• Energy Research

Abstract Newer generation sources and loads are posing new challenges to the conventional power system protection schemes. Adaptive and intelligent protection methodology, based on advanced measurement techniques and intelligent fault diagnosis such as machine learning (ML), is found to be useful to meet these challenges. A large number of research works are reported on ML-based power system fault diagnosis. However, ML techniques are evolving at a very fast pace, and an inclusive, as well as state-of-the-art review on ML-based power system fault diagnosis, is not available in the literature. Given this need and growing trend towards ML, the study presented in this paper aims to provide a comprehensive review of ML-based power system fault diagnosis. At first, efforts have been made to enlist the issues present in conventional fault diagnosis which led to the popularity of ML techniques. Also, a baseline framework and workflow for ML-based fault diagnosis are presented. Next, various unsupervised and supervised learning techniques have been discussed separately which have been used by several researchers for fault diagnosis. The discussion throughout is supported with tabulated facts for fault detection, classification and localization works with techniques used, different simulation tools used, and their application system. The advantages and disadvantages of all the techniques of fault diagnosis have also been discussed which will help the readers in the selection of techniques for their research. A brief review of reinforcement learning and transfer learning is also given as they are gaining popularity in power system-related studies and have the potential to be used for fault diagnosis. Finally, the research trends, some key issues, and directions for future research have been highlighted.

Selection of the most suitable drill bit type is an important task for drillers when planning for new oil and gas wells. With the advancement of intelligent predictive models, the automated selection of drill bit type is possible using earlier drilled offset wells’ data. However, real-field well data samples naturally involve an unequal distribution of data points that results in the formation of a complex imbalance multi-class classification problem during drill bit selection. In this analysis, Ensemble methods, namely Adaboost and Random Forest, have been combined with the data re-sampling techniques to provide a new approach for handling the complex drill bit selection process. Additionally, four popular machine learning techniques namely, K-nearest neighbors, naïve Bayes, multilayer perceptron, and support vector machine, are also evaluated to understand the performance degrading effects of imbalanced drilling data obtained from Norwegian wells. The comparison of results shows that the random forest with bootstrap class weighting technique has given the most impressive performance for bit type selection with testing accuracy ranges from 92% to 99%, and G-mean (0.84–0.97) in critical to normal experimental scenarios. This study provides an approach to automate the drill bit selection process over any field, which will minimize human error, time, and drilling cost.

The oil and gas industry plays a vital role in meeting the ever-growing energy demand of the human race needed for its sustainable existence. Newer unconventional wells are drilled for the extraction of hydrocarbons that requires advanced innovations to encounter the challenges associated with the drilling operations. The type of drill bits utilized in any drilling operation has an economical influence on the overall drilling operation. The selection of suitable drill bits is a challenging task for driller while planning for new wells. Usually, when it comes to deciding the drill bit type, generally, the data of previously drilled wells present in similar geological formation are analyzed manually, making it subjective, erroneous, and time consuming. Therefore, the main objective of this study was to propose an automatic data-driven bit type selection method for drilling the target formation based on the Optimum Penetration Rate (ROP). Response Surface Methodology (RSM) and Artificial Bee Colony (ABC) have been utilized to develop a new data-driven modeling approach for the selection of optimum bit type. Data from three nearby Norwegian wells have been utilized for the testing of the proposed approach. RSM has been implemented to generate the objective function for ROP due to its strong data-fitting characteristic, while ABC has been utilized to locate the global optimal value of ROP. The proposed model has been generated with a 95% confidence level and compared with the existing model of Artificial Neural Network and Genetic Algorithm. The proposed approach can also be applied over any other geological field to automate the drill bit selection, which can minimize human error and drilling cost. The United Nations Development Programme also promotes innovations that are economical for industrial sectors and human sustainability.