Method to Optimize Production from a Gas-lifted Oil Well

Abstract

A method for determining and reporting a general production state of a gas-lift system for an oil well, the oil well having associated tubing, casing, and gas-lift valves, and wherein sensor signals from the welt and its associated tubing and casing are input into mathematical models. The method comprises the steps of: a) extracting values from the mathematical models that indicate instantaneous states of production; b) supplying the sensor signals and the values to an associative memory agent; and c) using the associative memory agent to associate the sensor signals and the values to generate the general production state.

Description

CROSS-REFERENCES TO RELATED APPLICATIONS

None.

STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

None.

REFERENCE TO A “SEQUENCE LISTING,” A TABLE, OR A COMPUTER PROGRAM LISTING APPENDIX SUBMITTED ON A COMPACT DISC AND AN INCORPORATION BY REFERENCE OF THE MATERIAL ON THE COMPACT DISC

None.

BACKGROUND OF THE INVENTION

(1) Field of the Invention

The invention relates to methods to optimize liquid production from an oil well that is under gas lift.

(2) Description of the Related Art

As is well known in the art, gas-lift techniques are employed in oil wells which have difficulty in producing satisfactory levels of liquids based on natural formation pressure. Typically, such wells have formation pressures which are insufficient to impel liquids to the surface at economically acceptable volumes.

The gas-lift technique involves injecting gas into the casing of an oil well through one or more valves, typically located at varying heights along the well. Depending upon the technique being used, the gas may be injected substantially continuously into the column of fluid in the well, thereby lightening this column of liquid so as to enhance the natural formation pressure. Alternatively, gas can be injected intermittently in a repeated or cyclical process so as to produce successive slugs of liquid at the well head.

Although gas-lift techniques provide excellent results for certain types of oil wells, each well is different in terms of downhole or formation pressure, downhole or formation temperature, depth to the producing formation, geothermal gradient experienced along the vertical height of the well, and numerous other factors. In addition, gas-lift injection systems differ in terms of their valve characteristics, injection pressure, injection volumes and other factors. Thus, determining the optimal operating parameters for a gas-lift technique is often a time consuming trial and error process which may require extensive supervision and nevertheless provide less than ideal production.

Typically entire oil fields are produced using the gas-lift technique because the formation pressure everywhere in the field is insufficient to produce satisfactory quantities of oil. The gas-lift technique then must be applied to each well in the field. A complex system of gas supply and piping must connect each well to a central collection and distribution facility. Operating the gas-lift system then becomes a situation of complex interaction between the needs of each well and the overall need to enhance production of each well in the field. While several software programs and procedures exist that help control and optimize production from the field while utilizing the available gas supply efficiently, these programs assume that each well within the field is operating efficiently. This assumption is often violated because individual wells may not perform at their most efficient potential. Thus the field management software is not as effective as it needs to be to optimize field production.

There are many situations in which less than optimal production from the well may occur. These situations involve the pressure, temperature, production flow, gas injection rates, and the states of the several valves in the well. In order to diagnose a problem, it is necessary to consider many configurations of these parameters and the implications of their current values. Further, it is necessary to classify possible states of the oil well in order that the diagnostic can relate to the existing production state of the well.

Several attempts have been made to optimize oil well liquid production under gas-lift that are based on so-called expert systems that use a rules-based decision making process to identify problems with the way in which a gas-lift technique is performing on a given well. Such expert systems may not perform as well as needed because the full set of data values required for making an incontrovertible diagnosis may not be available. Accordingly the system must be able to diagnose problems using whatever data is available. Also, such expert systems may not diagnose lifting problems correctly because the parameters of the operation change during the life of the well. In order to account for the aging of the well, the expert system would require continuous or intermittent retuning to ensure effective diagnostic abilities. in addition, many factors that influence the ability to diagnose problems in a well under gas-lift are often overlooked by the expert system because the developers of the systems cannot know all possible conditions that may influence the operation at the time that they develop the software program.

An early expert system that used a rules-based decision making process which attempted to improve the rules based on the results obtained is disclosed in the following patent, which is incorporated herein by this reference: U.S. Pat. No. 4,918,620, which states in the abstract, “A computer software architecture and operating method for an expert system that performs rule-based reasoning as well as a quantitative analysis, based on information provided by the user during a user session, and provides an expert system recommendation embodying the results of the quantitative analysis are disclosed. In the preferred embodiment of the invention, the expert system includes the important optional feature of modifying its reasoning process upon finding the quantitative analysis results unacceptable in comparison to predetermined acceptance criteria.” However, the method disclosed in this patent does not allow for generating attributes from real time data to compare to known symptoms of poor well behaviors. Rather, it requires that an expert think of all the rules possible in the system in order to account for novel behavior, and it cannot adapt to data-drop-out when sensors fail in service.

Another expert system that uses a rules-based decision making process that attempts to improve the rules based on the results obtained is disclosed in the following patent, which is incorporated herein by this reference: U.S. Pat. No. 6,529,893, which states in the abstract, “The system uses an author interface, an inference generator, and a user interface to draw authoring and diagnostic inferences based on expert and user input. The inference generator includes a knowledge base containing general failure attribute information. The inference generator allows the expert system to provide experts and users with suggestions relating to the particular task at hand.” However, the method disclosed in this patent does not show how to deploy an expert system to diagnose problems with gas-lift wells, and it is furthermore subject to the limitations of rule-based-systems as described in the previous paragraph.

Another expert system that uses a rules-based decision making process which attempts to improve the rules based on the results obtained is disclosed in the following patent, which is incorporated herein by this reference: U.S. Pat. No. 6,535,863, which states in the abstract, “The method improves the performance of the system by evaluating how well the system's body of knowledge solves/performs a problem/task and verifying and/or altering the body of knowledge based upon the evaluation”. However, the method disclosed in this patent does not address monitoring and diagnosis. Also, it requires a human to evaluate the results of the analysis, and provide feedback to the software program regarding which rules to accept and which to keep based on performance.

Another expert system that uses a knowledge-based decision making process that attempts to improve the base of knowledge based on the results obtained is disclosed in the following published patent application, which is incorporated herein by this reference, U.S. Patent Application No. 20060025975, which states in the detailed description, “The weights of each network or expert are determined at the end of a learning stage; during this stage, the networks are supplied with a set of data forming their learning base, and the configuration and the weights of the network are optimized by minimizing errors observed for all the samples of the base, between the output data resulting from network calculation and the data expected at the output, given by the base.” However, the method disclosed in this patent requires an accurate model of flow in the system in order to train it, and it will not diagnose the origin of flow impairments.

Another expert system that uses a knowledge-based decision making process that attempts to improve the base of knowledge based on the results obtained is disclosed in the following patent, which is incorporated herein by this reference: U.S. Pat. No. 6,236,894, which states in the abstract, “A genetic algorithm is used to generate, and iteratively evaluate solution vectors, which are combinations of field operating parameters such as incremental gas-oil ratio cutoff and formation gas-oil ratio cutoff values. The evaluation includes the operation of an adaptive network to determine production header pressures, followed by modification of well output estimates to account for changes in the production header pressure.” However, the method disclosed in this patent does not address individual well productivity, and it requires iterative applications rather than recognizing and diagnosing problems from the data presented.

Another expert system that uses a knowledge-based decision making process that attempts to improve the base of knowledge based on the results obtained is disclosed in the following patent, which is incorporated herein by this reference: U.S. Pat. No. 6,434,435, which states in the abstract, “The systems and the methods utilize intelligent software objects which exhibit automatic adaptive optimization behavior. The systems and the methods can be used to automatically manage hydrocarbon production in accordance with one or more production management goals using one or more adaptable software models of the production processes.” However, the method disclosed in this patent requires production models of the production process, which is itself subject to errors. Therefore, the system disclosed in the '435 patent will not be fault tolerant of failed or missing sensor data. Furthermore, the system disclosed in the '435 patent does not produce a specific diagnosis of unsatisfactory behavior.

In light of the foregoing, a need remains for a more efficient method for diagnosing production problems from a gas-lift well.

BRIEF SUMMARY OF THE INVENTION

A method for determining and reporting a general production state of a gas-lift system for an oil well, the oil well having associated tubing, casing, and gas-lift valves, and wherein sensor signals from the well and its associated tubing and casing are input into mathematical models. The method comprises the steps of: a) extracting values from the mathematical models that indicate instantaneous states of production; b) supplying the sensor signals and the values to an associative memory agent; and c) using the associative memory agent to associate the sensor signals and the values to generate the general production state.

A system for diagnosing problems in, and reporting the general state of, the production mode of the gas-lift operations on a well, the well having associated gas-lift valves and sensors, the system comprising: a) a personal computer for receiving reports of signals from the sensors; b) means stored on the personal computer for generating mathematical models to deduce the states of the gas-lift valves and the states of the production mode by using as inputs both the sensor signals and a knowledge base, to generate multiple states over time of the production mode; and c) an associative memory agent stored on the personal computer, and responsive to the multiple states, for detecting and aggregating anomalies, and for reporting a general state of the production mode.

The method and system are readily adaptive to the widely varying conditions experienced at different gas-lifted oil wells. In another feature of the invention, the system learns well behaviors for the specific conditions of a particular well, and diagnoses gas injection and production problems in the well based on pattern recognition and past well behaviors so as to reduce the need for human operator involvement in the diagnosis of problems.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

FIG. 1 is a schematic diagram of an oil well on which gas-lift is being exercised.

FIG. 2 is a schematic diagram showing the inputs to the associative memory agent of the present invention.

FIG. 3 is a flow chart illustrating how raw data from the sensors is processed to yield a report.

DETAILED DESCRIPTION OF THE INVENTION

In FIG. 1, an oil well 12 has casing 14 and production tubing 16, through which production from perforations 18 is conveyed to the surface 19. When a formation 20 has a geological pressure that is insufficient to cause the liquids in the formation 20 to reach the surface 19 in sufficient economic quantities, an operator can assist the formation 20 by injecting gas from the casing 14 from a supply line 22, passing through an inlet valve 24, and from the casing 14 into the tubing 16, through gas-lift valves, 26, 28, 30, and 32. The injected gas mixes with the liquid in the well and reduces the density of the liquid so that the column of liquid is now light enough for the formation pressure to provide sufficient lifting force to cause liquids to be produced in economic quantities through a production valve 34 into a flow-line 36. Also disposed around the well 12 are pressure sensors 38 and 40 whose readings are conveyed to a control room where engineers and other field personnel can monitor the well's production. The flow of gas into the casing 14 can also be monitored by observing pressure differentials between the pressure sensor 40 and other pressure sensors disposed in other locations around the field. Differences in pressure between the pressure sensor 40 and other sensors indicate that gas is flowing in the lines. In addition, outflow of oil and other liquids can be inferred from the differential pressure between pressure sensor 38 and other pressure sensors in other locations (not shown) on the flow line 36.

Often the design of the gas-lift technique does not account for changing conditions during the productive life of the well, and the gas-lift operation begins to behave in a manner that reduces its lifting capabilities. This may occur for several reasons. The gas-lift technique is designed so that gas is injected through the lowest gas-lift valve 32. At that depth, the gas is most effective in reducing the density of the liquid column and hence in lifting the liquid to the surface. Less beneficial operation of the technique may occur when any of the following conditions occurs: a) insufficient gas is supplied to the valves 26-32 to lift the liquid to the surface 19 in economic quantities; b) gas is being injected through more than one valve; c) no gas is being injected because of insufficient pressure in the gas delivery system; d) no gas is being injected because the pressure of liquid in the tubing 16 is greater than the gas pressure in the casing 14; e) too much gas is being injected; f) one or more of the valves is faulty and doesn't respond to casing or tubing pressure correctly; and g) other reasons having to do with the mechanics of operating the gas-lift system.

Surveillance engineers over the years have identified most of the causes of poor gas-lift performance in individual wells. Each of these causes can be labeled with a name and associated with a set of circumstances the realization of which cause the performance. Most often, the causes for particular behaviors are not clearly demarcated so that a rules based diagnosis is inadequate to identify the cause of the problem. The set of names of the causes of poor performance are called “states” of the gas-lift operation. Hence, each state represents a particular set of conditions causing the known poor performance.

The present invention operates by using an associative memory technology such as that sold by Saffron Technologies, Inc. However, the present invention can use any associative memory technology, and is not restricted to the associative memory technology sold by Saffron Technologies. The associative memory technology uses mathematical and symbolic evidence from sensors and known event types to associate that set of sensor readings, mathematical model results, and symbolic evidence with the states of the gas-lift operation. In addition to the associative memory technology, other important elements of the invention are a knowledge base, a database, and simulation software.

The knowledge base includes the states of the system and their symptoms. Such knowledge base exists in the form of spreadsheets and written documents in which the states of the system and the associated symptoms are tabulated. This knowledge base is used for training the associative memory to recognize each of approximately 70 different states for a gas-lift system.

The database includes examples of the readings of sensor values at such times during which the system is exhibiting such behaviors. The database can also be a database kept by an operating oil company of the sensor readings for use by engineers and other company personnel to diagnose operations of gas-lift fields.

The output of simulation software represents the behavior of gas-lift systems when undergoing such behaviors. Such simulators have been developed by companies in the oil field services industry. In the preferred embodiment, the simulation software is the program “DynaLift” produced by Weatherford. As an illustration of the output of the simulation software, the simulation software provides information about the state (open/closed) of each injection valve in the system based on known pressures in the system and the valve's performance attributes that are provided in specification sheets from the valve manufacturer.

An engineer developing an intelligent agent for diagnosing gas-lift problems then combines the knowledge base and databases (real, simulated, or a combination of both) with the associative memory technology to produce an intelligent agent that is trained to recognize various states of the system. The combining activity is described below.

Referring now to FIG. 2, in the process of combining the various sources of information, an associative memory agent 50 receives input from four sources. Data 52 from sensors on the well and flow-lines includes pressures, temperatures, flow rates, valve states, injection rates, production rates, etc. The data 52 is conditioned by a signal conditioning process 54 to convert analog sensor readings into information about the well. In addition to the absolute values of the sensor readings, the signal conditioning process 54 computes the time averages of the sensor values, the standard deviations of the averages, Finite Fourier Transforms (FFT) of the data, long term and short term time derivatives of the sensor readings, and other mathematical functions of the data, to produce attributes of the data stream that are useful to describe the states of the system.

Mathematical models 56 of the dynamics of the fluids (gas and liquids) flowing in the well are also used to estimate quantities that are not measured directly, such as whether gas-lift valves in the well are open and flowing. Additional mathematical models include models for calculating the deepest point of injection, the inflow/outflow curves, and operating rates vs. lift gas injection rates. The results obtained from the mathematical models are then transmitted to the associative memory agent 50.

A library 58 of associative memory agents trained on both data from wells and on data generated by a simulator computer program is also an input to the associative memory agent 50. An analysis knowledge base 60 is used to configure the memory of the associative memory agent 50 to receive the data 52 and conditioned attributes from the other data sources, e.g., from simulators. Approximately seventy distinct anomalous well states exist that are covered by the analysis knowledge base 60.

An associative memory agent 50 is used for monitoring and diagnosis of gas lift systems by employing the following steps.

• • 1. Prior to operation of the diagnosis system, a knowledge base is constructed which identifies the set of conditions (normal or abnormal) which are to be recognized. Each condition is a system state which is defined by a vector of attribute values.
  • 2. Prior to operation of the diagnosis system, the knowledge base is used to train an associative memory. Training the memory means that we store in a memory each condition and the vector of attribute values that define the condition.
  • 3. During the operation of the diagnostic system, a vector of attribute values describing a well's current state (the input context) is compared to the set of conditions stored in the memory. The input context may or may not include all of the attribute values needed for fully characterizing the well's state. A scoring algorithm is used to determine the degree of similarity between each stored condition and the well's existing condition.
  • 4. Since the input context may or may not contain all of the attributes needed for fully characterizing the well's state, the diagnosis system produces a rank-ordered list of the most likely diagnosis.

A variety of implementation methods exist for storing information in an associative memory and for scoring input contexts. A sparse matrix implementation of an associative memory is described in “Comparative Analysis of Sparse Matrix Algorithms for information Retrieval”, Nazli Goharian, Ankit Jain, Qian Sun. Information Retrieval Laboratory, Chicago, Ill. This particular paper also describes an inverse document frequency algorithm for scoring. Note that although this paper addresses the use of associative memories for text processing applications, similar techniques can be applied to implement a gas lift diagnosis system.

Referring now to FIG. 3, in step 70 the method of the present invention begins. The method may be performed in real-time (i.e., as time-series data become available) or in a batch mode. Each data “snapshot” is processed one-at-a-time, though many of the system's algorithms look back at previous data values to calculate current system state. For each data sampling interval (i.e., for each data time stamp), the following steps are repeated.

In step 72 time series pressure data 52 from well sensors 38 and 40 are received. In step 74 the data 52 is processed using an algorithm that employs a Finite Fourier Transform (“an FFT”) and ordered statistics to generate a set of values that describe the periodicity of the pressure signals. The associative memory agent 50 classifies each input pressure channel, including tubing pressure and casing pressure, as either normal or abnormal. The associative memory agent 50 has been trained to distinguish between normal and abnormal states. Abnormal states include behaviors such as flow instability and heading. Thus, in step 74 the associative memory agent 50 assesses casing and tubing pressures to determine if heading is present. Heading is a cyclical variation in casing or tubing pressure that indicates potential problems in the gas-lift system.

In steps 76, 78, and 80 values in addition to pressure heading states are obtained or calculated in order to identify anomalous well conditions. In step 76 time series values such as well fluid level and gas flow rates are received. Step 76 includes analysis of data to identify conditions such as fluctuating input gas rates. Step 78 includes receipt of well configuration data such as the well depth, valve characteristics, and the location of injection valves in the system. Step 80 includes receipt of mathematical model outputs such as the calculated state of each valve (open/closed) and deepest point of injection. The values generated by these mathematical models depend on parameters such as gas injection rates and pressures.

For each data sampling interval, in step 82 the associative memory agent 50 combines heading information and the values from steps 76, 78, and 80 into a record that represents the well's current operating state. The associative memory agent 50 compares these operating conditions with patterns in the anomaly detection memory to determine whether an anomalous condition is present. In some cases, some input data values may be missing; and, in that case, the associative memory agent 50 computes a probabilistic value for the well condition. The anomaly detector provides the likelihood of several possible anomalies in a format where the most likely condition, A is assigned to the highest probability, the 2nd most likely condition, B, is assigned to the second most probable state, and the 3rd most likely condition, C, is assigned the third highest probability, etc. In order to obtain these events in rank order, the memory is queried by using software commands.

The representation of knowledge in the associative memory agent 50 enables queries using only partial data. Hence, if a data stream becomes unavailable, the query can still be executed, and the agent will return a set of probabilities the values of which take into account the missing data stream. Thus, in step 80 there may be more than a single probabilistic result, which is a clear advantage of the invention because, in the absence of complete data, it would otherwise be difficult to determine whether anomaly A or B or C exists.

Because the method of the present invention processes data one time-stamp at a time, in step 82 anomaly detection identifies anomalies that exist at a particular point in time. In step 84 anomaly states over time are aggregated to determine the system state. In both steps 82 and 84, the method of the present invention compares the received data with the knowledge base 60. In a simple case, a condition such as high separator pressure may exist over an uninterrupted period of time. In a boundary case, the well may cycle between an abnormal state (e.g., excessive injection) with intermittent periods of normalcy. In more complex cases, a well may cycle from one abnormal state such as multi-point injection to another abnormal state such as surging repeatedly. In this case, the well is best characterized as surging, because multi-point injection is a symptom of extreme surging, and this cycling between states can be indicative of the well's true problem. In step 84, anomaly aggregation “smoothes” anomaly detections over time in order to provide operators with a clear and understandable picture of the well's condition over time, not just its instantaneous state. Various implementations of the data aggregation function are possible, including rule-based approaches and pattern-matching methods such as associative memories or neural nets.

The system continuously processes time-series data, an interval at a time, and in step 86 generates a continuous picture of the gas-lift system's state over time. In a bounded period of time such as twenty-four hours, the system may be nominal for some periods of time, suffering from condition A during other periods of time, and suffering from condition B or C during yet other periods. Also in step 86, the method generates recommendations for steps to take to correct problems. This method repeats as long as data are available. In step 88, if there are no more data, then in step 90 the program ends.

Claims

1. A method for determining and reporting a general production state of a gas-lift system for an oil well, the oil well having associated tubing, casing, and gas-lift valves, and wherein sensor signals from the well and its associated tubing and casing are input into mathematical models, the method comprising the steps of:

a. extracting values from the mathematical models that indicate instantaneous states of production;

b. supplying the sensor signals and the values to an associative memory agent; and

c. using the associative memory agent to associate the sensor signals and the values to generate the general production state.

2. The method according to claim 1, wherein the operation of the associative memory agent comprises pattern recognition and use of knowledge of past well behaviors.

3. The method according to claim 1, wherein the step of extracting values includes a step of deducing instantaneous states of the gas-lift valves by using the mathematical models.

4. The method according to claim 2, wherein the step of associating uses probabilistic classification to generate the general production state.

5. The method according to claim 3, wherein the step of deducing includes determining if abnormal conditions exist based on the received sensor signals, and wherein an associative memory agent is used to make the determination.

6. The method according to claim 3, further comprising, after the step of generating the general production state, the step of reporting the general production state.

7. The method according to claim 5, wherein the step of determining if abnormal conditions exist uses a Finite Fourier Transform combined with ordered statistics.

8. A system for diagnosing problems in, and reporting the general state of, the production mode of the gas-lift operations on a well, the well having associated gas-lift valves and sensors, the system comprising:

a. a personal computer for receiving reports of signals from the sensors;

b. means stored on the personal computer for generating mathematical models to deduce the states of the gas-lift valves and the states of the production mode by using as inputs both the sensor signals and a knowledge base, to generate multiple states over time of the production mode;

c. an associative memory agent stored on the personal computer, and responsive to the multiple states, for detecting and aggregating anomalies, and for reporting a general state of the production mode.

9. The system of claim 8, wherein the associative memory agent learns well behaviors for the specific conditions of a particular well, and diagnoses gas injection and production problems in the well based on pattern recognition and past well behaviors.