Traffic flow simulation is used for prior evaluations to solve traffic jam problems. In this paper, we propose a fast and precise traffic flow simulation method to examine various measures repeatedly. By using the Payne-Whitham model, the traffic flow simulation model based on continuum approximation, it is solved at high speed even in a section with a large amount of traffic flow. A model based on continuum approximation makes it possible to simulate traffic flow over a wide area network, but it is generally considered to be less precise because it cannot express individual vehicle characteristics and interactions. We construct a macroscopic traffic flow model so as to reproduce a microscopic traffic flow model such as multi-agent traffic flow model. The multi-agent model has already been verified to be consistent with actual traffic flow. For this purpose, we use parameter identification through a neural network to compensate the defect and to verify actual traffic flow.