Modeling and Control of Renewable Energy Sources with Energy Storage System for Smart Grid Application

Even though electricity generated by fossil fuels (coal, natural gas, petroleum, and other gases) is recognized as one of the largest sources of greenhouse gas emissions, until 2016, 65 percent of the electricity generated in the United States was from these sources. Recently, a more environmental friendly energy source, renewable energy is increasingly incorporated with the electricity grid. The U.S Department of Energy mandates that by 2030, 20 percent of the total energy consumption is produced from wind energy. A major challenge of renewable energy sources is the fact that they are non-dispatchable ones, which cannot be dispatched freely to meet the electricity demands. Thus, massive energy storage system (ESS) is installed to the grid to meet the energy demand and shift the production from low-energy value periods to high-energy value periods to maximize profits.

There are two ways to improve the optimal cost of an electric grid with an ESS. The first one focuses on the placement and sizing of the ESS on the grid, and the second one focuses on the operation of the grid. These two approaches can be studied independently, but they can also be applied simultaneously to further improve efficiency. Adeodu and Chmielewski applied Economic Model Predictive Control (EMPC) and Economic Linear Optimal Control to solve the Optimal Sizing and Placement (OSP) of ESS. The focus of our PURE project in this Fall 2017 uses this OSP foundation to establish an appropriate grid operating policy. The tools that we use are EMPC and Multistage Stochastic Programming (MSP).

EMPC uses a state-space model to predict the future values of the cost of electricity generated from the current load demands and the realizations of the renewable energy generated randomly. However, since the expected values of the disturbances are zeros, EMPC ignores forecasts, so the optimal solutions can be ameliorated further. MSP allows the users to specify the scenarios of disturbances; hence it provides better controlling policy than EMPC. The disadvantage of MSP is the problem size increases geometrically with the size of future predicting horizon. To diminish the computational time of the MSP, some decomposition method such as Nested Bender Decomposition is applied.

Additionally, determining disturbance trends is critical to effective control. The disturbances, which are the load demands and the generated renewable power, are modeled using third-ordered shaping filter. The probability of each disturbance’s realization is quantified by an effective counting method. This filter simulation is run for a long time and the occurrence of each instance is counted, e.g Given 3 bins: low, mid, and high, how many times the sequence Low-Medium-High, High-High-High, High-Low-High... occur, etc.

The purpose of this project is to identify if MSP is an appropriate framework for economic dispatch on networks with energy storage. Additionally, this project will certify the hypothesis that Infinite Horizon EMP can be a reasonable approximation for the MSP.