Artificial lift methods typically drive Coal Seam Gas (CSG) wells, and Progressive Cavity Pump (PCP) is the preferred method of lift with Australian CSG operators. CSG wells in Australia are typically equipped with necessary instrumentation and automation systems to provide real-time data gathering for monitoring and control purposes. Real-time data gathered from CSG wells presents an opportunity to better understand PCP performance by identifying anomalous pump behavior.
However, before undertaking any real-time analytics exercise, it is pertinent to carry out Exploratory Data Analytics (EDA) to understand time series data behavior and extract relevant features; and this exercise is particularly important with multi-variate data sets. Obtaining significant data features from multivariate time series data can help define which analytics and machine learning methods could be exploited to analyze PCP performance in near real time.
This paper will discuss EDA methodologies that can help streamline time-series data normalization and feature extraction techniques. A three (3) year time-series dataset, gathered from forty-two (42) CSG wells, will be used to showcase EDA methodologies utilized as part of this research. All EDA activities covered in this paper are based on the Python programming language and its supporting libraries.