Scholarlink

<p>This paper presents a stochastic optimization framework for planning the allocation of solar energy credits under forecast uncertainty. A Hill Climbing-based algorithm was employed, incorporating a hybrid evaluation strategy that combines the Unscented Transform (UT) with selective Monte Carlo (MC) simulation using Sobol sampling. Candidate solutions are primarily evaluated via UT, which explicitly models uncertainty in forecasts while keeping computational costs manageable. When a candidate solution outperforms the current best, MC simulation refines the evaluation and adaptively recalibrates UT parameters, further enhancing accuracy. In contrast, the deterministic baseline relies solely on fixed point estimates, disregarding input variability. Although the observed improvements peak at around 1.2%, they translate into significant amounts of energy effectively compensated rather than lost, yielding tangible financial and operational benefits. For a system generating 10 GWh annually, this improvement corresponds to approximately 120 MWh of additional compensated energy, representing tens of thousands of reais in annual financial benefit. Furthermore, the adaptive integration of Monte Carlo refinements ensures scalability by avoiding the prohibitive cost of full sampling, making the approach both computationally viable and economically justifiable for strategic planning in distributed renewable energy systems.</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

Robust allocation of solar energy credits under forecast uncertainty via adaptive unscented transform

  • 期刊论文

  • 发表日期：

    2026年5月13日

查看全文
  • Pedro H. da Silva Palhares,
  • Thyago Carvalho Marques,
  • Paulo H. Cardoso de Souza,
  • Werikcyano Lima Guimarães,
  • Gabriel D. Brito,
  • Leonardo da Cunha Brito

摘要

This paper presents a stochastic optimization framework for planning the allocation of solar energy credits under forecast uncertainty. A Hill Climbing-based algorithm was employed, incorporating a hybrid evaluation strategy that combines the Unscented Transform (UT) with selective Monte Carlo (MC) simulation using Sobol sampling. Candidate solutions are primarily evaluated via UT, which explicitly models uncertainty in forecasts while keeping computational costs manageable. When a candidate solution outperforms the current best, MC simulation refines the evaluation and adaptively recalibrates UT parameters, further enhancing accuracy. In contrast, the deterministic baseline relies solely on fixed point estimates, disregarding input variability. Although the observed improvements peak at around 1.2%, they translate into significant amounts of energy effectively compensated rather than lost, yielding tangible financial and operational benefits. For a system generating 10 GWh annually, this improvement corresponds to approximately 120 MWh of additional compensated energy, representing tens of thousands of reais in annual financial benefit. Furthermore, the adaptive integration of Monte Carlo refinements ensures scalability by avoiding the prohibitive cost of full sampling, making the approach both computationally viable and economically justifiable for strategic planning in distributed renewable energy systems.