一种改进型的自适应预报算法

A modified adaptive forecast algorithm

采用一种改进型自适应预报算法，有效地解决了预测时间点与实际时间点不一致的预测问题．该算法的系统辨识技术，通过一种“复位”机制的引人，可以在确定的时间内完成参数估计及控制器、预报器的设计．自适应预报器采取基于最小二乘法的直接自适应ｄ步预报方案，利用“后退”方法实现其输出的最优预报．此外，系统通过自学习的方法使得预报模型给以较快的速率收敛．

A modified adaptive forecast algorithm presented in this paper has effectively solved the problem of incoordinate between forecase time and actualtime. The identitfcation in the algorithm, through the introduction of deducible mechanism, can fulfil the estimate of parameter and the design of controller and predictor within fixed time. The adaptive predictor adopts the direct adaptive d-step prediction program based on the least square method and achieve the bet prediction input by retreat. In addition the system enable, the predicting pattern to converge at relatively rapid sped through automatic study.