By studying the classical betweenness and the universal degree methods, we put forward a new model to control the spread of cascading failure on scale-free networks. The new model is based on defining the load of an edge with respect to the betweenness centrality of the two connected nodes. The iterative process of a cascading failure on scale-free networks is analysed by removing one edge.We find that the proposed new model can control the spread of cascading failure more significantly. To make our conclusions more convincing, we have explored the performance of new models in real network by the power grid of the western United States. We further give the following reasonable explanations: First, the reason why the new model shows a more stable performance than the others has been explained. Secondly, we have explored the reason why the new model shows different advantages depending on the load for different networks and lastly, we have studied the exact difference between these two methods and the network structure. This paper might be useful for preventing and mitigating the cascading failure in real life.