Fiber-optic distributed temperature sensors (DTSs) are used increasingly in the oil and gas industry to provide temperature logs in real time. Often, users need to know the temperature resolution—the noise level on the measured temperature trace. Temperature resolution can be difficult to specify because it depends on many factors, including averaging period, fiber length, excess attenuation from connectors and splices, and hydrogen darkening of the fiber. To overcome this difficulty, we developed a novel technique to enable the DTS to provide an estimated temperature-resolution trace to accompany each temperature trace. We modeled the noise generated by an optical receiver as constant background noise plus signal-dependent shot noise; parameters of this model were readily measured during system tests. Because the DTS measures the Stokes and anti-Stokes signal level at each point along the fiber, we were then able to calculate the corresponding noise at each point. We modified the subsequent data processing that calculates temperature from these signals to handle a Gaussian model (µ, s) instead of a single-valued signal. The resulting output provided an estimated temperature-error trace point-by-point along the fiber to accompany each temperature trace. Being derived from the optical signal levels rather than from statistics of the measured temperature, the estimate is not affected by the actual temperature varying with time or along the fiber. However, changes in signal level resulting from changing fiber loss immediately affect the error trace, giving a real-time indication of measurement quality. We have validated the technique against actual temperature noise measurements using a resolution test rig we designed for the purpose. We found good agreement between the estimated and measured temperature resolution, over a range of operating conditions. This new technique, by providing temperature error bounds based on conditions obtaining at the time of measurement, will enable the performance of a DTS to be continuously monitored and will give temperature analysts the information needed to determine the confidence levels of their temperature log interpretations.
The fiber-optic distributed temperature sensor (DTS) is increasingly being used in the oil and gas industry as a tool to understand flow behavior in the wellbore. Since the temperature changes caused by fluid flows are often small, users are concerned about the temperature resolution of the DTS or, equivalently, the noise level on the measured temperature trace. Consequently, there has been much recent interest in agreeing on a definition of the temperature resolution parameter and in laboratory measurement techniques (Hadley et al. 2005; Johnson 2007).
However, the temperature resolution achievable in practice depends on many unknown variables. These include such factors as losses in connectors and splices, variations in DTS laser output, and progressive hydrogen darkening of the fiber. Making worst-case assumptions can lead to expensive overspecifying of the DTS and other components for a given temperature resolution requirement.
A DTS works by measuring the levels of backscattered light at one or more wavelengths and computing from them the temperature at the point of scattering. When properly designed, the only significant source of temperature noise in the DTS measurement will be the uncertainty (noise) in the measured light levels. Consequently, if the uncertainty in the measured light level is known, we can, in principle, calculate the resulting temperature noise.
The uncertainty in the measured light level depends on the noise performance of the optical receiver, the stability of transmitted power of the laser pulse, and the number of laser pulses averaged. All these parameters are known, or measurable, enabling the uncertainty of the measured light level to be determined. From this, the temperature uncertainty at each point along the fiber can be determined and an estimated error output provided to the user.