In this paper, we discuss the general application of higher neural networks, this follows from the successful application of such networks to the identification of well test models. During the course of the research, it was discovered that these networks have a large range of applications in petroleum engineering. Hence the objective of this paper is to give a background of higher order neural networks and their potential uses. Conventional neural networks have activation functions that are linear correlations of their inputs, whereas higher order networks have a non-linear correlation of their inputs. Higher order neural networks do not have wide practical applications due to the enormous amount of parameters (weights) associated with them. However, for certain problems these vast amount of weights are greatly reduced by constraining the architecture of the network. That is, problems that need to be classified regardless of some transformation groups such as translation, scaling and rotation. A typical example is the identification of well test models where standard type curves are translated both horizontally and vertically with respect to field data plots.

You can access this article if you purchase or spend a download.