Abstract
Abstract
The process of validating and monitoring pressure and temperature data is a key element in production engineering as it ensures proper well evaluation. Consequently, wells are frequently surveyed for better reservoir monitoring and accurate measurement of productivity. This study explores a validation method using advanced Artificial Intelligence (AI) and Machine Learning (ML) classification models that were developed utilizing historical data to automatically validate conducted pressure and temperature measurement and communicate observations and alerts to engineers.
The proposed method validates pressure and temperature measurement using ML model based on previously conducted measurement using advanced algorithms. The developed model fed on pre-identified key production and pressure/temperature parameters that are used to classify surveys. Moreover, these parameters were selected based on historical data and measurement reports and then were analyzed and ranked to identify the most important parameters on the performance and accuracy of the model utilizing advance algorithm and correlation analysis. This is to predict and classify test measurement via the utilization of a non-linear relationship through the use of data-based analysis alongside physics-based analysis.
The data set of conducted pressure and temperature measurement was split into two groups i.e. training and testing. In addition, a K-fold cross-validation was performed on the training set to validate the performance of all considered and selected ML models. The results of each ML model were then compared for accuracy and the Random Forest Classification algorithm was selected. The developed classification model achieved an overall accuracy level of more than 95%. Validating and testing the model on several cases showed promising results as irregularities are detected in advance before engineers evaluate these conducted measurements. The developed model enabled an effective utilization of previous measurements to validate newly conducted ones and, consequently, alert engineers of any detected anomalies in advance. This yielded significant impact on cost and time savings due the model's ability to automatically predict and validate the conducted measurements.
The pressure and temperature validation model enhanced monitoring and interpreting the production/pressure and temperature measurements and resulted in a substantial improvement in timesaving. The model is developed to be run on the Cloud and it provides an automatic validation of the newly conducted measurements. In addition, it also delivers an alerting mechanism to engineers for any observed abnormalities.
Reference18 articles.
1. A gentle introduction to the gradient boosting algorithm for machine learning;Brownlee;Machine Learning Mastery,2020
2. High integrity wellbore surveys: methods for eliminating gross errors;Ekseth,2007
3. Support vector machine — Introduction to machine learning algorithms;Gandhi;Medium,2018
4. What's a deep neural network? Deep nets explained;Johnson;BMC Blogs,2020
5. Deep neural networks;Listlink;KDnuggets,2020
Cited by
3 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献