Neural network metamodelling in multi-objective optimization of a high latitude solar community

Research output: Contribution to journalArticleScientificpeer-review

Researchers

Research units

  • Michigan State University

Abstract

A solar community of 100 residential houses was optimized for Finnish conditions with the aim of achieving a 90% solar fraction for both space heating and domestic hot water. Optimization was done using a novel method based on neural network metamodelling and compared to the standard NSGA-II genetic algorithm. Compared to NSGA-II, the new method obtained a larger hypervolume by finding better solutions both in the center and edge of the non-dominated front. The combined non-dominated front of both methods was better than either one separately. The performance target was achieved as the optimal solar community designs had heating solar fractions ranging from 64% to 95%.

Details

Original languageEnglish
Pages (from-to)323-335
Number of pages13
JournalSolar Energy
Volume155
Publication statusPublished - 1 Oct 2017
MoE publication typeA1 Journal article-refereed

    Research areas

  • Neural network, Seasonal storage, Simulation-based optimization, Solar assisted heat pump, Solar community, Solar district heating

ID: 14711275