可再生能源出力的不确定性为微电网的优化调度带来了重大挑战。同时，传统的优化方法和调度时间尺度过于单一，导致调度结果存在较大误差，从而难以确保系统运行的可靠性与经济性。针对上述问题，提出了一种基于K-近邻（K-nearest neighbor，K-NN）算法、变模态分解（variational mode decomposition，VMD）、卷积神经网络（convolutional neural network，CNN）以及双向长短期记忆（bidirectional long short-term memory， BiLSTM）神经网络的微电网两阶段优化运行策略。首先，构建了基于K-近邻算法和混合BiLSTM功率预测模型，为两阶段优化调度模型提供准确的风光发电预测数据。其次，建立了两阶段优化调度模型。在日前调度阶段，引入阶梯式碳交易机制和激励型需求响应，以最小化系统总运行成本为目标制定日前调度计划；在日内调度阶段，则采用基于模型预测控制的方法，实现日内滚动优化调度策略，以调整量最小为目标对日前调度计划进行动态修正，从而降低因预测误差引起的功率波动。最后，以某微电网为例进行了仿真分析，结果表明：该方法不仅有效提高了预测精确性，同时也提升了微电网的经济性、环保性及稳定性。

The uncertainty of renewable energy output poses significant challenges to the optimization and scheduling of microgrids. At the same time， traditional optimization methods and scheduling time scales are too single， resulting in large errors in scheduling results， making it difficult to ensure the reliability and economy of system operation. A two-stage optimization operation strategy for microgrids based on K-nearest neighbor （K-NN） algorithm， variational mode decomposition （VMD）， convolutional neural network （CNN）， and bidirectional long short-term memory （BiLSTM） neural network is proposed to address the above issues. Firstly， a power prediction model based on K-nearest neighbor algorithm and hybrid BiLSTM neural network is established to provide accurate wind and solar prediction data for the two-stage optimization scheduling model. Secondly， a two-stage optimal scheduling model is established. In the day ahead scheduling phase， a stepped carbon trading mechanism and incentive demand response are introduced to develop a day ahead scheduling plan with the goal of minimizing the total operating cost of the system； In the intra day scheduling phase， an intra day rolling optimal scheduling strategy based on model predictive control is established to achieve rolling correction of the intra day scheduling plan with the goal of minimizing the adjustment of the intra day scheduling plan， and reduce the power fluctuation caused by the prediction error. Finally， taking a microgrid as an example for simulation analysis， the results show that the proposed method effectively improves the prediction accuracy while enhancing the economic， environmental， and stability of the microgrid.