In information-measurement systems (IMS) and information-control systems representing the state the plant (subsystem) being controlled, there exist problems that arise in conditions when some state parameters have no effect on subsystem measuring sensors, i. e. in conditions of incomplete information. This problem is solved based on analysis of the plant-IMS system dynamics equation (in the state parameters space), and neural network (NN) algorithms. The second (of three) paper parts considers the structure and learning algorithms for NN, designed on principles feedback and called recurrent NNs, which adequately simulate the input data dynamics.