Nowadays, the DC distribution network has been becoming an attractive topic in the power system because of its advantages over the conventional AC distribution system. The DC distribution system can easily host the Renewable Energy Sources (RES), such as photovoltaic panels (PV), gas-biomass, or fuel cell (FC), because they produce DC power. The RESs comfortably connect to the DC distribution system without any synchronization because of the absence of frequency, phase angle, and reactive power. Furthermore, the development of DC appliances in residential buildings is also one of the critical factors for the development of the DC distribution system. By using DC networks, the customer can reduce the power loss conversion caused by the power-electronic devices by up to 33% compared to the AC system. Without frequency, reactive power, and phase angle, the control and operation in the DC distribution system are more comfortable and more straightforward than the AC system. Therefore, the DC distribution system is the future distribution network, which defined as a low-voltage DC (LVDC) distribution network whose rated voltage of 1500VDC. The LVDC distribution system consists of loads and distributed energy resources (DERs). However, along with the benefits compare to the AC distribution system, LVDC systems are still facing the challenges of the increasing electricity demand and RES. LVDC has a low-rated voltage of 1500VDC. The small changes in loads and PVs can cause voltage problems in the distribution line, which is one of the most significant issues in the distribution system. According to the voltage quality standard, the voltage profiles at the customer''s point must be maintained to the ±5% of the rated voltage. Moreover, loads, EV charging demand, and RES are known as the uncertainty factors because they are affected by customer''s behavior and the weather, respectively. Hence, voltage problems become more serious when considering these factors
This dissertation presents the optimal voltage control and operation using predictive control algorithms to improve the resiliency of the voltage profiles considering the uncertainties in loads, PV, and unpredicted electric vehicles. The optimal process is divided into two-stage: 1) Day-ahead optimal scheduling algorithm and 2) real-time operation using predictive control algorithms. The day-ahead scheduling algorithm of the DC distribution system is developed so that determining the economic operation points of the controllable DER considering the day-ahead prediction of load and solar irradiance. The scheduling algorithm produces the optimal hourly power of DER and ON/OFF statement of the DGs. The real-time operation using model predictive control (MPC) algorithms is to determine the optimal plan in real-time action. By using the receding horizon concept, the online-optimization is run in every interval to find the economic operation points of controllable DER. The MPC model considers the uncertainties in load and PV, such as prediction errors in multi-step sliding window prediction or fluctuation of RES and electricity demand, or the unpredicted electric vehicle participation. The modification of the control horizon is proposed to optimize the charging and discharging of electric vehicle (EV) participation. The dissertation also produces an incentive for EV charging and discharging actions. The EVs that used for support-voltage in the intervals can receive the discount for charging in the next interval time. At the end of the dissertation, an LVDC distribution network consists of loads, and DERs are used to verify the performance of the proposed control algorithm. DERs include PV, electric vehicle supply equipment (EVSE) with the real-time EV arrival, the Energy storage system (ESS), and distributed generation (micro-combined head and power). The forecasting in load and solar irradiance is introduced using long-short term memory based recurrent neural network (LSTM-RNN). The forecasting results are compared to the previous work. Forecasting algorithms, day-ahead optimal scheduling, and real-time operation using model predictive control (MPC) framework are developed by MATLAB programming following scenarios and real data obtained from the public dataset on solar radiation research laboratory (SRRL) and US Department of Energy