Wind power, photovoltaic power generation and load are with natures of volatility and intermittent in the temporal and spatial distribution. This paper fits wind power prediction error distribution according to wind speed segment; photovoltaic output distribution is fitted using Beta distribution; and load distribution is fitted by scheduling time. Depending on wind speed and scheduling time, by convolution generalization method, give the distribution function of the event which the constraining condition is satisfied under each scenario. The multi-target micro-grid chance constrained dynamic scheduling model is built, with the minimum of micro-grid operation and maintenance costs and environmental costs. Then make the solutions of the kinds of controllable resources closed to globally optimal solution by improved multi-objective modified teaching-learning algorithm. What’s more, the simulation results verify the effectiveness of the method. Monte Carlo simulation method is found that the posterior confidence level of the multi scene chance constrained scheduling is higher, and it further ensures the security of the system and improves the robustness of the strategy.