Load forecasting and load management in smart grid

Smart grid have become a promising solution since the current grid infrastructure is inadequate to meet rapidly renewed, developing, growing technological and industrial needs. The key elements of the smart grid, such as advanced metering infrastructure, bi-directional communication and distributed production resources, challenge current energy management approaches. Although conventional power systems are well-established to address the varying load demand response, the additional variability and uncertainty of renewable energy sources incorporated into the system with smart grids pose a significant threat to network stability. In this study, it has been focused on developing the management strategy of smart grids that can be used by a power system operator, reducing power losses and increasing network operational speed. Three of the many problems that are expected to be solved with the introduction of smart grids have been addressed. As a result of combining these three problems, the proposed solution method provides a new insight into the solution of network problems. Firstly; load forecasting which has a very important role in power system management is discussed. The results of three different prediction models, whose reliability has been repeatedly proven and frequently studied in the literature, have been compared. The same data set was used in all of these models. estimates are made for short periods of time, such as one hour. In particular, attention was paid to the different logic of the prediction models. Regression Analysis Method which uses the statistical solution infrastructure, Artificial Neural Networks (ANN) method based on brain neural structure and artificial neural fuzzy inference system (ANFIS) method based on fuzzy logic approach have been preferred. ANFIS was the most successful method in these three methods. ANFIS was chosen as the load prediction model in the proposed solution method. Secondly; Optimal Reactive Power Dispatch (ORPD), which plays an important role in the economic and safe operation of power systems, has been studied. ORPD is a complex optimization problem that involves a nonlinear objective function and constraints. The main objective of ORPD is to determine the optimal settings of all control variables, such as generator setpoints, transformer step adjustment and reactive power compensation output. For this purpose, multiple solution methods have been presented by the researchers for the solution of the mentioned problem. In this thesis, meta-heuristic algorithmic methods which are successful in solving linear problems are used. The results of tree seed algorithm (TSA) which previously has never been adapted to the ORPD problem and shuffled frog leaping algorithm (SFLA) were compared with the methods studied in the literature. As a result of the comparisons, TSA which has a better success was preferred to be used in the proposed solution method. IEEE-30 and IEEE-118 bus systems were used to test the success of optimization algorithms. As the third; In the power system, if the load demand is higher than the energy production, the subject of load shedding, which is frequently applied, is discussed. Load shedding is preferred because of its rapid response and easy adaptation to sudden changes. In order to ensure the stability of power systems, optimized load shedding methods have gained importance in recent years. Therefore, the current method of intelligent load shedding (ILS) has been studied in this thesis. In this method, loads are divided into four classes according to their importance. Each class is divided into three groups in order of importance. Loads are switched on or off in order of their importance. According to estimated production value, loads that to be used in the following period are pre-determined and stored. When the actual production information was transmitted to the system, the operations were continued over the stored loads. This saved the system from wasting extra time. All methods used in this thesis are brought together for a single purpose, a new solution approach is presented. The proposed method was developed by using ANFIS, TSA and ILS methods. Therefore, the method can be defined as ANFIS+TSA+ILS hybrid model. If we take the whole system into consideration, the total targeted objectives are listed as realistic production forecasting, minimum power loss and increased operating capability.