This paper is concerned with using multi-objective optimization by learning automata (MOLA) for economic emission dispatching in the environment where wind power and loads vary. With its capabilities of sequentially dimensional search and state memory, MOLA is able to find accurate solutions while satisfying two objectives: fuel cost coupled with environmental emission and voltage stability. Its searching quality and efficiency are measured using the hypervolume indicator for investigating the quality of Pareto front, and demonstrated by tracking the dispatch solutions under significant variations of wind power and load demand. The simulation studies are carried out on the modified midwestern American electric power system and the IEEE 118-bus test system, in which wind power penetration and load variations present. Evaluated on these two power systems, MOLA is fully compared with multi-objective evolutionary algorithm based on decomposition (MOEA/D) and non-dominated sorting genetic algorithm II (NSGA-II). The simulation results have shown the superiority of MOLA over NAGA-II and MOEA/D, as it is able to obtain more accurate and widely distributed Pareto fronts. In the dynamic environment where the operation condition of both wind speed and load demand varies, MOLA outperforms the other two algorithms, with respect to the tracking ability and accuracy of the solutions.