Most governments are applying financial instruments and other polices to encourage distributed renewable electricity generation (DREG). DREG is less predictable and more volatile than traditional forms of energy generation. Closure of larger fossil-fuelled power plants and rising share of DREG is reducing system inertia on energy networks such that new methods of demand response are required. Usually participation in non-dynamic frequency response is reactive, affecting the duty cycle of thermostatically controlled loads. However, this can adversely affect building thermal efficiency. The research presented takes a proactive approach to demand response employing heat transfer dynamics. Here, thermal dynamics exhibit a significantly larger inertia than electrical power consumption. Thus, short-term fluctuations in energy use should have less effect on temperature regulation and user comfort in buildings than existing balancing services. A prototype frequency sensor and control unit for proactive demand response in building stock is developed. The paper reports on hardware-in-the-loop simulations, testing real thermal loads within a simulated power network. The instrumented approach adopted enables accurate real-time electrical frequency measurement, while the control method offers effective demand response, which suggest the feasibility of using decentralised frequency control regulation as a novel approach to existing demand response mechanisms. Keywords Frequency regulation; decentralised control; demand response. Nomenclature transport delay, s load damping constant, s Δ frequency deviation, Hz inertia time constant, s TCL controller integral gain ALFC secondary loop gain, p.u. MW/Hz s ℎ thermal load gain TCL controller proportional gain Δ step change in power demand, p.u. Δ change in hydraulic amplifier output Δ ℎ electrical power deviation, p.u. Δ change in turbine power output regulator, Hz/p.u. MW Δ temperature deviation, p.u. governor time constant, s ℎ thermal load time constant, s turbine time constant, s ̅ sample mean number of entries sample standard deviation standard error standardized test statistic