Rational development of oil and gas reservoirs is possible only with efficient monitoring by various well logging techniques. This paper presents algorithms for processing data acquired by spectral noise logging (SNL) in memory mode. The SNL technology is designed to identify flowing reservoir intervals, cross-flows behind casing and tubing and casing leaks by spectral analysis of recorded noise signals.
While moving through a reservoir, fluids and gases create turbulence and rock vibrations that in turn generate noise. This acoustic noise is recorded with a noise logging memory tool consisting of a high-sensitivity piezoelectric hydrophone sensor and an amplifier and data collection module. The tool records acoustic signals in the frequency range of 15 Hz to 60 kHz. The existing SNL technology excludes intense broadband noise created by the movement of the tool in the well.
Useful information is extracted from background noise using a technique based on wavelet thresholding. Spectral noise density in the depth-frequency plane undergoes a wavelet transform. At each measurement depth, several tens of noise signals are recorded to determine mean wavelet coefficients and their typical variance. Then, they are analysed to remove statistically insignificant details from the signal spectrum and to suppress noise components that are present throughout large depth intervals.
The processing of data acquired in tens of wells from various fields has show that the noise features identified by wavelet filtering correlate with open-hole data and are confirmed by conventional well logging techniques.