Based on the 2017-2018 high-resolution numerical forecast products of China Meteorological Administration (CMA),real-time urban forecast products of Gansu Province and data of national ground-based observation stations,the intelligent grid high and low temperature objective correction products of Gansu Province are developed by using wavelet analysis,sliding training,optimal fusion and other technologies.The test results show that the three correction products (urban forecast products,sliding training correction products and optimal fusion products) have the ability to correct CMA forecast,and the maximum temperature correction ability of the three objective correction products is stronger than the minimum temperature correction ability.The prediction results of the high and low temperature correction products produced by the sliding training method and the optimal fusion method are better than those of the model objective prediction in the areas with obvious systematic errors (Gannan,Longnan,etc.),while the high and low temperature urban prediction products have advantages in the areas with strong temperature localization or poor model objective prediction ability.The prediction ability of the high and low temperature products generated by the optimal fusion prediction method is slightly higher than that of the sliding training correction products,and is basically the same as that of the urban prediction products produced by the existing forecasters,which initially has the ability to replace the subjective prediction.