| Popis: |
Superheated steam temperature dynamic is statically non linear. The steam enthalpy exhibits nonlinear dependence on both steam pressure and temperature. Also, the heat transfer process on superheaters and attemperators is strongly non linear and it was, reportedly, very difficult to find a synthetic expression (model). The so-called Genetic Multivariate Polynomials (GMP) solve this problem by finding the coefficients of a multivariate polynomial for an arbitrary set of data. Although this regression problem has been tackled with success using neural networks (NN) the ‘black box’ characteristic of such models is frequently cited as a major drawback. Despite the restrictions of a polynomial basis, GPMs compete favorably with the NNs without the mentioned limitation. Therefore, a practical tool is proposed for temperature modeling on a wide range real plant operation and its static parameter estimations. Based on advanced simulation tools, the polynomial expression of enthalpy (on a wide range) and the empirical heat transfer equations in superheaters allow us to turn the static parameter estimation from a distributed to a lumped parameter system. |
