Deep Learning with Hybrid Attention for Power System Net Load Forecasting: A CNN-LSTM-GRU Approach

Yutong Zhou

doi:10.54097/m8g76683

Authors

Yutong Zhou

DOI:

https://doi.org/10.54097/m8g76683

Keywords:

Net Load Prediction, CNN-LSTM-GRU, Attention Mechanism.

Abstract

As the share of renewable energy in power systems continues to increase, net load forecasting has become significantly more challenging, with prediction errors under highly variable weather conditions often exceeding 300–600 MW using conventional methods. This paper introduces a hybrid attention model based on a Convolutional Neural Network (CNN) Long-Short-erm Memory (LSTM) -Gated Recurrent Units (GRU) architecture for net load forecasting. The model integrates a convolutional neural network for spatial feature extraction, long short-term memory and gated recurrent units for multi-scale temporal modeling, and an attention mechanism to dynamically weight relevant time steps. It processes raw net load data directly, avoiding the computational cost and inaccuracies associated with traditional decomposition techniques, which can introduce delays of several minutes and amplify errors during rapid changes. Validation under severe weather conditions shows that the proposed model achieves a mean absolute percentage error (MAPE) of 1.96% and a mean absolute error (MAE) of 42.85 MW, outperforming standalone LSTM, GRU, and CNN models by a margin of 18–32% in MAPE under similar conditions. This performance enables more reliable grid dispatch and contributes to maintaining system operational stability.

Downloads

Download data is not yet available.

References

[1] Van der Meer, D. W, Munkhammar, J. & Widén, J. Probabilistic forecasting of solar power, electricity consumption and net load: Investigating the effect of seasons, aggregation and penetration on prediction intervals. Solar Energy, 2018, 171: 397 – 413.

[2] Tziolis, G, Lopez-Lorente, J, Baka, M.-I., Koumis, et al. Direct short-term net load forecasting in renewable integrated microgrids using machine learning: A comparative assessment. Sustainable Energy, Grids and Networks, 2024.

[3] Sreenu, S, Sharma, K. C, Bhakar, R. Gumbel copula based aggregated net load forecasting for modern power systems. IET Generation, Transmission & Distribution, 2018, 12: 4348 – 4358.

[4] A. Wahab, M.A. Tahir, N. Iqbal, A. Ul-Hasan, F. Shafait, S.M. Raza Kazmi. A novel technique for short-term load forecasting using sequential models and feature engineering. IEEE Access, 2021, 9: 96221 - 96232.

[5] Zhang Yun, Zhou Quan, Caixin Sun, Shaolan Lei, Yuming Liu, and Song Yang. RBF neural network and ANFIS-based short-term load forecasting approach in real-time price environment. IEEE Transactions on Power Systems, 2002, 23 (3): 853 - 858.

[6] T.Y. Kim, and S.B. Cho. Predicting residential energy consumption using CNN-LSTM neural networks. Energy, 2019, 182: 72 - 81.

[7] Monowar Hossain, and Saad Mekhilef. Application of Extreme Learning Machine for short term output power forecasting of three grid-connected PV systems. J Cleaner Prod, 2017, 167: 395 - 405.

[8] Liu, C, Zhang, K, Xiong, H, Jiang, G, Yang, Q. Temporal self-attention network for medical concept embedding. IEEE International Conference on Data Mining, 2020.

[9] Md. Rayhan Ahmed, Salekul Islam, A.K.M. Muzahidul Islam, Swakkhar Shatabda. An ensemble 1D-CNN-LSTM-GRU model with data augmentation for speech emotion recognition. Expert Systems with Applications, 2023.

[10] Himmet ÖZARSLAN, Ihsan ULUOCAK. CNN-LSTM and CNN-GRU based estimation of NOx conversion efficiency in diesel engine exhaust system. Fuel, 2026.

[11] K.e. Yan, X. Wang, Y. Du, N. Jin, H. Huang, H. Zhou. Multi-Step Short-Term Power Consumption Forecasting with a Hybrid Deep Learning Strategy. Energies 11, 2018.

[12] Md. A. Istiake Sunny, M. M. S. Maswood, et al. Deep Learning Based Stock Price Prediction Using LSTM and Bi-Directional LSTM Model. 2nd Novel Intelligent and Leading Emerging Sciences Conference, 2020.

[13] Osaka Rubasinghe, Tingze Zhang, Xinan Zhang, San Shing Choi, Tat Kei Chau, Yau Chow, Tyrone Fernando, Herbert Ho-Ching Iu. Highly accurate peak and valley prediction short-term net load forecasting approach based on decomposition for power systems with high PV penetration. Applied Energy, 2023.

[14] P. Ravindran, A. Costa, R. Soares, and A.C. Wiedenhoeft. Classification of CITES-listed and other neotropical Meliaceae wood images using convolutional neural networks. Plant Methods, 2018.

[15] Vasileios Pentsos, Spyros Tra goudas, Senior Member, IEEE, Jason Wibbenmeyer, and Nasser Khdeer, A Hybrid LSTM-Transformer Model for Power Load Forecasting. IEEE Transactions on Smart Grid, 2025.