Short-Term Load Forecasting Using Channel and Temporal Attention Based Temporal Convolutional Network

Tang, Xianlun; Chen, Hongxu; Xiang, Wenhao; Yang, Jingming; Zou, Mi

doi:10.1016/j.epsr.2021.107761

Cited by 88 publications

(29 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These models have recently drawn considerable attention in the field of STLF as they can result to relatively higher accuracy, lower training times, and more interpretable forecasts [33,39]. Specifically, Tang et al [59] employed a TCN architecture with channel and temporal attention mechanisms to exploit the non-linear relationships between weather factors and load. Yin and Xie [65] applied a multi-temporalspatial-scale TCN for forecasting the load of a city in China, while Gu and Jia [15] compared the TCN architecture with three traditional models, suggesting its superiority when used to forecast the load of a certain region in Shanghai.…”

Section: Related Workmentioning

confidence: 99%

A comparative assessment of deep learning models for day-ahead load forecasting: Investigating key accuracy drivers

Pelekis¹,

Seisopoulos²,

Spiliotis³

et al. 2023

Preprint

View full text Add to dashboard Cite

Short-term load forecasting (STLF) is vital for the daily operation of power grids. However, the nonlinearity, non-stationarity, and randomness characterizing electricity demand time series renders STLF a challenging task. To that end, different forecasting methods have been proposed in the literature for day-ahead load forecasting, including a variety of deep learning models that are currently considered to achieve state-of-the-art performance. In order to compare the accuracy of such models, we focus on national net aggregated STLF and examine well-established autoregressive neural networks of indicative architectures, namely multi-layer perceptrons, N-BEATS, long short-term memory neural networks, and temporal convolutional networks, for the case of Portugal. To investigate the factors that affect the performance of each model and identify the most appropriate per case, we also conduct a post-hoc analysis, correlating forecast errors with key calendar and weather features. Our results indicate that N-BEATS consistently outperforms the rest of the examined deep learning models. Additionally, we find that external factors can significantly impact accuracy, affecting both the actual and relative performance of the models.

show abstract

Section: Related Workmentioning

confidence: 99%

A comparative assessment of deep learning models for day-ahead load forecasting: Investigating key accuracy drivers

Pelekis¹,

Seisopoulos²,

Spiliotis³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…A TCN model is a type of neural network designed for sequential data processing (Tang et al, 2022). It uses 1D convolutional layers to extract features from the input sequence and has been shown to be effective for different tasks.…”

Section: Temporal Convolutional Network (Tcn)mentioning

confidence: 99%

An Advanced Deep Learning model for Predicting Groundwater Level

Ehteram

Ghanbari-Adivi

2023

Preprint

View full text Add to dashboard Cite

Groundwater level prediction is important for effective water management. Accurately predicting groundwater levels allows decision-makers to make informed decisions about water allocation, groundwater abstraction rates, and groundwater recharge strategies. Groundwater level prediction can also be used to develop more effective drought preparedness plans to mitigate the impact of water scarcity. In this study, we introduce a new model called self-attention (SA) temporal convolutional network (SATCN)-long short term memory neural network (SATCN-LSTM) model to predict groundwater level. The new model combines the advantages of the SATCN model and the LSTM model to overcome the limitations of the LSTM model. The SATCN model uses skip connections and self-attention mechanisms to overcome the vanishing gradient problem of the LSTM model, identify relevant and irrelevant data, and capture short-, and long-term dependencies of time series data. The new model was used to predict GWL in a large basin. Meteorological data were used to predict GWL. The SATCN-LSTM model outperformed the other models. The SATCN-LSTM model had the lowest mean absolute error (MAE) of 0.06, followed by the self-attention (SA) temporal convolutional network (SATCN) model with an MAE of 0.09. The SALSTM model had an MAE of 0.12, while the TCN-LSTM, TCN, and LSTM models had MAEs of 0.14, 0.15, and 0.17, respectively. The SATCN-LSTM model had the lowest root mean square error (RMSE) of 0.08, followed by SATCN with an RMSE of 0.11. The results of the SATCN-LSTM model provide valuable insights into the dynamics of groundwater systems. By accurately predicting groundwater levels, the SATCN-LSTM model can help ensure that groundwater resources are used sustainably and efficiently.

show abstract

“…The authors reported that a simple convolutional architecture is more effective across diverse sequence modelling tasks than recurrent architectures such as LSTMs and GRUs. TCNs have been recently validated for STLF tasks [55], [56] and are employed throughout our case study.…”

Section: Model Selectionmentioning

confidence: 99%

In Search of Deep Learning Architectures for Load Forecasting: A Comparative Analysis and the Impact of the Covid-19 Pandemic on Model Performance

Pelekis

Karakolis

Silva³

et al. 2022

2022 13th International Conference on Information, Intelligence, Systems &Amp; Applications (IISA)

View full text Add to dashboard Cite

In power grids, short-term load forecasting (STLF) is crucial as it contributes to the optimization of their reliability, emissions, and costs, while it enables the participation of energy companies in the energy market. STLF is a challenging task, due to the complex demand of active and reactive power from multiple types of electrical loads and their dependence on numerous exogenous variables. Amongst them, special circumstances-such as the COVID-19 pandemic-can often be the reason behind distribution shifts of load series. This work conducts a comparative study of Deep Learning (DL) architectures-namely Neural Basis Expansion Analysis Time Series Forecasting (N-BEATS), Long Short-Term Memory (LSTM), and Temporal Convolutional Networks (TCN)-with respect to forecasting accuracy and training sustainability, meanwhile examining their out-of-distribution generalization capabilities during the COVID-19 pandemic era. A Pattern Sequence Forecasting (PSF) model is used as baseline. The case study focuses on day-ahead forecasts for the Portuguese national 15-minute resolution net load time series. The results can be leveraged by energy companies and network operators (i) to reinforce their forecasting toolkit with state-of-the-art DL models; (ii) to become aware of the serious consequences of crisis events on model performance; (iii) as a high-level model evaluation, deployment, and sustainability guide within a smart grid context.

show abstract

Short-Term Load Forecasting Using Channel and Temporal Attention Based Temporal Convolutional Network

Cited by 88 publications

References 35 publications

A comparative assessment of deep learning models for day-ahead load forecasting: Investigating key accuracy drivers

A comparative assessment of deep learning models for day-ahead load forecasting: Investigating key accuracy drivers

An Advanced Deep Learning model for Predicting Groundwater Level

In Search of Deep Learning Architectures for Load Forecasting: A Comparative Analysis and the Impact of the Covid-19 Pandemic on Model Performance

Contact Info

Product

Resources

About