MULTIVARIABLE PREDICTION SYSTEM DESIGN BASED ON AUTOREGRESSIVE MODEL WITH EXOGEN FEATURES FOR PRESSURE IN THE UPSTREAM OIL INDUSTRY COMPLEX STEAMFLOOD DISTRIBUTION NETWORK

• Main Author: Faris, Muhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73261
• Institution: Institut Teknologi Bandung

<p align="justify"> Driven by environmental interests and the need for more sustainable energy, the upstream oil industry is currently increasing the efficiency of its production performance through digital strategic transformation. With the development of digital technology, especially in data science, this can be done by implementing the AR (Autoregressive) machine learning method in making future predictions of production variables to maximize system production. However, this method still has some drawbacks. One of them is a model that has not been able to predict as many variable magnitudes as in the case study conducted in this study on well pressure variables in Steamflood production systems that have more than one well. In addition, the model is also not able to consider external factors from the predicted magnitude. Therefore, this final project aims to develop a model that can predict more than one variable and use determining external factors. The developed model adds a Multi-series approach to the previous method so that the prediction of the water vapor pressure entering the well can be made on a system that has more than one well. In addition, there are additional external factors or exogenous features in this model that physically determine pressure variables such as steam generators and valves. By applying the developed solution, the model can predict the pressure in each oil well with satisfactory results. This can be proven both visually where the forecasting results quite follow the test data well and in terms of the measurement metric values the model used has exceeded the Naive value as a comparison value with an average RMSE of 30.5 and a MAPE of 0.04.