Challenges of flexibility estimation from asset level data: insights from experimental studies

Flexible resources from buildings and distributed energy devices can deliver cost effective electricity market services. Flexibility availability estimates are required for planning, trading in the market and financial settlements. Flexibility is not a static value but depends on operating conditions and hence continuous estimation is required. However, accounting for the normal (baseline) operation behaviours of different types of assets in a way that is reliable, accurate, fair and can be deployed in a scalable way across potentially millions of assets is challenging. This study reports on explicit demand response trials conducted on four different types of assets; a commercial air-conditioning system, a lithium-ion battery, residential heat pump hot water system and kerbside electric vehicle chargers. A range of different baselining techniques are applied to evaluate the delivered flexibility for multiple increase-load and decrease-load events, and the results used to highlight both the identified challenges and a proposed pathway to address them. While the baseline behaviour of some of the assets was challenging to determine, limiting the duration of the explicit events to a shorter time for a given asset allows simple, efficient and device agnostic rolling horizon type models, especially the well-known Meter-Before-Meter-After method to be used. Over individual events estimated flexibility is found to have significant variability that originates from two sources: ‘unexplained’ variability of baseline power use, and the variable response to demand flexibility events. For some assets such as batteries and heat pumps normal operation is harder to predict but the flexibility response is more repeatable. On the other hand, for air-conditioning assets normal operation tends to be more predicable but the flexibility response is more variable. Aggregation of the flexibility response over multiple events for a given asset is suggested as the most plausible way of dealing with this variability for the purposes of awarding participants based on the flexible energy amount provided. These findings have implications for design of market based explicit flexibility schemes. • Challenges of estimating flexibility of different types of assets for settlement are summarised • Explicit demand flexibility trials are conducted over 4 different types of assets (EV, battery, a/c and hot water heat pump) • The performance of several common baselining methods is compared • Implications of the results for the design of explicit flexibility schemes are discussed.