Abstract
Monitoring of integrity of plugged and abandoned (P&A'ed) wells is of interest for the oil and gas industry and for CO2 storage. The purpose of this study is to develop artificial intelligence (AI)-based approaches to detect anomalies or defects when monitoring permanently plugged wells. The studied solution is based on the analysis of electromagnetic (EM) data. We consider an offshore setting where the EM signal is generated in presence of a P&A'ed well and the resulting electric field is recorded at the seafloor. Numerical simulations are used to train an AI algorithm to classify the modelled EM features into predefined well integrity classes. We consider four scenarios: (1) no well, (2) well with three 20 meters thick cement barriers of thickness, (3) well with three cement barriers of 60 meters thickness, and (4) well with three cement barriers of 100 meters thickness. Convolutional neural networks (CNNs) are tested as the AI algorithm in this study. After training the algorithm on 80% of the data, it shows an accuracy of 95.36% on the test data. P&A'ed well integrity monitoring currently remains limited to local observation and symptom identification, but this study shows that there is great potential for developing remote non-invasive well integrity monitoring techniques.