Wenjie 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
Wenjie Zhang has a total of 95 co-authors. Here are the top co-authors by number of shared publications:
- Dipti Srinivasan
National University of Singapore
16 shared publications - Oktoviano Gandhi
Solar Energy Research Institute of Singapore (SERIS), National University of Singapore (NUS), Singapore 117574, Singapore
10 shared publications - Chi Yung Chung
Hong Kong Polytechnic University
6 shared publications Dipti Srinivasan
6 shared publicationsCarlos D. Rodríguez-Gallegos
5 shared publications- Carlos David Rodriguez Gallegos
Solar Energy Research Institute of Singapore (SERIS), National University of Singapore (NUS), Singapore 117574, Singapore
5 shared publications
Latest Publications
- Open Access
Decision-Focused Wind Power Forecasting for Economic Dispatch Under Asymmetric Imbalance Costs
Wind power forecasts are commonly trained to minimize statistical errors, although the thermal schedule based on a forecast ultimately incurs asymmetric recourse costs. To connect forecast training with this dispatch consequence, a dispatch-value-oriented forecasting (DVOF) framework is developed by coupling a Mamba-style forecaster with a solver-based differentiable economic-dispatch layer. The forecast determines the first-stage thermal schedule; after wind realization, load shedding and wind curtailment settle the imbalance, and the resulting recourse-cost gradient is propagated to the forecaster by KKT-based implicit differentiation. On Global Energy Forecasting Competition 2014 (GEFCom2014) data, DVOF reduces the day-ahead realized imbalance penalty from to cost units per hour relative to an MSE Loss model with the same backbone. The resulting forecast approaches the conservative operating point implied by the asymmetric recourse model. In the simplified single-area, single-period case, this result shows that the dispatch gradient can guide the forecaster toward a cost-relevant adjustment without prescribing or searching over a quantile level.Open AccessDecision-Focused Wind Power Forecasting for Economic Dispatch Under Asymmetric Imbalance Costs
Wind power forecasts are commonly trained to minimize statistical errors, although the thermal schedule based on a forecast ultimately incurs asymmetric recourse costs. To connect forecast training with this dispatch consequence, a dispatch-value-oriented forecasting (DVOF) framework is developed by coupling a Mamba-style forecaster with a solver-based differentiable economic-dispatch layer. The forecast determines the first-stage thermal schedule; after wind realization, load shedding and wind curtailment settle the imbalance, and the resulting recourse-cost gradient is propagated to the forecaster by KKT-based implicit differentiation. On Global Energy Forecasting Competition 2014 (GEFCom2014) data, DVOF reduces the day-ahead realized imbalance penalty from to cost units per hour relative to an MSE Loss model with the same backbone. The resulting forecast approaches the conservative operating point implied by the asymmetric recourse model. In the simplified single-area, single-period case, this result shows that the dispatch gradient can guide the forecaster toward a cost-relevant adjustment without prescribing or searching over a quantile level. - 16 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 17 (1) , 746-765
A Heterogeneous Multiple-Experts Approach to Low-Frequency Nonintrusive Load Monitoring
Many nonintrusive load monitoring (NILM) approaches to decomposing household-level energy consumption into appliance-specific usage patterns using aggregated smart-meter data in conjunction with advanced deep learning techniques are able to leverage the complementary information existing between the power consumption estimation and appliance state detection tasks using multi-task frameworks. However, existing multitask frameworks suffer from three primary limitations, including a reliance on a single shared neural network that is unable to extract features for both tasks fully, particularly under long sampling intervals on the order of 15 min, the application of homogeneous neural network architectures that are unable to extract the diverse characteristics of NILM data, and the reliance on static weights to balance the competing task objectives. The present work addresses these issues by developing a multiple and heterogeneous deep learning expert-based approach to NILM conducted using low sampling frequency smart-meter data. Specifically, the proposed approach leverages a multi-gate mixture-of-experts architecture comprising five different neural network architectures, to capture both local and global time-series features effectively. Additionally, a novel multi-objective optimization strategy is designed in conjunction with the gradient descent algorithm to manage tradeoffs between the appliance state detection and power consumption estimation tasks dynamically by optimizing the weights for each task. The performance of the proposed approach is compared with that of existing state-of-the-art methods. Generalizability is validated through cross-household testing on unseen homes in the public UK-DALE and REDD datasets, while robustness to realistic data sparsity is demonstrated on a real-world dataset sampled at challenging 15-minute intervals. The results obtained for the proposed approach demonstrate significant improvements in state classification accuracy and power estimation error, particularly for appliances with complex power consumption patterns.
- 16 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 17 (1) , 746-765
A Heterogeneous Multiple-Experts Approach to Low-Frequency Nonintrusive Load Monitoring
Many nonintrusive load monitoring (NILM) approaches to decomposing household-level energy consumption into appliance-specific usage patterns using aggregated smart-meter data in conjunction with advanced deep learning techniques are able to leverage the complementary information existing between the power consumption estimation and appliance state detection tasks using multi-task frameworks. However, existing multitask frameworks suffer from three primary limitations, including a reliance on a single shared neural network that is unable to extract features for both tasks fully, particularly under long sampling intervals on the order of 15 min, the application of homogeneous neural network architectures that are unable to extract the diverse characteristics of NILM data, and the reliance on static weights to balance the competing task objectives. The present work addresses these issues by developing a multiple and heterogeneous deep learning expert-based approach to NILM conducted using low sampling frequency smart-meter data. Specifically, the proposed approach leverages a multi-gate mixture-of-experts architecture comprising five different neural network architectures, to capture both local and global time-series features effectively. Additionally, a novel multi-objective optimization strategy is designed in conjunction with the gradient descent algorithm to manage tradeoffs between the appliance state detection and power consumption estimation tasks dynamically by optimizing the weights for each task. The performance of the proposed approach is compared with that of existing state-of-the-art methods. Generalizability is validated through cross-household testing on unseen homes in the public UK-DALE and REDD datasets, while robustness to realistic data sparsity is demonstrated on a real-world dataset sampled at challenging 15-minute intervals. The results obtained for the proposed approach demonstrate significant improvements in state classification accuracy and power estimation error, particularly for appliances with complex power consumption patterns.
- 15 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 30 (2) , 594-608
Autonomous Multiobjective Optimization Using Large Language Model
Multi-objective optimization problems (MOPs) are ubiquitous in real-world applications, presenting a complex challenge of balancing multiple conflicting objectives. Traditional multi-objective evolutionary algorithms (MOEAs), though effective, often rely on domain-specific expertise for improved optimization performance, hindering adaptability to unseen MOPs. In recent years, the Large Language Models (LLMs) has revolutionized software engineering by enabling the autonomous generation and refinement of programs. Leveraging this breakthrough, we propose a new LLM-based framework that autonomously designs MOEAs for solving MOPs. The proposed framework includes a robust testing module to refine the generated MOEA through error-driven dialogue with LLMs, a dynamic selection strategy along with informative prompting-based crossover and mutation to fit textual optimization pipeline. Our approach facilitates the design of MOEA without the extensive demands for expert intervention, thereby speeding up the innovation of MOEA. Empirical studies across various MOP categories validate the robustness and superior performance of our proposed framework.
- 15 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 30 (2) , 594-608
Autonomous Multiobjective Optimization Using Large Language Model
Multi-objective optimization problems (MOPs) are ubiquitous in real-world applications, presenting a complex challenge of balancing multiple conflicting objectives. Traditional multi-objective evolutionary algorithms (MOEAs), though effective, often rely on domain-specific expertise for improved optimization performance, hindering adaptability to unseen MOPs. In recent years, the Large Language Models (LLMs) has revolutionized software engineering by enabling the autonomous generation and refinement of programs. Leveraging this breakthrough, we propose a new LLM-based framework that autonomously designs MOEAs for solving MOPs. The proposed framework includes a robust testing module to refine the generated MOEA through error-driven dialogue with LLMs, a dynamic selection strategy along with informative prompting-based crossover and mutation to fit textual optimization pipeline. Our approach facilitates the design of MOEA without the extensive demands for expert intervention, thereby speeding up the innovation of MOEA. Empirical studies across various MOP categories validate the robustness and superior performance of our proposed framework.
- 25 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (12) , 19955-19971
A Class Alignment Multisource Domain Adaptation for Partial Discharge Condition Assessment With Unknown Faults in GIS
Recently, domain adaptation has emerged as a powerful technique for on-site partial discharge (PD) condition assessment in gas-insulated switchgear (GIS). However, most existing methods face three major challenges: 1) Relying on a single source domain for model development poses difficulties in effectively utilizing source domain samples with distribution differences. 2) Limited condition assessment for unknown fault samples on-site, which faces distributional differences between multiple source domains. 3) Handling only a single task, which makes it challenging to generalize to multiple tasks simultaneously. To address these concerns, we propose a class alignment multi-source domain adaptation network (CLMSDAN) for GIS PD condition assessment with unknown faults. First, a diversity feature extractor is developed to extract diverse features while addressing the negative transfer issue caused by knowledge differences by mining both inter-domain and intra-domain features, thus enabling the transfer of rich knowledge at multiple levels. Second, a novel multi-source domain adaptation approach is employed from multiple perspectives to align distribution and distinguish between shared and unknown classes. Finally, a multi-classifier complementary strategy is introduced to recognize unknown faults, which automatically filters out source domain irrelevant class samples while distinguishing the contributions of different source domains to the target task. Experimental results show that CLMSDAN achieves 94.86% accuracy in diagnosis and 93.38% in severity assessment, outperforming baseline methods by over 10% in both tasks. This highlights its superior generalization and robustness across varying conditions and noise levels.
- 25 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (12) , 19955-19971
A Class Alignment Multisource Domain Adaptation for Partial Discharge Condition Assessment With Unknown Faults in GIS
Recently, domain adaptation has emerged as a powerful technique for on-site partial discharge (PD) condition assessment in gas-insulated switchgear (GIS). However, most existing methods face three major challenges: 1) Relying on a single source domain for model development poses difficulties in effectively utilizing source domain samples with distribution differences. 2) Limited condition assessment for unknown fault samples on-site, which faces distributional differences between multiple source domains. 3) Handling only a single task, which makes it challenging to generalize to multiple tasks simultaneously. To address these concerns, we propose a class alignment multi-source domain adaptation network (CLMSDAN) for GIS PD condition assessment with unknown faults. First, a diversity feature extractor is developed to extract diverse features while addressing the negative transfer issue caused by knowledge differences by mining both inter-domain and intra-domain features, thus enabling the transfer of rich knowledge at multiple levels. Second, a novel multi-source domain adaptation approach is employed from multiple perspectives to align distribution and distinguish between shared and unknown classes. Finally, a multi-classifier complementary strategy is introduced to recognize unknown faults, which automatically filters out source domain irrelevant class samples while distinguishing the contributions of different source domains to the target task. Experimental results show that CLMSDAN achieves 94.86% accuracy in diagnosis and 93.38% in severity assessment, outperforming baseline methods by over 10% in both tasks. This highlights its superior generalization and robustness across varying conditions and noise levels.
- 30 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Systems
- Vol. 40 (3) , 2061-2077
Frequency-Constrained Optimal Restoration Scheduling in Active Distribution Networks With Dynamic Boundaries for Networked Microgrids
This paper proposes a frequency security-constrained optimal restoration scheduling framework for active distribution networks (FRSDN). The approach leverages distributed energy resources (DERs), particularly the inverter-interfaced renewable energy generators (IIREGs). First, incorporating the equivalent aggregated frequency response from both the governors of synchronous thermal generators (TGs) and the virtual synchronous control manner of IIREGs, linearized and conservative constraints for dynamic frequency metrics are derived, in particular for load-pickup process in renewable-based distribution system restoration (DSR). Next, a sequential dynamic boundary identification model (SDIDM) is developed to identify the dynamic boundaries and the involved generator types of networked microgrids (MGs), which facilitates the precise imposition of frequency security constraints by conserving the serial numbers of dynamically clustered networked MGs. Finally, the frequency-constrained DSR method is formulated as a three-layer robust mixed-integer linear programming (MILP) problem and solved by the column-and-constraint generation (C&CG) algorithm. Numerical results on a modified IEEE 123-node system are leveraged to demonstrate the effectiveness and superiorities of the proposed method. It is manifested that the frequency responses of all boundary-aware MGs could be confined to secure conditions by implementing the proposed FRSDN.
- 30 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Systems
- Vol. 40 (3) , 2061-2077
Frequency-Constrained Optimal Restoration Scheduling in Active Distribution Networks With Dynamic Boundaries for Networked Microgrids
This paper proposes a frequency security-constrained optimal restoration scheduling framework for active distribution networks (FRSDN). The approach leverages distributed energy resources (DERs), particularly the inverter-interfaced renewable energy generators (IIREGs). First, incorporating the equivalent aggregated frequency response from both the governors of synchronous thermal generators (TGs) and the virtual synchronous control manner of IIREGs, linearized and conservative constraints for dynamic frequency metrics are derived, in particular for load-pickup process in renewable-based distribution system restoration (DSR). Next, a sequential dynamic boundary identification model (SDIDM) is developed to identify the dynamic boundaries and the involved generator types of networked microgrids (MGs), which facilitates the precise imposition of frequency security constraints by conserving the serial numbers of dynamically clustered networked MGs. Finally, the frequency-constrained DSR method is formulated as a three-layer robust mixed-integer linear programming (MILP) problem and solved by the column-and-constraint generation (C&CG) algorithm. Numerical results on a modified IEEE 123-node system are leveraged to demonstrate the effectiveness and superiorities of the proposed method. It is manifested that the frequency responses of all boundary-aware MGs could be confined to secure conditions by implementing the proposed FRSDN.
- Top Cited Papers
- 5 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 16 (1) , 45-61
Enabling High-Efficiency Economic Dispatch of Hybrid AC/DC Networked Microgrids: Steady-State Convex Bi-Directional Converter Models
The use of bi-directional converters (BDCs) is crucial for enhancing power exchange in hybrid AC/DC networked microgrids (NMGs). However, the dynamic nature of their conversion efficiency and the non-convex conversion direction expression of BDC models result in a highly non-convex programming problem, which leads to significant computational challenges. This paper proposes a least squares approximation approach to simplify the complex trigonometric function that characterizes the dynamic changes in BDC conversion efficiency with power. The simplified expression transforms the original non-convex relationship into a computationally efficient convex form. Subsequently, we systematically explore and validate the sufficient conditions to achieve non-simultaneous rectification and inversion behaviors of BDCs in a convex form. These explored conditions are further customized and extended to various practical application scenarios. Case studies are conducted on a hybrid AC/DC NMG with 66 nodes. The results demonstrate that our proposed least squares approximation method transforms the computationally intractable model into one that is solvable. Additionally, the explored conditions contribute to a substantial reduction in solution time by more than two orders of magnitude. These results verify the superiority of our proposed method, and showcase its applicability in practical scenarios.
- 5 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 16 (1) , 45-61
Top Cited PapersEnabling High-Efficiency Economic Dispatch of Hybrid AC/DC Networked Microgrids: Steady-State Convex Bi-Directional Converter Models
The use of bi-directional converters (BDCs) is crucial for enhancing power exchange in hybrid AC/DC networked microgrids (NMGs). However, the dynamic nature of their conversion efficiency and the non-convex conversion direction expression of BDC models result in a highly non-convex programming problem, which leads to significant computational challenges. This paper proposes a least squares approximation approach to simplify the complex trigonometric function that characterizes the dynamic changes in BDC conversion efficiency with power. The simplified expression transforms the original non-convex relationship into a computationally efficient convex form. Subsequently, we systematically explore and validate the sufficient conditions to achieve non-simultaneous rectification and inversion behaviors of BDCs in a convex form. These explored conditions are further customized and extended to various practical application scenarios. Case studies are conducted on a hybrid AC/DC NMG with 66 nodes. The results demonstrate that our proposed least squares approximation method transforms the computationally intractable model into one that is solvable. Additionally, the explored conditions contribute to a substantial reduction in solution time by more than two orders of magnitude. These results verify the superiority of our proposed method, and showcase its applicability in practical scenarios.
- Open Access
- 10 July 2024
- Published by MDPI AG in Sustainability
- Vol. 16 (14) , 5881
Unveiling the Dynamics of Residential Energy Consumption: A Quantitative Study of Demographic and Personality Influences in Singapore Using Machine Learning Approaches
In the pursuit of instigating a progressive transition towards a more sustainable future, policy officials all over the world are fervently advocating the use of energy conservation techniques targeted at residential customers. Keeping this in mind, a quantitative study was conducted in this work using the data from Singapore, which aims to investigate the relationships between a resident’s pattern of energy utilisation and numerous demographic parameters as well as personality attributes. Moreover, the study was conducted with existing machine learning and data analytics approaches, including k-prototype unsupervised learning and statistical hypothesis tests. The obtained results denote a persuasive correlation between the consumption behaviour of the consumer for different appliances and factors such as income, energy knowledge, usage frequency, personality, etc. For instance, there is a higher probability of a consumer acting frugally and sparingly if they believe their energy consumption is insignificant. These findings can help policymakers identify the appropriate target populations for raising energy awareness in Singapore.
- 10 July 2024
- Published by MDPI AG in Sustainability
- Vol. 16 (14) , 5881
Open AccessUnveiling the Dynamics of Residential Energy Consumption: A Quantitative Study of Demographic and Personality Influences in Singapore Using Machine Learning Approaches
In the pursuit of instigating a progressive transition towards a more sustainable future, policy officials all over the world are fervently advocating the use of energy conservation techniques targeted at residential customers. Keeping this in mind, a quantitative study was conducted in this work using the data from Singapore, which aims to investigate the relationships between a resident’s pattern of energy utilisation and numerous demographic parameters as well as personality attributes. Moreover, the study was conducted with existing machine learning and data analytics approaches, including k-prototype unsupervised learning and statistical hypothesis tests. The obtained results denote a persuasive correlation between the consumption behaviour of the consumer for different appliances and factors such as income, energy knowledge, usage frequency, personality, etc. For instance, there is a higher probability of a consumer acting frugally and sparingly if they believe their energy consumption is insignificant. These findings can help policymakers identify the appropriate target populations for raising energy awareness in Singapore.
- 9 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (10) , 11998-12009
Mutitask Learning Network for Partial Discharge Condition Assessment in Gas-Insulated Switchgear
Condition assessment for gas-insulated switchgear (GIS), which are crucial component of power systems, involves three interrelated aspects, i.e., partial discharge (PD) diagnosis, localization, and severity assessment. However, existing methods for GIS PD condition assessment perform these aspects as separate tasks, ignoring the mutual influence among them and leading to inferior performance. To settle the abovementioned issue, we propose a multitask learning network (MTLN) for GIS PD condition assessment. First, a multitask network was developed, taking severity assessment as the main task and diagnosis and localization as parallel auxiliary tasks. This model not only facilitates the extraction of the coupling relationship between diagnosis and localization but also furnishes pertinent feature information for severity assessment. Second, to deploy the developed model to label-free scenarios on-site, a novel subdomain adaptation is established. The process of subdomain adaptation considers the alignment of both intraclass and interclass information, incorporating a secondary filtering mechanism to mitigate the issue of feature mismatch caused by incorrect pseudo labels. Experimental results show that the proposed MTLN not only offers diagnosis and location information for severity assessment but also facilitates the exploration of the coupling relationship between diagnosis and localization, thereby enhancing the performance of GIS PD condition assessment.
- 9 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (10) , 11998-12009
Mutitask Learning Network for Partial Discharge Condition Assessment in Gas-Insulated Switchgear
Condition assessment for gas-insulated switchgear (GIS), which are crucial component of power systems, involves three interrelated aspects, i.e., partial discharge (PD) diagnosis, localization, and severity assessment. However, existing methods for GIS PD condition assessment perform these aspects as separate tasks, ignoring the mutual influence among them and leading to inferior performance. To settle the abovementioned issue, we propose a multitask learning network (MTLN) for GIS PD condition assessment. First, a multitask network was developed, taking severity assessment as the main task and diagnosis and localization as parallel auxiliary tasks. This model not only facilitates the extraction of the coupling relationship between diagnosis and localization but also furnishes pertinent feature information for severity assessment. Second, to deploy the developed model to label-free scenarios on-site, a novel subdomain adaptation is established. The process of subdomain adaptation considers the alignment of both intraclass and interclass information, incorporating a secondary filtering mechanism to mitigate the issue of feature mismatch caused by incorrect pseudo labels. Experimental results show that the proposed MTLN not only offers diagnosis and location information for severity assessment but also facilitates the exploration of the coupling relationship between diagnosis and localization, thereby enhancing the performance of GIS PD condition assessment.