Browsing by Author "Cai, Lei"

Now showing 1 - 5 of 5
  • Thumbnail Image
    ItemOpen Access
    A bi-objective low-carbon economic scheduling method for cogeneration system considering carbon capture and demand response
    (Elsevier, 2023-12-14) Pang, Xinfu; Wang, Yibao; Yang, Shengxiang; Cai, Lei; Yu, Yang
    Carbon capture and storage (CCS), energy storage (ES), and demand response (DR) mechanisms are introduced into a cogeneration system to enhance their ability to absorb wind energy, reduce carbon emissions, and improve operational efficiency. First, a bi-objective low-carbon economic scheduling model of a cogeneration system considering CCS, ES, and DR was developed. In this model, the ES and CCS remove the coupling between power generation and heating. The DR mechanism, which is based on the time-of-use electricity price and heating comfort, further enhanced the flexibility of the system. In addition, an improved bare-bones multi-objective particle swarm optimisation (IBBMOPSO) was designed to directly obtain the Pareto front of the low-carbon economy scheduling model. The particle position update mode was improved to balance global and local search capabilities in various search stages. The Taguchi method was used to calibrate the algorithm parameters. The inverse generational distance (IGD), hypervolume (HV), and maximum spread (MS) were used to evaluate the distribution and convergence performance of the algorithm. The improved technique for order preference by similarity to an ideal solution (TOPSIS) method was utilised to obtain the optimal compromise solution. Finally, the proposed method was tested on a cogeneration system in Northeast China. According to the comparison results, the average economic cost of the cogeneration system considering CCS, ES, and DR was reduced by approximately 1.13%, and carbon emissions were reduced by 6.79%. The IBBMOPSO is more competitive than the NSGA-II, MOWDO, MOMA, MOPSO, and BBMOPSO in low-carbon economic scheduling for the cogeneration system.
  • Thumbnail Image
    ItemEmbargo
    A synthetic data generation method and evolutionary transformer model for degradation trajectory prediction in Lithium-ion batteries
    (Elsevier, 2024-10-22) Jin, Haiyan; Ru, Rui; Cai, Lei; Meng, Jinhao; Wang, Bin; Peng, Jichang; Yang, Shengxiang
    Identifying the long-term degradation of lithium-ion batteries in their early usage phase is crucial for the battery management system (BMS) to properly maintain the battery for practical use. Nevertheless, this procedure is challenging due to variations in the production and operating conditions of the battery. In recent years, it has been empirically proven that the data-driven method is a promising solution for handling the prediction of degradation. However, the lack of appropriate data remains the main obstacle that impacts the ultimate performance of the prediction. Furthermore, the prediction is also influenced by the setup of the predictor, which covers the structure of neural networks and their hyperparameters. The challenge of automating this process remains unresolved. In this study, we propose a novel degradation trajectory prediction framework. First, synthetic data is generated via a conditional generative adversarial network (CGAN), providing the characterization of the battery’s degradation at an early stage and utilizing the argument data to alleviate the issue of insufficient data. Second, an evaluation method to evaluate the quality of the synthetic data is also provided. In addition, a selection method is proposed based on the diversity mechanism to further filter out the redundancy of synthetic data. These two sub-processes aim to promote the quality of the synthetic data. Finally, the synthetic data hybrid with real values is used for the training of a transformer model, whose architecture and hyper-parameters are automatically configured via an evolutionary framework. The experimental results show that the proposed method can achieve accurate predictions compared to its rivals, and its best configuration can be automatically configured without hand-crafted efforts.
  • Thumbnail Image
    ItemOpen Access
    A unified deep learning optimization paradigm for Lithium-ion battery state-of-health estimation
    (IEEE, 2023-07-12) Cai, Lei; Cui, Ningmin; Jin, Haiyan; Meng, Jinhao; Yang, Shengxiang; Peng, Jichang; Zhao, Xinchao
    State-of-Health (SOH) is critical to managing the lifespan of the battery energy storage system. For the data-driven-based method, explicit features have many benefits, but they cannot be constructed automatically. To maximize the use of explicit features and automate its construction and other configuration processes, this paper design a unified optimization paradigm for synergizing the three fundamental procedures in the data-driven model, that is, feature extraction, importance assignment, and model parameterization. An Evolutionary Multi-Objective (EMO) method is used to synchronously find a series of non-dominated solutions for the best combination of features, attention layer, and hyper-parameters of the network, which can enable a flexible SOH estimation in various operation conditions. A piecewise aggregate approximation is designed to compress the partial voltage curve while keeping its tendency characteristics. A Long Short-Term Memory (LSTM) is used to establish the data-driven model, especially an attention layer is added to finalize the task of feature selection. Experimental results prove the proposed model can achieve accurate SOH estimation and also provide flexible solutions for different scenarios. And results with multi-battery validation and transfer learning demonstrate that the proposed method not only has a high generalization ability but also easily transfers to a new task.
  • Thumbnail Image
    ItemOpen Access
    An empirical-informed model for the early degradation trajectory prediction of lithium-ion battery
    (IEEE, 2024-04-04) Meng, Jinhao; Cai, Lei; Yang, Shengxiang; Li, Junxin; Zhou, Feifan; Peng, Jichang; Song, Zhengxiang
    Early prediction of the lithium-ion (Li-ion) battery degradation trajectory is of great importance to arrange the maintenance of battery energy storage systems (BESSs). Although extensive data driven methods have achieved a super good performance in state of health (SOH) and remaining useful life (RUL) prediction, the nonlinear characteristics of the Li-ion battery degradation trajectory still prevent an accurate prediction once very limited cycling data known in advance. To solve this issue, this paper proposes an empirical-informed model for the degradation trajectory prediction with only few data from the Li-ion battery's early cycling stage, which integrates the experience based knowledge to train the data driven model. A novel experience based model is proposed to describe the battery degradation curve, which further guides the training procedure of the long-short term memory (LSTM) network. In addition, XGBoost is selected to use a perceptually important point (PIP) based feature for providing the reference capacities. In this way, the proposed method can implement an end-to-end early prediction of the battery trajectory using only partial charging voltage as the input. The performance of the proposed method is verified on three datasets.
  • Thumbnail Image
    ItemOpen Access
    Automatically constructing a health indicator for lithium-ion battery state-of-health estimation via adversarial and compound staked autoencoder
    (Elsevier, 2024-02-17) Cai, Lei; Li, Junxin; Xu, Xianfeng; Jin, Haiyan; Meng, Jinhao; Wang, Bin; Yang, Shengxiang
    Precisely assessing the state of health (SOH) has emerged as a critical approach to ensuring the safety and dependability of lithium-ion batteries. One of the primary issues faced by SOH estimate methods is their susceptibility to the influence of noise in the observed variables. Moreover, we prefer to automatically extract explicit features for data-driven methods in certain circumstances. In light of these considerations, this paper proposes an adversarial and compound stacked autoencoder for automatically constructing the SOH estimation health indicator. The compound stacked autoencoder consists of two parts. The first one is a denoising autoencoder that learns three different denoising behaviors. The second is a feature-extracting autoencoder that employs adversarial learning to improve generalization ability. The experimental results show that the proposed compound stacked autoencoder can not only get explainable explicit features but also can achieve accurate SOH estimation results compared with its rivals. In addition, the results with transfer learning demonstrate that the proposed method not only can provide high generalization ability but also be easily transferred to a new SOH estimation task.