Greenhouse agriculture faces increasing pressure to optimize crop production while minimizing energy consumption and environmental impact. Traditional ventilation control systems, relying on rule-based or classical proportional-integral-derivative (PID) controllers, often fail to adapt to dynamic environmental conditions and complex interactions between climate parameters. Deep reinforcement learning (DRL) has emerged as a promising paradigm for intelligent greenhouse ventilation control, offering adaptive, data-driven decision-making capabilities that can optimize multiple objectives simultaneously. This review examines the application of DRL algorithms to dynamic greenhouse ventilation systems, focusing on their potential to enhance energy efficiency, maintain optimal climate conditions, and support sustainable crop production. Key DRL techniques including Deep Q-Networks (DQN), Proximal Policy Optimization (PPO), and actor-critic methods are analyzed in the context of greenhouse climate control. The integration of these algorithms with sensor networks, IoT infrastructure, and simulation environments enables real-time optimization of temperature, humidity, and CO₂ levels while reducing energy costs. Performance evaluations demonstrate that DRL-based systems can achieve 15-30% energy savings compared to conventional controllers while maintaining or improving crop yield and quality. However, significant challenges remain, including data scarcity, training instability, safety constraints, and the gap between simulation and real-world deployment. Future research directions emphasize the integration of DRL with digital twins, transfer learning approaches, and explainable AI techniques to enhance system robustness, interpretability, and practical adoption in commercial greenhouse operations.