Real-Time Well-Integrity Monitoring Using Fiber-Optic Distributed Acoustic Sensing

Well integrity, defined according to ISO/TS 16530–2 (2014) as the containment of fluids (i.e., liquids or gases) and prevention of their escape to subterranean formations or to surface, relies on adequate cementation of the casing strings. For wells drilled in high–temperature/high–pressure regimes (e.g., high–temperature geothermal wells), large cyclic–load changes occur during drilling, well testing, fluid production or injection, and shut–in periods. Cyclic–load changes are known to cause a high risk of well failure (Goodwin and Crook 1992; Thiercelin et al. 1998; Southon 2005). In addition, recent studies indicate that the ability of the cement sheath to provide a mechanical seal against fluid migration deteriorates over time (Jackson 2014; Davies et al. 2014). To implement a modern cement–state downhole–monitoring system in the sustainable operation of subsurface reservoirs for production or injection of hydrocarbons and geothermal fluids, a passive real–time monitoring system with the incorporation of optical fibers will be discussed in this study.

Introduced by Grosmangin et al. (1961) as CBL, various implementations of sonic and ultrasonic source–and–receiver geometries are currently in use to evaluate the properties of the cemented annulus behind casing. Data acquisition and analysis rely on the signal propagation between source and receiver and its interactions with the wellbore fluid, casing, cement, and formation. The signals are influenced by (among others) the shear modulus between casing and cement, the volume of cement behind casing and its mechanical properties, the impedance contrast between casing and cement, and tool centralization (Pardue and Morris 1963; Jutten et al. 1989; Song et al. 2013).

In addition to conventional electronic sensors, fiber–optic–based sensing systems have been increasingly used in downhole applications. With the arrival of DTS technology (Hartog and Leach 1985), distributed sensing systems were increasingly used for geothermal studies (Hurtig et al. 1993; Förster et al. 1997). The optical fiber, embedded in a cable structure, acts as the sensing element. Interrogated with a read–out unit at the surface, temperature data can be retrieved with high temporal and spatial resolution along the entire length of the fiber. The fact that only the measurement cable is subject to potentially hostile downhole conditions makes such fiber–optic systems suitable for operation in high–temperature environments (Reinsch and Henninges 2010; Reinsch et al. 2013). In addition to DTS, techniques to detect and localize dynamic strain changes along an optical fiber have drawn attention in recent years. This technology is often referred to as DAS. DAS has demonstrated its applicability to detect dynamic strain changes in a number of applications, such as vertical seismic profiling (Daley et al. 2013; Götz et al. 2018), flow profiling (Bukhamsin and Horne 2016), near–surface geophysics (Dou et al. 2017), and seismology (Jousset et al. 2018). Measured strain changes are influenced by the coupling between the optical fiber and the medium in which the cable is installed (Daley et al. 2013; Reinsch et al. 2017).

Both technologies (DTS and DAS) typically use an incident laser pulse and analyze backscattered light from individual points along the optical fiber. DTS systems mostly use the temperature–sensitive Raman scattering (Dakin et al. 1985). DAS systems are often dependent on the detection of phase changes of Rayleigh backscattered light (Hartog 2017). A change in the phase difference of light (between individual laser pulses) scattered by two separate points along the fiber is linearly proportional to a change in fiber length separating the points (Masoudi et al. 2013). Determining these phase changes for consecutive fiber intervals and over time can therefore be used to record the dynamic strain evolution (strain rate) of a fiber–optic cable induced by a seismic signal.

To determine the dynamic phase difference, a number of optical architectures have been proposed and used; Liu et al. (2016) provide a summary. For the presented study, the DAS interrogator unit sends short, coherent light pulses into the optical fiber. The phase difference is determined by splitting the backscattered signal and feeding one part through an extra length of fiber (imbalanced Mach–Zehnder interferometer). The two signals are brought to interference. The measured intensity is therefore proportional to the phase difference between points separated by the length of the delay fiber, generally referred to as the gauge length (Masoudi et al. 2013). For seismic applications, the gauge length is on the order of 10 m (Jousset et al. 2018). Applying a dynamic force on a section of the fiber results in local length changes (strain ε), and therefore the measured phase difference. Comparing the phase differences between consecutive laser pulses provides, for a single DAS channel, an estimate of the dynamic strain evolution with time (⁠$∂ε/∂t$⁠) of that fiber section. Modern DAS interrogator units are capable of extracting distributed acoustic data over several kilometers of fiber with submeter channel spacing (Parker et al. 2014).

In this study, we investigate the application of the DAS technology along a fiber–optic cable permanently installed behind the anchor casing of a high–temperature geothermal well. Fiber–optic acoustic data acquired during drilling and testing the reservoir interval together with fiber–optic DTS data will be used to qualitatively evaluate the cement condition behind casing.

The fiber–optic cable was installed behind the 355–m 18⅝–in. anchor casing of Well RN–34 located in the Reykjanes Geothermal Field in Iceland (Fig. 1). All depth specifications in this paper will be given in measured depth with reference to the ground level (GLV); the cellar reference level is set at 3 m below the GLV. The installed cable contains two single–mode fibers (SM1, SM2) and one multimode (MM) fiber in a gel–filled metal tube.

As the casing was run in hole, the cable was fixed to it using metal straps. During installation, the cable was damaged at a depth of approximately 175 m.

Recording data from the MM and SM2 fibers was impossible below this point. The SM1 fiber also was damaged, but data acquisition was still possible to a distance of 217 m along the fiber. We suspect that the cable was sheared off the casing below the initial damage. Data recorded below 175 m are therefore discarded in the following.

During the cementation operation of the 18⅝–in. anchor casing and 21–in. hole, an approximately 61–hour–long DTS measurement was performed along the undamaged part of the MM fiber. The DTS measurement was started at 7:30 p.m. Pumping of the cement slurry started at approximately noon. The density of the cement slurry was given as 1.7 g/cm³. Primary cementation was performed using a stinger. This was followed by a remedial top–up job. Cementation, including top–up jobs, was finished at approximately 10 a.m. on the following day. After cementation and a wait–on–cement period of approximately 25 hours, a CBL was run. The CBL tool was run in the 18⅝–in. anchor casing without surface pressure. After completion of the CBL, operations were suspended for another 22 hours. DTS measurements continued for an additional 48 hours after completion of the cementing operations.

Subsequently, drilling began for the production casing down to 1019 m. The directional well trajectory was planned with a kickoff depth at 375 m, build rate of 1°/10 m, and an inclination of 17°. Once the production section was drilled, the 13⅜–in. production casing was run to the planned depth and cemented. The next 12¼–in. section was drilled to reach the reservoir interval for the installation of a 9⅝–in. perforated production liner. During drilling and testing operations of the 12¼–in. section, DAS data were continuously recorded for 9 days. The final depth of Well RN–34 was 2686 m.

The DTS data were recorded with a DTS Ultra system from Schlumberger. Data were acquired with a 0.5–m spatial resolution and a repetition rate of 37 seconds.

The DAS data were recorded along the second SM fiber with an iDAS unit by Silixa Limited. Data recording started at 1:37 p.m. on 17 March 2015 and lasted until 10:15 a.m. (UTC time) on 25 March 2015. Data were recorded with a sampling frequency of 1000 Hz, a 250–Hz low–pass filter, raw channel spacing of 25 cm, and a gauge length of 10 m, and were stored in 1–minute–long files. Before data analysis, noise–adaptive, spatial downsampling was performed, resulting in 1–m channel spacing.

To relate individual fiber–optic channels along the optical cable to the physical position within the well, the GLV was determined by tapping the cable for the DAS data. Silixa Limited estimates that the maximum error of this localization procedure is on the order of 2 m. For the DTS data, cold spray was used at the same location. Within the well, the cable is assumed to be straight down to 175 m.

We start our analysis by investigating the temperature evolution during cementation of the anchor casing registered along the MM fiber (Fig. 2a). The DTS data reveal an interval between 66– and 106–m depth (red dashed lines) where the temperature evolution, as measured by the fiber–optic cable, differs from that of the rest of the wellbore.

Before pumping of the cement slurry, approximately 7 hours after starting the DTS measurement, a temperature step is observed at 66–m depth. Constant temperature of approximately 10°C is measured from 0 to 66 m, and, below, temperatures of approximately 15°C are recorded. Below 66–m depth, a thermal gradient is observed. Temperature gradually increases from 15 to approximately 20°C at 175–m depth.

A temperature increase caused by the arrival of the cement slurry is observed during primary cementation. The temperature signal rises up to the bottom of the aforementioned anomalous interval. Above 106–m depth, the signal starts to fade, and vanishes at 100–m depth. Between 91– and 93–m depth, an isolated, lower–magnitude temperature increase is again observed. Above, again no temperature signal from the cement slurry is observed, up to 66–m depth. Above 66–m depth, the temperature evolution begins as expected as the cement slurry rises close to the surface.

Cold water was then pumped through the annulus (at ≈12 hours), as indicated by a sudden temperature decrease along the fiber–optic cable. Again, at 66–m depth, the temperature as measured by the fiber–optic cable becomes discontinuous. Temperature abruptly rises from approximately 10 to 15°C. Lower temperatures are again recorded between 82– and 92–m depth, similar to the reappearance of the temperature signal during the pumping of the cement slurry.

Because the initially calculated cement volume was not sufficient for the slurry to reach the surface, two remedial top–up jobs were performed using 9 and 1 m³ of cement slurry, respectively. This is seen in the data as a small temperature increase close to the surface at approximately 13 hours after the start of the measurement.

Approximately 20 hours after the start of the measurement, we observe a temperature increase along the fiber–optic cable caused by the exothermic hydration reaction of the cement slurry. The release of latent heat of hydration during cement setting led to a temperature increase along the entire well. Between 82– and 92–m depth, the released heat caused by hydration is again higher than above and below. Maximum heat release is seen approximately 18 hours after finishing the primary cement job.

During drilling for the surface–casing section, major circulation–loss zones were identified in 41– and 85.5–m depth. During drilling operations for the anchor–casing section, no significant circulation losses occurred.

Fig. 2b shows the results of a caliper log run before installation of the anchor casing. The shoe of the surface casing is marked by the black dashed line. The loss of the temperature signal is preceded by a zone with considerable increases in hole diameter.

Figs. 2c and 2d show the result of the conventional CBL performed within the anchor casing of Well RN–34. The 3– and 5–ft amplitudes are generally high. Low pipe amplitudes are recorded only in the topmost part of the casing, as well as a narrow interval at approximately 100–m depth. Considerable casing arrivals are recorded in the variable-density log (5–ft spacing) over most of the depth profile. For clarity, the travel–time curve has been shifted by 100 µs; the travel–time curve shows some cycle skipping, predominantly above 140–m depth. On–site, drilling engineers concluded that a continuous cement column could be found behind the casing string. It was suspected that the cement had not properly set yet.

Fig. 3  illustrates examples of signals recorded during the DAS survey, performed 1.5 months after installation of the fiber–optic cable. Plotted in Fig. 3 is every fourth channel of the data record along the 18⅝–in. anchor–casing section. The data series for both examples have been detrended and integrated with respect to time, to convert them from strain rate to strain. A 10– to 50–Hz Butterworth band–pass filter has been applied. Data processing made use of the Python toolbox ObsPy (Beyreuther et. al. 2010). Data plotted in Fig. 3a were recorded during injection testing of the reservoir. The flow rate was 113 L/s. Data shown in Fig. 3b was recorded after finishing the injection test with the drillstring still out of hole.

The signals in both examples differ distinctively by their propagation velocity. Each signal was analyzed by picking the time and value of maximum amplitudes in a predefined moving window in space and time. From the picks, a linear regression was calculated to determine the average velocity. The signals in Fig. 3a travel with a speed of approximately 1348 m/s and appear to be periodic in time. The signal observed in Fig. 3b travels at a significantly higher speed, reaching approximately 2342 m/s, and exhibits no periodicity.

Fig. 4  shows the picked maximum amplitudes vs. the depth of the aforementioned signals. The picking of the amplitude has been performed with the full data record, increasing the spatial density. The amplitude of the slow–propagating signal (V = 1348 m/s, shown in Fig. 3a) shows an unsteady behavior with sudden magnitude changes (e.g., between 95– and 110–m depth). In contrast to the unsteady behavior of the slow–propagating signal, the amplitude of the fast–propagating signal (V = 2342 m/s, shown in Fig. 3b) decays quadratically with depth. In addition, the theoretical amplitude decay for a body wave as a result of spherical divergence is indicated by the blue dotted line in Fig. 4. Overall, the amplitude decay of the fast–traveling signal closely follows the theoretical amplitude–decay prediction. Between approximately 30- and 50-m depth, higher amplitudes are observed. This interval coincides with a depth interval showing higher amplitudes for the slow–traveling signal.

To estimate the relative average–strain amplitude transferred to the fiber–optic cable, we calculate the root–mean–square (RMS) values for each channel in 1–minute–long data files using

where x_i is the strain–rate amplitude of the ith sample in an n–sample–long time series.

Figs. 5a and 5b and 6a and 6b  each show 20 minutes of the aforementioned RMS values computed for three distinct activities during the survey: drilling and 20– and 113–L/s fluid injection. For all three cases, computation has been performed in two different time windows (Time Windows 1 and 2). Table 1  summarizes the averaged rig–log data for the corresponding time windows.

RMS maps in the figure rows denoted by (a) have been computed from raw data. For RMS–maps denoted by (b), the raw data have been band–pass filtered between 30 and 60 Hz before computing the RMS values. The RMS maps have been normalized to a global maximum.

Injection testing was performed for approximately 12 hours. Flow rates were alternated between 20 and 113 L/s, each with a duration of approximately 45 minutes. Fluid injection was performed through the production casing, with the drillstring out of hole.

For the drilling cases, Time Window 1 starts at 18 March 2015 at 11:40:07 a.m. Time Window 2 starts at 24 March 2015 at 10:20:07 p.m., approximately 6 days later. For the 20–L/s–injection case, Time Window 1 starts at 21 March 2015 at 02:00:07 a.m. Time Window 2 starts at 21 March 2015 at 07:30:07 a.m., therefore separated by 5.5 hours. Likewise, for the 113–L/s–injection case, Time Window 1 starts at 21 March 2015 at 01:00:07 a.m. Time Window 2 starts at 21 March 2015 at 08:20:07 a.m., a separation of 7 hours and 20 minutes.

Starting times for the time windows have been selected arbitrarily. The shorter time separation for the injection scenarios is because of the limited duration of the injection test. During fluid injection, a very high vibrational load was observed close to the wellhead, resulting in high RMS amplitudes for the first 25 m of the profile. To be able to compare the data measured during the different operations, the first 25 m of the RMS profile had to be excluded from the analysis.

The RMS maps between the two time windows are generally similar and constant in time, except between the raw 20–L/s–injection cases. For Time Window 1, no structural pattern is apparent. For Time Window 2, the 20–L/s case is similar to the 113–L/s–injection case. We tested different band–pass–filter widows to investigate if the same pattern could be revealed in some frequency interval. The frequency range between 30 and 60 Hz ultimately proved to be the most successful to increase the similarity between the two different cases.

Table 2  summarizes the average discrepancy between all combinations of RMS maps from Figs. 5a and 5b and 6a and 6b, according to

where x_i and y_i are the individual pixel values of the RMS map and n × m are the dimensions of the image. In general, the discrepancy is reduced by the filtering operation. In particular for the 20–L/s–injection cases, the discrepancy reduces from 26.2 to 7.9%.

Figs. 5c and 6c show the amplitude spectra for the three cases from Figs. 5a and 6a. The frequency spectra have been computed from 10–second–long time sections, located in the middle of the corresponding RMS time window. The frequency spectra have been normalized for each individual depth channel. The three cases show distinct amplitude distributions. For the drilling case, the energy is distributed evenly over the whole frequency range, whereas distinct frequency bands are excited during fluid injection. Between the two time windows, the frequency spectra are again similar.

Fig. 7  illustrates the energy distribution for three channels located in 91– (1 m before the shoe of the surface casing), 110–, and 125–m depth for the three analyzed cases. The amplitude spectra between the two time windows are again similar.

In Figs. 5a and 5b and 6a and 6b, we observe a variable average–strain amplitude with depth during drilling for both time widows. For the shallow part (<90 m) in the unfiltered case and Time Window 1, we observe elevated energy at 50–, 70–, and 85–m depth. For Time Window 2, elevated amplitudes are recorded at 55–, 65–, and 75–m depth. Maximum amplitudes are recorded between 120– and 125–m depth and at approximately 135–m depth for both time windows.

For the band–pass–filtered data (Figs. 5b and 6b), amplitude variation with depth is generally higher than for the unfiltered case. Amplitudes above 50 m are attenuated by the filtering operation. Contrary to the unfiltered case, dominant amplitudes are recorded at 65– and 75–m depth for both time windows. For Time Window 1, significant (≈0.5) relative amplitudes are recorded at 45–m depth. Amplitudes between 120– and 125–m depth, as well as approximately 130–m depth, are reduced compared with the unfiltered case.

In general, the RMS profile is more or less constant in time. Relative amplitudes fluctuate between ≈0.6 and 1. Fluctuations in depth and time appear bigger for Time Window 1 compared with Time Window 2. The frequency spectra (Figs. 5c and 6c) show high variability with depth. Constant frequency content is recorded close to the surface (<50 m) and at approximately 100– and 160–m depth. This indicates that the energy content for different frequencies is dependent on depth.

The largest discrepancy in the RMS maps is observed between Time Windows 1 and 2 in the unfiltered 20–L/s–injection case (Figs. 5a and 6a). For Time Window 1, steady, low–variability amplitude distribution, dominated by individual peaks, is observed. Time Window 2 shows an amplitude distribution similar to that of the drilling case. Elevated amplitudes are recorded at approximately 45–m depth, between 120– and 125–m depth, and at approximately 130–m depth. The frequency spectra in Figs. 5c and 6c show elevated amplitudes in distinguished frequency bands between 1 and 10 Hz, as well as close to 50 Hz.

For the band–pass–filtered data (Figs. 5b and 6b), Time Window 2 shows little change compared with the unfiltered case. Relative amplitudes at approximately 45–m depth reduce from approximately 0.9 to 0.7. The resolution is increased in that section. For Time Window 1, a significant change is observed. The amplitude distribution becomes similar compared with Time Window 2. Elevated amplitudes appear at 45–, 125–, and 130–m depth. Similarity to the drilling case is increased as well, except between 60– and 75–m depth. Here, both time windows show a low–amplitude section, whereas maximum amplitudes are recorded for both drilling cases in both time windows. Again, elevated–amplitude sections are constant in time, with Time Window 2 showing less variability.

For the 113–L/s–injection case, variability is low between both time windows. Dominant amplitudes are recorded at the same depths as in of the drilling and 20–L/s–injection cases (125– and 130–m depth). The frequency spectra are dominated by a strong signal at approximately 60 Hz (Figs. 5c and 6c). Because the spectra are dominated by frequencies at 60 Hz, the band–pass–filtered data between 30 and 60 Hz do not differ much from the unfiltered data.

For all three operations, the average–strain amplitudes calculated from 1–minute data remained mostly constant over the 20 minutes that were analyzed. Time Window 2 shows greater stability with time as well as depth. When comparing the average–strain amplitudes for the three different cases, the main difference was found in the depth interval between 65 and 85 m, where higher amplitudes (>0.7) were observed for the drilling operation compared with lower values (<0.5) for both injection cases.

Fig. 8a  compares average–strain amplitudes computed from two different time windows during drilling and testing the reservoir interval with conventional CBL data from Fig. 2c. A 20–minute normalized temporal mean from the 1–minute RMS values in the 30– to 60–Hz frequency range from Figs. 5b and 6b was calculated for all three cases. For the CBL data shown in Fig. 2c, the normalized amplitude was calculated for the interval between 25– and 175–m depth. CBL data were recorded with a spatial resolution of 0.1 m. To make this comparable with the 1–m–spaced DAS records, the CBL amplitudes have been smoothed with a sliding average over 11 samples, centered at Sample 6. Fig. 8b shows the difference between the 3–ft CBL and mean RMS amplitude from Fig. 8a. For the difference computation, the 3–ft CBL data have been resampled to a spatial sampling of 1 m using the Fourier method.

For both time windows, when comparing the averaged strain amplitudes measured during the drilling phase with normalized CBL amplitudes, overall similar trends with depth can be observed along the fiber–optic cable and in the CBL data, although not all peaks match in both data sets. Time Window 2 shows a larger offset at shallow depth. An additional peak is observed at 67–m depth. In the depth interval from 75 to 85 m, the amplitude decrement along depth is stronger for the noise data from the drilling operation compared with the CBL amplitudes. Furthermore, the relative amplitudes from 85 to 100 m are lower for the average–strain data. The maximum relative amplitudes during drilling were detected at the same depth as the maximum values for the CBL data, at approximately 75 and 125 m.

The normalized average–strain amplitudes measured during injection of 20 L/s again match the same trends as the average–strain amplitudes generated during the drilling phase, although individual peaks have a different amplitude and slope with depth. A significantly different trend is again observed in the depth interval between 65 and 85 m, where average–strain amplitudes remain low. Although the maximum average–strain amplitudes during drilling are detected at the same depth as the maximum amplitudes for the CBL, at approximately 75 m, the maxima during fluid injection are observed at approximately 125–m depth. The data series remain similar between time windows.

For an increased injection rate, similar normalized average–strain amplitudes were recorded with a higher amplitude interval close to 125–m depth and a low amplitude interval between 66– and 100–m depth. Overall, the normalized average amplitudes for the high injection rate are slightly lower compared with the low injection rate. This indicates a higher peak value. This is more prominent in Time Window 1.

In an attempt to quantify the similarity between the trends observed between the normalized CBL amplitudes and the normalized average–strain amplitudes, a discrepancy was calculated using Eq. 2, with m = 1; x_i,y_i are the individual data samples from the CBL and the DAS data, respectively; and n is the number of data points with depth. The resulting data are displayed in Tables 3 and 4.Table 3 shows the average discrepancy for the whole depth profile, which results in 15 to 20%. Table 4 shows the discrepancy for the interval below 100–m depth. For the interval below 100 m, the error is on the order of 10 to 15%. Overall, data acquired during the drilling operation show the highest similarity over the entire analyzed interval and below 100 m. Major differences are observed between 65– and 100–m depth, close to the shoe of the surface casing (92 m).

It is noteworthy that the difference between CBL amplitudes and the average RMS value against depth is again similar for the two different time windows.

Similar trends have been observed for the normalized average–strain amplitudes during drilling and injection testing compared with the normalized CBL amplitudes. The largest differences were observed in the interval between 66– and 106–m depth. This interval coincides with the depth interval where a different thermal signature along the optical cable was observed during the cementation of the anchor casing. We will first address the DTS temperature information from the cementation and then address the DAS data.

Increasing temperatures measured along the optical cable indicate the presence of cement within the annulus behind the 18⅝–in. anchor casing. Latent heat released during the hydration of the cement led to an overall increase of the temperature within the well and in the vicinity of the wellbore. Quickly increasing temperatures indicate a direct contact between cable and cement; slowly increasing temperatures also indicate the presence of cement at depth. Possible reasons are that the amount of cement at that depth is small, that there is little cement in the immediate vicinity of the cable, that the annulus at the position of the cable is thin, or a combination of any of these aspects. A clear distinction cannot be made using the available information.

The 18⅝–in. anchor casing was cemented in two stages. For the primary cementation, the flow of warm cement slurry in the vicinity of the cable can be observed. The positive temperature variation rises from the bottom to the top, indicating the continuous placement of cement in the annulus. In the interval from 66 to 106 m, the temperature variation caused by flowing cement is smaller, and is not observed at all between 66 and 85 m. The bottom of this interval also corresponds to the depth of increased borehole breakouts below the shoe of the surface casing at 92–m depth. At shallower depth (<66 m), the positive temperature variation is again observed, indicating the presence of flowing cement around the cable.

Before remedial cementation, cold water was injected into the annulus. Although we see a strong temperature variation because of the presence of cold fluid flowing from the surface down to 66 m, only a smaller temperature variation was observed at greater depths. In addition, no temperature variation was observed at depths greater than 105 m. After cold–fluid injection, a top–up cementation was performed until the cement reached the surface. A small temperature signature can be observed at depth shallower than 66 m. Again, the displacement from the top to the bottom can be confirmed by the temperature signature.

After cementation, temperature increased along the whole depth of the well because of the latent heat of cement during hydration. This indicates that cement is present in all depths. The magnitude of the temperature increase can typically be seen as a measure of the amount of cement per depth interval. It is typically higher in the intervals of larger annular diameter compared with intervals with a narrower annulus.

Overall, the temperature evolution analyzed during primary and secondary cementation indicated that there seem to be less cement in the immediate vicinity of the cable for the interval from 66 to 106 m. In addition, elevated CBL amplitudes in this interval indicate less cement bonding.

The two exemplary signal types with different propagation velocities and their relative recorded amplitudes in Figs. 3 and 4 were recorded during and just after injection testing. The measured DAS amplitude shown in Fig. 3a is therefore likely caused by tube waves resulting from fluid injection into the wellbore. The speed of tube waves is related to mechanical fluid, well–completion, and formation properties. Their speed is typically slower than the speed of sound in water [Norris (1989) provides exemplary calculations], matching our observations. Because tube waves are guided by the well structure, no geometric attenuation is observed along the fiber–optic cable.

Because of the faster propagation velocity and quadratically decreasing amplitude, the signal from Fig. 3b is likely a compressional body wave resulting from some effect on the drillsite. Although the source and its position are unknown, the amplitude decay with depth indicates a source close to the well. The signal energy is therefore transmitted across the formation and through the cement sheet surrounding the fiber–optic cable.

The two waves mentioned previously reach the fiber–optic cable from different directions (inside and outside), but show a similar trend at close to 50–m depth. This indicates that local strain–amplitude variations are influenced by the coupling to and condition of the material surrounding the fiber–optic cable (i.e., casing ↔ cement ↔ cable and formation ↔ cement ↔ cable).

The average–strain amplitudes calculated for the three operations in different time windows show similar trends with depth, apart from the previously discussed interval between 66– and 106–m depth. It has to be stated that the behavior of the 20–L/s–injection case in Time Window 1 is not observed in other 20–L/s–flow–rate intervals during the injection test. Here, an additional noise source appears to be present, which cannot be identified from the rig–log data. When filtering to appropriate frequency bands, the discrepancy to Time Window 2 reduces to 7.9%. Assuming constant cable properties, locally changing properties of the medium surrounding the cable remain as the only explanation for the change in the average–axial–strain amplitudes recorded along the fiber–optic cable. Hence, the coupling between and mechanical properties of casing, cement, and formation appear to determine the magnitude of the measured strain amplitudes. In the presented case, the signal source is inside the well. Better coupling leads to more–effective energy transmission from inside the well to the formation, and hence measured average–signal amplitudes are lower. The lower the bonding, the more energy trapped inside the well, and the higher the measured amplitudes along the fiber–optic cable. Therefore, the recorded signal amplitude behind the casing can potentially be used to monitor changing mechanical properties of the well completion and hence the integrity of the cemented annulus.

To support the hypothesis that the average–axial–strain amplitude is a measure of the condition of the cemented annulus, the normalized average–axial–strain amplitude was compared with normalized amplitude values from a CBL. Overall, both amplitudes show similar trends, although the CBL was acquired during wait on cement for the anchor casing and the noise log was acquired 1 month later after installation and cementation of the production casing. When calculating the discrepancy between both amplitudes, we observed a high similarity between both sets of data, especially in the drilling cases. When excluding the interval where we observed the temperature anomaly during the cementation, discrepancy is reduced from up to 21 to less than 17% for the injection operations. For the strong noise source of the rotating drillstring during drilling, we observed a reduction from approximately 18 to less than 13%. The high similarity between both data sets indicates that the noise amplitude recorded along the fiber–optic cable is dependent on the material properties of the cemented annulus in a way similar to a conventional CBL tool. The similarity in the difference between CBL and average–strain amplitudes between the two time windows suggests that the measured average–strain amplitude is influenced by other components of the well construction (e.g., additional casing strings). Similar to a conventional CBL, the measured amplitude is likely influenced by different parameters of the well construction, such as the coupling strength between different casings and cement sheaths, the formation, or the compressional strength of the cement. In addition, the measured CBL amplitude, as mentioned previously, is influenced by a multitude of factors. Interpretation is therefore not always unambiguous. In particular, the transition from open to cased hole and the presence of additional casing strings—above 92–m depth, as in our case—increases the measurement uncertainty. The slurry density of 1.7 kg/cm³ is unsuitable to explain elevated CBL amplitudes along the investigated depth interval. Considerable washouts were recorded by the caliper tool below the shoe of the surface casing at 92–m depth, which might influence the CBL attenuation behavior in that section. However, a clear correlation between elevated CBL amplitudes and washout zones is not observed.

High CBL amplitudes in the depth interval from 66 to 85 m indicate a lower cement bonding for this depth section. This coincides with the top of the observed temperature anomaly during cementation. Varying average–strain amplitudes in this interval were observed during drilling and injection testing. In addition to the signal source, the wellbore pressure changed between both operations. Although the wellhead pressure during drilling was zero, 15– and 20–bar wellhead pressure was measured during injection testing with 20 and 113 L/s, respectively. The wellbore–pressure increase might have resulted in an improved coupling between casing, cement, and formation, and hence lower average axial amplitudes. The change in coupling as a result of pressure suggests nonoptimal cement conditions in this depth interval. CBL data acquired without additional wellhead pressure, on the other hand, show elevated amplitudes, similar to the average–axial–strain amplitudes during drilling.

From our analysis, we can draw the following conclusions:

1. DTS during cementation: The use of DTS data acquired during cementation of the anchor casing enabled us to study the cement placement. From the data, it was possible to identify an interval with a different thermal signature during cementation. Temperature data suggest nonoptimal cement placement. This interval also shows higher CBL amplitudes, which also point to nonoptimal cement bonding.

2. Real–time noise monitoring using DAS: We analyzed passively recorded continuous average–axial–strain amplitudes acquired during different well operations and can conclude that the recorded signal amplitudes correlate with the coupling properties of the well completion to the formation.

Further investigation is necessary to identify possible signal sources or increase the fiber/interrogator sensitivity. In our case, drilling and injection testing provided sufficient signal energy to reveal information about the coupling condition of the well completion. The energy recorded during fluid circulation through the drillstring or during shut–in periods was not sufficient to reveal the information about the coupling properties with the simple approach presented here. Although external signals as presented in Fig. 3 were not analyzed in greater detail, data suggest that sources from outside the wellbore can as well be used to analyze the coupling condition of the well completion.

Possible influencing factors (e.g., elastic parameters and additional cement sheaths and casing strings) should be investigated. A mathematical model should be developed, incorporating the effects of changing well architecture.

It is established that the optical properties of optical fibers deteriorate over time because of the reaction of silicon dioxide with free hydrogen. For long–term monitoring applications in harsh environments, this issue would need to be addressed using specially designed optical fibers. Current proposals to overcome this problem include special carbon coatings and advanced glass compositions.

DTS succeeded in observing the cement placement over the installation interval. This might offer a solution for investigation of low–density foam cements, which pose challenges for conventional cement–evaluation tools.

It has been shown that the average–axial–strain amplitude recorded along an optical fiber is closely related to the coupling of the well completion to the formation. Hence, signals traveling along or through the well completion are more attenuated, resulting in a lower average–axial–strain amplitude in the case of good coupling conditions. This might also have an effect on the amount of signal energy that is trapped inside a well. The better the coupling, the lower the resulting signal energy inside the well. If this conclusion holds true, an assessment of the coupling condition of the well completion could also be performed with DAS wireline or behind–tubing installations, opening the path to a wide implementation of this technology.

Given the ease of data acquisition, fiber–optic DAS measurements along a permanently installed fiber–optic cable behind casing are potentially suited to monitor the long–term coupling condition of the downhole completion (hence, the mechanical integrity of a wellbore). A clear limitation of the DAS technology compared with conventional logging techniques is the channel spacing and the lack of radial resolution in the current implementation. Emerging distributed–strain–sensing technologies might be able to overcome the spatial–sampling limitation of current DAS systems. The lack of radial resolution might be addressed by installing more than one fiber–optic cable or currently emerging cables that allow for the acquisition of three–component data. Distributed fiber–optic sensing systems, however, will likely not be able to match conventional sonic and ultrasonic well–logging tools in resolution and precision. They are candidate technologies, however, to monitor the mechanical properties of the cemented annulus in real time over the entire life cycle of a well. Real–time monitoring will thereby allow for analyzing load cycles during production or injection, optimizing costly maintenance operation and ultimately ensuring the longevity of downhole installations.

Data were acquired within the framework of project IMAGE (Integrated Methods for Advanced Geothermal Exploration), funded by the European Commission's Seventh Framework Programme under Grant No. 608553. This study has received funding from the EU's Horizon 2020 Framework Programme for Research and Innovation under Grant No. 654497 (GeoWell project), 676564 (EPOS IP), and 691728 (DESTRESS). We would like to thank our partners from the GeoWell project for the excellent collaboration, constant support during data acquisition and analysis, and fruitful discussions over the past years. We are especially grateful to Árni Ragnarsson, Ingólfur Örn þorbjörnsson, and Gunnar Skúlason Kaldal, and their colleagues from íSOR, as well as Guðmundur Ó. Friðleifsson and Ómar Sigurðsson and their colleagues from HS Orka. We would like to thank Andi Clarke and his colleagues from Silixa Limited for their efforts during data acquisition and analysis. At the GFZ German Research Centre for Geosciences, we would like to thank David Bruhn, Ernst Huenges, Philippe Jousset, Christian Cunow, Jörg Schrötter, and Ronny Giese, as well as all colleagues in the Geoenergy Section who contributed to this project.