The power generation of prosumers in peer-to-peer (P2P) transactions has high volatility and uncertainty. Also the real-time power generation often deviates from the day-ahead generation scheduling. Given the real-time imbalance problem of prosumers in P2P transactions using the continuous double auction (CDA) mechanism, this paper proposes a local energy balance strategy considering the timing characteristics of P2P transactions. The system operator (SO) uses model predictive control in the real-time balancing market. A local energy balancing strategy model based on the coordinated operation mechanism of the day-ahead P2P energy market and the real-time balancing market is developed. The model aims to minimize the overall balancing cost and considers the cost-benefit constraints of the SO and balancing service provider. In the real-time balancing market, the compensation fee charged by the SO to the responsible party is equivalent to purchasing flexible balanced services from the service provider. Simulation results show that the balancing service provider balances 78% of the localized real-time unbalanced electricity. Compared with the grid balancing strategy, the local real-time balancing market cost is one third of the grid balancing cost. At the same time, the strategy proposed in this paper guarantees the revenue of balancing service providers and SOs, and promotes PV energy consumption. This will be beneficial for the long-term development of the P2P trading market.