Full Operating Range Robust Hybrid Control of a Coal-Fired Boiler/Turbine Unit

Abstract

Multi-input-multi-output robust controllers recently designed for the megawatt output/throttle pressure control in a coal-fired power plant boiler/turbine unit have demonstrated performance robustness noticeably superior to that of the currently employed nonlinear PID-based controller. These controllers, however, have been designed only for the range of 150–185MW around the 185MW nominal operating point, exhibiting a significant loss of performance in the lower range of 120–150MW. Through system identification, the reason for this performance loss is demonstrated in the current work to be a pronounced dependence of the boiler/turbine unit steady state gains on the operating point. This problem is addressed via a hybrid control law consisting of two robust controllers and a robust switch between them activated by the set point change. The controllers are designed to cover the corresponding half-ranges of the full operating range. This permits attainment of the desired overall performance as well as reduction of modeling uncertainty induced by the operating point change to approximately 25% of that associated with the previous designs. Robust switching is accomplished through a novel hybrid mode of behavior—robustly controlled discrete transition. The latter mode is produced through realizing that the off-line transfer speedup suggested by and (2005, “ The L2(l2) bumpless Transfer Problem for Linear Parts: Its Definition and Solution,” Automatica, 41, pp. 1273–1280) can be taken to the limit and incorporating the result into a robust bumpless transfer technique recently developed by the authors. As demonstrated by simulation results, the proposed strategy provides an adequate solution to the problem of robust boiler/turbine unit performance over the full operating range. This fact combined with numerical algorithm tractability, relative ease of its design, its insensitivity to implementation nonidealities, and accompanying identification methodology for nominal model generation makes it a viable candidate for industrial acceptance.