CN109209361A - A kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique - Google Patents
A kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique Download PDFInfo
- Publication number
- CN109209361A CN109209361A CN201810996052.7A CN201810996052A CN109209361A CN 109209361 A CN109209361 A CN 109209361A CN 201810996052 A CN201810996052 A CN 201810996052A CN 109209361 A CN109209361 A CN 109209361A
- Authority
- CN
- China
- Prior art keywords
- tracer
- time domain
- output
- layer neuron
- hidden layer
- 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
- 230000035699 permeability Effects 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000015572 biosynthetic process Effects 0.000 title claims abstract description 26
- 239000000700 radioactive tracer Substances 0.000 claims abstract description 192
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 115
- 238000002347 injection Methods 0.000 claims abstract description 82
- 239000007924 injection Substances 0.000 claims abstract description 82
- 238000004519 manufacturing process Methods 0.000 claims abstract description 71
- 238000013528 artificial neural network Methods 0.000 claims abstract description 41
- 239000000126 substance Substances 0.000 claims abstract description 41
- 238000005192 partition Methods 0.000 claims abstract description 36
- 238000013178 mathematical model Methods 0.000 claims abstract description 24
- 238000003062 neural network model Methods 0.000 claims abstract description 21
- 238000012544 monitoring process Methods 0.000 claims abstract description 19
- 238000005516 engineering process Methods 0.000 claims abstract description 10
- 210000002569 neuron Anatomy 0.000 claims description 146
- 230000003068 static effect Effects 0.000 claims description 40
- 239000011435 rock Substances 0.000 claims description 38
- 238000004146 energy storage Methods 0.000 claims description 14
- 230000001186 cumulative effect Effects 0.000 claims description 13
- 239000006185 dispersion Substances 0.000 claims description 9
- 238000009826 distribution Methods 0.000 claims description 9
- 239000000047 product Substances 0.000 claims description 8
- 239000002002 slurry Substances 0.000 claims description 8
- 239000000203 mixture Substances 0.000 claims description 6
- 239000003129 oil well Substances 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 5
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 239000012466 permeate Substances 0.000 claims description 3
- 238000009738 saturating Methods 0.000 claims description 2
- 238000011161 development Methods 0.000 abstract description 6
- 238000012360 testing method Methods 0.000 description 8
- 239000000243 solution Substances 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 210000005036 nerve Anatomy 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
Classifications
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH OR ROCK DRILLING; MINING
- E21B—EARTH OR ROCK DRILLING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR A SLURRY OF MINERALS FROM WELLS
- E21B49/00—Testing the nature of borehole walls; Formation testing; Methods or apparatus for obtaining samples of soil or well fluids, specially adapted to earth drilling or wells
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Mining & Mineral Resources (AREA)
- Geology (AREA)
- Fluid Mechanics (AREA)
- Environmental & Geological Engineering (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geochemistry & Mineralogy (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to Fractured Development Technique For Extreme Low Permeability Reservoir field more particularly to a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction techniques, initially set up time domain chemical tracing physics model；Next establishes time domain chemical tracing mathematical model；Time domain tracer neural network model is established again；Finally according to time domain chemical tracing physics model, time domain chemical tracing mathematical model and time domain tracer neural network model, time domain chemical tracing mathematical model is solved using BP neural network technology, crack band number N, crack band mean permeability K, injection water swept volume V, the injection water partition coefficient f and water drive fltting speed v of Fractured Low-Permeability Reservoir are predicted according to solving result.Prediction technique of the invention can predict the following well formation parameter using all previous tracer monitoring explanation results of injection-production well group, realize the passage of Fractured ultralow permeable reservoir parameter in time, the variation tendency and rule of feedback parameter.
Description
Technical field
The present invention relates to Fractured Development Technique For Extreme Low Permeability Reservoir field more particularly to a kind of Fractured extra-low permeability oil
Hide formation parameter prediction technique.
Background technique
During Fractured ultralow permeable reservoir waterflooding development, apparent time domain change can occur for formation parameter between injection-production well
Change, dynamic rule and trend are difficult to Efficient Characterization and prediction.It is continuous with Fractured ultralow permeable reservoir waterflooding extraction
Deeply, formation parameter can occur time domainization variation, especially the crack in stratum by water injection intensity, injection pressure etc. variation and
Change entire injection-production well group water flooding effectiveness.How using injection-production well group all previous tracer monitoring explanation results will be predicted future
Well formation parameter, the passage of realization Fractured ultralow permeable reservoir parameter in time, the variation tendency and rule of feedback parameter,
Have become the key for realizing Fractured ultralow permeable reservoir waterflooding extraction dynamic realtime monitoring and prediction.
Summary of the invention
In order to overcome the drawbacks of the prior art, the purpose of the present invention is to provide a kind of Fractured extra-low permeability oil reservoirs stratum
Parameter prediction method can predict the following well formation parameter using all previous tracer monitoring explanation results of injection-production well group, realize
The passage of Fractured ultralow permeable reservoir parameter in time, the variation tendency and rule of feedback parameter.
In order to achieve the above object, the technical solution of the present invention is as follows:
A kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique, comprising the following steps:
Step 1, time domain chemical tracing physics model is established, explain parameter, producing well using all previous tracer of well group and is adopted
Oil well production dynamic parameter combination associated static parameter realizes that tracer explains the passage and prediction of parameter in time, chooses straight
The known dynamic history information for indicating Fractured extra-low permeability oil reservoirs formation parameter and known static information are seen, is clearly solved
Variable；
Step 2, time domain chemical tracing mathematical model is established, with the time domain chemical tracing physics model that step 1 is established, showing
The historical information of track agent monitoring looks for inherent complicated implication relation between parameters, establishing description should as learning sample
The mathematical expression of physical model；
Step 3, establish time domain tracer neural network model, using BP neural network carry out time domain tracer analysis, establish compared with
For suitable network structure, including input layer number, node in hidden layer and output layer number of nodes；
Step 4, according to time domain chemical tracing physics model, time domain chemical tracing mathematical model and time domain tracer neural network
Model, solves time domain chemical tracing mathematical model using BP neural network technology, predicts Fractured hyposmosis according to solving result
Crack band number N, crack band mean permeability K, injection water swept volume V, injection water partition coefficient f and the water drive of oil reservoir
Fltting speed v.
Time domain chemical tracing physics model and time domain chemical tracing mathematical model in the present invention belong to time domain tracer reason
By model, the time domain tracer be exactly according to tracer monitoring history curve explain parameter value, and combine Production development data and
Geology static data predicts the variation tendency of following tracer production curve explanation parameter to analyze, to instruct oil-field flooding
The adjustment of development plan.Historical summary is monitored as data source using chemical tracing between well, is regarded as the sequence of time, utilizes mind
Carrying out time passage tracer prediction and analysis through network technology needs to establish one first and meets the actual physical model sum number in oil field
Model is learned, while should determine its definite condition from production reality.
Time domain chemical tracing physics model in the step 1 includes known dynamic history information, known static letter
Breath solves variable and Predicting Technique, in which:
Known dynamic history information includes: production time t, yield, moisture content, well group cumulative voidage replacement ratio, well group stage
Injection-production ratio, last time crack band number, last time crack band mean permeability, last time injection water swept volume, last time inject water
Distribution coefficient and last time water drive fltting speed；
Known static information includes: production section basement rock static state reservoir capacity, production section basement rock static state energy storage coefficient and note
Adopt well spacing；
Solving variable includes: crack band number, crack band mean permeability, injection water swept volume, injection moisture
Distribution coefficient and water drive fltting speed；
Predicting Technique is nerual network technique.
In known dynamic history information, last time crack band number, last time crack band mean permeability, last time injection
Water swept volume, last time injection water partition coefficient and last time water drive fltting speed are classified using Fractured ultralow permeable reservoir tracer
Interpretation model is calculated, and Fractured ultralow permeable reservoir tracer classification interpretation model includes that single peak type tracer explains physics mould
Type, multimodal tracer explain that physical model or wide-bed-type press brake type tracer explain that physical model is calculated.
Single peak type tracer explains that physical model is equivalent to man-made fracture and alters flow-through, explains physics mould in single peak type tracer
In type, 1 crack band is the flowing bundle being made of the flow tube that n length is L, equivalent diameter is D, and tracer is in this crack
It is flowed in n flow tube in band, tracer interpretation model are as follows:
In formula: C is tracer output concentration；C0For tracer initial concentration；T is the time；fjFor the injection to producing well j
Water partition coefficient；N is equivalent flow tube total number；VdFor the total injected slurry volume of tracer slug；α is hydrodynamic dispersion degree；L is flow tube
Equivalent length；D is the equivalent diameter of equivalent flow tube；Q is average daily injection；
Multimodal tracer explains physical model as difference crack interactive, and multimodal tracer is explained to be had in physical model
There is many cracks band, and the i-th crack band is distributed as by niA length is Li, equivalent diameter DiFlow tube composition stream
Tube bank, tracer is the n in many cracks bandiIt is flowed in a flow tube, tracer interpretation model are as follows:
In formula: C is tracer output concentration；C0For tracer initial concentration；T is the time；fjFor the injection to producing well j
Water partition coefficient；niFor the equivalent flow tube number of i-th of crack band；VdFor the total injected slurry volume of tracer slug；Q is average day
Injection rate；Di is the equivalent diameter of i-th of crack band any equivalent flow tube；Li is that the equivalent flow tube of i-th of crack band is long
Degree；αiFor hydrodynamic dispersion constant of the tracer in i-th of equivalent flow tube of crack band；
Wide-bed-type press brake type tracer explains that physical model as the relatively uniform drive-in in crack, is equivalent to many cracks band, and i-th
Crack band is distributed by niA length is Li, equivalent diameter DiFlow tube composition flowing bundle, tracer is in different flow tubes
Successively reach oil well, tracer interpretation model are as follows:
In formula: C is tracer output concentration；C0For tracer initial concentration；T is the time；fjFor the injection to producing well j
Water partition coefficient；niFor the equivalent flow tube number of i-th of crack band；VdFor the total injected slurry volume of tracer slug；Q is average day
Injection rate；Di is the equivalent diameter of i-th of crack band any equivalent flow tube；Li is that the equivalent flow tube of i-th of crack band is long
Degree；αiFor hydrodynamic dispersion constant of the tracer in i-th of equivalent flow tube of crack band；M is effective resistance constant.
In known static information, production section basement rock static state reservoir capacity PC is equal to the effective permeability k of production section basement rock
With the product of the effective thickness h of production section basement rock, production section basement rock static state energy storage coefficient NH is equal to effective thickness of production section basement rock
The effecive porosity spent h, produce section basement rockWith the oil saturation S of production section basement rockoProduct.
When establishing time domain tracer mathematical model, regard tracer monitoring process as a time point, going through for tracer monitoring
History information is as learning sample, then the time domain chemical tracing mathematical model are as follows:
In formula, z is time domain tracer BP neural network learning sample serial number；N is crack band number；K is flat for crack band
Equal permeability；V is injection water swept volume；F is injection water partition coefficient；V is water drive fltting speed；Q is yield；fwIt is aqueous
Rate；R is well group stage injection-production ratio；RC is well group cumulative voidage replacement ratio；W is weight, the coefficient of connection between two adjacent layers.
The time domain tracer neural network model is Three Tiered Network Architecture, including input layer, hidden layer and output layer group are hidden
It is 1 layer of structure containing layer；Output layer includes 5 nodes, and 5 nodes are respectively as follows: crack band number N, crack band averagely permeates
Rate K, injection water swept volume V, injection water partition coefficient f and water drive fltting speed v.
Input layer includes prediction time corresponding 8 parameters, and 8 parameters are respectively as follows: yield q, moisture content fw, well group accumulation
Injection-production ratio RC, well group stage injection-production ratio R, production time t, injector producer distance L, production section basement rock static state reservoir capacity PC and production section
Basement rock static state energy storage coefficient NH；
Or input layer includes following parameter: yield q, moisture content fw, well group cumulative voidage replacement ratio RC, well group stage injection-production ratio
R, production time t, injector producer distance L, production section basement rock static state reservoir capacity PC, production section basement rock static state energy storage coefficient NH, last time
Crack band number N, last time crack band mean permeability K, last time injection water swept volume V, last time inject water partition coefficient f
With last time water drive fltting speed v.
It include as follows using the process that BP neural network technology solves time domain chemical tracing mathematical model in the step 4
Step:
Step 4.1, time domain tracer BP neural network learning sample is normalized, time domain chemical tracing number is set
Learning all weights of model is lesser random number；
Step 4.2, it gives input vector X and desired output vector d: X is sent to the defeated of time domain tracer neural network model
Enter layer；
Step 4.3, the output vector Y of time domain tracer neural network model hidden layer is calculated；
Step 4.4, the output vector O of time domain tracer neural network model output layer is calculated；
Step 4.5, the error signal δ of time domain tracer neural network model output layer is calculatedo；
Step 4.6, the error signal δ of time domain tracer neural network model hidden layer is calculatedy；
Step 4.7, the error signal δ of output layer is utilizedoWith the error signal δ of hidden layeryWeight is adjusted and is repaired
Just；
Step 4.8, the error of BP neural network system is calculated；
Step 4.9, judge whether the error of BP neural network system meets default precision or more than maximum study number
Condition；
If the error of BP neural network system meets default precision or the condition more than maximum study number, crack is exported
Property the crack band number N of low-permeability oil deposit, crack band mean permeability K, injection water swept volume V, injection moisture disposition
Number f and water drive fltting speed v；
If the error of BP neural network system is unsatisfactory for default precision or the condition more than maximum study number, repeat to walk
Rapid 4.3 to step 4.9, the default precision of error satisfaction up to BP neural network system or the condition more than maximum study number,
Finally export the crack band number N of Fractured Low-Permeability Reservoir, crack band mean permeability K, injection water swept volume V,
Inject water partition coefficient f and water drive fltting speed v.
In step 4.2, input vector X and desired output vector d are respectively as follows:
X=[qI fwI RCI RI TI L PC NH]T
D=[NI KI VI fI vI]T
Wherein, qIFor the yield of i-th time domain tracer learning sample, fwIFor the aqueous of i-th time domain tracer learning sample
Rate, RCIFor the well group cumulative voidage replacement ratio of i-th time domain tracer learning sample, RIFor the well group of i-th time domain tracer learning sample
Stage injection-production ratio, TIFor the production time of i-th time domain tracer learning sample, L is injector producer distance, and PC is that production section basement rock is static
Reservoir capacity, NH are production section basement rock static state energy storage coefficient, NIFor the crack band number of i-th time domain tracer learning sample, KI
For the crack band mean permeability of i-th time domain tracer learning sample, VIFor the injection water of i-th time domain tracer learning sample
Swept volume, fIFor the injection water partition coefficient of i-th time domain tracer learning sample, vIFor i-th time domain tracer learning sample
Water drive fltting speed；
In step 4.3, output vector Y are as follows:
Wherein, A is the threshold value of each neuron of hidden layer, viqFor the power between input layer i and hidden layer neuron q
Value；vfwkFor input layer i and hidden layer neuron fwBetween weight；vRCkFor input layer i and hidden layer mind
Through the weight between first RC；vRkFor the weight between input layer i and hidden layer neuron R；vTkFor input layer
Weight between i and hidden layer neuron T；vLkFor the weight between input layer i and hidden layer neuron L；vPCkIt is defeated
Enter the weight layer between neuron i and hidden layer neuron PC；vNHkBetween input layer i and hidden layer neuron NH
Weight；
In step 4.4, output vector O are as follows:
Wherein, B is the threshold value of each neuron of output layer, wjNFor the power between output layer neuron N and hidden layer neuron j
Value；wjKFor the weight between output layer neuron K and hidden layer neuron j；wjVFor output layer neuron V and hidden layer nerve
Weight between first j；wjfFor the weight between output layer neuron f and hidden layer neuron j；wjvFor output layer neuron v with
Weight between hidden layer neuron j；yjFor j-th of output valve of hidden layer；
In step 4.5, the error signal δ of output layeroAre as follows:
δo=[oN(NI-oN)(1-oN)oK(KI-oK)(1-oK)
oV(VI-oV)(1-oV)of(fI-of)(1-of)
ov(vI-ov)(1-ov)]T
Wherein, oNFor the output valve of output layer crack band number；oKFor the output of output layer crack band mean permeability
Value；oVThe output valve of water swept volume is injected for output layer；ofThe output valve of water partition coefficient is injected for output layer；ovFor output
The output valve of layer water drive fltting speed；
In step 4.6, the error signal δ of hidden layeryAre as follows:
Wherein, δk oFor the error signal of output layer；wqkFor the power between output layer neuron k and hidden layer neuron q
Value；wfwkFor output layer neuron k and hidden layer neuron fwBetween weight；wRCkFor output layer neuron k and hidden layer mind
Through the weight between first RC；wRkFor the weight between output layer neuron k and hidden layer neuron R；wTkFor output layer neuron
Weight between k and hidden layer neuron T；wNHkFor the weight between output layer neuron k and hidden layer neuron NH；wPCkFor
Weight between output layer neuron k and hidden layer neuron PC；wLkBetween output layer neuron k and hidden layer neuron L
Weight；yqIFor the output valve of hidden layer neuron q i-th sample；yfwIFor hidden layer neuron fwThe output of i-th sample
Value；yRCIFor the output valve of hidden layer neuron RC i-th sample；yRIFor the output valve of hidden layer neuron R i-th sample；
yTIFor the output valve of hidden layer neuron T i-th sample；yPCFor the output valve of hidden layer neuron PC；yLFor hidden layer nerve
The output valve of first L；
In step 4.7, the error signal δ of output layer is utilizedoWith the error signal δ of hidden layeryWeight is carried out using following formula
Adjustment and amendment:
Wherein, output layer neuron k=1,2 ..., l；wjkFor the power between output layer neuron k and hidden layer neuron j
Value；Output layer neuron j=1,2 ..., m；vijFor the weight between input layer i and hidden layer neuron j；ΔwjkWith
ΔvijRespectively weight wjkAnd vijAlong the negative gradient direction change value of output error E；S is frequency of training；η is learning rate；Mc is
Factor of momentum；
In step 4.8, the error of BP neural network system are as follows:
Wherein, dk(I) desired output for being i-th time domain tracer learning sample output layer neuron k；okIt (I) is I
The output valve of a time domain tracer learning sample output layer neuron k.
The value range of learning rate η is 0.01~0.8；The value range of factor of momentum mc is 0.85~0.95.
Compared with prior art, the invention has the following beneficial effects:
Fractured extra-low permeability oil reservoirs formation parameter prediction technique of the invention first establishes time domain chemical tracing physics model；
Resettle time domain chemical tracing mathematical model；Resettle time domain tracer neural network model；Finally according to time domain chemical tracer
Model, time domain chemical tracing mathematical model and time domain tracer neural network model are managed, solves time domain using BP neural network technology
Chemical tracing mathematical model predicts that the crack band number N of Fractured Low-Permeability Reservoir, crack band are average according to solving result
Permeability K, injection water swept volume V, injection water partition coefficient f and water drive fltting speed v.Prediction technique of the invention being capable of benefit
The following well formation parameter is predicted with all previous tracer monitoring explanation results of injection-production well group, realizes Fractured ultralow permeable reservoir parameter
Passage in time, the variation tendency and rule of feedback parameter.Compared with other reservoir monitoring technologies, the present invention can be by independence
Reservoir monitoring point realize time domain, can be predicted any time well formation parameter, be Fractured extra-low permeability oil reservoirs it is subsequent
Development plan adjustment and water breakthrough water logging comprehensive treatment provide foundation.
Detailed description of the invention
Fig. 1 is that the single peak type tracer that the present invention uses explains physical model；
Fig. 2 is the tracing pattern that the single peak type tracer that the present invention uses explains physical model；
Fig. 3 is that the multimodal tracer that the present invention uses explains physical model；
Fig. 4 is the tracing pattern that the multimodal tracer that the present invention uses explains physical model；
Fig. 5 is that the wide-bed-type press brake type tracer that the present invention uses explains physical model；
Fig. 6 is the tracing pattern that the wide-bed-type press brake type tracer that the present invention uses explains physical model；
Fig. 7 is 1 time domain tracer neural network structure figure of scheme in the embodiment of the present invention；
Fig. 8 is 2 time domain tracer neural network structure figure of scheme in the embodiment of the present invention；
Fig. 9 is that the time domain tracer BP neural network that the present invention uses solves flow chart.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.
Embodiment
Time domain is carried out so that Ordos Basin GGY oil field A fills the water block 1291-3 injection-production well group as an example in the present embodiment to show
Track analysis mode.
The Fractured extra-low permeability oil reservoirs formation parameter prediction technique of the present embodiment, comprising the following steps:
Step 1, time domain chemical tracing physics model is established, time domain chemical tracing physics model mainly includes known dynamic
Historical information, known static information solve 4 parts of variable and Predicting Technique, specific as follows:
(1) dynamic history information known to includes: production time t, yield, moisture content, well group cumulative voidage replacement ratio, well group rank
Section injection-production ratio, last time crack band number (, last time crack band mean permeability, last time injection water swept volume, last time injection
Water partition coefficient and last time water drive fltting speed；
(2) static information known to include: production section basement rock static state reservoir capacity, production section basement rock static state energy storage coefficient and
Injector producer distance；
(3) solving variable includes: crack band number, crack band mean permeability, injection water swept volume, injection water
Distribution coefficient and water drive fltting speed；
(4) Predicting Technique is nerual network technique.
In time domain chemical tracing physics model, it is known that dynamic history information: i.e. last time crack band number (, last time crack
This 5 band mean permeability, last time injection water swept volume, last time injection water partition coefficient and last time water drive fltting speed ginsengs
Number is calculated using following Fractured ultralow permeable reservoir tracer classification interpretation model, that is, uses single peak type tracer solution
Release physical model, multimodal tracer explains physical model or wide-bed-type press brake type tracer explains that physical model is calculated:
1. as shown in Figure 1, single peak type tracer explains that physical model is equivalent to man-made fracture and alters flow-through, i.e. 1 crack item
Band by n length is L, the flowing bundle that the flow tube that equivalent diameter is D forms, and tracer can be regarded as in this crack band
It is flowed in n flow tube, single peak type tracer explains the tracing pattern of physical model as shown in Fig. 2, single peak type tracer explains mould
Type are as follows:
In formula: C is tracer output concentration, unit mg/L；C0For tracer initial concentration, unit g/cm3；When t is
Between, unit is d (day)；fjFor to the injection water partition coefficient of producing well j, the tracer quantum of output that can be measured by each producing well is accounted for
The ratio of total tracer injection rate obtains；N is equivalent flow tube total number, and unit is a；VdBody is always injected for tracer slug
Product, unit m3；α is hydrodynamic dispersion degree, unit m；L is the equivalent length of flow tube, unit m；D is working as equivalent flow tube
Diameter is measured, unit is μm；Q is average daily injection, unit m3/d。
2. as shown in figure 3, multimodal tracer explain physical model as difference crack interactive, i.e. many cracks band,
And i-th crack band distribution can regard as by niA length is Li, equivalent diameter DiFlow tube composition flowing bundle, respectively
Seepage flow difference between the band of crack is larger, and tracer can regard the n in these crack bands asiIt is flowed in a flow tube
Dynamic, multimodal tracer explains the tracing pattern of physical model as shown in figure 4, multimodal tracer interpretation model are as follows:
In formula: C is tracer output concentration, unit mg/L；C0For tracer initial concentration, unit g/cm3；When t is
Between, unit d；fjFor to the injection water partition coefficient of producing well j, the tracer quantum of output that can be measured by each producing well accounts for total
The ratio of tracer injection rate obtains；niFor the equivalent flow tube number of i-th of crack band, unit is a；VdFor tracer slug
Total injected slurry volume, unit m3；Q is average daily injection, unit m3/d；Di is i-th of crack band any equivalent flow tube
Equivalent diameter, unit are μm；Li is the equivalent flow tube length of i-th of crack band, unit m；αiIt is split for tracer at i-th
Stitch the hydrodynamic dispersion constant in the equivalent flow tube of band, unit m.
3. as shown in figure 5, wide-bed-type press brake type tracer explains that physical model as the relatively uniform drive-in in crack, is equivalent to a plurality of split
Band is stitched, and the distribution of the i-th crack band is it is seen that by niA length is Li, equivalent diameter DiFlow tube composition stream
Tube bank, tracer successively reach oil well in different flow tubes, and wide-bed-type press brake type tracer explains tracing pattern such as Fig. 6 institute of physical model
Show.Wide-bed-type press brake type tracer interpretation model are as follows:
In formula: C is tracer output concentration, unit mg/L；C0For tracer initial concentration, unit g/cm3；When t is
Between, unit d；fjFor to the injection water partition coefficient of producing well j, the tracer quantum of output that can be measured by each producing well accounts for total
The ratio of tracer injection rate obtains；niFor the equivalent flow tube number of i-th of crack band, unit is a；VdFor tracer slug
Total injected slurry volume, unit m3；Q is average daily injection, unit m3/d；Di is i-th of crack band any equivalent flow tube
Equivalent diameter, unit are μm；Li is the equivalent flow tube length of i-th of crack band, unit m；αiIt is split for tracer at i-th
Stitch the hydrodynamic dispersion constant in the equivalent flow tube of band, unit m；M is effective resistance constant, and unit is m/ μm4。
In time domain chemical tracing physics model, it is known that in static historical information: basement rock static state reservoir capacity PC is equal to effective
The product of permeability k and effective thickness h, basement rock static state energy storage coefficient NH are equal to effective thickness h, effecive porosityIt is full with oil-containing
With degree SoProduct.
In time domain chemical tracing physics model, the variable parameter of solution are as follows: crack band number, crack band averagely permeate
Rate, injection water swept volume, injection water partition coefficient, water drive fltting speed.
In time domain chemical tracing physics model, Predicting Technique is nerual network technique.
Step 2, time domain tracer mathematical model is established:
When establishing time domain tracer mathematical model, what it is due to inter-well tracer test monitoring is injection fluid under the steady pressure of stratum point
Cloth, stratum filtration feature are held essentially constant, and tracer monitoring hardly influences the normal production of oil well, tracer monitoring process
Regard a time point as.Using the historical information of tracer monitoring as learning sample, look for inherent complicated between parameters
Implication relation then have time domain tracer mathematical model so that each tracer monitoring is continuously got up are as follows:
In formula, z is time domain tracer BP neural network learning sample serial number；N is crack band number, and unit is a；K is to split
Stitch band mean permeability, unit 10-3μm；V is injection water swept volume, unit m3；F is injection water partition coefficient, single
Position is f；V water drive fltting speed, unit m/d；Q is yield, unit t/d；fwFor moisture content, unit %；R is well group rank
Section injection-production ratio, unit f；RC is well group cumulative voidage replacement ratio, unit f；W is weight, the coefficient of connection between as two adjacent layers.
Step 3, time domain tracer neural network model is established
The network structure of time domain neural network model includes three layers: i.e. input layer, hidden layer and output layer.Hidden layer includes
1 layer；Output layer is arranged 5 nodes and is respectively as follows: crack band number N, crack band mean permeability K, injection water swept volume
V, water partition coefficient f and water drive fltting speed v is injected.Input layer designs two schemes, and 1 consideration prediction time of scheme is corresponding
8 parameters: i.e. yield q, moisture content fw, well group cumulative voidage replacement ratio RC, well group stage injection-production ratio R, production time t, injector producer distance
L, section basement rock static state reservoir capacity PC and production section basement rock static state energy storage coefficient NH are produced；As shown in fig. 7, by taking scheme 1 as an example, it is defeated
Enter vector number n=8 of layer input, the vector number of output layer output l=5, hidden layer 1, hidden number m=8 is a；
Scheme 2 is in addition to considering prediction time corresponding 8 parameters, it is also contemplated that 5 parameters that last time tracer monitoring is explained:
Also contemplate crack band number N, crack band mean permeability K, injection water swept volume V, injection water partition coefficient f and
Water drive fltting speed v, by taking scheme 2 as an example, the vector number of input layer input n=13, the vector number l=of output layer output
5, hidden layer 1, hidden number m=13.
Step 4, time domain chemical tracing mathematical model is solved using BP neural network technology, referring to Fig. 9, solution procedure is as follows
(being solved by taking scheme 1 as an example in step 4):
Step 4.1, time domain tracer BP neural network learning sample is normalized using minimax method, is arranged
All weights are lesser random number in time domain chemical tracing mathematical model；
Step 4.2, input vector X and desired output vector d are provided:
X=[qI fwI RCI RI TI L PC NH]T
D=[NI KI VI fI vI]T
X is sent to input layer, qIFor the yield of i-th time domain tracer learning sample, fwILearn sample for i-th time domain tracer
This moisture content, RCIFor the well group cumulative voidage replacement ratio of i-th time domain tracer learning sample, RILearn sample for i-th time domain tracer
This well group stage injection-production ratio, TIFor the production time of i-th time domain tracer learning sample, L is injector producer distance, and PC is production section
Basement rock static state reservoir capacity, NH are production section basement rock static state energy storage coefficient, NIFor the crack item of i-th time domain tracer learning sample
Band number, KIFor the crack band mean permeability of i-th time domain tracer learning sample, VIFor i-th time domain tracer learning sample
Injection water swept volume, fIFor the injection water partition coefficient of i-th time domain tracer learning sample, vIFor i-th time domain tracer
Practise the water drive fltting speed of sample；
Step 4.3, the output vector Y of hidden layer is calculated:
In formula, A is the threshold value of each neuron of hidden layer；viqFor the power between input layer i and hidden layer neuron q
Value；vfwkFor the weight between input layer i and hidden layer neuron fw；vRCkFor input layer i and hidden layer mind
Through the weight between first RC；vRkFor the weight between input layer i and hidden layer neuron R；vTkFor input layer
Weight between i and hidden layer neuron T；vLkFor the weight between input layer i and hidden layer neuron L；vPCkIt is defeated
Enter the weight layer between neuron i and hidden layer neuron PC；vNHkBetween input layer i and hidden layer neuron NH
Weight；
Step 4.4, the output vector O of output layer is calculated:
In formula, B is the threshold value of each neuron of output layer, and B is the threshold value of each neuron of output layer；wjNFor output layer neuron
Weight between N and hidden layer neuron j；wjKFor the weight between output layer neuron K and hidden layer neuron j；wjVIt is defeated
Weight between layer neuron V and hidden layer neuron j out；wjfFor the power between output layer neuron f and hidden layer neuron j
Value；wjvFor the weight between output layer neuron v and hidden layer neuron j；yjFor j-th of output valve of hidden layer；
Step 4.5, the error signal δ of output layer is calculatedo:
δo=[oN(NI-oN)(1-oN)oK(KI-oK)(1-oK)
oV(VI-oV)(1-oV)of(fI-of)(1-of)
ov(vI-ov)(1-ov)]T
Wherein, oNFor the output valve of output layer crack band number；oKFor the output of output layer crack band mean permeability
Value；oVThe output valve of water swept volume is injected for output layer；ofThe output valve of water partition coefficient is injected for output layer；ovFor output
The output valve of layer water drive fltting speed
Step 4.6, the error signal δ of hidden layer is calculatedy:
Wherein, δk oFor the error signal of output layer；wqkFor the power between output layer neuron k and hidden layer neuron q
Value；wfwkFor output layer neuron k and hidden layer neuron fwBetween weight；wRCkFor output layer neuron k and hidden layer mind
Through the weight between first RC；wRkFor the weight between output layer neuron k and hidden layer neuron R；wTkFor output layer neuron
Weight between k and hidden layer neuron T；wNHkFor the weight between output layer neuron k and hidden layer neuron NH；wPCkFor
Weight between output layer neuron k and hidden layer neuron PC；wLkBetween output layer neuron k and hidden layer neuron L
Weight；yqIFor the output valve of hidden layer neuron q i-th sample；yfwIFor the output of hidden layer neuron fw i-th sample
Value；yRCIFor the output valve of hidden layer neuron RC i-th sample；yRIFor the output valve of hidden layer neuron R i-th sample；
yTIFor the output valve of hidden layer neuron T i-th sample；yPCFor the output valve of hidden layer neuron PC；yLFor hidden layer nerve
The output valve of first L；
Step 4.7, the error signal δ of output layer is utilizedoWith the error signal δ of hidden layeryWeight is adjusted using following formula
Whole and amendment, i.e. the modified weight formula of adaptive-learning-rate with momentum BP neural network are as follows；
In formula, k=1,2 ..., l；wjkFor the weight between output layer neuron k and hidden layer neuron j；J=1,
2 ..., m；vijFor the weight between input layer i and hidden layer neuron j；ΔwjkWith Δ vijRespectively weight wjkWith
vijAlong the negative gradient direction change value of output error E；S is frequency of training；η is learning rate, generally takes 0.01~0.8；Mc is
The factor is measured, takes 0.85~0.95.
Step 4.8, the error of BP neural network system is calculated:
Wherein, dk(I) desired output for being i-th time domain tracer learning sample output layer neuron k；okIt (I) is I
The output valve of a time domain tracer learning sample output layer neuron k；
Step 4.9, judge whether the error of BP neural network system meets default precision or more than maximum study number
Condition；
If the error of BP neural network system meets default precision or the condition more than maximum study number, crack is exported
Property the crack band number N of low-permeability oil deposit, crack band mean permeability K, injection water swept volume V, injection moisture disposition
Number f and water drive fltting speed v；
If the error of BP neural network system is unsatisfactory for default precision or the condition more than maximum study number, repeat to walk
Rapid 4.3 to step 4.9, into the study of next round, until the error of BP neural network system meets default precision or is more than most
The condition of big study number, finally export the crack band number N of Fractured Low-Permeability Reservoir, crack band mean permeability K,
Inject water swept volume V, injection water partition coefficient f and water drive fltting speed v.
The corresponding time domain tracer neural network test sample of scheme 1 and the results are shown in Table 1:
Table 1
1 time domain tracer neural network test sample of scheme the result shows that: the absolute error of 4 test samples be 0.269, phase
It is 4.36% to error；The maximum absolute error of crack band number is 0.1, relative error 50%；Crack band averagely seeps
Saturating rate maximum absolute error is 0.105, relative error 14.1%；Injecting water swept volume maximum absolute error is 0.121, phase
It is 108.0% to error；Injecting water partition coefficient maximum absolute error is 0.18, relative error 128.6%；Water drive promotes speed
Spending maximum absolute error is 0.23, relative error 35.4%.
When being solved by taking scheme 2 as an example, the selection of time domain tracer neural network learning sample and test sample, model
Method for solving, and maximum frequency of training, minimum training error, factor of momentum and initial learning rate setting with scheme 1
It is identical.Hands-on 787 times, error 0.67%, test result is shown in Table 2:
Table 2
2 time domain tracer neural network test sample of scheme the result shows that: the absolute error of 4 test samples be 0.120, phase
It is 1.94% to error；The maximum absolute error of crack band number is 0；Crack band mean permeability maximum absolute error is
0.045, relative error 9.34%；Injecting water swept volume maximum absolute error is 0.008, relative error 7.08%；Note
Entering water partition coefficient maximum absolute error is 0.02, relative error 14.29%；Water drive fltting speed maximum absolute error is
0.023, relative error 13.37%.
By scheme 1 and scheme 2 the result shows that, can be into using time domain chemical tracer technique and BP neural network technology
The prediction of row Fractured extra-low permeability oil reservoirs formation parameter, and when being predicted, consider last prediction result, it can be more quasi-
True judgement prediction result next time.
Claims (10)
1. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique, which comprises the following steps:
Step 1, time domain chemical tracing physics model is established；
Step 2, time domain chemical tracing mathematical model is established；
Step 3, time domain tracer neural network model is established；
Step 4, according to time domain chemical tracing physics model, time domain chemical tracing mathematical model and time domain tracer neural network mould
Type, solves time domain chemical tracing mathematical model using BP neural network technology, predicts Fractured low-permeability oil according to solving result
Crack band number N, crack band mean permeability K, injection water swept volume V, injection water partition coefficient f and the water drive of hiding push away
Into speed v.
2. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 1, which is characterized in that institute
The time domain chemical tracing physics model in step 1 is stated to include known dynamic history information, known static information, solve variable
And Predicting Technique, in which:
Known dynamic history information include: production time t, yield, moisture content, well group cumulative voidage replacement ratio, the well group stage note adopt
Than, last time crack band number, last time crack band mean permeability, last time injection water swept volume, last time injection moisture match
Coefficient and last time water drive fltting speed；
Known static information includes: production section basement rock static state reservoir capacity, production section basement rock static state energy storage coefficient and injection-production well
Away from；
Solving variable includes: crack band number, crack band mean permeability, injection water swept volume, injection moisture disposition
Several and water drive fltting speed；
Predicting Technique is nerual network technique.
3. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 2, which is characterized in that
In the static information known, production section basement rock static state reservoir capacity PC is equal to the effective permeability k and production Duan Ji of production section basement rock
The product of the effective thickness h of rock, production section basement rock static state energy storage coefficient NH are equal to the effective thickness h of production section basement rock, production section
The effecive porosity of basement rockWith the oil saturation S of production section basement rockoProduct.
4. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 2, which is characterized in that
In the dynamic history information known, last time crack band number, last time crack band mean permeability, last time injection ripples and body
Product, last time injection water partition coefficient and last time water drive fltting speed are using Fractured ultralow permeable reservoir tracer classification interpretation model
It is calculated, Fractured ultralow permeable reservoir tracer classification interpretation model includes that single peak type tracer explains physical model, multimodal
Type tracer explains that physical model or wide-bed-type press brake type tracer explain that physical model is calculated.
5. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 3, which is characterized in that single
Peak type tracer explains that physical model is equivalent to man-made fracture and alters flow-through, explains in physical model that 1 splits in single peak type tracer
Stitching band is the flowing bundle being made of the flow tube that n length is L, equivalent diameter is D, n of the tracer in this crack band
It is flowed in a flow tube, tracer interpretation model are as follows:
In formula: C is tracer output concentration；C0For tracer initial concentration；T is the time；fjFor to the injection moisture of producing well j
Distribution coefficient；N is equivalent flow tube total number；VdFor the total injected slurry volume of tracer slug；α is hydrodynamic dispersion degree；L be flow tube etc.
Imitate length；D is the equivalent diameter of equivalent flow tube；Q is average daily injection；
Multimodal tracer explains physical model as difference crack interactive, and multimodal tracer is explained in physical model with more
Crack band, and the i-th crack band is distributed as by niA length is Li, equivalent diameter DiFlow tube composition flowing bundle,
Tracer is the n in many cracks bandiIt is flowed in a flow tube, tracer interpretation model are as follows:
In formula: C is tracer output concentration；C0For tracer initial concentration；T is the time；fjFor to the injection moisture of producing well j
Distribution coefficient；niFor the equivalent flow tube number of i-th of crack band；VdFor the total injected slurry volume of tracer slug；Q is average day injection
Amount；Di is the equivalent diameter of i-th of crack band any equivalent flow tube；Li is the equivalent flow tube length of i-th of crack band；αi
For hydrodynamic dispersion constant of the tracer in i-th of equivalent flow tube of crack band；
Wide-bed-type press brake type tracer explains that physical model is equivalent to many cracks band as the relatively uniform drive-in in crack, and i-th splits
Band distribution is stitched by niA length is Li, equivalent diameter DiFlow tube composition flowing bundle, tracer in different flow tubes successively
Reach oil well, tracer interpretation model are as follows:
In formula: C is tracer output concentration；C0For tracer initial concentration；T is the time；fjFor to the injection moisture of producing well j
Distribution coefficient；niFor the equivalent flow tube number of i-th of crack band；VdFor the total injected slurry volume of tracer slug；Q is average day injection
Amount；Di is the equivalent diameter of i-th of crack band any equivalent flow tube；Li is the equivalent flow tube length of i-th of crack band；αi
For hydrodynamic dispersion constant of the tracer in i-th of equivalent flow tube of crack band；M is effective resistance constant.
6. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 5, which is characterized in that build
Immediately when domain tracer mathematical model, regard tracer monitoring process as a time point, using the historical information of tracer monitoring as
Learning sample, then the time domain chemical tracing mathematical model are as follows:
In formula, z is time domain tracer BP neural network learning sample serial number；N is crack band number；K is that crack band averagely seeps
Saturating rate；V is injection water swept volume；F is injection water partition coefficient；V is water drive fltting speed；Q is yield；fwFor moisture content；R
For well group stage injection-production ratio；RC is well group cumulative voidage replacement ratio；W is weight, the coefficient of connection between two adjacent layers.
7. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 6, which is characterized in that institute
Stating time domain tracer neural network model is Three Tiered Network Architecture, including input layer, hidden layer and output layer group, and hidden layer is 1 layer
Structure；Output layer includes 5 nodes, and 5 nodes are respectively as follows: crack band number N, crack band mean permeability K, injection water
Swept volume V, injection water partition coefficient f and water drive fltting speed v.
8. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 7, which is characterized in that defeated
Entering layer includes prediction time corresponding 8 parameters, and 8 parameters are respectively as follows: yield q, moisture content fw, well group cumulative voidage replacement ratio RC,
Well group stage injection-production ratio R, production time t, injector producer distance L, production section basement rock static state reservoir capacity PC and production section basement rock are static
Energy storage coefficient NH；
Or input layer includes following parameter: yield q, moisture content fw, well group cumulative voidage replacement ratio RC, well group stage injection-production ratio R, life
Produce time t, injector producer distance L, production section basement rock static state reservoir capacity PC, production section basement rock static state energy storage coefficient NH, last time crack
Band number N, last time crack band mean permeability K, last time injection water swept volume V, last time injection water partition coefficient f and on
Secondary water drive fltting speed v.
9. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 8, which is characterized in that institute
It states in step 4, is included the following steps: using the process that BP neural network technology solves time domain chemical tracing mathematical model
Step 4.1, time domain tracer BP neural network learning sample is normalized, time domain chemical tracing mathematical modulo is set
All weights are lesser random number in type；
Step 4.2, it gives input vector X and desired output vector d: X is sent to the input layer of time domain tracer neural network model；
Step 4.3, the output vector Y of time domain tracer neural network model hidden layer is calculated；
Step 4.4, the output vector O of time domain tracer neural network model output layer is calculated；
Step 4.5, the error signal δ of time domain tracer neural network model output layer is calculatedo；
Step 4.6, the error signal δ of time domain tracer neural network model hidden layer is calculatedy；
Step 4.7, the error signal δ of output layer is utilizedoWith the error signal δ of hidden layeryWeight is adjusted and is corrected；
Step 4.8, the error of BP neural network system is calculated；
Step 4.9, judge whether the error of BP neural network system meets default precision or the condition more than maximum study number；
If the error of BP neural network system meets default precision or the condition more than maximum study number, it is low to export Fractured
Permeate the crack band number N of oil reservoir, crack band mean permeability K, injection water swept volume V, injection water partition coefficient f and
Water drive fltting speed v；
If the error of BP neural network system is unsatisfactory for default precision or the condition more than maximum study number, repeatedly step
4.3 to step 4.9, until the error of BP neural network system meets default precision or the condition more than maximum study number, most
The crack band number N, crack band mean permeability K, injection water swept volume V, note of Fractured Low-Permeability Reservoir are exported afterwards
Enter water partition coefficient f and water drive fltting speed v.
10. a kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique according to claim 9, which is characterized in that
In step 4.2, input vector X and desired output vector d are respectively as follows:
X=[qI fwI RCI RI TI L PC NH]T
D=[NI KI VI fI vI]T
Wherein, qIFor the yield of i-th time domain tracer learning sample, fwIFor the moisture content of i-th time domain tracer learning sample, RCI
For the well group cumulative voidage replacement ratio of i-th time domain tracer learning sample, RIIt is infused for the well group stage of i-th time domain tracer learning sample
Adopt ratio, TIFor the production time of i-th time domain tracer learning sample, L is injector producer distance, and PC is production section basement rock static state production capacity system
Number, NH are production section basement rock static state energy storage coefficient, NIFor the crack band number of i-th time domain tracer learning sample, KIFor I
The crack band mean permeability of a time domain tracer learning sample, VIFor i-th time domain tracer learning sample injection ripples and
Volume, fIFor the injection water partition coefficient of i-th time domain tracer learning sample, vIFor the water drive of i-th time domain tracer learning sample
Fltting speed；
In step 4.3, output vector Y are as follows:
Wherein, A is the threshold value of each neuron of hidden layer；viqFor the weight between input layer i and hidden layer neuron q；
vfwkFor the weight between input layer i and hidden layer neuron fw；vRCkFor input layer i and hidden layer neuron
Weight between RC；vRkFor the weight between input layer i and hidden layer neuron R；vTkFor input layer i with
Weight between hidden layer neuron T；vLkFor the weight between input layer i and hidden layer neuron L；vPCkFor input
Weight between layer neuron i and hidden layer neuron PC；vNHkBetween input layer i and hidden layer neuron NH
Weight；
In step 4.4, output vector O are as follows:
Wherein, B is the threshold value of each neuron of output layer, and B is the threshold value of each neuron of output layer；wjNFor output layer neuron N with
Weight between hidden layer neuron j；wjKFor the weight between output layer neuron K and hidden layer neuron j；wjVFor output
Weight between layer neuron V and hidden layer neuron j；wjfFor the power between output layer neuron f and hidden layer neuron j
Value；wjvFor the weight between output layer neuron v and hidden layer neuron j；yjFor j-th of output valve of hidden layer；
In step 4.5, the error signal δ of output layeroAre as follows:
δo=[oN(NI-oN)(1-oN) oK(KI-oK)(1-oK)
oV(VI-oV)(1-oV) of(fI-of)(1-of)
ov(vI-ov)(1-ov)]T
Wherein, oNFor the output valve of output layer crack band number；oKFor the output valve of output layer crack band mean permeability；
oVThe output valve of water swept volume is injected for output layer；ofThe output valve of water partition coefficient is injected for output layer；ovFor output layer water
Drive the output valve of fltting speed；
In step 4.6, the error signal δ of hidden layeryAre as follows:
Wherein, δk oFor the error signal of output layer；wqkFor the weight between output layer neuron k and hidden layer neuron q；wfwk
For output layer neuron k and hidden layer neuron fwBetween weight；wRCkFor output layer neuron k and hidden layer neuron RC
Between weight；wRkFor the weight between output layer neuron k and hidden layer neuron R；wTkFor output layer neuron k with it is hidden
Weight between the T of neuron containing layer；wNHkFor the weight between output layer neuron k and hidden layer neuron NH；wPCkFor output
Weight between layer neuron k and hidden layer neuron PC；wLkFor the power between output layer neuron k and hidden layer neuron L
Value；yqIFor the output valve of hidden layer neuron q i-th sample；yfwIFor the output valve of hidden layer neuron fw i-th sample；
yRCIFor the output valve of hidden layer neuron RC i-th sample；yRIFor the output valve of hidden layer neuron R i-th sample；yTIFor
The output valve of hidden layer neuron T i-th sample；yPCFor the output valve of hidden layer neuron PC；yLFor hidden layer neuron L's
Output valve；
In step 4.7, the error signal δ of output layer is utilizedoWith the error signal δ of hidden layeryWeight is adjusted using following formula
And amendment:
Wherein, output layer neuron k=1,2 ..., l；wjkFor the weight between output layer neuron k and hidden layer neuron j；
Output layer neuron j=1,2 ..., m；vijFor the weight between input layer i and hidden layer neuron j；ΔwjkAnd Δ
vijRespectively weight wjkAnd vijAlong the negative gradient direction change value of output error E；S is frequency of training；η is learning rate, value model
Enclose is 0.01~0.8；Mc is factor of momentum, and value range is 0.85~0.95；
In step 4.8, the error of BP neural network system are as follows:
Wherein, dk(I) desired output for being i-th time domain tracer learning sample output layer neuron k；ok(I) be i-th when
The output valve of domain tracer learning sample output layer neuron k.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810996052.7A CN109209361B (en) | 2018-08-29 | 2018-08-29 | Method for predicting stratum parameters of fractured ultra-low permeability reservoir |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810996052.7A CN109209361B (en) | 2018-08-29 | 2018-08-29 | Method for predicting stratum parameters of fractured ultra-low permeability reservoir |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109209361A true CN109209361A (en) | 2019-01-15 |
| CN109209361B CN109209361B (en) | 2022-02-18 |
Family
ID=64985286
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810996052.7A Active CN109209361B (en) | 2018-08-29 | 2018-08-29 | Method for predicting stratum parameters of fractured ultra-low permeability reservoir |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109209361B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800521A (en) * | 2019-01-28 | 2019-05-24 | 中国石油大学(华东) | A kind of oil-water relative permeability curve calculation method based on machine learning |
| CN113323618A (en) * | 2021-06-02 | 2021-08-31 | 德仕能源科技集团股份有限公司 | Ultra-low permeability reservoir artificial fracture water channeling treatment method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140025301A1 (en) * | 2012-07-20 | 2014-01-23 | Bruce H. Storm, Jr. | Determination of subsurface properties of a well |
| CN106526693A (en) * | 2016-09-30 | 2017-03-22 | 中国石油天然气股份有限公司 | Crack recognition method and device |
| US20180018561A1 (en) * | 2016-07-14 | 2018-01-18 | King Fahd University Of Petroleum And Minerals | Apparatuses, systems, and methodologies for permeability prediction |
| CN107989597A (en) * | 2017-11-15 | 2018-05-04 | 中国石油大学(北京) | A kind of crack data screening technique, device and storage medium |
| CN108301823A (en) * | 2018-01-19 | 2018-07-20 | 北京捷贝通石油技术股份有限公司 | A method of identification reservoir hydrocarbons dessert |
| CN108319756A (en) * | 2017-12-29 | 2018-07-24 | 西安石油大学 | A kind of compact reservoir volume fracturing seam net extended simulation and characterizing method |
-
2018
- 2018-08-29 CN CN201810996052.7A patent/CN109209361B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140025301A1 (en) * | 2012-07-20 | 2014-01-23 | Bruce H. Storm, Jr. | Determination of subsurface properties of a well |
| US20180018561A1 (en) * | 2016-07-14 | 2018-01-18 | King Fahd University Of Petroleum And Minerals | Apparatuses, systems, and methodologies for permeability prediction |
| CN106526693A (en) * | 2016-09-30 | 2017-03-22 | 中国石油天然气股份有限公司 | Crack recognition method and device |
| CN107989597A (en) * | 2017-11-15 | 2018-05-04 | 中国石油大学(北京) | A kind of crack data screening technique, device and storage medium |
| CN108319756A (en) * | 2017-12-29 | 2018-07-24 | 西安石油大学 | A kind of compact reservoir volume fracturing seam net extended simulation and characterizing method |
| CN108301823A (en) * | 2018-01-19 | 2018-07-20 | 北京捷贝通石油技术股份有限公司 | A method of identification reservoir hydrocarbons dessert |
Non-Patent Citations (2)
| Title |
|---|
| 吴洪彪 等: "四维试井理论研究", 《石油学报》 * |
| 蒲春生 等: "裂缝性特低渗透油藏水窜水淹逐级调控多级井间化学示踪技术", 《石油勘探与开发》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800521A (en) * | 2019-01-28 | 2019-05-24 | 中国石油大学(华东) | A kind of oil-water relative permeability curve calculation method based on machine learning |
| CN113323618A (en) * | 2021-06-02 | 2021-08-31 | 德仕能源科技集团股份有限公司 | Ultra-low permeability reservoir artificial fracture water channeling treatment method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109209361B (en) | 2022-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105095986B (en) | The method of stratified reservoir overall yield prediction | |
| Peters et al. | Extended Brugge benchmark case for history matching and water flooding optimization | |
| CN107829718A (en) | Oil reservoir well pattern and injection-production program Optimization Design based on balanced water drive theory | |
| CN105317407B (en) | A kind of development approach of ultra-high water cut stage Untabulated reservoirs | |
| CN106295095A (en) | New method based on Conventional Logs prediction low permeability sandstone reservoir production capacity | |
| CN106150477A (en) | A kind of method determining single well controlled reserves | |
| CN105822302A (en) | Oil-water distribution recognition method based on well-ground potential method | |
| Fonseca et al. | Introduction to the special issue: overview of OLYMPUS optimization benchmark challenge | |
| CN108984886A (en) | A kind of method of INVERSION OF MULTI-LAYER oil deposit inter-well dynamic connectivity | |
| CN109063403B (en) | Optimal design method for slickwater fracturing | |
| CN108643875A (en) | A kind of waterflooding extraction method of adjustment of hyposmosis clastic rock oil reservoir, apparatus and system | |
| CN109209361A (en) | A kind of Fractured extra-low permeability oil reservoirs formation parameter prediction technique | |
| RU2331761C1 (en) | Low-permeable oil reservoir development method | |
| Chen et al. | A flow feature clustering-assisted uncertainty analysis workflow for optimal well rates in waterflood projects | |
| Chen et al. | Optimization of production performance in a CO2 flooding reservoir under uncertainty | |
| CN108561129A (en) | Craven fault block reservoirs water to oil area remaining oil fast appraisement method | |
| Xie et al. | Numerical simulation of oil recovery after cross-linked polymer flooding | |
| Shelley et al. | Refork Completion Analysis with the Aid of Artificial Neural Networks | |
| Elmabrouk et al. | Reservoir management strategies for development of water injection planning project | |
| CN115248176A (en) | Shale gas three-dimensional development effect evaluation experiment system and method | |
| Zhan et al. | Deep Learning for Production Main Controlling Factors Analysis of Tight Oil | |
| RU2556649C1 (en) | Method of determining dynamics of extraction of hard to recover reserves of oil | |
| Liu et al. | Water-Flooding Incremental Oil Recovery Study in Middle Miocene to Paleocene Reservoirs, Deep-Water Gulf of Mexico | |
| CN114718529B (en) | Method for dividing boundary of separate-layer water injection of oil reservoir | |
| Rafiei et al. | A New Analytical Approach for Water Allocation Management by Determining Inter-well Connectivity: North Sea Case Study |
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 |