<![CDATA[Diesel generators (gensets) are essential in India for industries, construction, agriculture, and as backup power for hospitals and data centres. Common fuels include diesel, petrol, natural gas, and, increasingly, solar energy, with hybrid systems gaining popularity for improved efficiency and reduced emissions. Diesel gensets remain reliable and cost-effective, especially in remote areas, but growing environmental concerns are driving adoption of cleaner alternatives like compressed natural gas (CNG), bio-CNG, and dual-fuel systems. HCNG (hydrogen-enriched compressed natural gas) gensets are more efficient and environmentally friendly, though they require greater initial investment. Adding hydrogen enhances combustion and reduces emissions. In this study, various HCNG blends were tested on a multi-cylinder, single-speed gas engine. Experimental evaluation of combustion and performance characteristics is typically time and resource-intensive, so Machine Learning (ML) was applied to streamline the process, thereby minimizing the number of required experiments. The engine performance is assessed using the engine dynamometer, whereas the combustion characteristics are obtained from the High-Speed Data Acquisition (HSDA) system. A Random Forest (RF) regression model was developed to predict performance and combustion characteristics for higher HCNG blends from lower-blend data, with hyperparameter optimization used to improve accuracy and minimize overfitting. Predicted values were validated against experimental results, showing strong correlations. Key parameters like Brake-Specific Fuel Consumption (BSFC), Brake Mean Effective Pressure (BMEP), Exhaust Temperature, Maximum In-Cylinder Combustion Pressure (Pmax), Indicated Mean Effective Pressure (IMEP) and Combustion Duration were predicted, with evaluations showing strong correlations between predicted values and actual results.]]>
