RU2208153C2 - Drilling process control system - Google Patents

Abstract

FIELD: well drilling; applicable in automatic control of drilling process. SUBSTANCE: system has downhole motor, mud pump, downhole telemetric system, surface transducers of process parameters monitoring, information transmission unit and control computer with software and control database containing information on design well path. Introduced into control system is database of cad system which has design data on well structure, equipment characteristics, drilling process technology, geological and geophysical data. Software is made for self-learning by accounting of data contained in SAD system database of given well and consideration of experience gained before in drilling wells of same cluster or deposit in decision making, and realization with use of artificial neural networks. Artificial neural networks have similar cells simulating operation of brain neurons, possessing a group of synapses forming inlet links, and axons forming outlet link of given cell. EFFECT: higher accuracy in well path drilling and prevented emergency situations in drilling process. 2 cl, 2 dwg

Description

 The invention relates to drilling technology, particularly to automatic drilling control systems.

 A known system and method for controlling the drilling process according to the patent of the Russian Federation 2067170, containing sensors for monitoring the main parameters: axial force, well depth, vibration sensor of the drill string and electronic computer equipment for comparing the measured basic parameters with the set according to the control program. The main purpose of this system is to ensure an automatic transition from the Drilling mode to the Drilling mode. The transition from one mode to another is carried out according to the readings of the vibration sensor. Vibrations of the drill pipe string sharply increase when the surface of the rock block touches with a chisel. The system does not fully automate the drilling process.

Also known is a system and method for controlling the drilling process by AS USSR 1231946 (prototype), containing parameters control sensors, data transmission unit, data processing unit with software and a control database. The management database contains:
- design values of the load on the bit;
- rotor speed;
- mud flow rate;
- design drilling time;
- well depth.

 The disadvantages of the system are that it does not provide the required level of information reliability to determine the situation for making the optimal decision by automating the drilling process, mainly due to insufficient database capacity and the lack of a system for recording the experience of previous work and the system’s self-learning.

 The objective of the invention is to provide a high degree of reliability of determining the controlled situation for full automation and optimization of the drilling process.

 The solution to these problems was achieved due to the fact that the drilling process control system containing a downhole motor, a pump for pumping flushing fluid, a downhole telemetry system, ground sensors for monitoring process parameters, an information transfer unit and a control computer with software and a control database that contains information about the design trajectory of the well, additionally contains a CAD database, which contains design data on the design of the well, characteristics of the equipment drilling technology, geological and geophysical data, while the software is self-learning by taking into account the CAD data of the well and taking into account previous experience gained when drilling wells in the same cluster or field and implemented using artificial neural networks . Artificial neural networks contain cells of the same type that simulate the functioning of brain neurons, which have a group of synapses that form the input connections, and an axon that form the output connection of this cell element.

 The solution of these problems in the method of controlling the drilling process, including the operation of measuring downhole and ground technological parameters and making a decision by the control computer, is achieved by taking this decision taking into account the CAD database and previous experience in drilling wells in the same cluster or field and based on drilling experience wells, the CAD design is being adjusted.

 In FIG. 1 shows a general diagram of the system. In FIG. 2 - cell neural network.

 The drilling process control system comprises a downhole motor 2 mounted in the lower part of the drill pipe string 1 with a rock cutting tool 3 and a downhole telemetry system 4 with an electromagnetic or hydraulic communication channel. The downhole telemetry system contains sensors for monitoring parameters in the face 5, 6, 7 and 8. These sensors monitor, for example, azimuth angles, horizon, position of the diverter and logging indicators. Ground-based sensors for monitoring process parameters 9 and 10 can monitor any parameters, for example, the pressure of the flushing fluid in the pump line 11, the vibrations of the drill pipe string 1. The sensors 9 and 10 are connected to the control computer 13 through a receiving device 12.

 The control computer 13 contains software 14, a management database 15, and a CAD database 16.

 The management database 15 contains information about the design trajectory of the well. The CAD database (Drilling Complex Automated Design Systems) contains the design data for the entire complex: well design, equipment characteristics, drilling process technology, geology, geophysical data, etc. The outputs from the computer are connected to the respective actuators (actuators of the pump 11, winch, preventer, etc. (not shown in Figs. 1 and 2).

 Software 14 is implemented using artificial neural networks. Artificial neural networks (Fig. 2) contain cells of the same type that simulate the work of brain neurons, which have a group of synapses that form input links 18 ... 21 and the output link - axon 22. Each cell element is characterized by its current state by analogy with brain nerve cells brain that may be agitated or inhibited. Each synapse is characterized by the magnitude of the synaptic connection or its weight Wi, which is equivalent in physical meaning to electrical conductivity.

The current state of a neuron is defined as the weighted sum of its inputs
S = X • Wi.

The output of a neuron is a function of its state:
Y = f (s).

Очевидно, что процесс функционирования НС (нейронной сети), то есть сущность действий, которые она способна выполнять, зависит от величин синаптических связей, поэтому задавшись определенной структурой НС, отвечающей какой-либо задаче, при разработке сети необходимо найти оптимальные значения всех переменных весовых коэффициентов (некоторые синаптические связи могут быть постоянными).

Obviously, the process of functioning of the NS (neural network), that is, the essence of the actions that it is able to perform, depends on the values of synaptic connections, therefore, having given a certain structure of the NS corresponding to any task, when developing a network, it is necessary to find the optimal values of all variable weighting coefficients (some synaptic connections may be permanent).

 This stage is called "training of the National Assembly", and the ability to solve the problems posed to it during operation depends on how well it will be completed. The work of all networks is reduced to the classification, generalization of input signals belonging to n-dimensional hyperspace, according to a number of classes. From a mathematical point of view, this happens by splitting hyperspace with hyperplanes (entry for the case of a single-layer perceptron).

Learning Algorithm with a teacher:
1. Initialize the elements of the weight matrix (usually with small random values).

 2. Submit to the inputs one of the input vectors, which the network must learn to distinguish, and calculate its output.

 3. If the output is correct, go to step 4.

4. Otherwise, calculate the difference between the ideal and the obtained output values: δ = y ₁ -y.

Modify weights according to the formula:
W _ij (t + 1) = W _ij (t) + v (t) • δX _i ,
where t are iteration numbers;
v is the coefficient of learning speed;
i is the input number;
j is the number of neuron in the layer.

 5. Repeat the cycle from step 2 until the network stops making mistakes.

 During the operation of the system, sensors 5 ... 8, 9 and 10 register the most important parameters characterizing the drilling process. Information about the registered downhole parameters is transmitted through the transmitting device 4 to the receiving device 12. The parameters from the ground sensors 9 and 10 are transmitted to the receiving device 12. After converting the information about the parameters, it is transmitted to the control computer 13, where it is analyzed using software 14, and depending from the decision made by the neural network, a control signal is sent to the output of the control computer, which carries out specific actions to change the drilling mode, For example, changing the capacity of the pump 11, thereby adjusting the power and the rotational speed of the engine 2 with the downhole rock cutting tool 3. There may be any other form of influence on the drilling process, such as a winch, etc. preventer At the same time, it is advisable to display on the monitor screen the most important measurement results and commands generated by the system. The monitor of FIG. 1 and 2 are not shown.

 The implementation of the method is as follows.

 Downhole parameters are recorded by sensors 5 ... 8, and technological parameters are recorded by ground-based sensors for monitoring technological parameters 9 and 10. Information is transmitted to the receiving device 12 and then to the control computer 13. Software 14 is pre-installed on the hard drive of the control computer 13, database control 15 and CAD database 16. The control computer 13 makes a decision taking into account the information stored in the CAD database 16. Information about all parameters from sensors 5 ... 10 and the accepted control actions Rep also recorded on the hard disk 13. The control computer (HDD in Fig. 1 and 2 are not shown). When drilling subsequent wells, recorded information is taken into account for decision making. Decisions are made by the control computer 13 and taking into account previous experience in drilling wells of the same well or field that was recorded in the memory (on the hard disk) of the control computer 13. For example, in the same situations, the same decisions are made to correct the well path or to abort drilling operations. Experience in drilling wells of the same cluster is the most valuable and in our case it can be transferred from one drilling rig to another on a computer diskette or through communication channels without transferring qualified specialists to the newly developed well. This is one of the main elements of the system’s self-learning. The second element of self-learning is the technological features of neural networks and is described earlier in the section "Learning Algorithm with a Teacher".

 In addition, information recorded and obtained as a result of computer processing can be transmitted to correct the CAD design of the well. For example, according to the logging results obtained from the logging sensor 8, the presence of a significant amount of hydrocarbons, i.e. rock cutting tool 3 entered the oil reservoir at a depth significantly less than design. In this case, the CAD database (well design) is adjusted to reduce the design depth of the well and the drilling process is terminated.

 It is also possible emergency shutdown of systems in emergency situations, for example, according to the readings of the torque sensors on the rock cutting tool 3 or by a sharp change in axial force.

The application of the invention allowed:
1. Create a self-learning system by using a CAD database and neural networks.

 2. Fully automate the process of drilling wells.

 3. To increase the accuracy of wiring directional and horizontal wells.

 4. Optimize the well construction process.

 5. Correct the project (CAD) based on the results of well drilling.

 6. Provide emergency prevention.

 7. Use the drilling experience of wells of the same cluster to adjust the drilling process and make decisions in emergency situations.

 8. Promptly exchange information between drilling rigs while conducting drilling operations or with a slight time shift of technological cycles.

Claims (2)

 1. A drilling process control system comprising a downhole motor, a pump for pumping flushing fluid, a downhole telemetry system, ground-based process control sensors, an information transfer unit and a control computer with software and a control database that contains information about the project well trajectory the fact that the system additionally contains a CAD database (computer-aided design system), which contains design data on the well design , equipment characteristics, drilling process technology, geological and geophysical data, while the software is self-learning by taking into account the data contained in the CAD database of the well, and taking into account previous experience obtained earlier when drilling wells in the same cluster or field, with decision making, and implemented using artificial neural networks.  2. The drilling process control system according to claim 1, characterized in that the artificial neural networks contain cells of the same type that simulate the operation of brain neurons, having a group of synapses forming the input connections, and an axon forming the output connection of this cell element.

Images