SVR的限定记忆在线辨识算法及其应用

An On-line Identification Algorithm Based on SVR's Limited Memory and Its Application

目前,基于支持向量回归的辨识研究虽然涉及领域很广,但基本上都为离线辨识,其原因在于支持向量回归处理大批数据时存在耗时长及内存开销大的问题,不能满足实时计算的要求。针对这个问题提出了基于支持向量回归的限定记忆在线辨识算法,该算法有效地避免了内存开销大的问题,满足了在线辨识实时性的要求。利用该算法对具有动态、时变特性的主汽温系统进行了辨识研究,仿真结果表明:该算法可以获得很好的辨识效果,具有良好的可行性。

An on-line identification algorithm based on SVR＇s （support vector regression） limited memory is being proposed, which can effectively dodge the problem of occupying too large a volume of memory and satisfy the requirements of real time identification. The algorithm has been used for identification studies of fresh steam temperature, which is characterized by its dynamic and time varying behavior. Simulation results show the algorithm to be feasible and sound. Figs 6, table 1 and refs 9.