Abstract:Taking the contradictions between load shaving and new energy output encountered in the actual operation of Xinjiang power grid as a starting point,and taking the peaking capability and power balance constraints as the main means,the optimal solutions for the minimum start - up mode and the maximum new energy output to satisfy the requirements of peak shaving in Xinjiang power grid are determined through the prediction methods based on extreme learning machine ( ELM) to achieve the economic dispatch and intelligent scheduling.