A general diffusion system consisting of conduction and mass flow can be clearly modeled in the form of a state equation using network concepts. In the previous paper 1), a diffusion system comprised only of conduction was discussed. That the eigenvalue of the system is a real and negative number was proved, and an exact time integration scheme was given. In Section 3 of the present paper, the diffusion system is generalized including mass flow, and proof is provided that the real part of the eigenvalue of the system is negative. Also, a definition by formulation is given for system parameters, which are the three kinds of capacety mij, generalized conductance cij, and free input ratio rij. Next, two theories of system parameter identification are deduced from the modern method of least squares. One will be called "Batch Identification", which is implemented in a single step after a long period of observing state and input variables. Its formulation is indicated by (34), and the algorithm is described in Section 12. The other will be called "Sequential Identification", implemented successively at every short observation period. It is formulated as a parameter time discrete system indicated by (51a) (51b) (51c) and (50), with the algorithm described in Section 13. The recursive equations of the latter are deduced from (34) employing Woodbury's matrix inversion Lemma. The sufficient and necessary condition of identifiability and proof of the asymptotic stability of the parameter time discrete system (51a) are presented in Section 9. Error in an identified system parameter is considered to be caused by the residual of the observing equation or by the observing noise and is evaluated by co-variance matrix (61). Two indices are given in Section 11 enabling evaluation of the appropriateness of the state equation structured by parameters identified for real phenomena.