CN105257277B

CN105257277B - Dlagnosis of Sucker Rod Pumping Well underground failure prediction method based on Multi-variable Grey Model

Info

Publication number: CN105257277B
Application number: CN201510248922.9A
Authority: CN
Inventors: 韩莹; 李琨; 黄海礁; 魏泽飞; 杨柳; 杨一柳; 佘东生
Original assignee: Bohai University
Current assignee: Bohai University
Priority date: 2015-05-15
Filing date: 2015-05-15
Publication date: 2018-04-06
Anticipated expiration: 2035-05-15
Also published as: CN105257277A

Abstract

A kind of Dlagnosis of Sucker Rod Pumping Well underground failure prediction method based on Multi-variable Grey Model, surface dynamometer card is gathered by indicator card remote data acquisition system；Surface dynamometer card is converted into underground pump dynagraoph；The curve Character eigenvector of underground pump dynagraoph graphic feature can be characterized according to the theoretical extraction of Curve Moment；Each characteristic vector is predicted by Multi-variable Grey Model；The gray relation grades of forecast sample and each fault type in regular set are calculated, the fault type corresponding to maximum gray relation grades is the fault type of forecast sample.Can Reasonable adjustment pumpingh well working system, so as to improve the production efficiency of pumpingh well, can be achieved to the failure predication of Dlagnosis of Sucker Rod Pumping Well underground work situation；Calculate simply, required sample is few, and is all suitable for for the data set of Arbitrary distribution；The degree of accuracy of prediction is high, has higher precision of prediction especially for the fault type of gradation type.

Description

Dlagnosis of Sucker Rod Pumping Well underground failure prediction method based on Multi-variable Grey Model

Technical field

The invention belongs to Dlagnosis of Sucker Rod Pumping Well failure prediction method field, and in particular to one kind is based on Multi-variable Grey Model Dlagnosis of Sucker Rod Pumping Well underground failure prediction method.

Background technology

The failure of Dlagnosis of Sucker Rod Pumping Well is that oil field produces a faced subject matter, can influence the fortune of downhole pump The oil production of row situation and oil well.Because oil well pump is operated in thousands of meters of underground, working condition is sufficiently complex, working environment pole It is severe, and rate of breakdown is very high, can largely influence the normal production in oil field.Once pumpingh well can not there occurs failure Diagnosed in time, will result in the waste of the energy, and influence pumpingh well production, loss is brought to enterprise.Have at present very The method for diagnosing faults of Dlagnosis of Sucker Rod Pumping Well underground is focused in more researchs, but this is built upon the basis that failure has occurred and that On, the normal production to pumpingh well brings certain influence.

The content of the invention

The technical problem to be solved in the present invention is to provide a kind of Dlagnosis of Sucker Rod Pumping Well underground based on Multi-variable Grey Model Failure prediction method, realize the failure predication to Dlagnosis of Sucker Rod Pumping Well underground work situation.

The present invention technical solution be：

A kind of Dlagnosis of Sucker Rod Pumping Well underground failure prediction method based on Multi-variable Grey Model, it is characterized in that：

1), using indicator card remote data acquisition system, first, wireless indicator is arranged on taking out for Dlagnosis of Sucker Rod Pumping Well On the horse head of oil machine, the load and displacement data of polished rod are gathered；Wireless RTU wirelessly receives wireless indicator collection Data, by data acquisition module by wireless aps slave station module with wireless network teletransmission to wireless aps master station module；Show The load and displacement data for the polished rod that work(figure monitors and data processing server collects, draw the ground collected and show work( Figure figure；

2) surface dynamometer card, is converted into underground pump dynagraoph, to obtain truly reflecting showing for oil pumping working conditions of pump Work(figure figure；

3) the curve moment characteristics of underground pump dynagraoph, are extracted as characteristic vector, underground pump dynagraoph is divided into upstroke portion Point and down stroke part, then extract the 7 curve Character eigenvectors and undershoot of upstroke part respectively according to Curve Moment theory 7 curve Character eigenvectors of journey part, the variable using obtain 14 curve Character eigenvectors as next step prediction, specifically Step is as follows：

A) underground pump dynagraoph is divided into upstroke part and down stroke part, wherein upstroke partial trace has reflected The upstroke process of bar pumping oil well, it is that rod string pulls up plunger in the presence of power set, is sucked in oil well pump The process of fluid and well head discharge fluid；Down stroke partial trace reflects the down stroke process of Dlagnosis of Sucker Rod Pumping Well, is oil pumping Roofbolt pushes down on plunger in the presence of power set, and oil well pump discharges the process of fluid into oil pipe；Assuming that underground pump work Figure by N number of groups of samples into, then the point maximum from the 1st point to displacement be upstroke part, the maximum point vacation of the displacement It is set at l-th point；It is down stroke part from the l+1 point to n-th point；

B) the upstroke partial trace after dividing is by l groups of samples into the coordinate of each sampled point is (x_i,y_i), its s+ R rank Curve Moments m_srIt is defined as：

Wherein i=1 ..., l；S, r=0,1,2；Δz_iFor the distance between two continuous sampled points,

μ_srFor s+r rank central moments, it is defined as：

WhereinPointFor the barycentric coodinates of curve；

Calculate following each rank central moment：μ₀₀=m₀₀, μ₁₀=0, μ₀₁=0,

Standardization processing is carried out to each rank central moment of extraction, by η_srRepresent, be defined as：

η_sr=μ_sr/(μ₀₀)^s+r+1 (3)

7 curve Character eigenvectors of the upstroke part so constructed are defined respectively as：

Make 7 curve Character eigenvectors of the upstroke part of construction using following correction formulaSpan It is unified, it is designated as ψ₁—ψ₇, it is defined as：

Wherein q=0,1 ..., 7；

C) the curve Character eigenvector of down stroke part is extracted, the down stroke partial trace after division is by N-l sampled point Composition, the coordinate of each sampled point is (x_j,y_j), its s+r rank Curve Moment m'_srIt is defined as：

Wherein j=l+1, l+2 ..., N；S, r=0,1,2；Δz_jFor the distance between two continuous sampled points,

μ'_srFor s+r rank central moments, it is defined as：

WhereinPointFor the barycentric coodinates of curve；

Calculate following each rank central moment：μ'₀₀=m'₀₀, μ '₁₀=0, μ '₀₁=0,

Standardization processing is carried out to each rank central moment of extraction, by η '_srRepresent, be defined as：

η'_sr=μ '_sr/(μ'₀₀)^s+r+1 (14)

7 curve Character eigenvectors of the down stroke part so constructed are defined respectively as：

Make 7 curve Character eigenvectors of the down stroke part of construction using following correction formulaValue model Unification is enclosed, is designated as ψ₈—ψ₁₄, it is defined as：

Wherein q '=8,9 ..., 14；

The obtained 14 characteristic vector ψ that can characterize underground pump dynagraoph graphic feature will be extracted₁—ψ₁₄, as in next step The variable of prediction；

4) the changeable of Dlagnosis of Sucker Rod Pumping Well underground failure, is established according to Multi-variable Grey Model MGM (1, n) basic theories Grey forecasting model is measured, the dynamic correlation established using MGM (1, n) model between multiple variables, makes up univariate method Limitation；14 curve Character eigenvectors that pump dynagraoph in one section of continuous time is extracted are as Multi-variable Grey Model The input of (MGM (1,14)), establishes grey time series, a change of each curve Character eigenvector as grey time series Amount；Predict the feature at next time point of 14 curve Character eigenvectors respectively by the multivariable grey forecasting model established Vector, the curve Character eigenvector of 14 predictions is the characteristic vector that can characterize forecast sample feature；

The pump dynagraoph curve Character eigenvector ash time series established is represented as follows：

Wherein p=1,2 ..., 14；K=1,2 ..., m, m are the quantity of used pump dynagraoph；Represent pump dynagraoph K-th of influence factor in p-th of variable in curve Character eigenvector ash time series；

OrderFor corresponding one-accumulate formation sequence, have：

14 yuan of One first-order ordinary differential equations are established, are had：

Note

B=(b₁,b₁,…,b_n)^T (27)

A and B is identified parameters, then is abbreviated as formula (25)

Wherein

In [0, t] section upper integral, obtaining Continuous Time Response is

Ψ⁽¹⁾(t)=e^AtΨ⁽¹⁾(0)+A^-1(e^At-I)B (29)

Wherein

Formula (25) is subjected to discretization, obtained：

Wherein p=1,2 ..., 14；K=2,3 ..., m；

Note

E identifier E is obtained by least square method, had：

E=(L^TL)^-1L^TY (32)

Wherein,

Identified parameters A and B identifier A and B are obtained by formula (35), are respectively：

According to obtained identifier A and B, formula (29) is written as discrete form,

Ψ⁽⁰⁾(1)=Ψ⁽⁰⁾(1) (38)

Ψ⁽¹⁾(k)=e^A(k-1)Ψ⁽¹⁾(1)+A^-1(e^A(k)-I)B (39)

Wherein Ψ⁽¹⁾And Ψ (k)⁽¹⁾(1) formula (36) and formula (38) can be substituted into formula (30) to try to achieve；

Obtaining forecast model is：

Ψ⁽⁰⁾(k)=Ψ⁽¹⁾(k)-Ψ⁽¹⁾(k-1) (40)

Wherein k=2,3 ..., m；

5) it is, last, judge which class fault type forecast sample belongs to, each event is established by existing indicator card information Hinder the regular set of type, i.e., to having determined that a number of indicator card of fault type by the way that surface dynamometer card is converted into underground Pump dynagraoph and its 14 curve Character eigenvectors are extracted respectively, obtain each curve Character eigenvector in each fault type Interval；Resulting interval byRepresent, p=1,2 ..., 14, whereinRepresent the F_MThe lower limit of p-th of curve Character eigenvector of kind fault type,Represent F_MP-th of Curve Moment of kind fault type The higher limit of characteristic vector；

Each the curve Character eigenvector value Curve Moment corresponding with each fault type for calculating forecast sample is special The distance of vector value is levied, is calculated by following formula：

Then the grey relation coefficient of forecast sample and each fault type is calculated by following formula,

Wherein p=1,2 ..., 14；M=1,2,3 ....ρ ∈ [0,1] are resolution ratio, take 0.5 herein；

The gray relation grades of forecast sample and each fault type are calculated by following formula again,

Fault type corresponding to maximum gray relation grades is the fault type of forecast sample.

Further, the upstroke part of underground pump dynagraoph is the suspension point of Dlagnosis of Sucker Rod Pumping Well sucker rod by most in step 3) Low spot moves to the part of peak；Down stroke part is moved to minimum for the suspension point of Dlagnosis of Sucker Rod Pumping Well sucker rod by peak The part of point.

Further, employed in multivariable grey forecasting model of the step 4) for Dlagnosis of Sucker Rod Pumping Well underground failure Sample in one section of continuous time, the unit of continuous time can be hour or day, and the quantity of the sample is 50-200, To improve the computational efficiency of forecast model and the degree of accuracy.

Further, step 5) is established the regular set of each fault type by existing information, and existing information is by Know that the indicator card of fault type obtains, for establishing the quantity of the indicator card of the regular set of each fault type more than 2.

Further, Dlagnosis of Sucker Rod Pumping Well underground fault type has following 9 kinds：" normal ", " gases affect ", " supply Liquid deficiency ", " sucker rod, which breaks, to fall ", " travelling valve leakage ", " fixed valve leakage ", " being touched on pump ", " being touched under pump ", " sand production ".

The beneficial effects of the invention are as follows：

1st, Dlagnosis of Sucker Rod Pumping Well is a complicated mechanical system, have it is very strong non-linear, its failure have randomness, Uncertain and relativity.The present invention excavates and using the useful information in Dlagnosis of Sucker Rod Pumping Well production, to Dlagnosis of Sucker Rod Pumping Well Failure be predicted, and then the working system of Reasonable adjustment pumpingh well, so as to improve the production efficiency of pumpingh well, this is to ensureing Normally production tool is of great significance in oil field.

2nd, indicator card is the main method that Dlagnosis of Sucker Rod Pumping Well underground work situation is analyzed in the production of oil field, is collection polished rod One complete cycle internal load of change in displacement and displacement data and the PRL and the curve of polished rod displacement relation drawn, show work( The different shape of figure figure reflects the different working condition in Dlagnosis of Sucker Rod Pumping Well underground.Due to the production of Dlagnosis of Sucker Rod Pumping Well Journey is a continuous process, and the working condition of its underground is also the process of a gradual change, therefore, can by gather one section when The indicator card of interior Dlagnosis of Sucker Rod Pumping Well, realized according to the progressive formation of its graphics shape to Dlagnosis of Sucker Rod Pumping Well underground work The failure predication of situation.

3rd, the multivariable grey forecasting model for the Dlagnosis of Sucker Rod Pumping Well underground failure that the present invention is established, simple, institute is calculated The sample needed is few, and is all suitable for for the data set of Arbitrary distribution.Entered according to the regular set for each fault type established The judgement of row forecast sample, the degree of accuracy is high, has higher precision of prediction especially for the fault type of gradation type.

Brief description of the drawings

Fig. 1 is the indicator card remote data acquisition system schematic diagram of the present invention；

Fig. 2 is the fundamental diagram of the present invention；

Fig. 3 is that the underground pump work diagram after the surface dynamometer card and conversion gathered is intended to；

Fig. 4 is the division schematic diagram of underground pump dynagraoph；

Fig. 5 is curve Character eigenvector Ψ 1 Grey Model result schematic diagram；

Fig. 6 is curve Character eigenvector Ψ 2 Grey Model result schematic diagram；

Fig. 7 is curve Character eigenvector Ψ 3 Grey Model result schematic diagram；

Fig. 8 is curve Character eigenvector Ψ 4 Grey Model result schematic diagram；

Fig. 9 is curve Character eigenvector Ψ 5 Grey Model result schematic diagram；

Figure 10 is curve Character eigenvector Ψ 6 Grey Model result schematic diagram；

Figure 11 is curve Character eigenvector Ψ 7 Grey Model result schematic diagram；

Figure 12 is curve Character eigenvector Ψ 8 Grey Model result schematic diagram；

Figure 13 is curve Character eigenvector Ψ 9 Grey Model result schematic diagram；

Figure 14 is curve Character eigenvector Ψ 10 Grey Model result schematic diagram；

Figure 15 is curve Character eigenvector Ψ 11 Grey Model result schematic diagram；

Figure 16 is curve Character eigenvector Ψ 12 Grey Model result schematic diagram；

Figure 17 is curve Character eigenvector Ψ 13 Grey Model result schematic diagram；

Figure 18 is curve Character eigenvector Ψ 14 Grey Model result schematic diagram.

Embodiment

A kind of Dlagnosis of Sucker Rod Pumping Well underground failure prediction method based on Multi-variable Grey Model, its step are as follows：

1st, as depicted in figs. 1 and 2, using indicator card remote data acquisition system, be arranged on wireless indicator has first On the horse head of the oil pumper of bar pumping oil well, the load and displacement data of polished rod are gathered；Wireless RTU wirelessly receives nothing The data of line indicator collection, are remotely passed through interchanger 2 by data acquisition module by wireless aps slave station module with wireless network It is sent in wireless aps master station module, then indicator card monitoring and data processing server is reached through interchanger 1；Indicator card monitors and number The load and displacement data of the polished rod collected according to processing server, draw the surface dynamometer card figure collected.

2nd, as shown in figure 3, surface dynamometer card is converted into underground pump dynagraoph, to obtain truly reflecting oil pumping pump work The indicator card figure of situation；Due to being influenceed by factors such as the deformation of oil pumping post and vibrations, surface dynamometer card can not be truly anti- Reflect the practical working situation of Dlagnosis of Sucker Rod Pumping Well underground.Therefore, by the way that surface dynamometer card is converted into underground pump dynagraoph, to eliminate These influence.

3rd, because the graphics shape of underground pump dynagraoph reflects the working condition of Dlagnosis of Sucker Rod Pumping Well underground, therefore, to The prediction to underground failure is realized, some characteristic vectors that can characterize underground pump dynagraoph graphic feature can be first extracted, pass through In a period of time the variation tendency of each characteristic vector come predict future characteristic vector, so as to obtain corresponding to this feature vector Fault type.The curve moment characteristics of present invention extraction underground pump dynagraoph comprise the following steps that as characteristic vector：

A) the underground pump dynagraoph in Fig. 3 is divided into upstroke part and down stroke part, the upstroke of underground pump dynagraoph Part is moved to the part of peak for the suspension point of Dlagnosis of Sucker Rod Pumping Well sucker rod by minimum point；Taken out for sucker rod pump down stroke part The suspension point of well rod is moved to the part of minimum point by peak.As shown in figure 4, wherein upstroke partial trace reflects The upstroke process of Dlagnosis of Sucker Rod Pumping Well, it is that rod string pulls up plunger in the presence of power set, is inhaled in oil well pump Enter the process of fluid and well head discharge fluid；Down stroke partial trace reflects the down stroke process of Dlagnosis of Sucker Rod Pumping Well, is to take out Beam hanger post pushes down on plunger in the presence of power set, and oil well pump discharges the process of fluid into oil pipe.Assuming that down-hole pump Work(figure by N number of groups of samples into, then the point maximum from the 1st point to displacement be upstroke part, the point of the displacement maximum It is assumed to be at l-th point；It is down stroke part from the l+1 point to n-th point.

μ_srFor s+r rank central moments, it is defined as：

WhereinPointFor the barycentric coodinates of curve；

η_sr=μ_sr/(μ₀₀)^s+r+1 (3)

Wherein q=0,1 ..., 7.

μ'_srFor s+r rank central moments, it is defined as：

WhereinPointFor the barycentric coodinates of curve；

η'_sr=μ '_sr/(μ'₀₀)^s+r+1 (14)

Wherein q '=8,9 ..., 14.

The obtained 14 characteristic vector ψ that can characterize underground pump dynagraoph graphic feature will be extracted₁—ψ₁₄, as in next step The variable of prediction.The extraction of curve Character eigenvector, 14 obtained spies are carried out to the underground pump dynagraoph after the division shown in Fig. 4 Sign vector is as shown in table 1.

The pump dynagraoph curve Character eigenvector that table 1 extracts

Upstroke part	ψ₁	ψ₂	ψ₃	ψ₄	ψ₅	ψ₆	ψ₇
								Curve Character eigenvector value	1.7060	0.2577	1.6842	0.7475	1.8887	0.1878	1.6955
Down stroke part	ψ₈	ψ₉	ψ₁₀	ψ₁₁	ψ₁₂	ψ₁₃	ψ₁₄
								Curve Character eigenvector value	1.3639	0.4036	2.2777	1.3236	3.1241	1.1877	1.3260

4th, the changeable of Dlagnosis of Sucker Rod Pumping Well underground failure is established according to Multi-variable Grey Model MGM (1, n) basic theories Measure grey forecasting model.

The pump dynagraoph curve Character eigenvector ash time series established is as follows：

OrderFor corresponding one-accumulate formation sequence, have：

Note

B=(b₁,b₁,…,b_n)^T (27)

A and B is identified parameters, then formula (25) can be abbreviated as

Wherein

In [0, t] section upper integral, obtaining Continuous Time Response is

Ψ⁽¹⁾(t)=e^AtΨ⁽¹⁾(0)+A^-1(e^At-I)B (29)

Wherein

Formula (25) is subjected to discretization, obtained：

Wherein p=1,2 ..., 14；K=2,3 ..., m；

Note

E identifier E is obtained by least square method, had：

E=(L^TL)^-1L^TY (32)

Wherein,

According to obtained identifier A and B, formula (29) can be written as discrete form,

Ψ⁽⁰⁾(1)=Ψ⁽⁰⁾(1) (38)

Ψ⁽¹⁾(k)=e^A(k-1)Ψ⁽¹⁾(1)+A^-1(e^A(k)-I)B (39)

Wherein Ψ⁽¹⁾And Ψ (k)⁽¹⁾(1) formula (36) and formula (38), are substituted into formula (30) to try to achieve；

Then obtaining forecast model is：

Ψ⁽⁰⁾(k)=Ψ⁽¹⁾(k)-Ψ⁽¹⁾(k-1) (40)

Wherein k=2,3 ..., m.

In one section of continuous time employed in multivariable grey forecasting model for Dlagnosis of Sucker Rod Pumping Well underground failure Sample, the unit of continuous time can be hour or day, and the quantity of the sample is 50-200, to improve forecast model Computational efficiency and the degree of accuracy.50 width indicator cards of certain mouthful of Dlagnosis of Sucker Rod Pumping Well in 50 days in the present embodiment collection actual production As sample, surface dynamometer card is converted into underground pump dynagraoph first；Then 14 songs of 50 width underground pump dynagraophs are extracted respectively Line moment characteristics are as characteristic vector；The calculating of model is predicted further according to formula (23)-formula (40), obtained result is such as Shown in Fig. 5-Figure 18.By Fig. 5-Figure 18, using Multi-variable Grey Model respectively to 14 curves of the underground pump dynagraoph of extraction Character eigenvector is predicted, and predicted value is attained by satisfied effect.

51st sample is predicted by Multi-variable Grey Model using above-mentioned 50 samples, obtains its 14 curve moment characteristics Vector value is as shown in table 2.

The curve Character eigenvector of the forecast sample of table 2

Upstroke part	ψ₁	ψ₂	ψ₃	ψ₄	ψ₅	ψ₆	ψ₇
								Curve Character eigenvector value	1.1870	0.5246	1.4833	0.9148	1.0945	0.3579	1.0045
Down stroke part	ψ₈	ψ₉	ψ₁₀	ψ₁₁	ψ₁₂	ψ₁₃	ψ₁₄
								Curve Character eigenvector value	0.9699	0.6187	3.0128	1.7286	1.9872	1.5046	0.8967

5th, it is last, judge which class fault type predicted sample belongs to.Assuming that the class of Dlagnosis of Sucker Rod Pumping Well underground failure Type is by F_MRepresent, wherein M=1,2,3 ....The standard of each fault type can be established by existing indicator card information Collection, existing information is by having learned that the indicator card of fault type obtains, for establishing showing for the regular set of each fault type The quantity of work(figure is more than 2.I.e. to having determined that a number of indicator card of fault type by the way that surface dynamometer card is converted into Underground pump dynagraoph and its 14 curve Character eigenvectors are extracted respectively, obtain each curve moment characteristics in each fault type The interval of vector.Resulting interval byRepresent, p=1,2 ..., 14, whereinTable Show F_MThe lower limit of p-th of curve Character eigenvector of kind fault type,Represent F_MP-th of kind fault type is bent The higher limit of line Character eigenvector.The regular set of each fault type is established by existing information.

Dlagnosis of Sucker Rod Pumping Well underground fault type has following 9 kinds in the present invention：" normal ", " gases affect ", " feed flow is not Foot ", " sucker rod, which breaks, to fall ", " travelling valve leakage ", " fixed valve leakage ", " being touched on pump ", " being touched under pump ", " sand production ".Built The regular set of vertical each fault type is as shown in table 3.

The regular set of 3 each fault type of table

The curve Character eigenvector value for the 51st sample that Multi-variable Grey Model is predicted byRepresent, then prediction The distance of each curve Character eigenvector value curve Character eigenvector value corresponding with each fault type of sample can To be calculated by following formula,

According to formula (41)-formula (43), the gray relation grades of forecast sample and each fault type are calculated, are obtained As a result it is as shown in table 4.

The gray relation grades of the forecast sample of table 4 and each fault type

By table 4, the gray relation grades of forecast sample and " gases affect " fault type are maximum, it is believed that belong to " gases affect " event Hinder type.

The specific embodiment of the present invention is these are only, is not intended to limit the invention, for those skilled in the art For member, the present invention can have various modifications and variations.Any modification within the spirit and principles of the invention, being made, Equivalent substitution, improvement etc., should be included in the scope of the protection.

Claims

1. a kind of Dlagnosis of Sucker Rod Pumping Well underground failure prediction method based on Multi-variable Grey Model, it is characterized in that：

1.1) indicator card remote data acquisition system is used, first, wireless indicator is arranged on to the oil pumping of Dlagnosis of Sucker Rod Pumping Well On the horse head of machine, the load and displacement data of polished rod are gathered；Wireless RTU wirelessly receives the number of wireless indicator collection According to by data acquisition module by wireless aps slave station module with wireless network teletransmission to wireless aps master station module；Show work( The load and displacement data for the polished rod that figure monitoring and data processing server collect, draw the surface dynamometer card collected Figure；

1.2) surface dynamometer card is converted into underground pump dynagraoph, to obtain truly reflecting the indicator card of oil pumping working conditions of pump Figure；

1.3) extract underground pump dynagraoph curve moment characteristics be used as characteristic vector, by underground pump dynagraoph be divided into upstroke part with Down stroke part, then extract 7 curve Character eigenvectors and the down stroke portion of upstroke part respectively according to Curve Moment theory The 7 curve Character eigenvectors divided, the variable using obtain 14 curve Character eigenvectors as next step prediction, specific steps It is as follows：

A) underground pump dynagraoph is divided into upstroke part and down stroke part, wherein upstroke partial trace reflects sucker rod pump The upstroke process of pumpingh well, it is that rod string pulls up plunger in the presence of power set, fluid is sucked in oil well pump And the process of well head discharge fluid；Down stroke partial trace reflects the down stroke process of Dlagnosis of Sucker Rod Pumping Well, is rod string Plunger is pushed down in the presence of power set, oil well pump discharges the process of fluid into oil pipe；Assuming that underground pump dynagraoph is by N Individual groups of samples into, then the point maximum from the 1st point to displacement be upstroke part, and the point of the displacement maximum is assumed to be the L point；It is down stroke part from the l+1 point to n-th point；

B) the upstroke partial trace after dividing is by l groups of samples into the coordinate of each sampled point is (x_i,y_i), its s+r rank Curve Moment m_srIt is defined as：

<mrow> <msub> <mi>m</mi> <mrow> <mi>s</mi> <mi>r</mi> </mrow> </msub> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>l</mi> </munderover> <msup> <msub> <mi>x</mi> <mi>i</mi> </msub> <mi>s</mi> </msup> <msup> <msub> <mi>y</mi> <mi>i</mi> </msub> <mi>r</mi> </msup> <msub> <mi>&Delta;z</mi> <mi>i</mi> </msub> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow>

μ_srFor s+r rank central moments, it is defined as：

<mrow> <msub> <mi>&mu;</mi> <mrow> <mi>s</mi> <mi>r</mi> </mrow> </msub> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>l</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <mover> <mi>x</mi> <mo>&OverBar;</mo> </mover> <mo>)</mo> </mrow> <mi>s</mi> </msup> <msup> <mrow> <mo>(</mo> <msub> <mi>y</mi> <mi>i</mi> </msub> <mo>-</mo> <mover> <mi>y</mi> <mo>&OverBar;</mo> </mover> <mo>)</mo> </mrow> <mi>r</mi> </msup> <msub> <mi>&Delta;z</mi> <mi>i</mi> </msub> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>

WhereinPointFor the barycentric coodinates of curve；

η_sr=μ_sr/(μ₀₀)^s+r+1 (3)

Make 7 curve Character eigenvectors of the upstroke part of construction using following correction formulaSpan system One, it is designated as ψ₁—ψ₇, it is defined as：

Wherein q=0,1 ..., 7；

C) extract the curve Character eigenvector of down stroke part, the down stroke partial trace after division by N-l groups of samples into, The coordinate of each sampled point is (x_j,y_j), its s+r rank Curve Moment m'_srIt is defined as：

<mrow> <msubsup> <mi>m</mi> <mrow> <mi>s</mi> <mi>r</mi> </mrow> <mo>&prime;</mo> </msubsup> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>j</mi> <mo>=</mo> <mi>l</mi> <mo>+</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <msup> <msub> <mi>x</mi> <mi>j</mi> </msub> <mi>s</mi> </msup> <msup> <msub> <mi>y</mi> <mi>j</mi> </msub> <mi>r</mi> </msup> <msub> <mi>&Delta;z</mi> <mi>j</mi> </msub> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>12</mn> <mo>)</mo> </mrow> </mrow>

Wherein j=l+1, l+2 ..., N；S, r=0,1,2；Δz_jFor the distance between two continuous sampled points, μ′_srFor s+r rank central moments, it is defined as：

<mrow> <msubsup> <mi>&mu;</mi> <mrow> <mi>s</mi> <mi>r</mi> </mrow> <mo>&prime;</mo> </msubsup> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>j</mi> <mo>=</mo> <mi>l</mi> <mo>+</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>j</mi> </msub> <mo>-</mo> <mover> <msup> <mi>x</mi> <mo>&prime;</mo> </msup> <mo>&OverBar;</mo> </mover> <mo>)</mo> </mrow> <mi>s</mi> </msup> <msup> <mrow> <mo>(</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>-</mo> <mover> <msup> <mi>y</mi> <mo>&prime;</mo> </msup> <mo>&OverBar;</mo> </mover> <mo>)</mo> </mrow> <mi>r</mi> </msup> <msub> <mi>&Delta;z</mi> <mi>j</mi> </msub> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>13</mn> <mo>)</mo> </mrow> </mrow>

WhereinPointFor the barycentric coodinates of curve；

η′_sr=μ '_sr/(μ'₀₀)^s+r+1 (14)

Make 7 curve Character eigenvectors of the down stroke part of construction using following correction formulaSpan system One, it is designated as ψ₈—ψ₁₄, it is defined as：

Wherein q '=8,9 ..., 14；

The obtained 14 characteristic vector ψ that can characterize underground pump dynagraoph graphic feature will be extracted₁—ψ₁₄, predicted as next step Variable；

1.4) multivariable of Dlagnosis of Sucker Rod Pumping Well underground failure is established according to Multi-variable Grey Model MGM (1, n) basic theories Grey forecasting model, the dynamic correlation established using MGM (1, n) model between multiple variables, make up the office of univariate method It is sex-limited；14 curve Character eigenvectors that pump dynagraoph in one section of continuous time is extracted are as Multi-variable Grey Model (MGM (1,14)) input, establish grey time series, a variable of each curve Character eigenvector as grey time series；By The multivariable grey forecasting model established predicts the characteristic vector at next time point of 14 curve Character eigenvectors respectively, The curve Character eigenvector of 14 predictions is the characteristic vector that can characterize forecast sample feature；

Wherein p=1,2 ..., 14；K=1,2 ..., m, m are the quantity of used pump dynagraoph；Represent pump dynagraoph Curve Moment K-th of influence factor in p-th of variable in characteristic vector ash time series；

OrderFor corresponding one-accumulate formation sequence, have：

<mrow> <msubsup> <mi>&psi;</mi> <mi>p</mi> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </msubsup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>g</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>k</mi> </munderover> <msubsup> <mi>&psi;</mi> <mi>p</mi> <mrow> <mo>(</mo> <mn>0</mn> <mo>)</mo> </mrow> </msubsup> <mrow> <mo>(</mo> <mi>g</mi> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>24</mn> <mo>)</mo> </mrow> </mrow>

Note

B=(b₁,b₁,…,b_n)^T (27)

A and B is identified parameters, then is abbreviated as formula (25)

Wherein

In [0, t] section upper integral, obtaining Continuous Time Response is

Ψ⁽¹⁾(t)=e^AtΨ⁽¹⁾(0)+A^-1(e^At-I)B (29)

Wherein

Formula (25) is subjected to discretization, obtained：

<mrow> <msubsup> <mi>&psi;</mi> <mi>p</mi> <mrow> <mo>(</mo> <mn>0</mn> <mo>)</mo> </mrow> </msubsup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>w</mi> <mo>=</mo> <mn>1</mn> </mrow> <mn>14</mn> </munderover> <mfrac> <msub> <mi>a</mi> <mrow> <mi>p</mi> <mo>-</mo> <mi>w</mi> </mrow> </msub> <mn>2</mn> </mfrac> <mrow> <mo>(</mo> <msubsup> <mi>&psi;</mi> <mi>w</mi> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </msubsup> <mo>(</mo> <mi>k</mi> <mo>)</mo> <mo>+</mo> <msubsup> <mi>&psi;</mi> <mi>w</mi> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </msubsup> <mo>(</mo> <mrow> <mi>k</mi> <mo>-</mo> <mn>1</mn> </mrow> <mo>)</mo> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>30</mn> <mo>)</mo> </mrow> </mrow>

Wherein p=1,2 ..., 14；K=2,3 ..., m；

Note

E identifier E is obtained by least square method, had：

E=(L^TL)^-1L^TY (32)

Wherein,

Ψ⁽⁰⁾(1)=Ψ⁽⁰⁾(1) (38)

Ψ⁽¹⁾(k)=e^A(k-1)Ψ⁽¹⁾(1)+A^-1(e^A(k)-I)B (39)

Obtaining forecast model is：

Ψ⁽⁰⁾(k)=Ψ⁽¹⁾(k)-Ψ⁽¹⁾(k-1) (40)

Wherein k=2,3 ..., m；

1.5) finally, judge which class fault type forecast sample belongs to, each failure is established by existing indicator card information The regular set of type, i.e., to having determined that a number of indicator card of fault type by the way that surface dynamometer card is converted into down-hole pump Work(figure and its 14 curve Character eigenvectors are extracted respectively, obtain each curve Character eigenvector in each fault type Interval；Resulting interval byRepresent, p=1,2 ..., 14, whereinRepresent F_M The lower limit of p-th of curve Character eigenvector of kind fault type,Represent F_MP-th of Curve Moment of kind fault type is special Levy the higher limit of vector；

Calculate forecast sample each curve Character eigenvector value curve moment characteristics corresponding with each fault type to The distance of value, is calculated by following formula：

Wherein p=1,2 ..., 14；M=1,2,3 ...；ρ ∈ [0,1] are resolution ratio, take 0.5 herein；

<mrow> <msub> <mi>Grey</mi> <msub> <mi>F</mi> <mi>M</mi> </msub> </msub> <mo>=</mo> <mfrac> <mn>1</mn> <mi>p</mi> </mfrac> <munderover> <mo>&Sigma;</mo> <mrow> <mi>p</mi> <mo>=</mo> <mn>1</mn> </mrow> <mn>14</mn> </munderover> <msub> <mi>&xi;</mi> <mrow> <msub> <mi>pF</mi> <mi>M</mi> </msub> </mrow> </msub> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>43</mn> <mo>)</mo> </mrow> </mrow>

2. the Dlagnosis of Sucker Rod Pumping Well underground failure prediction method according to claim 1 based on Multi-variable Grey Model, its It is characterized in：The upstroke part of underground pump dynagraoph is moved for the suspension point of Dlagnosis of Sucker Rod Pumping Well sucker rod by minimum point in step 1.3) To the part of peak；Down stroke part is moved to the portion of minimum point for the suspension point of Dlagnosis of Sucker Rod Pumping Well sucker rod by peak Point.

3. the Dlagnosis of Sucker Rod Pumping Well underground failure prediction method according to claim 1 based on Multi-variable Grey Model, its It is characterized in：One section employed in multivariable grey forecasting model of the step 1.4) for Dlagnosis of Sucker Rod Pumping Well underground failure is continuous Sample in time, the unit of continuous time can be hour or day, and the quantity of the sample is 50-200, pre- to improve Survey computational efficiency and the degree of accuracy of model.

4. the Dlagnosis of Sucker Rod Pumping Well underground failure prediction method according to claim 1 based on Multi-variable Grey Model, its It is characterized in：Step 1.5) is established the regular set of each fault type by existing information, and existing information is by having learned that failure The indicator card of type obtains, for establishing the quantity of the indicator card of the regular set of each fault type more than 2.

5. the Dlagnosis of Sucker Rod Pumping Well underground failure prediction method according to claim 1 based on Multi-variable Grey Model, its It is characterized in：Dlagnosis of Sucker Rod Pumping Well underground fault type has following 9 kinds：" normal ", " gases affect ", " feed flow deficiency ", " take out Beam hanger, which breaks, to fall ", " travelling valve leakage ", " fixed valve leakage ", " being touched on pump ", " being touched under pump ", " sand production ".