Optimizing ESP Well Analysis Using Natural Language Processing and Machine Learning Techniques

Abstract The process of extracting and searching for well and field historical data is an essential element in production engineering as it helps capture anomalous events and mitigate measures during critical well operations and analysis. Thus, a new methodology utilizing Machine Learning (ML) and Natural Language Processing (NLP) was used to create an advance model to automate this process. This study explores the process of automating exploring artificially lifted well data via the utilization of ML and NLP algorithms. The proposed method introduces an intelligent petroleum engineering system that takes users’ input and analyze Electrical Submersible pumps (ESP) wells using an advanced ML and NLP model. Moreover, text extraction, cleaning and validation tasks are initially performed to ensure data quality prior to machine and language modeling. Also, word embedding techniques were used to train the model to learn semantic level relationships such as well and field names. The developed model takes the user's questions and transform them to Structured Query Language (SQL) to be executed on cloud servers and evaluated ESP wells using real time data in order to generate the targeted analysis. The NLP system was implemented utilizing engineers’ daily workflows, and evaluated on its ability to retrieve users’ targeted results in the data set based on query entered by the users. Moreover, performance of the system on extraction, mapping and mitigation attributes were evaluated using F1 scores performance matrix. Validating and testing the NLP model disclosed a promising outcome. It’ is worth noting that the advance NLP system scored significantly high F1 record which indicated high reliability of the model to retrieve critical well information. The developed model enabled engineer in the field to utilize such system to explore and extract targeted results from cloud solutions using advanced ML and NLP algorithms. This yielded significant impact on cost as well as time savings of more than 40% due the system ability to provide engineers with intended results during critical times. The developed NLP model enhanced the way engineers explore and study well historical data as the proposed system lead to a fast and substantial improvement in acquiring a desired result. Also, the system provides a detailed well description and analysis to its users. This resulted in significant money and time saving especially in offshore operations where a fast and reliable data is needed to make critical decision in a timely manner leading into avoiding production losses.