In order to overcome the shortcomings of the traditional similar day method that each factor and its weight need to be manually assigned, this paper establishes a similar day model based on intelligent optimization method to predict daily electricity consumption by fully considering several main conventional influencing factors such as date type, date distance and meteorological factors. All the parameters in the similarity calculation formula that used to be manually assigned are trained from historical data by using fruit fly optimization algorithm. The parameter values can be dynamically adjusted according to the load change characteristics of specific users, which enhances the accuracy and versatility of the similar day method. In order to solve the multi-dimensional optimization problem of parameter training and avoid the algorithm from falling into local extremum, this paper proposes an improved fruit fly optimization algorithm that introduces the concept of multiple groups to enhance the algorithm’s global searching ability. The simulation example shows that compared with the traditional similar day model, the prediction accuracy of the similar day model based on intelligent optimization method has been significantly improved. This work is supported by the Key Research and Development Program of Shandong Province (No. 2019JZZY020804).