Online algorithms for storage utilization under real-time pricing in smart grid

With the rapid proliferation of the advanced metering infrastructure, the smart grid is evolving towards increased customer participation. It is now possible for a utility to influence the customer demand profile via demand side management techniques such as real-time pricing and incentives. Energy storage devices play a critical role in this context, and must be optimally utilized. For instance, the peak power demands can be shaved by charging (discharging) the batteries during periods of low (high) demand. This paper considers the problem of optimal battery usage under real-time and non-stationary prices. The problem is formulated as a finite-horizon optimization problem, and solved via an online stochastic algorithm that is provably near-optimal. The proposed approach gives rise to a class of algorithms that utilize the battery state-of-charge to make usage decisions in real-time. The proposed algorithms are simple to implement, provably convergent for a wide class of non-stationary prices, easy to modify for a variety of use cases, and outperform the state-of-the-art techniques, such as those based on the theory of Markov decision processes or Lyapunov optimization. The robustness and flexibility of the proposed algorithms is tested extensively via numerical studies in MATLAB and real time digital simulator (RTDS).