With the popularization of clean renewable energy sources

the permeability of distributed power sources is increasing. Firstly

in order to improve the utilization rate of clean energy and reduce the cost of micro-grid operation

we established a double-layer wind power prediction error model based on comprehensive consideration of time-of-use price and operating characteristics of different types sources such as wind power

photo-voltaic

thermal power and tie line. Secondly

aiming at minimizing the economic cost

the environmental cost

and the asynchronous degree of output

we established a micro-grid collaborative optimization model in which the wind power forecast uncertainty and valve point effect are taken into account. Thirdly

to solve the established multi-objective optimization model

we proposed an improved intelligent optimization algorithm combined by co-evolution theory and non-dominated sorting genetic algorithm（Co NSGA-Ⅱ）. In this algorithm

a variety of groups in the non-dominated sorting genetic algorithm are employed to compete and correct with each other

therefore

the proposed algorithm can combine the excellent convergence of the non-dominated sorting genetic algorithm and the powerful searching ability of the co-evolutionary algorithm. Finally

a practical micro-grid system in Northwest China was simulated as a case study

and the improved algorithm was compared with the traditional fast non-dominated algorithm. The simulation results demonstrate the superiority of the global search performance and the fast convergence performance of the improved algorithm.