Drilling Data Quality and Reliability: A Novel Algorithm for Data Correction and Validation

The difficulty of drilling data analysis stems from the fact that such data often contain errors and disturbances such as missing samples, desynchronization, misinterpretation, ambient noise, and unfit equations and models. In this study, an algorithm is developed to identify different operations during the drilling processes and automate the procedure of validating data quality. The proper analysis and processing of drilling data is crucial in ensuring its quality. Initially, the algorithm separates drilling data reported in 5 sections from the Equinor Volve dataset into different intervals depending on their mode of operation. Additionally, the algorithm performs data correction and provide analysis on data quality. The data is corrected based on a series of physics- and operation-informed conditions. Key performance indicators (KPIs) are calculated for the rate of penetration (ROP), weight on bit (WOB), bit depth, measured depth, and torque. The KPIs include validity, accuracy, and Kuder-Richardson 20 (KR-20) scores. The KR-20 score, which is a measure of the lower bound of the reliability of data, is selected since it considers the difficulty of measurement of each attribute to be unequal, i.e. some attributes may be more prone to error than others. Based on the results, the corrected data displays better correlation between the aforementioned drilling data. The results prove that the intelligent software analysis provides an automated workflow that allows the separation of all operations and the required time in addition to the nigh instantaneous generation of corrected data and data uncertainty qualification. The detection of invalid data points was performed by investigating every operation i.e. rotating off bottom, drilling mode and evaluate the corresponding measurements. The analysis showed that approximately 76% of ROP values are invalid data, which validates the importance of data correction in order to be used for validation. The developed algorithm allows the rapid and reliable analysis of unprocessed drilling data to facilitate decision making and quality control. The key advantages of the intelligent algorithm are that it provides fast data assimilation and comprehensive analysis of drilling parameters by allowing rapid visualization and quantification of data quality and reliability. The correlation of drilling parameters for various operations can be evaluated and visualized with correlation matrices (heatmap). The primary qualification from the original set of data shows a low correlation i.e. rate of penetration values on rotating off bottom operation, which indicates low data validity. The current method and procedure show a significant correction of data point correlation from −1 to up to 1 depending on the operation. More importantly, for a specific set of the original data, ROP showed a validity of 0.23 in contrast 0.75 for bit depth and with accuracy of 0.58 and 0.99 respectively.