Protecting Agent could adopt the current protection based on neural network as the literature [4] narrated. The protection is composed of two neural sub-networks: ANN1(sub-system) is positive-direction judgment and failure type selection, ANN1(sub-system) whose judgment range can extend to adjacent route; ANN2(sub-system 2) is the failure instantaneous action ANN1(subsystem) within the range of 85% protection area. When ANN1 (sub-system) and ANN2 (sub-system 2) act at the same time, trip of Agent is activated and instantly jump the breaker in the side, meanwhile send the message to organize of Agent; When ANN1 (sub-system) acts and receive the trip command sent by organize of Agent, the trip of Agent is equally activated and instantly jump the breaker in the side. The interior of the organize of Agent has instantly made judgment that other breakers in the protection area need tripping and sent the command to relative protection sub-system when storing those breaker to constitute one message according to the mode of network connector .