模擬退火與基因演算於震測速度選取之最佳化之研究

Abstract

我們採用模擬退火與基因演算法來解決反射震測訊號中速度分析的問題。模擬退火與基因演算法具有搜尋全域最佳解的能力。傳統的震測速度選取是由地球物理專家透過觀察 semblance image 來進行，得出時間與速度之間的關係函式，然而此過程耗時。我們把整個速度選取問題轉化成為組合最佳化的問題，並建立能量函式，包含被選取點的semblance、數量、與interval velocity和velocity slope的限制條件，計算候選的time-velocity pairs的能量，進而利用模擬退火演算法與基因演算法找到全域最佳解，即最佳化的 stacking velocity與時間的對應關係。之後，可進一步執行動態修正及共中點疊加，以反映真實地層的原貌 (in time)。我們的研究有助於進一步的震測資料處理與解釋。

We adopt two global optimization methods, simulated annealing method (SA) and genetic algorithm (GA), for seismic velocity picking in reflection seismic data. Conventional seismic velocity picking was made by geophysical experts through looking at the semblance image to pick several peaks representing the relation of time and stacking velocity, however it was a time consuming task. In this study, we consider the velocity picking problem as a combinatorial optimization problem and define an energy function consisting of the semblance of picked points, picking number, and the constraints of interval velocity and velocity slope for GA and SA. By using SA and GA to calculate the energy of a polyline, the best solution can be obtained. Furthermore, we apply the obtained polyline to do the normal move-out (NMO) correction and stacking. Our research can improve the further seismic data processing and interpretation.