Parameter Optimization and Model Fitting of Thermal Models of Air-cooled Hydrogenerator

Parameter optimization plays a vital role in fitting a mathematical model with experimental data. In the prior model development, some of the parameters are either chosen based on an educated guess or hit and trial method or even at random. This causes the mathematical model to drift away with experimental data. The optimized value of parameters can be found formulating a least-squares data fitting problem. In this paper, parameters of four mathematical quasi-linear thermal models of an air-cooled hydrogenerator are optimized. We have formulated the data fitting problem by using two different measurement data vectors. First, we have used a measurement data vector containing measurement of two of the states for finding optimized parameters and second, we have used two of the states and two of the algebraic variables. The optimized parameters found by using two different measurement data vectors are then used for fitting mathematical models with experimental data. The performance of model fitting is then compared using root mean square errors (RMSE) of least square errors. We found that the choice of data affects model fitting.