“…The frequency weighted (FW) methods are extended versions of BT. They are based on either (1) direct application of input/output weighting functions [16,29,63,66,68] or (2) Gramians in a restricted frequency interval = [ω 1 , ω 2 ] in the frequency domain [5,10,22,28,31] or (3) Gramians in a restricted time interval τ = [t 1 , t 2 ] in the time domain [22,25,32,61]. These lead to a higher model accuracy in selected frequency ranges, in particular when the parameters of the weighting functions and frequency/time intervals are optimized [50].…”