Under the background of power system reform, it is necessary to mine the user electricity characteristic and develop corresponding packages to optimize the demand response strategy considering the problem of the deviation assessment of load server entities. Firstly, the user’s power consumption curve based on selfcoding neural network and fuzzy C means clustering method is classified. Then an optimization model of time of use electricity price based on the user response model of consumer psychology is established. Furtherly, a superimposed electricity price model for the peak period is established. Research shows that the development of package can effectively guide users to adjust their electricity consumption behavior, so as to reduce the difference between electricity consumption modes. From the perspective of deviation assessment, it is helpful to develop monthly electricity purchase strategy.