Intelligent Systems' Applications

Candidate Well Selection

Summary

It is a common occurrence in the oil and gas industry that certain operations will be performed on a group of wells and the results are studied and analyzed before the same operation or variations of it is performed on the rest of the wells in the field. Sometimes budget limitations only warrant the operations on a subset of the wells in the field each year. In all cases the fact remains the same; the objective is to select the group of wells for the operation that are most likely to produce good results. In other words, we are always facing the question, "which wells would benefit the most from a certain operation and can help the engineers, geo-scientists and management achieve their objectives from the operation". This process is referred to as the candidate well selection.

Identification of best candidate wells for all sorts of operations is a common practice in the oil and gas industry. The candidate well selection includes but is not limited to the following operations:

• Well Treatments
  • Workover
  • Hydraulic Fracturing
  • Re-stimulation
  • New Completions
• Steam Injection or SAGD
• Water Flooding
• Gas Injection
• Rate Increase (Bean Up)

Although a common practice, candidate well selection is not a straight forward process and up to now there has not been a well defined and unified approach to address this problem. The main reason for lack of a unified approach has been the view that each operation is different and therefore requires a certain set of wells that might be good candidate for a particular operation but not necessarily are good candidate for another operation. In other words, candidate wells are highly case specific. The goal of this white paper is to show that although different operations such as steam injection, water flooding, hydraulic fracturing, workover, etc. are quite different in nature, there can be a unified approach in identifying the specific wells (candidates) that would respond in a favorable fashion to the specific operation being performed.

The Unified Approach

The unified approach for candidate well selection being introduced here, defines the process of candidate selection as an optimization problem. As such, the process is inspired by the general approach to optimization in order to address the candidate well selection problem. The general optimization process requires identification of an objective function and an exhaustive search process. The complexity of the candidate selection process makes the development of a meaningful objective function for the development of a complete solution space very challenging. Furthermore, the search process, once a solution space is developed, is not very straight forward. Domain expertise usually plays an important role in success of many of such operations and cannot be discounted.

I would submit that due to the complexity and inherent non-linearity associated with this process Artificial Intelligence would be the only viable methodology to effectively achieve the objectives of candidate well selection process. This is due to the following facts:

1. Lack of extensive data required for numerical modeling in addition to the fact that simulation is a prohibitively expensive process leaves room only for two possible approach in developing an objective function. Analytical solutions and intelligent modeling. Most of these processes are too complex to be analytically modeled. Therefore, building an intelligent predictive model in many cases remains as the only realistic alternative. Artificial neural networks are the tool to be used in this step.
2. Once the objective function is developed, tested and verified, it usually includes a large number of parameters and variables. Effectively and efficiently searching through all the possible combinations that may exist (the comprehensive solution space) can be achieved by intelligent techniques such as genetic algorithms.
3. Domain expertise is human knowledge. Human knowledge can best be expressed in the vague and imprecise natural language that is the domain of fuzzy logic. The expertise may be coded into fuzzy rules that can be used in a fuzzy expert system and combined with the results of the optimization techniques accomplished in the previous step.

This unified approach is an integration of two steps. The first step is fully data driven and the second step provides the option to incorporate domain expertise in the selection process.

Intelligent Best Practices Analysis

Summary

Identification of best practices in the oil and gas operations is gaining unprecedented momentum. This is partly due to the realities of the new economy that ties the success of oil and gas companies to their performance in the stock market. Companies that have gathered large amounts of data now realize that they own a valuable commodity (above and beyond the hydrocarbon) that can play an important role in increasing efficiency in their day to day operations.

The question is how this vast amount of data can be used in order to help the company's bottom-line. This paper attempts to address this question by introducing a newly developed methodology that enables oil and gas companies to deduce information and knowledge from the existing data. The deduced information and knowledge can then be used in developing business rules and making decisions.

Many companies in the oil and gas industry have been collecting large amounts of data over the past several years. Hundreds of thousands of dollars have been invested in collecting and compiling various types of data. These databases cover all aspects of oil and gas business, from purely technical data that includes certain measurements from the reservoir or the surface facilities to non-technical data such as those related to economics or human resources issues. Now that all this data is available, following questions may be asked:

• "What can we do with this data?"
• "How can the company get a return on its data collection and preservation investment?"
• "Are there stories hidden in the megabytes, or sometime gigabytes of data?"
• "The collected data is a reflection of the history of the operations that have taken place and sometime are still taking place. What can we learn from our past practices?"

As the volume of data increases, human cognition is no longer capable of deciphering important information from it by conventional techniques. Data mining and machine learning techniques must be used in order to deduce information and knowledge from the raw data that resides in the databases. The Intelligent Best Practices Analysis (IBPA) that is introduced here incorporates the state of the art in data mining and machine learning to assist petroleum professionals in making the most of their existing data. Figure 1 is a schematic diagram of IBPA.