Electric Vehicle (EV) batteries are generally replaced as soon as their capacity declines below 70–80% of the initial capacity. However, they can be still useful for stationary applications. Additionally, a second-use approach reduces the ecological footprint. Stationary batteries could be connected to the electrical grid to compensate for supply and demand imbalances.
Battery storage systems are one technology investigated as a means for autonomous demand side management in the scope of the energy model library of the Josef-Ressel center. The operation strategy investigate d can be realized via an optimization procedure, which relies on a one-way communicated pseudo-cost function (PCF). The procedure searches for the optimal charge, discharge and idle states (switch states) of the battery by minimizing the PCF while keeping the battery’s state-of-charge (SOC) within its operational bounds. Autonomous control, which also has been successfully tested for domestic hot water heaters (cf. Kepplinger, P., Huber, G., & Petrasch, J. (2015). Autonomous optimal control for demand side management with resistive domestic hot water heaters using linear optimization. Energy and Buildings, 100, 50-55.), is applied to battery storage systems.
In our recent publication, (Faessler, B., Kepplinger, P., & Petrasch, J. (2017). Decentralized price-driven grid balancing via repurposed electric vehicle batteries. Energy, 118, 446-455.), different nonlinear and linear optimization approaches have been compared with respect to runtime and optimality. A sequential quadratic programming (SQP) approach was used for nonlinear optimization; a dynamic programming (DP) approach as well as an integer linear programming (ILP) approach were used for linear optimization.
The results obtained using historic 2015 day-ahead data as the driving PCF showed, that the SQP routine with its higher model accuracy leads to the highest earnings. The ILP approach provided a good approximation and requires the lowest computing time. The DP approach performed marginally worse than ILP. Furthermore, ILP was used to estimate the potential of autonomous grid balancing using a model of a physical EV battery storage system in the period 2003–2015 based on historical day-ahead stock market price data as PCF. It showed that a profitable operation is not possible since variations in price are currently too low. Simulations reveal a strong correlation between the earnings and the variation of the PCF during the same period. Higher price fluctuations lead to more dynamic battery operation. This results in higher earnings, efficiencies, and shorter idle times, which indicates a better storage system utilization.
A scaling of the modeled battery storage system should show if higher earnings per battery capacity are possible. To this end, the capacity to power ratio is varied. The results showed that the earnings per capacity exhibit a maximum since the system loses the freedom to realize all optimal switch states for low capacity to power ratios. For very high capacity to power ratios, the storage capacity is not fully utilized. Generally, large battery storage systems are preferable since relative thermal losses decrease as the surface to volume ratio goes down.
Due to the structure of the proposed optimization routines and simulation approaches, all calculations are well suited to be processed on the distributed execution framework (DEF), developed in-house by the team of the Josef-Ressel center.