Utilizing Neural Network for Predictive Production Forecasting and Environmental Emission Monitoring
Loading...
Date
2023-11-29
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This paper explores two distinct engineering domains where machine learning techniques are applied, with neural networks playing a pivotal role in both chapters. In the first chapter, atmospheric dispersion modeling is examined, comparing the Gaussian plume model with the more complex Gaussian puff model. The latter, while accurate, demands substantial computational resources. An Encoder-decoder architecture of the Long Short-Term Memory network is used to predict the Gaussian puff model's results based on the simpler Gaussian plume model, resulting in a significant reduction in computation time. In the second chapter, the focus shifts to production forecasting in the oil and gas upstream sectors. Traditional methods offer detailed insights but are computationally intensive, especially in complex reservoir scenarios. The proposed workflow integrates geological model compression techniques while maintaining forecasting accuracy with neural network-based regression. This approach significantly reduces the size of the geological model while maintaining forecasting accuracy. Achieving substantial compression ratios through dimensionality reduction techniques, machine learning predictions demonstrate remarkable efficiency, outperforming traditional simulation software in terms of processing speed. Overall, this work paves the way for future applications of machine learning in engineering, enabling rapid and efficient predictive modeling in complex scenarios.
Description
Keywords
Gaussian plume model, Gaussian puff model, encoder-decoder, Long Short-Term Memory network, production forecasting, geological model compression, dimensionality reduction, network-based regression, accuracy, computational time