Effect of Filtering Techniques on Interpretation of Production Log Data Mohammad Aghabeigi, aghabeigi60@yahoo.com Ali Reza Tabatabaei Nejad, tabalireza@yahoo.com Egbal Sahraei, sahraei@yahoo.com Mehdi Kalantari, kalantaridehaghi@yahoo.com Faculty of Chemical Engineering, Sahand University of Technology, Tabriz, Iran Abstract Production logging procedures are carried out in production and injection wells to diagnose well problems and identify flow profiles. These tasks can be used for optimization of productivity or injectivity of the well by identification of potential oil producing zones, blocking water producing zones, blocking thief zones, reperforation, etc. Production logs are run by means of a series of logging tools for measuring different properties of the fluid flowing in the well. Each tool measures a specific property and then sends data to the surface for recording. The raw data recorded by each tool can be directly used for interpretation, but usually raw data are mixed with noises, uncertainty, and errors. Proper filtering of raw data before processing and interpretation is essential in a good production log analysis. In this work, a program was written in Microsoft Visual Basic for plotting lag data, filtering, smoothing and statistical analysis of raw data. Filtering and smoothing techniques are applied to raw data of production log tests of an oil production well and effectiveness of each technique is investigated. Key Words: production log, spinner flow meter, flow profile, production zone, Log interpretation, Data processing 1
1. Introduction Production log is a record of one or more in-situ measurements that describe the nature and behavior of fluids in or around the borehole during production or injection. Production logs are run for the purpose of analyzing dynamic well performance and the productivity or injectivity of different zones, diagnosing problem wells, or monitoring the effectiveness of a stimulation or completion. The term is sometimes extended to include logs run to measure the physical condition of the well, for example cement bond and corrosion logs [1]. The theory of production logging is a combination of fluid mechanics, multiphase flow, heat transfer and radioactivity, therefore advancement in these topics has resulted in advancement of production logging. The earliest production logs consisted of temperature logs (1930 s) and flowmeters (1940 s), to which were soon added fluid density and capacitance logs (1950 s). Flow rate measurements were gradually improved by the development of tracer logs and improvement to the basic spinner flowmeter [2]. Production logging techniques were adequate for near-vertical wells with single or biphasic flow, but could be misleading in highly deviated, and especially horizontal, wells. New techniques were developed starting in the 1980 s which focused on local probes to measure liquid and gas holdup at different points in the borehole, nuclear techniques to analyze the total holdup of all three phases, and phase-velocity logs for the analysis of individual fluids. At the same time, complex flow structures and flow regimes have been studied more extensively using flow loops [3]. Production logging traditionally encompasses a number of well logging techniques run on completed injection or production wells, with the goal being to evaluate the well itself or the reservoir performance. In recent years, however, the role of production logging has expanded to include applications that start at the early stages of drilling and that last throughout the life of the well [2]. The purpose of production logging is to evaluate fluid flow inside and outside pipe or, in some cases, to evaluate the well completion directly. The most common application of production logging is the measurement of the well s flow profile, the distribution of flow into or out of the wellbore. Major applications of production logging are as follows: Detection of mechanical problems of the well, Evaluation of completion efficiency, Monitoring production and injection profiles, Determining reservoir characteristics, Evaluation of treatment effectiveness, and Detection of thief zones and channeled cement. Production logs are run by passing production log tools once or several times in the borehole while the well is on production. The velocity, density, water holdup, temperature, and pressure of flowing fluid are measured continuously. In addition, the natural Gamma ray (GR) emitted by formation rock, well bore and/or tubing diameter, cable speed, and casing collar locations are measured. These are basic measurement, however some new measuring techniques have been developed which add up to accuracy and certainty of measured parameters. Production log tools have a series of sensor and transducers for measuring the abovementioned properties directly or indirectly. 2
Like all other sensors, sensors used in production logging have some inaccuracy, error, and noise in their measurements. The errors result in incorrect interpretation and analysis. There are several methods for removing noises, smoothing raw data, and filtering out unwanted portions of data [4]. In this paper, several filtering and smoothing techniques are applied to raw data of production log test of a well producing from a naturally fractured reservoir in southwest of Iran. The effectiveness of any technique is compared with other methods and also with original raw data. 2. Theory Any data driven from log files are combined with environment noises to some extent. Log data are plotted against depth with the vertical axis being measured depth. Usually two or more curves are plotted together in a track. Usually some curves show very sharp spikes and discontinuity in some depths because the borehole is a harsh environment. Other sources of error and noises are tools inaccuracy, data transfer cables, electrical noises in the logging site, and random noises. Therefore some curves should be filtered before processing and interpretation. Note that casing collar locator (CCL) and Gamma ray data don t need filtering, because they are used just for depth matching. Continuous water holdup, fluid density, temperature and spinner data require filtering and smoothing. Some curves such as temperature, pressure, fluid density, and water holdup are similar for different log runs. These curves can be stacked together for additional accuracy and random noise removal, because these curves should be identical for any individual run (up or down pass, with any cable velocity). These curves can be compared for different runs to ensure the accuracy of the particular logging tool. Whenever a bad or poor quality curve was detected, it can be replaced by a synonym curve from other runs. Finally these curves are stacked together to provide a more reliable data for final interpretation. Other curves such as continuous spinner or full-bore spinner cannot be stacked together because they are velocity dependent. There are a variety of filtering techniques which can be categorized as smoothing, low cut and high cut filters. 2.1 Smoothing In statistics and data processing, to smooth a data set is to create a function that attempts to capture important patterns in the data, while leaving out noise [5]. Many different algorithms are used in smoothing. One of the most common algorithms is the moving average, which is performed by recursive or non-recursive methods. Recursive filter is a type of filter which re-uses one or more of its outputs as an input. This feedback typically results in an unending impulse response, characterized by exponentially growing, decaying, or sinusoidal signal output components [6]. 2.2 Low and High Cut Filters A low and high cut or band-pass filter is a filter that passes frequencies within a certain range and rejects or attenuates frequencies outside that range. These filters can also be created by combining a low-pass filter with a high-pass filter. An ideal filter would have a completely flat pass band and would completely attenuate all frequencies outside the pass band. The bandwidth of the filter is simply the difference between the upper and lower cutoff frequencies [5]. 3
2.3 Low Cut or High-pass Filter A high-pass filter is a filter that passes high frequencies well, but attenuates frequencies lower than the cutoff frequency. The actual amount of attenuation for each frequency varies from filter to filter. It is sometimes called a low-cut filter; the terms bass-cut filter or rumble filter are also used [6]. 2.4 High Cut or Low-pass Filter A low-pass filter is a filter that passes low frequencies well, but attenuates frequencies higher than the cutoff frequency. A low-pass filter is the opposite of a high-pass filter, and a band-pass filter is a combination of a high-pass and a low-pass. It is useful as a filter to block any unwanted low frequency components of a complex signal while passing the higher frequencies. Of course, the meanings of low and high frequencies are relative to the cutoff frequency chosen by the filter designer [6]. 3. Case Study The production log test of a well drilled in a naturally fractured reservoir (NFR) in southwest of Iran is discussed as the case study. This well produces oil, water and gas simultaneously at surface. The specific gravity of stock tank oil is 0.67 and producing gas-oil ratio is 650 SCF/STB. The oil formation volume factor (B o ) is 1.41 res. bbl/stb for this oil. The well has been drilled vertically. There are two perforated intervals in this well. The top perforations are located at 2597 to mdd and the bottom perforations are located at 2615 to 2618 mdd. The production log package for this well included Gamma Ray (GR) tool, casing collar locator (CCL) tool, Quartz Pressure (QP) tool, Temperature (TEMP) tool, Fluid Density (DENR) tool, Continuous Water Holdup (CWH) tool, X and Y Calipers (XCAL and YCAL), Inline Spinner (ILS), and Caged Fullbore Spinner (CFB), all manufactured by Sondex. All tools were calibrated according to specified test methods in their technical manuals. 4. Results A program was written in Microsoft Visual Basic for plotting lag data, filtering, smoothing and statistical analysis of raw data. Using this program, the raw data was filtered and then interpreted. Figs. 1 through 4 show the raw data curves of spinner rotation speed, continuous water holdup and fluid density, temperature, and calipers, respectively. As these figures show, there are lots of noises in the raw data. As shown in Fig. 1, the upper portion of spinner rotation data are generally in large errors. 4
2550 0 20 40 60 80 100 120 Spinner Rotation (rps) Figure 1: Raw data of spinner rotation speed. Fig.2 shows the variation of both fluid density and water holdup. All density data have noises, but only lower portion of water holdup data are affected by noises. 2550 DENR CWH 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Fluid Density (g/cc) and Water Holdup Figure 2: Raw data of fluid density and water holdup. In Fig.3, the raw values of fluid temperature show very sharp spikes thorough the data points, 5
2550 171.5 171.7 171.9 172.1 172.3 172.5 172.7 172.9 Temperature ( o F) Figure 3: Raw data of fluid temperature. and finally Fig. 4 shows that there are rapid changes in wellbore diameters and discontinuity in raw data. 2550 XCAL YCAL 3.5 3.6 3.7 3.8 3.9 4 4.1 4.2 4.3 4.4 4.5 Wellbore Diameter (in) Figure 4: Raw data of two diameters of the wellbore (X and Y calipers). The raw data was first filtered by low and high cut filters were then was smoothed using moving average technique. Figs. 5 through 8 shows the effect of smoothing and filtering. All noisy data points of spinner rotation speed, temperature, fluid density, water holdup and wellbore diameters have been removed and replaced by interpolated values. The filtered data was then used for final processing and interpretation. The interpretation can be grouped into two parts: qualitative and quantitative interpretation methods. 6
2550 0 20 40 60 80 100 120 Spinner Rotation (rps) Figure 5: Filtered and smoothed data of spinner rotation speed. 2550 DENR CWH 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Fluid Density (g/cc) and Water Holdup Figure 6: Filtered and smoothed data of fluid density and water holdup. 7
2550 171.5 171.7 171.9 172.1 172.3 172.5 172.7 172.9 Temperature ( o F) Figure 7: Filtered and smoothed data of fluid temperature. 2550 XCAL YCAL 3.5 3.6 3.7 3.8 3.9 4 4.1 4.2 4.3 4.4 4.5 Wellbore Diameter (in) Figure 8: Smoothed and filtered data of two diameters of the wellbore (X, Y calipers). 5. Qualitative Interpretation 5.1 Temperature data Temperature data are the main key for production log interpretation because it can be used to check and confirm other data [2]. Fig. 7 illustrates the filtered and smoothed temperature data. There are two main observations at 2615 mdd and 2597 mdd. In 2615 mdd, temperature increases and in 2597 mdd, the temperature decreases considerable. In case of this well, temperature log shows a point of water or oil entry at depth of 2615 mdd and a point of gas entry at depth of 2597 mdd. 8
5.2 Density data Fluid density data are valuable because they help us to determine the type of fluids in the wellbore and also fraction of heavy and light phases. Fig. 6 shows the fluid density data of well under study. As this figure shows, there is heavy brine and or remaining drilling mud below the bottom perforations (2618 mdd) which is static. At the bottom perforations (2615 to 2618 mdd), a fluid with density greater than 1 is produced which must be water. At top perforations (2597 to mdd), there is a rapid change in fluid density which reveals that oil and gas are produced in this zone. 5.3 Water holdup data For determination of amount of each fluid, water holdup data is needed in addition to density data. Fig. 6 shows the water holdup data recorded by continuous water holdup tool. Like density and temperature data, it can be seen that there is a static fluid below the bottom perforations, a heavy phase (mostly water) produced in bottom perforations, and light phase (oil and gas) produced at top perforations. There are three unknown parameters in fluid type determination, including: water holdup, oil holdup, and gas holdup. Using density and water holdup data, one can determine these three unknowns. 6. Quantitative Interpretation The raw log data file was input to the program and then data was filtered and smoothed. Finally 20 points was selected as stable zone. In-situ spinner calibration was made for every station using the program. Table 1 covers the spinner rotation speed, cable velocity, and calculated spinner slopes for 20 stations. Table 2 lists the interpreted data for each station. Table 1: Raw data of CFB spinner and calculated slopes. Station No. Top Depth Bottom Depth Tool Speed Spinner Slope (m) (m) (ft/min) (RPS) (min/ft.sec) 1 2618 2618.2-39.18-22.2 0.6628 2 2617.5 2617.8-39.26-21.8 0.6654 3 2616.8 2617-39.1-22.2 0.7008 4 2616.1 2616.5-38.98-21.2 0.6844 5 2612.4 2612.9-39.06-34.3 0.8577 6.6 2611-39.07-36.7 0.7919 7 2608.7 2609.3-38.95-37.4 0.776 8 2603.3 2603.9-38.63-37.8 0.7438 9 2601.2 2601.7-38.91-37.6 0.7507 10 2599 2599.4-38.79-36.7 0.7688 11 2597.9 2598.4-39.03-36.1 0.736 12 2596.9 2597.3-39.75-37 0.8414 13 2595.9 2596.3-40.01-48 0.7768 14 2594.8 2595.3-40.4-61.4 0.7756 15 2592.3 2592.8-40.87-82.5 0.7135 16 2589.8.4-41.17-86.5 0.7173 17 2582.1 2582.6-40.59-93.5 0.7263 18 2569.6-39.92-94 0.7127 19.3-40.52-93.5 0.7132 20 2548.4 2550.1-41.21-94.2 0.717 9
Table 2: Quantitative interpretation of production log data. Middle Depth Fluid Velocity Water Holdup Oil Holdup Gas Holdup Total Rate Water Rate Oil Rate Gas Rate (m) (m/min) (fraction) (fraction) (fraction) (bbl/day) (bbl/day) (bbl/day) (ft 3 /Day) 2618.1 0 0.959 0 0.0374 0 0 0 0 2617.65 0 0.957 0 0.0422 0 0 0 0 2616.9 0 0.962 0 0.038 0 0 0 0 2616.3 0 0.965 0 0.035 0 0 0 0 2612.65 0.2354 0.878 0.0298 0.0922 18.813 16.518 0.561 9.7373.8 1.8403 0.88 0.033 0.087 147.47 129.77 4.8656 72.044 2609 2.3391 0.883 0.0299 0.0871 187.66 165.71 5.6176 91.744 2603.6 3.0839 0.886 0.0281 0.0859 248.19 219.9 6.979 119.68 2601.45 2.8275 0.884 0.0281 0.0879 228.26 201.79 6.4105 112.68 2599.2 2.96 0.881 0.0355 0.0835 186.71 164.49 6.628 87.539 2598.15 2.5346 0.843 0.0834 0.0736 208.79 176.01 17.42 86.246 2597.1 3.07 0.485 0.5074 0.089 87.659 174 44.482 3.7205 2596.1 5.5105 0.347 0.545 0.0916 453.6 187 254.66 233.26 2595.05 9.8068 0.324 0.5405 0.1355 798.51 258.72 431.63 607.35 2592.55 18.9124 0.294 0.5583 0.1477 1521 447.17 849.17 1261.4.1 20.0923 0.294 0.5591 0.1469 1610.5 473.48 900.4 1328.4 2582.35 22.2992 0.297 0.5569 0.1461 1733.9 514.96 965.54 1422.8 2569.8 23.1159 0.296 0.5563 0.139 1795.2 531.38 997.18 1485.4.15 23.21 0.297 0.5566 0.1474 1794 530.93 998.67 1486.2 2549.25 23.344 0.297 0.5568 0.143 1795.1 530.23 999.13 1484.9 Figs. 9 to 12 show the results of interpretation of production logging in the well under study. Fig. 9 shows the fluid velocity in the borehole. It can be shown that the main fluid velocity comes from top perforations. Fig. 10 shows oil and gas holdups that are calculated using density and water holdup at each station. It seems that there is a small amount of gas and oil produced at bottom perforations and lots of oil produced at top perforations. Fig. 11 show oil and water flow rates and flow profile at the wellbore. And finally Fig. 12 shows the flow rate of gas interpreted from log data. 10
2550 Depth, m 0 5 10 15 20 25 30 Fluid Velocity, m/min Figure 9: Total fluid velocity calculated from spinner data. 2550 Oil Holdup Gas Holdup Depth, m 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fluids Holdups Figure 10: Oil and gas holdups calculated from density and water holdup data. 11
2550 Depth, m Oil Flow Rate Water Flow Rate 0 200 400 600 800 1000 1200 Flow Rates, bbl/day Figure 11: Flow contribution and flow rates of water and oil. 2550 Depth, m 0 200 400 600 800 1000 1200 1400 Gas Flow Rates, ft 3 /day Figure 12: Gas flow rate interpreted from spinner data. 12
7. Conclusions 1. The quality of production logs is strongly affected by tools calibration, accuracy and maintenance. Therefore, before any production logging operation, a detailed calibration for all tools are recommended. Using old, malfunctioning or inaccurate tools will result in large error in recorded data and consequence interpretations. 2. Spinner, fluid density, water holdup and cable speed data should be filtered and smoothed before using in processing and interpretation. 3. Non-recursive moving average technique was found to be effective in removing spikes and rapid changes in raw data. In case of very noisy data, this technique can be used two or three time to ensure smoothing data. 4. Before smoothing raw data, the low and high cut filters should be applied. Any property has a definite range. The range before and beyond this range should be filtered using band-pass filters. Nomenclature B o Oil formation volume factor (rbbl/stb) CWH Continuous Water Holdup tool mdd meter drill depth (meter) CCL Casing Collar Locator CFB Caged Fullbore Flowmeter RPS revolution per second (1/Sec.) ILS Inline Spinner NFR naturally fractured reservoir GR Gamma Rays QP Quartz Pressure tool X-YCAL X and Y caliper 8. References 1- Oil Field Glossary, Schlumberger Online Glossary, 2006. 2- Hill, A. D.: Production logging-theoretical and interpretive elements, SPE Monograph Vol. 14, Henry L. Doherty Series, Richardson, Texas, 1990. 3- Production Log Interpretation, 2 nd edition, Schlumberger, 1973. 4- Theys, P., Log Data Acquisition and Quality Control, 2 nd edition, Editions Technip, Paris, France, 1999. 5- Press, W.H., Teukolsky, S.A., Vetterling, W.T., and Flannery, B.P.: Numerical Recipes in C++, 2 nd edition, Cambridge University Press, 2002. 6- Jensen, J.L., Lake, L.W., Corbett, P.W.M, and Goggin, D.J., Statistics for Petroleum Engineers and Geoscientists, Elsevier, 1997. 13