本文關(guān)鍵詞：高溫高壓酸性天然氣中飽和含水量理論模型研究　出處：《西南石油大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 酸性天然氣 飽和含水量 支持向量機 BP神經(jīng)網(wǎng)絡(luò) 灰色關(guān)聯(lián)分析 遺傳算法 異常點檢測

【摘要】：在常規(guī)和高含H_2S、CO_2酸性天然氣藏開采、集輸、加工過程中,防堵、防腐和防水合物生成是貫穿整個天然氣藏開采工作的重中之重。而這三類問題的產(chǎn)生,都與天然氣飽和含水量有密切聯(lián)系。天然氣飽和含水量是天然氣生產(chǎn)、加工和運輸過程中最重要的參數(shù)之一,直接影響著天然氣加工工藝設(shè)計及天然氣開采、集輸與處理等過程中工藝計算的準(zhǔn)確性。本文將遺傳算法(GR)、灰色關(guān)聯(lián)分析、異常點檢測等方法與支持向量機(SVM)、BP神經(jīng)網(wǎng)絡(luò)(BP-ANN)相結(jié)合,采用GA-SVM和BP-ANN模型對常規(guī)天然氣,純H_2S和純CO_2氣體,H_2S、CO_2混合氣體及高含H_2S、CO_2酸性天然氣等多個體系飽和含水量進行了系統(tǒng)的研究。模型主要以體系的溫度和壓力,CH4、H_2S、CO_2摩爾分?jǐn)?shù)及天然氣相對密度等為輸入變量,飽和含水量為輸出變量。模型所有數(shù)據(jù)的適用范圍為:溫度 0～325℃,壓力 0～150MPa,H_2S 摩爾分?jǐn)?shù) 0～100%,CO_2 摩爾分?jǐn)?shù)0～100%。在GA-SVM和BP-ANN模型計算過程中,在316個文獻實驗數(shù)據(jù)點訓(xùn)練的基礎(chǔ)上,進行了 83個數(shù)據(jù)點的多體系飽和含水量預(yù)測。計算結(jié)果表明:采用GA-SVM模型計算常規(guī)天然氣,純H_2S和純CO_2氣體及高含H_2S、CO_2酸性天然氣飽和含水量的絕對平均偏差(AAD)分別為 1.38%,2.6%,4.3%和 6.28%;采用 BP-ANN 模型計算 H_2S、CO_2混合氣體飽和含水量的絕對平均偏差為11.93%。通過實例計算及與已用的ANN模型,半經(jīng)驗?zāi)Ｐ?Bukacek模型,Awad模型,AQUAlibrium軟件,簡化熱力學(xué)+Bahadori模型,簡化熱力學(xué)+Mohammadi模型,Bahadori+Khaled模型,王俊奇模型,修正熱力學(xué)模型等計算結(jié)果進行對比表明:GA-SVM和BP-ANN模型預(yù)測精度高、泛化能力強、穩(wěn)定性好,更能滿足工程精度要求。最終,基于杠桿方法開展了所有數(shù)據(jù)的異常點檢測。其檢測結(jié)果表明:GA-SVM和BP-ANN新模型所有檢測數(shù)據(jù)點中出現(xiàn)異常點的個數(shù)最少,模型的穩(wěn)定性最強,從而為常規(guī)天然氣,純H_2S和純CO_2氣體,H_2S、CO_2混合氣體及高含H_2S、CO酸性天然氣等多個體系飽和含水量準(zhǔn)確計算提供了一種新的綜合分析方法。
[Abstract]:Prevention of plugging in conventional and high H2S- CO2 acid natural gas reservoirs for recovery, gathering, transportation and processing. Anticorrosion and hydrate prevention are the most important problems in the whole exploitation of natural gas reservoir, and these three problems are closely related to the saturated water content of natural gas, which is the production of natural gas. One of the most important parameters in the process of processing and transportation has a direct impact on the accuracy of process calculation in the process of natural gas processing design, natural gas extraction, gathering, transportation and treatment. In this paper, genetic algorithm (GA) is introduced. The methods of grey correlation analysis and anomaly detection are combined with support vector machine (SVM) and BP neural network (BP-ANN). GA-SVM and BP-ANN models are used for conventional natural gas. Pure H2s and pure CO_2 gases H2SCOS2 mixed gas and high H2S content. The saturated water content of several systems, such as CO_2 acid natural gas, was systematically studied. The model was mainly based on the temperature and pressure of the system. The molar fraction of CO_2 and the relative density of natural gas are input variables, and the saturated water content is the output variable. The applicable range of all data of the model is: temperature 0 ~ 325 鈩，

本文編號：1405396