Optimal Operation Strategy of Electric Vehicle Battery Charging and Swapping Stations for Frequency Response Service

A value function approximation based strategy is presented to make intra-day operating decisions of electric vehicle battery charging and swapping stations for battery charging management and frequency response service. The intra-day operation is formulated as Markov decision process to reflect stochastic sequential decision feature. An appropriate piecewise linear function is adopted to represent the expected long-term cost at each time period, which is trained offline by mini-batch gradient temporal difference learning. During intra-day operations, the power baseline and frequency response reserve are obtained online, wherein the trained value function is invoked. Numerical results show that the proposed strategy can ensure the above dual provision. Moreover, the offline training by mini-batch gradient can reach 98.43% of the theoretical optimum.