Estimation of the Air/Gas Ratio in Heating Appliances: Application of Basic ML Techniques to the State of the Art

Despite the prevalent notion that household heating systems operate with straightforward dynamics, they harbour numerous complications. Specifically, their construction often lacks integration of comprehensive sensor arrays, thereby limiting available data. Nevertheless, these systems are required to exhibit robustness in diverse operational scenarios, while simultaneously addressing safety concerns inherent in combustion processes to mitigate risks of explosions and intoxication. Additionally, there is a pressing need for energy-efficient operation to minimize environmental footprint, necessitating a wide modulation range. Central to achieving safe and efficient combustion is the maintenance of an optimal air-fuel ratio, denoted by the parameter $\lambda$. This study endeavours to evaluate elementary machine learning methods to enable heating systems to ascertain adherence to desired safety and efficiency standards. The selected techniques must be computationally lightweight to accommodate the constrained resources typically available on the applications' microcontrollers. Moreover, the scope of data availability is contingent upon the acquisitions feasible during routine application development, which time impacts the developing process. Subsequently, all findings are contextualized concerning the feasibility of practical implementation.