| 摘要 |
In an element (FLT) for filtering a regulating variable (X, Y) of a regulating system (ACE, FLT, RGL, AFZ, AFP), a multi-layered neuronal network (MNN) is used with an adaptation arrangement (ADAPT) for adapting weight values (w) of the weighted connections of the neuronal network while minimizing discrepancies (E) between input values (X) and their corresponding output values (Y). A particularly advantageous property of this filter (FLT) is its capacity to generalize, as well as its tolerance as regards contradictory input values (X). Such a filter (FLT) may be used for example for filtering the regulating variable in a network supply regulating system. |
