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

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.