Implementation of a load side demand management with intelligent control methods for renewable energy supported electric vehicle charging stations

The widespread use of electric vehicles and the corresponding rapid increase in EV charging stations will significantly increase energy demand in today's world. This increase in demand will lead to overloading in the power system and the emergence of various power quality problems. This increase in energy demand has increased the importance of concepts such as smart srid, smart energy management, and demand-side management in current grid conditions. Approaches such as smart grid and DSM reduce the investment cost required to increase grid capacity by increasing the reliability of the power system and reducing peak demand in the long term. The increase in the use of EV has led to the emergence of a need for developing smart charging strategies for EV charging applications. EVs are not actively used for a significant portion of the day, which makes them suitable options for direct load control applications, one of the demand-side management methods. Direct load control applications can contribute to preventing the occurrence of peak loads by developing smart charging strategies. In this thesis, it is aimed to develop smart charging strategies and reduce the formation of peak charge by using DLC methods with case studies evaluating two different conditions. There are many and different examples in the literature for intelligent control methods for load-side demand management. User satisfaction is also among the parameters that should be evaluated in DLC applications. In the processes of controlling EV charging stations with direct load control applications, it is critical to evaluate user demands and perform load-side demand management according to these evaluations in the processes of performing controls such as load shifting and load shedding. In this thesis, the design and application of smart control method and its effect on the grid load profile for the realization of load side demand management in a renewable energy supported microgrid are investigated. The problem of load-side demand management becomes more complex as the constraints and variables increase to perform load-side demand management. Intelligent load-side demand management decision maker decides whether to apply demand-side methods such as direct load management and which direct load control to apply. Two different case studies have been defined that center the electric vehicle charging station in renewable energy supported microgrids. The two case studies represent, respectively, the selection of appropriate charging speed in multi-charging stations and the evaluation of suitability for direct load control for personal-use electric vehicle charging stations. In the thesis study, In the first case study, the decision-making structure that will make the selection of the appropriate charging speed to prevent the formation of peak loads was designed, and in the second case study, the current load situation was evaluated before the charging process was started, and it was evaluated whether it was suitable for the application of DLC methods in the charging process. For both case studies, current state of charge, next trip distance, battery capacity and time to leave the charging station will be used. The other data to be used in decision-making processes is instantaneous consumption value of the microgrid. In the study, data sets were created in line with certain restrictions to be used in machine learning training processes and these data sets were labeled with decision algorithms determined for each case study. Machine learning classification methods were used for two different models in order to perform load-side demand management without causing a negative situation for the user with the created data sets. For the first case study in the simulation processes, the data set was classified with an accuracy of 99.1% using an artificial neural network-based decision maker and it was seen that unnecessary fast charging station usage was prevented. For the second case study, the bagging tree method, which is one of the ensemble learning methods, was classified with an accuracy of 98.7% and it was seen that the load management decision could be made directly without causing any negativity for the users in the EV charging processes. For both case studies, grid load profile comparisons were made before and after the implementation of the decisionmaking structure and when renewable energy support was added. It has been seen that the proposed method in the thesis study can be applied in charging stations with more than one EV charging station and in domestic uses with a single charging station. In addition, with the thesis, a practical demand management system has been developed that minimizes the negative effects of EV charging units on the grid and takes into account load-side changes.