Surface coal mining causes extensive and persistent vegetation disturbance worldwide, posing significant challenges for environmental impact assessment and restoration management. However, limitations in long-term remote sensing detection accuracy and the reliance on single-dimensional evaluation approaches have hindered systematic impact assessment and national-scale characterization of vegetation disturbance and restoration processes in mining areas. This study proposes a spatio-temporally constrained vegetation disturbance reconstruction method (ST-VDR). By combining dual-scale constraints with land-cover classification, the method substantially improves disturbance identification reliability (overall accuracy = 96.42%, Kappa = 0.91). Using this method, the historical footprint of vegetation disturbance in surface coal mining areas across Australia from 1988 to 2024 is reconstructed. The results show that cumulative vegetation disturbance reached 2356.12 km2, of which 35.04% has been restored. The average restoration lag is 11.05 years, and 41.89% of restored areas exhibit vegetation structure upgrades characterized by transitions toward more complex vegetation forms. Based on these results, a three-dimensional evaluation framework incorporating area, type, and coordination is established, enabling the classification of restoration performance into four distinct patterns. Overall, restoration practices in Australia's mining areas appear to be transitioning from a coverage-oriented approach toward greater emphasis on restoration quality and alignment. A small number of mines have already achieved high-quality and timely vegetation structural recovery, demonstrating that precise and high-standard restoration is technically feasible but not yet widely implemented. This study provides a new remote sensing method and evaluation framework for the systematic diagnosis and differentiated regulation of mine site restoration performance, and offers empirical insights for sustainable environmental management in mining-intensive regions worldwide.

Abstract The global demand for coal is dramatically altering landscapes, yet the full impact on critical ecosystem functions like carbon sequestration and the efficacy of restoration efforts remains inadequately quantified. This is particularly critical for surface coal mining, where restoration lags can induce long-term carbon sink deficits. To overcome the limitations of current research in baseline setting and process analysis, we develop an accounting method that integrates multisource satellite data with dynamic baseline modeling to decouple mining and restoration impacts over 23 years in China’s coal mining areas. Our results reveal a net loss of 21,757.91 Gg CO2 in vegetation carbon sequestration (VCS). Restoration has compensated for 5738.48 Gg CO2, but a pronounced recovery decoupling exists: carbon function recovery (20.87%) is less than half of vegetation area recovery (46.91%). Analysis using a novel “VCS loss pentagon” model shows that the year of start restoration (YSR) exhibited superior explanatory power for VCS losses compared to restoration level (RL), emphasizing the primacy of preventive ecological management over intensive postdestruction restoration. We find that early restoration outweighs intensive but delayed efforts in mitigating carbon loss. Model projections further indicate that achieving vegetation carbon balance by 2060 is feasible through a synergistic zonal strategy: “RL ≥ 100% (humid regions) + RL ≥ 130% (other regions) + YSR ≤ 2 years”. This work provides a scalable framework for precise carbon accounting and targeted restoration, guiding sustainable mining transitions aligned with global climate goals.

This research develops an underground mining cable-driven parallel robot (UMCDPR) for satisfying the task requirements of hoisting hydraulic supports underground. However, the swivel pulley with a time-varying structure in UMCDPR increases the kinematic complexity of the end-effector and poses challenges for controller design. To address this issue, the research proposes a correction coefficient to optimize the workspace of UMCDPR. Firstly, this study introduces the basic structure and working principle of UMCDPR. Secondly, a correction coefficient is proposed to simplify the kinematic model of the pulleys. Finally, combining the correction coefficient, the workspace of UMCDPR is optimized, and it is verified that the influence of the swivel pulley on the end-effector is only 11.7

Accurate cloud detection in optical (OPT) remote sensing imagery remains a persistent challenge, particularly for semi-transparent thin clouds and high-reflectance terrestrial surfaces where spectral signatures often overlap. This article proposes SAR-OPT-cloud detection network (CDNet), a novel multimodal fusion framework that synergistically integrates optical spectral information with the cloud-penetrating structural cues from synthetic aperture radar (SAR). To effectively bridge the modality gap, we develop a multimodal feature fusion module (MFFM) that explicitly disentangles and aggregates joint representations and cross-modal differential features under a channel-spatial attention mechanism. Furthermore, we introduce Mamba-inspired visual state space (VSS) blocks to facilitate efficient multidirectional contextual modeling with linear computational complexity, significantly enhancing the delineation of irregular cloud boundaries and the detection of fragmented cloud structures. Experimental results on the newly established Nanjing University of Aeronautics and Astronautics (NUAA)-SAR-OPT-cloud detection (CD) dataset-comprising temporally paired Sentinel-1 and Sentinel-2 imagery-demonstrate that our method achieves a state-of-the-art overall accuracy (OA) of 99.13% and an F1-score of 0.9178. Qualitative and quantitative analyses confirm that the proposed framework substantially reduces omission errors in thin-cloud regions and suppresses false alarms over high-reflectance surfaces, offering a robust and efficient solution for complex remote sensing applications.

Open-pit coal mining in arid and desert regions causes extensive and persistent surface disturbance, yet long-term monitoring of disturbance and backfilling processes remains challenging. Existing time-series change detection approaches can identify spectral changes but provide limited information on mining disturbance types and their evolution pathways. To address this issue, an automated surface disturbance detection method (Auto-SD) was developed for open-pit coal mines in arid and desert environments. This method integrates disturbance-type identification and temporal information extraction using the tasseled cap brightness (TCB) component to characterize changes associated with surface material exposure and accumulation. Using Landsat imagery from 1990 to 2023, Auto-SD was applied to 89 open-pit coal mines in China’s arid and desert regions, achieving an overall classification accuracy of 0.84. The cumulative disturbed area reached 423.10 km2, while the internal dumping area reached 94.25 km2, indicating limited backfilling recovery. Disturbance intensified after 2006, whereas backfilling lagged behind, forming a trajectory of rapid expansion, delayed recovery, and gradual stabilization. Spatially, mining areas exhibited a progressive transition from external dumping to internal dumping and backfilling. Furthermore, cumulative pit area generally followed an S-shaped growth pattern with mining duration. These findings provide new insights into long-term mining landscape evolution and support ecological restoration assessment and sustainable resource management in arid mining regions.

Spatio-temporal fusion is an important technique used to address the trade-off between temporal and spatial resolution in remote sensing satellite images. However, existing deep learning-based methods often rely on a single fusion approach, which limits their ability to effectively balance the preservation of spatial details in fine images with the integration of temporal information from coarse images. In this paper, we proposed a hybrid model architecture STCAFormer which combines temporal and spatial fusion attention mechanism, leveraging their complementary strengths to maximize the utilization of temporal and spatial information from coarse and fine images. Specifically, we design a dual-input encoder structure that processes two input images simultaneously. Moreover, inspired by the unified workflow of traditional methods that implement spatio-temporal fusion in pixel space, STCAFormer introduces a fusion stage designed for spatio-temporal fusions in feature space, enabling the effective integration of spatial and temporal features from dual-input encoder. Three novel blocks are designed in this research: a swin-temporal fusion block utilizing cross-attention mechanism to capture and fuse temporal variation features, a multi-scale spatial fusion block that integrates weighted summation for temporal variation and fine spatial features at both local and global scales, and a dual-path up-sampling block designed to achieve high-fidelity image restoration. STCAFormer was compared with six different types of spatio-temporal fusion methods on remote sensing image pairs, including Landsat 8/MODIS and Sentinel-2/3 pairs, from various regions. Experimental results demonstrate that STCAFormer performs better than baseline methods when the regions and satellite sensors remain unchanged. The transfer ability of STCAFormer was also proved to be more competitive than baseline methods. The ablation studies show that the proposed blocks are effective for performance improvement of spatio-temporal fusion The proposed STCAFormer demonstrates robust generalization across Landsat 8/MODIS and Sentinel-2/3, as well as across land-cover types represented in the test datasets. The code of the proposed STCAFormer is available at: https://github.com/Neooolee/STCAFormer.

Accurate multi-decadal vegetation monitoring is essential for quantifying the cumulative ecological impacts of mining. Annual maximum NDVI is widely used as a proxy for peak vegetation vigor, but long-term records derived from Landsat observations remain affected by two major limitations: phenological mosaic seamlines caused by inconsistent satellite acquisition timing, and temporal discontinuities associated with spectral differences among sensors. Here, we present Long NDVIcb, an automated Google Earth Engine algorithm for reconstructing spatiotemporally consistent 30-m annual maximum NDVI records across China's 14 coal bases from 1987 to 2024. The algorithm integrates two key components: a quantile mapping model with piecewise alignment to reduce phenological seamlines, and a spatially stratified cross-sensor calibration framework to harmonize Landsat 5–9 observations. Validation using simulated data omission demonstrates high reconstruction fidelity across the 14 coal bases, with RMSE values ranging from 0.032 to 0.104. The calibration functions derived for individual coal bases showed good model fits, with R² values of 0.674–0.936 for TM–ETM+ and 0.789–0.929 for ETM+–OLI. Compared with the calibration functions developed at the continental scale by Roy et al., Long-NDVIcb reduces RMSE and absolute bias by 3.3%–25.4% and 3.4%–55.6%, respectively. We further confirm high consistency between Landsat 8 and 9 records across the study area, with R² ≥ 0.890, indicating that no additional calibration is required. Comparative evaluations show that Long NDVIcb produces seamless imagery and smooth, continuous temporal trajectories, effectively reducing phenological mosaic seamlines. The resulting annual maximum NDVI dataset for China's 14 coal bases is openly available through Science Data Bank, doi:10.57760/sciencedb.26116, and provides a benchmark for long-term ecological assessment, vegetation restoration monitoring, and cross-mine comparative analysis across nearly 3,000 coal mines.

Achieving sustainable management of surface coal mines requires comprehensive and comparable eco-environmental quality (EEQ) assessment methods. This is particularly important across diverse regions. However, existing remote sensing-based indices are generally developed for individual mining areas and fail to account for mining-specific disturbances like atmospheric coal dust, thus lacking regional comparability. This study proposes a comprehensive and spatiotemporally comparable Mine Eco-environmental Quality Index (MiecoI), featuring three innovations: (1) a partitioned assessment system integrating land (vegetation and soil), water, and atmospheric factors by separately processing water and atmospheric component, overcoming the limitations of land-only assessment. (2) A standardized methodology applicable to multiple scenarios that establishes data-driven thresholds from a representative dataset and applying data standardization along with feature extraction, achieving comparable EEQ across various mines. (3) An efficient cloud-deployable computational approach for automated, large-area monitoring. The results demonstrate that: (1) MiecoI exhibited high reliability through integration of 11 indicators, with high correlation coefficients in terrestrial and non-terrestrial areas. (2) Compared to the existingindices, MiecoI demonstrated superior sensitivity to environmental degradation, with scores decreasingby 14.7 % in highly disturbed zones (stope), and spatial alignment with actual boundaries confirms its accuracy. (3) The methodology effectively compares EEQ across different countries and varied climatic and topographic regions, identifying ecological vulnerability and recovery dynamics. By providing a reliable and comparable assessment, MiecoI offers clear guidance for targeted ecological restoration and supports sustainable mining advancement.

[Objective]The Shendong coal base,located within the arid and semi-arid regions in western China,is identi-fied as the region with the most pronounced contradiction between coal resource development and ecological conserva-tion in the country.Analyzing the spatiotemporal variations in the ecosystem and their influencing factors at the base scale provides a necessary foundation for green mine construction while also offering critical support for regional sus-tainable development.[Methods]In this study,the vegetation net primary productivity(NPP)was selected as an indicat-or for the ecological evaluation of mining areas.By integrating data on terrain,weather and climate,socio-economy,and human activities,this study analyzed the spatiotemporal variations and stability of vegetation NPP in the Shendong coal base using Theil-Sen median trend analysis,Mann-Kendall(MK)test,and coefficient of variation(CV).Additionally,factors influencing vegetation NPP were examined by introducing partial derivative residual analysis,an optimal para-meters-based geographic detector(OPGD),and structural equation modeling(SEM).[Results and Conclusions]From 2000 to 2020,the vegetation NPP in the Shendong coal base exhibited an upward trend with fluctuations,with mean and total interannual growth rate determined at 6.39 gC/(m2·a)and 0.15 TgC/a,respectively.In this period,the vegetation NPP of most areas in the base showed relatively low stability yet a sustained increase.The partial derivative residual analysis results indicate that climatic factors contributed more significantly to changes in vegetation NPP than human activities.Among climatic factors,precipitation exhibited a greater contribution than solar radiation,temperature,and potential evapotranspiration.In addition,dominant factor analysis revealed that the proportions of areas where NPP in-creases were driven by climate and human activities accounted for 77.12％and 22.24％,respectively.The OPEG-de-rived results show that climatic factors exerted the highest impacts on changes in vegetation NPP and that the terrain factors produced greater influence than human activity factors.Furthermore,interactions among different factors en-hanced the explanatory power of individual factors.The SEM results reveal that climatic factors produced the highest total and direct impacts on vegetation NPP,whereas human activity factors produced the lowest impacts.In contrast,the terrain factors primarily influenced vegetation NPP through indirect effects on climatic and human activity factors.Res-ults from various analytical methods consistently demonstrate that compared to the terrain and human activity factors,climatic factors,especially precipitation,acted as the dominant factors influencing changes in vegetation NPP in the Shendong coal base.The results of this study elucidate the influence mechanisms of vegetation growth in the ecosys-tems in mining areas under high-intensity mining disturbance,providing a scientific basis for formulating differential strategies for ecological restoration and conducting green mine construction in arid and semi-arid coal mining areas.

A carbon emission accounting framework covering the entire coal production process is a prerequisite for identifying carbon source characteristics, optimizing low-carbon pathways, and reducing energy consumption intensity, and it also forms the basis for the low-carbon transition of coal enterprises. Taking a typical open-pit coal mine in northern China as a case study, the production process is divided into eight major stages — such as ore/coal mining, hauling, and dumping — according to the characteristics of open-pit mining, and direct and indirect emission sources at each stage are systematically identified. Based on a life-cycle perspective, a dynamic monthly carbon emission accounting model is developed using the emission factor method, enabling monthly carbon emission accounting across the entire production chain. The results indicate that: Carbon emissions from mining area production activities exhibit a “fugitive-dominated and process-concentrated” pattern. Diesel-driven outsourced stripping, hauling and dumping constitute high-emission stages, whereas electricity-driven loading and coal washing demonstrate clear carbon emission advantages. Optimization of the energy structure leads to a reduction in carbon emission intensity, with decrease in diesel consumption contributing more than electricity substitution. Monthly carbon emissions are positively correlated with raw coal output (r=0.99, P≤0.001) and increase noticeably with capacity expansion. Meanwhile, carbon emission intensity continues to decline, with its fluctuation range narrowing to 48.95−54.98 kgCO2e/t raw coal, indicating that capacity expansion and energy efficiency improvement can be promoted synergistically. Emission reduction potential is mainly concentrated on improving the operational efficiency of outsourced activities, electrifying mobile equipment in high diesel-consumption stages such as hauling, dumping, enhancing ecological restoration quality, and forward-looking deployment of CO2 capture, utilization, and storage (CCUS) technologies. The carbon emission accounting framework developed herein not only provides a quantitative approach for refined carbon footprint management in open-pit coal mines, but also offers scientific support for formulating targeted emission reduction policies in similar mines, contributing to the coal industry’s achievement of “dual carbon” goals and holding significant practical value.

Accurate information on rice planting areas and spatial distribution is critical for agricultural management in China; however, mapping efforts in regions like Sichuan Province are severely constrained by persistent cloud cover and fragmented terrain. Existing phenology-based methods and coarse-resolution products often fail to provide precise paddy localizations or usable training labels in such complex environments. To address these limitations, this study proposes a robust framework integrating optical and Synthetic Aperture Radar (SAR) imagery. The methodology employs a phenology-driven strategy for rapid candidate area annotation, coupled with an asymmetric feature extraction mechanism that incorporates a Multi-Scale Semantic Enhancement Module (MSEM) and a Category-Balanced Feature Fusion Module (CBFM) to facilitate adaptive and effective cross-modal fusion. Applying this framework to the Tianfu New Area (2019-2023) yielded 10-meter resolution rice distribution maps with a rice Intersection over Union (IoU) of 83.31% and a statistical correlation (R-2) of 0.972. These results demonstrate the framework's capacity for selective multi-source fusion and cost-effective sampling, facilitating precise rice mapping in challenging agricultural landscapes.

The unavoidable presence of clouds and their shadows in optical satellite imagery hinders the true spectral response of the Earth's underlying surface. Accurate cloud and cloud shadow detection is therefore a crucial pre processing step for optical satellite images and any downstream analysis. Various methods have been developed to address this critical task and can be broadly categorized into physical rule-based methods and learning based methods. In recent years, machine learning based methods, particularly deep learning frameworks, have proven to outperform physical rule-based models. However, these approaches are mostly fully supervised and require a large amount of pixel-level annotations whose acquisition is costly and time consuming. In this work, we pro pose to address cloud and cloud shadow detection in optical satellite images using self-supervised representation learning, a machine learning paradigm that focuses on extracting relevant representations from unlabeled data, which can then be used as an effective starting point to fine-tune models with few labeled data in a supervised fashion. These approaches have been shown to perform competitively with fully supervised methods without the requirement of large annotation datasets. Specifically, we assessed two self-supervised representation learning methods that use different philosophies about self-supervision: Momentum Contrast (MoCo), based on contrastive learning and DeepCluster, based on clustering. Using two publicly available Sentinel-2 cloud datasets, namely WHUS2-CD+ and CloudSEN12, we show that MoCo and DeepCluster, trained with only 25 % of the annotated data, can perform better than physical rule-based methods such as FMask and Sen2Cor, weakly supervised meth ods and even several fully supervised methods. These results highlight the strong applicability of self-supervised representation learning methods to the task of cloud and cloud shadow detection with self-supervised pretraining leading to fine-tuned models that outperform industry standards and achieve near state-of-the-art performance with a fraction of the data.

Satellite-observed vegetation greening is regarded as direct evidence of reclamation in open-pit mines. However, improvements in vegetation greenness do not necessarily indicate ecosystem resilience recovery. Identifying hidden areas of fragile resilience beneath apparent greening remains a critical challenge. In mining areas, extracting vegetation disturbance signals required for resilience assessment is difficult, because varying reclamation cycles and high spatial heterogeneity often lead to spectral mixing in existing decomposition methods. Therefore, we proposed VegDecouple, a decoupling method that adaptively separates background vegetation dynamics from short-term disturbances in mining areas. Using Landsat NDVI data (1990-2025) from a typical open-pit coal mine on the Loess Plateau, we characterized the heterogeneous trajectories of the variance-based resilience proxy beneath greening, identified three reclamation patterns, and quantified the effects of key drivers. The results showed that (1) VegDecouple effectively separated disturbance and steady-state components from non-stationary vegetation signals in mining areas. Compared with existing methods, it delivered the best spectral decoupling performance, achieving the highest energy capture and the lowest leakage. (2) Vegetation greenness and the resilience proxy exhibited pronounced temporal asynchrony and spatial mismatch. Three reclamation modes were identified within greening areas, revealing distinct inferred ecological recovery pathways. (3) Background vegetation state, reclamation age, and mean growing-season temperature showed the largest cumulative SHAP contributions, exceeding 75 % across the three reclamation modes. Moreover, SHAP dependence plots revealed nonlinear associations between key drivers and the resilience proxy. Beyond the case study, VegDecouple performed robustly across mining areas under contrasting climatic conditions. These findings provide a transferable framework for identifying potential degradation risks and supporting differentiated management in reclaimed mining landscapes.

Open-pit coal mines are widely distributed in China, and the ecological monitoring of which is susceptible to geographical heterogeneity in hydrothermal conditions, vegetation types, and soil characteristics. To achieve accurate cross-regional monitoring, it is essential to evaluate the applicability of ecological quality assessment indices. Based on Google Earth Engine and uniform Landsat 8 OLI data, we used the standardized remote sensing ecological index (RSEIs), land surface ecological status composition index (LSESCI), new remote sensing-based ecological index (RSEInew), and surface coal mine ecological index (SurMEI) to evaluate ecological quality of open-pit coal mines in four typical climatic-geomorphologic zones (arid Gobi, semi-arid grassland, semi-arid plateau, and karst plateau). Through comparative analysis of the evaluation results of four indices, combined with correlation analysis, spatial distribution identification, land cover response assessment, and validation with synchronous field measurement data from typical mining areas, we analyzed the adaptability and limitations of each index in open-pit coal mines across different climatic-geomorphologic zones. The results showed that the ecological quality assessment based on SurMEI exhibited the best performance (mean correlation coefficient of 0.810) in the four typical climatic-geomorphologic zones, demonstrating good adaptability in cross-regional evaluations. The performance disparities among the other three indices arose from mismatches between their index structures and regionally dominant ecological processes. RSEIs overestimated disturbances in humid areas due to excessive sensitivity to the normalized difference bare soil index. LSESCI exhibited misclassification in complex terrains, owing to instability of the brightness component derived from the tasseled cap transformation. RSEInew failed in arid zones because of the weak discriminative power of its added PM2.5 indicator. Furthermore, spatiotemporal analysis based on SurMEI revealed significant differences in ecological restoration potential across different climatic-geomorphologic zones, with the highest potential in the karst plateau, intermediate in the semi-arid region (including grassland and plateau), and the lowest in the arid Gobi, indicating that ecological restoration should follow the principle of "zonal management". This study would provide a theoretical basis for accurate monitoring of ecological quality and model optimization for open-pit coal mines across different climatic-geomorphologic zones.

Large-scale vegetation loss induced by surface coal mining constitutes a critical driver of regional ecological degradation. However, the applicability of existing change detection methodologies based on remote sensing within complex mining areas under diverse climatic conditions remains systematically unverified. To address this gap and reveal nationwide disturbance patterns, this study systematically evaluates the performance of two algorithms—Continuous Change Detection and Classification (CCDC) and Landsat-based Detection of Trends in Disturbance and Recovery (LandTrendr)—in identifying vegetation loss across three major climatic zones of China (the humid, semi-humid, and semi-arid zones). Based on the optimal algorithm, the vegetation loss year and loss magnitude across all of China’s surface coal mining areas from 1990 to 2020 were accurately identified, enabling the reconstruction of the comprehensive, nationwide spatio-temporal pattern of mining-induced vegetation loss over the past 30 years. The results show that: (1) CCDC demonstrated superior stability and significantly higher accuracy (OA = 0.82) than LandTrendr (OA = 0.31) in identifying loss years across all zones. (2) The cumulative vegetation loss area reached 1429.68 km2, with semi-arid zones accounting for 86.76%. Temporal analysis revealed a continuous expansion of the loss area from 2003 to 2013, followed by a distinct inflection point and decline during 2014–2016 attributable to policy-driven regulations. (3) Further analysis revealed significant variations in the average magnitude of loss across different climatic zones, namely semi-arid (0.11), semi-humid (0.21), and humid (0.25). These findings underscore the imperative for region-specific restoration strategies to ensure effective conservation outcomes. This study provides a systematic quantification and analysis of long-term, nationwide evolution patterns and regional differentiation characteristics of vegetation loss induced by surface coal mining in China, offering critical support for sustainable development decision-making in balancing energy development and ecological conservation.

ABSTRACTAccurately and efficiently identifying the vegetation disturbance ranges in surface coal mines is of great significance for determining the scope of land degradation and mitigating land degradation. The objective of this article is to propose an automated method for identifying disturbance ranges of surface coal mines on vegetation based on the fitting of NDVI spatial trajectory (called Disran_SpaTFit). The process of the proposed method includes preparing the NDVI spatial trajectory dataset, designing the curve conceptual function model, fitting the spatial trajectory, and selecting the optimal model to identify disturbance ranges. With the Shendong coal base in China as the study area, the mining disturbance ranges of 106 surface coal mines were automatically identified. The results show that: (1) The accuracy of the automated identification of mining disturbance distances was 91.1%, with a mean absolute error of 109 m. (2) Disran_SpaTFit is widely applicable to various heterogeneous coal mines. 96.62% of the NDVI spatial trajectories (1229 out of 1272 in total) were confirmed to match one of the four curve models designed in Disran_SpaTFit. (3) The ranges of mining disturbance in the 106 surface mines exhibit significant spatial heterogeneity across different directions and extend a certain distance away from the open‐cut area. (4) Disran_SpaTFit is able to accurately identify the ranges of mining disturbances for different years, covering the changes before and during mining activities. The results in this article demonstrate that the proposed Disran_SpaTFit provides an effective tool for identifying disturbance ranges of various surface coal mines, which is of importance for ecological assessment and restoration management in mining areas.

Multi-energy load forecasting is a crucial technology for energy management in low-carbon integrated energy system (IES). However, the complexity and volatility of load demand present significant challenges for accurate multi-energy load forecasting. To address this issue, a novel short-term multi-energy load forecasting method based on variational mode decomposition (VMD), marine predators algorithm (MPA), and optimal Bagging ensemble learning is proposed. A VMD-MPA-based modal decomposition model is constructed for load decomposition. An optimal Bagging ensemble load forecasting model is developed based on historical forecast error. The method is tested on IES dataset of Tempe campus of Arizona State University. The results indicate that the proposed method outperforms other compared methods in prediction accuracy.