. "11"^^ . . "11"^^ . . . "GP102/07/P137" . "2015-01-22+01:00"^^ . "2007-01-01+01:00"^^ . . "0"^^ . "predictive control" . . "1"^^ . " artificial neural networks" . "Prediktivn\u00ED \u0159\u00EDzen\u00ED pomoc\u00ED um\u011Bl\u00FDch neuronov\u00FDch s\u00EDt\u00ED s online adaptac\u00ED prediktoru" . "."@en . "2009-12-31+01:00"^^ . . "predictive control; artificial neural networks; prediction"@en . . "Projekt se zab\u00FDv\u00E1 prediktivn\u00EDm \u0159\u00EDzen\u00EDm pomoc\u00ED um\u011Bl\u00FDch neuronov\u00FDch s\u00EDt\u00ED. Konkr\u00E9tn\u011B je zam\u011B\u0159en na problematiku prediktoru zalo\u017Een\u00E9ho na um\u011Bl\u00E9 neuronov\u00E9 s\u00EDti. V\u00A0sou\u010Dasn\u00E9 dob\u011B v\u011Bt\u0161ina metod prediktivn\u00EDho \u0159\u00EDzen\u00ED s\u00A0um\u011Blou neuronovou s\u00EDt\u00ED jako prediktorem pou\u017E\u00EDv\u00E1 pouze offline identifikaci prediktoru. Je to zp\u016Fsobeno t\u00EDm, \u017Ee pou\u017Eit\u00ED um\u011Bl\u00FDch neuronov\u00FDch s\u00EDt\u00ED v\u00A0\u0159\u00EDd\u00EDc\u00EDch algoritmech v\u00FDrazn\u011B navy\u0161uje v\u00FDpo\u010Detn\u00ED n\u00E1ro\u010Dnost cel\u00E9ho regul\u00E1toru a pr\u016Fb\u011B\u017En\u00E1 (online) identifikace prediktoru v\u00FDpo\u010Detn\u00ED n\u00E1ro\u010Dnost je\u0161t\u011B v\u00EDce zvy\u0161uje. Av\u0161ak pou\u017Eit\u00ED offline identifikovan\u00E9ho prediktoru m\u00E1 mnoho nev\u00FDhod. Jsou to p\u0159edev\u0161\u00EDm vysok\u00E9 po\u017Eadavky na jeho p\u0159esnost, resp. na kvalitu offline identifikace, d\u00E1le tak\u00E9 nevhodnost takov\u00E9hoto regul\u00E1toru pro \u0159\u00EDzen\u00ED t-variantn\u00EDch syst\u00E9m\u016F. Proto t\u011B\u017Ei\u0161t\u011B projektu spo\u010D\u00EDv\u00E1 v\u00A0nalezen\u00ED vhodn\u00E9ho prediktoru na b\u00E1zi um\u011Bl\u00E9 neuronov\u00E9 s\u00EDt\u011B, jeho\u017E v\u00FDpo\u010Detn\u00ED n\u00E1ro\u010Dnost umo\u017En\u00ED nasazen\u00ED v\u00A0prediktivn\u00EDch regul\u00E1torech p\u0159i\u00A0\u0159\u00EDzen\u00ED neline\u00E1rn\u00EDch syst\u00E9m\u016F. Sou\u010Dasn\u011B mus\u00ED struktura a vlastnosti um\u011Bl\u00E9 neuronov\u00E9 s\u00EDt\u011B" . "2009-04-22+02:00"^^ . "C\u00EDlem projektu bylo testovat st\u00E1vaj\u00EDc\u00ED typy neuronov\u00FDch s\u00EDt\u00ED s ohledem na jejich pou\u017Eitelnost pro n\u00E1vrh jednoduch\u00E9ho prediktivn\u00EDho \u0159\u00EDzen\u00ED nezn\u00E1m\u00FDch line\u00E1rn\u00EDch nebo neline\u00E1rn\u00EDch deterministick\u00FDch syst\u00E9m\u016F.Takov\u00E1 s\u00ED\u0165 byla vybr\u00E1na, pou\u017Eita. Byl navr\u017Een adapti"@cs . . . . . . . "0"^^ . . . . . "Predictive control using artificial neural networks with online adaptation of predictor"@en . "http://www.isvav.cz/projectDetail.do?rowId=GP102/07/P137"^^ . "Project involves predictive control using artificial neural networks. It is especially aimed to the predictor problem based on artificial neural network. Nowadays, most of predictive control methods with artificial neural network as a predictor use only offline identified predictor. It is caused by the fact that usage of artificial neural networks in control algorithms significantly increases computational demands of the controller as a whole. What is more, the online identification of the predictor computational cost increases even more. However usage of the offline identified predictor has many disadvantages. Main drawback is high demands for the predictor accuracy and offline identification is useless for t-variant processes. Therefore, this project will be focused on developing suitable predictor based on artificial neural network while its computational demands enable usage in predictive controllers of nonlinear processes. At the same time, the structure and properties of the artificial"@en .