Optimizing Customer Selection for Sustainable Demand Response

Demand Response (DR) is a widely used technique to minimize the peak to average consumption ratio during high demands. An effective DR scheduling algorithm should minimize the curtailment error the difference between the targeted and achieved curtailment values to minimize the costs to the utility provider and maintain system reliability. Several polynomial time heuristics have been proposed in the literature to achieve this goal, however their accuracy can be extremely low. We argue that minimizing the error alone is not enough as peaks can be concentrated in some intervals while consumption being heavily curtailed in other intervals. In this paper, we leverage the availability of smart meters to provide fine grained data and customer control and formally develop the notion of Sustainable DR as a solution that distributes the curtailment evenly across the DR event. We formulate both Traditional DR and Sustainable DR problems as Integer Linear Programs. For both problems, we first provide a very fast √ 2-factor approximation algorithm. We also propose a Polynomial Time Approximation Scheme (PTAS) for approximating the curtailment error to within an arbitrarily small factor of the optimal. We develop a novel ILP formulation that solves the sustainable DR problem while explicitly accounting for strategy switching overhead as a constraint. We perform experiments using real data acquired from a university’s smart grid and show that our sustainable DR model achieves near exact results (error in the order of 10−5). INTRODUCTION Recent technological advances have transformed traditional power grids to complex cyber-physical systems [1], [2]. The widespread use of bi-directional smart meters, in addition to reporting energy consumption, allows remote monitoring and intelligent grid control. Utility providers now have several tools at their disposal to dynamically meet energy demand while ensuring the reliability of the power grid. Reliable operation of a power grid requires utilities to constantly match (fluctuating) energy supply with (fluctuating) load. Demand levels of customers fluctuate with peak demands concentrated in some portions of the day. During such periods, the demand might exceed the generation capacity of the utility forcing them to buy energy from the spot market at high rates. Demand Response (DR) is a standard technique used by utilities to mitigate energy supply-demand mismatch. Customers are incentivised to enroll in the program and curtail their energy consumption during peak load times which is signaled by a DR event. Each customer is provided with various strategies to reduce consumption. e.g. a customer can reduce the air conditioning or turn off some number of lights. For each customer-strategy pair, the utility estimates the load curtailment. This information is then used to decide the optimal strategies that will achieve the desired curtailment value. We observe that Demand Response can be implemented in two modes. Sometimes fine grained control of customer strategies may not be available, for example, due to a lack of smart meters for control and the necessity of manual intervention for adjusting strategies (such as manually turning on/off the AC). Under such circumstances, load curtailment during a DR event can be achieved by optimally selecting customer-strategy pairs to curtail demand by the desired amount for the entire peak demand period. Conversely, when fine grained control is available, the entire peak period can be divided into smaller intervals where customer strategies can be micro-adjusted. Note that implementing the first approach (which we label Traditional DR) might produce customer strategy assignments which achieve the desired total curtailment value by aggressively curtailing demand over a few small intervals. This could create peaks and valleys of demand over intervals (while still technically achieving DR objectives). Such demand peaks could possibly exceed the instantaneous generation capacity, forcing the utility to pay for additional procurement of energy. This motivates our Sustainable DR approach: To leverage the availabilty of fine grained smart meter data and customer control by evenly distributing curtailment over the entire time period.