| Recently, traditional vehicles are likely to be replaced by electric vehicles (EV) which are characterized by high energy efficiency and low emission. Plug-in electric taxis (PET) and oth-er public EVs have gained a rapid development worldwide due to government support. PETs keep running and consume much more power. Their huge demand for electricity and unscheduled charging behavior may generate new peak periods. However, real time price (RTP) is regarded as an efficient tool to schedule the charging behavior of PET fleet, which is able to avoid the possible peaks and benefits to the power grid through peak load shifting and demand response.As the foundation of fleet scheduling, we first introduce the threshold-based scheduling of s-ingle PET which helps optimize its charging cost or operating profits. Based on the characteristics of RTP and PET, this algorithm is aimed to solve the Markov model of charging scheduling of single PET and obtains a series of thresholds. By means of comparing the thresholds with RTPs, reasonable charging decisions will be made to reduce the charging cost or increase the operating profits of single PET. The thresholds which indicate the future expected charging cost or operating profits are able to distinguish and control different PETs with diverse electric quantities. In addi-tion, threshold-based scheduling is fit for PETs in the fleet to make their own charging decisions and will improve the schedulability of PET.Afterwards, on the assumption that there exists a special electricity market for PETs and all the PETs follow the threshold-based scheduling, we aim at scheduling the charging behavior of PET fleet by pricing in order to track the target load. Considering the properties of thresholds, the PET fleet is classified in several types. Therefore, the dynamics of PET fleet is modeled as a system dynamic function whose state variables are the numbers of PETs in different types and control variables are the proportion of charging PETs in the corresponding types. Finally, the load scheduling of PET fleet is formulated as a nonlinear integer optimization which is hard to solve. Then, an algorithm employing genetic algorithm (GA) is proposed to obtain the optimal charging decisions. Afterwards, optimized RTPs which are relative to the optimal charging decisions are made to schedule the PET fleet to track the target load. Numerical simulations verify the proposed method.Next, considering that the owners of PETs always have their own charging decisions, we focus on scheduling the charging behavior of PET fleet which follows a probabilistic decision model through pricing. Based on the characteristics of PET fleet, the convex optimization is introduced to obtain the realizable load which could track the target load. Assume that all the PETs follow the probabilistic decision model which is a function correspondent with the difference between threshold and price. Then, an online decision maker is proposed to obtain the proper RTPs which will lead the PET fleet to track the realizable load profile. The algorithms of online decision maker include fitting the probabilistic decision model using real data and obtaining the proper RTP through dichotomy, respectively. At last, the performance of proposed algorithms in different scenarios is validated on MATLAB platform. |
PDF Full Text Request