An Intelligent Machine Learning Framework For Autonomous Control of Conventional Air Conditioners

Abstract

In recent years, the demand for smart air conditioners (ACs) has expanded swiftly due to the need for comfort, automation, and energy efficiency. However, upgrading conventional AC units to smart systems is often expensive. To overcome this, the present research proposes a cost-effective IoT-based device that transforms a non-smart AC into an intelligent, smart AC. The developed model automatically adjusts cooling based on room temperature, occupancy, humidity, and CO₂ concentration without requiring a remote control or mobile application. The idea behind this research is to design and implement a low-cost, power-saving system that functions automatically without human intervention. The model is built using an ESP32 microcontroller interfaced with two proximity sensors, two temperature sensors, two humidity sensors, and one CO₂ sensor. One temperature and humidity sensor pair are placed outside the room, while the CO₂ sensor is located inside. The system monitors these parameters in real time and utilizes a logistic regression algorithm to predict whether high cooling is necessary. When occupancy exceeds four persons and environmental factors indicate increased CO₂ and humidity, the device automatically reduces the AC’s temperature to an optimal value. Experimental results give up to 82.72% accuracy in occupancy detection and improved cooling efficiency under crowded conditions. This economically feasible model provides a practical solution for domestic environments, senior citizens, and educational institutions, enabling intelligent control of traditional AC units without manual operation. This model is useful in conference halls of school or colleges. Also useful for senior citizens on domestic platforms.

Conventional air conditioners are widely used in residential and institutional environments, but most existing units require continuous manual operation and do not have the sensing and decision-making capabilities of modern smart air-conditioning systems. This study introduces an intelligent machine learning framework that enables conventional air conditioners to perform autonomous cooling control using an embedded sensing and computing platform. The proposed framework employs an ESP32 microcontroller to acquire real-time information from temperature, humidity, CO₂, and occupancy sensors. Environmental parameters are monitored from the indoor and outdoor surroundings to identify changes in thermal conditions and occupancy-related cooling requirements. The collected data are processed at the edge, allowing the system to make control decisions without depending on continuous user interaction or a mobile application. Logistic Regression is incorporated as the machine learning model to classify the required cooling condition based on the monitored parameters. When occupancy and environmental conditions indicate a higher cooling requirement, the system automatically modifies the operating temperature of the conventional AC. Experimental testing of the developed prototype achieved an occupancy detection accuracy of 82.72% and demonstrated reliable automatic adjustment of cooling under different occupancy conditions. The proposed framework provides a low-cost approach for extending intelligent control capabilities to existing conventional air conditioners without replacing the complete AC unit. The system can be applied in classrooms, conference halls, domestic environments, and other shared spaces where autonomous temperature regulation and reduced dependence on manual AC operation are desirable.