Prediction of Gathering Pipeline Internal Corrosion Rate Based on Improved Wavelet Neural Network Model

Abstract

The accurate prediction of corrosion rate is significant for the safe operation of oil-gas gathering pipeline. The internal corrosion of pipeline is a multivariable nonlinear system, and furthermore because of the complex mapping relation between corrosion rate and its influencing factors, it is always difficult to predict the corrosion rate accurately. Aiming this problem, this paper established a wavelet neural network (WNN) and introduced genetic algorithm (GA) to optimize it. Taking one gathering pipeline in gas field as an example, prediction model of internal corrosion rate for gas pipeline was built by using corrosion factors (CO2 content, temperature, etc.) as input parameters and internal corrosion rate as output parameter. The model combines the global optimization searching performance of the genetic algorithm and the well time-frequency localization of the wavelet neural network, so it has higher learning accuracy and faster convergence rate. Prediction results fit simulating test data preferably. It indicates that GA-WNN model has high fitting accuracy. Results proved it should be a good tool to predict internal corrosion rate of gathering pipeline.