METHOD FOR CONTROLLING A TECHNICAL SYSTEM

For a parametrization method for parameterizing a control parameter (Pi) of a controller (2a) for controlling a technical system (3) that requires little overall computational time and effort a stochastic process model (SPM) is used to iteratively model a parameter-cost-relation (PJ) in the possible range (Ri) of parameter values (SPi), the parameter-cost-relation (PJ) defining a dependence of a defined cost function value (Jj) from the parameter value (SPi) of the at least one parameter (Pi). With the stochastic process model (SPM) a predicted mean value µ̂N+1 and a predicted variance σ^N+12 are determined and, form that, an upper confidence bound σU and/or a lower confidence bound σL around the predicted mean value µ̂N+1 for different parameter values (SPN+1j) are determined. A minimum (σmin) of the lower confidence bound (σL) or a maximum (σmax) of the upper confidence bound (σu) in the possible range (Ri) of parameter values (SPi) is searched in each iteration step j and the parameter value (SPij+1) associated to the minimum (σmin) or maximum (σmax) in the last iteration is used as the optimized parameter value (SPi,opt) of the at least one control parameter (Pi) with which the control parameter (Pi) is parameterized for controlling the technical system (1).