A load curve convergence optimal inflection point recognition method includes the following steps: S1, data collection; S2, data pretreatment including S2-1 data filling and S2-2 data correction through a wavelet de-noising method; S3, achievement of approximate fitting of a load curve based on a data pretreatment technique; and S4, performing mathematical treatment on the obtained fitting daily load curve to determine an optimal inflection point. After pretreatment is performed on the original data through data filling and the wavelet de-noising method, Matlab software is called to perform approximate fitting on polynomial, and in this way, a smooth continuous function format is obtained; an inflection point definition of the mathematics field is applied to the power load curve, the inflection point of the load curve can be rapidly found, and then power dispatching workers can timely determine the present inflection point and can make ready for corresponding dispatching works, and the stability and the reliability of power grid operation can be improved.