Neural Particle Swarm Optimization for casing damage prediction

Abstract

Casing damage of oil wells is a serious problem, which has caused enormous economic losses to oil field of china. The reasons of casing damage are very complex, the experts of oil production can only predicted the damage trend by their personal experiences. This paper put forward a new method called Neural Particle Swarm Optimization (NPSO), which has been applied to the prediction of casing damage. In this method, combined with feedforward neural network and PSO (Particle Swarm Optimization PSO), the feedforward neural network is regarded as particles in space, called neural particles. The learning processes of neural network are take cover in the movement of neural particles following individual best and swarm best. NPSO method was used to predict casing damage of oil wells, the experimental result show that the learning efficiency and converge rate of NPSO improved much more than tradition learning method (such as BP) and has great application value for prediction of casing damage. © Springer-Verlag Berlin Heidelberg 2005.