Zujing Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Zujing Zhang has a total of 46 co-authors. Here are the top co-authors by number of shared publications:
Jiri Zhou
14 shared publicationsHongwei Wu
13 shared publicationsRuiyong Mao
13 shared publicationsXing Liang
12 shared publications- Ruiyong Mao
Guizhou Univ, Coll Civil Engn, Guizhou Prov Key Lab Rock & Soil Mech & Engn Safet, Guiyang 550025, Peoples R China
9 shared publications - Hongwei Wu
School of Physics, Engineering and Computer Science, University of Hertfordshire, Hatfield AL10 9AB, UK
6 shared publications
Latest Publications
- Open Access
- 9 July 2026
- Published by Elsevier in SSRN Electronic Journal
Investigation on thermal performance of a combined compressed air and phase change temperature control system in underground refuge chamber
The development of underground spaces is recognized as an effective solution to the resource shortages and traffic issues resulting from rapid urbanization. However, for underground spaces characterized by high population density, ensuring the safety of the indoor thermal environment has emerged as a critical factor to developing the resources of underground spaces. This paper examines the underground refuge chamber (URC) as a case study and investigates a safe and cost-effective combined temperature control scheme utilizing compressed air and phase change (CA-PC). Through experimental validation, the reliability of the numerical simulation model was confirmed, followed by an analysis of the impact of various factors—including the number of phase change plates (PCPs), per capita ventilation rate (PCVR), ventilation temperature (VT), per capita building area (PCBA), and latent heat (LH)—on the performance of temperature control via numerical simulation. The research findings indicate: (1) When employing mechanical ventilation for temperature regulation, the variation of indoor air temperature (IAT) over time follows an exponential function. (2) Only 10 PCPs with the LH of 250 kJ/kg are needed to ensure that the effective temperature control duration of the coupled temperature control scheme over 96 h, while fully utilizing the PCPs. (3) A functional relationship was established between IAT over time, relative dimensions of chamber, ventilation parameters, and the number of PCPs. These research outcomes hold significant implications for guiding the installation of combined temperature control systems utilizing CA-PC in various URCs.
- 9 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessInvestigation on thermal performance of a combined compressed air and phase change temperature control system in underground refuge chamber
The development of underground spaces is recognized as an effective solution to the resource shortages and traffic issues resulting from rapid urbanization. However, for underground spaces characterized by high population density, ensuring the safety of the indoor thermal environment has emerged as a critical factor to developing the resources of underground spaces. This paper examines the underground refuge chamber (URC) as a case study and investigates a safe and cost-effective combined temperature control scheme utilizing compressed air and phase change (CA-PC). Through experimental validation, the reliability of the numerical simulation model was confirmed, followed by an analysis of the impact of various factors—including the number of phase change plates (PCPs), per capita ventilation rate (PCVR), ventilation temperature (VT), per capita building area (PCBA), and latent heat (LH)—on the performance of temperature control via numerical simulation. The research findings indicate: (1) When employing mechanical ventilation for temperature regulation, the variation of indoor air temperature (IAT) over time follows an exponential function. (2) Only 10 PCPs with the LH of 250 kJ/kg are needed to ensure that the effective temperature control duration of the coupled temperature control scheme over 96 h, while fully utilizing the PCPs. (3) A functional relationship was established between IAT over time, relative dimensions of chamber, ventilation parameters, and the number of PCPs. These research outcomes hold significant implications for guiding the installation of combined temperature control systems utilizing CA-PC in various URCs.
- Open Access
- 7 July 2026
- Published by Elsevier in SSRN Electronic Journal
Prior-Guided Multi-Task Joint Forecasting and Storage Dispatch for Public Building Energy Systems with Photovoltaic Generation and Flexible Loads: A Case Study in Chengdu, China
With the rapid deployment of distributed photovoltaics (PV), building-side energy storage systems, and electric vehicle charging facilities, urban public buildings are gradually evolving into integrated energy nodes for coordinated source-load-storage operation. In this context, garage lighting load and charging pile load represent typical flexible building loads with different temporal patterns and behavior-driven characteristics. To address the limitations of existing studies in single-object forecasting, insufficient modeling of the coupling relationships among PV generation, flexible loads, and energy storage, and the lack of dispatch-level evaluation, this paper proposes a forecast-driven low-carbon energy management framework for public buildings and constructs a PG-WA-AE-TCN-GRU-MLKA multi-branch time-series forecasting model. The model constructs task-specific multi-scale input representations by combining wavelet decomposition (WA) and autoencoder latent features (AE), introduces a prior-guided soft gating mechanism to enhance key feature representation, and uses TCN-GRU-MLKA to extract local fluctuations, long-term dependencies, and multi-scale contextual information. Finally, the prediction results are input into a rule-based energy storage dispatch strategy, and the model performance is comprehensively evaluated from the aspects of prediction error, PV curtailment rate, and PV self-consumption rate. Experimental results show that the proposed model achieves an RMSE of 2.28 kWh, MAE of 0.95 kWh, WAPE of 30.17%, and R2 of 0.87 in joint forecasting. At the dispatch level, the PV curtailment amount and curtailment rate are 240.26 kWh and 4.93%, respectively, and the PV self-consumption rate reaches 95.07%. The results verify the effectiveness of the proposed method in improving multi-target forecasting accuracy, supporting energy storage dispatch, and promoting PV consumption.
- 7 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPrior-Guided Multi-Task Joint Forecasting and Storage Dispatch for Public Building Energy Systems with Photovoltaic Generation and Flexible Loads: A Case Study in Chengdu, China
With the rapid deployment of distributed photovoltaics (PV), building-side energy storage systems, and electric vehicle charging facilities, urban public buildings are gradually evolving into integrated energy nodes for coordinated source-load-storage operation. In this context, garage lighting load and charging pile load represent typical flexible building loads with different temporal patterns and behavior-driven characteristics. To address the limitations of existing studies in single-object forecasting, insufficient modeling of the coupling relationships among PV generation, flexible loads, and energy storage, and the lack of dispatch-level evaluation, this paper proposes a forecast-driven low-carbon energy management framework for public buildings and constructs a PG-WA-AE-TCN-GRU-MLKA multi-branch time-series forecasting model. The model constructs task-specific multi-scale input representations by combining wavelet decomposition (WA) and autoencoder latent features (AE), introduces a prior-guided soft gating mechanism to enhance key feature representation, and uses TCN-GRU-MLKA to extract local fluctuations, long-term dependencies, and multi-scale contextual information. Finally, the prediction results are input into a rule-based energy storage dispatch strategy, and the model performance is comprehensively evaluated from the aspects of prediction error, PV curtailment rate, and PV self-consumption rate. Experimental results show that the proposed model achieves an RMSE of 2.28 kWh, MAE of 0.95 kWh, WAPE of 30.17%, and R2 of 0.87 in joint forecasting. At the dispatch level, the PV curtailment amount and curtailment rate are 240.26 kWh and 4.93%, respectively, and the PV self-consumption rate reaches 95.07%. The results verify the effectiveness of the proposed method in improving multi-target forecasting accuracy, supporting energy storage dispatch, and promoting PV consumption.
- Open Access
- 6 July 2026
- Published by Elsevier in SSRN Electronic Journal
Multi-object power forecasting for grid-interactive building photovoltaic-storage-charging systems using a VMD-LSTM-CNN-KAN model: Evidence from a smart building
Accurate multi-object power forecasting is essential for the coordinated operation of grid-interactive photovoltaic-storage-charging systems in smart buildings. In building photovoltaic-storage-DC-direct flexibility (PSDC) systems, photovoltaic (PV) generation is highly intermittent and weather-dependent, while charging-pile loads are affected by stochastic user behavior and exhibit strong nonlinearity and temporal randomness. These uncertainties may weaken source-load matching and affect building’s energy management. To address this issue, this study proposes a hybrid forecasting model integrating variational mode decomposition (VMD), long short-term memory (LSTM), convolutional neural networks (CNN), and Kolmogorov–Arnold networks (KAN), abbreviated as VMD-LSTM-CNN-KAN (VLCK). The proposed model combines multi-scale decomposition, temporal dependency learning, local pattern extraction, and nonlinear feature mapping to improve forecasting accuracy. It is applied to a real building PSDC system for both source-side PV generation and load-side charging-pile demand forecasting. Results show that, for PV power forecasting, VLCK achieves an MSE of 0.0045, RMSE of 0.0671, MAE of 0.0447, and R2 of 0.9552, outperforming eight benchmark models. Under sunny, cloudy, and rainy conditions, its predicted curves show the closest agreement with actual PV output. During the transitional-season test, VLCK maintains the best comprehensive accuracy across three periods with different fluctuation characteristics. For charging-pile power forecasting, VLCK achieves an MSE of 0.0023, RMSE of 0.0480, MAE of 0.0289, and R2 of 0.9620. Ablation experiments further confirm the complementary contribution of VMD and KAN. These results indicate that VLCK can provide reliable source-load forecasting information for building-level PV-storage-charging coordination and grid-interactive energy management.
- 6 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMulti-object power forecasting for grid-interactive building photovoltaic-storage-charging systems using a VMD-LSTM-CNN-KAN model: Evidence from a smart building
Accurate multi-object power forecasting is essential for the coordinated operation of grid-interactive photovoltaic-storage-charging systems in smart buildings. In building photovoltaic-storage-DC-direct flexibility (PSDC) systems, photovoltaic (PV) generation is highly intermittent and weather-dependent, while charging-pile loads are affected by stochastic user behavior and exhibit strong nonlinearity and temporal randomness. These uncertainties may weaken source-load matching and affect building’s energy management. To address this issue, this study proposes a hybrid forecasting model integrating variational mode decomposition (VMD), long short-term memory (LSTM), convolutional neural networks (CNN), and Kolmogorov–Arnold networks (KAN), abbreviated as VMD-LSTM-CNN-KAN (VLCK). The proposed model combines multi-scale decomposition, temporal dependency learning, local pattern extraction, and nonlinear feature mapping to improve forecasting accuracy. It is applied to a real building PSDC system for both source-side PV generation and load-side charging-pile demand forecasting. Results show that, for PV power forecasting, VLCK achieves an MSE of 0.0045, RMSE of 0.0671, MAE of 0.0447, and R2 of 0.9552, outperforming eight benchmark models. Under sunny, cloudy, and rainy conditions, its predicted curves show the closest agreement with actual PV output. During the transitional-season test, VLCK maintains the best comprehensive accuracy across three periods with different fluctuation characteristics. For charging-pile power forecasting, VLCK achieves an MSE of 0.0023, RMSE of 0.0480, MAE of 0.0289, and R2 of 0.9620. Ablation experiments further confirm the complementary contribution of VMD and KAN. These results indicate that VLCK can provide reliable source-load forecasting information for building-level PV-storage-charging coordination and grid-interactive energy management.
- Open Access
- 8 April 2026
- Published by Elsevier in Case Studies in Thermal Engineering
- Vol. 81
Numerical study on unsteady heat transfer characteristics of phase change plates optimized by fin structure
- 8 April 2026
- Published by Elsevier in Case Studies in Thermal Engineering
- Vol. 81
Open AccessNumerical study on unsteady heat transfer characteristics of phase change plates optimized by fin structure
- Open Access
- 31 March 2026
- Published by Elsevier in Case Studies in Thermal Engineering
- Vol. 81
Experimental investigation on thermal and moisture performance of direct evaporative cooling coupled with phase change plates for data centers
- 31 March 2026
- Published by Elsevier in Case Studies in Thermal Engineering
- Vol. 81
Open AccessExperimental investigation on thermal and moisture performance of direct evaporative cooling coupled with phase change plates for data centers