發(fā)布時間：2018-06-24 06:59

  本文選題：SNESIM + 緊湊型樹��； 參考：《中國地質(zhì)大學(xué)(北京)》2015年碩士論文

【摘要】：本論文通過對多點(diǎn)地質(zhì)統(tǒng)計學(xué)進(jìn)行研究,選取其中經(jīng)典序貫非迭代算法SNESIM(Single Normal Equation Simulation)進(jìn)行剖析,并對該算法建模過程中影響性較大的幾個因素分別進(jìn)行了分析。針對算法在建模過程中內(nèi)存消耗過多這個問題進(jìn)行深入研究,進(jìn)而提出優(yōu)化方案,通過“緊湊型搜索樹”進(jìn)行模式優(yōu)選與壓縮,編碼實施。并對優(yōu)化算法進(jìn)行實際工區(qū)建模應(yīng)用,最終對建模效率進(jìn)行記錄,進(jìn)而對記錄分析,最終對建模效果進(jìn)行評估。目標(biāo)比率與伺服系統(tǒng)修正參數(shù)、掃描模板節(jié)點(diǎn)個數(shù)、多級網(wǎng)格、最小重復(fù)個數(shù)四個因素對建模過程,具有重要的影響作用。前期算法分析的過程中,通過對目標(biāo)比率與伺服系統(tǒng)修正參數(shù)的實驗,可以對整體建模結(jié)果中各種模擬結(jié)果比例進(jìn)行量化控制。這兩個參數(shù)通過互補(bǔ)作用影響河道走向。掃描模板節(jié)點(diǎn)個數(shù)對速度以及模擬效果影響較大。通過對該參數(shù)實驗,分析,發(fā)現(xiàn)模板節(jié)點(diǎn)個數(shù)與模擬工區(qū)大小關(guān)系緊密,較大的工區(qū)需要進(jìn)行相應(yīng)增大,但是不可過大,否則會影響速度,但不會對模擬效果有所提升。多級網(wǎng)格方法的提出主要是解決在節(jié)點(diǎn)數(shù)一定的情況下,對大范圍數(shù)據(jù)結(jié)構(gòu)再現(xiàn)的一個問題。通過多級網(wǎng)格的應(yīng)用,在較小模板節(jié)點(diǎn)個數(shù)的情況下,能夠?qū)Υ蠓秶臻g結(jié)構(gòu)進(jìn)行再現(xiàn)。最小重復(fù)個數(shù)保證提取出來模式的有效性。通過最小重復(fù)個數(shù)的限制,把其中不重要的模式排除在外,使得模擬結(jié)果更加精確。SNESIM算法創(chuàng)新點(diǎn)主要通過搜索樹的引入使得原來多點(diǎn)地質(zhì)統(tǒng)計學(xué)建模方法變得實際可用。搜索樹大大優(yōu)化了內(nèi)存的消耗,使得模擬速度加快,同時對建模過程進(jìn)行了優(yōu)化,從而使多點(diǎn)地質(zhì)統(tǒng)計學(xué)的發(fā)展,從理論研究發(fā)展到了真正的實用階段。為三維模型的建立,提供了非常實用可靠的算法支撐。但隨著石油開采的發(fā)展,所產(chǎn)生的數(shù)據(jù)量在飛速的增長。導(dǎo)致大型三維建模對內(nèi)存的要求更高。本文通過對搜索樹結(jié)構(gòu)認(rèn)真研究的基礎(chǔ)上,提出緊湊型樹進(jìn)行優(yōu)化,主要通過對訓(xùn)練圖像中提取的模式進(jìn)行優(yōu)選與壓縮,從而使得內(nèi)存消耗大大降低,并進(jìn)行代碼編寫,使得模式優(yōu)化得以實現(xiàn),提升了計算速度。最終通過工區(qū)的實際應(yīng)用。在多種訓(xùn)練圖像為基礎(chǔ)的前提下,進(jìn)行了多次多點(diǎn)建模。發(fā)現(xiàn)這種算法能夠在保證信息完整的前提下,加快三維模型建模速度。
[Abstract]:In this paper, we select SNESIM (single normal equation Simulation), a classical sequential non-iterative algorithm, to analyze the multi-point geostatistics, and analyze several influential factors in the modeling process of the algorithm. In order to solve the problem of excessive memory consumption in the modeling process of the algorithm, the optimization scheme is put forward, and the pattern is optimized and compressed by "compact search tree", and the coding is implemented. Finally, the efficiency of modeling is recorded, then the record is analyzed, and the modeling effect is evaluated. The ratio of target to the modified parameters of servo system, the number of scanning template nodes, the number of multilevel grids and the minimum number of duplicates play an important role in the modeling process. In the process of early algorithm analysis, through the experiment of target ratio and servo system correction parameters, the proportion of simulation results in the whole modeling results can be quantitatively controlled. These two parameters influence the river course through complementary action. The number of scan template nodes has great influence on the speed and simulation effect. Through the experiment of this parameter, it is found that the number of template nodes is closely related to the size of the simulated work area, and the larger work area needs to be increased correspondingly, but not too large, otherwise it will affect the speed, but it will not improve the simulation effect. The multilevel grid method is proposed to solve the problem of reproducing the large scale data structure in the case of a certain number of nodes. With the application of multilevel grid, the large spatial structure can be reproduced under the condition of small number of template nodes. The minimum number of duplicates ensures the validity of the extracted pattern. By limiting the minimum number of duplicates, the unimportant patterns are excluded, which makes the simulation results more accurate. The innovation point of SNESIM algorithm makes the original multi-point geostatistical modeling method practical and practical mainly through the introduction of search tree. The search tree greatly optimizes the memory consumption, speeds up the simulation speed, and optimizes the modeling process, which leads to the development of multi-point geostatistics, from the theoretical research to the real practical stage. It provides a very practical and reliable algorithm support for the establishment of three-dimensional model. But with the development of petroleum exploitation, the amount of data generated is increasing rapidly. This leads to higher memory requirements for large-scale 3D modeling. Based on the careful study of the structure of the search tree, this paper puts forward the compact tree optimization, mainly through the optimal selection and compression of the pattern extracted from the training image, so that the memory consumption is greatly reduced, and the code is compiled. The mode optimization is realized and the calculation speed is improved. Finally, through the practical application of the work area. Based on multiple training images, multiple multi-point modeling is carried out. It is found that this algorithm can accelerate the modeling speed of 3D model on the premise of information integrity.
【學(xué)位授予單位】：中國地質(zhì)大學(xué)(北京)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2015
【分類號】：P618.13;P628

【參考文獻(xiàn)】

相關(guān)期刊論文 前3條

1 侯景儒;;地質(zhì)統(tǒng)計學(xué)在我國的應(yīng)用及其發(fā)展[J];地質(zhì)與勘探;1991年04期

2 王家華;趙巍;;基于地震約束的地質(zhì)統(tǒng)計學(xué)建模方法研究[J];海洋石油;2010年04期

3 熊哲;嚴(yán)錫;金楊波;;油藏地質(zhì)建模中多點(diǎn)訓(xùn)練圖像應(yīng)用研究[J];江漢石油職工大學(xué)學(xué)報;2012年04期

，

本文編號：2060498