Old well measure structure optimization method based on BP neural network and genetic algorithm

Abstract

The invention belongs to the technical field of oil field optimization, and particularly relates to an old well measure structure optimization method based on a BP neural network and a genetic algorithm, and the method comprises the following steps: S1, collecting historical measure yield increase data of a target block (oil reservoir) and various related dynamic and static geological oil reservoir parameters; s2, analyzing single well measure yield increase influence factors, carrying out correlation analysis according to oil reservoir engineering principles and measure yield increase historical data, and finally determining main influence factors of measure yield increase. According to the method, efficient and accurate measure yield increase effect prediction can be carried out on various oil reservoir types, reservoir conditions and measure types, a specific seepage mechanism and a specific yield increase mechanism do not need to be concerned, and the problems that measure yield increase prediction precision is not high and the workload is huge are solved; and efficient and accurate references and suggestions can be provided for oil field old well measure structure optimization and decision making.

Description

Technical Field

[0001] The invention belongs to the technical field of oil field optimization, and in particular relates to an old well measure structure optimization method based on BP neural network and genetic algorithm. Background Art

[0002] As the main means of stabilizing oilfield production and an important part of oilfield development, offensive measures for old wells must also give priority to benefits. Different types of measures have different adaptability to different reservoir geological conditions, and there are also large differences in input-output ratios. Therefore, there is an urgent need for a method and technology that comprehensively utilizes historical production increase data, geological reservoir parameters, economic parameters and other data of old well measures, and uses effective mathematical tools, advanced computers and artificial intelligence technology to scientifically optimize the structure of measures to achieve the best conversion from input to output.

[0003] Many scholars at home and abroad have conducted a lot of relevant research on the establishment of oilfield measure structure optimization model and model solution method, and established the correlation between the measure increase production and related influencing factors based on neural network. However, the influencing factors mainly consider macro factors such as the number of measure wells, the number of effective measure wells, water content, and remaining recoverable reserves. It is unable to reflect the influence of the geological characteristics of a single well reservoir on the effect of the measure increase production. Therefore, the prediction accuracy is limited, and it is also unable to well predict the effect of measures on unconventional reservoirs such as ultra-low permeability oil and gas reservoirs and shale oil and gas reservoirs.

[0004] Therefore, we propose an old well measure structure optimization method based on BP neural network and genetic algorithm to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a method for optimizing the structure of old well measures based on BP neural network and genetic algorithm in view of the above problems.

[0006] To achieve the above object, the present invention provides the following technical solution: an old well measure structure optimization method based on BP neural network and genetic algorithm, comprising the following steps:

[0007] S1. Collect historical production increase data of target blocks (reservoirs) and related dynamic and static geological reservoir parameters;

[0008] S2. Analysis of factors affecting production increase through single well measures

[0009] According to the reservoir engineering principles and historical data of production increase measures, correlation analysis is carried out to finally determine the main influencing factors of production increase measures;

[0010] S3. Through neural network learning and training, a mapping relationship between measures to increase production and influencing factors is established to predict the measures to increase production under different types of reservoir geological conditions and measure parameters in the future;

[0011] S4. Determine the objective function (economic benefit) and constraint conditions of the measure structure optimization model;

[0012] S5. Use the genetic algorithm to solve the constructed measure structure optimization model, and finally obtain the optimal measure structure distribution that meets the target conditions.

[0013] In the above method for optimizing the measure structure of old wells based on BP neural network and genetic algorithm, in step S3, according to historical production data, a mapping relationship between measure oil production and its influencing factors is established using the BP neural network.

[0014] In the above method for optimizing the measure structure of old wells based on BP neural network and genetic algorithm, in step S5, the penalty function method is used to transform the established constrained optimization problem into an equivalent unconstrained optimization problem.

[0015] In the above method for optimizing the measure structure of old wells based on BP neural network and genetic algorithm, in step S4, combined with historical dynamic data, the values of uncontrollable factors in the influencing factors of measure oil production in the planned year are predicted using neural network or grey theory.

[0016] In the above method for optimizing the measure structure of old wells based on BP neural network and genetic algorithm, in step S5, the equivalent unconstrained optimization objective function is used as the fitness function, and the optimal measure structure of the oilfield in the planned year is solved through the genetic algorithm.

[0017] In the above method for optimizing the measure structure of old wells based on BP neural network and genetic algorithm, the specific steps of neural network learning and training in step S3 include the following:

[0018] A. Prepare training data;

[0019] B. Determine the topological structure of the BP neural network;

[0020] C. Optimize the optimal initial weights and thresholds using the genetic algorithm;

[0021] D. Calculate the outputs of the hidden layer units and the output layer units in the forward direction;

[0022] E. Calculate the error of the output unit;

[0023] F. When the accuracy is not met, adjust the weights and thresholds of the hidden layer - output layer and the input layer - hidden layer in the reverse direction;

[0024] G. After reaching the number of training times, the training ends.

[0025] In the above method for optimizing the measure structure of old wells based on BP neural network and genetic algorithm, when the accuracy is met in step E, the training ends directly.

[0026] In the above method for optimizing the old well measure structure based on BP neural network and genetic algorithm, when the number of training times is not reached in step G, the program of step D is performed again.

[0027] In the above method for optimizing the old well measure structure based on BP neural network and genetic algorithm, the specific steps of using genetic algorithm to solve the constructed measure structure optimization model in S5 are as follows:

[0028] a. Population initialization;

[0029] b. Determine the fitness function;

[0030] c. Selection, crossover, and mutation;

[0031] d. Calculate the value of the fitness. When the termination condition is met, output the result.

[0032] In the above method for optimizing the old well measure structure based on BP neural network and genetic algorithm, when the termination condition is not met in step d, the work of step c is performed again.

[0033] Compared with the prior art, the present invention provides a method for optimizing the old well measure structure based on BP neural network and genetic algorithm, having the following beneficial effects:

[0034] The method for optimizing the old well measure structure based on BP neural network and genetic algorithm optimizes and innovates on the basis of the traditional oilfield measure structure optimization model and solution method. The influencing factors of measure incremental production comprehensively consider dynamic and static data such as reservoir physical property parameters, reservoir energy maintenance degree, reservoir stimulation parameters, remaining oil saturation, historical production data, etc. The neural network is used to establish their mutual relationship, making the prediction result more accurate and the prediction efficiency higher. It can efficiently and accurately predict the measure incremental production effect for various reservoir types, reservoir conditions, and measure types without caring about the specific seepage mechanism and stimulation mechanism, solving the problems of low accuracy of measure incremental production prediction and huge workload, and can provide efficient and accurate reference and suggestions for the optimization and decision-making of the old well measure structure in the oilfield.

[0035] In summary: The present invention can efficiently and accurately predict the measure incremental production effect for various reservoir types, reservoir conditions, and measure types without caring about the specific seepage mechanism and stimulation mechanism, solving the problems of low accuracy of measure incremental production prediction and huge workload, and can provide efficient and accurate reference and suggestions for the optimization and decision-making of the old well measure structure in the oilfield. Description of the Drawings

[0036] Figure 1It is a flow chart for establishing and solving an optimization model of an optimization method for the old well measure structure based on BP neural network and genetic algorithm proposed by the present invention. Specific embodiments

[0037] The following embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention.

[0038] Please refer to Figure 1 , an optimization method for the old well measure structure based on BP neural network and genetic algorithm, including the following steps:

[0039] S1. Collect historical production increase data of measures and various related dynamic and static geological reservoir parameters of the target block (reservoir);

[0040] S2. Analysis of influencing factors of single-well measure production increase

[0041] According to reservoir engineering principles and historical production increase data of measures, carry out correlation analysis, and finally determine the main influencing factors of measure production increase;

[0042] S3. Through neural network learning and training, establish a mapping relationship between measure production increase and influencing factors, which is used to predict measure production increase under different types of reservoir geological conditions and measure parameters in the future;

[0043] S4. Determine the objective function (economic benefit) and constraint conditions of the measure structure optimization model;

[0044] S5. Use genetic algorithm to solve the constructed measure structure optimization model, and finally obtain the optimal measure structure distribution that meets the target conditions.

[0045] In step S3, according to historical production data, use BP neural network to establish a mapping relationship between measure oil production and its influencing factors.

[0046] In step S5, use the penalty function method to transform the established constrained optimization problem into an equivalent unconstrained optimization problem.

[0047] In step S4, combined with historical dynamic data, use neural network or grey theory to predict the values of uncontrollable factors in the influencing factors of measure oil production in the planned year.

[0048] In step S5, take the equivalent unconstrained optimization objective function as the fitness function, and solve the optimal measure structure of the oilfield in the planned year through genetic algorithm.

[0049] The specific steps of neural network learning and training in step S3 include the following:

[0050] A. Prepare training data;

[0051] B. Determine the topological structure of the BP neural network;

[0052] C. Optimize the optimal initial weights and thresholds using a genetic algorithm;

[0053] D. Forward-calculate the outputs of the hidden layer units and the output layer units;

[0054] E. Calculate the error of the output unit;

[0055] F. When the accuracy is not met, adjust the weights and thresholds of the hidden layer-output layer and the input layer-hidden layer in reverse;

[0056] G. After reaching the number of training times, the training ends.

[0057] When the accuracy is met in step E, the training ends directly.

[0058] When the number of training times is not reached in step G, the program work of step D is carried out again.

[0059] The specific steps for using a genetic algorithm to solve the constructed measure structure optimization model in S5 are as follows:

[0060] a. Initialize the population;

[0061] b. Determine the fitness function;

[0062] c. Select, crossover, and mutate;

[0063] d. Calculate the fitness value, and when the termination condition is met, output the result.

[0064] When the termination condition is not met in step d, the work of step c is carried out again.

[0065] Based on the traditional oilfield measure structure optimization model and solution method, optimization and innovation have been carried out. For the influencing factors of measure incremental production, dynamic and static data such as reservoir physical property parameters, degree of reservoir energy maintenance, reservoir stimulation parameters, remaining oil saturation, and historical production data are comprehensively considered. A neural network is used to establish their mutual relationships, making the prediction result more accurate and the prediction efficiency higher. It can efficiently and accurately predict the measure incremental production effect for various reservoir types, reservoir conditions, and measure types without caring about the specific seepage mechanism and stimulation mechanism, solving the problems of low accuracy of measure incremental production prediction and huge workload, and can provide efficient and accurate reference and suggestions for the optimization and decision-making of the measure structure of old wells in the oilfield.

[0066] The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An optimization method for the structure of old well measures based on BP neural network and genetic algorithm, characterized in that The method includes the following steps: S1. Collect the historical production-increasing data of the target block (reservoir) and various related static and dynamic geological reservoir parameters; S2. Analyze the influencing factors of the production increase of single well measures According to the reservoir engineering principle and the historical production-increasing data, carry out correlation analysis, and finally determine the main influencing factors of the production increase of measures; S3. Through neural network learning and training, establish the mapping relationship between the production increase of measures and the influencing factors, which is used to predict the production increase of measures under different types of reservoir geological conditions and measure parameters in the future; S4. Determine the objective function (economic benefit) and constraint conditions of the measure structure optimization model; S5. Use the genetic algorithm to solve the constructed measure structure optimization model, and finally obtain the optimal measure structure distribution that meets the target conditions.

2. The optimized method for the old well measure structure based on the BP neural network and genetic algorithm according to claim 1, wherein In S3, according to the historical production data, use the BP neural network to establish the mapping relationship between the oil production of measures and its influencing factors.

3. The optimized method for the old well measure structure based on the BP neural network and genetic algorithm according to claim 1, characterized in that In S5, use the penalty function method to transform the established constrained optimization problem into an equivalent unconstrained optimization problem.

4. A method for optimizing the structure of old well measures based on BP neural network and genetic algorithm according to claim 1, characterized in that In S4, combined with the historical dynamic data, use the neural network or grey theory to predict the values of uncontrollable factors in the influencing factors of the measure oil production in the planned year.

5. A method for optimizing the structure of old well measures based on BP neural network and genetic algorithm according to claim 1, characterized in that, In S5, take the equivalent unconstrained optimization objective function as the fitness function, and solve the optimal measure structure of the oilfield in the planned year through the genetic algorithm.

6. The optimization method for the old well measure structure based on BP neural network and genetic algorithm according to claim 1, characterized in that The specific steps of the neural network learning and training in S3 include the following: A. Prepare the training data; B. Determine the topology structure of the BP neural network; C. Use the genetic algorithm to optimize the optimal initial weights and thresholds; D. Forward calculate the outputs of the hidden layer units and the output layer units; E. Calculate the error of the output unit; F. When the accuracy is not met, adjust the weights and thresholds of the hidden layer-output layer and the input layer-hidden layer in reverse; G. After reaching the training times, the training ends.

7. A method for optimizing the structure of old well measures based on BP neural network and genetic algorithm according to claim 6, characterized in that In step E, when the accuracy is met, the training ends directly.

8. A method for optimizing the structure of old well measures based on BP neural network and genetic algorithm according to claim 6, characterized in that, In step G, when the training times are not reached, perform the program work of step D again.

9. The optimization method for the old well measure structure based on the BP neural network and genetic algorithm according to claim 1, characterized in that, The specific steps of using the genetic algorithm to solve the constructed measure structure optimization model in S5 include the following: a. Initialize the population; b. Determine the fitness function; c. Select, crossover, and mutate; d. Calculate the value of the fitness. When the termination condition is met, output the result.

10. A method for optimizing the structure of old well measures based on BP neural network and genetic algorithm according to claim 9, characterized in that, In step d, when the termination condition is not met, re-perform the work of step c.

Images