Scholarlink

<p>This paper proposes a stochastic optimization method for sophisticated office building design and compares an Improved Snow Geese Algorithm (ISGA) with basic SGA, Gray Wolf Optimizer (GWO), and Manta Ray Foraging Optimization (MRFO). The methodology is validated in three scenarios: (i) CEC-2019 benchmark functions (Case I), (ii) a deterministic detailed building design (Case II), and (iii) a stochastic building design that models zone lighting energy consumption uncertainty with the 2m + 1 Point Estimate Method (PEM) (Case III). On CEC-2019, ISGA has the lowest mean on all but one function and the lowest (second-lowest) standard deviations; Friedman's test ranks ISGA best overall, consistent with boxplots (lower medians and narrower interquartile ranges) and with faster and more stable convergence. In the building design case, ISGA produced the most efficient and robust performances: In Case II, the optimum yearly energy consumption had 129.3305&#xa0;kWh/m<sup>2</sup>a, the average is 133.0116&#xa0;kWh/m<sup>2</sup>a, and the standard deviation was 0.0864, better than SGA, GWO, MRFO, and a previously published baseline. Adding zone lighting energy consumption uncertainty to Case III adds annual energy for all methods (ISGA + 7.68%, SGA + 8.06%, GWO + 8.17%, and MRFO + 8.53%), yet ISGA continues to exhibit the lowest energy and most stable performance. The gains stem from ISGA's chaos-augmented search—chaotic initialization, step perturbations, and adaptive tuning—sustaining diversity, decreasing premature convergence, and optimizing exploration–exploitation trade-off.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Improved Snow Geese Algorithm for Building Energy Stochastic Optimization Using an Uncertainty-aware Point Estimate Method

  • 期刊论文

  • 发表日期：

    2026年7月23日

查看全文
  • Mohammad Ali Karbasforoushha,
  • Thira Jearsiripongkul,
  • Suraparb Keawsawasvong,
  • Mohammad Khajehzadeh

摘要

This paper proposes a stochastic optimization method for sophisticated office building design and compares an Improved Snow Geese Algorithm (ISGA) with basic SGA, Gray Wolf Optimizer (GWO), and Manta Ray Foraging Optimization (MRFO). The methodology is validated in three scenarios: (i) CEC-2019 benchmark functions (Case I), (ii) a deterministic detailed building design (Case II), and (iii) a stochastic building design that models zone lighting energy consumption uncertainty with the 2m + 1 Point Estimate Method (PEM) (Case III). On CEC-2019, ISGA has the lowest mean on all but one function and the lowest (second-lowest) standard deviations; Friedman's test ranks ISGA best overall, consistent with boxplots (lower medians and narrower interquartile ranges) and with faster and more stable convergence. In the building design case, ISGA produced the most efficient and robust performances: In Case II, the optimum yearly energy consumption had 129.3305 kWh/m2a, the average is 133.0116 kWh/m2a, and the standard deviation was 0.0864, better than SGA, GWO, MRFO, and a previously published baseline. Adding zone lighting energy consumption uncertainty to Case III adds annual energy for all methods (ISGA + 7.68%, SGA + 8.06%, GWO + 8.17%, and MRFO + 8.53%), yet ISGA continues to exhibit the lowest energy and most stable performance. The gains stem from ISGA's chaos-augmented search—chaotic initialization, step perturbations, and adaptive tuning—sustaining diversity, decreasing premature convergence, and optimizing exploration–exploitation trade-off.