A Linear Value Function Approximation Approach to the Battery Storage Optimization Problem with Layered Markov Modeling of Renewables

Abstract

In this paper, we use a linear value function approximation method for optimizing battery storage on a grid that makes use of both renewable and nonrenewable electricity technologies, integrating a new stochastic model for wind and solar. Because of the variability of wind and solar resources, when battery storage is necessary to provide reliability to the grid. In this model, we are solving the storage management problem. The newly developed stochastic model for wind and solar not only tries to capture the distribution of the output below and above the forecast, but also the distribution of the amount of time during which the output is different from the forecast. By using this newly developed stochastic model for renewable resources, the sample paths used to determine the optimal policy can provide a more accurate representation of renewable behavior and the relationship between their behavior and their forecasts. This model can then be used to evaluate investment decisions across different electricity generating technologies, considering the need for grid reliability.