CN106934075A - Drilling fluid density determines that method and static equal yield density determine method - Google Patents

Abstract

The determination method of drilling fluid density and static equal yield density determine method in a kind of pit shaft, and the determination method of drilling fluid density includes in the pit shaft：Step one, acquisition drilling fluid initial density ρ₀, components of drilling liquid parameter and each analysis site vertical depth data and temperature data, and to drilling fluid initial density ρ₀It is normalized with component data, obtains the first normalization data collection；Step 2, according to drilling fluid density ρ_i, pressure P_iAnd i-th analysis site and i+1 analysis site vertical depth data, calculate pressure P_i+1；Step 3, according to pressure P_i+1With temperature data T_i+1, calculate pressure differential Δ P_i+1With temperature difference Δ T_i+1；Step 4, to pressure differential Δ P_i+1With temperature difference Δ T_i+1It is normalized, and the second normalization data collection is obtained with reference to the first normalization data collection；Step 5, according to the second normalization data collection, using default neural network model, obtain the drilling fluid density ρ of i+1 analysis site_i+1.The method can avoid substantial amounts of repetition experiment and data analysis, substantially increase the computational accuracy and efficiency of HTHP well shaft ESD.

Description

Drilling fluid density determines that method and static equal yield density determine method

Technical field

The present invention relates to oil-gas exploration and development technical field, specifically, be related to a kind of drilling fluid density determine method and Static equal yield density determines method.

Background technology

As China's oil-gas exploration and development is progressively marched to deep formation, for high temperature high pressure deep well, the exploitation of ultradeep well It is on the increase therewith, this also causes narrow ' Safe Density Windows drilling problem to be widely present, and HTHP well shaft static state equivalent Density (Equivalent Static Density, referred to as ESD) is accurately to calculate bottom pressure, ensure the base of safety drilling Plinth.

One of accurate basis for calculating HTHP well shaft ESD is that drilling fluid density under HTHP is entered exactly Row prediction.Domestic and international many scholars have carried out corresponding research to drilling fluid density with HTHP change, and give many not Same forecast model, being summed up mainly has experience formula computational methods and drilling fluid component computational methods.

1972, Methven et al. to oil base drilling fluid, studied at high temperature under high pressure by performance, and it is to two groups of drilling wells Liquid sample has carried out HTHP density test, has obtained substantial amounts of PVT data, and depict brill under three kinds of different geothermal gradients Well liquid density curve such that it is able to so that very easily obtaining any point drilling fluid density value in pit shaft.Nineteen eighty-two, Hoberock Et al. the HTHP Forecasting Model of Density based on drilling fluid composition is established according to phase balance principle, i.e. component method prediction is bored Well liquid density model.1987, Kemp et al. thought that the model Consideration that forefathers provide is less, established consideration more fully Drilling fluid component method calculate HTHP under the quiet density model of drilling fluid.

The beginning of the nineties in last century, Yan Jienian is taught on the basis of experimental study, based on Multiple Non-linear Regression Analysis side Method establishes HTHP oil base drilling fluid Forecasting Model of Density.But because its coefficient is excessive, recurrence gets up increasingly complex, while It is likely to result in a certain number of coefficients level too small, calculation error can be caused on the contrary.

Thereafter many domestic scholars have done further analysis to drilling fluid density computation model under HTHP, or according to There is experimental data or by high temperature and pressure experiment data, a series of models are established based on drilling fluid empirical formula method.These moulds Type is all similar, is mostly that the exponential form according to Data Management Analysis on basic experience predictor formula has done certain change Change.

But, with actual well drilled engineering, drilling fluid system used becomes increasingly complex, it is difficult to based on a certain drilling fluid The empirical model that system is obtained predicts the drilling fluid density at high temperature under high pressure of all drilling fluids, and this has also just had influence on well The computational accuracy of cylinder ESD, therefore existing method can not meet safety drilling needs.

Based on above-mentioned situation, a kind of method for being capable of accurate pit shaft static state equal yield density is needed badly.

The content of the invention

To solve the above problems, the invention provides a kind of determination method of drilling fluid density in pit shaft, methods described bag Include：

Step one, the drilling fluid initial density ρ for obtaining well to be analyzed₀, components of drilling liquid parameter and each analysis site hang down Deep data and temperature data, and to the drilling fluid initial density ρ₀It is normalized with component data, obtains first and return One changes data set；

Step 2, the drilling fluid density ρ according to the i-th analysis site_i, pressure P_iAnd i-th analysis site and i+1 analysis site Vertical depth data, calculate the pressure P of i+1 analysis site_i+1, wherein, i >=1；

Step 3, the pressure P according to the i+1 analysis site_i+1With temperature data T_i+1, calculate the i+1 analysis site Pressure differential Δ P_i+1With temperature difference Δ T_i+1；

Step 4, to the pressure differential Δ P_i+1With temperature difference Δ T_i+1It is normalized, and returns with reference to described first One change data set obtains the second normalization data collection of the i+1 analysis site；

Step 5, according to the second normalization data collection, using default neural network model, obtain the i+1 point Analyse the drilling fluid density ρ of point_i+1；

Repeating said steps two obtain the drilling fluid density at each analysis site to step 5.

According to one embodiment of present invention, in the step one：

Formation parameter, casing programme parameter, drilling tool structure parameter and drilling fluid according to the Jing Chu to be analyzed for getting Component parameter, determines the Wellbore Temperature Field of the well to be analyzed, obtains the well bore temperature distribution of the well to be analyzed；

According to the well bore temperature distribution, the pit shaft is divided into some sections, correspondence obtains some analysis sites.

According to one embodiment of present invention, the components of drilling liquid parameter includes：Oil phase volume fraction f_oWith water phase body Fraction f_w。

According to one embodiment of present invention, the pressure differential Δ P of the i-th analysis site is determined according to following expression_iAnd temperature Difference Δ T_i：

ΔP_i=P_i-P₀

ΔT_i=T_i-T₀

Wherein, P₀And T₀Initial liquid column hydrostatic pressure and initial temperature are represented respectively.

According to one embodiment of present invention, according to following expression to the temperature difference Δ T of the i-th analysis site_iCarry out normalizing Change is processed：

Wherein, Δ T '_iRepresent temperature difference Δ T_iCorresponding normalized temperature is poor, Δ T_maxWith Δ T_minTemperature is represented respectively Poor maximum and minimum value.

According to one embodiment of present invention, the default neural network model builds obtain as follows：

Neural network topology structure determines step, according to drilling fluid density influence factor, determines input layer in neutral net With the neuronal quantity of output layer, the neuronal quantity for determining hidden layer in the neutral net is trained by iterative network, obtained Neural network topology structure；

Neural metwork training step, obtains the historical test data relevant with drilling fluid density, and the historical data is entered Row normalization, and by normalization after the historical data be input to the neural network topology structure in come to the nerve net Network topological structure is trained, and obtains optimal network parameter, so as to build obtain the default neural network model.

According to one embodiment of present invention, the neuron for determining hidden layer in the neutral net is trained by iterative network During quantity, iteration ranges are determined according to following expression：

Wherein, m represents the neuronal quantity of neutral net hidden layer, and k and l represents the neuron of input layer and output layer respectively Quantity, α represents tentative calculation parameter.

According to one embodiment of present invention, the pressure P of the i+1 analysis site is calculated according to following expression_i+1：

P_i+1=P_i+ρ_igH_i

Wherein, g represents acceleration of gravity, H_iRepresent the vertical depth between i+1 analysis site and the i-th analysis site.

Present invention also offers a kind of determination method of static equal yield density in pit shaft, methods described includes：

Drilling fluid density determines step, and the brill of each analysis site in the pit shaft is determined using the method described in as above any one Well liquid density；

Static equal yield density determines step, according to the drilling fluid density of each point, determines each analysis site in the pit shaft Static equal yield density.

According to one embodiment of present invention, the static equivalent of each analysis site in the pit shaft is determined according to following expression Density：

Wherein, ESD_i+1The static equal yield density of i+1 analysis site in pit shaft is represented, WHP represents casing pressure, ρ_jRepresent The drilling fluid density of jth analysis site, H_jThe vertical depth between the analysis site of jth+1 and jth analysis site is represented, g represents that gravity accelerates Degree, n represents the sum of analysis site in pit shaft.

The present invention relates to the calculating of drilling fluid density and equivalent precision in pit shaft in HTHP drilling process, it includes Based on the neural network model method for designing that drilling fluid constituent is set up, and obtained by the network parameter computational methods for being given Neural network parameter.The method calculates temperature in wellbore always according to formation parameter, casing programme parameter and drilling tool structure parameter , detailed segmentation is carried out to pit shaft, each section of drilling fluid density is calculated with the neural network model set up, finally calculate The quiet density profile of equivalent in whole pit shaft.The method can avoid substantial amounts of repetition experiment and data analysis, substantially increase The computational accuracy and efficiency of HTHP well shaft ESD.

Other features and advantages of the present invention will be illustrated in the following description, also, the partly change from specification Obtain it is clear that or being understood by implementing the present invention.The purpose of the present invention and other advantages can be by specification, rights Specifically noted structure is realized and obtained in claim and accompanying drawing.

Brief description of the drawings

In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing There is the accompanying drawing wanted needed for technology description to do simple introduction：

Fig. 1 is the flow chart for building default neutral net according to an embodiment of the invention；

Fig. 2 is neural network topology structure schematic diagram according to an embodiment of the invention；

Fig. 3 is the flow chart of the method for determination pit shaft static state equal yield density according to an embodiment of the invention；

Fig. 4 is detailed segmentation illustraton of model in pit shaft according to an embodiment of the invention；

Fig. 5 is that drilling fluid density and ESD sections result of calculation are illustrated in certain well shaft according to an embodiment of the invention Figure.

Specific embodiment

Describe embodiments of the present invention in detail below with reference to drawings and Examples, how the present invention is applied whereby Technological means solves technical problem, and reaches the implementation process of technique effect and can fully understand and implement according to this.Need explanation As long as not constituting conflict, each embodiment in the present invention and each feature in each embodiment can be combined with each other, The technical scheme for being formed is within protection scope of the present invention.

Meanwhile, in the following description, many details are elaborated for illustrative purposes, to provide to of the invention real Apply the thorough understanding of example.It will be apparent, however, to one skilled in the art, that the present invention can be without tool here Body details or described ad hoc fashion are implemented.

In addition, can be in the such as one group department of computer science of computer executable instructions the step of the flow of accompanying drawing is illustrated Performed in system, and, although logical order is shown in flow charts, but in some cases, can be with different from herein Order perform shown or described step.

In order to overcome prior art defect, present invention introduces based on particle swarm optimization algorithm (Paritcle Swarm Optimization, PSO) BP neural network, it is considered to drilling fluid composition factor sets up a set of improved BP High-temperature high-pressure drilling fluid density is combined Forecasting Methodology, and the Forecasting Methodology has drilling fluid constituent and PSO-BP neutral nets Well-bound feature, therefore precision of prediction is higher, Forecasting Methodology is more reasonable.And then predicted in pit shaft according to the Forecasting Methodology Drilling fluid density change under different temperatures pressure, calculates in pit shaft liquid column hydrostatic pressure at each point, finally establishes a set of based on changing Enter the HTHP well shaft ESD computational methods of BP neural network.

The accurate of pit shaft (particularly HTHP pit shaft) static state equal yield density is realized in order to overcome prior art defect Calculate, it is to be analyzed that pit shaft static state equal yield density provided by the present invention determines that method determines first with default neural network model Drilling fluid density in well shaft at each analysis site, then recycles the drilling fluid density that an analysis is pointed out to determine each analysis site The static equal yield density at place.

Required for determining that the drilling fluid density in well shaft to be analyzed at each analysis site, the present embodiment are constructed first Neural network model (preset neural network model).Wherein, the neural network model constructed by the present embodiment is preferably BP neural network based on particle cluster algorithm (Partical Swarm Optimizaiton, referred to as PSO).

Neutral net is to simulate the working mechanism of human brain from physiological structure, its have very strong approximation capability, from Adaptability and self-learning capability.Wherein, BP (Back Propagation) neutral net is the god being most widely used at present Through network.

BP (Back Propagation) neutral net is a kind of multilayer feedforward neural network, and the network is mainly characterized by To transmission, error back propagation before signal.In theory, BP neural network can approach the non-linear relation of arbitrary continuation.When When, because the error cost function of BP neural network is a complicated function of all connection weights, therefore, BP neural network exists Will necessarily there is a problem of that optimizing parameter is excessive during optimization weights, and optimizing parameter will excessively cause convergence rate mistake Slowly.Meanwhile, BP neural network also has that error cost complicated function there may be multiple extreme points, and the problem will likely Searching process is caused to be absorbed in local extremum.

Particle cluster algorithm (Partical Swarm Optimizaiton, referred to as PSO) is obtained from birds predation problem Enlighten and for solving optimization problem.In PSO, the solution of each problem to be optimized is defined as a bird in space, Referred to as " particle ".All particles have an adaptive value determined by problem to be optimized, and each particle also has a speed to determine The direction of fixed its search and distance.PSO initializes the random ion of such a group (i.e. RANDOM SOLUTION) first, is then found by iteration Optimal solution.In iterative process each time, particle updates oneself by tracking two " extreme values ".Wherein, the first extreme value is exactly The optimal solution that particle is found, another extreme value in itself is then the optimal solution that up to the present whole particle colony is found. PSO-BP neutral nets can effectively solve the problem that the convergence rate slow and searching process excessively existing for traditional BP neural network is easy It is absorbed in the problem of local extremum.

It is pointed out that in other embodiments of the invention, god can also be built using other rational methods Through network, the invention is not restricted to this.For example in an embodiment of the present invention, constructed neutral net can also be based on optimization Yield value come realize optimize neuronal transmission function PSO-GainBP neutral nets.

Fig. 1 shows the flow chart of structure PSO-BP networks in the present embodiment.

As shown in figure 1, in the present embodiment, according to drilling fluid density influence factor first in step S101, it is determined that refreshing Through network topology structure.In order to improve computational efficiency and prevent model from dissipating, the PSO-BP neutral nets constructed by the present embodiment Employ 3 layer network structures, i.e., 1 network input layer, 1 hidden layer and 1 output layer.

The neuronal quantity of input layer and output layer is determined first.In view of drilling fluid actual conditions, drilling fluid is by liquid Phase, solid phase two parts composition, liquid phase are pressurized and compress, expansion is influenceed by high temperature, and solid phase is kept constant by temperature, pressure volume, because This calculates main consideration temperature and pressure for drilling fluid density under HTHP influences on the liquid phase in drilling fluid.

Because liquid phase is mainly made up of water phase and oil phase in drilling fluid, and influence of the temperature, pressure to water phase and oil phase have compared with Big difference, it is therefore desirable to consider influence of the HTHP to water phase and oil phase respectively.Therefore the present embodiment setting input layer has 5 Input neuron (i.e. k=5) and 1 thresholded neuron, this 5 input neuron difference corresponding pressure difference Δ P, temperature difference Δ T, Drilling fluid density (i.e. drilling fluid initial density) ρ under normal temperature and pressure₀, water phase volume fraction f_wWith oil phase volume fraction f_o。

The neutral net mainly for being predicted to drilling fluid density (especially drilling fluid density under HTHP), because The neuronal quantity that this neural network opens up middle output layer is 1 (i.e. l=1), neuron correspondence drilling fluid density predicted value.

In the present embodiment, the neuronal quantity of hidden layer in neutral net is determined by iterative network training.Wherein, iteration Scope can be calculated according to following expression：

Wherein, m represents the neuronal quantity of hidden layer, and k represents the neuronal quantity of input layer, and l represents the nerve of output layer First quantity, α represents tentative calculation parameter.

In the present embodiment, the value of the neuronal quantity k of input layer is 5, and the value of the neuronal quantity l of output layer is 1, The value of tentative calculation parameter alpha is 1 to 10, is understood according to expression formula (1), and the span of the neuronal quantity m of hidden layer is 4 to 13.

Trained by iterative network, the neuronal quantity of hidden layer is preferably 5 in the neutral net constructed by the present embodiment It is individual, thus also just build and obtained PSO-BP neural network topology structures as shown in Figure 2.Wherein, W_ijExpression hidden layer jth (1≤ J≤m) node (i.e. neuron) is to the weights between input layer i-th (1≤i≤k) node, V_jlRepresent output layer l nodes (i.e. Neuron) to the weights between hidden layer jth node, θ_jRepresent the threshold value of hidden layer jth node, θ_lRepresent the threshold of output layer l nodes Value.

Again as shown in figure 1, after the topological structure for determining PSO-BP neural network, the present embodiment is in step s 102 The historical test data relevant with drilling fluid density is obtained, and these historical test datas are normalized.In the present embodiment In, the acquired historical test data relevant with drilling fluid density includes：Drilling fluid initial density ρ₀, initial temperature T₀, it is initial Pressure P₀, different test experiments drilling fluid pressure P_iWith suffered temperature T_i, in pressure P_iWith temperature T_iWhen drilling well it is liquid-tight Degree ρ_i, water phase volume fraction f_wAnd oil phase volume fraction f_o。

After obtaining above-mentioned historical data, measuring point drilling fluid density ρ is calculated_iCorresponding pressure differential Δ P_iWith temperature difference Δ T_i, I.e.：

ΔP_i=P_i-P₀ (2)

ΔT_i=T_i-T₀ (3)

Wherein, drilling fluid initial density ρ₀, pressure differential Δ P_i, temperature difference Δ T_i, water phase volume fraction f_wAnd oil phase volume Fraction f_oIt is input layer parameter, ρ_iIt is output layer parameter.

In the present embodiment, it is normalized by above-mentioned input layer parameter and output layer parameter so that Suo Youcan It is interval that number is normalized to [- 1,1].Specifically, in the present embodiment according to following expression to above-mentioned input layer parameter and output layer Parameter is normalized：

Wherein, x_midThe intermediate value of certain parameter, x_maxAnd x_minThe maximum and minimum value of certain parameter are represented respectively, and x and x ' is respectively Represent the value before certain parameter normalization and after normalization.

X can represent drilling fluid initial density ρ₀, pressure differential Δ P_i, temperature difference Δ T_i, water phase volume fraction f_w, oil phase body Fraction f_oWith measuring point drilling fluid density ρ_i.For example for temperature difference Δ T_iFor, it can be returned according to following expression One change is processed：

Wherein, Δ T ' represents temperature difference Δ T_iCorresponding normalized temperature is poor, Δ T_maxWith Δ T_minRepresent respectively and counted Maximum and minimum value in each temperature difference data for obtaining.

Other specification can also be normalized using identical principle, be will not be repeated here.

After being normalized to historical data, the historical data after normalization is input to step in step s 103 The neutral net topology is trained in neural network topology structure constructed by S101, so as to obtain neutral net most There is network parameter, that is, obtain required neural network model.

After obtaining above-mentioned neural network model, just can using the neural network model come to each pressure and at a temperature of Drilling fluid density is predicted, and Fig. 3 to show and determine drilling fluid density and static equal yield density using the neural network model Flow chart.

As shown in figure 3, formation parameter, the casing programme of the present embodiment well to be analyzed that basis gets in step S301 Parameter, drilling tool parameter and components of drilling liquid parameter, determine the Wellbore Temperature Field of well to be analyzed, so as to obtain the well of well to be analyzed Cylinder thermo parameters method.

In the present embodiment, acquired components of drilling liquid parameter in step S301 includes oil phase volume fraction f_oWith water phase Volume fraction f_w.Wherein, in step S301, the drilling fluid normal temperature and pressure of the to be analyzed well relevant with drilling fluid density is also obtained (i.e. initial pressure P₀With initial temperature T₀) under density (i.e. initial density) ρ₀, and also to drilling fluid initial density ρ₀And Components of drilling liquid parameter is normalized, and has obtained the first normalization data collection.

After well bore temperature distribution is obtained, in step s 302 according to well bore temperature distribution, the pit shaft for being analysed to well is drawn It is divided into some sections, correspondence obtains some analysis sites.After obtaining these analysis sites, the method that the present embodiment is provided is just in step The vertical depth data and temperature data of each analysis site are obtained in S303.

Fig. 4 shows in the present embodiment detailed segmentation model in pit shaft.Figure 4, it is seen that the present embodiment will be treated point Analysis well shaft has been divided into n sections, and its vertical depth is respectively H₁、H₂、…H_n.Wherein, H_i-1Represent the i-th analysis site and the i-th -1 analysis site Between vertical depth, H_iRepresent the vertical depth between i+1 analysis site and the i-th analysis site.For example, H₁Represent the 2nd analysis site and Vertical depth between 1 analysis site (analysis site i.e. at well head).

In the present embodiment, for the 1st analysis site, its pressure P₁As casing pressure WHP, its suffered temperature is T₁, its drilling fluid density is drilling fluid density ρ₁.And for the i-th analysis site, its pressure is pressure P_i, suffered temperature Degree is temperature T_i, drilling fluid density is drilling fluid density ρ_i。

Be can be seen that by above-mentioned set up neural network model, if it is desired to ask for the drilling fluid density of the i-th analysis site ρ_i, it is necessary to determine the pressure P of the i-th analysis site_iAnd temperature T_i.And the temperature T of the i-th analysis site_iThe basis in step S301 Well bore temperature distribution can determine to obtain, thus now it needs to be determined that namely the i-th analysis site pressure P_i, and the i-th analysis The pressure P of point_iCan be by the casing pressure WHP and liquid column hydrostatic pressure P of the position_{i_s}.Casing pressure WHP is, it is known that then only need to meter The liquid column hydrostatic pressure that calculation is caused by drilling fluid density.

Because with the increase of analysis site vertical depth, its drilling fluid density is also change.Therefore in order to ensure resulting i The pressure P of analysis site_iAccuracy, the method that the present embodiment is provided employs the mode of iteration to determine each analysis site Pressure, that is, be present：

P_i+1=WHP+P_{i+1_s}=P_i+ρ_igH_i (8)

Wherein, P_i+1And P_iThe pressure of i+1 analysis site and the i-th analysis site is represented respectively, and g represents acceleration of gravity, H_iTable Show the vertical depth between i+1 analysis site and the i-th analysis site.

For example, for the 2nd analysis site, its pressure P₂Can be calculated according to following expression：

P₂=P₁+ρ₁gH₁ (9)

The rest may be inferred, and the drilling well of present analysis point just can be so determined using the drilling fluid density of previous analysis site Liquid-tight degree.

It is pointed out that for the 1st analysis site, its described pressure is known casing pressure WHP.

Specifically, as shown in figure 3, parameter i first is set into 1 in step s 302, i.e., initial analysis site is the 1st analysis site. In step s 304, according to the pressure P of the 1st analysis site₁(P₁Equal to casing pressure WHP) and temperature T₁, determine the 1st analysis site Drilling fluid density ρ₁。

Specifically, in step s 304, due to the temperature T of the 1st analysis site₁Equal to initial temperature T₀, so if the 1st point Analyse the pressure P of point₁(i.e. casing pressure WHP) is equal to 0, then the drilling fluid density ρ of the 1st analysis site₁Also it is equal to drilling fluid initial Density p₀.And if the pressure P of the 1st analysis site₁(i.e. casing pressure WHP) is not equal to 0, then will be counted first in step S303 Calculate the pressure differential Δ P of the 1st analysis site₁With temperature difference Δ T₁, then to pressure differential Δ P₁With temperature difference Δ T₁It is normalized The drilling fluid density ρ of the 1st analysis site is determined using the data after normalization and default neural network model afterwards₁。

Specifically, the method that the present embodiment is provided calculates i+1 analysis site in step S305 according to following expression The pressure P at place_i+1。

P_i+1=P_i+ρ_igH_i (10)

Wherein, g represents acceleration of gravity, H_iThe vertical depth between i+1 analysis site and the i-th analysis site is represented, that is, is existed：

H_i=h_i+1-h_i(i=1 ..., n) (11)

Obtain the pressure P at the i-th analysis site_iAfterwards, the pressure of the i-th analysis site just can be respectively calculated in step S306 Power difference Δ P_iWith temperature difference Δ T_i.Specifically, in the present embodiment, the pressure of the i-th analysis site is calculated advantageously according to following expression Power difference Δ P_iWith temperature difference Δ T_i：

ΔP_i=P_i-P₀ (12)

ΔT_i=T_i-T₀ (13)

In the present embodiment, to the pressure differential Δ P of the i-th analysis site in step S307_iWith temperature difference Δ T_iIt is normalized, And gather resulting the first normalization data collection in step S301 and define the second normalization data collection.Wherein, second returns One change data are to include drilling fluid initial density, oil phase volume fraction, water phase volume fraction, the i-th analysis site after normalization Pressure differential Δ P_iWith temperature difference Δ T_i。

As shown in figure 3, be input into for above-mentioned second normalization data collection in step S308 by the method that the present embodiment is provided To the advance neutral net for building, the drilling fluid density of the i-th analysis site just can be so determined according to the output of neutral net ρ_i。

Because now i is equal to 1, therefore calculates in step S305 to step S308 is the relevant parameter of the 2nd analysis site, is walked Resulting drilling fluid density is the drilling fluid density ρ of the 2nd analysis site in rapid S308₂。

Specifically, the drilling fluid density ρ of the 1st analysis site is obtained in step s 304₁Afterwards, in step S305, just can be with According to the pressure P of the 1st analysis site₁, drilling fluid density ρ₁And the 2nd analysis site and the 1st analysis site vertical depth difference H₁Using table The pressure P of the 2nd analysis site is calculated up to formula (8)₂。

Then in step S306, according to the pressure P of the 2nd analysis site₂With temperature T₂The pressure of the 2nd analysis site can be calculated Power difference Δ P₂With temperature difference Δ T₂, and to pressure differential Δ P in step S307₂With temperature difference Δ T₂It is normalized.To return The default neural network model of data input after one change, so also can be obtained by the drilling fluid density ρ of the 2nd analysis site₂。

In step S309, the method will determine that whether reaching expectation analysis site (for example expects to obtain its drilling fluid density Analysis site).If expectation analysis site (i.e. the i-th analysis site is to expect analysis site) has been reached, according to each in step S310 The drilling fluid density and vertical depth data of individual analysis site, determine the static equal yield density for expecting analysis site；If be not reaching to Expect analysis site, then make return to step S305 after i=i+1, to obtain the drilling well of i+1 analysis site (being now the 3rd analysis site) Liquid-tight degree, and the process is repeated until reaching expectation analysis site.

Specifically, in the present embodiment, the static equal yield density of each analysis site in pit shaft is calculated according to following expression：

Wherein, ESD_i+1Represent the static equal yield density of i+1 analysis site in pit shaft, ρ_jRepresent the drilling fluid of jth analysis site Density, H_jThe vertical depth between the analysis site of jth+1 and jth analysis site is represented, g represents acceleration of gravity, and n represents analysis in pit shaft The sum of point.

As can be seen that in order to overcome prior art defect, the method that the present embodiment is provided introduces base from foregoing description In the BP neural network of particle swarm optimization algorithm (Paritcle Swarm Optimization, PSO), it is considered to which drilling fluid is constituted Composition factor establishes the compound Forecasting Methodology of high-temperature high-pressure drilling fluid density of a set of improved BP, Forecasting Methodology tool Have drilling fluid constituent and the well-bound feature of PSO-BP neutral nets, thus precision of prediction is higher, Forecasting Methodology more Adduction is managed.And then drilling fluid density change under different temperatures pressure in pit shaft is predicted according to the Forecasting Methodology, calculate each in pit shaft Liquid column hydrostatic pressure at point, finally establishes a set of HTHP well shaft ESD computational methods based on improved BP.

The present embodiment is using one group of water-base drilling fluid and two groups of oil base drilling fluid measurement data as training sample to nerve Network topology structure is trained, so as to obtain the network parameter of neutral net topology.In the present embodiment, No.1 water-based drillings Liquid includes into composition：350g water, 20g bentonites, 3g CLSs and 3g lignite sodium hydroxide solutions, then use 4.22g/cm³Barite is aggravated to drilling fluid, is each configured to density for 1.28g/cm³Three kinds of water-base drilling fluid samples.

Oil base drilling fluid composition includes：2# diesel oil 298cm³、52cm³Concentration is 30% calcium chloride water, 5g is organic Soil, 5g emulsifying agents and 2g white limes, equally use proportion to be aggravated to drilling fluid for 4.22 barite, are respectively configured Density is respectively 1.31g/cm³No.2 oil base drilling fluids and 2.15g/cm³No.3 oil base drilling fluids.

In the present embodiment, the component data of above-mentioned three kinds of drilling fluids are as shown in table 1.

Table 1

And three kinds of drilling fluid density data are specific such as table under the conditions of 21.11-204.44 DEG C of temperature, pressure 0-96.46MPa Shown in 2.

Table 2

Then resulting data (i.e. training data input/output argument) are normalized.Specifically, this reality Apply example to be normalized the data shown in Tables 1 and 2 using expression formula (4) and expression formula (5), all data are returned One changes between [- 1,1].Wherein, the structure obtained by normalized is distinguished as shown in Table 3 and Table 4.

Table 3

Table 4

Then, flow is trained to be trained neutral net according to network parameter according to normalization data obtained as above, Optimum network structure is drawn as shown in Fig. 2 and the network parameter determined is as shown in table 5.

Table 5

After the training for completing neutral net, the prediction of drilling fluid density just can be carried out using the neutral net for training .In the present embodiment, certain well uses 1.90g/cm³The oil base drilling fluid of (20 DEG C, 0MPa) is crept into, and geothermal gradient is 18.1 DEG C/km, synthetic wells liquid constituent, oil phase volume fraction is 58.4%, and water phase volume fraction is 9.068%, wellhead casing pipe Pressure WHP is 0, shown in other data such as table 6 (i.e. wellbore construction tables of data) and table 7 (i.e. drilling tool structure tables of data).

Table 6

Table 7

Drilling tool type	Drilling tool explanation	External diameter (cm)	Internal diameter (cm)	Length (m)
					Drilling rod	5”	12.70	10.86	5931.7
Drilling rod	4”	10.16	8.48	1154.0
					Heavy weight drill pipe	4”	10.16	6.51	141.0
Drill collar	5”	12.70	5.72	198.0
					Drill bit *	PDC	16.83	-	-

When the static equal yield density of each analysis site in pit shaft is calculated, Wellbore Temperature Field is calculated according to given data first, The drilling fluid density in pit shaft at each analysis site is then calculated according to the neural network model for building, is finally calculated at each point successively Static equal yield density, final calculation result is as shown in Figure 5.

The present invention relates to the calculating of drilling fluid density and equivalent precision in pit shaft in HTHP drilling process, it includes Based on the neural network model method for designing that drilling fluid constituent is set up, and obtained by the network parameter computational methods for being given Neural network parameter.The method calculates temperature in wellbore always according to formation parameter, casing programme parameter and drilling tool structure parameter , detailed segmentation is carried out to pit shaft, each section of drilling fluid density is calculated with the neural network model set up, finally calculate The quiet density profile of equivalent in whole pit shaft.The method can avoid substantial amounts of repetition experiment and data analysis, substantially increase The computational accuracy and efficiency of HTHP well shaft ESD.

" one embodiment " or " embodiment " mentioned in specification means special characteristic, the structure for describing in conjunction with the embodiments Or characteristic is included at least one embodiment of the present invention.Therefore, the phrase " reality that specification various places throughout occurs Apply example " or " embodiment " same embodiment might not be referred both to.

Although above-mentioned example is used to illustrate principle of the present invention in one or more applications, for the technology of this area For personnel, in the case of without departing substantially from principle of the invention and thought, hence it is evident that can in form, the details of usage and implementation It is upper various modifications may be made and without paying creative work.Therefore, the present invention is defined by the appended claims.

Claims (10)

1. in a kind of pit shaft drilling fluid density determination method, it is characterised in that methods described includes：

Step one, the drilling fluid initial density ρ for obtaining well to be analyzed₀, components of drilling liquid parameter and each analysis site vertical depth data And temperature data, and to the drilling fluid initial density ρ₀It is normalized with component data, obtains the first normalization number According to collection；

Step 2, the drilling fluid density ρ according to the i-th analysis site_i, pressure P_iAnd i-th analysis site and i+1 analysis site vertical depth Data, calculate the pressure P of i+1 analysis site_i+1, wherein, i >=1；

Step 3, the pressure P according to the i+1 analysis site_i+1With temperature data T_i+1, calculate the pressure of the i+1 analysis site Power difference Δ P_i+1With temperature difference Δ T_i+1；

Step 4, to the pressure differential Δ P_i+1With temperature difference Δ T_i+1It is normalized, and combines described first and normalizes Data set obtains the second normalization data collection of the i+1 analysis site；

Step 5, according to the second normalization data collection, using default neural network model, obtain the i+1 analysis site Drilling fluid density ρ_i+1；

Repeating said steps two obtain the drilling fluid density at each analysis site to step 5.

2. the method for claim 1, it is characterised in that in the step one：

Formation parameter, casing programme parameter, drilling tool structure parameter and components of drilling liquid according to the Jing Chu to be analyzed for getting Parameter, determines the Wellbore Temperature Field of the well to be analyzed, obtains the well bore temperature distribution of the well to be analyzed；

According to the well bore temperature distribution, the pit shaft is divided into some sections, correspondence obtains some analysis sites.

3. method as claimed in claim 2, it is characterised in that the components of drilling liquid parameter includes：Oil phase volume fraction f_oWith Water phase volume fraction f_w。

4. the method as any one of claims 1 to 3, it is characterised in that the i-th analysis site is determined according to following expression Pressure differential Δ P_iWith temperature difference Δ T_i：

ΔP_i=P_i-P₀

ΔT_i=T_i-T₀

Wherein, P₀And T₀Initial liquid column hydrostatic pressure and initial temperature are represented respectively.

5. the method as any one of Claims 1 to 4, it is characterised in that according to following expression to the i-th analysis site Temperature difference Δ T_iIt is normalized：

${\mathrm{\ΔT}}_{mid}=\frac{{\mathrm{\ΔT}}_{max}+{\mathrm{\ΔT}}_{min}}{2}$

${\mathrm{\ΔT}}_i^{\′}=\frac{2({\mathrm{\ΔT}}_i-{\mathrm{\ΔT}}_{mid})}{{\mathrm{\ΔT}}_{\max }-{\mathrm{\ΔT}}_{min}}$

Wherein, Δ T '_iRepresent temperature difference Δ T_iCorresponding normalized temperature is poor, Δ T_maxWith Δ T_minTemperature difference is represented respectively Maximum and minimum value.

6. the method as any one of Claims 1 to 5, it is characterised in that the default neural network model is to pass through Following steps build what is obtained：

Neural network topology structure determines step, according to drilling fluid density influence factor, determines in neutral net input layer and defeated Go out the neuronal quantity of layer, the neuronal quantity for determining hidden layer in the neutral net is trained by iterative network, obtain nerve Network topology structure；

Neural metwork training step, obtains the historical test data relevant with drilling fluid density, and the historical data is returned One change, and by normalization after the historical data be input in the neural network topology structure to open up the neutral net Flutter structure to be trained, obtain optimal network parameter, so as to build obtain the default neural network model.

7. method as claimed in claim 6, it is characterised in that trained by iterative network and determine hidden layer in the neutral net Neuronal quantity when, iteration ranges are determined according to following expression：

$m=\sqrt{k+l}+\mathrm{\α}$

Wherein, m represents the neuronal quantity of neutral net hidden layer, and k and l represents the neuron number of input layer and output layer respectively Amount, α represents tentative calculation parameter.

8. the method as any one of claim 1~7, it is characterised in that the i+1 is calculated according to following expression The pressure P of analysis site_i+1：

P_i+1=P_i+ρ_igH_i

Wherein, g represents acceleration of gravity, H_iRepresent the vertical depth between i+1 analysis site and the i-th analysis site.

9. in a kind of pit shaft static equal yield density determination method, it is characterised in that methods described includes：

Drilling fluid density determines step, using each point in pit shaft as described in the method determination any one of claim 1~8 Analyse the drilling fluid density of point；

Static equal yield density determines step, according to the drilling fluid density of each point, determines the quiet of each analysis site in the pit shaft State equal yield density.

10. method as claimed in claim 9, it is characterised in that each analysis site in the pit shaft is determined according to following expression Static equal yield density：

${\mathrm{ESD}}_{i+1}=\frac{WHP+\underset{j=1}{\overset{i}{\Σ}}{\mathrm{\ρ}}_j{\mathrm{gH}}_j}{g\underset{j=1}{\overset{n}{\Σ}}{H}_j}$

Wherein, ESD_i+1The static equal yield density of i+1 analysis site in pit shaft is represented, WHP represents casing pressure, ρ_jRepresent jth point Analyse the drilling fluid density of point, H_jThe vertical depth between the analysis site of jth+1 and jth analysis site is represented, g represents acceleration of gravity, n tables Show the sum of analysis site in pit shaft.