Misra Shobhit and Nikolaou Michael
Gas leaks from natural gas wells can create severe environmental and safety problems, as the leaked gas may end up in the air or water. Poor cementing job and other well completion operations of gas wells may allow gas from one rock formation zone to migrate into another and eventually leave the well. Therefore, it is important to ensure that natural gas well construction prevents such unwanted events. The design objective is known as zonal isolation and has been the subject of intense study recently.
In well construction jobs that do not ensure zonal isolation, natural gas may move through the small channels created in the cement sheath in the annulus between the well casing and well wall, and may eventually reach the well head. To detect whether a continuous flow of gas reaches the wellhead a simple procedure is followed, namely a pressure gauge followed by a needle valve are installed, and the needle valve is temporarily opened to lower pressure by bleeding off a small amount of gas, if pressure increase is detected by the pressure gauge. If gas pressure builds up again after the gas bleed off and closing of needle valve, there is clear indication that there is communication between the well head and producing sections of the well through the cemented well wall. The resulting pressure is called sustained casing pressure (SCP) and is undesirable. The objective of this work is to determine what factors contribute toward SCP, and prevent such factors from creating measureable SCP values. Because the effect of all relevant factors on SCP is complicated, the proposed approach relies on statistical analysis of a large volume of data. Through such analysis, a model is built that captures quantitatively the effect of related factors on SCP. These factors are associated with well drilling, cementing and hydraulic fracturing parameters (Nelson 2006).
A statistical technique called partial least squares (PLS) regression was used to create the model. The model structure involves a number of latent variables, which are linear combinations of the original variables, and which are determined through leave-one-out cross-validation in such a way, that correlation between model inputs and outputs is maximized.
Variable Importance in Projection (VIP) variable selection method was used to exclude the less relevant variables. As the term suggests, VIP score of a variable indicates the contribution of a particular X variable to the predicted value of Y (SCP in this case). In our model we excluded the variables with VIP scores less than 0.5.
The contribution of various variables to SCP is shown in the bar chart (Figure 1).
Figure1. VIP scores for variables
Nelson, E. B. (2006). Well Cementing. Texas, USA.
See more of this Group/Topical: Computing and Systems Technology Division