New expert system for enhanced oil recovery screening in non-fractured oil reservoirs

Abstract

As the oil production from conventional oil reservoirs is decreasing, oil production through Enhanced Oil Recovery (EOR) processes is supposed to compensate for both the oil production reduction in matured oil reservoirs and the worldwide dramatic increase in oil demand. Therefore, developing a strategy to choose an optimized EOR technique is crucial to find a resolution for production decline in oil reservoirs. A screening tool recommending the most appropriate EOR method is proposed in this study. An expert fuzzy logic system is employed to screen four well-known EOR methods including miscible ${\text{CO}}_2$ injection, miscible HC gas injection, polymer flooding and steam injection based on the reservoir condition, fluids characteristics and rock properties. Since this expert system has incorporated the rules from successful past experience, it reduces the requirement of extensive laboratory and field data. The expert system can be used as a quick estimation for evaluating the success of the EOR processes for a particular reservoir before making any executive decision. Based on the approach proposed in this research, a screening program capable of ranking proper EOR techniques has been developed. The output results have been compared with the results of a Bayesian Belief Network (BBN) model and some field experience. This comparison indicated that not only is the proposed model in a good agreement with the field data, but it also provides a more robust approach than the BBN model.

Introduction

Enhanced oil recovery (EOR) screening techniques have been utilized in the past few decades to identify the most appropriate EOR method for a candidate reservoir based on proposed screening criteria. These criteria originated from field experience, experimental investigations, and simulation studies, thereby significantly reducing the application risks of these techniques.

Some useful technical recommendations were proposed by Taber and Martin [1] where the screening criteria were presented in tables and graphs. Taber et al. [2] proposed screening criteria for most EOR techniques by using available data and considering EOR mechanisms. However, the application of these recommendations in computer programming is difficult due to the wide overlapping ranges in the criteria of different processes.

Over the past few years, Artificial Intelligence techniques (AI) such as Artificial Neural Networks (ANN), fuzzy inference systems, and Bayesian Belief Networks (BBNs) have been successfully applied to a wide variety of complicated engineering problems [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14]. In particular, these methods have been considered as reliable decision-making tools in petroleum engineering. Since EOR screening is a great challenge for petroleum engineers, AI techniques have been utilized as an important tool in EOR method selection [15], [16], [17], [18], [19], [20], [21]. For instance, these techniques have been utilized in EOR project risk and economic analyses as well as EOR screening [15], [16]. Guerillot [17] utilized fuzzy logic to design a screening system that is capable of representing knowledge-based descriptions for EOR methods. The proposed model yielded satisfactory outcomes, particularly in thermal processes. However, this model failed in other EOR methods due to the lack of available data. Parkinson et al. [18] proposed a computer program based on an expert system to check the possibility of a few EOR techniques in terms of economic considerations. Also, Chung and Carroll [19] employed a fuzzy expert system to analyze the EOR project risk for a specific EOR method. Tapias et al. [20] utilized a neuro-fuzzy system to study EOR screening in the Colombian Petrolea field and to determine the potential hydrocarbon producing zone. Furthermore, a BBN method was recently applied in this research group to select the appropriate EOR method using field scale data from EOR projects [21].

In this study, we incorporated the fuzzy rules derived from previous successful field experience into a fuzzy inference system to screen four well-known EOR techniques in depleted non-fractured oil reservoirs. For a specific reservoir, the fuzzy system was employed to rank the success possibility of these EOR methods based on the reservoir and rock properties. In contrast to other AI techniques where the design and development relies mainly on rich data sets, the proposed fuzzy system is not influenced by either the scarcity of data in some ranges of screening parameters, or by the accumulation of data in other ranges. Furthermore, this method is easier to use than screening tables and graphs, since it considers the economic priority of methods as well as technical issues. Also, in this method, the solutions are obtained within a few milliseconds (0.82 seconds) of CPU time using an ordinary desktop type computer.