The Northeast China maize belt is one of the three major golden maize belts in the world and has been severely affected by climate change, however, the evapotranspiration (ET) partitioning is not clear. It is important to study ET and its components under climate change. In this paper, the water balance model SOILWAT2 was used to estimate ET partitioning in drought and humid years, seasons, and maize growth stages from 1989 to 2018 over rainfed maize farmland. The results indicated that the SOILWAT2 model performed well for the prediction of ET and its partitioning compared with eddy covariance measurements. The mean yearly ET, transpiration (T), soil evaporation (Es), and canopy interception evaporation (Int) were 432.3 mm, 197.6 mm, 204.7 mm and 19.2 mm, respectively, over 30 years. Es/ET was 6.3% lower in drought years than in humid years, T/ET was conversely higher (6.2% higher in drought years). There was no clear difference of Int/ET between humid and drought years. In the growing season, T/ET, Es/ET, and Int/ET varied from 40.0% to 75.0%, 22.8% to 55.7%, and 0.7% to 7.0%, respectively. T/ET decreased along with the growth of maize and was greatest at the greening–jointing stage. Es/ET was smallest at the greening–jointing stage. We found a power function relationship between T/ET, Es/ET, and leaf area index (LAI) and above-ground biomass. Our results indicated that for the rainfed farmland, drought may limit maize yield by increasing water loss of maize through increasing T under climate change conditions. Therefore, securing food yield will depend on increases in water-use efficiency and other adaptive strategies, such as drought-resistant varieties, and irrigation.