Temperature controlled loads (TCLs) are of significant importance in demand response (DR) research due to their remarkable thermal inertia and regulation potential. Addressing the limitations of existing models in fully capturing the time-varying characteristics and regulation capacity differences of TCLs, a data-driven parameter identification and demand response capability evaluation method is proposed. Based on the second-order equivalent thermal parameter (ETP) model, a linear identification equation is constructed, and a multi-layer recursive least squares algorithm with an adaptive forgetting factor is introduced to achieve online identification of the dynamic equivalent parameter matrix. Furthermore, tailored demand response capability evaluation strategies are designed for switching loads and continuously adjustable loads, considering their distinct regulation characteristics. Simulation results demonstrate that the proposed method can accurately identify load parameters and evaluate demand response capabilities, providing reliable support for TCL modeling and optimization of grid ancillary services.