本論文利用美國核能管制委員會委託聖帝亞國家實驗室所研究的MELCOR程式，分析我國第一座核能發電廠─金山電廠電廠全黑嚴重事故，並把結果與美國核能工業界所發展出的MAAP5程式比較，比較重點除了熱水流模式分析，還包括了氫氣產生與放射性物質外釋結果分析。 本研究內容主要包括：(1) 以核能研究所建立的MELCOR核一廠輸入檔為基礎，依據核研所提供的MAAP5程式輸入檔之計算書，建立新的MELCOR輸入數據，以便兩程式模擬結果之比較；(2)運用兩程式模擬電廠全黑之高壓爐心毀損、壓爐心毀損事故，低壓事故假設在水位降至燃料頂端時，手動開啟自動洩壓裝置；(3)比較兩程式預測之高壓與低壓爐心毀損事故的熱水流反應、氫氣產生量、與放射性物質外釋程度和外釋至環境的比例，並比較兩程式結果之差異與在高壓與低壓事故中之差異。 比較兩程式模擬結果後，發現兩程式較大之差異為：(1) MELCOR對爐心與壓力槽內部區域的模擬較詳細，徑向分為三環，軸向分為13層，MAAP5則視為一整體，此項差異造成模擬結果有極大的差異；(2)兩程式對於壓力槽底部破裂的模擬有差異；(3)兩程式對於flow blockage的設定與模擬方式不同，造成驢心內的氫氣產生量有極大的差異，flow blockage 對MAAP5的結果影響較顯著，MELCOR則較不顯著；(4)兩程式對於放射性物質的分類有所不同，前述的三項模式的差異，造成預測之放射性物質釋出有差異，外釋至外界環境的比例也有不同；(5) MELCOR程式時間間隔(Time Step Size) 的設定對結果具有明顯影響。 In the present study, a MELCOR input deck for the Chinshan Nuclear Power Station of Taiwan Power Company is developed. Chinshan nuclear power station employs a Boiling Water Reactor (BWR IV) designed by General Electric and Mark I containment. The input deck is used to analyze the station blackout sequence, and the results will be compared with the MAAP5. The work involved in the study includes: (1) Use the MELCOR input deck from INER as the basis. Build a new MELCOR input deck of Chinshan nuclear power station according to the MAAP5 input deck and the corresponding calculation sheets from INER. (2) Initialize the new MELCOR input deck to staeday state. (3) Simulate the SBO event of the plant using MELCOR and MAAP5 codes with the assumption that the core melt occurs under high pressure and low pressure. (4) Compare the results of these two codes. The major focus are the timing of major events, the thermal hydraulic responses of reactor coolant system and containment, hydrogen generation, the radionuclide releases from core during the core melt and during the molten core concrete interactions, and the fraction of radionuclide releasing to the environment. Compared the results, it has been found that: (1) MELCOR has a more detailed modeling of core and vessel internal regions. It consists of 3 radial rings and 13 axial levels. MAAP5 treats the core as a single volume. (2) The reactor vessel bottom attack model amd mode of its failure of these two codes are also significantly different. (3) The amount of hydrogen generation during the core melt as predicted by these two codes are significantly different. The impacts of flow blockage on the prediction of hydrogen generation of these two codes are different. MAAP5 is more sensitive to the assumpation of flow blockage. (4) The classification of radionuclide groups is different. Due to the difference in the modeling of core region, the predicted in-vessel releases of radionuclide is different. The predicted ex-vessel releases are also significantly different due to difference in the modeling of core concrete interactions. The fraction of each radionuclide released to the environment is different. (5) The MELCOR results are very sensitive to the time step size. If the time step size has not been set properly, the code stops calculation prematurely.