Electric Vehicle Charging Load Forecasting Method Based on Improved Long Short-Term Memory Model with Particle Swarm Optimization

World Electric Vehicle Journal, Год журнала: 2025, Номер 16(3), С. 150 - 150

Опубликована: Март 5, 2025

With the rapid global proliferation of electric vehicles (EVs), their integration as a significant load component within power systems increasingly influences stable operation and planning electrical grids. However, high uncertainty randomness inherent in EV users’ charging behaviors render accurate forecasting challenging task. In this context, present study proposes Particle Swarm Optimization (PSO)-enhanced Long Short-Term Memory (LSTM) network model. By combining search capability PSO algorithm with advantages LSTM networks time-series modeling, PSO-LSTM hybrid framework optimized for seasonal variations is developed. The results confirm that model effectively captures variations, providing high-precision, adaptive solution dynamic grid scheduling infrastructure planning. This supports optimization resource allocation enhancement energy storage efficiency. Specifically, during winter, Mean Absolute Error (MAE) 3.896, reduction 6.57% compared to 10.13% Gated Recurrent Unit (GRU) During winter–spring transition, MAE 3.806, which 6.03% lower than 12.81% GRU spring, 3.910, showing 2.71% improvement over 7.32%

Язык: Английский

An adaptive search mechanism with convolutional learning networks for online social media text summarization and classification model

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 1, 2025

The fast development of social media platforms has led to an unprecedented growth daily short text content. Removing valued patterns and insights from this vast amount textual data requires advanced methods provide information while preserving its essential components successfully. A summarization system takes more than one document as input tries give a fluent concise summary the most significant in input. Recent solutions for condensing reading are ineffective time-consuming, provided plenty is available online. Concerning challenge, automated have developed convincing choice, achieving important significance their growth. It was separated into two kinds according abstraction utilized: abstractive (AS) extractive (ES). Furthermore, automatic many applications spheres impact. This manuscript proposes Adaptive Search Mechanism Based Hierarchical Learning Networks Social Media Data Summarization Classification Model (ASMHLN-SMDSCM) technique. ASMHLN-SMDSCM approach aims present novel on using deep learning models. To accomplish that, proposed model performs pre-processing, which contains dissimilar levels employed handle unprocessed data. BERT used feature extraction process. moth search algorithm (MSA)-based hyperparameter selection process performed optimize results model. Finally, classification uses TabNet convolutional neural network (TabNet + CNN) efficiency method validated by comprehensive studies FIFA FARMER datasets. experimental validation illustrated superior accuracy value 98.87% 98.55% over recent techniques.

Язык: Английский

PoolKANNeXt: A new pooling-based Kolmogorov Arnold convolutional neural network

Alexandria Engineering Journal, Год журнала: 2025, Номер 128, С. 144 - 152

Опубликована: Май 26, 2025

Язык: Английский