Use of Machine Learning in Microseismic Monitoring for Thermal Operations in Cold Lake, AB, Canada

Abstract This paper presents a novel advancement in the microseismic data processing workflow for monitoring thermal operations at an unconventional bitumen reservoir in Cold Lake, Alberta, Canada. The new enhancement to the monitoring workflow enables fast processing of the microseismic data, by filtering upwards of 94% of noise files from the daily dataset before the files are reviewed and analyzed. The software consists of a deep learning-based application trained to recognize noise files and classify event files according to their predominant source. The tool was deployed in production on an Azure-hosted data manager for almost a year, after extensive testing and achieving adequate level of performance.