Abstract: Currently, highway congestion is a major challenge to the normal operation of the transportation system, and the congestion early warning method can provide accurate congestion prediction information for the traffic management department, enabling timely measures to be taken to mitigate traffic congestion. Therefore, a highway congestion early warning method based on a macro traffic flow model is proposed, which is applicable under limited data quality conditions. Firstly, the historical traffic flow data are analyzed in terms of spatio-temporal characteristics to grasp the time-varying pattern of congestion; subsequently, the three classical traffic flow models of Underwood, Pipes, and Van Aerde are considered, and the parameters are fitted using the least-squares method, and the degree of fit is described by using the indicators of RMSE, MAE, MRE, and R², and the model with the highest fitting accuracy is selected to identify the key traffic flow of parameters. Finally, the parameter values corresponding to the congestion warning points are obtained according to the change of the slope of the basic diagram, and the congestion warning is issued when the traffic flow parameters change to the warning value. The congestion warning method is tested by using the measured data of Guangxi Expressway, and the results show that the congestion warning can be issued in time 15-40 min before the moment of congestion, which indicates that the method can effectively analyze the trend of traffic flow changes and accurately issue congestion warning and helps to provide timely congestion prediction information for the traffic management department to reduce the occurrence and persistence of congestion.