On the Inclusion of Energy-Shifting Demand Response in Production Cost Models: Methodology and a Case Study

Abstract

In the context of future power system requirements for additional flexibility, demand response (DR) is an attractive potential resource. Its proponents widely laud its prospective benefits, which include enabling higher penetrations of variable renewable generation at lower cost than alternative storage technologies, and improving economic efficiency. In practice, DR from the commercial and residential sectors is largely an emerging, not a mature, resource, and its actual costs and benefits need to be studied to determine promising combinations of physical DR resource, enabling controls and communications, power system characteristics, regulatory environments, market structures, and business models. The work described in this report focuses on the enablement of such analysis from the production cost modeling perspective. In particular, we contribute a bottom-up methodology for modeling load-shifting DR in production cost models. The resulting model is sufficiently detailed to reflect the physical characteristics and constraints of the underlying flexible load, and includes the possibility of capturing diurnal and seasonal variations in the resource. Nonetheless, the model is of low complexity and thus suitable for inclusion in conventional unit commitment and market clearing algorithms. The ability to simulate DR as an operational resource on a power system over a year facilitates an assessment of its time-varying value to the power system. The modeling methodology is demonstrated through a case study of aggregated supermarket refrigeration systems providing balancing energy reserves in real-time markets at different levels of variable generation (VG). This DR resource is implemented in a test power system that represents a subset of the U.S Western Interconnection centered on Colorado. The value of DR from the population of supermarkets in Colorado is found to be $32.85 per kilowatt-year (kW-yr) presuming no other DR resources. The value decreases significantly (to $6.95/kW-year in the most extreme case) when we increase the capacity of the DR resource to naïvely represent the incorporation of DR from other flexible loads (in actuality, other DR resources will have different characteristics, such that the decrease in value will not be as steep). Refrigeration DR is found to offer greater value to the power system during the winter months than the summer months due to operational constraints that limit the flexibility of the resource during the summer. The value of DR is found to increase as the penetration of VG increases, reaching $46.05/kW-year for our baseline DR penetration and a variable generation (VG) penetration of 55%. We do see a plateau in the value of DR going from 45% to 55% VG. This is attributable to the inability of DR to provide energy storage on horizons longer than 24 hours. Overall, this work is a study in methodology. The case study is included primarily to show that the model is working properly and that this line of research is worthwhile. The reported numbers do not represent a true value of DR, but they do suggest orders of magnitude for a particular DR resource providing a particular grid service in a particular power system; they also confirm expected correlation directions between value and DR penetration (decreasing) and between value and VG penetration (increasing). Future work includes extending this method and developing new methods to be able to model physically realistic DR resources at scale. Some important aspects not studied here include capturing all possible value streams for a single resource (capacity, energy, and ancillary service values), simultaneously evaluating DR from multiple resources, and economically competing DR resources based on their costs of enablement and the trade-offs between end-user disutility and participation payments.