核能電廠爐心核燃料是核能發電能量及放射性物質的根本來源，因此運轉人員於控制機組升載時，係以在能確保核燃料完整之安全前提下，儘速達到額定功率運轉，獲致最大經濟效益為目標。為能確保核燃料於正常運轉、預期暫態運轉及發生類似冷卻水流失事故時之完整，機組於升載過程中之重要運轉參數，例如：最小臨界熱功率比、最大燃料單位長度之發熱率、最大平面單位長度平均發熱率等熱限制值、功率震盪等，均必須時刻符合較可能真正影響燃料完整之安全限制值更嚴格、保守之運轉限制值要求。但由於機組於功率變動期間，爐心各項參數，包括：溫度、空泡、反應度、氙濃度、功率分佈等之反應暫態交互作用，導致爐心物理現象極為複雜，因此實務上難有標準升載操控程序，可供運轉人員遵循，導致升載策略依人員經驗不同而有所不同，此現象除易造成運轉人員的負擔外，並可能因運轉人員過於保守而導致升載緩慢，或過於躁進致使安全餘裕降低等，而難以達到兼顧最安全及最大經濟效益之目標。
為能克服上述問題，本研究利用Simulate-3模擬爐心狀態，並以基因演算法將搜尋升載軌跡問題轉化成多目標求解問題，其中定義適應函數包含各項熱限制值、穩定度、最大控制棒負載線及升載時間等要求，利用基因世代演化，以搜尋出滿足該適應函數定義之最佳升載軌跡解。由於適應函數係以計算耗時之Simulate-3計算升載軌跡所得之爐心參數為函數變數，因此為能降低無謂演化計算時間的浪費，本研究除初步利用專家運轉經驗，將升載軌跡之參數（控制棒位、爐心流量）適當設限，以先刪除許多不合物理意義，明顯違反運轉限制之不適當解，俾合理縮小搜尋空間外，並將搜尋空間內可能之升載軌跡解，利用離線訓練完成之類神經網路選擇器篩選，初步判斷該解是否合理，合理者持續基因演算法之演化程序，不合理者則予以剔除，以進一步確保演化時解的品質。而建立類神經網路選擇器時，為能增加學習及辨識效能，故使用主成分分析法擷取輸入資料之特性，以維度較原參數少之綜合指標解釋原輸入資料，達到降低輸入向量維度之目的。總括而言，本研究應用基因演算法之多目標求解功能，並結合類神經網路之分類及主成分分析法之維度縮減特性，可自動搜尋出沸水式反應器在不同燃耗點符合適應函數定義下之最佳升載軌跡，模擬測試結果顯示，於燃料週期之初、中、末階段，皆可搜尋出適當之升載軌跡。

The nuclear fuel rods of the nuclear power plant are both the source of energy and radiation. Therefore, the thumb rule for the power ascension operation is to bring up the power to the rated level as efficiently as possible, however under the premise of fuel integrity, to maximize economic benefits. To ensure the fuel rod integrity at all time, including normal operation, anticipated transient operation, and loss of coolant type of accident, the crucial operational parameters such as minimum critical power ratio, maximum linear heat generation rate, maximum average planar linear heat generation rate, power oscillation, etc., must be kept within the operation limit value which are more restrict and conservative than safety limit value. It is difficult to complete the aforementioned task because of the complicated characteristics of the boiling water reactor core, such as relationship between temperature, void and reactivity, local power distribution and xenon transient. There is no standard operating procedure to guide the operator in performing control rod withdrawal and core flow rate changes; operators must determine their actions with on-site measurements and experiences, and these actions often lead to operational difficulties. As a result, the power ascension strategy differs by each operator. In addition to the extra work load, some operators may be too aggressive to narrow the safety margin, or too conservative to delay the power ascension process which is not economic and efficient. A pre-defined power ascension path will be beneficial to the operators.
To overcome the above issue, the Simulate-3 code was used to calculate the reactor core status. The requirements of power ascension path was formulated by fitness function as an optimization problem with power ascension time, thermal limits, core stability and maximum rod line, as the constraints. Being an efficient and stable global search algorithm, the genetic algorithm was adopted to search for the optimized power ascension path. The fitness value was based on the core status calculated by Simulate-3, which was very time consuming. To reduce the effort, this study incorporated in experts’ operation experience to define the preliminary constraints of power ascension parameters （control rod position and core flow）to confine the searching domain by eliminating some improper solutions that were against the operation requirement, and the offline-trained artificial neural network（ANN）selector was introduced to screen the power ascension path by excluding the improper solutions from further evaluations. The input vector’s dimension of ANN was relatively large, so that principal component analysis（PCA）is utilized to reduce the dimension. As the result, the ANN training became more efficient, and the pattern recognition capability was improved as well. In summary, the research combined the multi-object optimization of GA, pattern recognition of ANN and dimension reduction of PCA to search for the optimized power ascension path of boiling water reactor （BWR）. The simulation results showed that the developed algorithm can obtain the adequate power ascension path at beginning, middle, and end of cycle.

1. 蘇獻章、李綺思，”京都議定書施行週年之世界核能現況”，核能簡訊，中華民國核能學會、財團法人核能資訊中心，2006
2. M. Maqua, K. Kotthoff, and W. Pointner, “Neutron flux oscillations at German BWRs,” Das EUROSAFE-Forum 2002, Berlin (2002).
3. “Mechanical & Nuclear Training,” General Electric Company (1980).
4. 核能發電工程學，楊昭義、歐陽敏盛，國立編譯館，1997
5. M. Gen and R. Cheng, Genetic Algorithms & Engineering Design, Wiley-Interscience, New York (1997).
6. M. Gen and R. Cheng, Genetic Algorithms & Engineering Optimization, Wiley-Interscience, New York (2000).
7. D. Whitely, “A Genetic Algorithm Tutorial,” Technical Report CS-93-103, Colorado State University (1993).
8. E. K. P. Chong and S. H. Zak, An introduction to optimization, Wiley-Interscience (2001).
9. “Simulate-3 Manual,” Studsvik Scandpower, STUDSVIK/SOA-95/15 Rev 2. (2001).
10. M. F. Lin, Neural Networks in Computer Intelligence, McGRAW-HILL International Editions (1994).
11. S. Y. Kung, Digital Neural Networks, Prentice-Hall International Editions (1993).
12. B. Kosko, Neural Networks and Fuzzy Systems, Prentice-Hall International Editions (1992).
13. Y. Xin, “Evolving Artificial Neural Networks,” Proceedings of the IEEE, 87, 1423 (1990).
14. 神經網路入門與圖樣辨識，林昇甫、洪成安，全華科技，1993
15. 類神經網路理論與應用，盧炳勳、曹登發，全華科技，1992
16. 神經網路導論，傅心家，第3波，1991
17. A. Hojjat and S. L. Huang, Machine Learning, John Wiley & Sons, Inc (1995).
18. S. Sharma, Applied Multivariate Techniques, Wiley-Interscience, New York (1996).
19. 多變量解析與其應用，曾國雄，華泰書局，1978
20. J. Shlens, “A tutorial on Principal Component Analysis,” (2005).
21. A. Galperin, “Exploration of the Search Space of the IN-Core Fuel Management Problem by Knowledge-Based Techniques,” Nuclear Science & Engineering, 119 (1995).
22. H. Siegelmann, E. Nissan and A. Galperin, “A Novel Neural /Symbolic Hybrid Approach to Heuristically Optimized Fuel-Allocation and Automated Fuel-Allocation in Nuclear Engineering,” Advances in Engineering Software, 28 (1997).
23. Y. Tahara, K. Hamamoto and M. Takase, “Computer Aided System for Generating Fuel Shuffle Configuration Based On Knowledge Engineering,” Journal of Nuclear Science and Technology, 28 (1991).
24. B. M. Rothleder, G. R. Poetschhat, W. S. Faught and W. J. Eich, “The potential for expert system support is solving the prssurized water reactor fuel shuffling problem,” Nuclear Science and Engineering, 100 (1988).
25. Y. P. Mahlers, “Core Loading Pattern Optimization based on Simulated Annealing and Successive Linear Programming,” Ann. Nucl. Energy, 22, 29 (1995).
26. C. Lin, J. I. Yang, K. J. Lin, and Z. D. Wang, “Pressurized Water Reactor Loading Pattern Design Using the Simple Tabu Search,” Nucl. Sci. Eng., 129, 61 (1998).
27. K. J. Lin and C. Lin, “Pressurized Water Reactor Reload Design by Expert System,” Nucl. Sci. Eng., 130, 128 (1998).
28. S. Jagawa, T. Yoshii, and A. Fukao, “Boiling Water Reactor Loading Pattern Optimization Using Simple Linear Perturbation and Modified Tabu Search Methods,” Nucl. Sci. Eng., 138, 67 (2001).
29. J. J. Ortiz and I. Requena, “Using a Multi-State Recurrent Neural Network to Optimize Loading Patterns in BWRs,” Ann. Nucl. Energy, 31, 789 (2004).
30. J. J. Ortiz and I. Requena, “An Order Coding Genetic Algorithm to Optimize Fuel Reloads in a Nuclear Boiling Water Reactor,” Nucl. Sci. Eng., 146, 88 (2004).
31. J. Zhao, E. Nissan, B. Knight, and A. Soper, “FuelGen: a Genetic Algorithm Based System for Fuel Loading Pattern Design in Nuclear Power Systems,” Expert Systems with Applications, 14, 461 (1998).
32. Y. Kobayashi and E. Aiyoshi, “Optimization of Boiling Water Reactor Loading Pattern Using an Improved Genetic Algorithm,” Nucl. Technol., 143, 144 (2003).
33. A. Yamamoto, “Application of Neural Network for Loading Pattern Screening of In-Core Optimization Calculations,” Nucl. Technol., 144, 63 (2003).
34. A. Yamamoto and H. Hashimoto, “Application of the Distributed Genetic Algorithm for In-Core Fuel Optimization Problems under Parallel Computational Environment,” J. Nucl. Sci. Technol., 39, 1281 (2002).
35. J. L. Montes, J. J. Ortiz, I. Requena, and R. Perusquia, “Searching for Full Power Control Rod Patterns in a Boiling Water Reactor Using Genetic Algorithms,” Ann. Nucl. Energy, 31, 1939 (2004).
36. T. Kiguchi, “Optimization of Boiling Water Reactor Control Rod Patterns Using Linear Search,” Nucl. Technol., 67, 38 (1984).
37. C. Lin, “An Automatic Control Rod Programming Method for a Boiling Water Reactor,” Nucl. Technol., 92, 118 (1990).
38. L. S. Lin and C. Lin, “A Rule-Based Expert System for Automatic Control Rod Pattern Generation for Boiling Water Reactors,” Nucl. Technol., 95, 1 (1991).
39. J. L. Francois, C. Martin-del-campo, and A. Tavares, “Development of a BWR Control Rod Pattern Design System based on Fuzzy Logic and Knowledge,” Ann. Nucl. Energy, 31, 343 (2004).
40. J. A. Castillo, J. J. Ortiz, G. Alonso, L. B. Morales, and E. Valle, “BWR Control Rod Design Using Tabu Search,” Ann. Nucl. Energy, 32, 741 (2005).
41. H. Moon, A. Gu and R. G. Grummer, “Optimization of Bwr Control Rod Pattern for Relaxed Rod Sequence Exchange,” International Conference on the New Frontiers of Nuclear Technology Reactor Physics, Safety and High-Performance Computing, Korea (2002).
42. C. Lin and H. W. Lin, “Application of Fuzzy Logic Controller to Load-Follow Operations in Pressured Water Reactor,” J. Nucl. Sci. Technol., 31, 407 (1994).
43. S. G. Chi and N. I. Cho, “H∞ Control Theory Applied to Xenon Control for Load-Following Operation of a Nuclear Reactor,” Nucl. Technol., 137, 127 (2002).
44. M. G. Na, D. W. Jung, S. H. Shin, J. W. Jang, K. B. Lee, and Y. J. Lee, “A Model Predictive Controller for Load-Following Operation of PWR Reactors,” IEEE Trans. Nucl. Sci., 53, 1009 (2006).
45. M. Kinoshita, T. Fukuzaki, A. Nishimura, Y. Fukasawa and T. Matsuki, “A knowledge-based system for power maneuver planning for boiling water reactor,” Nucl. Technol. 96, 1 (1991).
46. C. Lin, Z. P. Lin, and W. J. Jiang, “Optimal Control of a Boiling Water Reactor Load Following Operation,” Nucl. Sci. Eng., 102, 134 (1989).
47. C. Lin and C. M. Shen, “ Neurocontrol of Pressurized Water Reactor in Load-Follow Operations,” Nucl. Technol., 132, 389 (2000).
48. S. Fukutomi, Y. Takigawa and H. Namba, “Operational margin monitoring system for boiling water reactor power plants,” Nucl. Technol., 100, 13 (1992).
49. T. Aldermir, G. Torri, M. Marseguerra, E. Zio and J. A. Borkowski, “Using point reactor models and genetic algorithms for on- line global Xenon estimation in nuclear reactors,” Nucl. Technol., 143, 247 (2003).
50. T. T. Chow, Z. Lin and C. L. Song, “Applying Neural Network and Genetic Algorithm in Chiller System Optimization,” Seventh International IBPSA Conference, Rio de Janeiro, Brazil (2001)
51. A. R. Burton and T. Vladimirova, “ A genetic algorithm for utilising neural network fitness evaluation for musical composition,’ Proceedings of the 1997 International Conference on Artificial Neural Networks and Genetic Algorithms, 220 (1997)
52. Y. Jin and B. Sendhoff, “Reducing Fitness Evaluations Using Clustering Techniques and Neural Network Ensembles” Proceedings of genetic and evolutionary computation conference, Seattle (2004).
53. M. Bielli, M. Caramia and P. Carotenuto, “Genetic algorithms in bus network optimization,” Transportation Research Part C, 10, 19 (2002).
54. P. P. B. Oliveira, O. Vogler and C. E. Matta, “A Combination of Genetic Algorithm and Neural Network for Diagnosing Arteriosclerotic Lesions,” 4th International Conference on Inverse Problems in Engineering, Rio de Janeiro, Brazil, (2002).
55. R. Hochman and J. P. Hudepohl, “Using the Genetic Algorithm to Build Optimal Neural Networks for Fault-Prone Module Detection,” The seventh International Symposium on Software Reliability Engineering, 152 (1996).
56. J. H. Holland, “Outline for a logical theory of adaptive system,” J. ACM, 3, 297 (1962).
57. K. De Jong, An Analysis of the Behavoir of a Class of Genetic Adaptive Systems, Ph.D. thesis, University of Michigan, Ann Arbor, (1975).
58. W. S. McCulloch and W. Pitts, “A Logical Calculus of Ideas Immanent in nervous Activity,” Bulletin of Mathematical Biophysics, 5, 115 (1943).
59. D. O. Hebb, The Organization of Behavior, New York: Addison-Wiley (1949).
60. M. L. Minsky and S. Papert, Perceptrons : An Introduction to Computational Geometry, Cambridge : MIT press (1969).
61. J. J. Hopfield, “Neural Networks and Physical Systems with Emergent Collective Computational Abilities,” Proceedings of the National Academy of Sciences, 79, 2554 (1982).
62. D. E. Rumelhart, J. L. McClelland and the PDP Research Group, Parallel Distributed Processing, Cambridge : MIT press (1986).
63. M. Moreira and E. Fiesler, “Neural Networks with Adaptive Learning Rate and Momentum Terms,” Technical Report, Institute Dalle Molle D’Intelligence Artificielle Perceptive (1995).
64. K. Pearson, “On lines and planes of closest fit to systems of points in space,” Philosophical Magazine, 2, 559 (1901).
65. H. Hotelling, “Analysis of a complex of statistical variables into principal components,” Journal of Educational Psychology, 24, 417 (1933).
66. G. B. Arfken, Mathematical Methods for Physicists, 3rd ed. Orlando, FL: Academic Press (1985).
67. H. F. Kaiser, “The application of electronic computers to factor analysis,” Educational and Psychological Measurement, 20, 141 (1960).
68. R. B. Cattell, “The Meaning and Strategic Use of Factor Analysis,” Handbook of Multivariate Experimental Psychology, Rand McNally, Chicago (1966).
69. L. W. Chan and F. Fallside, “An Adaptive Training Algorithm for Backpropagation Networks,” Computer Speech and Language, 2, 205 (1987).