Abstract: The problem of grid workflow technology.In traditional grid in assigned process, when multitasking competition, because the task of pathskilling transmission unbalanced load slow problem.This paper presents a genetic algorithm-based load balancing algorithm for grid work flow.By simulating the natural process of biological evolution were randomized to the task of the search space, according to the fitness function of scheduled tasks, and parallel with the global search for ways to find the best node, method is to avoid the traditional balance of the implementation of successive matching problems caused by inefficient.Experimental results show that the new algorithm is fast load balancing, grid workflow system to improve the work efficiency, and achieved satisfactory results.