A dynamic multivariate partial grey model based on the traffic flow parameter equation and its application

Accurate real-time traffic flow prediction can provide effective information for ITS, thus improving the efficiency of traffic management. This paper establishes a dynamic multivariate partial grey model based on the parametric equations of traffic flow in response to existing partial grey model studies, which are basically univariate in traffic flow, and addresses the existing problem in which the traffic flow is closely linked to the traffic speed and density. The model explores the dynamic relationship between traffic flow and speed in depth, captures the details and characteristics of traffic flow data, uses the input data in the form of a matrix to fully consider the spatial and temporal characteristics of traffic flow data, improves the stability of the model and the prediction accuracy, and broadens the modelling structure and the scope of application of grey models. The new model was applied to simulate and predict short-term traffic flow on roads in the UK. By utilizing the periodicity and spatiotemporal nature of traffic flow, data with cycles of weeks and days were selected to dynamically demonstrate the effectiveness of the model. The results revealed that the simulation error of the new model was less than 6 % and that the fitting effect was better than those of the other six prediction models, indicating that the model has good stability and can effectively predict short-term traffic flow. Moreover, the model with the best fitting effect was used to predict the traffic flow in the same period in the future, and the predicted results were consistent with the trend of the original data.