Stochastic Optimization-based Control of Energy Storage Systems

Session: Rapid-Fire Introductions by Poster PresentersDate: Wednesday, October 16, 2019 / 11:35AM - 12:05PM PDTTags: Poster / Performance, Reliability and Technological Advancements


Stochastic Optimization-based Control of Energy Storage Systems

Akhilesh Bakshi, Senior Computational Scientist, Enel X
Srinikaeth Sambandam, Controls Engineer, Enel X

Demand for electricity is ever increasing, especially in densely populated areas along coastal USA. These load centers are often not colocated with power generators, straining transmission and distribution networks significantly. This is particularly the case during peak demand times and, in order to maintain grid stability, utilities are increasingly penalizing power consumption during these times, such as through Time-Of-Use rates, and/or providing incentives for customers to curtail their demand. In this context, energy storage systems (ESS) such as batteries are viable means for minimizing rising costs by reducing and/or controlling the customer’s consumption: for instance, by charging from the grid when delivery or supply charges are relatively low, and discharging for on-site consumption when these are higher (usually during peak demand periods).

In order to maximize customer savings, an ESS controller must be based on an economic optimization model which makes intelligent decisions accounting for multiple value streams including but not limited to delivery and supply charges, ESS constraints and degradation, as well as load forecasts in the near future horizon. While mathematical models for the former are largely deterministic, uncertainties associated with load forecasts can often affect customer savings adversely. For instance, an ESS controller may be unable to mitigate demand charges if site loads forecasted ahead of time are significantly lower than the actuals during a peak demand period.

Stochastic optimization is a class of optimization methods which minimizes a cost function explicitly accounting for the underlying uncertainty. In the context of ESS control in multi-DER (Distributed Energy Resources) applications, a stochastic optimization-based controller makes decisions accounting for uncertainties associated with forecasts of non-controllable power generation (e.g., solar PV) and consumption units. This is unlike deterministic optimization methods, and makes stochastic optimization particularly beneficial in instances of high load or generation volatility.

In this presentation, we will first discuss the basics of stochastic optimization, including forecast uncertainties and their aggregation for economic optimization-based controls. Next, we will analyze high fidelity data from sites where Enel X’s stochastic optimization software is deployed. Through these analyses, we will highlight the advantages of stochastic optimization and its usefulness based on load characteristics such as predictability and volatility. These insights are particularly useful for furthering the development of model predictive controls in multi-DER applications.