Smart meter data-driven evaluation of operational demand response potential of residential air conditioning loads

Abstract Residential air conditioning loads (ACLs) are promising demand response (DR) resources with a certain flexibility and controllability that can enhance the operational flexibility and resource utilization of the power grid. The evaluation of DR potential is of great importance for estimating the power reduction, targeting appropriate DR customers, and obtaining constrained boundary for economic dispatch. To better reveal the multi-faceted factors and multi-uncertainties during DR, this paper present a novel definition and evaluation approach of operational DR potential from single customer to large-scale load centers. Aimed at resolving two main issues of evaluation: disaggregation of ACLs component and parameter estimation, an unsupervised load decomposition methodology considering load level difference and seasonal variation is proposed to disaggregate the whole-house energy consumption into ACLs and baseload components non-intrusively. Subsequently, based on the thermal dynamics model of ACLs, a segment analysis methodology is developed, including a constrained regression method for the static parameter estimation and a hybrid method for the dynamic parameter estimation. Data experiments based on ground truth data of residential customers in Austin, Texas, U.S., smart homes in Western Massachusetts and low voltage area in a developed city, Jiangsu province, China, validate the better performance of accuracy and robustness using the proposed methods. The proposed methods are further implemented to four application scenarios, including ACLs consumption behavior learning, operational DR potential analysis, customers targeting for different time-scale DR programs, and day ahead scheduling. These results also demonstrate that great difference in terms of ACLs usage patterns (a total of 19 patterns), DR potential (a maximum of 0.7 kW) results from different DR duration and operational conditions. The DR programs designers and load aggregators are suggested to consider the proposed 5 basic indicators to target customer and select participants in different DR scenarios. And the operational DR potential is more reliable and suitable to generate strategies for day ahead scheduling.