The generalized average consensus of heterogeneous multiagent system with discretetime on general strongly connected digraphs was investigated. Firstly, the definition of generalized average consensus was presented. Aiming at the study of average consensus confined to the balanced or symmetric network, a new linear distributed algorithm was proposed. The novelty is the augmentation of an additional variable for each agent, whose function is to record individual state update. By applying nonnegative matrices theory, graph theory and the eigenvalue perturbation theory, the consensus algorithm was studied. The result indicates that the algorithm guarantees state averaging on arbitrary strongly connected digraphs. Numerical simulations were given to demonstrate the effectiveness of the theoretical results.