Several oil and gas industry applications of artificial intelligence have been presented in the last decade. Machine learning techniques take center stage in most presentations. Prior to performing the actual modeling in data analytic project, the first three steps in the data analytic lifecycle involve planning, data preparation and model planning. Data visualization is an important aspect in the model planning phase as it aids in selecting variables that are important for modeling and in deciding the appropriate choice of ST model to use. Hence, it is important to invest time and resource to carry out this aspect of the project in order to delineate patterns in the dataset. Researches conducted in this area in the oil and gas industry have not maximized the power of visualizations, particularly when it comes to spatio-temporal data analysis.
In this paper, we present applications of spatio-temporal exploratory data analysis for oil and gas datasets using ST-plots, applied on dataset from four unconventional formations in the US: Bakken, Marcellus, Eagleford and Wolfcamp and Bone Spring formations in the Delaware Basin. Several plots are presented, including space-time plots, animations, Hovmöller diagrams,temporal histograms, and boxplots – capturing variations in oil and gas production across space and time.
In the first plot, we show variation in yearly oil production across space with different panels representing time domains. This plot shows the development of an area with few wells in the initial stages and additional wells in later stages. Colors can be added to capture the magnitude of the yearly oil/gas production, normalized with days produced. Areas of the field that contribute to higher production becomes apparent. Animations can be used to present this data monthly as is typical in production data reported to state agencies. For unconventional formations, further normalization can be done with the number of completion stages or lateral length. Hovmöller diagrams are good alternatives with space on the x-axis and time on the y-axis. This plot enables us capture spatial (in)dependence of the dependent variable and can suggest the use of space as a covariate in the final model. Another plot that could be useful is the temporal histogram, in which the third dimension is time. This plot could be useful in deciphering a suitable model for the data and clearly show if the distribution or average changes over time. We highlight how these plots aid in variable selection for spatio-temporal modeling of production data.
These plots are not the typical plots used in presenting oil and gas data and hence the application is new in this regard, in addition to the temporal histogram and boxplots which are new additions.