| 摘要: | 台灣西南部永安海嶺的天然氣水合物和游離氣存在於鬆散的海相沉積物中，對震測資料反應記錄的物理特性有很大影響。孔隙率、密度、縱波速度（Vp）、橫波速度（Vs）、Vp/Vs、泊松比(ν)、Lambda-Rho(λ??)、Mu-Rho(μ??)、Lambda/Mu比值(λ /μ)變化很大。因此，我們旨在分析鬆散海相沉積物與其中游離氣和天然氣水合物的物理性質，並嘗試對永安海嶺天然氣水合物和游離氣藏特徵進行研究。透過天然氣水合物和游離氣的指標進行地物的震測法探測，嘹解構造、形貌或沉積過程特徵、震測地層變化，甚至岩石或孔隙流體性質，都可以先經過地震屬性分析來完成。屬性研究成為標定天然氣水合物穩定帶與相關的潛在震波響應的初步理解手段，並為水合物帶以下可能存在高潛在 GH 濃度和持續供應天然氣的區域提供建議。接著是透過疊後和疊前反演來分析、估算天然氣水合物和游離氣的聲學與彈性岩石物理。
受限於沒有測井資料，疊後反演的處理是將每個CMP的處理結果當作假想的井測資料，我們的方法是先進行疊後建模。疊後建模試圖先定義一維速度模型，並透過手動模擬定義 P 波速度模型，將某些選定 CMP 位置的最後疊加數據視為單獨的偽測井。根據這些選定的偽測井來定義一維模型，首先,通過基於捲積(convolutional model)模型的反演(MBI)方法進行疊後反演。 MBI 產生 P-阻抗 (Zp)，通過 HRS 軟件中的線性方程，確定 Vp模型後，然後可以進一步估計密度。基於模型的反演法所得的 Vp 和密度可用於建立初步的聲學岩石物理模型，來進行疊前反演。同步反演(Simultaneous Inversion)是執行疊前反演的著名方法。 SI 嘗試根據 Fatti 方程計算其對應的聲學與彈性阻抗及對應的反射率，已角度震測聚集(gather)從事逆推的角度範圍為 0o 到 40o。透過反演獲得的反射率包括 P 阻抗的反射率、S 阻抗的反射率和密度的為主要參數代表。因此，在疊前反演中獲得 P 波、S 波、密度、Zp、Zs 和 Vp/Vs 等聲、彈性特性。其他相關聲、彈性岩石物理特性參數，如泊鬆比、 μ、 楊氏模數E=σ/ε、體積模數也稱為不可壓縮量K、λ/μ ratio、λ??和 μ?? 皆可以透過方程導出。對永安嶺天然氣水合物和游離氣非常敏感的參數是 λ??和 μ??。它們可以透過交叉比對分析所顯示的背景趨勢將岩石物理參數分開。岩性也可以透過交叉比對分析進行分離。為了增強這種分析逆推與伴隨地解釋結果，我們使用無監督式機器學習算法(例如 K-means 群集分析)來區分天然氣水合物和游離氣體的岩性特徵。通過使用淺海MSCL 數據顯示，永安海嶺至少有 七種不同的岩性。使用建議的 Elbow 方法所提供的附加信息來選擇最佳 K 值，然後可以執行 K-means 群集分析。Silhoutte Score 的值介於 -1 到 1 則可用於判斷 K-means 群集類別的結果。天然氣水合物可通過高 P-、S-、Zp、Zs 和略高的 Vs/Vp、略低的 λ/μ 和泊松比 (ν) 來識別，而游離氣可以通過低 VP-、AI、EI，無明顯可變 S 阻抗、高 Vs/Vp、低 λ/μ、 λ??和 ν來識別.
;Gas hydrate and free gas in Yung-An Ridge, Offshore SW Taiwan are existed in unconsolidated marine sediment and give a large effect on the elastic properties. The elastic properties such as porosity, density, P-wave velocity (Vp), S-wave velocity (Vs), Vp/Vs, Poisson’s Ratio, Lamdha-Rho (λρ), Mu-Rho (μρ), Lamdha/Mu (λ/μ) are varied a lot. So, we purpose to distinguish the physical properties of gas hydrate and other mineral bearing in unconsolidated marine sediment, and try to characterize the gas hydrate and free gas reservoir in Yung-An Ridge. Detecting gas hydrate and free gas’ indicator and knowing the structure, morphological or depositional processes features, stratigraphy, even for rock or pore fluid properties can be done through seismic attribute method. The attribute study become the initial effort to allocate the potential seismic responses associate with gas hydrate stability zone and provide the suggestion on area where high potential GH concentration and continuous supply of gas below the hydrate zone can exist. Next step is to characterizing the gas hydrate and free gas through post-stack and pre-stack inversion.
The approach to perform post-stack inversion with the limitation of budget, there is no well log information, we do post-stack modeling. The post-stack modeling tries to define a 1D velocity model and treat the stacked data at some selected CMP locations as a individual pseudo-log with efforts by manually define the P- wave velocity model. With 1D model defined from these selected pseudo-log, post-stack inversion were performed first through Model-Based Inversion (MBI) method. The MBI produces P-Impedance (Zp), and with the linear equation in the HRS Software, Vp can be determined, and then density can be estimated. Vp and density from Model-Based Inversion can be used to initial guess for pre-stack inversion. Simultaneous Inversion (SI) is the famous method to perform the pre-stack inversion. SI try to calculate reflectivity based on Fatti’s equation with angle range 0o to 40o. The reflectivity that obtained in this inversion is representative by reflectivity of the P-impedance, reflectivity of the S-Impedance, and reflectivity of the density. Hence, the elastic properties such as P-wave, S-wave, density, Zp, Zs, and Vp/Vs can be obtained in a single inversion. Other parameters of elastic properties such Poisson’s Ratio, λρ, μρ, and λ/μ can be derived through equations. The parameters which are very sensitive with the gas hydrate and free gas in Yung-An Ridge are λρ, and μρ. They are separated from background trend that can be shown through cross plot. The lithology can be separated through crossplot too. To enhance the result of this interpretation, we use unsupervised Machine Learning algorithm such as K-means cluster to distinguish gas hydrate-and free gas from the lithology. By using MSCL Data that show information that at least, there are 7 lithologies in the Yung-An Ridge. With additional information from Elbow method to choose the optimal K-value then K-means cluster can be performed. Silhoutte Score with value -1 to 1 can be used judge to K-means cluster’s result. Gas hydrate can be identified with high P-, S-, Zp, Zs, λ?? and slightly higher Vs/Vp, slightly lower λ/μ and Poisson’s ratio (ν) while free gas can be identified with low P-, AI, EI, none-varying S- impedance, high Vs/Vp, low λ/μ, λ?? and ν. |
