Application of neural network and fuzzy mathematic theory in evaluating the adaptability of inflow control device in horizontal well

• Single factor analysis of ICD adaptability is done by static evaluation method.
• Sensitivity analysis of ICD adaptability is done by orthogonal experiment.
• A fuzzy evaluation model of ICD adaptability is established.
• A neural network synthetic evaluation model of ICD adaptability is established.

Inflow control device (ICD) has gained popularity since its first introduction to oil industry. However, ICD completion isn’t suitable for all reservoirs because the economic value of ICD technology depends on various factors. This paper first establishes a mechanism model and summarizes the influences of single factors (permeability variation coefficient, reservoir thickness, average permeability, Kv/Kh (the ratio of vertical permeability to horizontal permeability), horizontal section length, base pipe diameter, segment number and ΔPICD/ΔPreservoir (the ratio of pressure drop caused by ICD to the pressure drop of the reservoir)) on ICD completion adaptability by static evaluation method. Next, sensitivity analysis of the influencing factors is conducted by orthogonal experiment. Then, different subordinate functions of each influencing factor are established according to the variation tendency of water yield reduction and inflow profile variation coefficient difference (compared with conventional screen completion) resulting from mechanism model. After that, six cases of ICD completion are introduced, and the weight sets and evaluation sets of the cases are established, according to which fuzzy evaluation model of ICD completion adaptability is obtained. What’s more, a neural network synthetic evaluation model with superior in self-learning ability and computing power is proposed by improving the fuzzy evaluation model. The results of case studies demonstrate that both fuzzy evaluation model and neural network synthetic evaluation model are practical and feasible.