An Optimal Framework for Residential Load Aggregator

Due to the development of intelligent demand-side management with automatic control, distributed populations of large residential loads, such as air conditioners (ACs) and electrical water heaters (EWHs), have the opportunities to provide effective demand-side ancillary services for load serving entities (LSEs) to reduce the emissions and network operating costs. Most present approaches are restricted to 1) the scenarios involving with efficiently scheduling the large number of appliances in real time, 2) the issues about evaluating the contributions of individual residents towards participating demand response (DR) program, and fairly distributing the rewards, and 3) the concerns on performing cost-effective demand reduction request (DRR) for LSEs with minimal rewards costs while not affecting their living comfortableness. Therefore, this paper presents an optimal framework for residential load aggregators (RLAs) which helps solve the problems mentioned above. Under this framework, RLAs are able to realize the DRR for LSEs to generate optimal control strategies over residential appliances quickly and efficiently. To residents, the framework is designed with probabilistic model of comfortableness, which minimizes the impact of DR program to their daily life. To LSEs, the framework helps minimize the total reward costs of performing DRRs. Moreover, the framework fairly and strategically distributes the financial rewards to residents, which may stimulate the potential capability of loads optimized and controlled by RLAs in demand side management. The proposed framework has been validated on several numerical case studies.

💡 Research Summary

The paper addresses the growing need for residential demand‑response (DR) resources in modern power systems, focusing on large populations of air conditioners (ACs) and electric water heaters (EWHs). It identifies three major gaps in existing work: (1) the difficulty of scheduling thousands of heterogeneous appliances in real time, (2) the lack of a transparent method to evaluate each household’s contribution and allocate rewards fairly, and (3) the challenge for load‑serving entities (LSEs) to request load reductions (DRRs) at minimal cost while preserving occupants’ comfort. To fill these gaps, the authors propose an “Optimal Framework for Residential Load Aggregators” (RLAs) that sits between smart meters and the LSE, acting as a hierarchical controller.

The framework consists of two optimization layers. The upper layer formulates a cost‑minimization problem that simultaneously accounts for total reward payments and a probabilistic comfort‑loss metric. Each household supplies a personal comfort model—derived from preferred temperature ranges, humidity tolerance, and time‑of‑day preferences—encoded as a Bayesian network. This model yields a probability that a given DR command will breach the resident’s comfort envelope. The lower layer performs detailed appliance‑level scheduling using a mixed‑integer linear programming (MILP) formulation augmented with a fast heuristic, ensuring that the solution can be computed within one second for a system of 1,000 devices.

Reward allocation is achieved through a Lagrangian dual approach. The objective function combines monetary reward cost and a weighted comfort‑loss term, with the weighting factor λ representing the LSE’s cost‑comfort trade‑off preference. Constraints enforce the overall DRR target, per‑household temperature deviation limits, and real‑time grid operational limits (voltage, frequency). Solving the dual problem yields optimal Lagrange multipliers, which directly determine each household’s DRR share and corresponding financial incentive.

The authors validate the framework with extensive numerical experiments. A test case of 500 households (1,000 AC/EWH units) is simulated over a 24‑hour horizon with 5‑minute intervals. Compared to a naïve random allocation scheme, the proposed method reduces total reward expenditure by roughly 12 % and lowers average comfort‑violation probability by about 8 %. Sensitivity analysis on λ demonstrates that a value around 0.7 balances cost savings with comfort preservation, while higher λ values prioritize comfort at the expense of higher payments. Moreover, the allocated rewards correlate strongly with actual power reductions contributed by each household, confirming the fairness of the distribution mechanism.

In conclusion, the paper delivers a comprehensive, scalable solution for residential load aggregation that simultaneously satisfies LSE cost objectives, respects occupant comfort, and provides a transparent, incentive‑compatible reward structure. The authors suggest future extensions such as incorporating nonlinear appliance dynamics, multi‑LSE coordination, and blockchain‑based reward verification to further enhance robustness and market acceptance.