The invention discloses a high-quality electric power customer data mining method based on hierarchical data envelopment analysis. According to the high-quality electric power customer data mining method, a comprehensive high-quality electric power customer evaluation model is established based on a hierarchical analysis method and a data envelopment analysis method; the model comprises three hierarchies; a first hierarchy is a target layer in which the top-quality electric power customers are evaluated according to the requirements of a power grid company; a second hierarchy is a criterion layer in which various criterions are set from the view of economic and stable operation of a power grid, and the criterions comprise the economical efficiency, the reliability and the contribution rate; a third hierarchy is an index layer in which various indexes are set from the view of load property and the like of electric power customers, and the indexes comprise the power purchase amount, the power use scale, the load fluctuation level, the average load rate, the electric charge recovery rate, the electric power purchase increase amount and the load curve similarity coefficient. According to the requirements of the power grid company, different criterion layers, index layers and weight coefficients are set to screen high-quality electric power customers, and the method is significant for operation of the power grid and policy making.