An adaptive fault diagnosis and compensation scheme for stochastic distribution system under 2-step neural network modeling frame

An adaptive sensor fault diagnosis (FD) and compensation scheme for stochastic distribution control (SDC) systems is studied under framework of 2-step neural networks in this paper. The 2-step neural networks are used for modeling purpose, where the static neural network (NN) is employed to model the output probability density function (OPDF) while the dynamic NN is employed to identify the nonlinearity, uncertainty of system and to refine the OPDF model. An interesting thing is that the dynamic NN designed here is also as a part of a filter for fault diagnosis purpose, where some adaptive rules are given to character the coefficient matrices and their boundary and an adaptive learning rule is given for fault estimation. Through such adaptive algorithm, nonlinear parameter estimation and sensor fault identification can be well dealt with simultaneously. A sensor compensation rule is consequently given to restore the plant with output feedback controller. A simulation example is given to verify the effectiveness of the presented algorithm.