Design Parameter Modeling of Solar Power Tower System Using Adaptive Neuro-Fuzzy Inference System Optimized with a Combination of Genetic Algorithm and Teaching Learning-Based Optimization Algorithm

Tutkimustuotos: Lehtiartikkelivertaisarvioitu

Tutkijat

  • Ali Khosravi

  • Mohammad Malekan
  • J. J. G. Pabon
  • Xiaowei Zhao
  • Mamdouh El Haj Assad

Organisaatiot

  • University of Sharjah
  • Universidade Federal de Itajubá
  • The University of Warwick
  • Aarhus University

Kuvaus

Determining the optimal sizing of a solar power tower system (SPTS) with a thermal energy storage system is subject to finding the optimum values of design parameters including the solar multiple (SM), design direct normal irradiance (DNI) and thermal storage hours. These design parameters are determined for each station separately and have remarkable effects on the thermo-economic performance of the system. This paper aims to demonstrate how artificial intelligence (AI) techniques may play an important role in addressing the above-mentioned need and help determine the optimum design parameters for different stations. For this purpose, we developed a thermo-economic model of a 100 MW SPTS with a molten salt storage system for five stations (two stations in India, and one each in Bangladesh, Pakistan, and Afghanistan). A method-based AI is utilized in this paper to ascertain the design parameters of the system. Additionally, a novel hybrid method based on adaptive neuro-fuzzy inference system optimized with a combination of genetic algorithm and teaching-learning-based optimization algorithm (ANFIS-GATLBO) is employed. The input parameters are latitude, longitude, design point DNI and SM, while the annual energy produced, levelized cost of energy and capacity factor are the target variables. The results of the study show that although the annual energy produced by SPTS rises by increasing the SM and decreasing design point DNI, optimum design parameters should be determined by the economic factors. In addition, it was found that the ANFIS-GATLBO method used in this study successfully predicted the targets with a correlation coefficient close to 1.

Yksityiskohdat

AlkuperäiskieliEnglanti
Artikkeli118904
Sivumäärä20
JulkaisuJournal of Cleaner Production
TilaSähköinen julkaisu (e-pub) ennen painettua julkistusta - 17 lokakuuta 2019
OKM-julkaisutyyppiA1 Julkaistu artikkeli, soviteltu

ID: 37714352