The objective of this paper is to evolve simple and effective methods for the optimal dispatch problem in thermal units, which are capable of obtaining optimal power generations for a large-scale system. In an improved λ-based genetic algorithm (GA), the encoding parameter is normalized system incremental cost λnm. This approach takes network losses, ramp rate limits, and prohibited zone(s) into account because of GA’s flexibility. The features of a refined genetic algorithm are also incorporated in the λ-based GA approach. The probability of mutation and crossover is exponentially varied in each generation. Probability of limits has been fixed for mutation and crossover. The results of the λ-based GA and improved λ-based GA are compared. The latter is simple and effective. In the proposed improved λ-based GA, an optimal solution is obtained neglecting losses in the first stage. The resulting solution is given to load flow, and the network transmission loss is computed from load flow in the next stage. Th...