Abstract: In order to improve the optimal allocation of renewable energy in microgrid and reduce the

uncertainty of source load in microgrid，a multi-objective coordinated scheduling optimization method for microgrid based on Pearson correlation coefficient meta-learning（PCC-ML）source load prediction was proposed. Firstly，Pearson correlation analysis was used to analyze the time series composed of original multidimensional input variables，and meta-learning（ML）was used to manage the five processes of feature extraction，candidate model preparation，labeling，off-line training and online prediction result evaluation for microgrid source load data. Then，based on the prediction results，a two-stage rolling scheduling optimization model of microgrid was established. The first stage is day-ahead pre-scheduling stage，and the optimization goal is to achieve the global economic optimization of microgrid region. The second stage is the real-time operation and regulation stage. Considering the uncertainties of the real-time output of new energy sources in microgrid，the prediction deviation is regulated in real time to realize the optimal operation cost of each equipment in microgrid. Then，the column constraint generation algorithm（C&CG）was used to decompose the main and sub-problems into interactive iterations to solve the two-stage optimization model. Finally，an example shows that the proposed method can meet the performance requirements of visible and invisible forecasting tasks and improve the economic benefits of microgrid.

      Key words: micro-grid；source load prediction；meta-learning（ML）；scheduling optimization；column constraint generation algorithm（C&CG）