Application of time-frequency entropy based on high-order synchrosqueezing transform in reservoir prediction

Abstract

Reservoir prediction is an essential component of seismic exploration. We have introduced a new technique for reservoir prediction based on time-frequency entropy that can detect seismic attenuation caused by reservoir fluid, such as gas and oil. The use of conventional methods of time-frequency analysis leads to low resolution due to inadequate calculation accuracy. To address this issue, we have adopted a novel method of calculating time-frequency entropy that computes the time-frequency distribution using a high-order synchrosqueezing transform to obtain time-frequency entropy with a higher resolution and better energy concentration. Numerical tests on synthetic signals show that the proposed method provides high accuracy and can detect ultrasmall variations (e.g., 1%) in amplitude, phase, or frequency. The model test shows that frequency entropy based on our proposed method can effectively distinguish an oil-bearing reservoir and is also robust to noise. Application on field data further demonstrates that it provides time-frequency entropy with high resolution and can identify weak seismic signal variations caused by dry, water-saturated, and oil-saturated sedimentary layers. Practical well drilling verifies these prediction results. The proposed method can be broadly applied in seismic reservoir prediction.