Model validation and control systems analysis for pressurized water reactor components

A study of the water level control in a U-tube recirculation type steam generator of a 900 MWe pressurized water reactor was undertaken for operation at full power and at 30% power level. The need to simulate the complete power plants and controls for load following transients required the use of low-order physical models for the core and the steam generator and simulation of the control system components presently in use.

The validation of the models, necessary for any control study, was accomplished at both power levels by performing model parameter identification and using experimental transients from 100-90% and 30-20% power demand step perturbations. The large, complex, industrial processes modeled required the design of a multi-input, multi-output time domain identification scheme, using the model-adjustment method with a leastsquares estimator and the simplex optimization scheme for its minimization. The results of the model validation work have demonstrated that the physical models used represent the dynamic behavior of the main components of the pressurized water reactor satisfactorily and realize a very good trade off between accuracy and complexity.

The importance of high quality experimental data in model validation work, and the existence of various problems associated with the electronics of the data transmition, acquisition and recording required careful treatment.

The implementation of the overall system simulation in real time on a minicomputer was achieved. Special care was taken for a detailed simulation of the control rod nonlinear movement. It was shown that the presently used control system is sufficient in controlling the water level after load following transients at the power levels studied. The use of a medium size minicomputer in this work showed that safety and functional analyses of a large process can be achieved with economy and flexibility as well as accuracy. The use of the Linear-Quadratic- Gaussian theory was attempted as a modern control theory application for an industrial process, but numerical problems were encountered and the approach was found to be impractical.