ABSTRACT

Accurate prediction or measurement of bottom hole pressure is of utmost importance in the petroleum Industry. Over the years, mechanistic, numerical and analytical models have been developed to estimate bottom hole pressure and also the use of down-hole pressure gauges have been used to tackle this problem in the petroleum Industry. Though some of the models developed are economic, they failed to predict bottom hole pressure to an acceptable accuracy. The down-hole gauges measured to an acceptable accuracy but are expensive to use and maintain.

This project aims to develop economic prediction models for bottom hole pressure based on input data obtained from Volve production field in Norway. Machine learning algorithm based on Artificial Neural Network, Multi Linear Regression, Robust Regression, Random Forest Regression and Stack regression were used to predict the bottom hole pressure to an acceptable accuracy and selecting the best out of the models for prediction. In developing this model, the initial data set was filtered, and processed to about 5555 data points. The data was normalized using python programming to prepare the data sets for input into the model. The data set was split into two sets, one set for training and the other for validation. The various models were then developed afterwards and used for prediction. The results show the Multi Linear regression model with an accuracy of 86.556788 percent, the Robust Regression model with an accuracy of 85.766543 percent, the Random Forest model with an accuracy of 96.324562 percent, the Artificial Neural Network model with an accuracy of 95.4325542 and Stacked regression model with an accuracy of 97.9994353 percent. These results indicated that the stack model gives a better prediction of bottom hole pressure when compared to other models.