Performance Assessment and Predictive Modeling of a Hybrid Hydrogen-Natural Gas Water Heater Using Experimental Data and Machine Learning

In response to the global need to reduce greenhouse gas emissions and advance the decarbonization of thermal energy systems, this study evaluates the performance of a tankless water heater operating with hydrogen–natural gas blends. The objective is to improve thermal efficiency and reduce pollutant emissions without requiring major modifications to existing equipment. Experimental tests were conducted at three thermal power levels (35, 40, and 45 kW) and four hydrogen volume fractions (0%, 20%, 40%, and 60%), analyzing operational variables such as temperatures, flow rates, efficiency, and NOx emissions. Results show that efficiency increases with hydrogen content, particularly at lower power levels, reaching a maximum of 56%. NOx emissions tend to rise with both power and hydrogen fraction, although this effect can be mitigated by controlling the water flow rate. In addition, machine learning models were trained to predict efficiency and emissions, with the scaled Support Vector Regression (SVR) model achieving R² values above 90% for both outputs. This approach not only enables system optimization but also represents a step toward the implementation of digital twins and opens the door to monitoring indirect variables, offering broad potential for predictive applications in thermal equipment.