Transformation, Analysis and Visualization of Distributed Temperature Sensing Data generated by Oil Wells

This research paper examines the full lifecycle of the case of turning the gas and oil industry to a data-driven operation model. The study presents the approach of re-engineering production data, building a predictive model for temperature forecast, statistical analysis, and visualization of Distributed Temperature Sensing (DTS) data provided by the oil-gas industry. For better analysis, the raw data have been pre-processed and organized according to the proper model. Furthermore, after the data organization, we proceed with observing relationships among three features (Date, Depth and Temperature), analyze vague upheavals and similarities utilizing plot histograms, scatterplots, box plots, heatmaps, violin plots for better visualization. Since the drastic temperature change indicates the anomaly, several alternative Outlier Detection Techniques are offered to predict early equipment failure and prevent production outage. Our results indicated a high correlation between depth and temperature, presence of trend in temperature distribution, and temperature drops in specific ranges. Proper analysis of the data allows the specialist to understand reservoir performance and prolong the production file of the wells.