Accurate prediction of electricity demand during hot seasons is essential for maintaining power system reliability, particularly in humid subtropical regions such as Guangdong, China, where high temperatures strongly influence consumption. However, many models rely primarily on air temperature and may not fully capture combined atmospheric effects. This study evaluates the potential of biometeorological thermal indices for improving summer electricity load forecasting. Daily maximum load and meteorological data during May–September 2019–2021 were analyzed using Back-Propagation Neural Network (BP), Random Forest (RF), and a Stacking ensemble model. Three indices—Effective Temperature (ET), Physiological Equivalent Temperature (PET), and the Universal Thermal Climate Index (UTCI)—were introduced as predictors. The ensemble model achieved the best performance, with Ensemble–UTCI yielding the highest accuracy (R² = 0.559, RMSE = 60.96 × 10⁴ kW, MAE = 45.10 × 10⁴ kW). Compared with temperature-based models, biometeorological indices consistently improved predictions, with UTCI performing best (average RMSE = 62.81 × 10⁴ kW). Bayesian analysis shows strong evidence of improvement in RF and ensemble models, but not in BP or linear models, indicating model dependence. During the July 2021 heat event, RF showed greater robustness, with PET–RF achieving the lowest error (MAPE = 3.03%). These results demonstrate the value of biometeorological indices for load forecasting in humid subtropical regions.

In the context of global warming, compound dry-hot events (CDHEs) are intensifying in Guangdong, yet CMIP6 projections remain uncertain. This study employs CMIP6 data and the Standardized Compound Event Indicator (SCEI) to quantify CDHEs severity, applying an observational constraint approach to reduce inter-model uncertainty. The results show that, after observational constraint, uncertainties decrease by about 63% and 77% in Period I and II under SSP126 and by about 57% and 59% under SSP585, greatly improving projection robustness. CDHE risk is highest in SSP585-Period II. Future dry-hot intensification in Guangdong generally increases from north to south, with western Guangdong most strongly affected. Although CDHEs weaken in other periods, western Guangdong shows persistent aggravation. Mechanism analyses indicate that SSP585-Period I is mainly linked to cold sea surface temperature (SST) anomalies in the South Atlantic and waters near Australia. After correction, dry-hot conditions show a marked weakening across Guangdong, although slight intensification persists over the Leizhou Peninsula. SSP585-Period II is primarily influenced by warm SST anomalies in the eastern Pacific and South Atlantic and cold anomalies in the North Pacific. The two SSP126 periods are associated with warm SST anomalies in the South Atlantic and waters near Australia and with cold anomalies in the South Atlantic, North Pacific, and North Atlantic, respectively. After correction, CDHEs generally weaken across Guangdong, although southern and south-central areas remain relatively severe. These findings indicate that historical key SST biases can strongly influence future CDHEs projections in Guangdong by modulating large-scale atmospheric circulation, including the Pacific-South American wave train, Indian Ocean SST anomalies, and the Western North Pacific Subtropical Anticyclone.

Global warming has intensified extreme heat events, posing growing climate risks. As a climate-sensitive region, China has experienced compound daytime and nighttime extreme heat, threatening health, ecosystems, and socioeconomic development. Accurate projections of future extreme heat in China under different emission scenarios are therefore essential for climate risk assessment, adaptation planning, and regional sustainable development. This study integrates the SimmEX extreme temperature index dataset and NEX–GDDP–CMIP6 climate model data to investigate six daytime and nighttime extreme temperature indices (TXx, TNx, TX90p, TN90p, SU, and TR). A spatiotemporal evaluation framework is developed to optimize the Multi-Model Ensemble (MME), and the Combined Nonstationary Cumulative Distribution Function matching (CNCDFm) method is applied for bias correction. The main findings are as follows: (1) CMIP6 models can generally reproduce the spatiotemporal characteristics of extreme heat over China. Intensity-based indices (TXx, TNx) showed the highest skill (spatial correlations 0.85–0.95), followed by absolute-threshold indices (SU, TR), while percentile-based indices (TX90p, TN90p) exhibited larger uncertainties. (2) MME performance did not improve monotonically with increasing ensemble size. The spatiotemporal evaluation framework identified optimal models for different indices, and the corresponding optimal MME sizes balancing accuracy and robustness were identified as 7 for TXx, 4 for TNx, 5 for TX90p, 4 for TN90p, 4 for SU, and 4 for TR. (3) CNCDFm improved MME–observation agreement, with greater improvements for TXx, TNx, SU, and TR than for the percentile-based TX90p and TN90p. Extreme heat is projected to increase across China, with stronger increases under higher-emission scenarios and distinct regional patterns among different indices. By the late 21^st century, TXx and TNx are projected to rise by 6.7 °C and 7.3 °C under SSP5-8.5, compared with 2.4 °C and 3.1 °C under SSP1-2.6. Nighttime indices exhibit stronger increases than daytime indices, highlighting greater extreme heat risks under higher emissions.