Aplicação de técnicas de inteligência artificial para modelagem de falhas de integridade nos sistemas de revestimento de poço

Abstract

Currently, oil is one of the most consumed sources of energy in the world, and this demand is constantly growing. However, hydrocarbon production is filled with challenges. These challenges include the complex nature of its operations, as well as the harsh environment to which its facilities are exposed. This is the case, for example, with offshore wells subject to extreme environmental conditions, under high levels of pressure and temperature. This work addresses anomalies that occur throughout the lifespan of well productions, which can lead to situations involving loss of integrity, resulting in environmental damage, risks to human lives, and financial loss. Thus, the classification of unwanted events during hydrocarbon production benefits the industry as a whole, supporting techniques capable of reducing costs associated with well interventions, minimizing human error, avoiding rework, and ensuring operational safety. Literature and practice demonstrate that artificial intelligence (AI) methods can contribute in various ways in this context. In this sense, the objective of this work is to develop an automated model using AI techniques to classify anomalies based on publicly available data series provided by Vargas et al. (2019). This work utilizes the K-nearest neighbors (KNN), Support Vector Machine (SVM), and Non-Myopic Early Classifier techniques, along with the sklearn machine learning library as a data analysis tool and Python for algorithm development. In this context, the techniques applied achieved the best performance in terms of accuracy with KNN, with a value of 97.56%. Additionally, the analysis is conducted using metrics such as balanced accuracy, precision, recall, and F1 score. The results obtained demonstrated a significant improvement compared to the public data.