Recently

the heat and electricity integrated energy system (HE-IES) has become a hot topic in both industry and academia. In the HE- IES

the potential flexibility of the buildings' thermal loads can be exploited to relax the heat power balance constraints and consequently allow a more flexible operation of the combined heat and power units. In this paper

model-driven and data-driven techniques are combined to quantify the demand flexibility of the buildings' thermal loads in a non-instructive way. First

the explicit analytical equivalent thermal parameter (ETP) model of the aggregated buildings is developed. The heat transfer coefficient (

$k$

) and thermal inertia coefficient (

$C$

) of the ETP model are designated to measure the potential demand flexibility. Second

the Particle Swarm Optimization optimized Radial Basis Function neural network (PSO-RBF) is used to identify the relationship between the values of

$k$

and

$C$

and the meteorological factors. To obtain the training data

an innovative two-stage regression method based on the adaptive temporal resolution is proposed to ext

ract

$k$

and

$C$

values from the historical thermal load data. Finally

a flexible thermal load model is built based on the predictions of the meteorological factors

which can be conveniently incorporated into the online dispatch of the HE-IES. A comprehensive simulation environment is designed to verify the accuracy and availability of the proposed technique.