Oil field productivity monitoring method using carbon quantum dot tracer agent

Abstract

The invention provides an oil field productivity monitoring method using a carbon quantum dot tracer agent, and the method comprises the steps: S100, obtaining the convection term and diffusion term characteristics of an oil field through the injection of the carbon quantum dot tracer agent and an inert tracer agent, and building an oil field monitoring characteristic set; step S200, selecting main characteristics of oil field monitoring from the oil field monitoring characteristic set by using a Spearman correlation coefficient and a random forest tree model method; and S300, inputting the main features into a neural network to monitor the productivity of the oil field, wherein the neural network is a particle swarm optimization-back propagation neural network for Wiener process optimization.

Description

TECHNICAL FIELD

[0001] The present application relates to an oilfield monitoring technology, in particular to an oilfield productivity monitoring method using carbon quantum dot tracers. BACKGROUND

[0002] Accurate prediction of productivity is essential for efficient and economic development of oil and gas reservoirs. The traditional oilfield productivity prediction method is the empirical formula and numerical simulation method, which has fewer influencing factors and makes a lot of ideal assumptions on the assumption conditions when predicting productivity. There are many factors affecting oilfield productivity, and the interaction between them presents complexity. Continuing to use the traditional reservoir productivity prediction method may lead to significant calculation errors and may cause economic losses in oilfield development. The patent with publication number CN115977599A uses tracers to monitor oilfields, but the features used are relatively few, including interwell connectivity analysis results, tracer test results, pressure drop test results, and water absorption profile test results. The patent with publication number CN103958643A provides a carbon-based fluorescent tracer for detecting oil reservoirs, which provides the possibility for carbon quantum dot tracer monitoring of oilfields. However, the oilfield features obtained by carbon quantum dot tracers are numerous and complex, and not all features are suitable for oilfield monitoring for a specific oilfield. SUMMARY

[0003] The present application aims to provide an oilfield productivity monitoring method using carbon quantum dot tracers, comprising:

[0004] Step S100, obtaining the convection term and diffusion term features of the oilfield by injecting carbon quantum dot tracers and inert tracers, and establishing an oilfield monitoring feature set;

[0005] Step S200, selecting the main features of oilfield monitoring from the oilfield monitoring feature set using the method of Spearman correlation coefficient and random forest model;

[0006] Step S300, inputting the main features into a neural network to monitor the oilfield productivity, wherein the neural network is a particle swarm optimization-back propagation neural network optimized by a Wiener process.

[0007] Further, the convection term features in step S100 include {porosity, effective permeability, oil volume factor, oil saturation, reservoir pressure, clay volume, oil viscosity}; and the diffusion term features include {flowback proppant volume, proppant ratio, fracture cross section, fracture spacing}.

[0008] Further, the following parameters of the carbon quantum dot tracer and the inert tracer are recorded in step S100 to obtain a detection feature set of the oilfield: the first breakthrough time and the peak concentration time of the carbon quantum dot tracer, the first breakthrough time and the peak concentration time of the inert tracer, the total amount of the carbon quantum dot tracer injected and the total amount of the inert tracer injected, the total amount of the inert tracer injected and the total amount of the inert tracer injected, the main flow line length of the inert tracer between the injection well and the production well, and the flow cross-sectional area of the carbon quantum dot tracer.

[0009] Further, the specific process of step S200 includes:

[0010] In step S201, the monotonic relationship between each feature and the oilfield productivity is evaluated by the Spearman correlation coefficient.

[0011] In step S202, the importance of each feature to the oilfield productivity is determined by the random forest tree model.

[0012] In step S203, the main features for monitoring the oilfield are selected according to the monotonic relationship and the importance.

[0013] Further, in step S203, the features with strong positive correlation of the Spearman correlation coefficient and the top K correlation of the random forest tree model are selected as the main features for monitoring the oilfield.

[0014] Further, in step S300, the random disturbance d(f) of the feature porosity, effective permeability, oil saturation, reservoir pressure, oil viscosity, volume of proppant, fracture cross section, and fracture spacing is obtained in the neural network using the Wiener process.

[0015] d(f) = μ f dt+σ f dW f

[0016] Wherein, f represents the feature, μ f is the deterministic drift term of the feature f, σ f is the randomness parameter of the feature f, dW f is the Wiener increment.

[0017] Further, in the neural network, the particles are updated using the random disturbance d(f)

[0018] X t+Δt = X t +d(f)

[0019] Wherein, X t is the feature parameter value at time t, and Δt is the time step.

[0020] Compared with the prior art, the present application has the following advantages: (1) carbon quantum dot tracers and inert tracers are used to obtain oil field characteristics, and the inert tracers can make up for the problem of the adsorption of the carbon quantum dot tracers affecting the accuracy of the characteristics; (2) the most critical characteristics of the oil field are selected by comprehensively evaluating all the characteristics by using the Spearman correlation coefficient and the random forest tree, so as to reduce the operation speed of the subsequent model; and (3) the PSO is optimized by using the Wiener process, the local optimal deadlock of the PSO algorithm is broken by the random disturbance of the Wiener process, and the global search ability and adaptability of the PSO algorithm are improved.

[0021] The present application will be further described below in combination with the drawings of the specification. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 It is a method flowchart of the present application.

[0023] Figure 2 It is a heat map of the Spearman correlation coefficient analysis.

[0024] Figure 3 It is a WP-PSO flowchart.

[0025] Figure 4 It is a comparison diagram of the WP-PSO-BP and PSO-BP prediction model results. DETAILED DESCRIPTION

[0026] In combination Figure 1 An oil field monitoring method using carbon quantum dot tracers, mainly for monitoring the productivity of the oil field, the method uses carbon quantum dot tracers to obtain oil field characteristics, uses the random forest tree method to screen the characteristic factors that affect the productivity of the oil field, and uses the neural network of the Wiener-particle swarm optimization algorithm to establish a monitoring model. In the development of the oil field, according to the homogeneity of the reservoir, it can be divided into homogeneous reservoir (Homogeneous Reservoir) and heterogeneous reservoir (Heterogeneous Reservoir), for the homogeneous reservoir, when using carbon quantum dot tracers for detection, the error is caused by adsorption, which is not easy to cause large error; but for the heterogeneous reservoir, the carbon quantum dot tracers are also disturbed by the heterogeneous reservoir. The present embodiment mainly aims at monitoring the heterogeneous reservoir oil field by using carbon quantum dot tracers. The method involved in the present embodiment includes the following steps:

[0027] Step S100, collecting oil field characteristics by carbon quantum dot tracers, and constructing an oil field monitoring characteristic set;

[0028] Step S200, selecting the main characteristics of the oil field monitoring from the oil field monitoring characteristic set by using the method of the Spearman correlation coefficient and the random forest tree;

[0029] Step S300, input the required characteristic factors of the oilfield into the neural network for productivity detection, the neural network is an oilfield monitoring model established by using particle swarm optimization back propagation neural network, and the model is optimized by using Wiener process in the model establishment process.

[0030] The carbon quantum tracer has high sensitivity and long monitoring advantages when monitoring the oilfield. Although the adsorbability of the carbon quantum tracer is small, it may be affected by the adsorption of the formation minerals in some cases, which may cause the concentration of the tracer to change and affect the accuracy of the monitoring results. Therefore, in step S100, carbon quantum dot tracer (CODs) and inert tracer (such as NaBr solution) are added to the injection well of the oilfield. Due to the differences in molecular size, seepage capacity, adsorption capacity, and diffusion capacity, the inert tracer only reflects the fluid motion itself under ideal conditions, and the breakthrough time of the inert tracer is the pure fluid migration time. The time difference between the two can be used to infer the degree of carbon quantum dot retention. Especially in the case of heterogeneous reservoirs, the inert tracer is used as a control group to calculate the relevant characteristic parameters of the oilfield, and the carbon quantum tracer is used for correction to obtain the relevant characteristic parameters. The accuracy of the former is about 3 times higher than that of the latter. The error source of the former is only measurement noise, but the noise source of the latter is due to the adsorbability error of CQDs and the interference of heterogeneous reservoirs.

[0031] The characteristics to be obtained in step S100 include porosity (P), effective permeability (EP), oil volume factor (VF O ), oil saturation (S O ), reservoir pressure (P R ), clay volume (V C ), oil viscosity (V O ), flowback proppant volume (V P ), proppant ratio (R P ), fracture cross-section (S F ), fracture spacing (D F ). Among them, the oil viscosity (V O ) is not measured by the carbon quantum dot tracer, but is obtained by PVT measurement. Porosity (P), effective permeability (EP), oil volume factor (VF O ), oil saturation (S O ), reservoir pressure (P R ), clay volume (V C ), oil viscosity (V O ) are convection term characteristics, representing the process of fluid bulk motion carrying tracer migration; flowback proppant volume (V P ), proppant ratio (R P ), fracture cross-section (S F ), fracture spacing (D F) is the characteristic of the dispersion term, describing the spontaneous dispersion of the tracer due to random collisions.

[0032] Specifically, carbon quantum dot tracers (CODs) and inert tracers are added to the injection well of the oilfield at a concentration ratio of 1:1, fluid samples are collected periodically (<0.5h) in the production well, and the first breakthrough time t i,CQDs , t i,inert of CQDs and inert tracers are recorded respectively, the peak concentration time t p,CQDs , t p,inert of CQDs and inert tracers are recorded, the returned CQDs are quantitatively analyzed by a high-precision fluorescence spectrophotometer, the returned inert tracers are analyzed by a chromatograph, and the following characteristics of the oilfield are calculated:

[0033] (1) Porosity (P)

[0034] Porosity refers to the ratio of pore volume to total volume in rock. The larger the porosity, the more oil that can be stored in the reservoir, thereby providing a material basis for higher production capacity.

[0035]

[0036] Wherein, ρ in,CQDs is the injection flow rate of carbon quantum dot tracers, L inert is the main flow line length between injection and production wells of inert tracers, A is the flow cross-sectional area of carbon quantum dot tracers, i.e. the effective cross-sectional area of carbon quantum dot tracers flowing between injection and production wells; since the breakthrough time of inert tracers is purely fluid migration time, t i,inert is used to correct the breakthrough time of carbon quantum dot tracers, and L inert is used to determine the main flow line length between injection and production wells.

[0037] (2) Effective permeability (EP)

[0038] Effective permeability refers to the actual permeability of a certain phase fluid flowing in the reservoir. The higher the effective permeability, the stronger the flow ability of the fluid in the reservoir; effective permeability is irrelevant to carbon quantum dot tracer adsorption and is related to the true flow rate of oil. The effective permeability EP is obtained by the following formula:

[0039]

[0040] Wherein, μ is the fluid viscosity, is the average pressure gradient between injection and production wells.

[0041] (3) Oil volume factor (VF O )

[0042] The oil volume factor is the ratio of the volume of formation oil under standard conditions on the ground to the volume under formation conditions, which reflects the volume change of formation oil under different pressure and temperature conditions, and has a significant impact on productivity prediction.

[0043]

[0044] wherein A k is the well control cross-sectional area, Q O is the oil production of the production well, is the oil / water viscosity (test data), is the carbon quantum dots inversion value (obtained by laboratory calibration).

[0045] (4) Oil saturation (S O )

[0046] The oil saturation refers to the proportion of the volume of oil in the reservoir to the effective pore volume of the reservoir, and is an important parameter for measuring the reserves and flow capacity of oil in the reservoir.

[0047]

[0048] wherein R water is the resistivity under pure water conditions in laboratory calibration, γ is the core calibration index, α is the core coefficient, R t is the true resistivity of the formation.

[0049] (5) Reservoir pressure (P R )

[0050] The reservoir pressure refers to the pressure borne by the fluid in the reservoir, which reflects the storage state and flow capacity of the fluid in the reservoir.

[0051]

[0052] wherein ρ in,inert is the inert tracer injection flow rate, and h is the effective vertical thickness of the fluid flowing in the reservoir.

[0053] (6) Clay volume (V C )

[0054] The type, content, occurrence and physical properties of clay minerals have a great influence on the physical properties of the reservoir. The higher the content of clay minerals, the lower the porosity and permeability of sandstone, and the poorer the reservoir performance.

[0055]

[0056] wherein M in,CQDs and M out,CQDs are the total amount of CQDs injected and the total amount of flowback, respectively, k CQDsis the adsorption coefficient of CQDs on the surface of clay, p clay is the density of clay, V out,CQDs is the flowback volume of CQDs, C out,CQDs is the average concentration of CQDs in flowback fluid.

[0057] (7) The proppant flowback volume (V P )

[0058] During the flowback process, the flowback of proppants can cause the conductivity of artificial fractures to decrease. If the flowback speed is too high, proppants will flow back to the wellbore with the flowback fluid in large quantities, or even be brought to the ground surface, which not only reduces the conductivity of the fracture, but also can cause sand accumulation at the bottom of the well, burying the gas layer and affecting the oil and gas production.

[0059] In the laboratory, a "CQDs fluorescence intensity-proppant concentration" standard curve is established, which is obtained by measuring the fluorescence intensity of a known volume of labeled proppant sample in a simulated flowback fluid environment; the total proppant flowback volume V

[0060] V P =∑C·Q·Δt

[0061] Wherein, C is the proppant concentration obtained by the "CQDs fluorescence intensity-proppant concentration" standard curve in the sampling, Q is the flowback flow rate at the sampling time, and Δt is the sampling time interval.

[0062] (8) Proppant ratio (R P )

[0063] The proppant ratio refers to the ratio of the amount of proppants to the volume of the fracture. With the increase of the proppant ratio, the conductivity of the fracture will increase, and more proppants can more effectively maintain the open state of the fracture, reduce the closure of the fracture, and thus improve the flow capacity of the fluid.

[0064] The fracture volume V f

[0065]

[0066] Wherein, M in,inert , M out,inert are the total amount of inert tracer injection and the total amount of flowback, respectively, and C in,inert is the initial concentration of inert tracer in the fracture;

[0067] The volume of retained proppants V retained,prop

[0068] V retained,prop =V in,prop -V P

[0069] Among them, V in,prop is the initial injection volume of proppant;

[0070] Calculating proppant ratio

[0071] R P =V retained,prop / V f .

[0072] (9) Crack cross section (S F )

[0073] Fractures are the main channels for fluids, and the size of their cross-sectional area directly affects the flow velocity and flow rate of the fluid. A larger fracture cross-section can reduce the resistance to fluid flow, thereby increasing production capacity.

[0074] Through the crack top depth h top and bottom depth h bottom Calculate the crack height h f

[0075] h f =h top -h bottom

[0076] Calculate the crack cross section S F

[0077] S F =V f / h f .

[0078] (10) Crack spacing (D F )

[0079] The fracture spacing is the distance between adjacent fracture clusters in a horizontal well section. When the spacing is too large, the unfractured area becomes a mining blind area. When the spacing is too small, the fracture width is suppressed, resulting in sand plugging or uneven distribution of proppant. Therefore, it is necessary to calculate the maximum and minimum values ​​of the spacing, i.e., D F ∈[D Fmin ,D Fmax ],in

[0080]

[0081] Among them, v is Poisson's ratio, E is Young's modulus, t is the expected production time, P e 、P w are the boundary pressure and bottom hole pressure, c t is the comprehensive compression coefficient.

[0082] In step S200, different characteristic factors have different effects on the production capacity monitoring of different oilfields. When constructing an oilfield monitoring model, it is crucial to analyze the correlation between the production capacity of a specific oilfield and the characteristics. Too many characteristics can increase the complexity of the model, which may lead to overfitting; while the least influential characteristics may reduce the accuracy of the model. By evaluating and selecting the characteristics that have a major impact, weakly correlated characteristics can be eliminated while strongly correlated characteristics are retained, improving the accuracy of the model's predictions. Based on this, the present embodiment first uses the Spearman correlation coefficient to evaluate the monotonic relationship between each characteristic and the purpose of oilfield monitoring, obtaining a preliminary feature impact order; then uses a random forest tree model to train the data to evaluate the comprehensive impact of the characteristics on oilfield monitoring. The Spearman correlation coefficient can solve the influence of the correlation between features on the evaluation in the random forest model, and the random forest model can solve the problem that the Spearman correlation coefficient cannot capture complex non-monotonic relationships. Based on this, the specific steps of step S200 are as follows:

[0083] In step S210, the Spearman correlation coefficient is used to evaluate the monotonic relationship between each characteristic and the purpose of oilfield monitoring for multiple samples.

[0084] In step S220, the production capacity is set as the dependent variable to train the random forest regression model and calculate the feature importance.

[0085] In step S230, the feature impact order is evaluated.

[0086] In step S210, the characteristics in each sample are calculated for multiple samples, the Spearman correlation coefficient r between each characteristic and the production capacity is calculated, and the absolute values are sorted to determine the correlation strength between the characteristics and the production capacity, wherein the Spearman correlation coefficient r is calculated by the following formula

[0087]

[0088] where M is the total number of samples, x m is the feature value in the mth sample, y m is the value of the production capacity in the mth sample, The greater the absolute value |r| of the Spearman correlation coefficient, the stronger the monotonic correlation between the characteristics and the oilfield production capacity. Figure 2 The figure shows the analysis heat map of the Spearman correlation coefficient calculated by the data obtained by the present embodiment for monitoring an oilfield.

[0089] In step S220, a new data set of characteristics is randomly and with replacement selected each time, and each decision tree in the random forest model is independently trained. The final oilfield production capacity maximum displacement and final displacement regression prediction results are obtained by averaging the prediction results of all decision trees. Specifically

[0090] Step S221, determine the number of decision trees, the way of feature selection, the way of growing decision trees;

[0091] Step S222, build root node: put all training data in root node, select an optimal feature, and take the optimal feature as the splitting criterion of the current node;

[0092] Step S223, data segmentation: segment the training data set into subsets according to the selected feature, each subset corresponding to a value of the feature; if these subsets can be correctly classified, build leaf nodes and divide these subsets into corresponding leaf nodes; if there are still subsets that cannot be correctly classified, then reselect an optimal feature for these subsets and continue to segment them to build corresponding nodes;

[0093] Step S224, recursively build sub-tree: repeat data segmentation to build sub-tree for each subset of data on the leaf node until the maximum tree size is met to stop building;

[0094] Step S225, generate leaf node: generate leaf node and assign classification or regression result.

[0095] The optimal feature acquisition method in step S222 is as follows:

[0096] Step S2221, define data set D m ={D m,1 ,D m,2 ,...,D m,Nm} randomly sampled from the feature set {P, EP, VF O , S O , P R , V C , V O , V P , R P , S F , D F};

[0097] Step S2222, obtain the information entropy H(D m,n ) of the nth feature D m,n )

[0098]

[0099] wherein, is the number of samples of the jth class of the nth feature, and J m,n is the total number of classes of samples of the nth feature;

[0100] Step S2223, obtain the conditional entropy H(D m,n |D m ) of the nth feature D m,n ​

[0101]

[0102] Among them, N m For dataset D m The total number of features;

[0103] Step S2224, obtain the nth feature D m,n For dataset D m The degree of influence G(D m ,D m ,n)

[0104] G(D m ,D m,n )=H(D m,n )-H(D m |D m,n )

[0105] Step S2225, select max(G(D m ,D m ,n)) is taken as the optimal eigenvalue.

[0106] In step S230, the Spearman correlation coefficient and the feature influence order obtained by the random forest algorithm are compared and comprehensively evaluated. Figure 2 The Spearman correlation coefficient analysis obtained the porosity (P), effective permeability (EP), oil saturation (S O ), crack cross section (S F ), reservoir pressure (P R ), flowback proppant volume (V P ), crack spacing (D F ) is strongly positively correlated with the target production capacity, and the proppant ratio (R P ) is positively correlated with the target production capacity, and the clay volume (V C ), oil volume factor (VF O ), oil viscosity (V O ) is weakly negatively correlated with the target production capacity. From Table 1, we can analyze that the porosity (P), fracture spacing (D F ), effective permeability (EP), oil saturation (S O ), flowback proppant volume (V P ), reservoir pressure (P R ) is the core feature of the model prediction.

[0107] Table 1 Correlations obtained by random forest

[0108] Features P EP VF O ]] [SA O ]]> P R ]]> V C ]]> V O ]]> V P ]]> [R P ]]> [SA F ]]> D F ]]> Relevance 0.92 0.87 -0.2 0.83 0.81 -0.2 -0.2 0.82 0.45 0.66 0.91

[0109] according toFigure 2 And as shown in Table 1, a comprehensive analysis is made to obtain Table 2, and the importance of the characteristics is ranked as porosity (P) > effective permeability (EP) > oil saturation (S O )> reservoir pressure (P R )> proppant volume (V P )> fracture section (S F )> fracture spacing (D F ).

[0110] Table 2 Comprehensive analysis table

[0111]

[0112] The WP-PSO-BP network model established in step S300 is used to monitor the oilfield. The BP neural network is a multi-layer feedforward network that uses error backpropagation for nonlinear fitting, enabling it to handle high-dimensional data and complex nonlinear relationships. The particle swarm optimization (PSO) algorithm is a swarm intelligence-based search algorithm suitable for nonlinear function optimization in multidimensional space. The algorithm operates by randomly generating a swarm of particles in space, which are evaluated based on the fitness values determined by the objective function, and continuously move in space at a certain speed. Through multiple iterations, global optimization is achieved. In each iteration, the individual optimal value and group optimal value are obtained by comparing the fitness values of each particle, and the speed and position of the particles are updated based on the positions of the individual optimal value and group optimal value. The particle swarm optimization (PSO) algorithm has issues such as fast convergence speed, poor global search ability, strong weight dependence, and consistent iteration population information during use. The Wiener process (WP) is a continuous-time stochastic process. Since the motion of the carbon quantum dot tracer in the fluid is affected by the random collisions of liquid molecules, these collisions are independent, continuous, and random. The diffusion process can be approximated as a Wiener process, and the above characteristics are obtained through the relevant parameters of the carbon quantum dot tracer. In practical applications, there is random fluctuation of the Wiener process. The Wiener process (WP) can break the deadlock of the particle swarm optimization (PSO) algorithm falling into local optimization through random disturbance, improving the global search ability and adaptability of the particle swarm optimization (PSO) algorithm. Therefore, a WP-PSO-BP network is constructed in this embodiment to monitor the oilfield, and the random parameters in PSO are updated using WP.

[0113] Specifically, the PSO particle is defined by two core vectors, namely the position vector and the velocity vector where the position vector represents the direct encoding of the optimized vector, and the velocity vector represents the direction and step length of the control position update. In this embodiment, the parameters to be optimized in the position vector include {porosity, effective permeability, oil volume factor, oil saturation, reservoir pressure, clay volume, oil viscosity, proppant volume, proppant ratio, fracture cross section, fracture spacing}, i.e., {P, EP, VF O O R C O P P F F} in this embodiment, WP significantly optimizes the performance of PSO (particle swarm optimization algorithm) in reservoir parameter inversion by introducing a randomness parameter. The randomness parameter is mainly generated by the combined effects of the diversity of the sedimentary environment, the influence of diagenesis, the influence of tectonic activity, the influence of biological activity, the influence of fluid flow, and random geological processes. The randomness parameter is as follows:

[0114] P : random fluctuation of porosity around the mean value in spatial distribution;

[0115] EP : random disturbance intensity of effective permeability, indicating random disturbance of spatial variability of permeability or porosity, with randomness from formation stress sensitivity (fracturing / closing), particle migration plugging, and dynamic changes in fracture network;

[0116] : random fluctuation intensity of oil saturation, indicating non-uniform displacement caused by microscopic pore scale phenomena, with randomness from non-uniform displacement front, random effects of capillary force, and evolution of water channeling channels;

[0117] : random fluctuation intensity of reservoir pressure, indicating random changes of pressure over time or space, with randomness from fluid production / injection disturbance, formation connectivity change, and edge and bottom water invasion;

[0118] : random fluctuation coefficient of proppant volume, indicating uncertainty of proppant distribution, with randomness from the influence of random fluctuation of production system (flow rate / pressure) on proppant flowback rate;

[0119] : random fluctuation coefficient of oil viscosity, indicating viscosity mutation under conditions such as temperature fluctuation, with randomness from non-uniform changes in the formation temperature field and asphaltene precipitation;

[0120] ​​​​​​​​​​Fracture cross-section random fluctuation coefficient, representing the ratio of the standard deviation to the average value of the fracture cross-section size, the randomness comes from proppant embedment / breakage, stress reorientation, fracturing fluid flowback rate fluctuation;

[0121] Fracture spacing random fluctuation coefficient, representing the ratio of the standard deviation to the average value of the fracture spacing, the randomness comes from proppant embedment / breakage, stress reorientation, fracturing fluid flowback rate fluctuation;

[0122] wherein, θ * is the random parameter corresponding to the convection term feature, σ * is the random parameter corresponding to the diffusion term feature. The oil volume factor (VF O ), the clay volume (V C ), and the proppant ratio (R P ) do not have corresponding random parameters. The oil volume factor (VF O ) randomness mainly comes from the random fluctuation of pressure rather than an independent random process. The clay volume (V C ) and the proppant ratio (R P ) are often treated as constants in PSO.

[0123] A random disturbance model is constructed for the above random parameters by a Wiener process:

[0124] d(P) = μ P dt + θ P dW P

[0125] d(EP) = μ EP dt + θ EP dW EP

[0126]

[0127] wherein, μ is a deterministic drift term determined by the change trend of the obtained features {porosity, effective permeability, oil volume factor, oil saturation, reservoir pressure, clay volume, oil viscosity, proppant volume, proppant ratio, fracture cross-section, fracture spacing}; dW is a Wiener increment, a random noise subject to a normal distribution N(0, dt).

[0128] In each iteration process of PSO, a Wiener process disturbance is added to each particle to simulate the randomness of the parameters. The fitness function of the PSO optimization algorithm in this embodiment is based on the MSE loss function, and the random disturbance of the Wiener path is added, i.e.

[0129]

[0130] wherein, N is the number of Wiener paths, T is the monitoring period, is the ith Wiener path.

[0131] The updating process of the particle by using the Wiener process is

[0132] X t+Δt = X t + d(f)

[0133] Wherein, X t is the characteristic parameter value at time t, and Delta t is the time step, and d(f) represents the random disturbance of the corresponding feature. Dt is the continuous time differential symbol, which represents an infinitesimal time increment; in practice, replace dt with a finite step length Delta t for discretizing the random update.

[0134] The process of the PSO algorithm optimized by the Wiener process is as shown in Figure 3 The WP-PSO-BP network is trained through the training set, and the authenticity and applicability of the WP-PSO-BP network are tested through the test set. Through Figure 4 It can be seen that the WP-PSO-BP model involved in the embodiment is closer to the actual capacity than the BP neural network model using only PSO.

Claims

1. A method for monitoring oil field productivity using carbon quantum dot tracers, characterized in that: include: Step S100, obtaining the convection and diffusion characteristics of the oil field by injecting carbon quantum dot tracers and inert tracers, and establishing an oil field monitoring feature set; Step S200, using the Spearman correlation coefficient and random forest model method to select the main features of oilfield monitoring from the oilfield monitoring feature set; In step S300, the main features are input into a neural network to monitor the oil field productivity. The neural network is a particle swarm optimization-back propagation neural network optimized by Wiener process.

2. The method according to claim 1, characterized in that In step S100, the convection item features include {porosity, effective permeability, oil volume factor, oil saturation, reservoir pressure, clay volume, oil viscosity}; the diffusion item features include {flowback proppant volume, proppant ratio, fracture cross section, fracture spacing}.

3. The method according to claim 2, characterized in that In step S100, the following parameters of the carbon quantum dot tracer and the inert tracer are recorded to obtain the detection feature set of the oil field: the first breakthrough time and peak concentration time of the carbon quantum dot tracer, the first breakthrough time and peak concentration time of the inert tracer, the total amount of carbon quantum dot tracer injected and the total amount of flowback, the total amount of inert tracer injected and the total amount of flowback, the length of the main flow line between the injection and production wells of the inert tracer, and the flow cross-sectional area of ​​the carbon quantum dot tracer.

4. The method according to claim 2 or 3, characterized in that The specific process of step S200 includes: Step S201 , evaluating the monotonic relationship between each feature and oilfield productivity using the Spearman correlation coefficient; Step S202, determining the importance between each feature and oilfield productivity through a random forest tree model; Step S203: Select the main features of oilfield monitoring based on the monotonic relationship and importance.

5. The method according to claim 4, characterized in that In step S203 , the features with strong positive correlation in the Spearman correlation coefficient and the top K correlations in the random forest tree model are selected as the main features for oilfield monitoring.

6. The method according to claim 4, characterized in that In step S300, the random perturbation d(f) of characteristic porosity, effective permeability, oil saturation, reservoir pressure, oil viscosity, flowback proppant volume, fracture cross section, and fracture spacing is obtained using the Wiener process in the neural network. d(f)=μ f dt+σ f dW f Among them, f represents the feature, μ f is the deterministic drift term of feature f, σ f is the randomness parameter of feature f, dW f is the Wiener increment.

7. The method according to claim 6, characterized in that In the neural network, random perturbations d(f) are used to update particles. X t+Δt =X t +d(f) Among them, X t is the characteristic parameter value at time t, and Δt is the time step.

Images