Scholarlink

<p>This paper proposes RAPTOR (Risk-Aware Probabilistic Temperature-Driven Framework) for mid-term probabilistic peak load forecasting under temperature uncertainty. The framework models temperature dynamics using deterministic seasonal–trend components and stochastic AR(1)-based residuals, enabling realistic representation of short-term persistence and clustered high-load behavior. Temperature sensitivity is estimated from recent data, and probabilistic peak load trajectories are generated via Monte Carlo simulation. Seasonal maximum load risk is quantified through parametric extreme-value modeling with a focus on reliability-relevant quantiles. A case study using Korean power system data shows that RAPTOR accurately captures both central tendency and upper-tail behavior of peak load distributions. In particular, the Gumbel distribution provides a more conservative and empirically consistent representation of upper-tail risk than the Normal distribution at the 95% reliability level. The proposed framework delivers quantile-based outputs directly applicable to reliability-oriented operational planning, offering a transparent and practical tool for mid-term power system planning under increasing climate-driven uncertainty.</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

Risk-Aware Mid-Term Peak Load Forecasting Using Probabilistic Temperature Modeling and Extreme-Value Analysis

  • 期刊论文

  • 发表日期：

    2026年3月12日

查看全文
  • Young-Min Wi

摘要

This paper proposes RAPTOR (Risk-Aware Probabilistic Temperature-Driven Framework) for mid-term probabilistic peak load forecasting under temperature uncertainty. The framework models temperature dynamics using deterministic seasonal–trend components and stochastic AR(1)-based residuals, enabling realistic representation of short-term persistence and clustered high-load behavior. Temperature sensitivity is estimated from recent data, and probabilistic peak load trajectories are generated via Monte Carlo simulation. Seasonal maximum load risk is quantified through parametric extreme-value modeling with a focus on reliability-relevant quantiles. A case study using Korean power system data shows that RAPTOR accurately captures both central tendency and upper-tail behavior of peak load distributions. In particular, the Gumbel distribution provides a more conservative and empirically consistent representation of upper-tail risk than the Normal distribution at the 95% reliability level. The proposed framework delivers quantile-based outputs directly applicable to reliability-oriented operational planning, offering a transparent and practical tool for mid-term power system planning under increasing climate-driven uncertainty.