Related problems analysis of electric submersible pump system control in oilfield

Abstract: Based on the engineering practice, the problems existing in the control of electric submersible pump in oil field are summarized. According to the characteristics of the controlled object, the available control strategies are analyzed, and the control of the system with humanoid intelligent control algorithm shows that the proposed control method is satisfactory.

1 Introduction

As we all know, electric submersible pump is one of the more oil recovery equipment in oilfield. In essence, the electric submersible pump is a multi-stage centrifugal pump working in the well, with the tubing into the downhole, the ground power through the transformer, control panel and electric submersible pump dedicated cable to the underground electric submersible pump motor, the motor driven Multi-stage centrifugal pump rotation, the electrical energy into mechanical energy, the liquid wells up to the ground.

There are two main problems with ESP application, one is how to save energy, and the other is how to control ESP to make it work in the best conditions. As the electric submersible pump is below the ground 2Km bottom of the work, the working environment is very harsh (high temperature, strong temperature, etc.), the general use of traditional power supply, that work under the full power frequency, and thus frequent failures, high operating costs. On the one hand, when the electric submersible pump starts at power frequency, the starting current is large and the voltage drop of the motor cable is large, so that the reverse voltage of the motor cable in the starting process is higher and the insulation performance of the cable is reduced. Each time the power is turned on, life. Electric submersible pump repair costs only as much as a project as much as 50,000 yuan, the value of 100,000 yuan on average put on the cable to be replaced 5 times to be replaced, the average electric submersible pump to repair once every 10 months, the maintenance cost to be 80,000 yuan, so that operating costs increase. Electric submersible pump on the other hand, under normal work, the prevalence of the motor load rate than teach low case, "big horse car" phenomenon is serious, resulting in a huge waste of electricity. In addition, the electric submersible pump to reduce power factor, power consumption, work frequency, ESP always work at rated speed, if the amount of fluid in short supply, easily lead to "dead wells" in the event of dead wells, the heavy losses. The correct solution is that the ESP should be able to adjust the pumping capacity according to the geological conditions to balance supply with demand. However, the traditional adjustment method is to change the nozzle to adjust the output, which not only results in the waste of energy but also can not be precisely controlled. Sometimes makes the motor and pump running at high pressure for a long time; sometimes make the well out of the sand serious, shorten the life of the equipment, so the field of electric submersible pump system control related technical issues a brief discussion is necessary. This article mainly discusses the characteristics of the controlled object, the selection of control strategy, the control algorithm and the application of VVVF technology.

2 Oilfield characteristics

Oil production, which controls the field as a special controlled object, must first understand certain characteristics of the field during its production. One of the most important is time-variability, due to the complex and changeable geological conditions, to sum up the following features. For the ESP system, from a macro point of view, the characteristics of controlled objects are mainly reflected in the following aspects:

(1) the unknown, time-varying, randomness and dispersion of system parameters;

(2) the system lag unknown and time-varying;

(3) serious non-linear system;

(4) the relationship between the system variables;

(5) The unknown, diversity and randomness of environmental interference.

These features bring many problems to system modeling and control.

3 control problems

Faced with the above characteristics, traditional control is powerless due to its control over complex objects (or processes) that are subject to uncertainty:

(1) Certainty issues

Traditional control (such as PID) is based on mathematical model control, that is, control, object and interference models are known or can be obtained by identification. However, many of the control problems in oilfield systems are uncertain and often even abruptly change. Control problems that are "unknown", uncertain, or poorly understood can not be modeled by traditional methods and therefore can not be effectively controlled.

(2) highly nonlinear

In the traditional control theory, although some non-linear methods are available to control objects with high degree of nonlinearity, on the whole, nonlinear theory is far less mature than linear theory because the method is too complicated and difficult to be applied. There are a lot of nonlinear problems in oilfield system.

(3) semi-structured and unstructured issues. The traditional control theory mainly uses the differential equation, the state equation and the various mathematical transformations as the research tools. The essence of the traditional control theory is a kind of numerical calculation method, belonging to the quantitative control category. It requires that the control problem has a high degree of structuring and is easy to be described by a quantitative mathematical method or Modeling. However, the most concerned and needed support in the oilfield system is sometimes the semi-structured and unstructured problems.

(4) system complexity issues

According to the viewpoint of system engineering, the generalized object should include the operating object and the environment in the usual sense. However, the relationship among subsystems in the oilfield system is complicated. The coupling and mutual restraint of each element are highly complicated and sometimes unpredictable. Traditional control lacks an effective solution.

(5) reliability issues

The conventional mathematical model-based control problem tends to be an interdependent one. Although the system based on this method often has the contradiction between robustness and sensitivity, the reliability of the control of the simple system is not conspicuous. For oilfield systems, if the above method is used, the entire control system may collapse due to the change of conditions.

Thus, the traditional method can not be effective for oilfield system control, we must explore more effective control strategies and methods.

4 system modeling problems

Oilfield systems are characterized by classical mathematics that have not been considered. Although probability theory deals with uncertainty, it has its own underlying assumptions. These basic assumptions limit its use in expert systems and of course limit its use in other areas. It can only deal with problems that contain randomness, neither do I know nor blur. In fact, the four kinds of uncertainty information people find at present are gray information and unidentified information besides random information and fuzzy information. These four kinds of uncertain information are often presented at the same time in an oilfield system or simultaneously. At the same time, they affect people's understanding of system features and functions and affect people's research, management and control of oilfield systems. Moreover, no matter from the connotation of the concept or the axiomatic system and the theory of set theory, the four kinds of uncertainty information have the necessary connection. Therefore, to establish the basic control model describing the oilfield system and realize the comprehensive treatment of the oilfield information in the control system is a difficult and urgent problem to be solved urgently.

Control is the need to model, but the control model is not the same mathematical model described by strict mathematical expressions. The essence of the model is to describe the nature of things, the description can be described in strict mathematical way, called mathematical model; also can be described in language, called the language model; as well as framework models, logical models, etc., according to the object The complexity of the decision whether to choose which description more reflects the nature of the object. In complex oilfield systems, methods of combining qualitative and quantitative methods are often used. Such models have the following characteristics:

(1) The integrity of system information

Known information and unknown information live together, all kinds of uncertainty information together; certainty information and uncertainty information together. They are interrelated, mutually influential, mutually restrictive, and under certain conditions, mutual transformation, but the total quantity will not change.

(2) the dynamics of system development

Like ordinary things, the system of uncertainty and its factors are all functions of time. They all change over time, develop, decay, and transform.

(3) Observability of system information

The process of human cognition of things is not only the process of acquiring information, but also the process by which humans measure each factor in the system by using objective criteria and scales (which may be collectively referred to as scales) formed in practice. Because the generation of uncertainty information is the inevitable result of material movement, it must be followed by law, observable, and recognizable.

(4) the hierarchy of system information

The system can be divided into different levels. In the macroscopic view, it is uncertain information, and at the micro level, relative certainty information can be separated. With the deepening of levels, people's understanding of the system is more profound.

(5) gray system information

Uncertainty information is observable and can increase observability as the hierarchy deepens. But "uncertain information is inevitable." Therefore, the uncertainty information is not completely known, only partially known, partially unknown, although the unknown part can be deepened and narrowed with the measurement level; and there is also information loss in the known information. This part of the known (white), part of the unknown (black) phase relationship called gray sex. 5 control strategy selection

Modern control theory developed in the 1950s, whether it is state space law or black box method based on I / O description, accurate mathematical description is the basis of its analysis and design system. If the mathematical model of the object (or process) is not known, then it must first be modeled mathematically, but whether it is optimal control or adaptive control, the premise of the discussion is to require an accurate mathematical model, and obviously not for complex systems in the field The above conditions, for the author's discussion of the control system, should not be chosen as a control strategy.

(1) Artificial Neural Network (ANN)

Originated in the 1940s, it reflects the basic features of the human brain from some aspects, but it is not the true description of the human brain but only its abstraction, simplification and simulation. The information processing of the network consists of inter-neuronal interactions to fulfill. The key to neural network control is to select a suitable neural network model and train and learn it until it meets the requirements, that is to find the optimal neural network structure and weights. However, neural network learning requires a certain amount of experimental samples, but it also needs to be run thousands of times to obtain the best structure. Sometimes obtained is a local optimal solution, rather than the global optimal solution, due to the limitations of the method, it is also difficult to achieve effective control of the oil field object discussed in this article.

(2) Expert Control System (ECS)

It is a set of computer programs that are based on knowledge and attempt to simulate people's thinking and behavior in a special field and can process various qualitative, quantitative, precise and fuzzy information. Therefore, based on the experience of the controlled process and Knowledge acquisition, take different forms of description, in order to more reflect the characteristics of the object, to provide control strategies and control modalities. After the dynamic information of the controlled process is extracted and processed and the pattern recognition is performed on the feature information, it is sent to the reasoning institution on the one hand and useful information is added to the knowledge base on the other hand. Reasoning agencies based on feature information and knowledge provided by the knowledge base to judge, reason, and the results sent to the control agencies, so as to give the appropriate control output, the controlled process implementation of control. However, because of the collection of feature information, the expression of feature information and the establishment of a complete knowledge base, it is difficult to implement. Therefore, the expert system is not necessarily a good choice.

In the actual project, a very skilled operator, with his rich practical experience, can obtain more satisfactory control effects by judging the various phenomena in the field. If the experience of the measures taken into the corresponding control rules, and the development of a controller instead of these rules, which can achieve the control of complex industrial processes. Practice has proved that fuzzy controller based on fuzzy control theory (FC) can accomplish this task.

Man's control experience is summarized and described in human language. The language is the shell of thinking, it has a great ambiguity. For example, to maintain the water level in a water tower, the water level can be stabilized at a fixed point by adjusting the water pump valve opening. According to human experience can have the following control rules:

If the water level is higher than the fixed point, the drainage, if the difference is greater, then the drain valve open large, the faster the drain, if the difference is smaller, the drain valve open small drainage more slowly;

If the water level is lower than the fixed point, the water supply, if the difference is greater, then the water valve to open large, water faster, if the difference is smaller, the water valve to open small, water supply is more slow.

In the above description of the operating experience of the language, "higher than", "lower than", "open big", "open small" and other words with some ambiguity, it must be fuzzy set of fuzzy mathematics to describe These vague language, and use IF condition THEN action statement to be realized. Its core is to establish a mathematical model of linguistic analysis of complex systems or processes so that the natural language in daily life can be directly translated into the algorithmic language accepted by the computer. It provides a powerful tool for dealing with ambiguities that already exist in the objective world. The application of fuzzy control technology in China has achieved remarkable results. Although it is still in the stage of continuous improvement and development, its control quality and effect are satisfactory. It is an alternative strategy for oilfield systems.

6 control model and control algorithm

6.1 Control model

Due to the complexity and uncertainty of the controlled object, it is impossible to establish a rigorous mathematic model of the oilfield system according to the traditional method. A non-mathematical generalized control model based on knowledge representation can be adopted as shown in FIG. 1.

In fact, people do not know nothing about the object, but do not know it, that is, the object information has the gray information. In practical engineering, relying on the knowledge and experience of control experts, the system can be effectively controlled and satisfactory results obtained. It is based on human intelligence, the controller's experience and skills, to complete the scheduled control tasks. So it is a man-machine control

Model, that is, the controller (person) model and the controlled object (machine) combination of the model, such as the controller's knowledge model and the controlled object mechanism model combined with the generalized control model.

6.2 Control Algorithm

The basic idea is to mimic the general behavior of experienced operators in process control systems, such as when the system's error tends to increase (closed-loop control); when the system error tends to decrease, the control action is canceled and waiting Observe and so on. The more people understand the state, dynamic characteristics and behaviors of the accused system, the better the control will be. If en represents discretization of the current sampling time error value, en-1 and en-2, respectively, said the previous and the first two sampling time error, then there

Δen = en-en-1

Δen-1 = en-1-en-2

From the two basic features of error e and error change Δe, more characteristic information can be obtained from the dynamic process.

(1) e · Δe

The product of the error e and the error Δe constitutes a new characteristic variable describing the dynamic process of the system. By using whether the value of the characteristic variable is greater than zero, the trend of the dynamic process of the system can be described.

When en · Δen <0, it indicates that the dynamic process of the system is changing in the direction of decreasing the error, that is, the absolute value of the error gradually decreases.

When en · Δen> 0, it indicates that the dynamic process of the system is changing in the direction of increasing error, that is, the absolute value of error increases gradually.

In the process of control, recognizing the sign of en · Δen can grasp the behavioral characteristics of the dynamic process of the system in order to better formulate the next control strategy.

(2) Δen · Δen-1

The product of two adjacent errors Δen · Δen-1 constitutes a feature quantity that indicates the extreme value of the error. If Δen · Δen-1 <0 indicates the extreme value, Δen · Δen-1> 0 expresses the infinite value .

(3) │ Δe / e│

The magnitude of the absolute value of the error change, Δe, to the error, e, describes the attitude of the error in the system dynamics.

Using │Δe / e│ together with e · Δe, the dynamic process can be further divided. Through this division, different gestures of the dynamic process can be captured.

(4) Δ (Δe)

The rate of change of the error, that is, the second-order difference, describes that the dynamic process is in the overshoot or callback segment. When Δ (Δe)> 0, the overshoot segment: Δ (Δe) <0 is in the callback segment.

Summarize the above characteristics, the basic control algorithm can be summarized as:

Where: Un is the output of the n-th sampling time of the controller;

Vo is the nth holding value of the controller;

e and, respectively, the system error and its rate of change;

emi is the ith extremum of the error;

Kp is the proportional gain of the controller;

k for the control of gain suppression (attenuation) coefficient, generally take o <k <1.

The above control algorithm is very suitable for the setting control of a large lag self-balancing object. In order to adapt to the abrupt change of traffic or further improve the anti-interference ability, the Bang-Bang control can also be added to make the dynamic quality better.

7 Conclusion

Above from the control point of view discussed the oilfield electric submersible pump system-related issues. In order to save energy in the physical realization, it is necessary to combine the control technology and the variable frequency speed regulation technology, and the electric transmission device adopts the frequency converter, that is, to convert the electric submersible pump to a variable frequency so as to realize the soft start, soft stop and effective parking of the electric submersible pump To protect the electric submersible pump and cable. In this way, the controller can conveniently adjust the hydraulic pressure by adjusting the frequency, thus avoiding the long-term operation of the electric submersible pump under high pressure, prolonging the service life of the electric submersible pump, saving the cost of maintenance and repair of the oil well, Electric submersible pump oil production efficiency is greatly improved, and improve the power factor, improve the power supply capacity of the grid, energy-saving effect is obvious. Facts have proved that as long as the frequency control and control technology is well integrated, oil field submersible pump system will certainly get a good social and economic benefits.