Reaserch On Optimal Scheduling Of Electric Taxi With Real-time Prices

With the prosperity of smart grid, electric vehicles (EVs), emerging as a typical clean energy transportation, achieve a fast development. Meanwhile, the time-varying electricity prices make it possible for the scheduling of EVs charging as well as the interaction between EVs and the distribution grid. Essentially, an EV can be regarded as energy storage. Selecting proper charging slot will contribute to the cost reduction for an individual EV. Based on the charging behavior, EVs can be divided into two categories:private and public. Compared with the private ones, public EVs possess more complicated charging behavior and consume much more power since they keep serving the public all the time. Consequently, it is important to schedule public EVs charging.In this study, we focus on the scheduling of Plug-in Electric Taxies (PET) charging. First-ly, it is assumed that the PET operates ideally with a linear battery model targeting at minimiz-ing charging cost. Considering the limitation of battery capacity, the problem is modeled as a Markov decision process. Due to the randomness of real-time prices, we utilize stochastic dynam-ic programming and propose a backward induction threshold computing algorithm to calculate the thresholds in advance. The PET can make decision by simply comparing the actual price with the threshold. The theoretical analysis of the algorithm is also provided. Secondly, based on the previous scenario, the additional cost and battery loss are considered and a realistic battery model is applied leading to a nonlinear charging process. Then, the previous algorithm is employed here and its effectiveness is verified through simulation. Finally, to obtain a global optimal solution, we utilize the periodicity of mean prices to transfer the problem from infinite time horizon to finite time duration. Then the optimal system cycle, control sequence and initial capacity are obtained by using binary integer programming. The optimality of our algorithm is also proven rigorously. At last, a method for converting an arbitrary initial capacity into the optimal one is proposed and the performance of local optimal algorithm and global optimal is contrasted through MATLAB.