Subspace Modeling , Neural Decoupling and Robust Nonlinear Controller Design for Two-Time-Scale Nuclear Power Plant

In this paper, a novel Hybrid Linear and Nonlinear (HLN) technique of modeling and control design for two-time-scale nuclear power plant has been devised in LabVIEW environment. A higher order linear model for two-time-scale nuclear power plant power control is developed based on DeterministicStochastic Subspace Identification via Principal Component Analysis (DSSIPCA) technique. The eighth order linear model is identified using innovative time domain plant data. The eighth order model is decoupled into two modes of dynamics using Nonlinear Recurrent Artificial Neural Network (NRANN). The first reduced order model is a second order model capturing slow dynamics of plant while second reduced order model is a sixth order model capturing fast dynamics of plant. The slow dynamics model is so decoupled that it mimics the original higher order model of nuclear power plant. A sliding surface is designed in state space for slow dynamics model and full order model. Based on sliding surface, a Robust Discrete Nonlinear Controller (RDNC) is designed for decoupled two-time-scale nuclear power plant model in Triangular Block Structure (TBS) form using Sliding Mode Control (SMC) technique. The design, optimization, testing, validation and analysis work is carried out in most modern graphical programming environment LabVIEW 7.0. The performance of proposed of HLN technique is tested in reference tracking mode for an operating unit of pressurized heavy water reactor based nuclear power plant in Pakistan and found satisfactory and within design limits.