A proactive 2-stage indoor CO2-based demand-controlled ventilation method considering control performance and energy efficiency

•The proposed proactive 2-stage demand-controlled ventilation (P2S-DCV) method improves the DCV performance.•A deep neural network -based prediction model is proposed to predict future indoor CO2 and applies ventilation control in advance.•A reinforcement learning-based method is proposed to achieve accurate and rapid airflow control.•The fan pressure and damper positions are optimized to minimize energy consumption.•The proposed P2S-DCV method has good generalization ability under various conditions.