The most of the recent models of directed weighted network evolution capture the growth process based on two conventional assumptions: constant average degree assumption and slowly growing diameter assumption. Such evolution models cannot fully support and reflect the dense power law and diameter shrinkage in the process of evolution of real networks. In this paper, a new evolution model, called BBVd, is proposed for directed weighted networks by extending BBV model with the idea of the Forest Fire model. In BBVd, new directed edges are established with probabilities computed based on in/our-strength of nodes, with dynamical evolution of weights for local directed edges. The experimental result shows that the generated networks using BBVd display power-law behavior for the node strength distributions, and moreover, it satisfies the densification power laws and has shrinking diameter.