In real financial markets, investors are often affected by subjective psychological cognition when making decision. We study the portfolio selection problem in fuzzy environment by considering the psychological characteristics such as reference dependence, diminishing sensitivity and loss aversion that affect investment decisions. We assume that the assets' returns are trapezoidal fuzzy numbers, and transform the portfolio return into perceived value that reflects investors' psychological characteristics according to the value function in prospect theory. A fuzzy portfolio optimization model considering the psychological characteristics is established with the objectives of maximizing the possibilistic mean value and minimizing the lower possibilistic semi-variance of the perceived value. Furthermore, in order to solve the proposed model effectively, we design a multiple population genetic algorithm. Finally, we conduct numerical analysis to illustrate the effectiveness of the model and the algorithm. The results show that the multiple population genetic algorithm designed solves the proposed model more effectively than the traditional genetic algorithm, and the fuzzy portfolio optimization model considering the psychological characteristics improves investors' satisfaction level, and thus can provide decision support for actual investment activities.