<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>Accurate short-term power load forecasting is essential to balance energy supply and demand, thus minimizing operating costs. However, power load data possesses temporal and nonlinear characteristics, and to mitigate the effects of these factors on the prediction results, we introduce the Long Short-Term Memory neural network (LSTM, Long Short-Term Memory). However, the performance of the LSTM algorithm is highly dependent on the pre-set parameters, and relying on empirically set parameters will make the model have low generalization performance and reduce the prediction effect. In this regard, a prediction model (CWOA-LSTM) combining improved whale optimization algorithm and LSTM is proposed. The whale population is initialized using Circle chaotic sequences; nonlinear time-varying factors, inertial weight balance and Corsi variance are introduced. CWOA optimized the parameters of LSTM, and the experimental results showed that the MAE, MAPE, and RMSE of CWOA-LSTM were reduced by 13.1775 MV, 0.18423%, and 17.415 MV, respectively, compared with LSTM, which verified the accuracy and stability of CWOA-LSTM model.</p>
</abstract>

Accurate short-term power load forecasting is essential to balance energy supply and demand, thus minimizing operating costs. However, power load data possesses temporal and nonlinear characteristics, and to mitigate the effects of these factors on the prediction results, we introduce the Long Short-Term Memory neural network (LSTM, Long Short-Term Memory). However, the performance of the LSTM algorithm is highly dependent on the pre-set parameters, and relying on empirically set parameters will make the model have low generalization performance and reduce the prediction effect. In this regard, a prediction model (CWOA-LSTM) combining improved whale optimization algorithm and LSTM is proposed. The whale population is initialized using Circle chaotic sequences; nonlinear time-varying factors, inertial weight balance and Corsi variance are introduced. CWOA optimized the parameters of LSTM, and the experimental results showed that the MAE, MAPE, and RMSE of CWOA-LSTM were reduced by 13.1775 MV, 0.18423%, and 17.415 MV, respectively, compared with LSTM, which verified the accuracy and stability of CWOA-LSTM model.

Short-term power load forecasting model based on multi-strategy improved WOA optimized LSTM

Key Laboratory of Advanced Manufacturing and Automation Technology

Applied Mathematics and Nonlinear Sciences

This work is licensed under the Creative Commons Attribution 4.0 International License.

Accurate short-term power load forecasting is essential to balance energy supply and demand, thus minimizing operating costs. However, power load data...

Short-term power load forecasting model based on multi-strategy improved WOA optimized LSTM

Pubblicato online: 26 feb 2024

Pagine: -

Ricevuto: 02 gen 2024

Accettato: 09 gen 2024

DOI: https://doi.org/10.2478/amns-2024-0323

Parole chiave
Short-term electric load forecasting, Long and short-term memory networks, Whale optimization algorithm, Machine learning, Adaptive Weights

© 2024 Qian Liang et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Short-term power load forecasting model based on multi-strategy improved WOA optimized LSTM

Pubblicato online: 26 feb 2024

Pagine: -

Ricevuto: 02 gen 2024

Accettato: 09 gen 2024

DOI: https://doi.org/10.2478/amns-2024-0323

Parole chiaveShort-term electric load forecasting, Long and short-term memory networks, Whale optimization algorithm, Machine learning, Adaptive Weights

© 2024 Qian Liang et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Parole chiave
Short-term electric load forecasting, Long and short-term memory networks, Whale optimization algorithm, Machine learning, Adaptive Weights