Analysis And Dynamic Dispatch Of Energy Storage Systems In Electricity Markets Under Uncertainty

With the increasing penetration levels of renewable energy resources in power systems, to realize efficient energy management, energy storage systems have become indispensable. This research aims at analyzing the impact of energy storage systems on electricity markets, especially the impact on electricity prices, and developing strategies to dynamically dispatch the available storage resources for energy management. Marginal generation units as the generators that determine the system-wide electricity prices are theoretically analyzed in this research. The electricity price of any node in the system is approximately an affine combination of the bidding prices of the marginal generation units if the power losses are ignored and the generation shifting factors are considered constant. By allocating energy storage at a node, the capacity and charge (or discharge) power have an aggregate impact on the change of the price of that node. Based on the analysis, a chance-constrained optimization model has been developed to determine the optimal storage capacities and their allocation in a power system. Unlike common commodities, the generation and consumption of electricity should be balanced all the time. Net power demand may change unanticipatedly, while electrical power generation is not flexible enough to follow the change of demand. A two-settlement system of electricity markets becomes a critical solution to this dilemma. In the whole sale markets, most of the power is balanced in the day-ahead market based on day-ahead prediction, and only a fraction of power is balanced in the real-time market. However, intermittent renewable energy resources are less predictable than traditional load profiles. The day-ahead scheduling is unlikely to compensate very well the net load profiles influenced by renewable energy such as solar or wind in a precautionary manner. In this research, we developed dynamic dispatch strategies for two situations. In the first one, we consider the cooperation between a flexible load and a wind farm, where the flexible load reacts like an energy storage system according to the changes of wind generation. They participate in the whole sale market as well as bilateral transactions out side the market. They determine their selling prices or power levels in a dynamic manner so that to reach an equilibrium of the benefits for both of them. In the second case, we study a power system node that has load and renewable energy resources connecting to it. To mitigate the difference between day-ahead scheduling and real-time dispatch of net power consumption of the node, an energy storage system is installed at the node and the decisions on charge/discharge power are made in a real-time market dynamically. Unlike model predictive control, the long-term performance of the strategy is considered through the Lyapunov optimization technique. The algorithm is simple for application, since it is based only on observed events, and has a provable worst-case performance guarantee. Simulation study cases show that the proposed strategy has a better performance than model predictive control in terms of long term costs, no matter the study case is based on real-world data or artificially created independent identically distributed data