Contents lists available at ScienceDirect Electrical Power and Energy Systems journal homepage: www.elsevier.com/locate/ijepes A scalable and robust approach to demand side management for smart grids with uncertain renewable power generation and bi-directional energy trading Ren-Shiou Liu ⁎ , Yu-Feng Hsu Dept. of Industrial and Information Management, National Cheng Kung University, Taiwan ARTICLE INFO Keywords: Smart grid Demand side management Real-time pricing Robust optimization Column & constraint generation ABSTRACT This paper investigates the energy cost minimization problem for smart grids with distributed renewable energy resources. Unlike earlier research studies that either have assumed all the appliance jobs are interruptible or power-shiftable and that the electricity prices as well as the availability of renewable resources are known, this paper focuses on more challenging scenarios in which appliance jobs are non-interruptible and non-power- shiftable, the electricity prices vary with the overall load of the entire grid in real-time, and the renewable power generation is uncertain. Because home solar systems are widely available, this paper assumes that each consumer in the grid can have a photovoltaic system and a side battery. Collected solar energy can be used to meet a consumer’s individual power demand, stored in the battery for future use, or sold back to the grid during peak hours to lower electricity bills and the overall load on the entire grid. To solve this problem, a two-stage robust optimization model is proposed, and the C&CG method is utilized to solve it. However, to solve the problem more efficiently when the number of consumers and appliance jobs is large, a second approach called SRDSM is proposed. The SRDSM algorithm consists of two parts: The first part is a job scheduling algorithm that minimizes electricity costs for all consumers. The second part is a power management algorithm based on dynamic pro- gramming that reduces the energy cost further by utilizing renewable energy. The numerical results show that, although the C&CG method produces optimal solutions, the SRDSM algorithm is much more scalable and effi- cient when the problem size is large. 1. Introduction With climate change and a rising awareness of environmental pro- tection, legacy power grids face many challenges. For example, the general public is becoming increasingly opposed to the use of fossil fuels because they are one of the biggest sources of air and water pol- lution. Legislations and/or regulations are also becoming increasingly restrictive regarding the construction and operation of new grid facil- ities. This not only increases the cost of power generation but may also lower grid adequacy because utility companies are forced to set aside more of their budget for cleaning the pollutants released from burning fossil fuels and thus have less money to invest in capacity expansion. In contrast to fossil fuels, generating electricity from renewable energy produces little to no air and water pollutants or global warming emis- sions. As such technologies have matured, they have become one of the most effective and prevalent tools for environmental protection. Increasing numbers of consumers are installing renewable power gen- eration systems locally in their homes. Several states and local governments in the US are also advancing policies to encourage greater deployment of renewable technologies. Take the City of Lancaster, California, for instance. It has required newly-built houses to feature a home photovoltaic system since January 1st of 2014 [1]. However, this greatly changed the centralized power generation and dispatching model of legacy grids. In addition, each form of renewable energy has its own disadvantages. For example, wind and solar energy are inter- mittent in nature since they rely on the weather for their source of power. Slow wind or abundant rain can result in significant reductions in energy production. Using batteries can mitigate the problem to some extent, but the usage of battery power must be carefully planned. In response to these challenges, a significant amount of effort has been devoted to the development of technologies for grid moderniza- tion, which forms the foundation of smart grids. A smart grid can be defined as an electricity network that can intelligently integrate the actions of all consumers connected to it – generators, consumers, and those that do both – in order to efficiently deliver sustainable, in- expensive, and secure electricity [2]. By adopting the advanced https://doi.org/10.1016/j.ijepes.2017.11.023 Received 21 June 2017; Received in revised form 4 October 2017; Accepted 19 November 2017 ⁎ Corresponding author. E-mail addresses: rsliu@mail.ncku.edu.tw (R.-S. Liu), h34005038@mail.ncku.edu.tw (Y.-F. Hsu). Electrical Power and Energy Systems 97 (2018) 396–407 0142-0615/ © 2017 Elsevier Ltd. All rights reserved. T