Forecasting Petroleum Production Using the Time-Series Prediction of Artificial Neural Network

Abstract

The purpose of this paper is to present a special back-propagation neural network (BPNN) with two techniques of the optimal learning time count (OLTC) and the time-series prediction (TSP) for forecasting petroleum production in Chinese oilfields, as well as algorithm applicability. In general, when different algorithms are used to solve a real-world problem, they often produce different solution accuracies, and an algorithm is used to solve real-world problems, it often produces different solution accuracies. Toward this issue, the solution accuracy is expressed with the total mean absolute relative residual for all samples, R(%); and it is proposed that an algorithm is applicable if R(%) ≤ 5, otherwise this algorithm is inapplicable. Two case studies of China have been used to validate the proposed approach. The application results of this special BPNN are R(%) = 2.18 in Case study 1 while R(%) = 2.05 in Case study 2. From these results, it is concluded that: (a) this special BPNN for forecasting petroleum production in Chinese oilfields is feasible and practical; and (b) the definition of solution accuracy R(%), and the threshold of algorithm applicability (R(%) ≤ 5) for an algorithm, are feasible and practical, too.