Bilevel load-agent-based distributed coordination decision strategy for aggregators

This paper addresses the risk faced by load aggregators (LAs) participating in a demand response (DR), owing to the response uncertainty, and proposes a load-agent-based bilevel distributed coordination decision-making strategy for LAs considering response uncertainty. DR uncertainty models of reducible and transferable loads are established based on the evidence theory. The agent DR uncertainty model is established through individual user DR characteristic clustering. The conditional value at risk is used as a measure of risk, and a two-hierarchical-decision model of the LA is established. The model is decoupled by introducing decoupling variables and is solved using the analytical target cascading method, owing to the coupling relationship of the bilayer decision model. The results of the example show that the computing time of the distributed scheduling method proposed in this paper is 91.54 % shorter than that of the centralized scheduling method. Compared with the traditional distributed scheduling strategy, under the condition of comparable computing efficiency, the risk cost expectations of LAs is lower after clustering processing. The proposed model can reasonably evaluate uncertain events in the DR and realize decentralized coordination optimization between the upper and lower layers of decision-making, which helps LAs to effectively improve the decision-making efficiency.(c) 2021 Elsevier Ltd. All rights reserved.

Automated summary

This paper proposes a load-agent-based bilevel distributed coordination decision-making strategy for load aggregators participating in demand response, considering the risk caused by response uncertainty. Uncertainty models for reducible and transferable loads are established based on evidence theory, and an agent uncertainty model is created through clustering of individual user characteristics. The proposed strategy reduces computation time and lowers risk cost expectations after clustering processing.