Hydrogen is becoming a more and more crucial resource for chemical industries. Hydrogen network integration is an effective tool in hydrogen resource conservation. Hydrogen sources flow in and flow out from different processing units where consumption, generation, mixing, and separation occur. Consequently, there is a dependence on both hydrogen and contaminant species among those streams, especially when throughputs of processing units fluctuate. Hydrogen network integration should account for such a dependence so that real properties including the flow rate and concentration can be obtained to perform reliable network integration. This article proposes the processing unit mass balance to describe the dependence of hydrogen streams and establishes a corresponding superstructure as problem illustration. A mixed integer nonlinear programing (MINLP) model is built to synthesize hydrogen networks for fresh hydrogen minimization. This method considers the tolerable disturbance of the processing unit and thus a reliable fresh hydrogen target and internal sources profile can be obtained. Two cases are employed to demonstrate its application.