The invention relates to a method for predicting lightning motion by space density clustering. The method comprises the following steps of: processing automatic lightning monitoring data through a computer, establishing a clustering analysis database, combining a geographic information system (GIS) to establish an analysis platform, performing clustering analysis on lightning data at different moments by establishing a density clustering algorithm model, setting lightning data sample screening conditions in the database, describing basic parameters of lightning density to acquire lightning sample data according to a time sequence and classifying according to a time period; establishing a lightning data space position relationship table through lightning data points and distance values; determining position characteristic points of different lightning storms; and calculating motion direction and speed along with displacement change during time lapse according to the position characteristic points of the lightning storms, and predicting positions of lightning storms to be generated in the next time period. The invention has the advantages that: the method has novel conception, a method for processing the lightning data is advanced and reasonable, and the authenticity of the lightning motion can be reflected.