While most types of logs are used to characterize the wellbore, formation, and fluids prior to well completion, a number of logging tools are available to provide information during production operations and beyond. This article discusses the various types of production logs and how they can often be used together to provide crucial information for understanding and resolving problems.. Production Logging is one of a number of cased hole services that includes cement monitoring, corrosion monitoring, monitoring of formation fluid contacts (and saturations), perforating and plug and packer setting. Services performed in dead, overbalanced, conditions can use relatively simple surface pressure control equipment and are often performed using large open hole style logging cables. Wells with surface pressure typically have a completion tubing of relatively small internal diameter, ID, compared to the casing size across the reservoir. This reduced ID means that cased hole toolstrings for live wells are typically sized at 1-11/16" in order to pass through the smallest nipple in a 2-3/8" tubing.
When combined with relatively mature subsea production technologies (see subsea chapter on well systems, manifold, pipeline, power and control umbilical, and so on), it can reduce development cost, enhance reservoir productivity, and improve subsea system reliability and operability. Over the period from 1970 to 2000, millions of dollars have been spent to develop subsea separation and pumping systems. But because of unresolved technical issues, along with a lack of confidence and clear understanding of the costs and benefits, industry has not rushed to deploy the technology on a commercial basis. However, as the industry moves into remote deep and ultradeep water, various degrees of subsea processing are becoming more common. In deep water, the technology can enable hydrocarbon recovery from small reservoirs that are subeconomic by conventional means, making small fields economically viable and large fields even more profitable. Subsea processing refers to the separation of produced ...
Artificial lift is a method used to lower the producing bottomhole pressure (BHP) on the formation to obtain a higher production rate from the well. This can be done with a positive-displacement downhole pump, such as a beam pump or a progressive cavity pump (PCP), to lower the flowing pressure at the pump intake. It also can be done with a downhole centrifugal pump, which could be a part of an electrical submersible pump (ESP) system. A lower bottomhole flowing pressure and higher flow rate can be achieved with gas lift in which the density of the fluid in the tubing is lowered and expanding gas helps to lift the fluids. Artificial lift can be used to generate flow from a well in which no flow is occurring or used to increase the flow from a well to produce at a higher rate.
Jones, Drew (Deep Imaging) | Pieprzica, Chester (Apache Corporation) | Vasquez, Oscar (Deep Imaging) | Oberle, Justin (Deep Imaging) | Morton, Peter (Deep Imaging) | Trevino, Santiago (Deep Imaging) | Hickey, Mark (Deep Imaging)
We used a new, large-scale, surface-based, controlled-source electromagnetics (CSEM) approach to map the locations of frac fluid during flowback following a three-well hydraulic fracture stimulation in the Permian Basin. CSEM records and analyzes electric field signals induced in the electrically conductive frac fluids by a surface-based transmitter. For this study, we placed a grounded dipole transmitter directly above the central horizontal well of three parallel neighboring wells. The transmitted signal was a broadband pseudo-random binary sequence. To record the frac fluid response signal, we placed an array of 161 receivers on the surface covering the three horizontal wells. We recorded the induced, response signals of the flowback fluids in three-hour intervals (three on, three off) for 228 hours. The CSEM recording started eleven days after flowback began on the central well and four days after flowback began in the two outer wells. From this time-lapse recording we captured the spatial and temporal change in electrical conductivity within the fractured reservoir, allowing us to infer the location of flowback fluid and its movement. During the stimulations chemical tracers had been included in the frac fluid. Analysis of the tracers captured during flowback agreed well with the mapped fluid locations and movement found in the CSEM data.
Flowback monitoring and its interpretation offer another valuable tool for frac and reservoir engineers. This understanding is especially critical in developing and managing unconventional reservoirs. Here, the stimulation responses are not simple, more and more evidence show complex fracturing and complex fracture networks (e.g., Rassenfoss, 2018). Characterizing a fracture network or networks in shale (i.e., an unconventional reservoir) is a challenging task. It is complicated by multiphase and complex flow regimes, non-static permeability and porosity, natural fracture and flow systems, heterogeneities and complex stress, changing stress with production, liquid loading, and a host of operational concerns (Zolfaghari et al., 2016). In the past, to determine hydraulic fracture properties, operators used production data in a variety of models to manage wells and reservoirs. Garnering production data can take months or even years delaying, for example, upgrades to well and stimulation designs and designing infill drilling (Williams-Kovacs, Clarkson, & Zanganeh, 2015). In contrast, a flowback occurs during the transition between stimulation and bringing the well online. Understanding the flowback provides significant improvements in determining early production rates enabling estimates of the effective size of stimulations, distinguishing key reservoir properties, and predicting long-term production rates (Jacobs, 2016). In addition, there can be direct savings if, for example, flowback interpretation identifies an underproducing play in time to redirect funds into a more lucrative play before infill drilling (Williams-Kovacs et al., 2015).
Hydraulic fracturing stimulation designs are moving towards tighter spaced clusters, longer stage length, and more proppant volumes. However, effectively evaluating the hydraulic fracturing stimulation efficiency remains a challenge. Distributed fiber optic sensing, which includes Distributed Acoustic Sensing (DAS) and Distributed Temperature Sensing (DTS), can continuously monitor the hydraulic fracturing stimulation downhole and be compared with other monitoring technology such as microseismic. The DAS and DTS data, when integrated with the microseismic, highlight processes relevant to the completion design and allow for a better understanding and interpretation of each dataset.
This paper outlines a workflow to improve processing and interpretation of DAS and DTS data. In addition, an estimate of the slurry distribution can be made. These methods will be demonstrated for a horizontal Wolfcamp well in the Permian Basin. Here we compare key aspects of the microseismic, DAS, and DTS results in several fracture stages to understand the downhole geomechanical processes. In order to interpret the DTS data a thermal model is developed (using DTS data) to simulate the temperature behavior after pumping has ceased. A slurry distribution is obtained by matching the simulated temperature with the measured temperature from DTS. In addition, the DAS data signal is studied in the frequency domain and the dominant frequencies are identified that are mostly related to fluid flow and to reduce the background noise. This time frequency analysis enhances the ability to monitor and optimize well treatments.
After reducing the background noise, the acoustic intensity is correlated to the slurry distribution. The fluid distribution data from DAS and DTS are compared with the microseismic and near field strain to better understand the completion processes. We utilized fiber optic microseismic to better understand and compare it to conventional microseismic.
Finally, we highlight the dynamics of strain and microseismic signature as fluid moves from an offset well completion into the prior stimulated fiber well to better understand the reservoir and far field effects of the completion.
This work presents a novel methodology to predict multiphase flowrates using early-time flowback data (flowrates and measured bottomhole pressures) from wells in unconventional reservoirs. The methodology is entirely data-driven using the total liquid productivity index [JTL(t)] as a function of time, both diagnostically for the identification of flow regimes as well as fitting the JTL(t) data trend with a "hyperbolic" decline function, which is later used as a forecast relation. The second major component of this methodology is the use of measured flowing bottomhole pressure data [pwf(t)], which is also correlated, or fit, with a "hyperbolic" decline function for forecasting purposes. The JTL(t) model is extrapolated in time, as is the independent pwf(t) model, and both models are then combined to forecast the total liquid flowrate. The individual phase flowrates are then predicted from the total liquid flowrate using the average water cut and average GOR observed during early-time flowback.
This work is not bound by any physical constraints, and the "physics" of the process are implicit in the use of the independent hyperbolic decline models used for the JTL(t) and pwf(t) functions. Other, perhaps more rigorous models could be used, but such applications are beyond the scope of the present work. To calibrate the proposed JTL(t) and pwf(t) hyperbolic models, we initially focus solely on the use of early-time flowback data, but we also demonstrate that continuous updating of the calibrations with 30, 60, 90, and 120 days of production history provides a more robust prediction of long-term performance. The most significant, or limiting, assumption is the character and nature of the flowing bottomhole pressure behavior [pwf-model(t)], of which the model implicitly assumes the production operations remain constant (i.e., a constant choke and no shut-ins). This limitation is generally less important in practice, however, cases with longer term shut-ins are problematic, particularly when such shut-ins occur at early-times or during flowback.
Forecasting production performance for horizontal wells with multiple fractures in low to ultra-low permeability unconventional reservoirs is challenging due to the long duration of transient flow regimes. The widely-used Arps decline curve relationships were proposed for conventional reservoirs that quickly reach pseudosteady-state (or more appropriately, boundary-dominated flow). In this work we present conceptual decline curve models which conform to the long-term transient flow behavior observed in unconventional reservoirs. Validation is shown by matching production rate, loss ratios, and the b-factors.
Historically, decline curve models were derived based on observations of production, loss ratios, and b-factors. Specifically, we observed quadratic changes in loss ratio and power law changes in inverse loss ratio as the primary motivation in deriving these decline curve models. One of our proposed models is an extension of the existing stretched exponential model where the primary difference is that this extension accounts for curvature in the inverse loss ratio plot while the stretched exponential model is represented by a straight line on the inverse loss ratio plot. Detailed descriptions are provided for each model.
Using production data from the Barnett Shale, the proposed models are compared to existing models using statistical methods (i.e., total sum-of-squared deviations for production, loss ratios, and b-factors). Results show that the b-factor varies in time, as would generally be expected for transient flow. As a result, the constant b-factor assumption put forth by Arps for the traditional hyperbolic decline curve fails to properly match (and forecast) the production observed in low to ultra-low permeability reservoirs.
In contrast, our results show that the power-law based models conformvery well to production performance trends observed for unconventional gas wells. These power-law models include the stretched exponential and the proposed multi-segment or variable extension of the stretched exponential model. Through this work, it was found that an empirical understanding of the loss ratio and b-factor are essential in creating a robust decline curve. Any mischaracterization of the loss ratio behavior can result in inadequate matches of production, loss ratios, and b-factor; and result in erroneous production forecasts and inaccurate reserve estimates.
The novelty of the proposed decline curve models is in the simultaneous understanding of the production, loss ratio, and b-factor relationships for wells in unconventional reservoirs and the corresponding impact on reserve analysis. Specifically, the variable power-law model for the b-factor provides a unique level of flexibility in fitting production data.
This paper presents a new type curve that can be used to estimate reservoir and completion properties and forecast future production from multi-fractured wells in unconventional, low-permeability reservoirs. In addition, we present a methodology to construct statistical typical well production profiles (type wells) over a complete range of probabilities, including P90, P50, P10 type wells. The type curve is a modification of the familiar Fetkovich type curve, with the transient stems having Arps b-factors from 1 to 4, and the boundary-dominated flow (BDF) stems with b-factors varying from 0 to 1. This new type curve allows us to analyze transient data other than linear flow (b = 2) and to determine the b-factor for BDF based on log-log rate-time plots.
We constructed the type curve using the same definitions of dimensionless time and rate variables as in the familiar Wattenbarger type curve for transient linear flow followed by BDF. However, we included many transient and BDF stems and transformed and generalized that type curve into a form similar to the Fetkovich type curve. From a type curve match, we can estimate apparent or effective matrix permeability and an apparent average fracture half-length. Given these parameters and other more readily available parameters (e.g., lateral length, number of fracture stages) we can then scale a well's production profile to a chosen set of reference conditions. Based on the shape of the production profile (dominated by transient and BDF b-factors), we can place all the wells from a given data set into a small number of individual “bins.” We can then construct type wells for each bin. To forecast future production, we can select appropriate design lateral lengths, fracture half-lengths, stage spacing, and mean or other permeabilities determined from analysis of field data.
We validated our new type curve with analysis of simulated well performance. We then analyzed a set of production data from the Permian Basin and found that all wells could be placed into a single bin, and type wells could be constructed. We observed two practical problems: (1) a majority of the wells were still in transient flow, which means that a unique fit on the type curve is not possible, and (2) our scaling procedure fails when fracture stage spacing is unknown, as it often is with publicly available data. However, for wells still in transient flow, auxiliary plots allow us to determine pairs of fracture length and matrix permeability values consistent with observed data.
The new type curve allows us to estimate apparent matrix permeability and fracture length. Given these reservoir and completion properties, we can scale wells whose production profiles have common shapes to a common set of reference conditions and thus generate typical well production profiles. We can then rescale from the reference conditions to selected design conditions and forecast performance of undrilled wells.