Application of Hybrid ARIMA – Decision Tree Algorithm in Reservoir Pressure Surveillance

Abstract

Abstract Bottom hole pressures are valuable source of information for reservoir surveillance and management and are the heart of reservoir engineering. Real – time pressure measurements record pressure data at 5 second interval resulting in enormous accumulation of data. The size and volume of the accumulated data limit the capability of existing analysis software to load and interpret data. This paper presents an improved methodology for data quality checking and data optimization in determining reservoir pressure depletion via Autoregressive Integrated Moving Average (ARIMA) and Decision Tree Model. Dataset was gathered from a representative reservoir from Malay Basin. The ARIMA algorithm presented was designed for quick and efficient data quality checking. The Decision Tree Model in other hand was utilized to select maximum buildup pressure for reservoir depletion point via well status parameters. The maximum pressures were selected from buildup up data when the decision tree conditions were met. Versus classical methods, the algorithm has obtained around 90% similarity. The resulting data were then can fully optimized for reserve reporting and forecasting study i.e. analysis and numerical simulation. The paper also reports on the advantages in the application of ARIMA – Decision Tree Algorithm in pressure surveillance revealing few key advantages namely minimize the need of well intervention and optimized workflow for reservoir engineer to view, utilize, and detect reservoir depletion data. ARIMA – Decision Tree Algorithm is targeted to be installed and integrated in field historian for better overall data analysis and visualization. Results produced from the ARIMA – Decision Tree Algorithm which consist of reservoir pressure depletion data will then improve more advance analysis such as simulation and forecasting in terms of overall speed and accuracy. As a conclusion, this paper presents the importance and application of incorporating Big Data Analytics Algorithm in reservoir management and reporting. Future work, deliverability calculations can be incorporated in the model to identify and rectify any abnormal reservoir behavior.