Bringing Demand-side Flexibility to Ancillary Service Markets

Demand-side flexibility is envisioned to have a crucial role in the future power system to accommodate the green transition with an increase of renewable, intermittent generation. The power grid and Transmission System Operators (TSOs) face challenges in balancing the grid due to the variability and uncertainty of renewable generation. Flexible demands can provide a solution to these challenges by adjusting their consumption patterns in response to grid conditions.

This thesis investigates the potential of demand-side flexibility in the power system from different perspectives. Through the lens of an aggregator, it is possible to investigate and outline the incentives for aggregators and flexible demands to utilize their flexibility for ancillary service markets. Taking the system perspective, we also assess current grid codes and their impact on flexibility procurement. The majority of the thesis is otherwise concerned with the perspective of an aggregator and its flexible demands. The aggregator aims to maximize profits by exploiting flexibility of its flexible demands. This can be done in ancillary service markets operated by the TSO, or as load shifting according to spot prices. The second perspective taken is that of the TSO and regulator.

Regarding the first perspective, we investigate the potential of demand-side flexibility in the Danish power system by outlining how flexibility provision is conducted status quo and in the future ecosystem as outlined by the Danish Energy Agency through the so-called Market Model 3.0. We demonstrate how the new market model lowers business barriers for new aggregators who wish to enter ancillary service markets but at the expense of increased technical complexity with respect to verification and settlement of flexibility provision.

Regarding the second perspective, monetary value of demand-side flexibility is of utmost importance to aggregators and flexible demands. This thesis develops mathematical optimization models for thermostatically controlled loads, as exemplified by two cases from IBM’s Flex Platform. The first case is a supermarket freezer that can exploit its thermal inertia in frozen food to either provide Manual Frequency Restoration Reserve ( mFRR) or to load shift. The second case is a zinc galvanizing furnace that can become flexible and provide Frequency Containment Reserve (FCR ) which we compare to mFRR. By estimating the temperature dynamics of both assets using real data, three optimization models are developed. For mFRR, this results in a two-stage stochastic mixed-integer linear program that makes optimal bidding decisions in a novel manner, adhering to the actual sequence of events in the market. For FCR and load shifting, two deterministic linear programs are developed. The optimization models maximize profit for the aggregator. We show how mFRR and load shifting have similar potential for profit for the supermarket freezer, while FCR also shows significant potential for profit for the zinc galvanizing furnace with next to no impact on its temperature.

The aggregator perspective is further investigated by including a portfolio of assets to assess the synergy effect of having multiple flexible demands. We show that the synergy effect is significant when aggregating assets without rebound effects through a simulated case study, and the use of historical data for electric vehicles participating in Frequency Containment Reserve - Disturbance (FCR-D) markets. However, for ancillary services with non-negligible energy delivery and assets with rebound effects, the synergy effect is less significant.

The thesis concludes with an investigation into a new grid code, labeled the P90 requirement, from the Danish TSO specifying a probabilistic framework for offering flexibility from stochastic flexible resources. We develop a mathematical model for bidding while adhering to and exploiting this requirement using joint chance constraints. We also provide two approximations to the model, one using conditional value-at-risk and another using ALSO-X algorithm. This is exemplified by a case study using historical data of electric vehicles participating in FCR-D up and down markets, showing annual savings of 6-10% for owners of electric vehicles. Lastly, we generalize our formulation to become distributionally robust, allowing an aggregator to bid in ancillary service markets even with a misspecified distribution of its available flexibility. We develop a tractable formulation in the form of a mixed-integer linear program and demonstrate its effectiveness in a simulated case study. We focus on the non-stationary power consumption of electric vehicles, showcasing how the inclusion of a distributional robustness constraint aligns with meeting the P90 requirement.

Returning to the perspective of the TSO, the P90 requirement is assessed through a bi-level optimization problem, the outer being the TSO maximizing its flexibility procurement, and the inner being the aggregator maximizing its profit. The TSO sets the P90 requirement, and the aggregator bids in the market while adhering to this requirement. Furthermore, the TSO also imposes distributional robustness on the aggregator by setting the Wasserstein distance within the tractable formulation developed earlier. As such, we show how different levels of the P90 requirement and Wasserstein distance affect the procurement of flexibility for the TSO by solving the bi-level optimization problem using a grid search.