Advanced Screening in Oil Recovery Processes using Attention Mechanisms Combined with Deep Convolutional Neural Networks

Detecting oil spill is essential for monitoring the environment and disaster management thus, helps to minimize ecological damage. Traditional methods, such as satellite imagery analysis and radar sensing, often struggle with low spatial resolution, weather interference, and not suitable for distinguishing oil from similar substances. These issues lead to inaccurate detection and delayed response time. To overcome these issues, this paper proposes an integrated topology combining attention mechanisms with Deep Convolutional Neural networks (DCNN) for Enhanced Oil Recovery (EOR) screening. The raw data is preprocessed by Exploratory Data Analysis (EDA), scaling and normalization approaches. The preprocessed data is visualized graphically to identify the patterns or features underlying in the complex dataset. A Deep Learning (DL) based classification model is integrated with attention mechanism that effectively detects and classifies the EOR chemicals by analyzing its features. The proposed model is implemented in the Python software and it achieves the accuracy of 98% allowing more precise screening for ROR.