This paper develops a novel energy trading system with a community energy storage (CES) device for demandside load management within a neighborhood area network. The energy users in the proposed system that have their own photovoltaic power generation are allowed to trade energy from their personal surplus with the grid and the CES device. We adopt a dynamic noncooperative repeated game with Paretoefficient pure strategies as the decentralized approach for the users to determine optimal energy trading amounts for the next day. This decentralized model needs minimal information exchange or communication between users. Simulation results show that our system is able to provide peak load leveling for the grid, while providing financial benefits to users. Moreover, the performance benefits of our system are robust to both inaccuracy in day-ahead power forecasts and CES battery inefficiencies.