Data mining/analysis

Big Data for Advanced Well Engineering Holds Strong Potential To Optimize Drilling Costs

Considering the significant weight of drilling costs in upstream ventures, saving even a few hours of drilling could lead to substantial cost savings on the overall capital expenditures (Capex). Thanks to the big data revolution, cost optimization still has strong potential in drilling operations.

jpt-2018-02-techupf1.jpg
Fig. 1—The potential reduction in drilling time enabled by the use of big-data solutions.
Source: kwantis.

Considering the significant weight of drilling costs in upstream ventures, saving even a few hours of drilling could lead to substantial cost savings on the overall capital expenditures (Capex). Thanks to the big data revolution, cost optimization still has strong potential in drilling operations.

Traditionally, drilling performance has been addressed through the analysis of drilling reports written once a day by the operator’s staff. However, high-frequency data are already collected continuously by surface and downhole sensors to address operational safety and continuity. These data contain a huge amount of valuable information that can be used for drilling operations performance analysis (Fig. 1 above).

High-frequency data analysis allows a new level of performance monitoring through the identification of potential invisible lost time and the anticipation of well problems that could be minimized to reduce drilling costs.

Data Valorization, Combination

More than 60 sensors record high-frequency data during drilling, creating sets of millions of data for a single well. These data can be reprocessed through a set of algorithms to identify each single activity with the highest possible level of accuracy and granularity. As an example, reaming is based on the following seven surface logging parameters: bit depth, well depth, rate of penetration (ROP), weight on bit (WOB), torque, mud flow-in, and standpipe pressure.

However not all activities can be automatically detected through surface log-data interpretation. It they cannot be, daily reporting becomes a relevant source of information. Thus, a set of preeminence rules is applied so that the most relevant data can be used and combined as needed at any time to build a complete and precise time breakdown for the analysis of drilling and flat time.

This innovative approach is the basis of the Integrated Drilling Data Discovery (ID3) system, developed by kwantis. ID3 manages one single big-data platform to integrate surface logging data with reporting information (e.g., daily drilling reports and bit reports) and measures lithology, trajectories, and other key data while drilling to create a complete set of drilling performance analyses.

The system is able to break down the drilling sequence to the most detailed activities (e.g., bit on bottom, reaming, circulating, or on slip) with a 5-­seconds accuracy. The information regarding phases, troubles, and equipment are provided by the daily reporting and plotted on the common time or depth bases to create meaningful analytics on a single well or group of offset wells.

The applications of these analyses are multiple and provide new capabilities at any stage of the well life cycle: planning, operating, or post-well analysis.

Analytics for Planning

Usually well planning starts with offset well analysis to estimate the duration of the next well. The higher the level of detail and the availability of past performance measurement, the higher the capability of accurately estimating the well duration. The system provides distributions on a full set of performance factors, such as ROP, tripping in/out (open/cased hole for bottomhole assembly, casing, and wireline), weight to weight, and cementing, which are directly usable for assumptions in probabilistic well planning.

These key performance indicators (KPIs) are free from any subjective interpretation, and the given distributions are a direct input as performance variables for time estimates of future similar wells for statistical sampling or being fitted with normal/log-normal distributions.

These statistics can be further filtered by phase, lithology, and equipment (e.g., bit type) to identify the most cost-­efficient equipment (Fig. 2).

jpt-2018-02-techupf2.jpg
Fig. 2—The ID3 system combines operations reporting and sensor data. Source: kwantis.

 

The use of ID3 for well planning has demonstrated a much higher accuracy in time estimates because of the availability of precise performance analytics from offset wells. This additional accuracy allows operators to better allocate financial resources on the various investment initiatives. The saving generated by the optimization of financial resources has been measured on a 33-well offshore campaign at 1.4% of the Capex.

Moreover, for better understanding of how an issue originated, nonproductive time (NPT) recorded in daily drilling reports is plotted on main-sensor data curves to give an easy and rapid insight. In this way, patterns or trends of specific data in the few hours preceding the problem can be analyzed, thus giving evidence of the occurrence factors and enabling a better capability to anticipate the problem in the future. As an example, the combined analysis of hookload, torque, and standpipe pressure can provide evidence of an imminent sticking event.

Analytics for Operations

Performance monitoring and optimization. During the operations, sensor data are transferred from the rig through wellsite information transfer standard markup language (WITSML) protocol and analyzed on a near-real-time basis (a batch of records covering 10 minutes is needed for consistent statistical calculations). These analyses are used to measure the actual performance of the operations and benchmark it with the best performance measured on the same basis (offset wells, phase, operation, lithology, and equipment).

The comparison can be broken down to the single operating parameter, such as WOB or rotations, to optimize the drilling performance in real time.

Prevention. Another real-time application is the anticipation of well problems by means of highlighting patterns that have been previously identified as possible signals that precede the occurrence of a problem. These patterns can be directly set, and specific warnings can alert the crew on a situation becoming more critical.

Post-Well Analytics

Service performance. Detailed statistics provided by the analysis of multiple sources allow an accurate and impartial measurement of the operations once the well is finished. The impartiality of the analyses is the key to the evaluation of the contractor’s performance and the optimization of the contractual strategy. Here are some examples:

  • Rig contractor: Connection time and tripping speed in cased hole are pertinent indicators of the performance of a given rig crew that can be used as parameters in a specific rig contractor KPI.
  • Bit selection: The actual ROP, combined with mechanical specific energy and dull grading, provide unequivocal information about the bit performance in a specific section, formation, and lithology. Such information can be advantageous in the future sourcing of drilling equipment.
  • Surface/downhole logging service: The quality of the data transferred is one of the parameters that can provide an objective indication of the contractor performance. Such indication, combined with other qualitative indication, can build a proper KPI.

Lessons learned. A pertinent learning indicator is the number of hours to drill 100 m, if limited to a number of activities that are driven by the well length. Again, the use of rig-sensor data guarantees the accuracy of the indicator and provides a good evaluation of the learning factor in a well campaign. This kind of analysis can be fine-tuned to the single phase or to specific activities to address possible accelerations based on the best performance among offset wells (considered as the technical limit).

The calculation of this statistical technical limit and the implementation of associated actions enable operators to increase the performance on the overall drilling phase. In a recent case, the recommendation to reduce specific activities and the selection of the appropriate equipment together enabled an 8% reduction in drilling time for an offshore well.

Conclusion

The added value of the analytics continuously increases with the consolidation of additional data. Also, it can be extended to the prediction of specific NPT areas through the introduction of machine-learning algorithms and to the training of the well engineers by leveraging the rich knowledge base.

Current trends in scalable databases enable much easier handling of such a huge volume of information, allowing a scaleup to petabytes of data on commodity hardware. Data transmission is increasingly robust as a result of the wider adoption of the WITSML format as a data exchange standard. Data science algorithms can be smoothly applied on top of this information by using increasingly available open source libraries.

Systems such as ID3 make a right step toward the goal of a digital oil field. Data already collected on the rigsite should be leveraged through the use of advanced analytics to maximize savings through increased operational efficiency. In this historic moment of oil price uncertainty, the oil and gas industry needs to embrace the big-data revolution to ensure the sustainability of its exploration and production investments.