Thermostatically Controlled Loads (TCLs) represent a valuable source of flexibility for the system. Depending on network needs, these devices could alter their nominal energy consumption and provide multiple ancillary services, facilitating the cost-effective transition to a low-carbon power system. Previous work mainly focused on investigating single service provision from TCLs, while intersections among different services have not been considered. Furthermore, the intrinsic energy payback effect was not fully included within optimisation models for TCL scheduling. This paper presents a novel Demand Side Response Model (DSRM), which enables the optimal scheduling of energy/power consumption of a heterogeneous population of TCLs and the simultaneous allocation of multiple ancillary services. The model explicitly considers the effect of the energy recovery after delivering the services so that the deliverability of scheduled services from TCLs is always guaranteed. The proposed DSRM is integrated into an Advanced Stochastic Unit Commitment model (ASUC) to investigate the system benefits of the flexibility from TCLs. Case studies demonstrate that 1) time-varying provision of multiple services from TCLs significantly increases their benefits, 2) TCL operation which aims to minimise the amplitude of the energy recovery causes sub-optimal utilisation of the devices, 3) ignoring the energy payback leads to overestimate the TCL value.