On-Line Identification of Power Generation Plant

Abstract

Mathematical models of electric power plants are needed to predict the response of the plant to events which occur on the power system, and in the development of improved control schemes. Analytical models suffer from difficulty in obtaining para-meters, and in updating them with changing plant conditions. The present work involves deriving models of boilers and turbogenerators directly from plant measurements during tests. These consist of injecting small pseudo-random disturbances to the plant, and recording its responses to these known inputs. A model is then derived by processing the data using identification algorithms. Extensive tests have been conducted on full-scale plant, to establish acceptable data-acquisition and identification procedures, identify suitable model structures, and determine their range of validity. The paper describes the methods adopted. Test and identification results are presented, showing that accurate models of plant may be obtained.