032
When the energy composition item exceeds the value range of the normal category, the attributes have no similarity; the attribute c is generalized according to the energy generalization hierarchy GxIs replaced by GxIn (c)xTo obtain a generalized representation of the abnormal energy consumption sample Ci(ii) a On the contrary, no generalization operation is performed and the attribute c is calculatedxDegree of dissimilarity ofδx(x1)；
S24, repeatedly executing the steps S22 and S23 until the dissimilarity degree delta of all the energy composition item attributes is calculated, accumulating the dissimilarity degrees delta of all the attributes, and representing an abnormal energy consumption sample CiDissimilarity of (a):
Figure FDA0003602459330000031
d(Ci[cx]) The dissimilarity of the samples of abnormal energy consumption represented for generalization.
4. The method for locating abnormal energy consumption of the horseshoe flame glass melting furnace based on root cause analysis hierarchical clustering as claimed in claim 3, wherein the hierarchical clustering algorithm in step S30 comprises the following steps:
a. obtaining an original abnormal energy consumption sample set A ═ C1,C2,…,CkH, selecting a certain attribute cxAll the abnormal energy consumption sample attributes c in AxValue replacement is GxIn (c)xA parent value of;
b. repeating the step a, continuously generalizing the abnormal energy consumption until a subset L of A is found, wherein the number of elements in the subset L is | L | ≧ minisize, and H (L) is minimum, and terminating the generalization operation;
c. the generalized representation of the energy composition item set L is output.
5. The method for positioning abnormal energy consumption of the horseshoe flame glass melting furnace based on the root cause analysis hierarchical clustering, according to claim 4, is characterized in that the step S30 is carried out according to the following steps:
s31, importing an abnormal energy consumption sample set A, and calculating the attribute c of each energy composition itemxMathematical expectation ofxSum variance σx；
S32, selecting an abnormal energy consumption sample C from the abnormal energy consumption sample set Ai；
S33, selecting attributes c from bottom to top in sequence according to the energy generalization hierarchical structurex(ii) a Universal jointExcess Chebyshev inequality, judging attribute cxValue of (d) Dom (c)x) Whether the reasonable interval requirement of the Chebyshev inequality is met: if yes, let cxReplacement to Gx in cxTo obtain a generalized representation of the abnormal energy consumption sample CiCumulative equivalence and CiAnd is stored in A and the original C is deletediCalculating the attribute dissimilarity δx；
S34, repeatedly executing the step S33 to determine the dissimilarity delta of all the attributesxCumulatively represent energy consumption samples CiDegree of dissimilarity d ofx；
S35, repeatedly executing the steps S32-S34 until the generalization process of all the abnormal energy consumption samples and the dissimilarity calculation are completed; and L is searched, all the same samples of the abnormal energy consumption represented by the generalization are merged, and the average dissimilarity degree of the clusters of the abnormal energy consumption represented by the generalization
Figure FDA0003602459330000041
Representing the current generalization effect;
s36, searching whether the element number of the abnormal energy consumption cluster represented in the generalization mode is larger than the minsize parameter or not, and the average dissimilarity degree of the current abnormal energy consumption cluster
Figure FDA0003602459330000042
Is the minimum value; if the condition is not met, the steps S32-S36 are repeatedly executed;
and S37, terminating the hierarchical clustering iteration condition, and outputting the energy represented in a generalization manner to form an item set A.
6. The method for locating abnormal energy consumption of the horseshoe flame glass melting furnace based on root cause analysis hierarchical clustering as claimed in claim 5, wherein in step S36, the minisize parameter is self-adapted according to the following method: setting initial value minszie to ms0Then, a smaller value w is defined, w ∈ [0,1 ]](ii) a When ms is0、ms0*(1-w)、ms0When the clustering effect of (1+ w) is the same, ms is considered to be0The method is the best value of the current data scale.
7. The method for locating abnormal energy consumption of the horseshoe flame glass melting furnace based on root cause analysis hierarchical clustering as claimed in claim 6, wherein the initial value of minisize is 1/5 of the size of the abnormal energy consumption sample set.
8. The method for locating abnormal energy consumption of horseshoe flame glass melting furnace based on root cause analysis hierarchical clustering as claimed in any one of claims 4 to 7, wherein in step S23, a reasonable interval of each characteristic attribute is determined according to a range value e of a deviation mean value, and if e is set to be 2 times of a standard deviation, then P (| X-m |. kii.2s) ═ 0.9544 is provided, which means that the probability of taking values outside (m-2S, m +2S) is only 0.5%.
CN202010771841.8A 2020-08-04 2020-08-04 Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering Active CN112102891B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010771841.8A CN112102891B (en) 2020-08-04 2020-08-04 Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010771841.8A CN112102891B (en) 2020-08-04 2020-08-04 Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering

Publications (2)

Publication Number Publication Date
CN112102891A CN112102891A (en) 2020-12-18
CN112102891B true CN112102891B (en) 2022-06-14

Family

ID=73749974

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010771841.8A Active CN112102891B (en) 2020-08-04 2020-08-04 Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering

Country Status (1)

Country Link
CN (1) CN112102891B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113191678B (en) * 2021-05-21 2023-01-10 联洋国融(北京)科技有限公司 Safe production index abnormity rapid sensing method based on Internet of things and artificial intelligence
CN117435676B (en) * 2023-07-13 2024-06-07 南京电力设计研究院有限公司 Building energy management method based on subsequence mining and directed weighted graph clustering
CN117150244B (en) * 2023-10-30 2024-01-26 山东凯莱电气设备有限公司 Intelligent power distribution cabinet state monitoring method and system based on electrical parameter analysis

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5944863A (en) * 1997-04-11 1999-08-31 Owens-Brockway Glass Container Inc. Apparatus for controlling flow of casing glass in a cased glass stream

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9366451B2 (en) * 2010-12-24 2016-06-14 Commonwealth Scientific And Industrial Research Organisation System and method for the detection of faults in a multi-variable system utilizing both a model for normal operation and a model for faulty operation
FR3032786B1 (en) * 2015-02-17 2017-03-24 Schneider Electric Ind Sas DATA PROCESSING AND MODELING SYSTEM FOR ANALYZING THE ENERGY CONSUMPTION OF A SITE
CN105812177B (en) * 2016-03-08 2019-10-18 华为技术有限公司 A kind of network failure processing method and processing equipment
US11221353B2 (en) * 2018-07-06 2022-01-11 Schneider Electric USA, Inc. Systems and methods for analyzing power quality events in an electrical system
CN110147387B (en) * 2019-05-08 2023-06-09 腾讯科技（上海）有限公司 Root cause analysis method, root cause analysis device, root cause analysis equipment and storage medium
CN110208696A (en) * 2019-06-25 2019-09-06 华北电力大学 Based on model from the coal pulverizer method for diagnosing faults of typical sample of drawing
CN110536258B (en) * 2019-08-09 2021-07-16 大连理工大学 Trust model based on isolated forest in UASNs

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5944863A (en) * 1997-04-11 1999-08-31 Owens-Brockway Glass Container Inc. Apparatus for controlling flow of casing glass in a cased glass stream

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于GA-SVR的挤压机能耗异常检测模型研究;杨海东 等;《机床与液压》;20190315;第163-168页 *

Also Published As

Publication number Publication date
CN112102891A (en) 2020-12-18

Similar Documents

Publication Publication Date Title
CN112102891B (en) Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering
CN107016455B (en) Prediction system and method for oxygen content of flue gas at hearth outlet of circulating fluidized bed household garbage incineration boiler
Yang et al. Community mining from signed social networks
CN110990461A (en) Big data analysis model algorithm model selection method and device, electronic equipment and medium
Pan et al. A virtual metrology system for predicting end-of-line electrical properties using a MANCOVA model with tools clustering
CN112016175A (en) Water supply pipe network pressure measuring point optimal arrangement method based on tree hierarchical clustering
CN109073753B (en) System and method for generating an energy model and tracking energy model evolution
CN112445844B (en) Financial data management control system of big data platform
CN115115796A (en) Building digital twin frame based on semantic Web technology and modeling method
CN117056688A (en) New material production data management system and method based on data analysis
Ye et al. Hydrologic time series anomaly detection based on flink
Nagwani et al. A data mining model to predict software bug complexity using bug estimation and clustering
CN112765746B (en) Turbine blade top gas-thermal performance uncertainty quantification system based on polynomial chaos
CN112487053B (en) Abnormal control extraction working method for mass financial data
CN117743803A (en) Workload perception instant defect prediction method based on evolutionary feature construction
CN108399284A (en) It is a kind of about subtracted based on deviation big data Trading Model analysis and restorative procedure
CN112800672B (en) Evaluation method, system, medium and electronic equipment for boiler fouling coefficient
CN112463853B (en) Financial data behavior screening working method through cloud platform
CN114971258A (en) Industrial process optimization method and system based on coke quality prediction
Zhang et al. Attack graph-based quantitative assessment for industrial control system security
CN112992285A (en) IPSO-HKELM-based blast furnace molten iron silicon content prediction method
CN113362911A (en) IPSO-HKELM-based blast furnace molten iron silicon content prediction method
CN111783850A (en) Load clustering method for optimizing Bisegmentation K-means based on Kd tree and Canopy
Jassar et al. Impact of data quality on predictive accuracy of ANFIS based soft sensor models
Zhang et al. Efficient index-free SimRank similarity search in large graphs by discounting path lengths

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