Multi-step wind power forecasting based on long-short-term Mamba

JING Huitian; HAN Li; WANG Xiaojing; YUAN Zhixin

doi:10.19783/j.cnki.pspc.260099

English Version Journal Login

您当前的位置：

首页 >

文章列表页 >

Multi-step wind power forecasting based on long-short-term Mamba

更新时间：2026-10-15

- Multi-step wind power forecasting based on long-short-term Mamba
- Power System Protection and Control Vol. 54, Issue 20, (2026)
- 作者机构：
  
  中国矿业大学电气工程学院,江苏,徐州,221116
- 作者简介：
- 基金信息：
- DOI：10.19783/j.cnki.pspc.260099
  CLC：
- Published：2026
- 稿件说明：
移动端阅览
JING Huitian, HAN Li, WANG Xiaojing, et al. Multi-step wind power forecasting based on long-short-term Mamba[J]. Power System Protection and Control, 2026, 54(20).
DOI：

JING Huitian, HAN Li, WANG Xiaojing, et al. Multi-step wind power forecasting based on long-short-term Mamba[J]. Power System Protection and Control, 2026, 54(20). DOI： 10.19783/j.cnki.pspc.260099.

摘要

准确的风电功率预测对风电场运行与电网调度优化至关重要。针对风电功率时间序列蕴含长期趋势与短期波动等多尺度动态特性，长期趋势与短期波动对模型记忆能力的要求不同，现有方法难以同时高效捕捉的问题，提出一种基于长短时Mamba(long-short-term Mamba

LSTMA)的风电功率多步预测方法。首先，将Mamba结合两种高阶多项式投影算子(high-order polynomial projection operator

HiPPO)初始化策略构建双通道状态空间模块，分别实现对长期趋势的记忆保持与对短期波动的精细化跟踪。然后，引入专家混合(mixture of experts

MoE)机制构建动态路由网络，自适应选择最优patch大小以处理短期波动模式。最后，设计门控函数动态融合双通道输出，实现多尺度预测结果的自适应加权。基于中国某风电场实际数据的实验表明，所提模型在4个季节的预测平均绝对百分比误差(mean absolute percentage error

MAPE)稳定在8%以下，决定系数保持在97%以上，较对比模型精度有所提升，验证了其在多尺度特征捕捉上的优越性。

Abstract

Accurate wind power forecasting is crucial for wind farm operation and grid dispatch optimization. Wind power time series exhibit multi-scale dynamic characteristics

including long-term trends and short-term fluctuations

which impose different requirements on the memory capacity of the model. Existing methods struggle to efficiently capture both types of dynamics simultaneously. To address this issue

a multi-step wind power forecasting method based on long-short-term Mamba (LSTMA) is proposed. First

Mamba is combined with two initialization strategies based on high-order polynomial projection operator (HiPPO) to construct a dual-channel state-space module

enabling the retention of long-term patterns and the fine-grained tracking of short-term fluctuations

respectively. Then

a mixture of experts (MoE) mechanism is introduced to build a dynamic routing network that adaptively selects the optimal patch size to process short-term fluctuation patterns. Finally

a gating function is designed to dynamically fuse the dual-channel outputs

achieving adaptive weighting of multi-scale forecasting results. Experiments based on real data from a wind farm in China show that the proposed model maintains a mean absolute percentage error (MAPE) below 8% and a coefficient of determination above 97% across all four seasons. Compared with the benchmark models

the proposed model improves forecasting accuracy

validating its superiority in multi-scale feature capturing.

关键词

Keywords

references

LIU H, CHAO C, LÜ X, et al. Deterministic wind energy forecasting: a review of intelligent predictors and auxiliary methods[J]. Energy Conversion and Management, 2019, 195: 328-345.

TANG Zao, LIU Jia, NI Jielong, et al. Power prediction of wind farm considering the wake effect and its boundary layer compensation[J]. Power System Protection and Control, 2024, 9(6): 19-29.

AASIM, SINGH S, MOHAPATRA A. Repeated wavelet transform based ARIMA model for very short-term wind speed forecasting[J]. Renewable Energy, 2019, 136: 758-768.

LIU C, ZHANG X, MEI S. Numerical weather prediction enhanced wind power forecasting: rank ensemble and probabilistic fluctuation awareness[J]. Applied Energy, 2022, 313.

ALY H. A hybrid optimized model of adaptive neuro- fuzzy inference system，recurrent Kalman filter and neuro-wavelet for wind power forecasting driven by DFIG[J]. Energy, 2022, 239.

HONG Y, SATRIANI T. Day-ahead spatiotemporal wind speed forecasting using robust design-based deep learning neural network[J]. Energy, 2020, 209.

HE X, HE B, QIN T, et al．Ultra-short-term wind power forecasting based on a dual-channel deep learning model with improved coot optimization algorithm[J]. Energy, 2024, 305.

JAVAID A, SAJID M, UDDIN E, et al. Sustainable urban energy solutions: forecasting energy production for hybrid solar-wind systems[J]. Energy Conversion and Management, 2024, 302.

ALI A, ZIMERMAN I, WOLF L. The hidden attention of mamba models[J]. arXiv preprint arXiv: 2403.01590, 2024.

GU A, DAO T. Mamba: linear-time sequence modeling with selective state spaces[J]. arXiv preprint arXiv: 2403.01590, 2024.

ZHU L, LIAO B, ZHANG Q, et al. Vision mamba: efficient visual representation learning with bidirectional state space model[C] // Proceedings of the 41st International Conference on Machine Learning, July 21-27, 2024, Vienna, Austria: 62429-62442.

WANG Honglin, XU Yaohua, ZHU Cheng. Enhancing medical image classification with BSDA-Mamba: integrating Bayesian random semantic data augmentation and residual connections[J]. Computers, Materials and Continua, 2025, 83(3): 4999-5018.

LIEBER O, LENZ B, BATA H, et al. Jamba: a hybrid Transformer-Mamba language model[J]. arXiv preprint arXiv: 2403.19887, 2024.

AHAMED A, CHENG Q. Time machine: a time series is worth 4 mambas for long-term forecasting[C] // 27th European Conference on Artificial Intelligence, October 19-24, 2024, Santiago de Compostela, Spain: 1688-1695.

WANG Z, KONG F, FENG S, et al. Is Mamba effective for time series forecasting[J]. Neurocomputing, 2025, 619.

BHETHANABHOTLA K, SWELAM O, SIEMS J, et al. Mamba4Cast: efficient zero-shot time series forecasting with state space models[C] // Thirty-eighth Annual Conference on Neural Information Processing Systems, December 10-15, 2024, Vancouver, Canada: 1-12.

SHAZEER N, MIRHOSEINI A, MAZIARZ K, et al. Outrageously large neural networks: the sparsely-gated mixture-of-experts layer[C] // International Conference on Learning Representations, April 24-26, 2017, Toulon, France: 1-19.

GU A, DAO T, ERMON S, et al. Hippo: recurrent memory with optimal polynomial projections[J]. Advances in Neural Information Processing Systems 2020, 33: 1474-1487.

XU X, CHEN C, LIANG Y. SST: multi-scale hybrid Mamba-Transformer experts for long-short range time series forecasting[J]. arXiv preprint arXiv: 2404. 14757, 2024.

GU A, JOHNSON I, SAAB K, et al. Combining recurrent, convolutional, and continuous-time models with linear state space layers[C] // Proceedings of the 35th International Conference on Neural Information Processing Systems, December 6-14, 2021, New York, USA(Virtual): 572-585.

ZHOU H, ZHANG S, PENG J, et al. Informer: beyond efficient transformer for long sequence time-series forecasting[C] // Proceedings of the AAAI Conference on Artificial Intelligence, February 2-9, 2021, Vancouver, Canada (Virtual): 11106-11115.

NIE Y, NGUYEN N, SINTHONG P, et al. A time series is worth 64 words: long-term forecasting with transformers[C] // 11th International Conference on Learning Representations, May 1-5, 2023, Kigali, Rwanda: 1-24.

LI Z, QI S, LI Y, et al. Revisiting long-term time series forecasting: an investigation on linear mapping[J]. arXiv preprint arXiv: 2305.10721, 2023.

ZENG A, CHEN M, ZHANG L, et al. Are transformers effective for time series forecasting?[C] // Proceedings of the AAAI Conference on Artificial Intelligence, February 7-14, 2023, Washington, USA: 11121-11128.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

基于联盟合作博弈的风电数据定价方法

Market-oriented dual-type demand response operation strategy for electric vehicle aggregators considering temperature effects

Maximum hosting capacity evaluation method for distributed generations under fixed overcurrent protection settings

A bi-level leader-follower game optimization model based on CVaR constraints and its application in multi-policy scenarios

Voltage control of single-phase grid-forming inverters based on a model-assisted extended state observer with a quasi-proportional resonant controller

Related Author

HUO Wei

ZHANG Yao

ZHAO Hanting

WANG Jianxue

Related Institution

No data

AI问答

⁰