For rectangular Li-ion batteries, this paper optimized the heat dissipation of the cold plate embedded with leaf-like network according to the structural theory. Not only the distributed network layer was established, but also the collection network layer was designed to mitigate the inherent temperature gradient problem of the conventional single channel. The effects on the split point of symmetrical/asymmetrical bifurcation, the width ratio of symmetrical bifurcation, the inlet mass flow rate, the number of feed-out channels and the width of main channel on the thermal performance of cooling plate were numerically simulated. The results demonstrated that the heat dissipation was better when the split point is near the limit position of the outer channel. Compared with the offset bifurcation network, the symmetric bifurcation network can achieve better heat transfer characteristics, but its pressure drop was slightly higher than that of offset bifurcation network. The increase of width ratio and inlet mass flow can meet both maximum temperature and temperature uniformity objectives, but at the expense of pressure drop. Compared with the single feed-out channel, the double feed-out channels can meet both the thermal performance and the pressure drop requirements. As the width of the main channel (D-1) was elevated, pressure drop greatly declined on the basis of heat transfer. The simulation results can provide a reference for the design of cold plate embedded fractal-like network based on battery thermal management systems. (C) 2019 Elsevier Ltd. All rights reserved.