CN104295286A - Intelligent identification method for operation condition of sucker rod type oil pumping unit - Google Patents

Abstract

The invention discloses an intelligent identification method for the operation condition of a sucker rod type oil pumping unit. The intelligent identification method comprises the steps that firstly, on-well indicator diagrams are collected and converted into underground indicator diagrams; qualitative analysis and quantitative calculation are conducted on the underground indicator diagrams, so that the feature values of the indicator diagrams of various working conditions are obtained; finally, the working state of the sucker rod type oil pumping unit is analyzed through the feature values of the indicator diagrams of various working conditions. By the adoption of the intelligent identification method for the operation condition of the sucker rod type oil pumping unit, the operation state of a current system can be accurately and rapidly judged when the operation condition of the sucker rod type oil pumping unit is identified in real time, the actual operation condition is provided, implementation is easy, operation is stable, the purpose that the working condition of the sucker rod type oil pumping unit is automatically identified, and the operation state of the system is adjusted in real time can be achieved easily , the overall efficiency is improved, and manpower and material resources are saved greatly.

Description

One has rod beam-pumping unit operation condition intelligent identification Method

Technical field

The invention belongs to field produces automation and information system management technical field, being specifically related to one has rod beam-pumping unit operation condition intelligent identification Method.

Background technology

Existing have rod beam-pumping unit operation mode recognition method, is all that namely technician or expert, by observing the indicator card measured by polished rod pump dynamograph, then carry out analysis interpretation to indicator card, to judge the duty of oil well rig manually.It is observed often with subjectivity, and different personnel are different with perception due to experience, cause and also have certain error to the accuracy of operating mode's switch, and makes sucker rod pumping machine operation mode recognition need to drop into a large amount of manpower and materials.

Summary of the invention

The object of this invention is to provide one and have rod beam-pumping unit operation condition intelligent identification Method, can accurate Quick current system running status, ensure the high efficiency work of oil pumper.

The technical solution adopted in the present invention is, one has rod beam-pumping unit operation condition intelligent identification Method, comprises the following steps:

Step 1: indicator card in production wells, is converted to downhole dynagraph;

Step 2, carries out qualitative analysis to downhole dynagraph and quantitatively calculates the characteristic value obtaining indicator card under various operating mode;

Step 3, has the duty of rod beam-pumping unit by the Eigenvalues analysis of indicator card under various operating mode.

Feature of the present invention is also,

The concrete steps of step 1 are as follows:

Step 1.1, the sample frequency of analog input signal is set in 13Hz ~ 40Hz, analog input signal (load F, displacement S) adopts the storage mode of two-dimensional vector, i.e. [Fn, Sn] vector of the n-th load F of gathering for input signal and displacement S, wherein the span of n is [1,200];

Step 1.2, be X along the displacement of well depth direction in the process that the sucker rod being provided with rod beam-pumping unit pumps, suffered gravitational load is Y; Then in the following way standardized data is obtained to input signal

$\stackrel{\‾}{X}=(X-X_{\mathrm{MIN}})/(X_{\mathrm{MAX}}-X_{\mathrm{MIN}})$

$\stackrel{\‾}{Y}=(Y-Y_{\mathrm{stdMIN}})/(Y_{\mathrm{stdMAX}}-Y_{\mathrm{stdMIN}});$

Wherein, X _mAXand X _mINmaximum value and the minimum value of displacement respectively, Y _stdMAXand Y _stdMINmaximum value and the minimum value of normal loading respectively;

Step 1.3, rejects the noise of analog input signal, and reflection has the operating mode of rod beam-pumping unit automatic control system more really; Gibbs one-order wave equation and fringe conditions is utilized to convert gathered vector;

$\left\{\begin{array}{c}\frac{{\∂}^{2}S(x,t)}{{\∂t}^2}=c^2\frac{{\∂}^{2}S(x,t)}{{\∂x}^2}-C\frac{\∂S(x,t)}{\∂t}\\ S(x,t){|}_{x=0}=S\left(t\right)\\ F(x,t){|}_{x=0}=L\left(t\right)-W_r=D\left(t\right)\\ {\left(F_{i,j}\right)}_1={\left(F_{i,j}\right)}_2\\ {\left(s_{i,j}\right)}_1={\left(s_{i,j}\right)}_2\end{array}\right.;$

Wherein, S (x, t) is that a is the spread speed of stress wave in rod string at the displacement function of x section not t in the same time, C is viscous damping coefficient, S (t) is measured indicator diagram displacement, and L (t) is load vector, and Wr is the weight of rod string in well liquid, F (x, t) for sucker rod is at x section not elastic force function in the same time, D (t) to move upward the elastic force received at piston for sucker rod from bottom dead centre, F _ijand s _ijelastic force and the displacement at multistage bar tie point place respectively.The vector that input vector obtains through the conversion of above-mentioned equation is downhole dynagraph vector.

The concrete steps of step 2 qualitative analysis and quantitatively calculating are as follows:

Step 2.1, extract area features:

Image data exists with two-dimensional vector form, comprises two elements, i.e. load, displacement in a vector; In a jig frequency, indicator card area adopts following computational methods:

$S=\underset{n\→\∞}{\lim }\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}({F_i}^u-{F_i}^d)(X_{i+1}-X_i),$

Wherein, (F _i ^u, X _i) and (F _i ^d, X _i) be the data of upstroke in the two-dimensional vector of image data in corresponding indicator card and down stroke respectively, n gets 200;

Step 2.2, extracts curvature feature: calculating curvature is the flex point in order to look on indicator card, i.e. opening point, the closing point of standing valve and travelling valve, any point P on indicator card curve _i(S _i, f _i) curvature K _iaccording to 5 adjacent with it some P _i-2, P _i-1, P _i, P _i+ 1, P _igeometrical relationship between+2 calculates, P _idiscrete point curvature calculate:

K _i＝Δθ _i/l _i

In formula, Δ θ _istraight line P _i-2P _ito straight line P _i+ 2P _ithe oriented anglec of rotation;

$\left\{\begin{array}{c}\mathrm{\Δ}{\mathrm{\θ}}_i=\arctan \left(\frac{(f_i-f_{i+2})(S_i-S_{i-2})-(f_i-f_{i-2})(S_i-S_{i-2})}{(S_i-S_{i+2})(S_i-S_{i-2})+(f_i-f_{i+2})(f_i-f_{i-2})}\right)\\ l_i=\sqrt{{(S_{i+1}-S_i)}^2+{(f_{i+1}-f_i)}^2}+\sqrt{{(S_i-S_{i-1})}^2+{(f_i-f_{i-1})}^2}\end{array}\right.;$

In formula, f _iand S _iload and the displacement of i-th point on indicator card curve respectively;

Step 2.3, extract invariant moment features:

Bending moment is not stablized due to clear concept, discrimination, has good consistency and anti-interference, effectively can reflect the substantive characteristics of figure to the figure with rotation and ratio change, utilize 7 of Hu.M.K not bending moment obtain invariant moment features value:

In formula, for not bending moment (i=1,2 ... 7), η _pqfor center, p+q rank square (p=0,1,2,3; Q=0,1,2,3).

Step 3 concrete steps are as follows:

First expert system processes the characteristic value extracted, and according to knowledge rule, namely the reasoning of expert system adopts the mode of forward reasoning to draw large class belonging to this indicator card, and dopes most probable operating mode for final conclusion reference; After the first order identification of expert system completes, BP neutral net corresponding under the large class that expert system provides starts the second level and identifies;

Characteristic value passes to the input of BP neutral net, and after whole network, namely output obtains the vector of recognition result, and system is resolved this vector and drawn final recognition result according to weights with reference to the prediction operating mode that provided by expert system.

The invention has the beneficial effects as follows,

1. the present invention has rod beam-pumping unit operation condition intelligent identification Method to can accurate Quick current system running status during sucker rod pumping machine operation condition Real time identification, provides operation actual condition.

2. the present invention has rod beam-pumping unit operation condition intelligent identification Method, is easy to realize, stable, is conducive to realizing sucker rod pumping machine and automatically identifies that operating mode adjusts in time to system running state, promotes whole efficiency, saves a large amount of manpower and materials simultaneously.

Accompanying drawing explanation

Fig. 1 is the flow chart that the present invention has rod beam-pumping unit operation condition intelligent identification Method;

Fig. 2 is the flow chart that indicator card is drawn in two-dimensional vector data collection;

Fig. 3 is expert system reasoning flow chart.

Detailed description of the invention

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.

As Fig. 1 has rod beam-pumping unit operation condition intelligent identification Method, comprise the following steps:

Step 1: indicator card in production wells, is converted to downhole dynagraph, and concrete steps are as follows:

Step 1.1, for the integrality and Complete Characterization that ensure input signal data information have the operating mode feature of rod beam-pumping unit automatic control system, the variation tendency of the input signal perfect representation indicator card curve namely gathered, and the resolution ratio of indicator card curve is finer and smoother, four region indicator card shapes are clear, reduce the information of primary signal by discrete sampling point; The sample frequency of analog input signal should be set in 13Hz ~ 40Hz, analog input signal (load, displacement) adopts the storage mode of two-dimensional vector, be [Fn, Sn] vector of the n-th load F of gathering for input signal and displacement S, wherein the span of n is [1,200].

Step 1.2, for ensuring reliability and the accuracy of rod beam-pumping unit operation mode recognition, by conversion collection signal become algorithm for pattern recognition can standard digital form, make all image data in a unified yardstick and space dimensionality.Being provided with in the process that rod beam-pumping unit sucker rod pumps along the displacement of well depth direction is X, and suffered gravitational load is Y; Then in the following way standardized data is obtained to input signal

$\stackrel{\‾}{X}=(X-X_{\mathrm{MIN}})/(X_{\mathrm{MAX}}-X_{\mathrm{MIN}})$

$\stackrel{\‾}{Y}=(Y-Y_{\mathrm{stdMIN}})/(Y_{\mathrm{stdMAX}}-Y_{\mathrm{stdMIN}})$

Wherein, X _mAXand X _mINmaximum value and the minimum value of displacement respectively, Y _stdMAXand Y _stdMINmaximum value and the minimum value of normal loading respectively.

Step 1.3, rejects the noise of input signal, and reflection has the operating mode of rod beam-pumping unit automatic control system more really; Gibbs one-order wave equation and fringe conditions is utilized to convert gathered vector:

$\left\{\begin{array}{c}\frac{{\∂}^{2}S(x,t)}{{\∂t}^2}=c^2\frac{{\∂}^{2}S(x,t)}{{\∂x}^2}-C\frac{\∂S(x,t)}{\∂t}\\ S(x,t){|}_{x=0}=S\left(t\right)\\ F(x,t){|}_{x=0}=L\left(t\right)-W_r=D\left(t\right)\\ {\left(F_{i,j}\right)}_1={\left(F_{i,j}\right)}_2\\ {\left(s_{i,j}\right)}_1={\left(s_{i,j}\right)}_2\end{array}\right.$

Wherein, S (t) is displacement vector, and L (t) is load vector, and Wr is the weight of rod string in well liquid, and the vector that input vector obtains through the conversion of above-mentioned equation is downhole dynagraph vector.

Step 2: carry out qualitative analysis to downhole dynagraph and quantitatively calculate the characteristic value obtaining indicator card under various operating mode, the concrete steps of qualitative analysis and quantitative calculating are as follows:

Step 2.1, extract area features:

Image data exists with two-dimensional vector form, comprises two elements (load, displacement) in a vector.In a jig frequency, indicator card area adopts following computational methods:

$S=\underset{n\→\∞}{\lim }\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}({F_i}^u-{F_i}^d)(X_{i+1}-X_i)$

Wherein, (F _i ^u, X _i) and (F _i ^d, X _i) be the data of upstroke in the two-dimensional vector of image data in corresponding indicator card and down stroke respectively, n gets 200;

Step 2.2, extract curvature feature:

Calculating curvature is the flex point in order to look on indicator card, i.e. opening point, the closing point of standing valve and travelling valve.Any point P on indicator card curve _i(S _i, f _i) curvature K _ican according to 5 adjacent with it some P _i-2, P _i-1, P _i, P _i+ 1, P _igeometrical relationship between+2 calculates, P _idiscrete point curvature calculate:

K _i＝Δθ _i/l _i

In formula, Δ θ _istraight line P _i-2P _ito straight line P _i+ 2P _ithe oriented anglec of rotation:

$\left\{\begin{array}{c}\mathrm{\Δ}{\mathrm{\θ}}_i=\arctan \left(\frac{(f_i-f_{i+2})(S_i-S_{i-2})-(f_i-f_{i-2})(S_i-S_{i-2})}{(S_i-S_{i+2})(S_i-S_{i-2})+(f_i-f_{i+2})(f_i-f_{i-2})}\right)\\ l_i=\sqrt{{(S_{i+1}-S_i)}^2+{(f_{i+1}-f_i)}^2}+\sqrt{{(S_i-S_{i-1})}^2+{(f_i-f_{i-1})}^2}\end{array}\right.;$

In formula, f _iand S _iload and the displacement of i-th point on indicator card curve respectively.

Step 2.3, extract invariant moment features:

Bending moment is not stablized due to clear concept, discrimination, has good consistency and anti-interference, effectively can reflect the substantive characteristics of figure to the figure with rotation and ratio change.Utilize 7 of Hu.M.K not bending moment obtain invariant moment features value:

In formula, for not bending moment (i=1,2 ... 7), η _pqfor center, p+q rank square (p=0,1,2,3; Q=0,1,2,3).

Different operating mode graph of a correspondence shapes is different, and the characteristic value extracting different graphic so just can utilize algorithm to identify corresponding operating mode.

Step 3: by the duty of the Eigenvalues analysis oil pumper of indicator card under various operating mode, concrete steps are as follows:

1. one-level expert system identification.Using expert system as first order recognition system, neutral net is as second level recognition system and provide result.First expert system processes the characteristic value extracted, and according to knowledge rule, the inference machine of expert system adopts the mode of forward reasoning to draw large class belonging to this indicator card, and dopes most probable operating mode for final conclusion reference.After the first order identification of expert system completes, BP neutral net corresponding under the large class that expert system provides starts the second level and identifies.

2. Secondary Neural Networks identification.Characteristic value passes to the input of BP neutral net, and after whole network, output can obtain the vector of recognition result, and system is resolved this vector and drawn final recognition result according to weights with reference to the prediction operating mode that provided by expert system.

Because classification reduces by first order recognition system, therefore the nerve net of the second level does not need very complicated, which improves BP neural network learning and recognition speed.The characteristic value extracted the normal data of different operating modes exports as input and its mapping and repeatedly trains and learn.For ensureing that the generalization ability of neutral net does not occur Expired Drugs and considers the stability of system, anticipation error is set to 0.001, and learning rate is set to 0.75.

If UPSflag in Fig. 2 is up stroke end mark, S is the displacement vector of Real-time Collection, and F is the load vector of Real-time Collection, and Sm is the maximum value of displacement vector, and Fstr is the value of indicator card starting point load vector, and Fend is indicator card terminal load vector value.Because oil pumper is ceaselessly pumping, so setting horse head is angular displacement initial point when bottom dead centre, upstroke direction is the positive direction of displacement.With displacement at initial point and pulling force non-vanishing be indicator card data start flag, be indicator card end-of-data mark when initial point is got back in displacement again thus the data of one group of complete jig frequency can be found out.But, because each stroke not necessarily can get back to the initial point of demarcation completely, and not necessarily gather when oil pump piston reaches and demarcates initial point due to frequency acquisition, thus can only by finding out displacement and being reduced to minimum value time as initial and end mark.

There is rod beam-pumping unit data message as first read from parameter library in Fig. 3, comprising characteristic value and other electrical quantitys, by mating the large rule-like of first floor one by one, indicator card being belonged to which large class and differentiating out.Then mate one by one in the little rule-like comprised under this large class, until to reach a conclusion or rule uses until exhausted completely.

Claims (4)

1. there is a rod beam-pumping unit operation condition intelligent identification Method, it is characterized in that, comprise the following steps:

Step 1: indicator card in production wells, is converted to downhole dynagraph;

Step 2, carries out qualitative analysis to downhole dynagraph and quantitatively calculates the characteristic value obtaining indicator card under various operating mode;

Step 3, has the duty of rod beam-pumping unit by the Eigenvalues analysis of indicator card under various operating mode.

2. have rod beam-pumping unit operation condition intelligent identification Method as claimed in claim 1, it is characterized in that, the concrete steps of step 1 are as follows:

Step 1.1, the sample frequency of analog input signal is set in 13Hz ~ 40Hz, and analog input signal load F, displacement S adopt the storage mode of two-dimensional vector, i.e. [Fn, Sn] vector of the n-th load F of gathering for input signal and displacement S, wherein the span of n is [1,200];

Step 1.2, be X along the displacement of well depth direction in the process that the sucker rod being provided with rod beam-pumping unit pumps, suffered gravitational load is Y; Then in the following way standardized data is obtained to input signal

$\stackrel{\‾}{X}=(X-X_{\mathrm{MIN}})/(X_{\mathrm{MAX}}-X_{\mathrm{MIN}})$

$\stackrel{\‾}{Y}=(Y-Y_{\mathrm{stdMIN}})/(Y_{\mathrm{stdMAX}}-Y_{\mathrm{stdMIN}});$

Wherein, X _mAXand X _mINmaximum value and the minimum value of displacement respectively, Y _stdMAXand Y _stdMINmaximum value and the minimum value of normal loading respectively;

Step 1.3, rejects the noise of analog input signal, and reflection has the operating mode of rod beam-pumping unit automatic control system more really; Gibbs one-order wave equation and fringe conditions is utilized to convert gathered vector:

$\left\{\begin{array}{c}\frac{{\∂}^{2}S(x,t)}{{\∂t}^2}=c^2\frac{{\∂}^{2}S(x,t)}{{\∂x}^2}-C\frac{\∂S(x,t)}{\∂t}\\ S(x,t){|}_{x=0}=S\left(t\right)\\ F(x,t){|}_{x=0}=L\left(t\right)-W_r=D\left(t\right)\\ {\left(F_{i,j}\right)}_1={\left(F_{i,j}\right)}_2\\ {\left(s_{i,j}\right)}_1={\left(s_{i,j}\right)}_2\end{array}\right.;$

Wherein, S (x, t) is that a is the spread speed of stress wave in rod string at the displacement function of x section not t in the same time, C is viscous damping coefficient, S (t) is measured indicator diagram displacement, and L (t) is load vector, and Wr is the weight of rod string in well liquid, F (x, t) for sucker rod is at x section not elastic force function in the same time, D (t) to move upward the elastic force received at piston for sucker rod from bottom dead centre, F _ijand s _ijelastic force and the displacement at multistage bar tie point place respectively; The vector that input vector obtains through the conversion of above-mentioned equation is downhole dynagraph vector.

3. have rod beam-pumping unit operation condition intelligent identification Method as claimed in claim 1, it is characterized in that, the concrete steps of step 2 qualitative analysis and quantitatively calculating are as follows:

Step 2.1, extract area features:

Image data exists with two-dimensional vector form, comprises two elements, i.e. load, displacement in a vector; In a jig frequency, indicator card area adopts following computational methods:

$S=\underset{n\→\∞}{\lim }\underset{i=0}{\overset{n-1}{\mathrm{\Σ}}}({F_i}^u-{F_i}^d)(X_{i+1}-X_i),$

Wherein, (F _i ^u, X _i) and (F _i ^d, X _i) be the data of upstroke in the two-dimensional vector of image data in corresponding indicator card and down stroke respectively, n gets 200;

Step 2.2, extract curvature feature:

Calculating curvature is the flex point in order to look on indicator card, i.e. opening point, the closing point of standing valve and travelling valve, any point P on indicator card curve _i(S _i, f _i) curvature K _iaccording to 5 adjacent with it some P _i-2, P _i-1, P _i, P _i+ 1, P _igeometrical relationship between+2 calculates, P _idiscrete point curvature calculate:

K _i＝Δθ _i/l _i

In formula, Δ θ _istraight line P _i-2P _ito straight line P _i+ 2P _ithe oriented anglec of rotation:

$\left\{\begin{array}{c}\mathrm{\Δ}{\mathrm{\θ}}_i=\arctan \left(\frac{(f_i-f_{i+2})(S_i-S_{i-2})-(f_i-f_{i-2})(S_i-S_{i-2})}{(S_i-S_{i+2})(S_i-S_{i-2})+(f_i-f_{i+2})(f_i-f_{i-2})}\right)\\ l_i=\sqrt{{(S_{i+1}-S_i)}^2+{(f_{i+1}-f_i)}^2}+\sqrt{{(S_i-S_{i-1})}^2+{(f_i-f_{i-1})}^2}\end{array}\right.;$

In formula, f _iand S _iload and the displacement of i-th point on indicator card curve respectively;

Step 2.3, extract invariant moment features:

Bending moment is not stablized due to clear concept, discrimination, has good consistency and anti-interference, effectively can reflect the substantive characteristics of figure to the figure with rotation and ratio change, utilize 7 of Hu.M.K not bending moment obtain invariant moment features value:

In formula, for not bending moment (i=1,2 ... 7), η _pqfor center, p+q rank square (p=0,1,2,3; Q=0,1,2,3).

4. have rod beam-pumping unit operation condition intelligent identification Method as claimed in claim 1, it is characterized in that, step 3 concrete steps are as follows:

First expert system processes the characteristic value extracted, and according to knowledge rule, namely the reasoning of expert system adopts the mode of forward reasoning to draw large class belonging to this indicator card, and dopes most probable operating mode for final conclusion reference; After the first order identification of expert system completes, BP neutral net corresponding under the large class that expert system provides starts the second level and identifies;

Characteristic value passes to the input of BP neutral net, and after whole network, namely output obtains the vector of recognition result, and system is resolved this vector and drawn final recognition result according to weights with reference to the prediction operating mode that provided by expert system.