Academic

From ARIMA to Attention: Power Load Forecasting Using Temporal Deep Learning

arXiv:2603.06622v1 Announce Type: new Abstract: Accurate short-term power load forecasting is important to effectively manage, optimize, and ensure the robustness of modern power systems. This paper performs an empirical evaluation of a traditional statistical model and deep learning approaches for predicting short-term energy load. Four models, namely ARIMA, LSTM, BiLSTM, and Transformer, were leveraged on the PJM Hourly Energy Consumption data. The data processing involved interpolation, normalization, and a sliding-window sequence method. Each model's forecasting performance was evaluated for the 24-hour horizon using MAE, RMSE, and MAPE. Of the models tested, the Transformer model, which relies on self-attention algorithms, produced the best results with 3.8 percent of MAPE, with performance above any model in both accuracy and robustness. These findings underscore the growing potential of attention-based architectures in accurately capturing complex temporal patterns in power con

S
Suhasnadh Reddy Veluru, Sai Teja Erukude, Viswa Chaitanya Marella · · 1 min read · 9 views
#cs.LG #cs.AI

arXiv:2603.06622v1 Announce Type: new Abstract: Accurate short-term power load forecasting is important to effectively manage, optimize, and ensure the robustness of modern power systems. This paper performs an empirical evaluation of a traditional statistical model and deep learning approaches for predicting short-term energy load. Four models, namely ARIMA, LSTM, BiLSTM, and Transformer, were leveraged on the PJM Hourly Energy Consumption data. The data processing involved interpolation, normalization, and a sliding-window sequence method. Each model's forecasting performance was evaluated for the 24-hour horizon using MAE, RMSE, and MAPE. Of the models tested, the Transformer model, which relies on self-attention algorithms, produced the best results with 3.8 percent of MAPE, with performance above any model in both accuracy and robustness. These findings underscore the growing potential of attention-based architectures in accurately capturing complex temporal patterns in power consumption data.

Executive Summary

This article presents an empirical evaluation of four models - ARIMA, LSTM, BiLSTM, and Transformer - for short-term power load forecasting using the PJM Hourly Energy Consumption data. The study reveals that the Transformer model, leveraging self-attention algorithms, outperforms other models in terms of accuracy, robustness, and MAPE. With a MAPE of 3.8 percent, the Transformer model showcases its potential in capturing complex temporal patterns in power consumption data. The findings highlight the growing significance of attention-based architectures in the field of power load forecasting. This study contributes to the development of more accurate and robust forecasting models, essential for modern power systems' optimization and management.

Key Points

  • ▸ The study evaluates four models - ARIMA, LSTM, BiLSTM, and Transformer - for short-term power load forecasting.
  • ▸ The Transformer model outperforms other models in terms of accuracy, robustness, and MAPE.
  • ▸ The study highlights the potential of attention-based architectures in capturing complex temporal patterns in power consumption data.

Merits

Strength

The study's empirical evaluation of multiple models provides a comprehensive comparison of their performance.

Methodological rigor

The use of MAE, RMSE, and MAPE metrics enables a thorough assessment of the models' forecasting accuracy.

Relevance

The study's focus on short-term power load forecasting is crucial for modern power systems' optimization and management.

Demerits

Limitation

The study's reliance on a single dataset, PJM Hourly Energy Consumption, may limit its generalizability to other regions or power systems.

Data preprocessing

The study's data processing steps, including interpolation and normalization, may not be universally applicable or optimal for all datasets.

Expert Commentary

This study contributes significantly to the development of accurate and robust forecasting models for power load forecasting. The Transformer model's performance highlights the potential of attention-based architectures in capturing complex temporal patterns in power consumption data. However, the study's reliance on a single dataset and limited data preprocessing steps may limit its generalizability and applicability. Future studies should aim to replicate these findings using diverse datasets and explore the potential of attention-based architectures in other areas of power systems, such as grid management and renewable energy integration.

Recommendations

  • ✓ Future studies should evaluate the performance of attention-based architectures on diverse datasets to assess their generalizability and robustness.
  • ✓ Researchers should explore the potential of attention-based architectures in other areas of power systems, such as grid management and renewable energy integration, to maximize their impact and applicability.

Sources

Related Articles