Event-driven-based adaptive control for thermostatically-controlled loads with online identification of thermal parameters

Abstract

The participation of thermostatically-controlled loads (TCLs) in demand response (DR) can effectively alleviate the power supply pressure during extreme weather conditions. However, current control methods often assume constant thermal parameters, neglecting their variability among load devices and dynamic changes with the environment, leading to inaccurate assessment of the TCL adjustability. Furthermore, the differentiated user preferences are not effectively utilized to exploit the TCL adjustability. These factors impact the precision of TCL clusters in tracking the target power. Therefore, in this paper, an event-driven-based adaptive control strategy for TCLs with online identification of thermal parameters is proposed. Firstly, the aggregator uses a conversion function to convert the target power into the target voltage. When a load control event is triggered, the control signal is broadcast to each load agent. The agents then perform initial screening based on adaptive action thresholds to limit the number of devices acting simultaneously. An improved Transformer neural network is used for rapid online identification of thermal parameters in different load devices. Leveraging heterogeneous thermal parameters and personalized user preferences, the TCLs’ adjustability is deeply explored for autonomous decision-making. Simulation results demonstrate that the proposed strategy effectively enhances the speed and accuracy of thermal parameter identification. Under the premise of safeguarding user preferences and ensuring control fairness, more precise power tracking results are obtained.