Optimal selection of control parameters for automatic drilling with constant drilling pressure based on mrpDMOPSO algorithm

Abstract

The constant drilling pressure automatic drilling control parameter optimization based on the mrpDMOPSO algorithm introduces the detection and response mechanism of environmental changes, which solves the problems of easy falling into local optimality and low diversity. The implementation process of the present invention is: (1) initialization; (2) problem change detection mechanism; (3) environmental change response mechanism of multi-representative point prediction. It mainly includes: representative point selection, calculation prediction direction and environmental change response mechanism of multi-representative point prediction; (4) static optimization algorithm iosMOPSO. It mainly includes: external file update, global optimal particle alternating selection strategy, individual optimal particle selection and other strategies and establishment of drilling pressure automatic drilling PID objective function. (5) Determine whether the optimization termination condition is met. If so, output the optimal front end of the constant drilling pressure automatic drilling PID parameters. Otherwise, go to (2). The present invention improves the dynamic multi-objective optimization performance and solves the problem of stable and accurate control of drilling pressure in the constant drilling pressure automatic drilling system.

Description

Technical Field

[0001] The invention belongs to the technical field of automatic drill feeding control, and relates to optimization of constant drilling pressure automatic drill feeding control parameters based on dynamic multi-objective particle swarm optimization of multi-representative point prediction. Background Art

[0002] In the 2021 Global Energy Consumption Statistics Report, oil and natural gas consumption accounted for as much as 55.9%. At the same time, oil and gas exploration and development are expanding from conventional oil and gas resources to unconventional oil and gas resources such as low permeability, deep and ultra-deep layers, deep ocean water, and shale oil and gas, facing challenges such as resource degradation, exploration diversification, development complexity, and environmental harshness. Therefore, during the drilling process, the use of the automatic drill delivery system can not only reduce the labor intensity of workers, but also be easy to operate and safe to use, which can improve drilling quality, speed up drilling speed, and extend the service life of the drill bit, becoming a key technology to improve drilling equipment technology.

[0003] In the actual drilling process, the automatic drill bit feeding with constant drilling pressure automatically adjusts the speed of the winch drum according to the drilling process requirements, so that the drilling pressure, drilling speed and other parameters of the drill bit can be controlled during drilling, ensuring the stability of the drilling pressure at the drill bit. This drill feeding method is the most widely used.

[0004] However, my country's oil and gas production, on the one hand, improves the recovery rate, develops the remaining oil resources and special economic marginal oil reservoirs such as low permeability, ultra-thin, heavy oil and super-heavy oil, as well as scarce resources such as shale gas and coalbed methane; on the other hand, domestic oil resources are increasingly concentrated in deep formations and deep waters, and must also rely on deep / ultra-deep well vertical drilling for exploration and development. In addition, the controlled object in the drilling process has the characteristics of multivariable, time-varying, strong disturbance, and time lag, which restricts the application of automatic drilling technology in the actual complex well drilling process.

[0005] Considering the very complex working conditions at the actual drilling site, such as the elastic deformation of the wire rope in the traveling system of the oil drilling rig and the drill pipe in the drilling tool; the saturated output phenomenon of the actuators such as the disc brake during operation; the dead zone of the overhead crane and the traveling block due to friction and mechanical structure; the time-varying friction coefficient of the brake disc, the wear of the drill bit, and the vibration of the derrick; in the actual constant drilling pressure automatic drilling control, dynamic multi-objective control decisions should be made according to the changes in the formation environment encountered and the drilling tools. Many factors make it difficult for the existing constant drilling pressure automatic drilling control to achieve stable and accurate constant drilling pressure automatic drilling control. Therefore, it is urgent to find a dynamic multi-objective optimization algorithm based on the environmental change detection and response mechanism to guide the actual automatic drilling process and improve the drilling efficiency and quality. Summary of the invention

[0006] The purpose of the present invention is to address the defects of the above-mentioned prior art and propose a digital PID control method for automatic drilling with constant drilling pressure based on a dynamic multi-objective particle swarm optimization algorithm based on representative point prediction. Dynamic Multi Objective Particle Swarm Optimization based on MultiRepresentative-points Prediction (mrpDMOPSO) algorithm. An innovative design of the global optimal particle alternating selection strategy balances the diversity and local search capabilities of the algorithm according to the crowding distance and dominant strength, and constructs a multi-objective particle swarm optimization algorithm for optimal particle selection (Multi Objective Particle Swarm Optimization with Improved Optimal-particle Selection, iosMOPSO); at the same time, an environmental change response mechanism for multi-representative point prediction is designed. According to the severity of environmental changes, a certain number of representative points are adaptively allocated to track environmental changes and guide particles to search. A more uniform and better distributed optimal frontier and optimal solution set are searched. Taking hydraulic disc brake constant drilling pressure automatic drilling as the object, the mrpDMOPSO algorithm is used for dynamic multi-objective constant drilling pressure automatic drilling digital PID parameter optimization to achieve fast and accurate dynamic adaptive control of constant drilling pressure automatic drilling.

[0007] To achieve the above purpose, the technical solution of the present invention is implemented as follows:

[0008] (1) Initialization: Including mrpDMOPSO algorithm parameter definition and population initialization.

[0009] (1.1) mrpDMOPSO parameter definition. For example, the population size n p , external file capacity n A , inertia weight ω, acceleration factor c1=c2 and maximum number of iterations T max wait.

[0010] (1.2) Population initialization. Population initialization includes the initial positions of particles. Initial Speed The current optimal position is The global optimal position of the population is wait.

[0011] (2) Problem change detection mechanism. The problem change degree is calculated as δ according to the formula t , determine whether it is greater than the corresponding threshold ε. If so, go to (3), otherwise go to (4).

[0012] Assume that the degree of environmental change at time t is δ t , the calculation method is as shown in formula (1).

[0013]

[0014] In the formula, Represents individual x before and after the tth iteration i The change in the jth target value is calculated as shown in formula (2).

[0015]

[0016] In the formula, It represents the mean change of the j-th target value of the environmental change detection group before and after the t-th iteration, and its calculation is as shown in formula (3).

[0017]

[0018] In the formula, and are the maximum and minimum values ​​of the jth target, respectively, m is the number of targets, n is D is the number of detected individuals. t If it is greater than the set change threshold ε, the problem is considered to have changed.

[0019] (3) Environmental change response mechanism predicted by multiple representative points. t is the external archive, i.e. the currently obtained Pareto optimal solution set; n r is the number of representative points, and m is the number of targets.

[0020] (3.1) Representative point selection.

[0021] Representative points n r calculate:

[0022] n r =n r min +δ t (n r max -n r min ) (4)

[0023] In the formula, δ t is the degree of environmental change; n r max and n r min n r The maximum and minimum values ​​of r min =m+1,n r max =3m. From formula (4), we can know that n under different environmental change levels r The values ​​are different, so that the number of representative points can change adaptively with the environment.

[0024] in, The solution is as shown in formula (5).

[0025]

[0026] Where n A A t The capacity size, A t The i-th individual in .

[0027] The mth boundary solution of the PF searched The calculation is as shown in formula (6).

[0028]

[0029] In the formula, is the minimum value of the mth target.

[0030] (3.2) Calculate the predicted direction;

[0031] set up Any representative point First, select the representative point set from the previous moment Select distance The nearest representative point The prediction direction for:

[0032]

[0033] Then, the available set of evolutionary guidance directions at time t is express.

[0034] (3.3) Based on the environmental change response mechanism predicted by multiple representative points, the predicted new population is obtained. The predicted individual set S P1 Individuals in The calculation is as shown in formula (8).

[0035]

[0036] In the formula, for Guidance direction; N(0, σ t ) is The same-dimensional quantity, each dimension of which satisfies the mean of 0 and the variance of δ t Gaussian perturbation.

[0037] (4) The static optimization algorithm iosMOPSO is used to obtain the Pareto optimal solution set at time t, that is, the external archive A. Assume T max is the maximum number of iterations, n p is the number of particles in the population, n Ais the external archive capacity. The specific steps of the multi-objective particle swarm optimization algorithm iosMOPSO for optimal particle selection are as follows:

[0038] (4.1) Initialization. Let the iteration counter t = 0; randomly generate n zero initial positions and velocities in the decision space. p particles; according to MOP, n p Each particle calculates m objective function values ​​respectively; initializes individual external files for each particle And add the i-th particle to the corresponding make The set of initial particles and S p With A t Input to (4.2) and update, and assign the result to A t ;

[0039] (4.2) External file update. With external archive A t Merge into A′, let Then calculate the individual crowding distance C i . Calculate individual The crowding distance Among them, individual The crowding distance C of the jth target i，j The calculation is as follows:

[0040]

[0041] In the formula, C i，j For individuals The crowding distance at the jth target.

[0042] (4.3)t=t+1:

[0043] (4.4) Global optimal particle Alternate selection strategy.

[0044] Among them, A t Any particle The convergence of can be measured using the dominant strength S i To measure, the calculation method is:

[0045]

[0046] In the formula, m represents the number of targets; n A Indicates the external file capacity; means rounding up; |g| means taking the absolute value; S i Reflects Compared to A tThe degree to which other particles in the PF approach the true PF can be used as a convergence measure, S i The larger it is, the better the convergence.

[0047] Global optimal particle Alternating selection strategy, calculated according to formula (11)

[0048]

[0049] In the formula, rand represents a random value among 1, 2, .., m, m represents the number of targets; S represents the set of particles with the best convergence performance, S rand represents the randth particle in S; C represents the optimal particle set for diversity, C rand represents the randth particle in C.

[0050] (4.5) Individual optimal particle We designed a method for selecting the best individual particles with diversity preservation and fast tracking. First, we used the individual external archives Save the Pareto optimal solution set in the search history of individual particles, and its update still adopts the external archive update strategy. Capacity A t 1 / 4 of the capacity, i.e. n A / 4. Then from the individual external archive of the i-th particle Select the global optimal particle with the i-th particle The nearest particle is taken as the individual optimal particle of the i-th particle

[0051] (4.6) Update each particle Speed ​​and position The particle motion equations (12) and (13) are used to update each particle Speed ​​and position

[0052]

[0053]

[0054] (4.7) Based on the multi-objective constant drilling pressure automatic drilling PID parameters, each particle is optimally evaluated. m objective function values ​​(m=2);

[0055] The first step is to establish a multi-objective mathematical model for optimizing the PID parameters of the constant drilling pressure automatic drilling:

[0056]

[0057] In the formula, k = [kp1 k i1 , k p2 ] T is the decision vector; t is the running time of the current system; J1=t r (S(k, t)) is the rise time of the S(k, t) WOB step response; J2 represents the cumulative absolute value error of the S(k, t) WOB step response; g i (k), h j (k) represent the i-th inequality and j-th equality constraints on decision variable k, respectively.

[0058] In the second step, the interference signal is added, and the constant drilling pressure automatic drilling PID parameters are dynamically optimized with equation (14) as the optimization target. The interference is added to the system by setting the friction coefficient μ between the brake caliper and the brake disc to change with time:

[0059]

[0060] (4.8) For each particle Will, and Input to (4.2), update the individual external file to obtain

[0061] (4.9) The set of all particles and S p With A t Input to (4.2), update the external file to get A t+1 ;

[0062] (4.10) External file Cauchy distribution perturbation strategy, update A t+1 . t Middle particles The perturbation method of Cauchy distribution is as follows:

[0063]

[0064] (4.11) If t>Tmax, the Pareto optimal solution set at time t is obtained, that is, the external archive A t Then the output A=A t ; Otherwise, jump to (4.3);

[0065] (5) Determine whether the optimization termination condition is met. If so, output the optimal front end of the constant drilling pressure automatic drilling PID parameters. Otherwise, go to (2).

[0066] The present invention innovatively designs a global optimal particle alternating selection strategy, balances the algorithm's diversity and local search capabilities according to the crowding distance and dominance intensity, introduces a global optimal particle selection method that balances convergence and diversity, and an individual optimal particle selection method that maintains diversity and fast tracking; an external archive perturbation strategy is used to construct a multi-objective particle swarm optimization algorithm iosMOPSO for optimal particle selection; through an environmental change detection mechanism and an environmental change response mechanism predicted by multiple representative points, a certain number of representative points are adaptively allocated to track environmental changes according to the severity of environmental changes, guiding particles to search. A dynamic multi-objective particle swarm optimization (mrpDMOPSO) algorithm based on multi-representative point prediction is designed. A multi-objective constant drilling pressure automatic drilling PID parameter optimization model is established, and the mrpDMOPSO algorithm is applied to the rise time J1=t of the drilling pressure step response. r (S(k, t)) and the cumulative absolute value error J2 of the drilling pressure step response are optimized to achieve the automatic drilling of constant drilling pressure under complex working conditions. PID parameter k = [k p1 k i1 , k p2 ] T Optimization can adapt well to different formation conditions, achieve stable and precise control of drilling pressure, and improve drilling success rate, drilling efficiency and quality.

[0067] The experimental results of using mrpDMOPSO to optimize the PID parameters of the automatic drilling with constant drilling pressure show that the traditional multi-objective optimized automatic drilling PID parameters have large deviations in the actual drilling pressure control. The PID parameter results optimized by the method proposed in the present invention overcome the defects of the existing methods and obtain multiple groups of PID parameters with better control performance. The obtained PID parameter optimal solution set can not only track the changes in the drilling environment, but also be closer to the optimal frontier. At the same time, the absolute value integral of the deviation IAE and the dynamic response overshoot of the system are significantly reduced. It meets the optimization requirements of the PID parameters of the actual complex drilling conditions and the constant drilling pressure control system with multivariable, time-varying, strong disturbance and time lag, and is more conducive to the stable and accurate control of the drilling pressure at the drill bit under different formation conditions.

[0068] A set of optimized PID parameters are randomly selected to observe the drilling pressure step response of the constant drilling pressure automatic drilling under different drilling environments. The response results further prove that the optimal solution set of PID parameters obtained based on the constant drilling pressure automatic drilling PID parameters optimized by mrpDMOPSO can not only track the changes in the drilling environment, but also be closer to the optimal frontier. At the same time, the absolute value integral of the error IAE is reduced by up to 68% compared with the suboptimal IAE, and the overshoot is reduced by up to 63% compared with the suboptimal overshoot. Therefore, the constant drilling pressure automatic drilling based on the PID parameter optimization of mrpDMOPSO can achieve a better control of drilling pressure.

[0069] Through the study of the mrpDMOPSO method for automatic drilling under constant drilling pressure, a theoretical foundation is laid for further realizing intelligent decision-making control of automatic drilling during drilling, intelligent drilling rig equipment, and automation of the drilling process, thereby improving drilling quality, drilling success rate, and drilling efficiency, reducing the occurrence of drilling accidents, and reducing drilling costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a flow chart of the dynamic multi-objective particle swarm optimization algorithm (mrpDMOPSO) for multiple representative point prediction.

[0071] Figure 2 It is a schematic diagram of the environmental change detection mechanism.

[0072] Figure 3 It is a schematic diagram of the global optimal particle alternating selection strategy.

[0073] Figure 4 It is a hydraulic disc brake constant drill pressure automatic drill feeding control block diagram.

[0074] Figure 5 It is the optimal solution set of PID parameters searched by mrpDMOPSO and MOPSO under different drilling environments.

[0075] Figure 6 It is the control system step response of the present invention and the four algorithms of relay feedback setting, fuzzy adaptive PID setting and MOPSO at different times.

[0076] Explanation of the attached table

[0077] Table 2 shows the parameters of hydraulic disc brake constant drilling pressure automatic drilling.

[0078] Table 3 shows the drilling pressure control performance indicators under different setting methods at time t=10.

[0079] Table 4 shows the drilling pressure control performance indicators under different setting methods at time t=20.

[0080] Table 5 shows the drilling pressure control performance indicators under different setting methods at time t=30. DETAILED DESCRIPTION

[0081] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0082] The optimization of the control parameters of the constant drilling pressure automatic drilling based on the mrpDMOPSO algorithm includes the following steps:

[0083] (1) Initialization: This includes the definition of mrpDMOPSO algorithm parameters and the initialization of the population.

[0084] (1.1) mrpDMOPSO parameter definition, such as population size n p , external file capacity n A , inertia weight ω, acceleration factor c1=c2 and maximum number of iterations T max Etc., as shown in Table 1.

[0085] Table 1 Parameter settings of mrpDMOPSO

[0086]

[0087] (1.2) Population initialization. Population initialization includes the initial positions of particles. Initial Speed The current optimal position is The global optimal position of the population is Etc. Randomly generate n p Initial particles and their velocities, n A The external archive A of capacity size generates n for each particle A / 4 capacity size of individual free particle external archives.

[0088] (2) Problem change detection mechanism. The problem change degree is calculated as δ according to the formula t , determine whether it is greater than the corresponding threshold ε, if so, go to (3), otherwise go to (4). The schematic diagram of the environmental change detection mechanism is shown in Figure 2 shown.

[0089] Assume that the degree of environmental change at time t is δ t , the calculation method is as shown in formula (17).

[0090]

[0091] In the formula, Represents individual x before and after the tth iteration i The change in the j-th target value is calculated as shown in formula (18).

[0092]

[0093] In the formula, It represents the mean change of the j-th target value of the environmental change detection group before and after the t-th iteration, and its calculation is as shown in formula (19).

[0094]

[0095] In the formula, and are the maximum and minimum values ​​of the jth target, respectively, m is the number of targets, n is D is the number of detected individuals. tIf it is greater than the set change threshold ε, the problem is considered to have changed.

[0096] (3) Environmental change response mechanism predicted by multiple representative points. t is the external archive, i.e. the currently obtained Pareto optimal solution set; n r is the number of representative points, and m is the number of targets.

[0097] (3.1) Representative point selection.

[0098] The first step is to obtain the representative points n according to formula (20) r , determines the number of prediction directions. Represents the number of points n r calculate:

[0099] n r =n r min +δ t (n r max -n r min ) (20)

[0100] In the formula, δ t is the degree of environmental change; n r max and n r min n r The maximum and minimum values ​​of r min =m+1,n r max =3m. From formula (20), we can see that n under different environmental change levels r The values ​​are different, so that the number of representative points can change adaptively with the environment.

[0101] The second step is to select representative points using the binary clustering method. According to formula (21), the center occupation The m boundary solutions obtained by equation (22) is the initial representative point, forming the initial representative point set In, n Sr for The current number of representative points; for each representative point Set the representative point member set C i， Where i = 1, 2, ..., n Sr Indicates the number of the representative point;

[0102] in, The solution is as shown in formula (21).

[0103]

[0104] Where n A A t The capacity size, A t The i-th individual in .

[0105] The mth boundary solution of the Pareto front found The calculation is as in formula (22).

[0106]

[0107] In the formula, is the minimum value of the mth target.

[0108] Step 3: Calculate A t Middle particles and The Euclidean distance between each representative point in Assign to the nearest representative point The member set C i middle;

[0109] Step 4: If n Sr ≥n r , output The algorithm ends. Otherwise, go to step 5;

[0110] Step 5: From the member set C i Find the individual farthest from each representative point, denoted as Assume that it belongs to the representative point The member set C i In From C i Take it out and store it in the representative point set Let n Sr =n Sr +1 and and As the representative point, i All individuals in the class are classified according to the principle of closest distance.

[0111] (3.2) Calculate the prediction direction

[0112] set up Any representative point First, select the representative point set from the previous moment Select distance The nearest representative point The prediction direction for:

[0113]

[0114] In the formula, and are the representative points at time t and t-1 respectively. and in, Number of and Number of Not necessarily equal.

[0115] Then, the available set of evolutionary guidance directions at time t is express.

[0116] (3.3) Based on the environmental change response mechanism predicted by multiple representative points, the predicted new population is obtained. K r is the random particle ratio, n Sp is the number of individuals in the new population.

[0117] The first step is to initialize a new initial particle population

[0118] The second step is to obtain the predicted individual set S according to formula (24): P1 . Predict individual set S P1 Individuals in Calculations such as:

[0119]

[0120] In the formula, for Guidance direction; N(0, σ t ) is The same-dimensional quantity, each dimension of which satisfies the mean of 0 and the variance of δ t Gaussian perturbation.

[0121] The third step is to start from S P1 Randomly pick K r (1-n Sp ) particles are stored in S p middle.

[0122] Step 4: Generate K r ·n Sp A random particle set S P2 , S P2 Deposit S p In, output S p .

[0123] (4) The static optimization algorithm iosMOPSO is used to obtain the Pareto optimal solution set at time t, that is, the external archive A. Assume T max is the maximum number of iterations), n p is the number of particles in the population, n A is the external archive capacity. The specific steps of the multi-objective particle swarm optimization algorithm iosMOPSO for optimal particle selection are as follows:

[0124] (4.1) Initialization. Let the iteration counter t = 0; randomly generate n zero initial positions and velocities in the decision space. p particles; according to MOP, n p Each particle calculates m objective function values ​​respectively; initializes individual external files for each particle And add the i-th particle to the corresponding make The set of initial particles and S p With A t Input to (2) to update, and assign the result to A t ;

[0125] (4.2) External file updates.

[0126] The first step is to set the new solution Merge with the external file Ar into A′, let

[0127] The second step is to perform non-dominated classification on A′ and obtain each individual The non-dominant level R i , and then calculate the individual crowding distance C i .

[0128] The third step is to calculate the value of the matrix of each individual in A′. Sort from best to worst, R i The smaller the individual, the better. i In the same case, C i The larger the individual, the better. Take out the items from A' and store them in A in descending order. t+1 , until or A t+1 Archive storage limit reached.

[0129] Step 4: Individual crowding distance C i Calculation

[0130] The capacity at time t is n A External file A t Medium The steps for calculating the crowding distance are as follows:

[0131] 1) Initialize the crowding distance, that is, any individual The crowding distance is assigned an initial zero value: C i =0, i=1, 2, ..., n A ;

[0132] 2) The set of j-th objective function values ​​of all individuals Sort, j = 1, 2, ..., m. Find the individual The objective function value is at the position k after sorting, individual The crowding distance C of the jth target i，j Calculate as formula (25);

[0133]

[0134] In the formula, C i，j For individuals The crowding distance at the jth target.

[0135] Step 5: Calculate individual The crowding distance

[0136] (4.3)t=t+1

[0137] (4.4) Global optimal particle Alternating selection strategy. The schematic diagram of the principle of the global optimal particle alternating selection strategy is as follows Figure 3 shown.

[0138] The first step is to calculate Is it less than the threshold ε? If not, keep unchanged, that is Go to (5).

[0139] The second step is to t Each solution According to formula (25) (26), the congestion distance C is calculated respectively j and dominant strength S j .

[0140] Among them, the dominant strength S j The calculation of is as follows:

[0141] A t Any particle The convergence of can be measured using the dominant strength S i To measure, the calculation method is:

[0142]

[0143] In the formula, m represents the number of targets; n A Indicates the external file capacity; means rounding up; |g| means taking the absolute value; S i Reflects Compared to A t The degree to which other particles in the PF approach the true PF can be used as a convergence measure, S i The larger it is, the better the convergence.

[0144] The third step is to t middle According to C j and S j Sort in descending order and store in the best particle set C for diversity and the best particle set S for convergence.

[0145] Step 4: Global Optimal Particle Alternating selection strategy, calculated according to formula (27)

[0146]

[0147] In the formula, rand represents a random value among 1, 2, .., m, m represents the number of targets; S represents the set of particles with the best convergence performance, S rand represents the randth particle in S; C represents the optimal particle set for diversity, C rand represents the randth particle in C;

[0148] The fifth step is to output the global optimal position:

[0149] (4.5) Individual optimal particle We designed a method for selecting the best individual particles with diversity preservation and fast tracking. First, we used the individual external archives Save the Pareto optimal solution set in the search history of individual particles, and its update still adopts the external archive update strategy. Capacity A t 1 / 4 of the capacity, i.e. n A / 4. Then from the individual external archive of the i-th particle Select the global optimal particle with the i-th particle The nearest particle is taken as the individual optimal particle of the i-th particle

[0150] (4.6) Update each particle Speed ​​and position The particle motion equations (28) and (29) are used to update each particle Speed ​​and position

[0151]

[0152]

[0153] (4.7) Based on the multi-objective constant drilling pressure automatic drilling PID parameters, each particle is optimally evaluated. m objective function values ​​(m=2);

[0154] In the first step, the mathematical model for optimizing the PID parameters of the automatic drilling feed with constant drilling pressure based on multiple objectives is established as follows:

[0155]

[0156] In the formula, k = [k p1 k i1 , k p2 ] T is the decision vector; t is the running time of the current system; J1=t r (S(k, t)) is the rise time of the S(k, t) WOB step response; J2 represents the cumulative absolute value error of the S(k, t) WOB step response; g i (k), h j (k) represent the i-th inequality and j-th equality constraints on decision variable k, respectively.

[0157] In the second step, the interference signal is added, and the constant drilling pressure automatic drilling PID parameters are dynamically optimized with equation (29) as the optimization target. The friction coefficient between the brake caliper and the brake disc is set to change every 10 seconds to simulate the fluctuation of drilling pressure caused by interference factors such as formation changes, changes in the friction coefficient of the brake disc, and drill bit wear.

[0158] Assume that the friction coefficient μ between the brake caliper and the brake disc changes with time in the form of:

[0159]

[0160] Since the constant drilling pressure automatic drilling system has a large hysteresis, the object hysteresis and time constant are large, the interference is strong and frequent, and the load changes greatly, cascade PID control is adopted. The outer ring is the drilling pressure ring, and the inner ring is the drilling speed ring. The hydraulic disc brake constant drilling pressure automatic drilling control block diagram is as follows Figure 4 The parameter values ​​of hydraulic disc brake constant drilling pressure automatic drilling are shown in Table 2.

[0161] Table 2 Hydraulic disc brake constant drilling pressure automatic drilling parameter values

[0162]

[0163]

[0164] The introduction of the inner ring drilling speed ring is to improve the characteristics of the object, speed up the control process, and have a leading control effect, thereby effectively overcoming the lag and improving the control quality. It belongs to follow-up control, so the inner ring adopts pure P control.

[0165] The outer ring drill pressure ring is used to achieve fixed value control of drilling pressure. The outer ring adopts PI control. The constant drilling pressure automatic drilling system has a certain adaptive ability due to the addition of a drilling speed ring, and can be used in occasions with large changes in load and operating conditions.

[0166] (4.8) For each particle Will, and Input to (4.2), update the individual external file to obtain

[0167] (4.9) The set of all particles and S p With A t Input to (4.2), update the external file to get A t+1 ;

[0168] (4.10) External file Cauchy distribution perturbation strategy, update A t+1 .

[0169] The first step is to t Each particle Perform the operation of formula (32) to obtain the set S p .

[0170] The second step is to t and S p Input into (4.2) and assign the result to A t .

[0171]

[0172] (4.11) If t>Tmax, the Pareto optimal solution set at time t is obtained, that is, the external archive A t Then the output A=A t ; Otherwise, jump to (4.3);

[0173] (5) Determine whether the optimization termination condition is met. If so, output the optimal front end of the constant drilling pressure automatic drilling PID parameters. Otherwise, go to (2).

[0174] The PID parameter solutions found by mrpDMOPSO and MOPSO under different drilling conditions are shown in the figure below: Figure 5 The drilling pressure control performance indicators set by different methods at different times are shown in Tables 3 to 5.

[0175] Table 3 Performance indicators of drilling pressure control under different setting methods (t=10)

[0176]

[0177] Table 4 Performance indicators of drilling pressure control under different setting methods (t=20)

[0178]

[0179] Table 5 Performance indicators of drilling pressure control under different setting methods (t = 30)

[0180]

[0181] The present invention innovatively designs an environmental change detection mechanism and an environmental change response mechanism predicted by multiple representative points. According to the severity of environmental changes, a certain number of representative points are adaptively allocated to track environmental changes and guide particles to search. At the same time, a global optimal particle alternating selection strategy is designed. According to the crowding distance and dominant strength, the diversity and local search capabilities of the algorithm are balanced, and a global optimal particle selection method that balances convergence and diversity and an individual optimal particle selection method that maintains diversity and fast tracking are introduced. An external archive disturbance strategy is used to construct a multi-objective particle swarm optimization algorithm iosMOPSO for optimal particle selection. A dynamic multi-objective particle swarm optimization (mrpDMOPSO) algorithm based on multi-representative point prediction is designed. A multi-objective constant drilling pressure automatic drilling PID parameter optimization model is established, and the mrpDMOPSO algorithm is applied to the rise time J1=t of the drilling pressure step response. r (S(k, t)) and the cumulative absolute value error J2 of the drilling pressure step response are optimized to achieve the automatic drilling of constant drilling pressure under complex working conditions. PID parameter k = [k p1 k i1 , k p2 ] T Optimization can adapt well to different formation conditions, achieve stable and precise control of drilling pressure, and improve drilling success rate, drilling efficiency and quality.

[0182] The experimental results of using mrpDMOPSO to optimize the PID parameters of the automatic drilling with constant drilling pressure show that the traditional multi-objective optimized automatic drilling PID parameters have large deviations in the actual drilling pressure control. The PID parameter results optimized by the method proposed in the present invention overcome the defects of the existing methods and obtain multiple groups of PID parameters with better control performance. The obtained PID parameter optimal solution set can not only track the changes in the drilling environment, but also be closer to the optimal frontier. At the same time, the absolute value integral of the deviation IAE and the dynamic response overshoot of the system are significantly reduced. It meets the optimization requirements of the PID parameters of the actual complex drilling conditions and the constant drilling pressure control system with multivariable, time-varying, strong disturbance and time lag, and is more conducive to the stable and accurate control of the drilling pressure at the drill bit under different formation conditions.

[0183] A set of optimized PID parameters are randomly selected to observe the drilling pressure step response of the constant drilling pressure automatic drilling under different drilling environments. The response results further prove that the optimal solution set of PID parameters obtained based on the constant drilling pressure automatic drilling PID parameters optimized by mrpDMOPSO can not only track the changes in the drilling environment, but also be closer to the optimal frontier. At the same time, the absolute value integral of the error IAE is reduced by up to 68% compared with the suboptimal IAE, and the overshoot is reduced by up to 63% compared with the suboptimal overshoot. Therefore, the constant drilling pressure automatic drilling based on the PID parameter optimization of mrpDMOPSO can achieve a better control of drilling pressure.

[0184] Through the study of the mrpDMOPSO method for automatic drilling under constant drilling pressure, a theoretical foundation is laid for further realizing intelligent decision-making control of automatic drilling during drilling, intelligent drilling rig equipment, and automation of the drilling process, thereby improving drilling quality, drilling success rate, and drilling efficiency, reducing the occurrence of drilling accidents, and reducing drilling costs.

Claims

1. A method for optimizing control parameters of automatic drilling feed with constant drilling pressure based on mrpDMOPSO algorithm, characterized in that: The following steps are involved: (1) Initialization: including mrpDMOPSO algorithm parameter definition and population initialization; mrpDMOPSO parameter definition is: population size n p , external file capacity n A , inertia weight ω, acceleration factor c1=c2 and maximum number of iterations T max ; Population initialization: Population initialization includes the initial positions of particles Initial Speed The current optimal position is The global optimal position of the population is (2) Problem change detection mechanism: According to formula (1), the problem change degree is calculated as δ t , determine whether it is greater than the corresponding threshold ε, if so, go to step (3), otherwise go to step (4); In formula (1), m is the number of targets, n is D To detect the number of individuals, Represents individual x before and after the tth iteration i The change in the jth target value; It represents the mean change of the jth target value of the environmental change detection group before and after the tth iteration; (3) Environmental change response mechanism predicted by multiple representative points: The number of representative points is calculated according to formula (2), the prediction direction is selected according to formula (3), and the predicted new population is obtained according to formula (4) based on the environmental change response mechanism predicted by multiple representative points; n r =n r min +d t (n r max -n r min ) (2) In formula (2), n r is the number of representative points, n r max and n r min n r The maximum and minimum values ​​of n r min =m+1,n r max =3m; in formula (3), Guiding the direction of evolution at time t The set of prediction directions, The prediction direction and Representative points of each and for Any representative point in; In formula (4), is the individual set S at time t P1 The i-th individual in N(0,σ t ) is The same-dimensional quantity, each dimension of which satisfies the mean of 0 and the variance of δ t Gaussian perturbation of The predicted individual set S for the calculated time t+1 P1 The i-th individual in ; (4) The static optimization algorithm iosMOPSO is used to obtain the Pareto optimal solution set at time t, that is, the external archive A; the specific steps of the multi-objective particle swarm optimization algorithm iosMOPSO for optimal particle selection are as follows: (4.1) Initialization: Set the iteration counter t = 0; randomly generate n objects with initial position and velocity of zero in the decision space. p Particles; (4.2) External file update: New solution With external archive A t Merge into A′, let Then calculate the individual crowding distance C i ; (4.3)t=t+1; (4.4) Global optimal particle Alternating selection strategy: A t Any particle The convergence of the dominant strength S using formula (5) i To measure; According to formula (6) to achieve the global optimal particle Alternating selection strategy; In formula (5), m represents the number of targets; n A represents the external archive capacity; in formula (6), S represents the optimal particle set for convergence, S rand represents the randth particle in S; rand represents a random value among 1, 2, .., m, and m represents the number of targets; C represents the optimal particle set for diversity, C rand represents the randth particle in C; (4.5) Individual optimal particle choose; (4.6) Update each particle according to equations (7) and (8) Speed and location In formula (7), is the position of the particle at time t, ω is the inertia weight, which is set as a parameter; is the velocity of the particle at time t; (4.7) Based on the multi-objective constant drilling pressure automatic drilling PID parameters, each particle is optimally evaluated. The two objective function values ​​are as follows, i.e., the rise time and the cumulative absolute value error are taken as the target min J(k, t) = [J1(k, t), J2(k, t)] T ; According to the interference signal added in formula (9), the dynamic response of the system is analyzed; (4.8) For each particle Will, and Input to (4.2), update the individual external file to obtain in, is the individual external file corresponding to the addition of the i-th particle at time t The corresponding individual external file after adding the i-th particle at time t+1 (4.9) The set of all particles and S p With A t Input to (4.2), update the external file to get A t+1 ; (4.10) External archive Cauchy distribution perturbation strategy; (4.11) If t>T max , and obtain the Pareto optimal solution set at time t, that is, the external archive A t Then the output A=A t ; Otherwise, jump to (4.3); (5) Determine whether the optimization termination condition is met. If so, output the optimal front end of the constant drilling pressure automatic drilling PID parameters, that is, output multiple sets of constant drilling pressure automatic drilling PID parameters k = [k p1 k i1 , k p2 ] T Optimize the result; otherwise go to step (2); The optimized PID parameter optimal solution set can not only track the changes in the drilling environment, but also be closer to the optimal frontier; at the same time, it significantly reduces the absolute value integral of the deviation IAE and the dynamic response overshoot of the system, meeting the optimization requirements of the PID parameters of the constant drilling pressure control system with multivariable, time-varying, strong disturbance and time lag in the complex drilling process, and is more conducive to the stable and precise control of the drilling pressure at the drill bit under different formation conditions.

2. The method for optimizing the control parameters of the constant drilling pressure automatic drilling based on the mrpDMOPSO algorithm according to claim 1 is characterized in that an environmental change response mechanism predicted by multiple representative points is established, and a certain number of representative points are adaptively allocated to track environmental changes according to the severity of environmental changes, so as to realize dynamic multi-objective optimization of the constant drilling pressure PID parameters; In step (4.7), the objective function established is: In formula (10), k = [k p1 k i1 , k p2 ] T is the decision vector; t is the running time of the current system; J1=t r (S(k, t)) is the rise time of the S(k, t) drilling pressure step response; J2 represents the cumulative absolute value error of the S(k, t) drilling pressure step response; g i (k), h j (k) represents the i-th inequality and j-th equality constraints on decision variable k respectively; Formula (1) realizes the problem change detection mechanism, Formulas (2) to (4) realize the environment change response mechanism of multi-representative point prediction, Formulas (5) to (8) realize the multi-objective particle swarm optimization algorithm iosMOPSO for optimal particle selection, and Formulas (1) to (10) search for T max The Pareto optimal solution set at the time, that is, the external archive A t Then the output A=A t , and obtain multiple groups of constant drilling pressure automatic drilling PID parameters k = [k p1 k i1 , k p2 ] T Optimal front end; select any group of optimal front ends k = [k p1 k i1 , k p2 ] T As the PID control parameter of hydraulic disc brake constant drilling pressure automatic drilling, it can realize stable and accurate control of drilling pressure at the drill bit under different formation conditions.

Images