Fast extremum seeking for optimization of brake specific fuel consumption

Extremum seeking control is a well-established technique for dealing with the online optimization of unknown dynamic systems. However, in many applications the convergence rate of traditional extremum seeking approaches is a limiting factor, largely arising from the need for full time scale separation between the elements of the closed loop system. Recent theoretical developments have shown that for plants with a Hammerstein structure, it is possible to exploit different tuning regimes for the algorithms, leading to significantly faster convergence speeds. This approach is extended here to a class of Hammerstein-Wiener plants and demonstrated experimentally in a stationary power plant calibration problem, where it is shown that optimal brake specific fuel consumption in a compressed natural gas engine can be reached an order of magnitude faster than with the more conventional extremum seeking approaches.