Abstract: In this paper， the Markov probability matrix was introduced to reduce the influence of the high fluctuation data columns on the GM（1，1） model to improve its prediction accuracy. Moreover，the monthly data of the residential electricity consumption in Suzhou from January 2015 to October 2020 were selected as the model's training data. The monthly data of the residential electricity consumption from November 2020 to March 2021 were used as the test data to measure the model's prediction accuracy and forecast the residential electricity consumption in April 2021. The empirical results showed that the average relative error of the GM（1，1）model was 24. 70%，while the average relative error of the GM（1，1） model optimized by the Markov chain was 11. 62%. The prediction performance superiority of the GM（1，1） model optimized by the Markov chain over the traditional GM（1，1）model is demonstrated in this paper.

Key words: prediction on the residential electricity consumption, grey GM（1，1） model, Markov chain