The invention discloses an NRIET rainstorm intelligent similar analysis method. In a meteorological big data similar forecasting method, a similar criterion is obtained by means of machine learning, asimilar calculation scheme can be established more accurately, meanwhile, similar deviation is used as a measurement standard, such that not only numerical similarity of elements is considered, but deformation factors of a weather system are also added. In addition, the level of similarity of different variable fields in the weather process is quantitatively measured by combining image big data similarity level, element fields are gradually introduced for similar filtering, and finally the dates of similar cases are obtained, and the historical live scenes corresponding to the similar cases are taken as a basis for making a weather forecast. According to the method, a graph big data technology and a traditional similar algorithm are combined, the atmospheric circulation evolution trend provided by a numerical forecast product is utilized, the three-dimensional structural features of the weather development process are considered, the linear and nonlinear change rules of the atmosphereare considered, and the method has the characteristics of clear idea and strong intuition and is used for assisting in forecast and analysis.