In the real-time pricing process of a smart grid, different utility functions were adopted for household residential users, large industrial users and commercial users respectively in order to improve the limitation caused by that for all the users only a single utility function was condisered in the past. The consumption preference of different users was simulated with multi-utility functions, and the social welfare maximization model considering multi-user coexistence was studied. According to the principle that the social welfare can be maximized only by balancing supply and demand in the market, using the KKT condition, the optimal power generation on the power demand side and the optimal power consumption on the power demand side were calculated respectively, and then a corresponding distributed real-time electricity price updating algorithm was put forward. The privacy of the user was guaranteed. A numerical simulation shows that, comparing with the fixed electricity price algorithm, the better efficiency and convergency of the distributed real-time electricity price algorithm can be proved.