壓水式反應器暫態辨識之研究

The Transient Identification for Pressurized Water Reactor

蕭德勇(Te-Yung Hsiao)

指導教授 ： 林強

國立清華大學/原子科學院/工程與系統科學系/博士(2012年)

摘要

一旦壓水式反應器發生暫態事件，如果運轉員可以辨識其類型，將有助於使系統回到安全運轉狀態。監測反應器之熱狀態，反應器冷卻系統(Reactor Cooling System，RCS)之溫度與壓力，可以幫助運轉員掌握飽和餘裕及事件的發展。在本研究中，先後發展了決定性與隨機性方法做為異常事件辨識之用。決定性方法除上述溫度與壓力外，另以反應器冷卻系統冷端與熱端溫差做為第3個系統關鍵參數；隨機性方法則增加蒸汽產生器壓力為第4個系統關鍵參數。
事件數據使用RELAP5最佳估算模式進行模擬，由其中可觀察到每個事件之暫態反應系統關鍵參數都有其特性，決定性方法利用在特有之辨識時間內之變動範圍及熱狀態變化趨勢做為事件辨識。隨機性方法以每個事件其數據的機率累加分配，每0.05為增量，從0.05到1.00，做為事件之基本資料，用作辨識區分之用。考慮量測存在雜訊，模擬的數據以亂數方式加以處理，而得到95%的信賴區間，做為異常事件資料庫之數據。辨識一個事件必須符合80個數值之比較，這些數據分別屬於4個系統關鍵參數。
由於辨識過程簡單，在非常短的時間即可完成。驗證結果皆能有效地辨識暫態事件，此兩種方法將有助於運轉員更有效地執行緊急操作程序。

關鍵字

壓水式反應器 ； 暫態辨識

並列摘要

Once an event in a pressurized water reactor (PWR) occurs and the operator has identified it, the system may then be returned to safe operational status. Usually, monitoring the thermal state of the reactor, e.g., the temperature and pressure of the reactor cooling system (RCS), the operator can realize the margin to the safety limit and the progress of the event. A deterministic approach and a stochastic approach are developed in this study. In deterministic approach, the temperature difference between hot and cold legs is additionally adopted as the third parameters used for identification. And, the steam generator pressure is adopted as the forth parameter in stochastic approach.
The event data are generated using the best estimated model RELAP5. In deterministic approach, since the variation ranges of system key parameters at a specific time duration represent the specific character of each initiating event, the identification procedure can easily determine the case by comparing the variation range of on-line data and the event data in data pool. In the stochastic approach, which accounts for measurement noise, the cumulative distribution function (CDF) is constructed and system parameters variables at the cumulative probability from 0.05 to 1.00, with 0.05 increments, are chosen for identification. The random value is added to the simulated data, and then a 95% confidence interval is obtained. To identify an event, eighty data points, i.e., twenty data points for each parameter and four parameters for each event, should match the stored data.
Since the identification procedures are simple, the computation is very fast and the results show that the event can be properly identified. These two methods will be beneficial in the context of executing an emergency operating procedure more effectively.

並列關鍵字

Pressurized Water Reactor ； Transient Identification

參考文獻

1. Embrechts, M.J., Benedek, S., “Hybrid identification of nuclear power plant transients with artificial neural networks”, IEEE Trans. on Industry Electron. 51,2004,686 – 693.

2. Lee, S.J., Seong, P.H., “A dynamic neural network based accident diagnosis advisory system for nuclear power plants”, Annals of Nuclear Energy 46, 2007, 268-281.

3. Santosh, T.V., Vinod, G., Saraf, R.K., Ghosh, A.K., Kushwaha,H.S., “Application of artificial neural networks to nuclear power plant transient diagnosis”, Reliability Engineering & System Safety 92, 2007, 1468-1472.

4. Marseguerra, M., Zoia, A., “The auto associative neural network in signal analysis: II. Application to on-line monitoring of a simulated BWR component”, Annals of Nuclear Energy 32, 2005, 1207-1223.

5. Roverso, D., “Soft computing tools for transient classification”, Information Science: an International Journal 127, 2000, 137 – 156.