Method for the recognition of the fuel composition in CNG engines fed with natural gas/biofuel/hydrogen blends

Abstract

The production of biofuels from biomass gasification results in a wide range of fuel mixture compositions, which have a notable impact on performance and emissions of internal combustion engines. Thus, it is crucial that the composition of the blend being used is automatically identified by the engine control unit, after refueling. Standard ECU hardware is not able to recognize which fuel is actually present in the vehicle tank, hence combustion parameters cannot be changed accordingly. A methodology to perform an estimation of the composition of natural gas/biofuel/hydrogen blends, as it is fueled into CNG engines, is here presented and assessed. Experimental data have been collected in an SI engine, working under different steady-state conditions. Five different fuel blends of known composition have been tested with the same ECU calibration, relative to CNG. The injector has been characterized by means of a linear regression model aiming to predict the injection duration from data readily available in the ECU. A candidate set of 10 different sample blends, representative of the possible wide variety of blends that can be fed to the engine at the fuel pump station, was given as input to the regression model and an error in terms of injection duration has been evaluated. The method has proved to correctly recognize the actual fuel blend within the candidate set, with a maximum error of 5% on hydrogen volume content when HCNG mixtures are considered. The recognition algorithm converges after less than 10 different engine working conditions, in terms of speed and load. The candidate set has been extended to a full factorial set of 2 million blends, to validate the ability of the recognition method to correctly match the real fuel.