Crop evapotranspiration (ETCML) is usually divided into two processes: crop transpiration (T) and soil evaporation (E). The transpiration process, which is influenced by crop characteristics, meteorological and environmental factors and field management measures, is the key to study the change of crop water demand and water conversion in farmland. Soil evaporation is also an important part of the water cycle in farmland and plays a very important role in the calculation of water balance and energy balance of farmland, however, the research on evapotranspiration and its composition under the conditions of water-saving irrigation is very few. An experiment was carried out from 2015 to 2016 to explore the characteristics of water consumption under the conditions of water-saving irrigation, and to obtain a deeper understanding of the composition of evapotranspiration in water-saving irrigated rice fields. Daily/seasonal variation characteristics of evapotranspiration, evaporation and transpiration, and the change of the ratio of E or T to ETCML during typical weather and each growth period were analyzed by using a micro-lysimeter system. The results illustrated that ETCML, E and T indicated significant daily variation characteristics in different kinds of typical weather. On typical sunny days, changes of ETCML and T showed basically the same trend and they both showed an inverted “U” curve, while E was mostly influenced by crop growth stage and environmental factors. In addition, because water vapor was condensed on the surface of crops as well as soil, ETCML showed positive and negative alternation at night. ETCML and T on cloudy days were demonstrated as unstable trend with multi-peak variation, while E had no obvious daily variation characteristics. The micro-lysimeter was sensitive to the weight changes produced by rainfall, which caused inaccurate measurement data. In consequence, it was important to exclude the data when it was raining heavily. The daily variation of ETCML and T stood out an inverted “U” curve in all rice growth stages, while E showed an inverted “U” only in the early stage, and the cumulative daily value of E was less than 1.00mm/d at the end of the growth stage. After the middle tillering stage, T became the decisive factor of rice evapotranspiration and determined the magnitude and pattern of ETCML. The seasonal variation of T and ETCML was increased and then decreased with Rn, and reached the maximum at the middle or late tillering stage, with obvious phenological characteristics. The proportion of E and T to ETCML was restricted by each other. The proportion of T to ETCML was increased gradually with the growth of rice, reaching a maximum at the milk ripening stage and decreasing slightly at the yellow ripening stage. The proportion of E was decreased gradually from about 0.95 at the beginning of rice growth to an average of about 0.24 at the end of rice growth period, with the average of 0.37 in rice season. The proportion of E to ETCML in paddy fields under irrigation conditions could be accurately expressed by its logarithmic relationship with rice transplanting days and leaf area index (LAI). The study of evapotranspiration process characteristics had important practical significance for both field water management and irrigation guidance.