Budget optimization is an important issue faced by advertisers in search auctions, and has significant impact on the design of various advertising strategies. Given a limited budget on a search market during a certain period, an advertiser has to distribute her budget to a series of sequential temporal slots (i.e., days, weeks, or months), during which advertisers must avoid the budget being used up quickly, so as to keep the budget for potential clicks with better performance in the future. Considering the optimal budgets over these temporal slots as fuzzy variables, we establish a two-stage fuzzy budget allocation model, and use particle swarm optimization (PSO) algorithm to solve it in case when these optimal budgets are characterized by discrete fuzzy variables. We also conduct experiments to validate our model and algorithm. The experimental results show that our model can outperform other five budget allocation strategies in terms of reducing the revenue loss of the advertiser.