next up previous contents
Next: Section NAMEN Up: Input File of PAREST Previous: Input File of PAREST

Section GENERAL

This section contains the description of the dimensions of the problem and of optional parameters of the optimization method. They are set by assigning values to the following variable names:

tabular275

The selection between parameter identification and optimal control problems is done by setting

tabular286

Free or prescribed initial and final time are set by using the statements

tabular292

in this section of the input file.

Three different methods can be selected for the numerical integration of the dynamical equations (1)

  1. DASSL: the initial-value problem solver for differential-algebraic equations of [13] (backward difference formulas) with the extensions of [1] for a numerical sensitivity analysis.
  2. DOPRI8: the initial-value problem solver for ordinary differential equations of [10] (8th order extrapolation method) with the extensions of [6] for a numerical sensitivity analysis.
  3. RKF4: an initial-value problem solver for ordinary differential equations (4th order Runge/Kutta method) with the extensions of [6] for a numerical sensitivity analysis.

tabular311

As a rule of thumb, the tolerances for integration should be less or equal one tenth of the tolerances for optimization (see below), for example,
RTOL tex2html_wrap_inline1551 and OPTTOL tex2html_wrap_inline1553 (see description below).

Three iterative methods for optimization are available:

  1. NLSCON (Nowak, Weimann [11], [12]):
    a generalized Gauß-Newton method for nonlinear least squares problems subject to nonlinear equality constraints [8]
    (NLSCON can only be used for parameter identification problems and not for optimal control problems),
  2. NPSOL (Gill, Murray, Saunders, Wright [9]):
    a Sequential Quadratic Programming (SQP) method for general nonlinearly constrained optimization problems [14]
    (NPSOL is the only possible selection for optimal control problems. It may also be selected for parameter identification problems but will usually be less efficient than NLSSOL or NLSCON),
  3. NLSSOL (Gill, Murray, Saunders, Wright [9]):
    an SQP-method for nonlinear least squares problems subject to general nonlinear equality and inequality constraints
    (NLSSOL can only be used for parameter identification problems and not for optimal control problems).

tabular351

tabular371


next up previous contents
Next: Section NAMEN Up: Input File of PAREST Previous: Input File of PAREST

Oskar von Stryk
Tue Feb 1 13:50:42 CET 2000