Abstract: According to failure of the optimum because of varying environment,a method of various alternatives for coal blending optimization is proposed.A novel swarm intelligence algorithm-multivariant optimization algorithm the method being adopted is applied.The search atoms of the algorithm were stored in a upper triangular data structure form that was providing memory and information sharing.Then the search atoms made full use of optimization process information,and implemented process memorise.At last the global optimal solution could be found meanwhile various alternatives retention.A case study was performed by taking a thermal power plant as an engineering background.By comparing the optimization results of the algorithm with other five swam intelligence algorithms,it was proved that multivariant optimization algorithm provided various alternatives and had the advantages in the solutions’ quality for coal blending optimization.It was effective to coal blending optimization using multivariant optimization algorithm.The method of various alternatives for coal blending optimization with integrated performance index has important practical significance for thermal power plants.