Abstract: With the gradual formation of sales side market competition pattern, the electricity companies can improve the safety level of power grid and the quality of power supply through load combinatorial optimization to improve the load rate and reduce the cost of electricity purchase on the users side, and to improve the equipment utilization rate and reduce the line loss on the power supply side. In order to integrate and optimize user resources according to the complementarity and diversity of power load, a user load combinatorial optimization model is proposed respectively for the scenarios of maximum purchase and sale benefits, maximum electricity transaction, and maximum comprehensive load rate based on the analysis of user load characteristics. And then, the greedy search algorithm is improved by introducing random factors to remedy the defects of the algorithm in determination of the initial estimation and optimization effects, thus improving the quality and accuracy of the algorithm. Finally, the improved greedy search algorithm is used to solve the load combinatorial optimization model under different scenarios. The improved algorithm has been verified to have stronger global search ability and adaptability in solving the combinatorial optimization problems.