Research Information System
Thesis Type: Doctorate
Institution Of The Thesis: Gazi Üniversitesi, Fen Bilimleri Enstitüsü, Turkey
Approval Date: 2018
Student: ALPER KEREM
Supervisor: ALİ SAYGIN
Abstract:The wind has an intermittent structure that can show instantaneous change. For this reason, it is so hard to make the wind speed and wind power forecasting correctly. The wind speed and wind power forecasting are vital due to the reasons such as the determination of the position of the wind fields, preliminary shaping of energy unit cost, efficient energy investments, network security and so on. For this purpose, it is aimed to develop a fast and stable prediction model with high accuracy for wind speed and wind power prediction studies. The study was started by training Artificial Neural Networks (ANNs) with some meta-heuristic algorithms (Evolutionary Strategy (ES), Genetic Algorithm (GA), Ant Colony Optimization (ACO), Probability Based Incremental Learning (PBIL), Particle Swarm Optimization, PSO) and Radial Movement Optimization (RMO)) in the literature. The success of each model (ANNs trained with ES, ANNs trained with GA, ANNs trained with ACO, ANNs trained with PBIL, ANNs trained with PSO, ANNs trained with RMO) is recorded in graphs. In order to make the closest prediction to the accurately and to increase system stability a new meta-heuristic hybrid model was developed using PSO and RMO, and the training of YSA was performed with this new model. In the application part, actual data provided from two different Wind Measurement Stations (WMS-1: 63 m and WMS-2: 60, 3 m) was used. Data sets were analyzed in two main groups for wind speed and wind forecasting studies, and each group was divided into different scenarios. In order to make easy use of the created model, the user interface was designed in MATLAB/ App Designer (R2017b) and the entire system was visualized. The success of ANNs trained with PSO + RMO has been compared separately with the success of each of the other hybrid models (ANNs trained with ES, ANNs trained with GA, ANNs trained with ACO, ANNs trained with PBIL, ANNs trained with PSO, ANNs trained with RMO). According to analyses results, it has been found that the individual performance of the proposed meta-heauristic hybrid model (ANNs trained PSO+RMO) is much more successful and reliable than the other hybrid models that were performed.