A method that includes a genetic algorithm (GA), principal component analysis (PCA), and an artificial neural network (ANN) is adopted in the search for the power ascension path of a boiling water reactor that used to rely solely on an operator's experiences. The power ascension path is formulated as an optimization problem with thermal limits, e.g., minimum critical power ratio, maximum linear heat generation rate, and maximum average planar linear heat generation rate, and with the stability requirement serving as a constraint. The Simulate-3 code is used to calculate the reactor core status, while the optimization problem is solved through the use of the GA. Since the search domain of the GA is relatively large, the ANN, which models the power ascension path, is developed in order to quickly select the candidate solutions for further Simulate-3 calculations, permitting the algorithm to converge effectively. Meanwhile, PCA is used to reduce the ANN input vector's dimension, which improves the ANN training efficiency and pattern recognition capability. The results show that this method efficiently obtains the proper power ascension path required for meeting all constraints at different fuel exposure levels.