The use of artificial neural networks in completion stimulation and design

Optimizing well-completion procedures is often difficult because of (1) var iables related to reservoir quality and (2) a general lack of understanding regarding the complex interactions between well-completion/stimulation pro cedures and the reservoir. In many situations, well completion is a comprom ise between reservoir conditions and operational procedures. However, artif icial neural networks (ANNs), can be used to help improve well economics. A NN models trained to interpret reservoir, well, and completion information can predict well cumulative production with an acceptable degree of accurac y. Sensitivity studies of these networks indicate that, for a given reservo ir quality, completion methods can significantly affect well production. Th is paper presents case histories in which ANN sensitivity analyses were use d to justify changes in completion/stimulation procedures. ANN-enhanced com pletions resulted in better overall well production than standard completio n optimization methods normally used in these fields. This paper discusses well-completion analyses performed with ANN technology. The economic advant ages of various completion techniques are also discussed. ANN technology ha s been used in areas where conventional engineering methods produced unacce ptable results. These completions were performed in (1) the Red Fork format ion of Roger Mills and Custer County, Oklahoma, USA (2) the Frontier format ion of Lincoln County, Wyoming, USA and (3) the Granite Wash formation of R oberts County, Texas, USA. Areas of interest include quantifiable well, res ervoir, and completion characteristics, such as porosity, pressure, and com pletion/stimulation procedures, that affect well production. Net present va lue is used to quantify the economic effect of ANN enhancement on well comp letion. (C) 2000 Elsevier Science Ltd. All rights reserved.