Decentralized Control of Thermostatic Loads for Flexible Demand Response

Abstract

Abstract-Thermostatically controlled loads (TCLs) such as refrigerators, air-conditioners and space heaters offer significant potential for short-term modulation of their aggregate power consumption. This ability can be used in principle to provide frequency response services, but controlling a multitude of devices to provide a measured collective response has proven to be challenging. Many controller implementations struggle to manage simultaneously the short-term response and the longterm payback, whereas others rely on a real-time commandand-control infrastructure to resolve this issue. In this work we propose a novel approach to the control of TCLs that allows for accurate modulation of the aggregate power consumption of a large collection of appliances through stochastic control. By construction, the control scheme is well suited for decentralised implementation, and allows each appliance to enforce strict temperature limits. We also present a particular implementation that results in analytically tractable solutions both for the global response and for the device-level control actions. Computer simulations demonstrate the ability of the controller to modulate the power consumption of a population of heterogeneous appliances according to a reference power profile. Finally, envelope constraints are established for the collective demand response flexibility of a heterogeneous set of TCLs. D. Dependent variablestemperature of warmest appliance (≤ T max ) θ(t) temperature of particular appliancē T (t) ensemble average temperature f (T, t) probability density of all appliances f s (T, t) density of appliances in state s ∈ {on, off} Φ(T, t) net on-off density flux r on off (T, t) stochastic switch-off rate r off on (T, t) stochastic switch-on rate

Leaky storage model for optimal multi‐service allocation of thermostatic loads

“…The model has been introduced in the context of cold appliances, but it applies equally to heating appliances. The leaky storage model is exact for large TCL populations with linear thermal models that are controlled by the controller in [8], and it guarantees the deliverability of a compatible power consumption pattern by a heterogeneous collection of appliances.…”

“…This system is typically modelled by a first order temperature model ( [3], [4], [11], [13]), giving rise to the following ODE for the evolution of the appliance temperature with a variable cooling rate. Note that the temperature distribution of a population of identical but independent appliances may also be interpreted as a probability distribution for the temperature of a single appliance with unknown initial conditions [4,8].…”

“…In this work, we assume that the TCL actions are coordinated by a TCL controller that is based on the design proposed in [8].…”

“…The temperature-linear controller in [8] was designed with analytical simplicity in mind, limiting it to control actions that consist of initial reductions in aggregate power consumption, optionally followed by an increase In the absence of a control signal the controller acts as a regular deadband controller that keeps the appliance temperature within…”

Value of thermostatic loads in future low-carbon Great Britain system

“…The dynamics of the leaky-storage unit are governed by a decentralized hybrid control strategy [11] and its extensions in [10]; under the considered strategy, each appliance individually targets the desired relative power curve ( ), so that the total power consumption equals (2). ̂0 [MW] is the average steady state population consumption and the superscript is used for single device parameters.…”