Assessing the Effect of Storage Devices and a PHEV Cluster on German Spot Prices by Using Model Predictive and Profit Maximizing Agents

In this paper, the German electricity wholesale market is modeled by individual profit maximizing agents. The agents use a learning approach called “model predictive bidding” to choose bidding curves for the upcoming day which are sent to the spot market. The chosen bidding curves thereby maximize the agents’ expected discounted profits. In contrast to traditional agent learning approaches such as Q-learning or learning classifier systems, the learning approach in this paper is predictive and model based. This predictive approach allows to model storage agents, which anticipate higher and lower future spot prices to charge and discharge their storage. Simulations with different types of storage agents are carried out to assess the impact of those agents on spot prices, price volatility and market-power. The storage agent in one simulation uses a high power output whereas the storage agent in a second simulation has a high storage capacity. Furthermore, a scenario is simulated in which a cluster of PHEVs is simulated as an agent who can be charged or discharged via the electric grid and has a variable charging/discharging power capacity depending on the amount of PHEVs connected to the grid and their current state of charge. In contrast to electric vehicles, PHEVs can drive with either electricity or gasoline. This introduces a higher flexibility in charging/discharging these vehicles. The paper shows that the temporal variable charging/discharging rate of the PHEV cluster results in less market-power by the PHEV cluster compared to a storage device and to a more favorable market output.