Abstract: Monthly coal power demand forecasting is important for guiding the development of coal power and securing reliable energy supply under carbon peak ＆ carbon neutrality goals,but the monthly coal power demand variation is non-stationary and non-linear. In order to accurately forecast future coal power demand,based on the idea of decomposition-integration,improved singular spectrum analysis（ISSA）is introduced to decompose and reconstruct the original demand series to obtain several sub-series with different frequencies,and sparrow search algorithm（SSA）is applied to optimize the extreme learning machine（ELM）model to forecast each sub-series,and then superimposes them to obtain the final coal power demand forecast. Taking Jiangsu province as an example,the proposed method is compared with EMD-SSA-ELM model based on the ensemble empirical modal decomposition（EMD）and SSAELM model without decomposition,and the results show that the proposed method can effectively remove the influence of noise components with lower error values,and the average absolute percentage error is 8.0% and 17.6% lower than that of EMD-SSA-ELM and SSA-ELM respectively,with higher prediction accuracy and better applicability.