关键词: 二叠系茅口组, 白云岩, 高保真, 高分辨率, 多次波, 共偏移距矢量片域, 压缩感知

Abstract:

The deep Permian Maokou Formation in the Penglai area, central Sichuan, is located in the Sichuan Basin. It is a sedimentary formation primarily composed of carbonate rocks, with dolomite being the main lithology. With good reservoir properties, it represents an important oil and gas reservoir in the Sichuan Basin. However, due to characteristics such as deep burial, thin reservoirs, strong heterogeneity, minimal impedance contrast with surrounding rocks, and weak reflection signals, seismic prediction becomes challenging. Conventional frequency enhancement processing either fails to identify thin reservoirs or results in a large area of coherent seismic events, making seismic prediction of thin reservoirs in this region difficult. Therefore, the application of “high-fidelity and high-resolution” processing technology for identifying thin reservoirs in the Maokou Formation was investigated. In response to the adverse factors such as deep burial, weak signals, thin reservoirs, and small impedance contrast with surrounding rocks in the Maokou Formation, research on weak-signal processing techniques for deep carbonate formations was conducted. In addition, the “high-fidelity and high-resolution” processing technology involving the concept of “protecting low frequencies and enhancing high frequencies” throughout the processing workflow was proposed. Methods such as pre-stack fidelity denoising, high-frequency residual static correction, multiple-wave suppression, pre-stack time migration in the Offset Vector Tile (OVT) domain, anisotropic correction, and frequency enhancement based on compressed sensing were studied to highlight the weak reflection signals of the Permian dolomite reservoirs. The research results effectively identified favorable reservoirs in the Maokou Formation as discontinuous, medium-to-strong amplitude reflections, with a well-seismic calibration match rate of 100%, thereby achieving effective identification and precise characterization of heterogeneity of thin dolomite reservoirs in the Maokou Formation. Based on the new findings, the drilled well PY001-H1 successfully reached high-quality reservoirs. Therefore, the results demonstrate that the studied “high-fidelity and high-resolution” processing technology is beneficial for deep reservoir prediction by improving seismic data resolution, amplitude preservation, heterogeneity analysis, and weak signal recovery.

Key Technical Descriptions

1. Pre-stack Fidelity Denoising Technology

Fidelity processing aims to preserve both amplitude and phase integrity. The current approach primarily employs the “six-step method” for denoising in the F-X, F-K, and Tau-P domains, which involves classification and segmentation by time, frequency, domain, step, and region. The principle of fidelity ensures that the chosen modules protect effective signals while maintaining relative amplitude relationships, particularly preserving low-frequency signals and improving the signal-to-noise ratio of weak high-frequency signals.

The data in this area is primarily affected by impulsive noise, surface waves, and linear noise. While methods for suppressing impulsive and surface waves are well-established, this study used frequency-domain suppression and localized surface wave suppression to maintain the original data’s frequency range. Due to wide-azimuth acquisition, residual surface waves and linear noise often remain at non-vertical offsets. In these cases, cross-spread domain suppression was applied post-static correction. This method considers interference wave frequency, apparent velocity, and non-vertical offsets for effective suppression.

2. Multiple-Wave Suppression Technology

A combination of Radon transforms and curvelet transform achieved effective suppression of interlayer multiples. Specifically, high-precision Radon transform was used to obtain multiple models, enhancing signal identification during curvelet transform. The curvelet transform better differentiates energy by frequency, dip, and position. By controlling the strength of simulated multiples through thresholds and applying adaptive subtraction, this approach improves multiple-wave suppression and enhances data quality for low signal-to-noise ratio datasets.

3. Compressed Sensing Frequency-Expansion Technology

Compressed sensing is a novel signal sampling theory. The method used here employs a robust compressed sensing spectral inversion algorithm. It determines initial reflection coefficients using a thin-layer matching pursuit algorithm and then performs post-stack sparse inversion based on the L0-norm compressed sensing theory. The final reflection coefficient model is obtained by applying regularization. Subsequently, wavelet decomposition and high-frequency wavelet replacement are conducted to expand the high-frequency spectrum while maintaining amplitude and fidelity, significantly improving seismic data resolution.

Application Results

The test area spans 200 km², using an observation system with 24 lines, 7 sources, 270 receivers, and 180-fold coverage. The bin size is 20 × 20 meters, with a maximum offset of 6 332.46 meters and an aspect ratio of 0.62, representing typical high-density and wide-azimuth acquisition. The raw data exhibits high noise levels, low dominant frequencies in deep layers, and narrow frequency bandwidths. Interferences include surface waves, impulsive noise, interlayer multiples, and anisotropy effects. After applying “Double-High Processing,” multiple-wave interferences were eliminated, and the quality of gathers significantly improved. Well-to-seismic profiles achieved good alignment, yielding favorable results.

For example, well PY1, with a burial depth of 6 040 meters and a reservoir thickness of 7 meters, showed a weak reflective base on synthetic seismograms and seismic profiles, with consistent matching between synthetic records and seismic waveforms. Clear reservoir characteristics were identified. Following these advancements, a new development well was drilled to a depth of 6 103 meters. In the second member of the Maokou Formation, a dolomite reservoir was encountered with a slanted thickness of 23.1 meters, a vertical thickness of 8.5 meters, and an average porosity of 3.8%. This high-quality dolomite reservoir achieved excellent drilling results.

Key words: Permian Maokou Formation, dolomite, high-fidelity, high-resolution, multiple wave, OVT domain, compressed sensing