Online estimation of variable parameters of synchronous machines using a novel adaptive algorithm: Principles and procedures
Zhengming Zhao,Longya Xu,Jianguang Jiang
1997-01-01
IEEE Power Engineering Review
Abstract:The frequently used adaptive algorithms include Least Mean Square (LMS) and Recurrence Least Square (RLS). LMS uses the gradient method to estimate parameters based on the data sampled at present. The ability of LMS algorithm is limited because of its poor tracking performance and large algorithm noise. RLS algorithm is based on an accumulated least square weighted by forgetting factor within a sliding time window. Essentially, the error sequence is weighted by the power of a Forgetting Factor �� and used as the driving force for an adaptive mechanism. Since this method can take previous data into consideration at present and emphasize the present data, the tracking performance of RLS is improved compared to that of LMS. However, the smaller �� is (forgetting more), the larger the deviation of parameter tracking and algorithm noise. On the other hand, if �� is selected larger (from 0.95 to 1), the tracking ability of RLS is reduced substantially. Although the formulation of weight factor in RLS is in question, the important effect of previous data to the adaptive online estimating SM parameters is evident. In this paper, a concept of Synthesized Information Factor (SIF) is proposed to replace the weight factor in RLS. A novel adaptive algorithm for online estimating SM parameters is then formed based on SIF. With a step-by-step derivation of the SIF based adaptive algorithm, this paper lays out the theoretical foundation for the adaptive online estimation method. The paper is organized as follows. First, conventional LMS and RLS adaptive algorithms are reviewed. Then, a novel adaptive identification algorithm is proposed and analyzed. A concept of Synthesized Information Factor is proposed as the core of the adaptive algorithm, and the physical meanings of SIF are explained. In particular, interface of SIF to the adaptive algorithm, and interface of the adaptive algorithm to the overall estimation process are described. The procedures of synchronous machine (SM) parameter estimation based on the novel adaptive algorithm are established. Furthermore, the paper presents computer simulation results to highlight the effectiveness of the adaptive algorithm. Finally, the paper completes with a brief summary and conclusion. Additional mathematical derivations are provided in the Appendix as the rigorous proof of the adaptive algorithm. The results of online estimating variable parameters of an actual 100 MVA turbogenerator and experimental verifications are detailed in a companion paper.