CN107642347A - Accident forecast method for early warning and device under shale gas fractured well - Google Patents
Accident forecast method for early warning and device under shale gas fractured well Download PDFInfo
- Publication number
- CN107642347A CN107642347A CN201710812110.1A CN201710812110A CN107642347A CN 107642347 A CN107642347 A CN 107642347A CN 201710812110 A CN201710812110 A CN 201710812110A CN 107642347 A CN107642347 A CN 107642347A
- Authority
- CN
- China
- Prior art keywords
- time
- accident
- operating mode
- shale gas
- trend feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the present application provides accident forecast method for early warning and device under a kind of shale gas fractured well, and this method includes：The online monitoring data for specifying monitoring parameters is obtained in real time；According to the online monitoring data and corresponding parametric prediction model, obtain the specified monitoring parameters and specifying the predicted value of prediction step；Determine the trend feature vector of the predicted value, and according to the trend feature is vectorial and corresponding producing condition classification device, the operating mode type corresponding to the predicted value is predicted, and when the operating mode type predicted is shale gas pressure break down-hole accident, sends corresponding warning information.The embodiment of the present application can real-time estimate shale gas pressure break down-hole accident.
Description
Technical field
The application is related to down-hole accident real-time estimate technical field in rock gas pressing crack construction process, more particularly, to a kind of page
Rock air pressure splits down-hole accident prediction and warning method and device.
Background technology
Shale gas pressure break is the agent technology of shale gas exploitation.However, in fracturing process, accident once occurs for underground may
The high pressure that can cause to be formed in high risks, especially oil pipe can damage ground installation, such as fracturing pump, wellhead assembly in turn
Deng, or even stratum filtration can be destroyed, cause pressing crack construction to fail.If Accurate Prediction can be carried out to shale gas down-hole accident,
It can be that Field Force takes control measure to set apart, so as to be advantageous to slow down damage sequence, reduce economic loss.Therefore, page
It is significant that rock air pressure splits down-hole accident prediction.
It is pattern recognition problem on shale gas fracturing process down-hole accident the essence of prediction.It is live in shale gas pressing crack construction,
The artificial operating mode according to the trend future time instance underground of monitoring parameters in data collecting system.Due to individualized medicine,
The reason such as experience and sense of responsibility difference, causes to predict in time and often occurs with the situation of process accident.There is scholar qualitative
Analyze hydraulic fracture operating curve morphological feature, can as site operation monitor foundation.However, shale gas fracturing process well
The real-time estimate of lower accident (including stratum forms crack, pressure alters accident and sand plug accident) still could not be resolved.
In summary, need badly at present can real-time estimate early warning shale gas pressure break down-hole accident technical scheme.
The content of the invention
The purpose of the embodiment of the present application is to provide accident forecast method for early warning and device under a kind of shale gas fractured well, with
Realization can real-time estimate early warning shale gas pressure break down-hole accident.
To reach above-mentioned purpose, on the one hand, it is pre- that the embodiment of the present application provides accident forecast under a kind of shale gas fractured well
Alarm method, including：
The online monitoring data for specifying monitoring parameters is obtained in real time；
According to the online monitoring data and corresponding parametric prediction model, obtain the specified monitoring parameters and specifying in advance
Survey the predicted value of step-length；
Determine the trend feature vector of the predicted value, and according to the trend feature is vectorial and corresponding producing condition classification
Device, the operating mode type corresponding to the predicted value is predicted, and when the operating mode type predicted is shale gas pressure break down-hole accident,
Send corresponding warning information.
Preferably, the parametric prediction model of the specified monitoring parameters is established beforehand through in the following manner：
From shale gas pressure break down-hole accident off-line data, it is determined that per the N*H groups that monitoring parameters are specified corresponding to class operating mode
Time series, wherein, N is operating mode categorical data, and H is the time series group number per class operating mode type；Every group of time series includes
The sampled data of each sampling instant of the whole evolution of the corresponding accident of reflection；
For every group of time series, L is chosen from this group of time series by step-by-step movement forward slip time windown,h-k
Group sample；Then when the sliding time window often slides into a position, each sampled point that sliding time window is covered
It is corresponding for one group of input sample, in this group of time series, with being apart adopting for k in front of the sliding time window of the opening position
Sampling point exports sample as corresponding to this group of input sample；Wherein, Ln,hFor the length of this group of time series, k is prediction step；
Input data set is built according to the input sample selected from all groups of time serieses, and according to from all groups of times
The output sample structure output data set determined in sequence；
Using the input data set as input, and exported using the output data set as target, it is default to train
Machine learning model, obtain the parametric prediction model of the specified monitoring parameters.
Preferably, the machine learning model includes support vector regression.
Preferably, after input data set and output data set is constructed, in addition to：
The input data set and output data set constructed is normalized.
Preferably, the punishment parameter of the support vector regression and nuclear parameter optimize to obtain by particle cluster algorithm.
Preferably, the producing condition classification device is established beforehand through in the following manner：
From shale gas pressure break down-hole accident off-line data, it is determined that specified corresponding to per class operating mode monitoring parameters it is multigroup when
Between sequence；Every group of time series includes the sampled data of each sampling instant of the whole evolution of the corresponding accident of reflection；
The time window of same time length is set to each group time series, and time window is divided into multiple equal lengths
Time slice；The width of the time slice is identical with the width of the prediction step；
Linear fit is carried out to the time series data in each time slice, is changed over time corresponding to acquisition linear
Relation curve；
The trend feature primitive according to corresponding to determining the linear relationship slope of a curve of each time slice；The trend is special
Levying the value of primitive includes steady, raising and lowering；
Trend feature vector, and the work according to corresponding to the trend feature primitive are built according to the trend feature primitive
Condition type builds operating mode type set；
Using trend feature vector as input data set, and using the operating mode set of types cooperation as output data set,
Default neural network model is trained, obtains producing condition classification device.
Preferably, it is described according to the online monitoring data and corresponding parametric prediction model, obtain the specified monitoring
Parameter is specifying the predicted value of prediction step, including：
The parametric prediction model is substituted into using the online monitoring data as input, the specified monitoring parameters is obtained and exists
Specify the predicted value of prediction step.
Preferably, the trend feature vector for determining the predicted value, including：
The time series data that multiple predicted values that specified time window ranges interior prediction goes out are formed is divided into multiple
The time slice of equal length；The width of the time slice is identical with the width of the prediction step；
Linear fit is carried out to the time series data in each time slice, is changed over time corresponding to acquisition linear
Relation curve；
The trend feature primitive according to corresponding to determining the linear relationship slope of a curve of each time slice；The trend is special
Levying the value of primitive includes steady, raising and lowering；
The trend feature vector of the time series data is built according to the trend feature primitive.
Preferably, it is described according to the trend feature is vectorial and corresponding producing condition classification device, predict that the predicted value institute is right
The operating mode type answered, including：
The producing condition classification device is substituted into using trend feature vector as input, obtains the work corresponding to the predicted value
Condition type.
On the other hand, the embodiment of the present application additionally provides accident forecast prior-warning device under a kind of shale gas fractured well, including：
Online data acquisition module, the online monitoring data of monitoring parameters is specified for obtaining in real time；
Monitoring parameters prediction module, for according to the online monitoring data and corresponding parametric prediction model, obtaining institute
State specified monitoring parameters and specify the predicted value of prediction step；
Operating mode prediction and warning module, for determining the trend feature vector of the predicted value, and according to the trend feature
Vectorial and corresponding producing condition classification device, the operating mode type corresponding to the predicted value is predicted, and be in the operating mode type predicted
During shale gas pressure break down-hole accident, corresponding warning information is sent.
The technical scheme provided from above the embodiment of the present application, can be according to obtaining in the embodiment of the present application in real time
The online monitoring data of monitoring parameters and corresponding parametric prediction model are specified, obtains and specifies monitoring parameters in specified prediction step
Predicted value；Then it is determined that on the basis of the trend feature vector of predicted value, according to trend feature is vectorial and corresponding operating mode point
Class device, the operating mode type corresponding to predicted value is predicted, and when the operating mode type predicted is shale gas pressure break down-hole accident, hair
Go out corresponding warning information, it is achieved thereby that the real-time estimate early warning of shale gas pressure break down-hole accident, is subsequent treatment shale air pressure
Split down-hole accident and reserved valuable time, so as to be advantageous to avoid the generation of shale gas pressure break down-hole accident or reduce shale
Air pressure splits the harm caused by down-hole accident.
Brief description of the drawings
, below will be to embodiment or existing in order to illustrate more clearly of the embodiment of the present application or technical scheme of the prior art
There is the required accompanying drawing used in technology description to be briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments described in application, for those of ordinary skill in the art, do not paying the premise of creative labor
Under, other accompanying drawings can also be obtained according to these accompanying drawings.In the accompanying drawings：
Fig. 1 is the flow chart of accident forecast method for early warning under shale gas fractured well in the embodiment of the application one；
Fig. 2 is the flow chart of accident forecast method for early warning under shale gas fractured well in another embodiment of the application；
Fig. 3 is the relation signal between prediction step k, time window width n and current time t in the embodiment of the application one
Figure；
Fig. 4 is the value schematic diagram of trend feature primitive in the embodiment of the application one；
Fig. 5 is that fragment division and trend feature primitive determine schematic diagram in the embodiment of the application one；
Fig. 6 is the parameter prediction and trend character extraction schematic diagram in the embodiment of the application one；
Fig. 7 is the implementation prediction and warning result that stratum forms crack accident in the embodiment of the application one；
Fig. 8 is the implementation prediction and warning result that near wellbore zone pressure alters accident in the embodiment of the application one；
Fig. 9 is the structured flowchart of accident forecast prior-warning device under shale gas fractured well in the embodiment of the application one.
Embodiment
In order that those skilled in the art more fully understand the technical scheme in the application, it is real below in conjunction with the application
The accompanying drawing in example is applied, the technical scheme in the embodiment of the present application is clearly and completely described, it is clear that described implementation
Example only some embodiments of the present application, rather than whole embodiments.It is common based on the embodiment in the application, this area
The every other embodiment that technical staff is obtained under the premise of creative work is not made, it should all belong to the application protection
Scope.
With reference to shown in figure 1, accident forecast method for early warning can include following under the shale gas fractured well of the embodiment of the present application
Step：
S101, the online monitoring data for specifying monitoring parameters is obtained in real time.
In the embodiment of the present application, the specified monitoring parameters are the monitoring parameters of reaction shale gas pressure break underground working.
In the exemplary embodiment of the application one, the specified monitoring parameters are such as can be with well head pressure, discharge capacity and casing pressure.
S102, according to the online monitoring data and corresponding parametric prediction model, obtain the specified monitoring parameters and exist
Specify the predicted value of prediction step.
, can be using online monitoring data as parameter prediction mould corresponding to input substitution in some embodiments of the present application
Type, the predicted value of prediction step is being specified to obtain the specified monitoring parameters.Such as cut-off current time t, when past
Between in the range of window width n, obtain online monitoring data (i.e. time series data xt,xt-1,…,xt-(n-1)), then it is corresponding
The predicted value of monitoring parameters is specified described in the t+k momentIt can be expressed as：
Wherein, the relation between prediction step k, time window width n and current time t can be as shown in Figure 3.Thus may be used
See, in the case of known multiple sampled datas, sampling corresponding to one can be predicted by corresponding parametric prediction model
Value.
In some embodiments of the present application, the parametric prediction model can pre-establish, for one finger of people
Determine monitoring parameters, its corresponding parametric prediction model can be established beforehand through in the following manner：
First, from shale gas pressure break down-hole accident off-line data, it is determined that per specified monitoring parameters corresponding to class operating mode
N*H group time serieses, wherein, N is operating mode categorical data, and H is the time series group number per class operating mode type；Every group of time series
The sampled data of each sampling instant including the whole evolution of the corresponding accident of reflection, to ensure that machine learning model can be more
Data characteristics is captured exactly.In the exemplary embodiment of the application one, the machine learning model for example can be to support
Vector regression (Support Vector Regression Machine, abbreviation SVR) etc..
Then, for every group of time series, chosen by step-by-step movement forward slip time window from this group of time series
Ln,h- k organizes sample；Then when the sliding time window often slides into a position, what sliding time window was covered each adopts
Sampling point is one group of input sample, corresponding, in this group of time series, with being apart k in front of the sliding time window of the opening position
Sampled point export sample as corresponding to this group of input sample；Wherein, Ln,hFor the length of this group of time series, k walks for prediction
It is long.Such as the total input sample quantity of finger monitoring survey parameter is：
Use Tm, 1≤m≤M represents m-th of sample, then the input data set comprising M group samples can be shown with matrix Input table,
As shown in formula (3), the row vector of matrix represents the time series in time window；Corresponding output data set can use Output
Represent, as shown in formula (4).
Output={ xi| i=n+1, n+2 ..., M+n } (4)
Secondly, input data set is built according to the input sample selected from all groups of time serieses, and according to from all
The output sample structure output data set determined in group time series.
Finally, using the input data set as input, and exported using the output data set as target, it is pre- to train
If machine learning model, obtain the parametric prediction models of the specified monitoring parameters.
Thus, the parametric prediction model of each specified monitoring parameters can be pre-established using above method, for example with
Above method can establish well head pressure, discharge capacity and each self-corresponding parametric prediction model of casing pressure respectively.
In some embodiments of the present application, pressing crack construction regular job, environmental disturbances etc. easily cause sensor sample
Singular value in data be present, can be first by data set (input in order to reduce the influence that unusual sample value is brought to parametric prediction model
Data set and output data set) make normalized, then the input data set after normalized and output data set are distinguished
As SVR input item and target output item, SVR models are trained.Further to improve the precision of prediction of model, particle can be used
Group's algorithm (Particle Swarm Optimization, abbreviation PSO) enters to the punishment parameter C in SVR models and nuclear parameter g
Row optimizing.
Certainly, in some embodiments of the present application, for certificate parameter forecast model, other shale gas can also be used
Whether pressure break down-hole accident off-line data builds test sample, meet to require with the performance of inspection parameter forecast model.
S103, the trend feature vector for determining the predicted value, and according to the trend feature is vectorial and corresponding operating mode
Grader, predict the operating mode type corresponding to the predicted value, and the thing in the case where the operating mode type predicted is shale gas fractured well
Therefore when, send corresponding warning information.
In some embodiments of the present application, the producing condition classification device can be established beforehand through in the following manner：
1), from shale gas pressure break down-hole accident off-line data, it is determined that per the more of specified monitoring parameters corresponding to class operating mode
Group time series；Every group of time series includes the sampled data of each sampling instant of the whole evolution of the corresponding accident of reflection,
To ensure that machine learning model can more accurately capture data characteristics.
2) time window (Time of same time length (Time length, abbreviation TL), is set to each group time series
Window, abbreviation TW), and time window is divided into the time slice of multiple equal lengths；The width of the time slice can be with
The width of the prediction step is identical, in order to realize, such as shown in Fig. 6.
3) linear fit, is carried out to the time series data in each time slice, is changed over time corresponding to acquisition
Linear relationship curve.In the exemplary embodiment of the application one, such as can be by linear least square to each time slice
Interior time series data carries out linear fit, so as to obtain shown in corresponding linear relationship curve such as following formula (5)：
xΛ(t)=p (t-t0)+x0 (5)
Wherein t0Between representing at the beginning of each fragment, p represents slope, x0Represent monitoring parameters in t0The sampled value at moment.
4), the trend feature primitive according to corresponding to determining the linear relationship slope of a curve of each time slice.
In some embodiments of the present application, the trend feature primitive according to corresponding to slope can recognize that each time slice,
Such as shown in Fig. 5.Shown in corresponding relation equation below (6) between slope and trend feature primitive：
As can be seen that the value of the trend feature primitive can include steady, three kinds of raising and lowering from formula (6)
Trend, these three trend are then more intuitively shown in Fig. 4.
5) trend feature vector, is built according to the trend feature primitive, and according to corresponding to the trend feature primitive
Operating mode type structure operating mode type set.
In some embodiments of the present application, such as N*H group time series datas, respectively in every group of time series number
According to the nominal situation stage randomly choose KnormalIndividual sampled point, K is randomly choosed in accident stage of developmentincidentIndividual sampled point, then
R=H* (K can be obtained per class accidentnormal+Kincident) group sampled point, the wherein sampling number of nominal situation is Rnormal=H*
Knormal, the sampling number of Accident phase is Rincident=H*Kincident.Then r-th of trend feature vector of the n-th class accident can
Using with equation below (7 tables as:
Wherein, a1ZThe 1st monitoring parameters are represented in trend feature primitive corresponding to the Z time slice, remaining class successively
Push away.
In some embodiments of the present application, all trend feature vectors can be built into the matrix D as shown in formula (8), D
Expression is made up of N*R sample, and each sample is expressed as row vector, row vector by each instance sample point trend feature
Vector sum type labelComposition,Represent operating mode type, wherein label corresponding to r-th of instance sample point of the n-th class accident0 represents underground nominal situation (i.e. non-accident condition), the different accident pattern of remaining tag representation.Then in D
Left column for trend feature vector set, input data set can be used as；And the right row in D are corresponding operating mode set of types
Close, target output data set can be used as.
6), using trend feature vector as input data set, and using the operating mode set of types cooperation as output data
Collection, default neural network model is trained, obtain producing condition classification device.
In the exemplary embodiment of the application one, the neural network model for example can be BP (back
Propagation) neutral net.BP neural network is by up of three layers, input layer, hidden layer and output layer.Wherein input layer section
Points are equal to the quantity of total trend feature vector, i.e. 3*Z；Output layer nodes are that sample type is total, i.e. 1+N；Hidden layer
Nodes are input layer and output layer nodes summation, i.e. 3*Z+1+N.Sigmoid may be selected in the activation primitive of BP neural network
Function, the function are nonlinear function.After determining network structure model and activation primitive, by the trend feature in data set to
Amount and type label trains BP neural network model, finally establishes work respectively as the input item and output item of BP neural network
Condition grader.
In the application some embodiments, the trend feature vector for determining the predicted value can include：
First, the time series data that multiple predicted values that specified time window ranges interior prediction goes out are formed is divided into
The time slice of multiple equal lengths；The width of the time slice is identical with the width of the prediction step；
Secondly, linear fit is carried out to the time series data in each time slice, is changed over time corresponding to acquisition
Linear relationship curve；
Then, the trend feature primitive according to corresponding to determining the linear relationship slope of a curve of each time slice；It is described
The value of trend feature primitive includes steady, raising and lowering；
Finally, the trend feature vector of the time series data is built according to the trend feature primitive.
The realization of producing condition classification device is established because above-mentioned implementation process is similar to, its detail refers to producing condition classification device
Part is established, will not be repeated here.
It is described according to trend feature is vectorial and corresponding producing condition classification device in the application some embodiments, described in prediction
Operating mode type corresponding to predicted value can be：The producing condition classification device is substituted into using trend feature vector as input, is obtained
Obtain the operating mode type corresponding to the predicted value.
In the application some embodiments, after parametric prediction model and producing condition classification device is established, shale gas fractured well
Lower accident forecast method for early warning can be as shown in Figure 2.
In the application some embodiments, when an accident can be by the development trend concentrated expression of multiple monitoring parameters,
It is then described to refer to for shale gas pressure break down-hole accident in the operating mode type predicted：The each specified monitoring parameters predicted
Corresponding operating mode type belongs to the accident, or operating mode type corresponding to most specified monitoring parameters belongs to the accident.
In the application some embodiments, if having been predicted that down-hole accident, warning information is issued in time.Issue early warning letter
The rule of breath can include as follows：
Rule 1：Accident is predicted every time, starts an early warning (one-level early warning or two level early warning).
Rule 2：If monitoring down-hole accident first, start one-level early warning；
Rule 3：(a) in the case of continuous early warning, if the one-level early warning duration exceedes (such as 0.5 point of setting time
Clock), then warning level is promoted to two level early warning；(b) in the case of interval early warning, if the setting time before the current early warning moment
In (such as 1 minute), (such as one-level is pre- in the setting time of 1 minute for the half of the setting time that surpasses total time of one-level early warning
Alert total time was more than 0.5 minute), then warning level is promoted to two level early warning, otherwise, remains in that one-level early warning.
Rule 4：If current warning level be in two level alert status, and the setting time (such as 1 before the currently early warning moment
Minute) in, it is no more than total time of two level early warning half (such as the two level early warning in the setting time of 1 minute of the setting time
Be no more than 0.5 minute total time), then two level early warning is lowered one's standard or status as one-level early warning, otherwise maintains two level early warning.
Although procedures described above flow includes the multiple operations occurred with particular order, it should however be appreciated that understand,
These processes can include more or less operations, and these operations sequentially can be performed or performed parallel (such as using parallel
Processor or multi-thread environment).
For ease of understanding the application, illustrated with reference to an exemplary embodiment：
The present exemplary embodiment chooses the hydraulic fracture operating curve database of certain block as data source, and have selected such as following table
5 kinds of operating modes (a kind of non-accident and 4 kinds of accidents) are used as research object in shown in 1, to verify the applicability of this method and accurate
Property.The sampling period of known monitoring parameters is 5s, therefore, can obtain the sampling number (being shown in Table 1) of every group of time series data.
The accident pattern of table 1 and sample data information
Pre-establish monitoring parameters forecast model
A：Establish data set (being used for training parameter forecast model)
Shale gas pressure break down-hole accident can cause the abnormal change of well head pressure in ground monitoring system, discharge capacity and casing pressure
Change, therefore, select research object of above-mentioned 3 monitoring parameters as the present exemplary embodiment.
By taking well head pressure parameter as an example, if time window width n=12, prediction step k=12, it is known that well head pressure is adopted
The sample cycle is 5s, therefore, time span t corresponding to prediction stepK=12=60s.
As known from Table 1, the time series data of well head pressure is 20 groups.According to formulaIt is available
Include the input data set of M=4440 group time series datas；Accordingly, according to prediction step, establish comprising M=4440 sampling
The output data set of point.
B：Establish forecast model
After PSO is to SVR parameter optimizations, the punishment parameter for obtaining well head pressure forecast model can be CW=22.64
Can be g with nuclear parameterW=75.25.In order to verify the effect of well head pressure forecast model, from the time of one section of well head pressure
Sequence data establishes corresponding test data set as test data.Wherein：The form of test data set and training dataset
Identical, test data set also includes input data set and output data set；Unlike, when verifying modelling effect, for defeated
Although going out data set to be brought into forecast model in form, actually and computing is not involved in.
Both similarly, the forecast model according to corresponding to step A and step B can establish discharge capacity and casing pressure respectively, wherein
Sampling period be 5s, for convenience extract trend feature, both time window width and prediction step and well head pressure
Model it is identical.Wherein, the punishment parameter of discharge capacity forecast model can be CD=7.34, nuclear parameter can be gD=25.52；Set
The punishment parameter of pipe pressure forecast model can be with CC=21.11, nuclear parameter can be gC=5.08.
Pre-establish underground working grader
Take TL=1.0min, Z=3, Knormal=10, Kincident=40, then per the sampling instant point sum R=of class operating mode
250, the wherein sample number in nominal situation stage is Rnormal=50, the sample number of accident stage of development is Rincident=200.Root
According to formula (8), it can obtain including the data set D of 1000 samples, partial data collection is as shown in table 2.
The partial data collection of table 2
Because monitoring parameters have 3, and Z=3, then the input layer number of BP neural network model is 9；And BP nerve nets
The output layer nodes of network model are 5, i.e., the label of 5 kinds operating mode types；Node in hidden layer is 14.Bring data set D into BP
Neural network model, underground working grader can be established.
Down-hole accident monitoring and early warning in real time
Parametric prediction model and producing condition classification device based on above-mentioned foundation, with " stratum formation crack " and " press near wellbore zone
Alter accident " exemplified by, checking utilizes the accuracy and real-time of the method for the embodiment of the present application.
Case 1：Stratum forms crack
As shown in fig. 7, when forming crack for stratum monitoring parameters change procedure, and give and utilize the embodiment of the present application
Method in the prediction result and accident early warning result of the monitoring parameters for locating to obtain at different moments, while table 3 below gives specifically
Warning information.As can be seen that operating mode evolution can be divided into nominal situation stage, accident omen rank from Fig. 7 and table 3
Section and accident outburst stage.During 2.5 minutes within the nominal situation stage, it is predicted that stratum forms crack, and it is pre- to provide one-level
It is alert；At 3.2 minutes, one-level early warning upgraded to two level early warning.Two level early warning is maintained all the time in premonition stage；From the angle of early warning
Analysis, as long as stratum is predicted in nominal situation stage and premonition stage forms crack, belong to early warning category.Utilize this Shen
Please the pre-warning time that provides of method of embodiment be advanced by 1 minute than true traffic injury time.
The warning information of table 3 (stratum formation crack)
Case 2：Near wellbore zone pressure is altered
To alter accident pre- what is put at different moments for the near wellbore zone pressure that Fig. 8 illustrates using the method acquisition of the embodiment of the present application
Alert result.Specific warning information is listed in the table below in 4.From Fig. 8 and table 4 as can be seen that using the method for the embodiment of the present application just
Normal operating mode stage forecast starts one-level early warning to underground " near wellbore zone pressure alters accident "；In accident premonition stage, one-level early warning
Upgrade to two level early warning；At 12.7 minutes, two level was lowered one's standard or status as one-level early warning, at 13.6 minutes, one-level early warning disappearance.
Using the pre-warning time that the method for the embodiment of the present application provides 1.2 minutes are advanced by than true traffic injury time.
The warning information of table 4 (near wellbore zone pressure is altered)
With reference to shown in figure 9, accident forecast prior-warning device can include under the shale gas fractured well of the embodiment of the present application：
Online data acquisition module 91, it can be used for obtaining the online monitoring data for specifying monitoring parameters in real time；
Monitoring parameters prediction module 92, it can be used for according to the online monitoring data and corresponding parametric prediction model,
Obtain the specified monitoring parameters and specify the predicted value of prediction step；
Operating mode prediction and warning module 93, be determined for the predicted value trend feature vector, and according to it is described become
Gesture characteristic vector and corresponding producing condition classification device, predict the operating mode type corresponding to the predicted value, and in the operating mode predicted
When type is shale gas pressure break down-hole accident, corresponding warning information is sent.
For convenience of description, it is divided into various units during description apparatus above with function to describe respectively.Certainly, this is being implemented
The function of each unit can be realized in same or multiple softwares and/or hardware during application.
Each embodiment in this specification is described by the way of progressive, identical similar portion between each embodiment
Divide mutually referring to what each embodiment stressed is the difference with other embodiment.It is real especially for device
For applying example, because it is substantially similar to embodiment of the method, so description is fairly simple, related part is referring to embodiment of the method
Part explanation.
The present invention is described with reference to the flow charts of method and apparatus according to embodiments of the present invention, block diagram.Ying Li
Solution can by each flow in computer program instructions implementation process figure and/or block diagram and/or square frame and flow chart and/
Or the flow in block diagram and/or the combination of square frame.These computer program instructions can be provided to all-purpose computer, dedicated computing
The processor of machine, Embedded Processor or other programmable data processing devices is to produce a machine so that passes through computer
Or the instruction of the computing device of other programmable data processing devices is produced for realizing in one flow or multiple of flow chart
The device for the function of being specified in one square frame of flow and/or block diagram or multiple square frames.
These computer program instructions, which may be alternatively stored in, can guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works so that the instruction being stored in the computer-readable memory, which produces, to be included referring to
Make the manufacture of device, the command device realize in one flow of flow chart or multiple flows and/or one square frame of block diagram or
The function of being specified in multiple square frames.
These computer program instructions can be also loaded into computer or other programmable data processing devices so that counted
Series of operation steps is performed on calculation machine or other programmable devices to produce computer implemented processing, so as in computer or
The instruction performed on other programmable devices is provided for realizing in one flow of flow chart or multiple flows and/or block diagram one
The step of function of being specified in individual square frame or multiple square frames.
In a typical configuration, computing device includes one or more processors (CPU), input/output interface, net
Network interface and internal memory.
Internal memory may include computer-readable medium in volatile memory, random access memory (RAM) and/or
The forms such as Nonvolatile memory, such as read-only storage (ROM) or flash memory (flash RAM).Internal memory is computer-readable medium
Example.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method
Or technology come realize information store.Information can be computer-readable instruction, data structure, the module of program or other data.
The example of the storage medium of computer includes, but are not limited to phase transition internal memory (PRAM), static RAM (SRAM), moved
State random access memory (DRAM), other kinds of random access memory (RAM), read-only storage (ROM), electric erasable
Programmable read only memory (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc read-only storage (CD-ROM),
Digital versatile disc (DVD) or other optical storages, magnetic cassette tape, the storage of tape magnetic rigid disk or other magnetic storage apparatus
Or any other non-transmission medium, the information that can be accessed by a computing device available for storage.Define, calculate according to herein
Machine computer-readable recording medium does not include temporary computer readable media (transitory media), such as data-signal and carrier wave of modulation.
It should also be noted that, term " comprising ", "comprising" or its any other variant are intended to nonexcludability
Comprising so that process, method, commodity or equipment including a series of elements not only include those key elements, but also wrapping
Include the other element being not expressly set out, or also include for this process, method, commodity or equipment intrinsic want
Element.In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that wanted including described
Other identical element also be present in the process of element, method, commodity or equipment.
It will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer program product.
Therefore, the application can be using the embodiment in terms of complete hardware embodiment, complete software embodiment or combination software and hardware
Form.Deposited moreover, the application can use to can use in one or more computers for wherein including computer usable program code
The shape for the computer program product that storage media is implemented on (including but is not limited to magnetic disk storage, CD-ROM, optical memory etc.)
Formula.
The application can be described in the general context of computer executable instructions, such as program
Module.Usually, program module includes performing particular task or realizes routine, program, object, the group of particular abstract data type
Part, data structure etc..The application can also be put into practice in a distributed computing environment, in these DCEs, by
Task is performed and connected remote processing devices by communication network.In a distributed computing environment, program module can be with
In the local and remote computer-readable storage medium including storage device.
Embodiments herein is the foregoing is only, is not limited to the application.For those skilled in the art
For, the application can have various modifications and variations.All any modifications made within spirit herein and principle, it is equal
Replace, improve etc., it should be included within the scope of claims hereof.
Claims (10)
- A kind of 1. accident forecast method for early warning under shale gas fractured well, it is characterised in that including：The online monitoring data for specifying monitoring parameters is obtained in real time；According to the online monitoring data and corresponding parametric prediction model, obtain the specified monitoring parameters and specifying prediction step Long predicted value；Determine the trend feature vector of the predicted value, and according to the trend feature is vectorial and corresponding producing condition classification device, in advance The operating mode type corresponding to the predicted value is surveyed, and when the operating mode type predicted is shale gas pressure break down-hole accident, is sent Corresponding warning information.
- 2. accident forecast method for early warning under shale gas fractured well as claimed in claim 1, it is characterised in that the specified monitoring The parametric prediction model of parameter is established beforehand through in the following manner：From shale gas pressure break down-hole accident off-line data, it is determined that per the N*H group times that monitoring parameters are specified corresponding to class operating mode Sequence, wherein, N is operating mode categorical data, and H is the time series group number per class operating mode type；Every group of time series includes reflection The sampled data of each sampling instant of the whole evolution of corresponding accident；For every group of time series, L is chosen from this group of time series by step-by-step movement forward slip time windown,h- k organizes sample This；Then when the sliding time window often slides into a position, each sampled point that sliding time window is covered is one Group input sample, it is corresponding, in this group of time series, the sampled point with front of the sliding time window of the opening position being apart k Sample is exported as corresponding to this group of input sample；Wherein, Ln,hFor the length of this group of time series, k is prediction step；Input data set is built according to the input sample selected from all groups of time serieses, and according to from all groups of time serieses The output sample structure output data set of middle determination；Using the input data set as input, and exported using the output data set as target, to train default machine Learning model, obtain the parametric prediction model of the specified monitoring parameters.
- 3. accident forecast method for early warning under shale gas fractured well as claimed in claim 2, it is characterised in that the machine learning Model includes support vector regression.
- 4. accident forecast method for early warning under shale gas fractured well as claimed in claim 2 or claim 3, it is characterised in that constructing After input data set and output data set, in addition to：The input data set and output data set constructed is normalized.
- 5. accident forecast method for early warning under shale gas fractured well as claimed in claim 3, it is characterised in that the supporting vector The punishment parameter and nuclear parameter of regression machine optimize to obtain by particle cluster algorithm.
- 6. accident forecast method for early warning under shale gas fractured well as claimed in claim 1, it is characterised in that the producing condition classification Device is established beforehand through in the following manner：From shale gas pressure break down-hole accident off-line data, it is determined that per multigroup time sequence that monitoring parameters are specified corresponding to class operating mode Row；Every group of time series includes the sampled data of each sampling instant of the whole evolution of the corresponding accident of reflection；To each group time series set same time length time window, and by time window be divided into multiple equal lengths when Between fragment；The width of the time slice is identical with the width of the prediction step；Linear fit, the linear relationship changed over time corresponding to acquisition are carried out to the time series data in each time slice Curve；The trend feature primitive according to corresponding to determining the linear relationship slope of a curve of each time slice；The trend feature base The value of member includes steady, raising and lowering；Trend feature vector, and the operating mode class according to corresponding to the trend feature primitive are built according to the trend feature primitive Type builds operating mode type set；Using trend feature vector as input data set, and using the operating mode set of types cooperation as output data set, training Default neural network model, obtain producing condition classification device.
- 7. accident forecast method for early warning under shale gas fractured well as claimed in claim 1, it is characterised in that described in the basis Online monitoring data and corresponding parametric prediction model, obtain the specified monitoring parameters and specifying the predicted value of prediction step, Including：The parametric prediction model is substituted into using the online monitoring data as input, the specified monitoring parameters is obtained and is specifying The predicted value of prediction step.
- 8. accident forecast method for early warning under shale gas fractured well as claimed in claim 1, it is characterised in that described in the determination The trend feature vector of predicted value, including：The time series data that multiple predicted values that specified time window ranges interior prediction goes out are formed is divided into multiple identical The time slice of length；The width of the time slice is identical with the width of the prediction step；Linear fit, the linear relationship changed over time corresponding to acquisition are carried out to the time series data in each time slice Curve；The trend feature primitive according to corresponding to determining the linear relationship slope of a curve of each time slice；The trend feature base The value of member includes steady, raising and lowering；The trend feature vector of the time series data is built according to the trend feature primitive.
- 9. accident forecast method for early warning under shale gas fractured well as claimed in claim 1, it is characterised in that described in the basis Trend feature is vectorial and corresponding producing condition classification device, predicts the operating mode type corresponding to the predicted value, including：The producing condition classification device is substituted into using trend feature vector as input, obtains the operating mode class corresponding to the predicted value Type.
- A kind of 10. accident forecast prior-warning device under shale gas fractured well, it is characterised in that including：Online data acquisition module, the online monitoring data of monitoring parameters is specified for obtaining in real time；Monitoring parameters prediction module, for according to the online monitoring data and corresponding parametric prediction model, obtaining the finger Determine monitoring parameters and specify the predicted value of prediction step；Operating mode prediction and warning module, for determining the trend feature vector of the predicted value, and according to trend feature vector And corresponding producing condition classification device, the operating mode type corresponding to the predicted value is predicted, and be shale in the operating mode type predicted When air pressure splits down-hole accident, corresponding warning information is sent.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710812110.1A CN107642347B (en) | 2017-09-11 | 2017-09-11 | Shale gas fracturing underground accident prediction and early warning method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710812110.1A CN107642347B (en) | 2017-09-11 | 2017-09-11 | Shale gas fracturing underground accident prediction and early warning method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107642347A true CN107642347A (en) | 2018-01-30 |
| CN107642347B CN107642347B (en) | 2019-12-20 |
Family
ID=61110308
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710812110.1A Active CN107642347B (en) | 2017-09-11 | 2017-09-11 | Shale gas fracturing underground accident prediction and early warning method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107642347B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109815542A (en) * | 2018-12-21 | 2019-05-28 | 东软集团股份有限公司 | Recognition methods, device, storage medium and the electronic equipment of sand plugging after fracturing |
| CN109948935A (en) * | 2019-03-21 | 2019-06-28 | 重庆工程职业技术学院 | A kind of coal mine risk intelligent analysis system |
| CN110969315A (en) * | 2018-09-28 | 2020-04-07 | 北京国双科技有限公司 | Monitoring method of water injection pump and related equipment |
| CN111127235A (en) * | 2019-11-28 | 2020-05-08 | 东软集团股份有限公司 | Fracturing sand blocking early warning method and device and related products |
| CN111931323A (en) * | 2019-04-28 | 2020-11-13 | 中国石油化工股份有限公司 | Memory, hydrocracking equipment fault prediction method, device and equipment |
| CN112307619A (en) * | 2020-10-29 | 2021-02-02 | 北京航天测控技术有限公司 | Construction method of early warning model, and equipment fault early warning method and device |
| CN113825890A (en) * | 2019-04-05 | 2021-12-21 | 施耐德电子系统美国股份有限公司 | Autonomous failure prediction and pump control for well optimization |
| CN114626615A (en) * | 2022-03-21 | 2022-06-14 | 江苏仪化信息技术有限公司 | Production process monitoring and management method and system |
| CN114912789A (en) * | 2022-05-10 | 2022-08-16 | 中国石油大学(北京) | Drilling underground risk early warning method, equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5675088A (en) * | 1995-04-03 | 1997-10-07 | Serata; Shosei | Method and apparatus for automatic monitoring of tectonic stresses and quantitative forecast of shallow earthquakes |
| CN102865059A (en) * | 2012-09-26 | 2013-01-09 | 中国石油天然气股份有限公司 | Method and device for predicting capacity of fissure-hole type oil deposit |
| US9134453B1 (en) * | 2011-04-01 | 2015-09-15 | Simpson Holdings, Inc. | Forecasting hydrocarbon production |
| CN104948173A (en) * | 2014-03-25 | 2015-09-30 | 中国石油化工股份有限公司 | Fracturing field data long-distance real-time monitoring method and system |
| CN105507875A (en) * | 2015-12-01 | 2016-04-20 | 中国石油天然气股份有限公司 | Method and device for predicting oil, gas and water well production parameters in real time |
-
2017
- 2017-09-11 CN CN201710812110.1A patent/CN107642347B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5675088A (en) * | 1995-04-03 | 1997-10-07 | Serata; Shosei | Method and apparatus for automatic monitoring of tectonic stresses and quantitative forecast of shallow earthquakes |
| US9134453B1 (en) * | 2011-04-01 | 2015-09-15 | Simpson Holdings, Inc. | Forecasting hydrocarbon production |
| CN102865059A (en) * | 2012-09-26 | 2013-01-09 | 中国石油天然气股份有限公司 | Method and device for predicting capacity of fissure-hole type oil deposit |
| CN104948173A (en) * | 2014-03-25 | 2015-09-30 | 中国石油化工股份有限公司 | Fracturing field data long-distance real-time monitoring method and system |
| CN105507875A (en) * | 2015-12-01 | 2016-04-20 | 中国石油天然气股份有限公司 | Method and device for predicting oil, gas and water well production parameters in real time |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110969315A (en) * | 2018-09-28 | 2020-04-07 | 北京国双科技有限公司 | Monitoring method of water injection pump and related equipment |
| CN109815542A (en) * | 2018-12-21 | 2019-05-28 | 东软集团股份有限公司 | Recognition methods, device, storage medium and the electronic equipment of sand plugging after fracturing |
| CN109948935A (en) * | 2019-03-21 | 2019-06-28 | 重庆工程职业技术学院 | A kind of coal mine risk intelligent analysis system |
| CN113825890A (en) * | 2019-04-05 | 2021-12-21 | 施耐德电子系统美国股份有限公司 | Autonomous failure prediction and pump control for well optimization |
| CN111931323B (en) * | 2019-04-28 | 2022-08-12 | 中国石油化工股份有限公司 | Memory, hydrocracking equipment fault prediction method, device and equipment |
| CN111931323A (en) * | 2019-04-28 | 2020-11-13 | 中国石油化工股份有限公司 | Memory, hydrocracking equipment fault prediction method, device and equipment |
| CN111127235B (en) * | 2019-11-28 | 2022-05-10 | 东软集团股份有限公司 | Fracturing sand blocking early warning method and device and related products |
| CN111127235A (en) * | 2019-11-28 | 2020-05-08 | 东软集团股份有限公司 | Fracturing sand blocking early warning method and device and related products |
| CN112307619A (en) * | 2020-10-29 | 2021-02-02 | 北京航天测控技术有限公司 | Construction method of early warning model, and equipment fault early warning method and device |
| CN112307619B (en) * | 2020-10-29 | 2024-06-11 | 北京航天测控技术有限公司 | Construction method of early warning model, equipment fault early warning method and device |
| CN114626615A (en) * | 2022-03-21 | 2022-06-14 | 江苏仪化信息技术有限公司 | Production process monitoring and management method and system |
| CN114626615B (en) * | 2022-03-21 | 2023-02-03 | 江苏仪化信息技术有限公司 | Production process monitoring and management method and system |
| CN114912789A (en) * | 2022-05-10 | 2022-08-16 | 中国石油大学(北京) | Drilling underground risk early warning method, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107642347B (en) | 2019-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107642347A (en) | Accident forecast method for early warning and device under shale gas fractured well | |
| Feng et al. | Tunnel boring machines (TBM) performance prediction: A case study using big data and deep learning | |
| EP1984860B1 (en) | Methods, systems, and computer-readable media for fast updating of oil and gas field production models with physical and proxy simulators | |
| CN109508827B (en) | Drilling accident early warning method based on time recursion neural network | |
| AU2011283109B2 (en) | Systems and methods for predicting well performance | |
| Wang et al. | An integrated parameter prediction framework for intelligent TBM excavation in hard rock | |
| US8229880B2 (en) | Evaluation of acid fracturing treatments in an oilfield | |
| CN104769215A (en) | System and method for characterizing uncertainty in subterranean reservoir fracture networks | |
| MX2015003997A (en) | Analyzing microseismic data from a fracture treatment. | |
| CN107633325A (en) | A kind of shale gas pressure break down-hole accident early diagnosis alarm method and device | |
| US11512573B2 (en) | Stimulation using fiber-derived information and fracturing modeling | |
| Mahmoodzadeh et al. | Prediction of Mode-I rock fracture toughness using support vector regression with metaheuristic optimization algorithms | |
| Udegbe et al. | From face detection to fractured reservoir characterization: Big data analytics for restimulation candidate selection | |
| Lacayo et al. | Pressure Normalization of Production Rates Improves Forecasting Results | |
| CN114398786B (en) | Fracturing construction pressure prediction model establishing method and device and prediction method and device | |
| Zhou et al. | Data driven modeling and prediction for reservoir characterization using seismic attribute analyses and big data analytics | |
| Madasu et al. | Deep recurrent neural network DRNN model for real-time multistage pumping data | |
| CN114002129B (en) | High-water-pressure-crack rock mass seepage test platform | |
| Yu et al. | A data driven approach in screenout detection for horizontal wells | |
| Ma et al. | Integration of artificial intelligence and production data analysis for shale heterogeneity characterization in SAGD reservoirs | |
| Zheng et al. | Prioritizing Wells for Repurposing or Permanent Abandonment Based on Generalized Well Integrity Risk Analysis | |
| Hoang et al. | Successful case study of machine learning application to streamline and improve history matching process for complex gas-condensate reservoirs in Hai Thach field, offshore Vietnam | |
| Yong et al. | Development Strategy Optimization of Ning201 Block Longmaxi Shale Gas | |
| Sugai et al. | An integrated approach to reservoir performance monitoring and analysis | |
| Asante et al. | Data-driven modeling for forecasting oil recovery: A timeseries neural network approach for tertiary CO2 WAG EOR |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |