Automatic Classification of Energy Consumption Profiles in Processes of the Oil & Gas Industry in Colombia

Smart meters provide detailed information about energy consumption behavior for different types of users. With the aim to enable automatic decisions that directly impact the performance of the energy network and the consumption of customers, machine learning (ML) methods emerge as a reasonable alternative. The accelerated growth in the implementation of Advanced Metering Infrastructure (AMI) in Colombia allows the exploration of different ML methods to contribute in the process of automating the analysis, control, and operation of different energy systems including those related with oil and gas exploitation. This paper presents a methodology to automatically discriminate information extracted from 72 smart meter currently operating in energy networks that provide service to different processes of the oil and gas business in Colombia. The obtained results indicate that it is possible to automatically discriminate between transport vs. extraction energy systems with accuracies of up to 69.1%. Similarly, the classification between transport vs. other kinds of systems yields accuracies of up to 65.3%.