New Dimension to Field Wide Cement Evaluation Based on AI & ML for Pro-Active Well Integrity Management in Matured Oil and Gas Fields

Cement evaluation can be performed when well integrity log measurements are combined with the mud and cementing data. However, a truly comprehensive cement evaluation can only be performed when well integrity log measurements are combined with all available well data, including cementing, mud, drilling, geological, and petrophysical data. However, well logging results are complex and can be ambiguous, and decisions associated with significant risks may be taken based on their interpretation based on physics applied in the logging tools. This paper describes a successful application of artificial intelligence and machine learning for cement evaluation which is independent of various logging providers and reservoir conditions. This in-house solution uses cement bond logs along with other data sets like cement quality parameters, mud parameters, drilling parameters, geological/formation markets information and well type as an input. Historical cement bond log was selection was based on leak off tests. Machine Learning Model was trained separately for each vendor such as Schlumberger, Halliburton, Baker tools with more than 30km of interpreted dataset with high confidence respectively. The ensembled Random Forest and convolution neural network models have been built and analyzed. Cement data quality process has been automated before it is used in model. Based on predictions, well cement ranking and well integrity recommendations were designed respectively for light weight and heavy weight casing designs based on lift type. All workflows’ steps have been automated to aid process standardization across company. The cement evaluation information is integrated GIS maps for finding aerial and reservoir cross section uncertainties. The system has resulted a good database for cement evaluation and well integrity that helped to identify potential integrity issues like possible cross flow, communication between reservoir zones with acquirers, possible fluid communication to surface This approach has replaced manual methods and achieved 95%-time efficiency in decision making cycle. The well retirement process has been improved based on accurate results from this system. This system has helped to challenge the conventional process and been evaluated for ranking the candidates for corrosion monitoring and possible challenges related to well integrity. Workovers have been planned to carry out cement repair jobs in one of matured field for the wells with the age of more than 10 years. This AI based tool is very cost effective, with minimum human bias, improves consistency and allows proactive well integrity management. This tool can be used to any wells/fields irrespective of their geological setting and geographic locations.