Abstract: Because annual city electricity demand is influenced by such factors as economy, society and climate, it possesses certain grey characteristics. As to the big forecasting errors of traditional GM (1, 1), the direct grey model (DGM) is introduced in city electricity demand forecasting, which can avoid selecting the background value when identifying the parameters of grey differential equation. To minimize the average relative error between restored value and real value of DGM, optimize model is built and the algorithm based on differential evolution is proposed. Annual electricity demand of 5 cities with different increasing rates is forecasted, and is compared with that forecasted by other methods. The results show the practicability of the proposed method.