publications

2025

1. Deep learning predicts real-world electric vehicle direct current charging profiles and durations

   Siyi Li, Mingrui Zhang , Robert Doel , and 2 more authors

   Nature Communications, 2025

   Abs DOI

   Accurate prediction of electric vehicle charging profiles and durations is critical for adoption and optimising infrastructure. Direct current fast charging presents complex behaviours shaped by many factors. This work introduces a deep learning framework trained on 909,135 real-world sessions, capable of predicting charging profiles and durations from minimal input with uncertainty estimates. The model initiates predictions from a single point on the power and state-of-charge profile and incrementally refines them as new observations arrive, enabling real-time updates. The model generalises across vehicle types and charging scenarios. It achieves 90% accuracy in predicting charging duration from a single point, and 95% accuracy with an absolute error under one minute using six points within five minutes. This work shows that using readily available input data at charge time enables accurate prediction of charging behaviour and offers a practical, scalable solution for deployment, energy planning, and infrastructure reliability.

2. Machine learning for modelling unstructured grid data in computational physics: A review

   Sibo Cheng , Marc Bocquet , Weiping Ding , and 20 more authors

   Information Fusion, 2025

   Abs DOI

   Unstructured grid data are essential for modelling complex geometries and dynamics in computational physics. Yet, their inherent irregularity presents significant challenges for conventional machine learning (ML) techniques. This paper provides a comprehensive review of advanced ML methodologies designed to handle unstructured grid data in high-dimensional dynamical systems. Key approaches discussed include graph neural networks, transformer models with spatial attention mechanisms, interpolation-integrated ML methods, and meshless techniques such as physics-informed neural networks. These methodologies have proven effective across diverse fields, including fluid dynamics and environmental simulations. This review is intended as a guidebook for computational scientists seeking to apply ML approaches to unstructured grid data in their domains, as well as for ML researchers looking to address challenges in computational physics. It places special focus on how ML methods can overcome the inherent limitations of traditional numerical techniques and, conversely, how insights from computational physics can inform ML development. For this purpose, we mainly focus in this review on recent papers from the past decade that reflect strong interactions between computational physics and deep learning methods. To support benchmarking, this review also provides a summary of open-access datasets of unstructured grid data in computational physics. Finally, emerging directions such as generative models with unstructured data, reinforcement learning for mesh generation, and hybrid physics-data-driven paradigms are discussed to inspire future advancements in this evolving field.

3. Facing & mitigating common challenges when working with real-world data: The Data Learning Paradigm

   Jake Lever , Sibo Cheng , César Quilodrán Casas , and 8 more authors

   Journal of Computational Science, 2025

   Abs DOI

   The rapid growth of data-driven applications is ubiquitous across virtually all scientific domains, and has led to an increasing demand for effective methods to handle data deficiencies and mitigate the effects of imperfect data. This paper presents a guide for researchers encountering real-world data-driven applications, and the respective challenges associated with this. This article proposes the concept of the Data Learning Paradigm, combining the principles of machine learning, data science and data assimilation to tackle real-world challenges in data-driven applications. Models are a product of the data upon which they are trained, and no data collected from real world scenarios is perfect due to natural limitations of sensing and collection. Thus, computational modelling of real world systems is intrinsically limited by the various deficiencies encountered in real data. The Data Learning Paradigm aims to leverage the strengths of data improvement to enhance the accuracy, reliability, and interpretability of data-driven models. We outline a range of methods which are currently being implemented in the field of Data Learning involving machine learning and data science methods, and discuss how these mitigate the various problems associated with data-driven models, illustrating improved results in a multitude of real world applications. We highlight examples where these methods have led to significant advancements in fields such as environmental monitoring, planetary exploration, healthcare analytics, linguistic analysis, social networks, and smart manufacturing. We offer a guide to how these methods may be implemented to deal with general types of limitations in data, alongside their current and potential applications.

2024

1. Towards universal mesh movement networks

   Mingrui Zhang , Chunyang Wang , Stephan Kramer , and 5 more authors

   In Proceedings of the 38th International Conference on Neural Information Processing Systems, Vancouver, BC, Canada, 2024

   Abs

   Solving complex Partial Differential Equations (PDEs) accurately and efficiently is an essential and challenging problem in all scientific and engineering disciplines. Mesh movement methods provide the capability to improve the accuracy of the numerical solution without increasing the overall mesh degree of freedom count. Conventional sophisticated mesh movement methods are extremely expensive and struggle to handle scenarios with complex boundary geometries. However, existing learning-based methods require re-training from scratch given a different PDE type or boundary geometry, which limits their applicability, and also often suffer from robustness issues in the form of inverted elements. In this paper, we introduce the Universal Mesh Movement Network (UM2N), which - once trained - can be applied in a non-intrusive, zero-shot manner to move meshes with different size distributions and structures, for solvers applicable to different PDE types and boundary geometries. UM2N consists of a Graph Transformer (GT) encoder for extracting features and a Graph Attention Network (GAT) based decoder for moving the mesh. We evaluate our method on advection and Navier-Stokes based examples, as well as a real-world tsunami simulation case. Our method out-performs existing learning-based mesh movement methods in terms of the benchmarks described above. In comparison to the conventional sophisticated Monge-Ampère PDE-solver based method, our approach not only significantly accelerates mesh movement, but also proves effective in scenarios where the conventional method fails. Our project page can be found at https://erizmr.github.io/UM2N/.

2. Learning to optimise wind farms with graph transformers

   Siyi Li, Arnaud Robert , A. Aldo Faisal , and 1 more author

   Applied Energy, 2024

   Abs DOI

   This work proposes a novel data-driven model capable of providing accurate predictions for the power generation of all wind turbines in wind farms of arbitrary layout, yaw angle configurations and wind conditions. The proposed model functions by encoding a wind farm into a fully connected graph and processing the graph representation through a graph transformer. The resultant graph transformer surrogate demonstrates robust generalisation capabilities and effectively uncovers latent structural patterns embedded within the graph representation of wind farms. The versatility of the proposed approach extends to the optimisation of yaw angle configurations through the application of genetic algorithms. This evolutionary optimisation strategy facilitated by the graph transformer surrogate achieves prediction accuracy levels comparable to industrially standard wind farm simulation tools, with a relative accuracy of more than 99% in identifying optimal yaw angle configurations of previously unseen wind farm layouts. An additional advantage lies in the significant reduction in computational costs, positioning the proposed methodology as a compelling tool for efficient and accurate wind farm optimisation.

2023

1. End-to-end wind turbine wake modelling with deep graph representation learning

   Siyi Li, Mingrui Zhang , and Matthew D. Piggott

   Applied Energy, 2023

   Abs DOI

   Wind turbine wake modelling is of crucial importance to accurate resource assessment, to layout optimisation, and to the operational control of wind farms. This work proposes a surrogate model for the representation of wind turbine wakes based on a state-of-the-art graph representation learning method termed a graph neural network. The proposed end-to-end deep learning model operates directly on unstructured meshes and has been validated against high-fidelity data, demonstrating its ability to rapidly make accurate 3D flow field predictions for various inlet conditions and turbine yaw angles. The specific graph neural network model employed here is shown to generalise well to unseen data and is less sensitive to over-smoothing compared to common graph neural networks. A case study based upon a real world wind farm further demonstrates the capability of the proposed approach to predict farm scale power generation. Moreover, the proposed graph neural network framework is flexible and highly generic and as formulated here can be applied to any steady state computational fluid dynamics simulations on unstructured meshes.

2020

1. ATRX/EZH2 complex epigenetically regulates FADD/PARP1 axis, contributing to TMZ resistance in glioma

   Bo Han , Xiangqi Meng , Pengfei Wu , and 8 more authors

   Theranostics, 2020

   DOI
2. Optimization of Mining–Mineral Processing Integration Using Unsupervised Machine Learning Algorithms

   Siyi Li, Yuksel Asli Sari , and Mustafa Kumral

   Natural Resources Research, 2020

   DOI

2019

1. Dimensioning a stockpile operation using principal component analysis

   Siyi Li, Marco Werk , Louis St-Pierre , and 1 more author

   International Journal of Minerals, Metallurgy, and Materials, 2019

   Abs DOI

   Mineral processing plants generally have narrow tolerances for the grades of their input raw materials, so stockpiles are often maintained to reduce material variance and ensure consistency. However, designing stockpiles has often proven difficult when the input material consists of multiple sub-materials that have different levels of variances in their grades. In this paper, we address this issue by applying principal component analysis (PCA) to reduce the dimensions of the input data. The study was conducted in three steps. First, we applied PCA to the input data to transform them into a lower-dimension space while retaining 80% of the original variance. Next, we simulated a stockpile operation with various geometric stockpile configurations using a stockpile simulator in MATLAB. We used the variance reduction ratio as the primary criterion for evaluating the efficiency of the stockpiles. Finally, we used multiple regression to identify the relationships between stockpile efficiency and various design parameters and analyzed the regression results based on the original input variables and principal components. The results showed that PCA is indeed useful in solving a stockpile design problem that involves multiple correlated input-material grades.

2. Risk of Adverse Vascular Events in Patients with Malignant Glioma Treated with Bevacizumab Plus Irinotecan: A Systematic Review and Meta-Analysis

   Jiawei Dong , Xiangqi Meng , Siyi Li, and 4 more authors

   World Neurosurgery, 2019

   DOI
3. New approaches to cognitive work analysis through latent variable modeling in mining operations

   Siyi Li, Yuksel Asli Sari , and Mustafa Kumral

   International Journal of Mining Science and Technology, 2019

   DOI