| Abstract | This thesis focuses on the development of electric vehicle (EV) charging protocols under a dynamic environment using artificial intelligence (AI), to achieve Vehicle-to-Grid (V2G) integration and promote automobile electrification. The proposed framework comprises three major complementary steps. Firstly, the DC fast charging scheme is developed under different ambient conditions such as temperature and relative humidity. Subsequently, the transient performance of the controller is improved while implementing the proposed DC fast charging scheme. Finally, various novel techno-economic scenarios and case studies are proposed to integrate EVs with the utility grid.
The proposed novel scheme is composed of hierarchical stages; In the first stage, an investigation of the temperature or/and relative humidity impact on the charging process is implemented using the constant current-constant voltage (CC-CV) protocol. Where the relative humidity impact on the charging process was not investigated or mentioned in the literature survey. This was followed by the feedforward backpropagation neural network (FFBP-NN) classification algorithm supported by the statistical analysis of an instant charging current sample of only 10 seconds at any ambient condition. Then the FFBP-NN perfectly estimated the EV’s battery terminal voltage, charging current, and charging interval time with an error of 1% at the corresponding temperature and relative humidity. Then, a nonlinear identification model of the lithium-polymer ion battery dynamic behaviour is introduced based on the Hammerstein-Wiener (HW) model with an experimental error of 1.1876%.
Compared with the CC-CV fast charging protocol, intelligent novel techniques based on the multistage charging current protocol (MSCC) are proposed using the Cuckoo optimization
algorithm (COA). COA is applied to the Hierarchical technique (HT) and the Conditional random technique (CRT). Compared with the CC-CV charging protocol, an improvement in the charging efficiency of 8% and 14.1% was obtained by the HT and the CRT, respectively, in addition to a reduction in energy losses of 7.783% and 10.408% and a reduction in charging interval time of 18.1% and 22.45%, respectively. The stated charging protocols have been implemented throughout a smart charger. The charger comprises a DC-DC buck converter controlled by an artificial neural network predictive controller (NNPC), trained and supported by the long short-term memory neural network (LSTM). The LSTM network model was
utilized in the offline forecasting of the PV output power, which was fed to the NNPC as the training data. The NNPC–LSTM controller was compared with the fuzzy logic (FL) and the
conventional PID controllers and perfectly ensured that the optimum transient performance with a minimum battery terminal voltage ripple reached 1 mV with a very high-speed response
of 1 ms in reaching the predetermined charging current stages.
Finally, to alleviate the power demand pressure of the proposed EV charging framework on the utility grid, a novel smart techno-economic operation of an electric vehicle charging station (EVCS) in Egypt controlled by the aggregator is suggested based on a hierarchical model of multiple scenarios. The deterministic charging scheduling of the EVs is the upper stage of the model to balance the generated and consumed power of the station. Mixed-integer linear programming (MILP) is used to solve the first stage, where the EV charging peak demand value is reduced by 3.31% (4.5 kW). The second challenging stage is to maximize the EVCS profit whilst minimizing the EV charging tariff. In this stage, MILP and Markov Decision Process Reinforcement Learning (MDP-RL) resulted in an increase in EVCS revenue by
28.88% and 20.10%, respectively. Furthermore, the grid-to-vehicle (G2V) and vehicle-to-grid (V2G) technologies are applied to the stochastic EV parking across the day, controlled by the aggregator to alleviate the utility grid load demand. The aggregator determined the number of EVs that would participate in the electric power trade and sets the charging/discharging capacity level for each EV. The proposed model minimized the battery degradation cost while maximizing the revenue of the EV owner and minimizing the utility grid load demand based on the genetic algorithm (GA). The implemented procedure reduced the degradation cost by an average of 40.9256%, increased the EV SOC by 27%, and ensured an effective grid stabilization service by shaving the load demand to reach a predetermined grid average power across the day where the grid load demand decreased by 26.5% (371 kW). |
|---|