Electricity Demand Forecasting With a Modified Extreme-Learning Machine Algorithm DOI Creative Commons
Chen Chen,

Chuangang Ou,

Mingxiang Liu

et al.

Frontiers in Energy Research, Journal Year: 2022, Volume and Issue: 10

Published: Aug. 15, 2022

To operate the power grid safely and reduce cost of production, power-load forecasting has become an urgent issue to be addressed. Although many load models have been proposed, most still suffer from poor model training, limitations sensitive outliers, overfitting forecasts. The current load-forecasting methods may lead generation additional operating costs for system, even damage distribution network security related systems. address this issue, a new prediction with mixed loss functions was proposed. is based on Pinball–Huber’s extreme-learning machine whale optimization algorithm. In specific, Pinball–Huber loss, which insensitive outliers largely prevents overfitting, proposed as objective function (ELM) training. Based ELM, algorithm added improve it. At last, effect hybrid verified using two real datasets (Nanjing Taixing). Experimental results confirmed that can achieve satisfactory improvements both datasets.

Language: Английский

Prediction of shear stress induced by shoaling internal solitary waves based on machine learning method DOI
Zhuangcai Tian,

Hanlu Liu,

Shaotong Zhang

et al.

Marine Georesources and Geotechnology, Journal Year: 2022, Volume and Issue: 41(2), P. 221 - 232

Published: Oct. 21, 2022

Recently, the interactions between internal solitary waves (ISWs) and seabed have directed increasing attention to ocean engineering offshore energy. In particular, ISWs induce bottom currents pressure fluctuations in deep water. this paper, we propose a method for predicting shear stress induced by shoaling based on machine learning, developed approach can be used quickly determine safety stability of engineering. First, provided basic dataset model training. Four learning models were selected predict under different trim conditions. The results indicated that performance convolutional neural network-long short-term memory (CNN-LSTM) forest prediction was significantly better than three other tested models, including long (LSTM), support vector regression (SVR) network (DNN) models. Therefore, CNN-LSTM optimal ISWs. Specifically, each metric smaller three, root mean squared error standard deviation ratio closest 0.7. addition, outperformed SVR DNN terms length time. predicted values consistent with experimental values. paper ISWs, guide future field experiment designs, reduce damage caused promote development

Language: Английский

Citations

5
A Seasonal-Trend Decomposition and Single Dendrite Neuron-Based Predicting Model for Greenhouse Time Series DOI
Qianqian Li,

Houtian He,

Chenxi Xue

et al.

Environmental Modeling & Assessment, Journal Year: 2023, Volume and Issue: 29(3), P. 427 - 440

Published: Sept. 19, 2023

Language: Английский

Citations

1
Three-dimensional numerical modeling of sediment transport in a highly turbid estuary with pronounced seasonal variations DOI Creative Commons
Anh T. K., Nicolas Huybrechts, Isabel Jalón‐Rojas

et al.

International Journal of Sediment Research, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 1, 2024

Language: Английский

Citations

0
Electricity Demand Forecasting With a Modified Extreme-Learning Machine Algorithm DOI Creative Commons
Chen Chen,

Chuangang Ou,

Mingxiang Liu

et al.

Frontiers in Energy Research, Journal Year: 2022, Volume and Issue: 10

Published: Aug. 15, 2022

To operate the power grid safely and reduce cost of production, power-load forecasting has become an urgent issue to be addressed. Although many load models have been proposed, most still suffer from poor model training, limitations sensitive outliers, overfitting forecasts. The current load-forecasting methods may lead generation additional operating costs for system, even damage distribution network security related systems. address this issue, a new prediction with mixed loss functions was proposed. is based on Pinball–Huber’s extreme-learning machine whale optimization algorithm. In specific, Pinball–Huber loss, which insensitive outliers largely prevents overfitting, proposed as objective function (ELM) training. Based ELM, algorithm added improve it. At last, effect hybrid verified using two real datasets (Nanjing Taixing). Experimental results confirmed that can achieve satisfactory improvements both datasets.

Language: Английский

Citations

1