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． An improved adaptive prediction algorithm is adopted to effectively solve the prediction problem that the predicted time point is not consistent with the actual time point. The system identification technology of the algorithm, through the introduction of a “reset” mechanism, can complete parameter estimation and design of controllers and predictors within a certain time. The adaptive predictor adopts a direct adaptive d-step prediction scheme based on the least square method, and uses the “backwards” method to achieve the optimal prediction of its output. In addition, the system allows the forecast model to converge at a faster rate through a self-learning approach.