ニューラルネットワークを用いた多段予測制御方式

DOI Web Site Web Site 参考文献8件

書誌事項

タイトル別名
  • A Multiple-Step Predictive Control Algorithm Using Neural Networks
  • ニューラル ネットワーク オ モチイタ タダン ヨソク セイギョ ホウシキ

この論文をさがす

抄録

This paper is focused on the model-based control system by neural networks. There are two types of neural network-based models for dynamical systems, i.e., one is the series parallel model, and the other is the parallel model (external recurrent model). The former is generally used in predictive control. But it is not appropriate for the long-range (multiple-step) prediction because of its error accumulation through iterative application of the model. The latter is complex in terms of learning and control calculation, but it is favorable for the multi-ple-step predictive control. Many learning methods of parallel models were developed, but any control algo-rithm for this model has not been found yet. We developed the control algorithm by minimizing the multiple-step cost function based on multiple-step prediction. The iterative inverse method was used in the calculation of control. To illustrate the usefulness of the techniques presented in this paper, the comparison of the conven-tional (1step series parallel model) prediction/control and multiple step prediction/control are presented.

収録刊行物

参考文献 (8)*注記

もっと見る

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ