Medium and long-term load forecasting is a necessary prerequisite for distribution network planning, and is of great significance for the improvement of economic and social benefits of the power system. Aiming at the defects of low prediction accuracy and poor applicability of traditional grey forecasting model, a medium and long-term load forecasting method of power system based on Markov residual correction and improved grey theory is proposed. Based on the classical grey forecasting GM(1,1) model, this method first divides the state according to the relative residuals, calculates the Markov residual transition matrix, and then uses the equal-dimensional innovation method to construct the grey Markov correction forecast. Model, and finally use residual processing method to revise the prediction results. The simulation results based on an actual example of load forecasting of annual power consumption in Guangxi Province show that compared with the traditional grey forecasting model, the grey Markov residual correction model proposed in this paper has significant advantages in forecasting accuracy and applicability.