In order to solve the problems in real systems where inputs and outputs are time-varied continuous functions, a process neural network model based on expansion of base functions is proposed in this paper. In this model, the continuous input-output mapping of the system is realized by nonlinear mapping capability to the time variable of process neural networks. A learning algorithm is also given in this paper. In order to simplify the algorithm, orthogonal functions are selected as base functions, and the effectiveness of the model and the algorithm is proved by simulation of oil reservoir exploitation.