Abstract: The traditional grey prediction model has the defect of low prediction accuracy. Based on the GM (1,1) model in the grey system, this paper analyzes the limitations of the model and gives solutions to improve the accuracy of load forecasting. Aiming at the problem that there may be abnormal data in the original data, which leads to the low accuracy of the prediction results, this paper adopts the method of smoothing the original data to reduce the influence of abnormal data on the prediction results. Considering that the initial value of the traditional GM (1,1) model usually selects the data of the earliest one year in the historical data, which is not closely related to the future and has low regularity, this paper uses the data of the latest one year in the historical data as the initial value of the prediction model. The residual processing method is used to modify the original model to improve the prediction accuracy. Considering the above three improvement methods, the GM (1,1) model of the traditional system is improved. The simulation results based on a practical example in a certain area show that compared with the traditional grey prediction model, the prediction accuracy of the improved prediction model is greatly improved, which verifies the effectiveness of the improvement measures.

Key words: distribution network planning, prediction model, load forecasting, grey model