A multi-microgrid in an Active Distribution Network (ADN) will influence the economy and reliability of an ADN system. When the traditional unified optimal dispatching method of a multi-microgrid and the whole ADN network is adopted, it is difficult to calculate the optimal power flow of the system efficiently and accurately once the wind power output in the microgrid fluctuates, and may even cause no solution being obtained. Therefore, an optimal scheduling model for a grid-connected multi-microgrid based on bi-level programming is proposed. Each microgrid in the upper model is incorporated into the ADN as a power source. The optimal power flow model is established with the ADN system power flow balance as the constraint. Using the second-order cone relaxation technique, the non-convex non-linear power flow model is transformed into a second-order cone programming model for the convex feasible region and the solution is obtained using the Gurobi solver. The lower model takes the power of the line optimized from the upper model as the constraint, sets up within the MG interconnection of a controllable power supply scheduling model, and uses the combination of Tent map chaotic and NDX cross technology improved Genetic Algorithm (GA) to solve the problem. The feasibility of the algorithm is verified by the example of the adjustment IEEE-33 nodes system incorporated into the multi-microgrid. The simulation results show that the scheduling model and algorithm of bi-level programming are feasible and the ADN system with multi-microgrid is more economical. When the wind/sun output fluctuates, the lower model can still be locally adjusted and optimized to reduce the influence of microgrid fluctuation on the ADN system and improve the reliability and robustness of the system. This work is supported by National Natural Science Foundation of China (No. 61364027) and Natural Science Foundation of Guangxi (No. 2019GXNSFAA185011).