Power load prediction method and device based on machine learning algorithm, and medium

The invention provides a power load prediction method and device based on a machine learning algorithm and a medium, and belongs to the technical field of load prediction, and the method specifically comprises the steps: dividing user groups into an industrial load user group, a commercial load user group and a resident load user group based on different load types of the user groups, constructing an input set based on the date type, the temperature, the humidity and the wind speed of the predicted date, and reconstructing the input set according to different user groups to obtain an industrial reconstruction input set, a commercial reconstruction input set and a resident reconstruction input set; industrial load prediction data, commercial load prediction data and resident load prediction data are obtained by adopting a load prediction model based on a machine learning algorithm, and a power load prediction result is obtained by combining an industrial fluctuation coefficient, a commercial fluctuation coeffi