Scaling Up Demand Response with Residential Thermal Loads: A Combined Control, Optimization, and Regulatory Perspective

The accelerating roll-out of Renewable Energy Sources and the rampant electrification
of energy end-uses are inducing more and more variability and unpredictability in the
power grid. These changes make it more challenging to balance electricity production
and consumption at all times, which is crucial to guarantee operational security and avoid
problems such as blackouts. Consequently, there is a growing need for flexible capacity
that can adapt its production or consumption in response to unforeseen changes.
Residential water-heating appliances such as Electric Water Heaters and Heat Pump
Water Heaters are excellent candidates to provide such flexible capacity. Their hot water
vessel constitutes an energy buffer that allows them to shift their consumption in time.
These appliances are massively present in the residential sector and represent a significant potential for electricity balancing. Water-heating appliances can be aggregated at a large scale to participate in wholesale energy and balancing markets similarly to generators. However, this is a challenging task, as it entails large-scale control problems subject to a lot of uncertainties, while the supply of hot water needs to be guaranteed for the end-user. This work examines how aggregations of water-heating appliances can be performed in large-scale commercial deployments. To do this, this thesis brings together three areas of knowledge: the regulation on residential Demand Response, the state of the art on aggregate control, and the literature on energy market participation. A three-step
control approach is used throughout this work. This approach consists of an aggregation
step in which a reduced-order representation of the cluster of appliances is constructed,
a scheduling step in which the aggregate power consumption is planned, and a dispatch
step in which the aggregate setpoints are tracked by switching heaters ON and OFF. This
thesis makes innovative contributions to each of these steps as well as to the existing
regulation and validates the overall approach for different applications in simulation
studies and real-life deployments.

This report starts with several contributions that are transversal across Demand Response applications. First, a comparative study of aggregate models for clusters of water-heating appliances is performed. The models are benchmarked according to their
conservativeness, reliability, and computational tractability. The heterogeneity of the
appliance parameters is shown to have an important influence on the available cluster-
level flexibility. The geometric Virtual Battery Model is identified as the model with the
most attractive features, though the application in which the model is used should be
taken into account.

Then, this work makes a regulatory contribution by proposing a framework for the
participation of independent aggregators of residential appliances in balancing markets
in the absence of head metering with sub-hourly resolution. A proposition is made
to subtract the Balance Responsible Party’s energy volumes from the distribution grid
infeed to prevent them from propagating through the allocation process and affecting
several Balance Responsible Parties. Arguments are also laid out for the omission of
retailer compensation, which is often deemed required in current market rules. These
recommendations are specifically designed to accelerate the deployment of residential
flexibility in the short term. After these service-agnostic developments, this work develops scheduling methods and performs validation studies for four use cases of large-scale aggregate Demand Response with clusters of water-heating devices. In the first use case, the provision of Frequency Containment Reserves is tackled with chance-constrained optimization and is validated in a simulation study. In the second use case, chance-constrained optimization is applied to the provision of automatic Frequency Restoration Reserves. The approach is validated in real life with a cluster of up to 600 Electric Water Heaters. In the third use case, a method to perform price arbitrage with a cluster of Electric Water Heaters in the Belgian single-price imbalance settlement is developed. The problem is treated as a reserves provision problem and price uncertainty is handled with chance-constrained programming. The method includes a Real-Time policy that leverages the fast reactivity of the appliances to respond to the latest available price information. The approach is validated in real life with a pool of 237 devices. Finally, the last use case entails the participation of a cluster of Heat Pump Water Heaters in the Californian wholesale
energy markets. The problem is tackled with mixed-integer stochastic programming
and large-scale simulation studies allow an evaluation of the valorization potential.
The performed validation studies show that the methods used and developed in this
work are adequate to tackle the different Demand Response participation problems under
realistic conditions and market rules.