Jiahan Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jiahan Chen has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Chengqi Yin
School of Automation, Wuxi University, Wuxi 214105, China
2 shared publications - Haojin Sun
School of Automation, Wuxi University, Wuxi 214105, China
2 shared publications - Mingxuan Mao
School of Automation, Wuxi University, Wuxi 214105, China
2 shared publications Yuanzhi Zhong
Fujian Huawang Information Technology Co., Ltd., Fujian, People’s Republic of China
2 shared publications- Yuhao Tang
School of Automation, Wuxi University, Wuxi 214105, China
2 shared publications Zhenfei Tang
Fujian Climate Center, Fujian, People’s Republic of China
2 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 14626-14635
HRU: Harmonizing Reading and Understanding for Parameter-Efficient Scene-Text Aware Cross-Modal Retrieval
Scene-text aware cross-modal retrieval requires models to jointly capture visual text content and object semantics. However, existing methods often exhibit an inherent imbalance: they either over-rely on fine-grained textual cues and become sensitive to noise, or overlook critical textual information, resulting in suboptimal alignment. To address this issue, we propose HRU, a parameter-efficient framework that Harmonizes Reading and Understanding via a two-stage progressive learning strategy and orthogonality constraints. In the first stage, visual prompts disentangle object-level and text-level representations, enabling adaptive scene text encoding while preserving visual semantics. In the second stage, a fusion mechanism progressively aligns and refines these features, achieving coherent heterogeneous visual information integration. With a parameter-efficient design, HRU can be seamlessly adapted to multimodal backbones of varying scales with minimal fine-tuning overhead. Extensive experiments demonstrate that HRU significantly improves alignment and retrieval robustness, achieving RSUM scores of 474.5, 380.3, and 546.3 on CTC-1k, CTC-5k, and TextCaps, respectively, with fewer than 30K additional trainable parameters. These results establish HRU as an effective and unified solution for text-rich visual understanding and retrieval.
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 14626-14635
Open AccessHRU: Harmonizing Reading and Understanding for Parameter-Efficient Scene-Text Aware Cross-Modal Retrieval
Scene-text aware cross-modal retrieval requires models to jointly capture visual text content and object semantics. However, existing methods often exhibit an inherent imbalance: they either over-rely on fine-grained textual cues and become sensitive to noise, or overlook critical textual information, resulting in suboptimal alignment. To address this issue, we propose HRU, a parameter-efficient framework that Harmonizes Reading and Understanding via a two-stage progressive learning strategy and orthogonality constraints. In the first stage, visual prompts disentangle object-level and text-level representations, enabling adaptive scene text encoding while preserving visual semantics. In the second stage, a fusion mechanism progressively aligns and refines these features, achieving coherent heterogeneous visual information integration. With a parameter-efficient design, HRU can be seamlessly adapted to multimodal backbones of varying scales with minimal fine-tuning overhead. Extensive experiments demonstrate that HRU significantly improves alignment and retrieval robustness, achieving RSUM scores of 474.5, 380.3, and 546.3 on CTC-1k, CTC-5k, and TextCaps, respectively, with fewer than 30K additional trainable parameters. These results establish HRU as an effective and unified solution for text-rich visual understanding and retrieval.
- Open Access
- 4 March 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (2)
UPD-Diff: A unified precipitation downscaling method based on multi-stream elucidating diffusion model
Precipitation downscaling plays a pivotal role in improving the resolution of coarse-grained weather and climate datasets, offering significant benefits for the issuance of localized precipitation warnings, which are indispensable for agricultural operations and economic planning in vulnerable regions. However, current downscaling approaches face several critical challenges: many deep learning methods, often adapted from computer vision, struggle to accurately model the long-tailed distribution of precipitation, leading to physically inconsistent outputs; reliance on regional a priori information severely limits model transferability, while inadequate merging of multivariate data yields ambiguous results; and prohibitive training costs and computational loads hinder real-time, large-scale deployment.To address these limitations, we propose the UPD-Diff, a unified multi-stream Elucidating Diffusion Model (EDM) framework designed for robust and efficient precipitation downscaling. Firstly, to promote physical consistency by meteorological variable conditioning and to more accurately model the long-tailed distribution of precipitation, the core EDM architecture of UPD-Diff employs refined noise scheduling and learns precipitation bias based on various physical auxiliary variables, thereby enhancing its alignment with meteorological dynamics. Secondly, to enhance model transferability and facilitate the effective fusion of multi-modal data inputs, each stream integrates a newly designed Mixed-Attention UNet (MA-UNet). This MA-UNet synergistically blends channel attention and local-importance attention, effectively capturing shared atmospheric features while preserving fine-grained local details across different geographical areas. Finally, to address the challenges associated with data and computational requirements, our method leverages its inherent architectural efficiency and bias-learning strategy to achieve superior SSIM and Corr scores compared to baseline models, even when trained on a relatively small dataset, thereby substantially reducing computational overhead. Trained in this globally sampled dataset, our method outperforms the leading SOTA baselines, showcasing competitive results with a PSNR of 29.02, a SSIM of 0.7942, a CSI of 0.867 (0.1 mm threshold) and improved event skill at higher thresholds(50/100 mm). Notably, our model supports both training and inference on a single GPU like RTX 4090, with local-region inference at about 5.1 seconds per 224×224 tile (100 steps) on the same GPU.
- 4 March 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (2)
Open AccessUPD-Diff: A unified precipitation downscaling method based on multi-stream elucidating diffusion model
Precipitation downscaling plays a pivotal role in improving the resolution of coarse-grained weather and climate datasets, offering significant benefits for the issuance of localized precipitation warnings, which are indispensable for agricultural operations and economic planning in vulnerable regions. However, current downscaling approaches face several critical challenges: many deep learning methods, often adapted from computer vision, struggle to accurately model the long-tailed distribution of precipitation, leading to physically inconsistent outputs; reliance on regional a priori information severely limits model transferability, while inadequate merging of multivariate data yields ambiguous results; and prohibitive training costs and computational loads hinder real-time, large-scale deployment.To address these limitations, we propose the UPD-Diff, a unified multi-stream Elucidating Diffusion Model (EDM) framework designed for robust and efficient precipitation downscaling. Firstly, to promote physical consistency by meteorological variable conditioning and to more accurately model the long-tailed distribution of precipitation, the core EDM architecture of UPD-Diff employs refined noise scheduling and learns precipitation bias based on various physical auxiliary variables, thereby enhancing its alignment with meteorological dynamics. Secondly, to enhance model transferability and facilitate the effective fusion of multi-modal data inputs, each stream integrates a newly designed Mixed-Attention UNet (MA-UNet). This MA-UNet synergistically blends channel attention and local-importance attention, effectively capturing shared atmospheric features while preserving fine-grained local details across different geographical areas. Finally, to address the challenges associated with data and computational requirements, our method leverages its inherent architectural efficiency and bias-learning strategy to achieve superior SSIM and Corr scores compared to baseline models, even when trained on a relatively small dataset, thereby substantially reducing computational overhead. Trained in this globally sampled dataset, our method outperforms the leading SOTA baselines, showcasing competitive results with a PSNR of 29.02, a SSIM of 0.7942, a CSI of 0.867 (0.1 mm threshold) and improved event skill at higher thresholds(50/100 mm). Notably, our model supports both training and inference on a single GPU like RTX 4090, with local-region inference at about 5.1 seconds per 224×224 tile (100 steps) on the same GPU.
- Open Access
Monitoring and Suppression Strategies of the Interharmonic in the Grid-Connected Photovoltaic System: A Review
With the continuous advancement of the green and low-carbon transformation of energy structures in countries around the world, renewable energy sources such as solar energy are receiving increasingly widespread attention. Driven by this trend, photovoltaic (PV) power generation, with its unique advantages, has continuously increased the scale of PV grid connection. However, the inherent characteristics of solar energy and the control features of grid-connected PV systems result in the presence of a large amount of interharmonic components in their output current, seriously endangering the safety and stable operation of the power system. Since interharmonics are difficult to obtain in an intuitive way, this paper first introduces the monitoring methods of interharmonics. Based on accurately monitoring the interharmonics, further discussions were carried out around the interharmonic suppression strategies. Through the analysis of the advantages and disadvantages of various interharmonic monitoring methods and suppression strategies, guidance can be provided for the selection of interharmonic monitoring methods and suppression strategies in different scenarios.
Open AccessMonitoring and Suppression Strategies of the Interharmonic in the Grid-Connected Photovoltaic System: A Review
With the continuous advancement of the green and low-carbon transformation of energy structures in countries around the world, renewable energy sources such as solar energy are receiving increasingly widespread attention. Driven by this trend, photovoltaic (PV) power generation, with its unique advantages, has continuously increased the scale of PV grid connection. However, the inherent characteristics of solar energy and the control features of grid-connected PV systems result in the presence of a large amount of interharmonic components in their output current, seriously endangering the safety and stable operation of the power system. Since interharmonics are difficult to obtain in an intuitive way, this paper first introduces the monitoring methods of interharmonics. Based on accurately monitoring the interharmonics, further discussions were carried out around the interharmonic suppression strategies. Through the analysis of the advantages and disadvantages of various interharmonic monitoring methods and suppression strategies, guidance can be provided for the selection of interharmonic monitoring methods and suppression strategies in different scenarios.
- Open Access
An Overview of Solar Photovoltaic Power Smoothing Control Strategies Based on Energy Storage Technology
Countries around the world are actively promoting the low-carbon transformation of the energy system, and renewable energy represented by solar photovoltaic (PV) power generation will occupy a greater proportion of the power system. The power of PV power generation is characterized by randomness and volatility, so an energy storage system (ESS) is needed for smooth control of fluctuating power to improve the quality of electric energy and the stability of the system. First of all, through the comparative analysis of various energy storage technologies, this paper finds that the battery-supercapacitor hybrid energy storage system (HESS) has both steady-state and dynamic response capabilities. Secondly, the power smoothing control strategy comprises centralized control strategies and distributed control strategies, corresponding control algorithms based on filter and optimization, and droop control strategy, respectively. This paper introduces them in turn and analyzes their advantages and disadvantages. Finally, according to the characteristics of the two control strategies, the analysis of the applicable scenarios is given, and it can guide future applications.
Open AccessAn Overview of Solar Photovoltaic Power Smoothing Control Strategies Based on Energy Storage Technology
Countries around the world are actively promoting the low-carbon transformation of the energy system, and renewable energy represented by solar photovoltaic (PV) power generation will occupy a greater proportion of the power system. The power of PV power generation is characterized by randomness and volatility, so an energy storage system (ESS) is needed for smooth control of fluctuating power to improve the quality of electric energy and the stability of the system. First of all, through the comparative analysis of various energy storage technologies, this paper finds that the battery-supercapacitor hybrid energy storage system (HESS) has both steady-state and dynamic response capabilities. Secondly, the power smoothing control strategy comprises centralized control strategies and distributed control strategies, corresponding control algorithms based on filter and optimization, and droop control strategy, respectively. This paper introduces them in turn and analyzes their advantages and disadvantages. Finally, according to the characteristics of the two control strategies, the analysis of the applicable scenarios is given, and it can guide future applications.
- 17 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
A Review of Super Resolution Technology for Extended-Range Precipitation Forecasting
In the field of meteorology, precipitation holds significant importance for human activities and environment. Extended-range precipitation forecasting, covering a timeframe of weeks to months, is crucial across various sectors, including water resource management, agriculture, and disaster prevention. However, the conventional meteorological forecasting models often fail to satisfy the precise demand for accurate and detailed precipitation predictions due to the inherent limitations in resolution and data collection frequency associated with these methods. Recently, super resolution technology has emerged as a promising solution with the potential to transform the resolution of meteorological data. By leveraging advanced algorithms and computational prowess, this technology can generate high-resolution images from low-resolution inputs. In this comprehensive review, we meticulously dissect the state-of-the-art super resolution techniques utilized in extended-range precipitation forecasting tasks. By classifying and analyzing the literature based on various methodologies, including convolutional neural networks (CNNs), generative adversarial networks (GANs), and diffusion models, we uncover their performance characteristics, strengths, and limitations. Our in-depth exploration reveals that while these methods have shown promise, they also face significant challenges. For example, CNNs struggle with capturing ultrafine details, GANs exhibit training instability, and diffusion models require further optimizations in terms of their computational efficiency. Future research efforts should focus on enhancing model architectures, seamlessly incorporating physical constraints to ensure the physical consistency of the generated data, and optimizing the current data utilization strategies. Furthermore, it is crucial to develop comprehensive evaluation metrics, conduct extensive long-term and regional validation studies, and establish effective uncertainty quantification methods. Additionally, the seamless integration of super resolution forecasts into practical applications such as decision support systems for flood prevention and water resource management, as well as their applications in climate change studies, will be pivotal for unlocking the full potential of this technology and addressing the longstanding challenges faces in precipitation forecasting scenarios.
- 17 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
A Review of Super Resolution Technology for Extended-Range Precipitation Forecasting
In the field of meteorology, precipitation holds significant importance for human activities and environment. Extended-range precipitation forecasting, covering a timeframe of weeks to months, is crucial across various sectors, including water resource management, agriculture, and disaster prevention. However, the conventional meteorological forecasting models often fail to satisfy the precise demand for accurate and detailed precipitation predictions due to the inherent limitations in resolution and data collection frequency associated with these methods. Recently, super resolution technology has emerged as a promising solution with the potential to transform the resolution of meteorological data. By leveraging advanced algorithms and computational prowess, this technology can generate high-resolution images from low-resolution inputs. In this comprehensive review, we meticulously dissect the state-of-the-art super resolution techniques utilized in extended-range precipitation forecasting tasks. By classifying and analyzing the literature based on various methodologies, including convolutional neural networks (CNNs), generative adversarial networks (GANs), and diffusion models, we uncover their performance characteristics, strengths, and limitations. Our in-depth exploration reveals that while these methods have shown promise, they also face significant challenges. For example, CNNs struggle with capturing ultrafine details, GANs exhibit training instability, and diffusion models require further optimizations in terms of their computational efficiency. Future research efforts should focus on enhancing model architectures, seamlessly incorporating physical constraints to ensure the physical consistency of the generated data, and optimizing the current data utilization strategies. Furthermore, it is crucial to develop comprehensive evaluation metrics, conduct extensive long-term and regional validation studies, and establish effective uncertainty quantification methods. Additionally, the seamless integration of super resolution forecasts into practical applications such as decision support systems for flood prevention and water resource management, as well as their applications in climate change studies, will be pivotal for unlocking the full potential of this technology and addressing the longstanding challenges faces in precipitation forecasting scenarios.