CN104200059B - An oil-water well behavior analysis and prediction device and method - Google Patents

Abstract

The invention discloses an oil-water well behavior analysis and prediction device and method. The technical solution adopted in the present invention includes the oil wells to be measured and the water wells arranged around the oil wells, characterized in that: the oil wells and the water wells are respectively equipped with a crude oil water content sensor and a water injection sensor, which are respectively used to collect the water content and water content of the oil wells. The water injection volume of the well, and the collected analog quantity is converted into a digital quantity through the A/D converter, and then transmitted to the host computer through the data cable, and the host computer performs stability identification on the collected data. The invention provides a stability identification algorithm for the data used for modeling, and has the advantages of clear steps, simple algorithm, satisfactory effect and the like.

Description

An oil-water well behavior analysis and prediction device and method

technical field

The invention belongs to the field of oilfield development prediction, in particular to an oil-water well behavior analysis and prediction device.

Background technique

Analysis and prediction of single well behavior is an important content in oilfield production management. At present, conventional analysis methods are generally used for oil wells, such as productivity analysis, production decline analysis, etc. These methods all isolate the oil well without considering the impact of surrounding water wells. Some unconventional methods of cross-well analysis, such as interference well test, tracer test and other technologies, are often rarely used because of the shutdown of the oil well or the high cost. Numerical simulation is mostly used to study the behavior of well groups, such as multivariate time series analysis method.

The system state is divided into stable state, unstable state and transition state between the two. For the prediction of oil and water well behavior, modeling is usually adopted. The collected samples are assumed to be obtained in a stable state of the system, and then a mathematical model is established to predict the behavior of oil and water wells. Existing predictive algorithms cannot discern whether the collected data used for modeling is reliable. Therefore, when the collected data deviates from the original stable state due to equipment aging, system instability, etc. in the middle and transition stages of the stable process and the unstable state, it is bound to make the prediction model constructed by using data containing unstable information at this time The prediction effect deviates from the stable system direction, which leads to misjudgment. However, there is no relevant literature to identify the algorithm for modeling data, so it is necessary to identify and analyze the modeling data, that is, to identify the stability of the initial matrix for modeling.

Contents of the invention

The technical problem to be solved by the present invention is to provide an oil-water well behavior analysis and prediction device and method.

In order to solve the above problems, the technical solution adopted by the present invention includes the oil wells to be measured and the water wells arranged around the oil wells, which are characterized in that: the oil wells and the water wells are respectively equipped with a crude oil water content sensor and a water injection sensor, which are respectively used to collect The water content of the oil well and the water injection volume of the water well, and the collected analog quantity is converted into a digital quantity through the A/D converter, and then transmitted to the host computer through the data cable, and the host computer performs stability identification on the collected data.

The method realized by the device is characterized in that the host computer performs stability identification on the collected data and includes the following steps:

Collection amount of data , enough samples are collected and divided into indivual matrix, is a natural number;

(1) order when order matrix is the steady state, take its covariance matrix , then the characteristic equation is: ;

(2) It can be considered that the remaining indivual Stability Judgment of Order Matrix in Normal Process Data If the disturbance has little effect on the original process, it is considered to be a benign disturbance, otherwise it is a vicious disturbance. The former can be used to build a prediction model or to update the model. It is expressed in mathematical language as: indivual The order matrix covariance is , the corresponding eigenvalues and eigenvectors change as: , , then the first indivual The characteristic equation of the order matrix covariance is: , solve the above formula to get ;

(3) Take the 2 norm of the above formula to get , which is the first indivual The relative error of the covariance matrix of the order matrix to the normal process data matrix, since the covariance matrix is a symmetric matrix, so , since the matrix in this formula is a real symmetric matrix with a complete eigenvector system, the maximum eigenvalue of the matrix can be obtained iteratively by using the power method;

(4) Repeat (2), (3) until , the smaller norm value is considered as normal process data, otherwise it can be considered as initial fault data or fault data and should be discarded.

The invention provides a stability identification algorithm for the data used for modeling, and has the advantages of clear steps, simple algorithm, satisfactory effect and the like. The application of the invention to the behavior prediction and analysis of oil and water wells will further improve the prediction accuracy, thereby indirectly improving the oil field management efficiency. In addition, the device of the present invention has the advantages of simple operation, convenient installation, light weight and the like.

Description of drawings

The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

Fig. 1 is a schematic diagram of the analysis and prediction device of the present invention;

Fig. 2 is a schematic diagram of the change of facies change of oil-water well group in the present invention.

detailed description

As shown in Figure 1, the oil-water well behavior analysis and prediction device of the present invention includes a measured object oil well 1 and a water well 2 arranged around the oil well. The sensor 4 is used to collect the water content of the oil well and the water injection rate of the water well respectively, and converts the collected analog quantity into a digital quantity through the A/D converter 5, and then transmits it to the host computer 6 through a data cable, and the host computer is to the host computer 6. Collect data for stability identification.

Stability identification of the collected data by the host computer includes the following steps:

Collection amount of data , enough samples are collected and divided into indivual matrix, is a natural number;

(1) order when order matrix is the steady state, take its covariance matrix (still use the code), then the characteristic equation is: ;in for Order matrix, composed of collected data; for the matrix eigenvalues; for order unit array; is the eigenvector; the equation is the solution matrix Features are worth formulas.

(2) It can be considered that the remaining indivual Stability Judgment of Order Matrix in Normal Process Data If the disturbance has little effect on the original process, it is considered to be a benign disturbance, otherwise it is a vicious disturbance. The former can be used to build a prediction model or to update the model. It is expressed in mathematical language as: indivual The order matrix covariance is ( is the perturbation matrix, ,and ), the corresponding eigenvalues and eigenvector changes (or solution changes) are: , ( ,and ), then the indivual The characteristic equation of the order matrix covariance is: , solve the above formula to get ;

(3) Take the 2 norm of the above formula to get , which is the first indivual The relative error of the covariance matrix of the order matrix to the normal process data matrix, since the covariance matrix is a symmetric matrix, so , because the matrix in this formula is a real symmetric matrix and has a complete eigenvector system, the maximum eigenvalue of the matrix can be obtained iteratively by using the power method; the meaning of each letter is the same as step (1). The power method is an iterative algorithm for solving the largest eigenvalue and corresponding eigenvector of a matrix. It is a numerical analysis method, and its basic idea is:

supposing order matrix The eigenvalues and eigenvectors of , take an initial dimension vector ,and (Note: ask for the infinite norm of ). algorithm termination constant , set ; Calculation steps are as follows (subscript is the number of iterations):

a. Calculate ;

b. seek ;

c.

d.if , then the output approximate eigenvector and approximate eigenvalues , to terminate the algorithm. Otherwise set , return to step a.

(4) Repeat (2), (3) until , the smaller norm value is considered as normal process data, otherwise it can be considered as initial fault data or fault data and should be discarded.

Experimental example:

A certain oilfield block has a five-point area well pattern, and the injection-production well spacing is 500m. Take a five-point well group to analyze the injection-production correspondence. The relative positions of oil and water wells in the five-point well group are: an oil well in the middle, 4 water wells around it, water well 1 and water well 2 in the northwest and northeast corners, and water well 3 and water well 4 in the southwest and southeast corners.

If the oil-water well group is regarded as a multi-input and multi-output system, the production indicators of one oil well and the surrounding water wells or one water well and the surrounding oil wells constitute a multivariate time series, and the water content of the oil well and the water injection volume of the four water wells are collected respectively , collected once a month, a total of 26 times, forming historical data. The raw data of oil well water cut and water injection volume are shown in Table 1 below:

The system model uses (autoregressive (AR) model is also called time series model) model, the recursive least squares method of parameter estimation establishes the following model, here only data stability analysis is performed.

Measurement data Respectively represent the water content of the oil well, the water injection volume of the water well 1, the water injection volume of the water well 2, the water injection volume of the water well 3 and the water injection volume of the water well 4, subscript Indicates the current moment. The five quantities are subject to white noise The impact is as follows:

Under normal circumstances, 1000 samples are collected and then Gaussian white noise is added to the 1-200 samples. According to the method of the above technical solution, the samples are divided into 200 5th-order square matrices, and the relative changes are calculated respectively. The schematic diagram is shown in Figure 2. Due to the addition of Gaussian noise to the first 40 square matrices, the 2-norm fluctuates greatly compared with the subsequent ones. This shows that the algorithm is effective.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention are included in the protection scope of the present invention. Inside.

Claims (1)

1. a kind of oil-water well behavioural analysis prediction meanss, including measurand oil well（1）And it is arranged in the well around oil well （2）, it is characterised in that：The oil well（1）, well（2）On be respectively equipped with crude oil water content sensor（3）With water filling quantity sensor （4）, it is respectively used to gather the moisture content of oil well and the water injection rate of well, and by the analog quantity of collection through A/D converter（5）Conversion Transmitted into digital quantity, then by data cable to host computer（6）, the host computer is to gathered data progress stability identification；

The host computer carries out stability identification to gathered data and comprised the following steps：

If collection Individual data volume , collect enough samples and be divided into It is individual Matrix, For nature Number；

（1）Order When Rank matrix For stable state, its covariance matrix is taken , then characteristic equation be： ；

（2）Think remaining It is individual The judgement of stability of rank matrix is in normal processes data On the basis of plus shock wave and Into if disturbance influences little to former process, then it is assumed that be benign disturbance, be otherwise pernicious disturbance, the former is used to build prediction mould Type or for more new model, is expressed as with the language of mathematics：The It is individual Rank matrix covariance is , accordingly Characteristic root and characteristic vector, which become, to be turned to： , , then It is individual The characteristic equation of rank matrix covariance For： , above formula neutralizing is obtained ；

（3）2 norms are taken to obtain to above formula , as It is individual The covariance matrix of rank matrix is to normal The relative error of process data matrix, because covariance matrix is symmetrical matrix, so , Because matrix belongs to real symmetric matrix and has complete characteristic vector-system in the formula, the maximum feature of matrix is asked for using power method iteration Value；

（4）Repeat（2）、（3）Until , norm value is less to think normal processes data, otherwise it is assumed that being primary fault number According to or fault data, should abandon.