Louis Li

I graduated from the University of Manchester with a major in automatic control (QS Rank 27), and during my academic career, I achieved numerous accomplishments, including publishing 4 international papers and inventing 13 patents.

With 10 years of experience in algorithm engineering and 5 years of experience leading algorithm teams, I excel in image algorithms and combinatorial optimization algorithms.

Email  /  Github  /  Linkedin  /  Instagram  /  Blog  /  Medium  /  WCA  /  Lab

My research interests include exciting topics such as Diffusion, NerF, and AIGC. In the future, I hope to apply my skills and knowledge in emerging industries like blockchain, gaming, and e-commerce.

This paper focuses on predicting the State of Charge (SOC) for electric vehicle batteries, which is crucial for efficient energy management. Accurate estimation of SOC helps optimize charging and discharging control, driving performance, and battery service life. The paper proposes a method that uses principal component analysis for dimensionality reduction and the XGBoost algorithm to build an SOC prediction model based on battery terminal voltage, discharge current, and temperature. The model considers the complex internal structure of the battery to improve prediction accuracy.

This paper presents a method for dividing peak-valley periods and designing orderly charging strategies to ease pressure on the power grid. The approach uses optimal division, considers power supply stability and user satisfaction, and employs a Bi-level programming model with a genetic algorithm to generate a reasonable power supply plan. Experimental results indicate that this approach offers benefits such as peak-shaving, valley-filling, and load smoothing for the state grid company.

This paper proposes a multi-in-one orderly charging model that integrates vehicle, pile, road network, and power grid information to optimize vehicle charging and relieve traffic congestion. The model addresses real-time changes in charging and grid load information, incorporating location, battery, and SOC data as well as charging station capacity and queue size. By leveraging road network information, the model calculates optimal station locations and shortest routes for each vehicle, optimizing allocation during peak hours and improving overall efficiency. This promising solution offers a path to more efficient and organized vehicle charging.

This paper introduces a new method for modeling and reconstructing visual data using Bayesian density estimation. The method is based on mixtures of scaled Dirichlet distributions, which allows for parameter estimation, model selection, and accounting for uncertainty. Three key issues related to Bayesian mixture learning are addressed, including the choice of prior distributions, parameter estimation, and component selection. The paper also presents a principled algorithm for using scaled Dirichlet mixtures. The method is tested on a real-life visual scene reconstruction task.