New Automated Work Flows Enhance Formation Evaluation
This paper bridges the gap between operational petrophysicists and FTS specialists, introducing an automated work flow by which petrophysicists can conduct FTS jobs.
Petrophysical work flows are designed primarily to process static data for traditional openhole logs, which can provide estimates of porosity, saturation, lithology, and mineralogy. However, these estimates normally have a high degree of uncertainty and formation testing and sampling (FTS) data often are required for reservoir-condition calibration. This paper bridges the gap between operational petrophysicists and FTS specialists, introducing an automated work flow by which petrophysicists can conduct FTS jobs.
Wireline formation testers (WFTs) were commercialized in the late 1950s; drilling formation testers (DFTs) were introduced more recently. The primary benefits for FTS by means of wireline (e.g., WFT) or drillpipe (e.g., DFT) tools always has been a link between openhole log static volumetric formation evaluation and dynamic reservoir properties such as reservoir pressure, mobility, and fluid-sample composition, with many more applications under development using advanced testing tools and methods. However, FTS always has faced challenges with the integration of formation-tester data into the petrophysical work flow, limiting the ability to take full advantage of this valuable source of dynamic data.
Current work flows normally involve having an FTS specialist evaluate openhole log analyses, plan testing and sampling jobs, monitor data acquisition and quality-control (QC) results, and report on the final data interpretation. Thus, an FTS specialist should hold a wide spectrum of expertise, including FTS-tool use and data analysis and openhole logging and log analysis, as well as reservoir engineering and reservoir dynamics. Consequently, few petrophysicists are truly qualified to be FTS specialists. Therefore, automating FTS work flows is desirable.
Automated FTS Work Flow
In recent years, new automated methods have been introduced to speed up FTS data delivery. In automated QC data processing, real-time pressure tests are given a rating on the basis of criteria such as pressure stability, temperature stability, drawdown mobility, radius of investigation, and supercharging. More recently, methods have been published that can identify test sequences automatically. Similar methods have been implemented in at least one commercial software offering that demonstrates the benefit of automating the tedious process of selecting testing events manually. By combining the automated QC with the test-event selections, the analysis of data can be automated for real-time monitoring and post-job data processing objectively. With the testing data being processed automatically and the valid or best results being selected from each test sequence, automatically identifying reservoir fluid-flow behaviors is possible.
One of the more critical steps in the FTS work flow is job planning. In this paper, high-level automation methods are introduced that enable petrophysicists to evaluate factors that must be considered to optimize a testing string and determine the overall testing time for the pressure testing and sampling stages of the job. The factors considered are the operational parameters, quality criteria, drilling and formation conditions, and the testing technology used. Tradeoffs can be made quickly to evaluate the role that these factors can play in obtaining the best results.
Automated Job Planning
Automated job planning always is a challenge because of the potential for many unforeseen variables. FTS job planning can be started by evaluating the quality criteria for both pressure testing and sampling and by performing initial simulations to evaluate which tool technology will be the most effective for the targeted job objectives. After a few alternative technologies have been considered, more simulations can be run to estimate the testing and sampling times. Test plans then can consider operational issues and cost tradeoffs. It can be a very time-consuming process.
Because these simulations require detailed inputs concerning formation parameters and an in-depth knowledge of FTS-tool capabilities, options, and limitations, job planning has been the domain of FTS-operations specialists. To make FTS job planning automatic, a simple high-level planner is developed for pressure testing as well as sampling. Both testing- and sampling-automation methods rely on setting best-practice standards for operations and test quality. The planner enables the evaluation of a wide variety of technologies to determine the fit-for-purpose applications, considering QC standards, operational best practices, and total time and cost of the operation within expected uncertainties.
Uncertainties and Automatic Data Analysis
Automation methods are being developed throughout the industry to lower costs and optimize results in all aspects of operations. Recently, methods have been developed for formation testing; these same methods may soon be applied to well testing and production monitoring.
Another important development in automation is event detection. When a formation test starts, the data stream can be used to identify when typical events occur, such as when a pretest drawdown starts and the beginning and end of the buildup. While event-detection methods using a tool’s internal signals, such as when a motor starts or a valve opens, have existed for some time, these signals have not proved accurate. Algorithms recently developed for monitoring the dynamic changes in the pressure and flow data have significantly improved event detection. Now, software is available that is more than 95% reliable in detecting events correctly. When a pretest event is detected, it can be automatically analyzed to determine drawdown mobility. By combining the automatic QC, event detection, and analysis, the entire process has been automated.
The next task to be automated is data interpretation. The first of these methods to be developed are for flow-unit identification with automatic gradients and fluid contacts. These methods are built on statistical methods of analyzing gradient quality. The first step is the QC of the pressure tests. When pressure tests of the required quality are identified, they are included as an initial grouping for analysis. Data points are added progressively to the grouping, and statistical trends are observed to identify intervals, gradient, and contacts. The automation process and statistical methods are illustrated using an example in the following section; another example is discussed in the complete paper.
Example: Automatic Gradient Contact and Uncertainties
This example is illustrated in Fig. 1 (processed openhole logs and formation testing results). All pressure points were selected on the basis of QC ratings and could have been grouped into a single gradient. In this case, the measured variance bars were outside of the expected variance bars—evidence that there were two gradients.
To automatically detect the two gradients shown in the figure, points are added progressively to a single gradient until the variance exceeds the expected range. Then, a new gradient is started for the remainder of the interval. After the two gradients are initially identified, the points near the intersection are switched between the oil and water gradients until the minimum measured variances are determined for each gradient. The same process can be used when a flow barrier is encountered and an offset exists between the gradients.
While the oil/water contact was identified, the oil gradient is of lower quality, which can add to the uncertainty of the contact-depth estimate. This is evident from the measured variance listed in the header, where the relative variance for oil gradient is 13.5% and that for water is 3.6%. Also, as shown in the residual plot, the expected variance red bars are outside of the expected blue bars. The lower quality for the oil gradient is caused primarily by the fact that there are only three pressure points over the limited depth interval. The water gradient has a much better quality because of the number of points and the longer interval. This is verified by comparing the measured gradient variance with the expected gradient variances in Fig. 1, where the water gradient is expected to have a higher variance from Monte Carlo analysis.
The measured fluid contact uncertainty is determined by considering the measured gradient variances by performing a Monte Carlo analysis. For the expected fluid-contact variance, the expected gradient uncertainties are used. The measured contact variance of ±2.80 ft is within the expected variance of ±4.39 ft. This fluid contact was verified by taking fluid samples in the oil gradient interval, confirming the gradient analysis, which generally is consistent with openhole log analysis within uncertainties. The sample closest to the contact did have a small fraction of formation water.
This paper documents methods used to automate the analysis and interpretation of formation-testing data. A number of methodologies have been in development during the past several years and are beginning to be implemented. New algorithms are developed to automate the pressure-testing job planning, which includes the combining of tool technologies to obtain an optimal job design. New algorithms are also developed for the sample planning function, where past experience can be used to estimate sampling times that consider changes in formation conditions and sample quality. The complete automation process is demonstrated with an example. The methods shown in this paper are just starting to be implemented and relate primarily to pressure-testing data. However, automation for sampling and contamination analysis may follow soon. These new methods also can be applied to automation of real-time operations to optimize data acquisition for testing and sampling, which would complete the automation of the entire FTS work flow (i.e., Fig. 1).
This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 187040, “Dynamic Data Analysis With New Automated Work Flows for Enhanced Formation Evaluation,” by M.A. Proett, SPE, S.M. Ma, SPE, N.M. Al-Musharfi, SPE, and M. Berkane, SPE, Saudi Aramco, prepared for the 2017 SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, 9–11 October. The paper has not been peer reviewed.