基于Q-learning的空气源热泵制冷系统无模型优化控制在传感器误差下的鲁棒性分析OA

The operation of HVAC systems is highly dependent on sensor data,and measurement errors can significantly affect the optimization effect and stability of control strategies.Taking the air source heat pump(ASHP)system of an office building in Shanghai City as the research object,a simulation environment is constructed based on real operation data,and different levels of sensor noise errors are introduced to quantitatively analyze their impact on the performance of model-free optimization control based on Q-learning and model-based optimization control based on particle swarm optimization(PSO).The results show that within the range of conventional sensor errors in industry,both optimization methods can achieve an improvement in system energy efficiency,with Q-learning performing better in terms of overall utility stability.Under fault-level sensor noise conditions,the performance of PSO deteriorates,but Q-learning still maintains better robustness.