CN100511041C - Petroleum well drilling engineering accidents early-warning system based on layered fuzzy system - Google Patents
Petroleum well drilling engineering accidents early-warning system based on layered fuzzy system Download PDFInfo
- Publication number
- CN100511041C CN100511041C CNB2007100551786A CN200710055178A CN100511041C CN 100511041 C CN100511041 C CN 100511041C CN B2007100551786 A CNB2007100551786 A CN B2007100551786A CN 200710055178 A CN200710055178 A CN 200710055178A CN 100511041 C CN100511041 C CN 100511041C
- Authority
- CN
- China
- Prior art keywords
- variable
- fuzzy
- input
- signal
- early warning
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Alarm Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses an oil drilling engineering accident early warning system based on a hierarchy fuzzy system. First, a stimulant database for oil drilling engineering accident early warning is created; second, an input variable for oil drilling engineering accident early warning; third, the characteristic information of input signals is analyzed; fourth, a structural model of the hierarchy fuzzy system is created; fifth, an interval mapping and a hierarchy fuzzy reasoning of a unified model are done for the input variable in the third step; sixth, the output quantity of the oil drilling engineering accident early warning system is confirmed; seventh, the self-adjustment of the early warning result is done. The present invention has the advantages that the hierarchy fuzzy system is used to translate the excessive physical variable in the oil drilling engineering accident early warning system into a hierarchy system formed by connecting hierarchically with a reduced dimensionality fuzzy unit, and make the regular number of the fuzzy system increase with the number of the input variable linearly, thereby putting an end to the phenomenal of dimension calamity, and realizing the purpose of early warning for the oil drilling engineering accident accurately, efficiently and conveniently.
Description
Technical field
The present invention relates to a kind of application of Fuzzy Inference Model in petroleum drilling engineering of many input physical descriptors, especially relate to petroleum well drilling engineering accidents method for early warning based on layered fuzzy system.
Background technology
Petroleum drilling is an excessive risk, high expensive systems engineering, in the wellbore construction process, exist a large amount of ambiguities, randomness and uncertain problem, the possibility that engineering accident takes place exists at any time, in case take place to cause fund and temporal huge waste and loss.The generation of petroleum well drilling engineering accidents and development are the multifactorial processes of a complexity, are difficult to be described and early warning with classical mathematical model; And people can not guarantee accurately and timely to carry out early warning according to the experience and knowledge of putting into practice with the cognitive differentiation drilling failure of being set up.Therefore, the inference pattern that utilizes fuzzy theory to set up accident be the effective ways of describing petroleum well drilling engineering accidents also be the basis of carrying out the engineering accident early warning.Petroleum well drilling engineering accidents early-warning system is a uncertain system with the input of super large dimension, output physical descriptor, can occur dimension calamity problem when using existing fuzzy theory inevitably: the petroleum drilling sensor information of accident early warning institute foundation and derived information quantity thereof surpass 40 (M〉40), if fuzzy subset's number of each input variable is 7 (N=7), accident (output) kind is 10 (T=10), and then the regular sum of early warning system will reach S=T*N
M=10*7
40, typical " dimension calamity " problem that Here it is.Therefore, early warning is unpractical to petroleum well drilling engineering accidents to utilize existing traditional fuzzy theory.
Summary of the invention
The object of the invention is to provide a kind of petroleum well drilling engineering accidents method for early warning based on layered fuzzy system that is suitable for commercial Application.
For achieving the above object, the present invention can take following technical proposals:
Petroleum well drilling engineering accidents method for early warning based on layered fuzzy system of the present invention, it is made of following step:
The first step, set up the model database of petroleum well drilling engineering accidents early warning:
Various sensor signals and derivation parameter data from the field integrated logging equipment collection of petroleum drilling, analyze by data message long-term accumulation,,, set up petroleum well drilling engineering accidents Early-warning Model database in conjunction with petroleum drilling operating personnel and relevant expert's experimental knowledge; Whole process to petroleum drilling is determined different duties according to the logical combination relation and the drilling process standard of sensor situation of change, these duties have different characteristics each other, and kind that may have an accident and sensor variable sign also have nothing in common with each other;
Second goes on foot, sets up the input variable of petroleum well drilling engineering accidents early warning:
According to the information in the model database in the first step, utilize corresponding sensor acquisition data, the data of being gathered mainly contain: utilize pressure transducer to gather signals such as weight on hook, standpipe pressure, casing pressure, moment of torsion; Utilize temperature sensor to gather the temperature signal of gateway; Utilize conductivity sensor to gather the signal of gateway conductivity; Utilize flow sensor to gather the rate of discharge signal; Utilize density sensor to gather the gateway density signal; Utilize pulse transducer to gather that hook height, pump dash, the rotary speed signal; Utilize liquid level sensor acquired volume signal; Utilize hydrogen sulfide sensor to gather the sulfureted hydrogen gas concentration signal;
The characteristic information of the 3rd step, analysis input variable
Determined sensor variable of second step is determined the characteristic informations such as fluctuating range, vibration frequency, rate of change and variation tendency of sensor variable in each drilling well course of work by adopting statistical dependence analysis, wavelet transformation, spectrum analysis, time-domain analysis method; Is corresponding different intermediate features variable with same sensor variable according to the different decomposition of its variation characteristic by these information; According to state-contingent in the drilling process and progressive relationship, actual measured results in conjunction with various parameter sensors, according to the requirement of production technology, drilling process is divided into some typicalnesses, and its kind corresponding sensor variable is incorporated into to pairing state model handle.To the data decomposition in each drilling state is steady state data and Temporal Data; Input variable signal in each process is carried out AR model filtering and least square optimization, wavelet de-noising filtering and the dynamically calculating of average, determine the normal variation scope of signal;
The 4th goes on foot, sets up the structural model of layered fuzzy system:
To the 3rd step determined n input variable x
1, x
2... x
n, being respectively the feature input variable in the accident pre-alarming system, the basic blur unit of each layer all has two input variables, wherein x
1, x
2As the input of the basic blur unit of ground floor, it exports y
1And the 3rd input variable x
3As the input of the basic blur unit of the second layer, all the other each layers by that analogy; Each layer all has only two input variables;
The 5th goes on foot, the input variable of determining in the 4th step is carried out the hierarchical fuzzy reasoning of interval mapping and unified model:
To the 4th step determined n different types of input variable, by statistical study to the actual physics variable, determine the variation range and the variation characteristic of this input variable, utilize and analyze the mapping relations that obtain, these input variables are mapped on the unified interval, on this basis the basic blur unit of each layer of layered fuzzy system are carried out the fuzzy reasoning of unified model;
The 6th goes on foot, determines the petroleum well drilling engineering accidents early-warning system output quantity:
The final output quantity of petroleum well drilling engineering accidents early-warning system is the basic blur unit output of each layer result's a weighted sum, promptly
Wherein k is a hierarchy number, and α is a weighting coefficient,
y
kBe the result behind each layer output ambiguity solution;
The 7th step, early warning result's self study and adjustment:
Realize that by the mode that input quantity domain mapping is adjusted promptly the evaluation index to the early warning result is set at four kinds: analyzes suitable, wrong report, low newspaper, the high newspaper; Except that first kind of evaluation do not need systematic parameter adjusted, other evaluation result was all revised relevant parameter by system.
In the 4th step, take vacant substitute mode to handle to the input variable of hierarchical fuzzy model:
Concrete processing procedure is: when having defined the off-note variable that the characteristic variable number that causes a certain accident occurs in more than actual engineering in model database, unusual characteristic variable do not occur and do not enter the blur unit layer, by occurring unusual characteristic variable substitute in the actual engineering of the next one as input quantity.
In the 5th step, fuzzy reasoning to layered fuzzy system adopts the mode of unified model to handle: detailed process is for passing through the unified domain interval of definition, analyze the variation range of actual physics input quantity, the input quantity unification is mapped on the defined interval, and this interval is promptly as the basic domain of fuzzy system; Each basic blur unit adopts identical fuzzy reasoning mode, and promptly identical domain, fuzzy method and unified fuzzy inference rule, fuzzy reasoning adopt and calculate easy Mamdani composition algorithm.
The invention has the advantages that and utilize layered fuzzy system, many input physical descriptors in the petroleum well drilling engineering accidents early-warning system are converted into the hierarchical system that is connected and composed by the layering of low-dimensional blur unit, make regular several of fuzzy system to count linear growth with input variable, stopped the generation of " dimension calamity " phenomenon, utilize signal processing technology to obtain to reflect the signal of physics input variable variation tendency, and carry out fuzzy reasoning by the layered fuzzy system of unified model, thereby realize accurate and effective, petroleum well drilling engineering accidents is carried out the purpose of early warning easily.
Description of drawings
Fig. 1 is the structural representation of layered fuzzy system of the present invention.
Fig. 2 is a membership function curve map of the present invention.
Fig. 3 is the petroleum well drilling engineering accidents method for early warning block diagram based on layered fuzzy system of the present invention.
Fig. 4 is the layered fuzzy system block diagram based on unified model of the present invention.
Embodiment
Petroleum well drilling engineering accidents method for early warning based on layered fuzzy system of the present invention is made of following step:
The first step, set up the model database of petroleum well drilling engineering accidents early warning:
At first, by summing up petroleum drilling operating personnel and relevant expert's experimental knowledge, set up the model database of petroleum well drilling engineering accidents early warning.Whole process to petroleum drilling is determined five different duties according to the logical combination relation and the drilling process standard of sensor states, these duties have different characteristics, and kind that may have an accident and sensor variable sign also have nothing in common with each other.The model database of having set up is as shown in table 1, and this database is an open system, can make amendment and replenishes according to actual conditions and new expertise.
Second goes on foot, sets up the input variable of petroleum well drilling engineering accidents early warning:
According to the information in the model database, utilize corresponding sensor acquisition data, mainly contain: utilize pressure transducer to gather signals such as weight on hook, standpipe pressure, casing pressure, moment of torsion; Utilize temperature sensor to gather the temperature signal of gateway; Utilize conductivity sensor to gather the signal of gateway conductivity; Utilize flow sensor to gather the rate of discharge signal; Utilize density sensor to gather the gateway density signal; Utilize signals such as pulse transducer collection hook height, pump dash, rotary speed; Utilize liquid level sensor acquired volume signal; Utilize hydrogen sulfide sensor to gather the sulfureted hydrogen gas concentration signal;
The characteristic information of the 3rd step, analysis input variable
According to state-contingent in the drilling process and transforming relationship, actual measured results in conjunction with various parameter sensors, according to the requirement of production technology, drilling process is divided into some typicalnesses, and its kind corresponding sensor parameter is incorporated into to pairing state model handle.To the data decomposition in each drilling state is steady state data and Temporal Data; The sensor variable of being gathered is analyzed and handled, by adopting filtering, statistical dependence is analyzed, wavelet transformation, spectrum analysis, methods such as time-domain analysis are determined the fluctuating range of sensor variable in each drilling well course of work, vibration frequency, characteristic information such as rate of change and variation tendency, is corresponding different intermediate variable with same physical sensors variable according to the different decomposition of its variation characteristic by these information, can be decomposed into weight on hook as this sensor variable of weight on hook rises, weight on hook descends, the weight on hook fluctuation, different characteristic variables such as the weight on hook rate of change is unusual, these characteristic variables belong to intermediate variable, but they have clear and definite physical meaning, corresponding with the changing features of sensor variable in the actual engineering, be convenient to the field technician and extract and understand, the technician also can make amendment to these characteristic variables by model database simultaneously, deletion, operations such as interpolation.
(1) the AR model filtering device of input signal and least square optimization thereof
Filtering algorithm is y (i)=a
1X (i)+a
2X (i-1)+Λ+a
nX (i-n+1), x is the sampled value of parameter in the formula, i is current sampling instant, a is a filter factor, z is the filter value of this input signal, and the item number of right-hand member x can change in the formula, such as only getting preceding two or first three items, be referred to as the order of x, order is high more, and the filtering degree is high more, and is severe more to the inhibition of sampled value, order is low, and parameter changes sensitive more; Determining of order can be according to determining comprehensively that to the requirement of sampled data fluctuation and to the requirement of sampling delay order selects 3 ~ 5 generally speaking.In petroleum well drilling engineering accidents early-warning system, the purpose of filtering is the variation tendency that will reflect input signal, so order is defined as 5.
In the model parameter a determine utilize least square method optimization to obtain, its computation process is:
y(i)＝H(i)*A
Wherein, H (i)=[x (i), x (i-1), x (i-2), x (i-3), x (i-4)]
A＝[a(1)，a(2)，a(3)，a(4)，a(5)]
Because y is filter value, be actually a unknown quantity constantly at i, but this amount should be known in formula, so we are with its parameter as expectation, the numerical value of this parameter is to utilize the given data of off-line to carry out 5 average smoothing methods to obtain:
Result after i is level and smooth constantly is y (i), and sampled value is x, then
Data for the not enough smooth length in data segment two ends (smooth length is 5 here) are then no longer carried out smoothing processing, directly get sampled value as smooth value.
After handling like this,, can obtain filter factor and be according to the least square optimum theory:
A＝(H
T*H)
-1*H
T*Y
(2) wavelet de-noising wave filter
The wavelet de-noising wave filter is by carrying out wavelet decomposition to the input variable signal, high frequency detail coefficients after decomposing is suppressed, high frequency coefficient and low frequency coefficient after will suppressing then are reconstructed, and obtain filtered signal, and this signal can truly reflect the variation tendency of actual signal.
(3) steady state data signature analysis
Each input variable information enter a new drilling state and data are relatively stable, do not have serious vibration after data between next state transition point be called the stable data of this state.Stablize the important evidence that the abnormal occurrence of the information characteristics of data such as filtering average, square error, variation tendency, change threshold, harmonic component and power spectrum thereof etc. will take place as the judgement accident.
(4) Temporal Data signature analysis
Each heat transfer agent gets starting point is called this state to the data between the data stabilization point Temporal Data entering a new drilling state.The abnormal occurrence of the information characteristics of Temporal Data such as overshoot, oscillation peak, oscillation frequency, the number of oscillation, rise time etc. will be as the important evidence of judgement accident generation.
(5) analysis of threshold
By above-mentioned Filtering Processing, can obtain the variation tendency signal of input signal, in steady-state process and transient state process, this signal is asked for dynamic average, signal reference value, can obtain in this signal normal variation process variation range up and down by statistical study with respect to reference value, thereby can determine the last lower threshold value of this input signal normal variation, can obtain signal normal variation scope by calculating dynamic average and signal threshold value, when actual signal numerical value surpass this scope the time, promptly handle as abnormal signal.
The 4th goes on foot, sets up the structural model of layered fuzzy system:
The sensor variable of petroleum well drilling engineering accidents early-warning system reaches dozens of, the characteristic variable of extracting is more, also to consider the real-time requirement of system when will carry out accident early warning, therefore, adopt the hierarchical fuzzy model of following structure according to these information.
As shown in Figure 1, to second step determined n input variable x
1, x
2... x
n, being respectively the feature input variable in the accident pre-alarming system, the basic blur unit of each layer all has two input variables, wherein x
1, x
2As the input of the basic blur unit of ground floor, it exports y
1And the 3rd input variable x
3As the input of the basic blur unit of the second layer, all the other each layers by that analogy; Each layer all has only two input variables.
Consider that the variable that causes a certain accident in the petroleum drilling engineering of reality is not unalterable, may be less than determined variable in the model database, has certain uncertainty, adopt the mode of vacant substitute to handle in the hierarchical fuzzy model, concrete processing procedure is: suppose to have defined the feature input variable that causes a certain accident and be followed successively by x in model database for this reason
1, x
2, x
3, x
4, can constitute one three layers Fuzzy Inference Model like this by above-mentioned hierarchical fuzzy structure of models, the input variable of each layer is respectively x
1And x
2, x
3, x
4, but in actual engineering, have only characteristic variable x
1, x
2, x
4Occurred unusually, adopted the mode of vacant substitute will originally be in the 3rd layer variable x
4The original x of the second layer of substituting
3The position, the hierarchical fuzzy model becomes two-layer structure thus, each layer input variable is respectively x
1And x
2, x
4
Each layer blur unit in the hierarchical fuzzy model set up unified fuzzy model, use same fuzzy inference rule, to solve real-time and the unusual uncertain problem of input variable in the actual use.Therefore system has the simplicity of good opening, extendability and calculating, also make the Fuzzy Inference Model of accident early warning have good versatility, be convenient to according to reality require to carry out the adjustment of input variable and according to expertise to the fuzzy rule adjustment, reduced the complicacy of system.
The interval mapping of the 5th step, input variable:
In petroleum well drilling engineering accidents early-warning system, the kind difference of input variable, physical unit also has nothing in common with each other, carry out fuzzy reasoning for ease of the unified model that uses layered fuzzy system, by statistical study to the actual physics variable, determine the variation range and the variation characteristic of this input variable, utilize and analyze the mapping relations that obtain, these input variables are mapped on the unified interval, carry out carrying out fuzzy reasoning by the blur unit of unified model after the obfuscation.If the normal variation scope of some input variables (or characteristic variable) v is v
1～v
2, the passing threshold analysis can determine that the threshold range up and down of its ANOMALOUS VARIATIONS is [v
Min-, v
Max-], [v
Min+, v
Max+], v
Max-≤ v
1, v
Min+〉=v
2, like this by linear mapping with actual threshold scope [v
Min-, v
Max-], [v
Min+, v
Max+] be mapped to a unified interval [d respectively
1, d
2] on, its mapping relations are d=k
1V+k
2, k
1, k
2Be respectively mapping coefficient, can be defined as
k
2＝d
1-k
1*v
min。
Be real-time, the minimizing operand of enhanced system simultaneously, the mode that system activates when ANOMALOUS VARIATIONS is adopted in the processing of input variable, just when data processing module judging characteristic variable is in normal variation, it is not carried out fuzzy operation, have only and judge this characteristic variable when data processing module and be in when unusual, just activate corresponding fuzzy model and carry out rational analysis.
The fuzzy reasoning of the 6th step, unified model:
Layered fuzzy system adopts the fuzzy reasoning of unified model, its principal character is that the fuzzy reasoning process of each basic blur unit is adopted in a like fashion: each input variable adopts identical obfuscation mode, on identical domain interval, carry out obfuscation and ambiguity solution, adopt identical fuzzy inference rule, therefore has good opening, the simplicity of extendability and calculating, also make the Fuzzy Inference Model of accident early warning have good versatility, be convenient to require to carry out the adjustment of input variable and fuzzy rule is adjusted, reduced the complicacy of system according to expertise according to reality.
The specific implementation process prescription is as follows:
The obfuscation of two input variables and output variable is all carried out on unified domain interval in the basic blur unit, for petroleum well drilling engineering accidents early-warning system, be defined as [1 between this map section, 10], input variable is mapped on this interval by different mapping relations, and obtain the fuzzy quantity of input variable with unified subordinate function, specific algorithm is as follows:
1, definitional language variable: to the description of accident degree, define four linguistic variables in producing according to petroleum drilling: " little ", " in ", " greatly ", " very big ".The exception level of accident and characteristic variable just represents that with these four fuzzy language variablees its membership function is the discrete type subordinate function on domain interval [1,10].
2, definition domain interval: the numerical value of each characteristic variable is mapped on the interval [1 ~ 10] according to different mapping coefficients, though the numerical values recited difference of each physical descriptor like this, all mapping is distributed on the interval 1 ~ 10.
3, definition membership function: according to the description of producers to the accident degree, can adopt general trapezoidal subordinate function, promptly for linguistic variable " in ", " greatly " all adopts unified membership function, it is trapezoidal profile, and linguistic variable " little " is adopted Z type subordinate function, linguistic variable " very big " is adopted S type subordinate function, as shown in Figure 2.
The analytical expression of these subordinate functions is:
Trapezoidal membership function is:
X is a mapping value, and a, b, c, d are respectively trapezoidal four flex points from left to right;
Z shape membership function is:
X is a mapping value, and it is 1 and 0 pairing interval mapping value that a, b are respectively the degree of membership value.
S shape membership function is:
X is a mapping value, and it is 0 and 1 pairing interval mapping value that a, b are respectively the degree of membership value.
4, determining of fuzzy rule:
According to the description of physical descriptor and accident relation in the petroleum drilling production, determine fuzzy rule as shown in table 2, the form of this fuzzy rule is: if x
1Be A and x
2Be B so y be C.X wherein
1, x
2Represent two input quantities of basic blur unit respectively, y represents the output quantity of basic blur unit, and this tittle all is actual physics characteristic quantity mapping value on the domain interval, and A, B, C represent the fuzzy language value of input quantity and output quantity respectively.
Table 2
Each characteristic variable of each layer of hierarchical fuzzy model is also all carried out according to this rule when reasoning.Fuzzy rule is 16 altogether in the table 2, and each bar is all blured compose operation.If A and B are respectively the input fuzzy sets, C is the output fuzzy set, and then the ternary fuzzy relation R that determined of rule " if A and B then C " is
R＝(A×B)×C
If each discrete element in the domain (i=1 ~ 10) is μ corresponding to the degree of membership of fuzzy set A
A(i), the degree of membership corresponding to fuzzy set B is μ
B(j), (j=1 ~ 10) are μ corresponding to the degree of membership of fuzzy set C
C(m), (m=1 ~ 10), then the concrete operation process is:
A×B＝μ
A(i)＾μ
B(j)
R
k＝μ
A(i)＾μ
B(j)＾μ
C(m)
Little computing, i.e. μ are got in symbol ^ representative
A(i) ^ μ
B(j)=min (μ
A(i), μ
B(j)).
Each bar rule is all carried out such compose operation, then can obtain the fuzzy relation of 16 rules.Fuzzy relation R with these 16 rules
k, k=1,2 ..., 16, merge, then obtain the pairing fuzzy relation matrix of above-mentioned fuzzy rule base.
5, the fuzzy subset's that fuzzy rule activated unification:
Each input variable for basic blur unit, its mapping value can activate two fuzzy subsets at most, it is linguistic variable, for real-time and the simplicity calculated are considered, with two fuzzy subset's unifications that activate is a comprehensive fuzzy subset, has promptly reflected two fuzzy subsets' that activated effect among this correction fuzzy subset.Concrete grammar is:
If the mapping value of a certain input variable on the domain interval is x, this mapping value is at most corresponding to two fuzzy subsets (being linguistic variable) under above-mentioned definite subordinate function, if these two fuzzy subsets are respectively A and B, its discrete degree of membership on the domain interval is respectively [a
i] and [b
i], i=1,2 ... 10, the degree of membership that mapping value x is under the jurisdiction of these two fuzzy subsets is respectively μ
A(x) and μ
B(x), C=[c is then comprehensively arranged later
i]=a
i* μ
A(x)+b
i* μ
B(x), if [c
i] in element greater than 1 appears, then by normalization with c
iChange the numerical value between 0 ~ 1 into, so just C is carried out fuzzy reasoning as the obfuscation result of input variable.
6, fuzzy inference rule:
The fuzzy reasoning of each basic blur unit adopts common mamdani inference rule, establishes the degree of membership of two input variables after carrying out comprehensive obfuscation on the domain interval and is distributed as C
1And C
2, then The reasoning results is U=(C
1* C
2) oR, wherein R is the fuzzy relation matrix that the front is tried to achieve by (3).Symbol o represents compose operation.
7, ambiguity solution algorithm:
The The reasoning results of basic blur unit is the distribution that is subordinate to of a Discrete Distribution on the domain interval, promptly, can obtain this The reasoning results pairing mapping value on 1 ~ 10 domain interval by the weighted method ambiguity solution by the distribution value of the degree of membership of two unusual variablees through obtaining this result behind the fuzzy reasoning on interval 1 ~ 10.Gravity model appoach ambiguity solution algorithm is
I is the discrete value on the domain interval, μ
iDegree of membership for corresponding discrete point.
8, determining of system's output quantity:
For embodying the difference of different variablees, based on the weighted sum that finally is output as each layer output result of the petroleum well drilling engineering accidents early-warning system of layered fuzzy system, that is: to the accident impact degree
Wherein k is a hierarchy number, and α is a weighting coefficient,
y
kBe the result behind each layer output ambiguity solution.The definite of weighting coefficient obtains by fuzzy Hierarchy Analysis Method, so just can avoid artificially determining the subjectivity and the inconsistency of weight coefficient; Pass through simultaneously of the influence degree analysis of basic blur unit input variable output variable, can determine the locations of structures of each accident input variable in the hierarchical fuzzy model, thereby the size of different variablees to the accident impact degree obtained embodying on the position of hierarchical fuzzy model.
The self study of 9, early warning process and adjustment process
For making system have stronger dirigibility, extendability and adaptability, the mode that the present invention adjusts by the mapping to the input quantity domain realizes the self study and the adjustment of early warning process.
In the accident early warning layered fuzzy system, fuzzy subset's the subordinate function and the adjustment of fuzzy inference rule mainly rely on the master that is adjusted into of technician's off-line, online self-adjusting function is then finished by the adjustment of the fuzzy domain of input quantity, reduce the complexity of system so on the one hand, also can reduce online real-time operation amount on the other hand.For guaranteeing the versatility and the applicability of Early-warning Model, each layer uses unified fuzzy domain in the accident early warning layered fuzzy system, system is mapped on this domain interval by the measured value of different mapping relations with input parameter, self-adjusting function just is to utilize the difference of mapping relations, measured value is mapped on the different fuzzy quantities handles.Utilization is to the study of field condition and data, and early warning system can be set up the mapping relations of corresponding accident early warning model gradually in this way.
The self-adjusting function of petroleum well drilling engineering accidents early-warning system realizes by the early warning result is carried out comprehensive evaluation.Evaluation index to the early warning result has four kinds: analyze suitable, wrong report, low newspaper, high newspaper.Except that first kind of evaluation do not need systematic parameter adjusted, other evaluation result all can be revised relevant parameter by system.
In the self-adjusting process, the adjustment of accident is divided into two levels, at first be adjustment process to the accident early warning result, secondly be study adjustment to the off-note variable-definition.
A, to accident early warning result's adjustment
By data processing and fuzzy reasoning, can carry out the fuzzy early warning of accident for drilling process.According to the definition of fuzzy variable, incident classification be divided into " little, in, greatly, bigger " level Four, when The reasoning results is some grades, by with the contrast of actual conditions, can provide four kinds of evaluations: analyze suitable, wrong report, low newspaper, high newspaper.When being evaluated as " suitable ", need not revise reasoning process.When being evaluated as " high newspaper ", the degree height of the analysis result of algorithm than actual accidents is described then, need revise this moment to the mapping coefficient of correlated variables, and the reduction algorithm makes it to be consistent with actual conditions to the forecast grade of accident; When being evaluated as " low newspaper ", illustrate that then the analysis result of algorithm is lower than the degree of actual accidents, need revise this moment to the mapping coefficient of correlated variables, and the raising algorithm makes it to be consistent with actual conditions to the forecast grade of accident; When being evaluated as " wrong report ", the analysis that algorithm then is described does not conform to actual conditions, this moment is according to the size of off-note variable mapping value, successively the threshold value coefficient that constitutes the off-note variable is revised, and the correlated characteristic variable that causes the accident made one by one judge, these characteristic variables are made further evaluation again,, then the threshold value coefficient of this variable is revised if assert that some variable reality is not unusual.
To the accident early warning evaluation of result is " low newspaper ", then need the correlated characteristic variable that causes the accident is made judgement one by one, these characteristic variables are made whether needing judgement that its mapping value is heightened again, if the mapping value of assert some unusual variablees than actual expectation for low, then the mapping coefficient of this variable is revised, its mapping value is improved, finally, reach the order of heightening the accident early warning result by the adjustment of characteristic variable mapping coefficient.
To the accident early warning evaluation of result is " high newspaper ", then need the relevant ` characteristic variable that causes the accident is made judgement one by one, these characteristic variables are made whether needing judgement that its mapping value is turned down again, if assert the be height of the mapping value of some characteristic variables than actual expectation, then the mapping coefficient of this variable is revised, its mapping value is reduced, finally, reach and turn down accident early warning purpose as a result by the adjustment of characteristic variable mapping coefficient.Adjustment process to characteristic variable
B, to the adjustment of characteristic variable
Mainly be that whether unusual threshold value coefficient and the unusual big or small mapping coefficient of measurement variable adjusted to defining variable.
In fuzzy reasoning process be to the mapping coefficients adjustment method:
1, when being evaluated as " high newspaper ", at first current fuzzy language variable is reduced one-level, such as the forecast grade is " greatly ", be evaluated as " high newspaper ", then will forecast accordingly grade be reduced to " in ", can obtain the discrete type membership function of this grade fuzzy quantity by the membership function of definition, carry out the ambiguity solution computing by the maximum membership degree method and obtain its mapping value in interval 1 ~ 10, this mapping value is a desired value, by to the mapping coefficient k
1, k
2Adjustment, make the mapping value of current characteristic quantity become the target mapping value.
By new mapping coefficient, the forecast result of fuzzy reasoning will reduce a grade, still be high if estimate, and then repeat above-mentioned adjustment process again, and the net result that makes is consistent with actual.
2, when being evaluated as " low newspaper ", at first current fuzzy language variable is improved one-level when adjusting, such as the forecast grade be " in ", be evaluated as " low newspaper ", to forecast accordingly that then grade rises to " greatly ", the membership function by definition can obtain the discrete type membership function of this grade fuzzy quantity, carries out the ambiguity solution computing by the maximum membership degree method and obtains its mapping value in interval 1 ~ 10, this mapping value is a desired value, by to the mapping coefficient k
1, k
2Adjustment, make the mapping value of current characteristic quantity become the target mapping value.
3, when being evaluated as " wrong report ", then the definition of characterization variable need be adjusted, and mainly is the unusual threshold value coefficient of defining variable is revised.
By correction to the threshold value coefficient, make variable no longer be defined as unusually, the forecast result of fuzzy reasoning also will obtain revising like this, make net result be consistent with actual.
The invention has the advantages that and utilize layered fuzzy system, many input physical descriptors in the petroleum well drilling engineering accidents early-warning system are converted into the hierarchical system that is connected and composed by the layering of low-dimensional blur unit, make regular several of fuzzy system to count linear growth with input variable, stopped the generation of " dimension calamity " phenomenon, utilize signal processing technology to obtain to reflect the signal of physics input variable variation tendency, and carry out fuzzy reasoning by the layered fuzzy system of unified model, thereby realize accurate and effective, petroleum well drilling engineering accidents is carried out the purpose of early warning easily.
Appendix:
Table 2 petroleum well drilling engineering accidents experts database
Illustrate: the first behavioral engineering accident title in each form, as leakage, well kick etc.
The technological process of the second behavior petroleum drilling in each form, as creep into, from the shaft bottom, trip-out, down brill, reaming etc.
Other each behavior causes the ANOMALOUS VARIATIONS process of the collection physical quantity of this accident, wherein ~ the expression unusual fluctuations, ↑ the expression abnormal ascending, ↓ expression decline unusually.
Claims (4)
1, a kind of petroleum well drilling engineering accidents method for early warning based on layered fuzzy system, it is characterized in that: it is made of following step:
The first step, set up the model database of petroleum well drilling engineering accidents early warning:
Various sensor signals and derivation parameter data from the field integrated logging equipment collection of petroleum drilling, analyze by data message long-term accumulation, the data message that provides according to the employed various sensors of the petroleum drilling of long-term accumulation, in conjunction with petroleum drilling operating personnel and relevant expert's experimental knowledge, set up the petroleum well drilling engineering accidents model database; Whole process to petroleum drilling is determined different duties according to the logical combination relation and the drilling process standard of sensor situation of change, these duties have different characteristics each other, and kind that may have an accident and sensor variable sign also have nothing in common with each other;
Second goes on foot, sets up the input variable of petroleum well drilling engineering accidents early warning:
According to the information in the model database in the first step, utilize corresponding sensor acquisition data, the data of being gathered mainly contain: utilize pressure transducer to gather weight on hook, standpipe pressure, casing pressure, torque signal; Utilize temperature sensor to gather the temperature signal of gateway; Utilize conductivity sensor to gather the signal of gateway conductivity; Utilize flow sensor to gather the rate of discharge signal; Utilize density sensor to gather the gateway density signal; Utilize pulse transducer to gather that hook height, pump dash, the rotary speed signal; Utilize liquid level sensor acquired volume signal; Utilize hydrogen sulfide sensor to gather the sulfureted hydrogen gas concentration signal;
The 3rd step, analysis characteristic information of input signals:
According to state-contingent in the drilling process and transforming relationship, actual measured results in conjunction with various parameter sensors, according to the requirement of production technology, drilling process is divided into some typicalnesses, and its corresponding sensor parameter is incorporated into to pairing state model handle; To each drilling state, be decomposed into steady-state process and transient state process; The sensor variable of being gathered is determined fluctuating range, vibration frequency, rate of change and the variation tendency characteristic information of sensor variable in each drilling well course of work by adopting statistical dependence analysis, wavelet transformation, spectrum analysis, time-domain analysis method; Is corresponding different intermediate features variable with same sensor variable according to the different decomposition of its variation characteristic by these information; The normal variation scope of determining variable by the calculating and the analysis of threshold of dynamic average;
The 4th goes on foot, sets up the structural model of layered fuzzy system:
To the 3rd step determined n input variable x
1, x
2... x
n, being respectively the feature input variable in the accident pre-alarming system, the basic blur unit of each layer all has two input variables, wherein x
1, x
2As the input of the basic blur unit of ground floor, it exports y
1And the 3rd input variable x
3As the input of the basic blur unit of the second layer, all the other each layers by that analogy; Each layer all has only two input variables;
The 5th goes on foot, the input variable of determining in the 3rd step is carried out the hierarchical fuzzy reasoning of interval mapping and unified model:
To the 3rd step determined n different types of input variable, by statistical study to the actual physics variable, determine the variation range and the variation characteristic of this input variable, utilize and analyze the mapping relations that obtain, these input variables are mapped on the unified domain interval, on this basis the basic blur unit of each layer of layered fuzzy system are carried out the fuzzy reasoning of unified model;
The 6th goes on foot, determines the petroleum well drilling engineering accidents early-warning system output quantity:
The final output quantity of petroleum well drilling engineering accidents early-warning system is the basic blur unit output of each layer result's a weighted sum, that is:
Wherein k is a hierarchy number, and α is a weighting coefficient,
y
kBe the result behind each layer output ambiguity solution;
The self study and the adjustment of the 7th step, early warning process:
Realize the self study and the adjustment of early warning process by the mode that input quantity domain mapping is adjusted, promptly the evaluation index to the early warning result is set at four kinds: analyzes suitable, wrong report, low newspaper, the high newspaper; Except that first kind of evaluation do not need systematic parameter adjusted, other evaluation result was all revised relevant parameter by system.
2, according to the described petroleum well drilling engineering accidents method for early warning of claim 1 based on layered fuzzy system, it is characterized in that: in the 3rd step, the signal of input variable is carried out the calculating of AR model filtering and least square optimization, wavelet de-noising filtering and dynamic average:
Concrete processing procedure is: to the data decomposition in each drilling state is steady state data and Temporal Data, utilize known off-line historical data to obtain desired value to input signal by 5 smoothing algorithms, and utilize least square method to obtain the filter factor of AR model, utilize this wave filter that the input data are handled the signal that obtains reflecting the input variable variation tendency; The wavelet de-noising wave filter is by carrying out wavelet decomposition to the input variable signal, high frequency detail coefficients after decomposing is suppressed, high frequency coefficient and low frequency coefficient after will suppressing then are reconstructed, obtain filtered signal, this signal can truly reflect the variation tendency of actual signal, and dynamic average and signal threshold value analysis by signal after the calculation of filtered can obtain signal normal variation scope.
3, according to the described petroleum well drilling engineering accidents method for early warning of claim 1, it is characterized in that: in the 4th step, take vacant substitute mode to handle the hierarchical fuzzy model based on layered fuzzy system:
Concrete processing procedure is: when having defined the off-note variable that the characteristic variable number that causes a certain accident occurs in more than actual engineering in model database, unusual characteristic variable do not occur and do not enter the blur unit layer, by occurring unusual characteristic variable substitute in the actual engineering of the next one as input quantity.
4, according to the described petroleum well drilling engineering accidents method for early warning of claim 1 based on layered fuzzy system, it is characterized in that: in the 5th step, fuzzy reasoning to layered fuzzy system adopts the mode of unified model to handle: detailed process is for passing through the unified domain interval of definition, analyze the variation range of actual physics input quantity, the input quantity unification is mapped on the defined interval, and this interval is promptly as the basic domain of fuzzy system; Each basic blur unit adopts identical fuzzy reasoning mode, and promptly identical domain, fuzzy method and unified fuzzy inference rule, fuzzy reasoning adopt and calculate easy Mamdani composition algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2007100551786A CN100511041C (en) | 2007-09-18 | 2007-09-18 | Petroleum well drilling engineering accidents early-warning system based on layered fuzzy system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2007100551786A CN100511041C (en) | 2007-09-18 | 2007-09-18 | Petroleum well drilling engineering accidents early-warning system based on layered fuzzy system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101118420A CN101118420A (en) | 2008-02-06 |
| CN100511041C true CN100511041C (en) | 2009-07-08 |
Family
ID=39054565
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2007100551786A Expired - Fee Related CN100511041C (en) | 2007-09-18 | 2007-09-18 | Petroleum well drilling engineering accidents early-warning system based on layered fuzzy system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100511041C (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102095590B (en) * | 2010-11-29 | 2013-07-31 | 中国石油大学(北京) | Safe monitoring method of storage power unit under dynamic environment |
| CN102128022B (en) * | 2010-12-30 | 2013-06-12 | 中国电子科技集团公司第二十二研究所 | Drilling engineering early warning method and system thereof |
| CN103488080B (en) * | 2013-09-09 | 2015-11-11 | 河北科技师范学院 | Moon craft based on layered fuzzy system is coordinated to drive adaptive fusion method |
| CN105089620B (en) * | 2014-05-14 | 2018-03-09 | 中国石油天然气集团公司 | Monitoring system, the method and device of bit freezing |
| CN104537446B (en) * | 2015-02-09 | 2017-12-22 | 四川大学 | Two layers of band fuzzy stochastic time window vehicle routing optimization method |
| CN104715326A (en) * | 2015-03-04 | 2015-06-17 | 中国石油化工股份有限公司 | Accident prewarning method by means of petrochemical industry occupational safety performance data |
| CN105466600B (en) * | 2015-12-16 | 2017-11-28 | 东南大学 | A kind of distributed capstan winch hydraulic fluid temperature monitoring method based on fuzzy algorithmic approach |
| CN109359871B (en) * | 2018-10-19 | 2022-03-15 | 中国海洋石油集团有限公司 | Oil-gas well casing reliability evaluation method |
| CN109711632B (en) * | 2018-12-29 | 2023-04-18 | 辽宁工程技术大学 | Coal and gas outburst prediction method based on gas emission abnormal sensitive index |
| CN109989740B (en) * | 2019-04-10 | 2022-09-23 | 中煤科工集团西安研究院有限公司 | Coal measure stratum drilling intelligent identification system and method based on multi-source information fusion |
| CN113393334B (en) * | 2020-03-11 | 2024-05-14 | 中国石油化工股份有限公司 | Drilling parameter optimization recommendation method and system |
| CN111724290B (en) * | 2020-06-24 | 2023-09-26 | 山东建筑大学 | Environment-friendly equipment identification method and system based on depth layering fuzzy algorithm |
| CN117927289B (en) * | 2024-03-21 | 2024-05-31 | 中煤北京煤矿机械有限责任公司 | Self-adaptive regulation and control system based on detachable rotating beam of large-inclination-angle advanced support |
-
2007
- 2007-09-18 CN CNB2007100551786A patent/CN100511041C/en not_active Expired - Fee Related
Non-Patent Citations (2)
| Title |
|---|
| 钻井工程事故监测和预警方法研究. 吴俊杰.录井工程，第17卷第1期. 2006 |
| 钻井工程事故监测和预警方法研究. 吴俊杰.录井工程，第17卷第1期. 2006 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101118420A (en) | 2008-02-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100511041C (en) | Petroleum well drilling engineering accidents early-warning system based on layered fuzzy system | |
| Xue et al. | Fault detection and operation optimization in district heating substations based on data mining techniques | |
| CN102128022B (en) | Drilling engineering early warning method and system thereof | |
| Wei et al. | Predicting injection profiles using ANFIS | |
| ElNozahy et al. | A probabilistic load modelling approach using clustering algorithms | |
| Su et al. | Identification model for dam behavior based on wavelet network | |
| Li et al. | Prediction of the failure point settlement in rockfill dams based on spatial-temporal data and multiple-monitoring-point models | |
| CN112857658B (en) | Pressure detection system based on cloud platform | |
| CN106199174A (en) | Extruder energy consumption predicting abnormality method based on transfer learning | |
| CN113807570A (en) | Reservoir dam risk level evaluation method and system based on XGboost | |
| CN108204895A (en) | A kind of Wind turbines change gear box fault diagnosis method and system | |
| CN110826237A (en) | Bayesian belief network-based wind power equipment reliability analysis method and device | |
| CN111859814A (en) | Rock aging deformation prediction method and system based on LSTM deep learning | |
| Schmelzer | On solutions of stochastic differential equations with parameters modeled by random sets | |
| Tong et al. | The prediction of concrete dam displacement using Copula-PSO-ANFIS hybrid model | |
| Zhang et al. | A hierarchical early kick detection method using a cascaded GRU network | |
| CN107992112A (en) | A kind of control loop performance estimating method and system based on arrangement entropy | |
| Kajornrit et al. | Rainfall prediction in the northeast region of Thailand using modular fuzzy inference system | |
| CN117034169A (en) | Power grid main transformer equipment abnormal state prediction method based on time sequence causality network | |
| CN111367255A (en) | Performance evaluation test system and method for multi-variable control system | |
| CN112819378B (en) | River ecological flow regulation and control method based on pollution condition | |
| Zheng et al. | fault diagnosis model and application of water injection well based on SPC rules and real-time data fusion | |
| Tsaur | Forecasting analysis by fuzzy grey model GM (1, 1) | |
| Chakrabarti et al. | Diagnosis using a first-order stochastic language that learns | |
| CN102393819A (en) | MATLAB (matrix laboratory)-based fuzzy controller HDL (hardware description language) code automatic generation method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090708 Termination date: 20100918 |