The invention discloses a load modeling and optimal control method based on demand response. The method comprises the steps that loads are classified into rigid loads and flexible loads in a decoupling mode, load models are established, further, according to the participating demand response, the flexible loads are classified into three types which are an electrovalence item, a direct load control item and an interruptible load item, a electrovalence model, a direct load control model and an interruptible load model are established in a classified mode, and therefore powerful data resources are provided for following load optimization through establishment of the models; according to the maximum reduction load which can be provided by a controlled variable and a relation among the maximum reduction load of the direct load control item, the maximum reduction load of the interruptible load item and the maximum value of a power supply gap, optimal control models of the three conditions are established, solutions are obtained through a genetic algorithm, and therefore optimal execution results are obtained. Thus, more resources and strategies are provided for actual operation of an electric power system.