In recent years, the field of humidity control in air-conditioning systems has emerged as a prominent area of research, with traditional humidity sensors facing challenges in terms of accuracy and stability. This study proposes a virtual humidity sensing technique based on convolutional neural networks (CNNs) to accurately predict indoor relative humidity by analyzing multiple air conditioner operating parameters. Nine air conditioner parameters, including compressor exhaust temperature and evaporator temperature, are used as inputs, and the humidity prediction is carried out using a two-layer CNN network structure. The findings demonstrate that the model’s prediction errors are predominantly within a 5%RH (relative humidity) range in typical home environments, indicating its robust generalization ability and potential for practical applications. This approach offers a novel solution to overcome the limitations of traditional humidity sensors and has a diverse range of applications in intelligent air conditioning system control.