A sudden load change may bring a sudden change to operating state of a power system and has serious impacts on the bad data identification in the state estimation with predicting functions. Considering engineering practices， an innovation graph approach based identification method for sudden load change is further studied. The concepts of sudden change route and sudden change subnet are proposed and the identification method for sudden load change is presented. After successful identification of the suspicious nodes with large injecting innovation values， sudden change routes can be found among the nodes. These sudden change routes can make up a sudden change subnet and all the sudden change nodes can be identified in cases when multi generators are supplying a sudden change load. The nodes injection innovation and branches innovation difference can be used for cross validation to identify the bad data occurring simultaneously with the sudden load changes. The IEEE 30-bus system example is shown to explain the identification process by sudden change subnet. This method has also been validated in practice.