In this paper, a neural network (NN)-based adaptive dynamic programming (ADP) algorithm is employed to solve the optimal temperature control problem in the water-gas shift (WGS) process. Since the WGS process has characteristics of nonlinearity, multi-input, time-delay and strong dynamic coupling, it is very difficult to establish a precise model and achieve optimal temperature control using traditional control methods. We develop an NN model of the conversion furnace using data gathered from the WGS process, and then establish an NN controller based on dual heuristic dynamic programming (DHP) to optimize the temperature control in the WGS. Simulation results demonstrate the effectiveness of the neuro-controller.