Haiyu Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Haiyu Zhang has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Yaqi Wu
Yantai Institute, China Agricultural University, Yantai 264670, China
4 shared publications - Boyu Sun
College of Engineering, China Agricultural University, Beijing 100083, China
3 shared publications - Xinqing Xiao
College of Engineering, China Agricultural University, Beijing 100083, China
3 shared publications - Zetian Fu
College of Engineering, China Agricultural University, Beijing 100083, China
3 shared publications - Delin Zhang
College of Water Resources and Civil Engineering, China Agricultural University, Beijing 100083, China
2 shared publications - Du Chen
Yantai Institute, China Agricultural University, Yantai 264670, China
2 shared publications
Latest Publications
- Open Access
- 24 April 2026
- Published by MDPI AG in Informatics
- Vol. 13 (5) , 66
Intelligent Question-Answering System for New Energy Vehicles Integrating Deep Semantic Parsing and Knowledge Graphs
The new energy vehicle (NEV) industry generates massive multi-source heterogeneous data. To overcome traditional database limitations in terminology disambiguation and multi-hop reasoning, this paper proposes a knowledge graph (KG)-based question-answering (QA) architecture. Three primary domain challenges are addressed: First, to tackle the poor semantic extraction of informal diagnostic texts, a deep semantic parsing network (BERT-BiLSTM-CRF) is integrated to extract high-precision knowledge from 150,000 real-world maintenance records. Second, to solve topological redundancy, the Labeled Property Graph (LPG) specification is employed to encapsulate parameters of 2157 vehicle models as internal attributes, significantly streamlining complex multi-hop reasoning. Finally, to enhance limited reasoning capabilities, an intent classification module (TextCNN) automatically translates natural language into graph queries, enabling deep fault tracing across up to five semantic levels. Experimental results demonstrate 98% and 93% accuracy in entity-relation recognition and intent classification, respectively. The resulting KG (8274 nodes, 14,488 edges) establishes a scalable paradigm for intelligent diagnostic reasoning in complex vertical domains.
- 24 April 2026
- Published by MDPI AG in Informatics
- Vol. 13 (5) , 66
Open AccessIntelligent Question-Answering System for New Energy Vehicles Integrating Deep Semantic Parsing and Knowledge Graphs
The new energy vehicle (NEV) industry generates massive multi-source heterogeneous data. To overcome traditional database limitations in terminology disambiguation and multi-hop reasoning, this paper proposes a knowledge graph (KG)-based question-answering (QA) architecture. Three primary domain challenges are addressed: First, to tackle the poor semantic extraction of informal diagnostic texts, a deep semantic parsing network (BERT-BiLSTM-CRF) is integrated to extract high-precision knowledge from 150,000 real-world maintenance records. Second, to solve topological redundancy, the Labeled Property Graph (LPG) specification is employed to encapsulate parameters of 2157 vehicle models as internal attributes, significantly streamlining complex multi-hop reasoning. Finally, to enhance limited reasoning capabilities, an intent classification module (TextCNN) automatically translates natural language into graph queries, enabling deep fault tracing across up to five semantic levels. Experimental results demonstrate 98% and 93% accuracy in entity-relation recognition and intent classification, respectively. The resulting KG (8274 nodes, 14,488 edges) establishes a scalable paradigm for intelligent diagnostic reasoning in complex vertical domains.
- Open Access
- 8 April 2026
- Published by MDPI AG in Electronics
- Vol. 15 (8) , 1547
Agent Technology for Agricultural Intelligence: Methodological Framework and Applications
Agricultural intelligent agent technology features autonomy in multimodal perception, scalability for cross-scenario collaboration and adaptability via closed-loop optimization, serving as a core technological pillar for industrial intelligent upgrading and refined production management. This paper systematically elucidates its technical essence and methodological framework, focusing on five key aspects: multimodal heterogeneous data perception and fusion, scenario-oriented knowledge modeling and dynamic memory, intelligent decision-making and planning, embodied artificial intelligence, and closed-loop feedback optimization. On this basis, the paper outlines its core agricultural applications in four domains: crop cultivation, efficient utilization of agricultural resources, intelligent upgrading of agricultural technologies and equipment, and collaborative governance of the entire agricultural industry chain. From an interdisciplinary “AI + Agriculture” perspective, the paper further analyzes its future development directions, aiming to provide insights for improving agricultural intelligent agent technologies and promoting their industrial application to accelerate agricultural intelligent transformation. This study constructs a three-dimensional integrated methodological framework encompassing technological analysis, application mapping and trend forecasting, systematically summarizes its agricultural application scenarios and technological evolution characteristics, enriches the theoretical system and methodological construction of agricultural intelligent agent research, and provides a reusable analytical paradigm for agricultural intelligent agent research and practice.
- 8 April 2026
- Published by MDPI AG in Electronics
- Vol. 15 (8) , 1547
Open AccessAgent Technology for Agricultural Intelligence: Methodological Framework and Applications
Agricultural intelligent agent technology features autonomy in multimodal perception, scalability for cross-scenario collaboration and adaptability via closed-loop optimization, serving as a core technological pillar for industrial intelligent upgrading and refined production management. This paper systematically elucidates its technical essence and methodological framework, focusing on five key aspects: multimodal heterogeneous data perception and fusion, scenario-oriented knowledge modeling and dynamic memory, intelligent decision-making and planning, embodied artificial intelligence, and closed-loop feedback optimization. On this basis, the paper outlines its core agricultural applications in four domains: crop cultivation, efficient utilization of agricultural resources, intelligent upgrading of agricultural technologies and equipment, and collaborative governance of the entire agricultural industry chain. From an interdisciplinary “AI + Agriculture” perspective, the paper further analyzes its future development directions, aiming to provide insights for improving agricultural intelligent agent technologies and promoting their industrial application to accelerate agricultural intelligent transformation. This study constructs a three-dimensional integrated methodological framework encompassing technological analysis, application mapping and trend forecasting, systematically summarizes its agricultural application scenarios and technological evolution characteristics, enriches the theoretical system and methodological construction of agricultural intelligent agent research, and provides a reusable analytical paradigm for agricultural intelligent agent research and practice.
- Open Access
- 31 December 2025
- Published by Springer Nature in Agricultural and Food Economics
- Vol. 13 (1) , 99
Spatial and temporal heterogeneity of ENSO impacts on rice production and price dynamics in japan under climate change
Amid escalating global climate change, the El Niño–Southern Oscillation (ENSO) has emerged as a critical driver of variability in agri-food systems. Using high-frequency panel data from 46 prefectures in Japan spanning 2000–2025, this study establishes a multi-scale analytical framework to investigate how ENSO influences the rice supply chain. The main findings are as follows: (1) Fixed-effect models combined with climate attribution reveal that ENSO exerts phase-specific effects on rice-growing degree days and effective precipitation. El Niño episodes lengthen the growing season in Kyushu by an average of 11.3 days ( p < 0.01), whereas La Niña conditions intensify precipitation variability in the Tohoku region by 22%. (2) A dynamic threshold panel model demonstrates a pronounced inverted-U relationship between rice yield and climatic factors, with an optimal mean temperature of 26.5 °C (95% CI 25.8–27.2). Beyond this threshold, yields decline by approximately 4.7% per 1 °C increase. (3) By employing a simultaneous-equation model integrating production shocks and policy adjustments, the analysis indicates that ENSO-driven supply responses account for 31% of rice price fluctuations. Furthermore, price elasticity during La Niña years (0.42) is significantly greater than during El Niño years (0.29). This research provides the first quantitative assessment of ENSO sensitivity across Japan’s agro-climatic zones, offering evidence-based guidance for region-specific climate adaptation strategies. The findings carry important implications for advancing UN SDG 2 (Zero Hunger) and SDG 13 (Climate Action).
- 31 December 2025
- Published by Springer Nature in Agricultural and Food Economics
- Vol. 13 (1) , 99
Open AccessSpatial and temporal heterogeneity of ENSO impacts on rice production and price dynamics in japan under climate change
Amid escalating global climate change, the El Niño–Southern Oscillation (ENSO) has emerged as a critical driver of variability in agri-food systems. Using high-frequency panel data from 46 prefectures in Japan spanning 2000–2025, this study establishes a multi-scale analytical framework to investigate how ENSO influences the rice supply chain. The main findings are as follows: (1) Fixed-effect models combined with climate attribution reveal that ENSO exerts phase-specific effects on rice-growing degree days and effective precipitation. El Niño episodes lengthen the growing season in Kyushu by an average of 11.3 days ( p < 0.01), whereas La Niña conditions intensify precipitation variability in the Tohoku region by 22%. (2) A dynamic threshold panel model demonstrates a pronounced inverted-U relationship between rice yield and climatic factors, with an optimal mean temperature of 26.5 °C (95% CI 25.8–27.2). Beyond this threshold, yields decline by approximately 4.7% per 1 °C increase. (3) By employing a simultaneous-equation model integrating production shocks and policy adjustments, the analysis indicates that ENSO-driven supply responses account for 31% of rice price fluctuations. Furthermore, price elasticity during La Niña years (0.42) is significantly greater than during El Niño years (0.29). This research provides the first quantitative assessment of ENSO sensitivity across Japan’s agro-climatic zones, offering evidence-based guidance for region-specific climate adaptation strategies. The findings carry important implications for advancing UN SDG 2 (Zero Hunger) and SDG 13 (Climate Action).
- Open Access
- 7 July 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (13) , 7611
Joint Entity–Relation Extraction for Knowledge Graph Construction in Marine Ranching Equipment
The construction of marine ranching is a crucial component of China’s Blue Granary strategy, yet the fragmented knowledge system in marine ranching equipment impedes intelligent management and operational efficiency. This study proposes the first knowledge graph (KG) framework tailored for marine ranching equipment, integrating hybrid ontology design, joint entity–relation extraction, and graph-based knowledge storage: (1) The limitations in existing KG are obtained through targeted questionnaires for diverse users and employees; (2) A domain ontology was constructed through a combination of the top-down and the bottom-up approach, defining seven key concepts and eight semantic relationships; (3) Semi-structured data from enterprises and standards, combined with unstructured data from the literature were systematically collected, cleaned via Scrapy and regular expression, and standardized into JSON format, forming a domain-specific corpus of 1456 annotated sentences; (4) A novel BERT-BiGRU-CRF model was developed, leveraging contextual embeddings from BERT, parameter-efficient sequence modeling via BiGRU (Bidirectional Gated Recurrent Unit), and label dependency optimization using CRF (Conditional Random Field). The TE + SE + Ri + BMESO tagging strategy was introduced to address multi-relation extraction challenges by linking theme entities to secondary entities; (5) The Neo4j-based KG encapsulated 2153 nodes and 3872 edges, enabling scalable visualization and dynamic updates. Experimental results demonstrated superior performance over BiLSTM-CRF and BERT-BiLSTM-CRF, achieving 86.58% precision, 77.82% recall, and 81.97% F1 score. This study not only proposes the first structured KG framework for marine ranching equipment but also offers a transferable methodology for vertical domain knowledge extraction.
- 7 July 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (13) , 7611
Open AccessJoint Entity–Relation Extraction for Knowledge Graph Construction in Marine Ranching Equipment
The construction of marine ranching is a crucial component of China’s Blue Granary strategy, yet the fragmented knowledge system in marine ranching equipment impedes intelligent management and operational efficiency. This study proposes the first knowledge graph (KG) framework tailored for marine ranching equipment, integrating hybrid ontology design, joint entity–relation extraction, and graph-based knowledge storage: (1) The limitations in existing KG are obtained through targeted questionnaires for diverse users and employees; (2) A domain ontology was constructed through a combination of the top-down and the bottom-up approach, defining seven key concepts and eight semantic relationships; (3) Semi-structured data from enterprises and standards, combined with unstructured data from the literature were systematically collected, cleaned via Scrapy and regular expression, and standardized into JSON format, forming a domain-specific corpus of 1456 annotated sentences; (4) A novel BERT-BiGRU-CRF model was developed, leveraging contextual embeddings from BERT, parameter-efficient sequence modeling via BiGRU (Bidirectional Gated Recurrent Unit), and label dependency optimization using CRF (Conditional Random Field). The TE + SE + Ri + BMESO tagging strategy was introduced to address multi-relation extraction challenges by linking theme entities to secondary entities; (5) The Neo4j-based KG encapsulated 2153 nodes and 3872 edges, enabling scalable visualization and dynamic updates. Experimental results demonstrated superior performance over BiLSTM-CRF and BERT-BiLSTM-CRF, achieving 86.58% precision, 77.82% recall, and 81.97% F1 score. This study not only proposes the first structured KG framework for marine ranching equipment but also offers a transferable methodology for vertical domain knowledge extraction.
- Open Access
- 6 April 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (7) , 4034
Large Language Model Based Intelligent Fault Information Retrieval System for New Energy Vehicles
In recent years, the rapid development of the new energy vehicle (NEV) industry has exposed significant deficiencies in intelligent fault diagnosis and information retrieval technologies, especially in intelligent fault information retrieval, which faces persistent challenges including inadequate system adaptability and reasoning bottlenecks. To address these challenges, this study proposes a Retrieval-Augmented Generation (RAG) framework that integrates large language models (LLMs) with knowledge graphs (KGs). The framework consists of three key components: fault data collection, knowledge graph construction, and fault knowledge model training. The primary research contributions are threefold: (1) A domain-optimized fine-tuning strategy for LLMs based on NEV fault characteristics, verifying the superior accuracy of the Bidirectional Encoder Representations from Transformers (BERT) model in fault classification tasks. (2) A structured knowledge graph encompassing 122 fault categories, developed through the ChatGLM3-6B model completing named entity and knowledge relation extraction to generate fault knowledge and build a paraphrased vocabulary. (3) An intelligent fault information retrieval system that significantly outperforms traditional models in NEV-specific Q&A scenarios, providing multi-level fault cause analysis and actionable solution recommendations.
- 6 April 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (7) , 4034
Open AccessLarge Language Model Based Intelligent Fault Information Retrieval System for New Energy Vehicles
In recent years, the rapid development of the new energy vehicle (NEV) industry has exposed significant deficiencies in intelligent fault diagnosis and information retrieval technologies, especially in intelligent fault information retrieval, which faces persistent challenges including inadequate system adaptability and reasoning bottlenecks. To address these challenges, this study proposes a Retrieval-Augmented Generation (RAG) framework that integrates large language models (LLMs) with knowledge graphs (KGs). The framework consists of three key components: fault data collection, knowledge graph construction, and fault knowledge model training. The primary research contributions are threefold: (1) A domain-optimized fine-tuning strategy for LLMs based on NEV fault characteristics, verifying the superior accuracy of the Bidirectional Encoder Representations from Transformers (BERT) model in fault classification tasks. (2) A structured knowledge graph encompassing 122 fault categories, developed through the ChatGLM3-6B model completing named entity and knowledge relation extraction to generate fault knowledge and build a paraphrased vocabulary. (3) An intelligent fault information retrieval system that significantly outperforms traditional models in NEV-specific Q&A scenarios, providing multi-level fault cause analysis and actionable solution recommendations.
- Open Access
- 18 March 2025
- Published by MDPI AG in Preprints.org
Large Language Model Based Intelligent Fault Information Retrieval System for New Energy Vehicles
In recent years, the rapid development of the new energy vehicle (NEV) industry has exposed significant deficiencies in intelligent fault diagnosis and information retrieval technologies, especially in intelligent fault information retrieval, which faces persistent challenges including inadequate system adaptability and reasoning bottlenecks. This study proposes a retrieval-augmented generation (RAG) framework that integrates large language models (LLMs) with KGs (KGs) through three key components: fault data collection, knowledge graph construction, and fault knowledge model training. The primary research contributions are three-fold: (1) A domain-optimized fine-tuning strategy for LLMs based on NEV fault characteristics, verifying the superior accuracy of the Bidirectional Encoder Representations from Trans-formers (BERT) model in fault classification tasks. (2) A structured knowledge graph encompassing 122 fault categories, developed through the ChatGLM3-6B model completing named entity and knowledge relation extraction to generate fault knowledge and build a paraphrase vocabulary. (3) An intelligent fault information retrieval system that significantly outperforms traditional models in NEV-specific Q&A scenarios, providing multi-level fault cause analysis and actionable solution recommendations.
- 18 March 2025
- Published by MDPI AG in Preprints.org
Open AccessLarge Language Model Based Intelligent Fault Information Retrieval System for New Energy Vehicles
In recent years, the rapid development of the new energy vehicle (NEV) industry has exposed significant deficiencies in intelligent fault diagnosis and information retrieval technologies, especially in intelligent fault information retrieval, which faces persistent challenges including inadequate system adaptability and reasoning bottlenecks. This study proposes a retrieval-augmented generation (RAG) framework that integrates large language models (LLMs) with KGs (KGs) through three key components: fault data collection, knowledge graph construction, and fault knowledge model training. The primary research contributions are three-fold: (1) A domain-optimized fine-tuning strategy for LLMs based on NEV fault characteristics, verifying the superior accuracy of the Bidirectional Encoder Representations from Trans-formers (BERT) model in fault classification tasks. (2) A structured knowledge graph encompassing 122 fault categories, developed through the ChatGLM3-6B model completing named entity and knowledge relation extraction to generate fault knowledge and build a paraphrase vocabulary. (3) An intelligent fault information retrieval system that significantly outperforms traditional models in NEV-specific Q&A scenarios, providing multi-level fault cause analysis and actionable solution recommendations.
- Open Access
- 17 March 2025
- Published by MDPI AG in Preprints.org
Joint Entity-Relation Extraction for Knowledge Graph Construction in Marine Ranching Equipment
The construction of marine ranching is a crucial component of China’s Blue Granary strategy, yet the fragmented knowledge system in marine ranching equipment impedes intelligent management and operational efficiency. This study proposes a novel knowledge graph (KG) framework tailored for marine ranching equipment, integrating hybrid ontology design, joint entity-relation extraction, and graph-based knowledge storage: (1) The limitations in existing KG are obtained through targeted questionnaires for diverse users and employees; (2)A domain ontology was constructed through a combined top-down and bottom-up approach, defining seven core concepts and eight semantic relationships; (3) Semi-structured data from enterprises and standards, combined with unstructured data from literature were systematically collected, cleaned via Scrapy and regular expression, and standardized into JSON format, forming a domain-specific corpus of 1,456 annotated sentences; (4)A novel BERT-BiGRU-CRF model was developed, leveraging contextual embeddings from BERT, parameter-efficient sequence modeling via BiGRU, and label dependency optimization using CRF. The TE+SE+Ri+BMESO tagging strategy was introduced to address multi-relation extraction challenges by linking theme entities to secondary entities; (4)The Neo4j-based knowledge graph encapsulated 2,153 nodes and 3,872 edges, enabling scalable visualization and dynamic updates. Experimental results demonstrated superior performance over BERT-BiLSTM-CRF, achieving 86.58% precision, 77.82% recall, and 81.97% F1-score(1.94% improvement, p < 0.05) . This study not only pioneers the first structured KG framework for marine ranching equipment but also offers a transferable methodology for vertical domain knowledge extraction.
- 17 March 2025
- Published by MDPI AG in Preprints.org
Open AccessJoint Entity-Relation Extraction for Knowledge Graph Construction in Marine Ranching Equipment
The construction of marine ranching is a crucial component of China’s Blue Granary strategy, yet the fragmented knowledge system in marine ranching equipment impedes intelligent management and operational efficiency. This study proposes a novel knowledge graph (KG) framework tailored for marine ranching equipment, integrating hybrid ontology design, joint entity-relation extraction, and graph-based knowledge storage: (1) The limitations in existing KG are obtained through targeted questionnaires for diverse users and employees; (2)A domain ontology was constructed through a combined top-down and bottom-up approach, defining seven core concepts and eight semantic relationships; (3) Semi-structured data from enterprises and standards, combined with unstructured data from literature were systematically collected, cleaned via Scrapy and regular expression, and standardized into JSON format, forming a domain-specific corpus of 1,456 annotated sentences; (4)A novel BERT-BiGRU-CRF model was developed, leveraging contextual embeddings from BERT, parameter-efficient sequence modeling via BiGRU, and label dependency optimization using CRF. The TE+SE+Ri+BMESO tagging strategy was introduced to address multi-relation extraction challenges by linking theme entities to secondary entities; (4)The Neo4j-based knowledge graph encapsulated 2,153 nodes and 3,872 edges, enabling scalable visualization and dynamic updates. Experimental results demonstrated superior performance over BERT-BiLSTM-CRF, achieving 86.58% precision, 77.82% recall, and 81.97% F1-score(1.94% improvement, p < 0.05) . This study not only pioneers the first structured KG framework for marine ranching equipment but also offers a transferable methodology for vertical domain knowledge extraction.
- Open Access
- 19 February 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (4) , 2674
Fuzzy-PID-Based Atmosphere Packaging Gas Distribution System for Fresh Food
The regulation process of gas distribution systems for atmosphere packaging has the characteristics of being nonlinear time varying and having hysteric delay. When the conventional proportional-integral-derivative (PID) control algorithm is applied to this kind of system, it is difficult to set the parameters as the process is time consuming and has poor reliability. For these reasons, this paper designs a gas distribution system for fresh food atmosphere packaging based on a fuzzy PID controller. The step response method is used to construct the system’s mathematical model under the given conditions and to optimize the gas distribution control flow. A simulation experimental platform to compare between the fuzzy PID controller and a conventional PID controller is designed, and the effectiveness of the fuzzy PID control strategy is verified, which proves that it can improve the performance of the monitoring system. The system can realize the remote monitoring of the gas distribution processes through the use of a mobile phone communication network. The data transmission is reliable, the operation is convenient, and, at the same time, the overall efficiency is improved. The results of the system simulation and the gas distribution for atmosphere packaging show that the fuzzy PID algorithm has a faster gas distribution speed and good environmental adaptability as the controller of the gas distribution system. The results show that the stability time of the fuzzy PID controller is about 38 s, while the stability time of the conventional PID controller is about 85 s. The concentration error of fresh gases is ±0.25% floating, the accuracy is increased by 12 times, and the gas distribution speed is increased by about 50% when the system is stable.
- 19 February 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (4) , 2674
Open AccessFuzzy-PID-Based Atmosphere Packaging Gas Distribution System for Fresh Food
The regulation process of gas distribution systems for atmosphere packaging has the characteristics of being nonlinear time varying and having hysteric delay. When the conventional proportional-integral-derivative (PID) control algorithm is applied to this kind of system, it is difficult to set the parameters as the process is time consuming and has poor reliability. For these reasons, this paper designs a gas distribution system for fresh food atmosphere packaging based on a fuzzy PID controller. The step response method is used to construct the system’s mathematical model under the given conditions and to optimize the gas distribution control flow. A simulation experimental platform to compare between the fuzzy PID controller and a conventional PID controller is designed, and the effectiveness of the fuzzy PID control strategy is verified, which proves that it can improve the performance of the monitoring system. The system can realize the remote monitoring of the gas distribution processes through the use of a mobile phone communication network. The data transmission is reliable, the operation is convenient, and, at the same time, the overall efficiency is improved. The results of the system simulation and the gas distribution for atmosphere packaging show that the fuzzy PID algorithm has a faster gas distribution speed and good environmental adaptability as the controller of the gas distribution system. The results show that the stability time of the fuzzy PID controller is about 38 s, while the stability time of the conventional PID controller is about 85 s. The concentration error of fresh gases is ±0.25% floating, the accuracy is increased by 12 times, and the gas distribution speed is increased by about 50% when the system is stable.