Application of adaptive control to reduce cyclic dispersion near the lean limit in a small-scale, natural gas engine

Predictive feedback control is applied to achieve reductions in cyclic dispersion in an analytical, lean, spark-ignition model and a two-cylinder, four-stroke, natural gas Kohler Command 25 engine operating at lean conditions. Recent observations of the combustion dynamics are used to define a desired target point for control and to predict future combustion events which may stray from the target point. Fueling perturbations are applied to steer the system back toward the desired behavior. Overall control perturbations are constrained to maintain a constant average fuel-to-air ratio. We present two methods for obtaining the predictions of future combustion events. In the first method, the recent history of cycle heat release is used to construct an adaptive, low-order map which relates the current-cycle heat release to the next-cycle heat release. The second method uses symbolic analysis to determine the relative frequency of successive-cycle combustion events and predict the most probable successor to the current cycle. Results are presented which show a moderate reduction in cycle-to-cycle variation near the lean limit in both the model and the engine. Similarities in behavior have been shown to exist during stressed (dilute) combustion in a wide variety of spark-ignition engines suggesting that a similar prediction strategy could be successfully applied to control cyclic dispersion in large-scale reciprocating engines.