Today many thin layer quantitative description methods are usually based on the Widess model. Seismic attributes can be used to identify isolated thin layers with the formation thickness of less than λ/4. However, what is usually encountered in actual work is a set of thin interlayer, the thickness of a single thin layer is often much less than λ/4, and it is difficult to resolve each thin layer using existing seismic methods. In this paper, we propose a new method based on support vector machines (SVM) algorithm to predict thin interlayer net thickness. Firstly, the paper proposes the tuning law of thin interlayer and the concept of apparent net-to-gross ratio. Then a detuning factor based on SVM algorithm is constructed, and the thin interlayer seismic response is converted into apparent net-to-gross ratio after detuning. Finally, the net thickness is calculated by multiplying the apparent net-to-gross ratio and apparent thickness. What’s more, 90°phase wavelet seismic data with the meaning of formation impedance is used to interpret thin interbedded reservoirs, and a detuning factor based on 90° phase wavelet is constructed. The 90° phase wavelet seismic amplitude can be more accurately translated into apparent net-to-gross ratio. The method is successfully tested by model data and applied in the Bohai Oilfield, the actual data application shows net thickness prediction results of the thin interbedded layers are agree well with the drilled results. It illustrates that the proposed method is feasible and effective, has certain industrial application value.
Note: This paper was accepted into the Technical Program but was not presented at the 2020 SEG Annual Meeting.
