This paper establishes a mathematic model about the cost of dispatch and regulation of AGC units, uses a plant growth simulation algorithm with genetic operators, and applies the algorithm to the research of economical unit dispatch. In order to tackle the problem of low speed in full-scale search while having full simulation of plant growth, it adopts the method of reserving the best and reserving the better after the hybridization of the worst, which helps to accelerate the calculation speed while restraining precocity and avoiding from local optimum. In addition, to prevent excessive aggregation of growing points, it applies K-means clustering algorithm to classify for the units, and then chooses the variation of bigger otherness for the units of the same kind when growing new growing points. As for the multiple optimal results, it distributes equally for the unit to ensure the equality of units dispatch. By comparing the calculation results, the rationality and validity of this algorithm are verified.