Abstract: The paper analyzes two main types of irrigation control which are based on sensing soil moisture or calculating water requirement, and discusses their application to micro-irrigation.As a synthesis of them, a new micro-irrigation control model with self learning ability is proposed.This model enhances calculating precision, because it eliminates the open loop control error, excludes the difficulties arising from rain and deep drainage, and facilitates evaluation of the crop coef ficient.A greenhouse status coefficient is introduced for calculating the water requirement of greenhouse irrigation.Though soil moisture sensors are adopted, their distribution density is quite low, and the control failure arising from lag of measure is avoided.The new model shows a large-scale application prospect.