Facility agriculture, which involves agricultural production in controlled environments such as greenhouses, indoor farms, and vertical farms, aims to maximize efficiency, yield, and quality while minimizing resource consumption and environmental impact. Energy-saving technologies are essential to the green and low-carbon development of facility agriculture. Recently, phase change heat storage (PCHS) systems using phase change materials (PCMs) have gained significant attention due to their high thermal storage density and excellent thermal regulation performance. These systems are particularly promising for applications in facility agriculture and related buildings, such as solar thermal utilization, greenhouse walls, and soil insulation. However, the low thermal conductivity of PCMs presents a challenge for applications requiring rapid heat transfer. This study aims to provide a comprehensive review of the types, thermophysical properties, and various forms of PCMs, including macro-encapsulated PCMs, shape-stabilized PCMs, and phase change capsules (PCCs), as well as their preparation methods. The research methodology involves an in-depth analysis of these PCMs and their applications in active and passive PCHS systems within facility agriculture and related buildings. The major conclusion of this study highlights the critical role of PCMs in advancing energy-saving technologies in facility agriculture. By enhancing PCM performance, optimizing latent heat storage systems, and integrating intelligent environmental control, this work provides essential guidelines for designing more efficient and sustainable agricultural structures. The article will serve as the fundamental guideline to design more robust structures for facility agriculture and related buildings.

Machine learning (ML) models for alkali-activated concrete (AAC) are almost universally evaluated with random train–test splits, yet the literature-compiled datasets are strongly clustered by source study, and the reliability of such evaluations has rarely been quantified. The novelty of this study is a systematic quantification of out-of-campaign generalisation—via Leave-One-Study-Out (LOSO) cross-validation—for ML models trained on the largest curated public AAC dataset (1630 mixtures compiled from 106 published sources), together with model interpretation and an exploratory strength–carbon analysis. Four models (Linear Regression, Random Forest, Gradient Boosting, and optimised extreme gradient boosting, XGBoost) were benchmarked for predicting 28-day compressive strength (CS₂₈). XGBoost performed best under conventional random splitting, with test-set coefficient of determination R² = 0.801 and root-mean-square error (RMSE) = 7.21 MPa (5-fold cross-validation R² = 0.758 ± 0.050). Under LOSO validation across 85 study folds, however, the median R² collapsed to −0.328, with 49 of 85 folds negative: random-split metrics on literature-compiled AAC datasets are substantially inflated by within-study clustering, and study-stratified evaluation should become standard practice in this field. Within these limits, SHapley Additive exPlanations (SHAP) identified ground granulated blast-furnace slag (GGBFS) content, specimen geometry, CaO fraction, curing time, and sodium silicate (Na₂SiO₃) content as the five most influential predictors; because the oxide descriptors are derived from the declared binder proportions and the carbon-footprint values are inherited estimates from the source dataset, these attributions are associational rather than causal. No practically meaningful overall linear association was observed between estimated CO₂ footprint and CS₂₈ (Pearson r = −0.113, 95% CI [−0.175, −0.050], R² = 0.013), and a Pareto analysis identified 14 candidate low-carbon, high-strength formulations for further experimental and life-cycle assessment. The developed models are suitable for within-dataset feature attribution and exploratory screening restricted to the represented feature domain; they should not be used as external mix-design tools without validation on independent experimental campaigns.

Machine learning offers promising tools for predicting the compressive strength of alkali-activated concrete, but high accuracy on literature-compiled datasets does not necessarily translate into reliable predictions for unseen experimental studies. This study investigates the sources and practical limits of this generalisation gap using 963 mixtures with reported 28-day compressive strength from 86 studies. A fixed XGBoost model was evaluated through random, repeated grouped and leave-one-study-out cross-validation, complemented by systematic analyses of missing-value handling, model capacity, hyperparameter selection, performance metrics and predictive uncertainty. Random cross-validation yielded R² = 0.79, whereas leave-one-study-out evaluation produced a pooled R² of 0.28. Missingness indicators alone predicted strength under random validation (R² = 0.20), revealing study-specific reporting signatures, but alternative imputation strategies and feature removal did not resolve the transfer gap. Stronger regularisation and tuning under grouped validation likewise provided little improvement. Negative per-study R² values were concentrated in studies with narrow strength ranges, demonstrating the need to interpret these scores alongside absolute prediction errors. Feature-space novelty was associated with poorer transfer, while restricting predictions to the quarter of mixtures closest to the training distribution increased pooled R² to 0.66. Adaptive conformal intervals achieved 91% mean coverage, with half-widths ranging from approximately 9 MPa for familiar mixtures to 69 MPa for the most novel. These findings point to between-study heterogeneity as a persistent constraint and support using literature-trained models for screening within well-represented composition regions, accompanied by study-aware validation and uncertainty estimates.