Objective Under the guidance of the “dual carbon” target, wind power, as a critical component of clean energy, plays a crucial role in its efficient utilization. Short-term wind power forecasting helps improve grid stability, optimize wind farm power generation plans, and reduce operating costs, enhancing the economic benefits of wind power and supporting the goals of low-carbon development. In addition, the prediction results provide valuable reference information for wind farms, assist the dispatching department in adjusting the generation plan in advance, reduce the impact of wind power grid connection on the grid, and ensure the safe operation of the power system. Given the instability and high volatility of wind power generation, this study proposes a short-term wind power prediction method based on BWO‒VMD and TCN‒BiGRU to improve the accuracy of wind power prediction and better support the energy transition under the “dual carbon” strategy. Methods A short-term wind power generation prediction model based on the beluga whale optimization (BWO) algorithm, variational mode de-composition (VMD), temporal convolutional network (TCN), and bidirectional gated recurrent unit (BiGRU) was carefully proposed to improve the prediction accuracy of wind power generation, particularly considering its inherent instability and high volatility. Firstly, considering the comprehensive and complex impact of various meteorological factors on wind power generation, the random forest (RF) method was employed. This involves a comprehensive process of carefully determining the importance of various meteorological factor characteristics, systematically and accurately ranking them, and then extracting the truly optimal features that have a significant impact on subsequent predictions. Secondly, VMD is effectively utilized to decompose raw power data, which is originally in a non-stationary sequence, into relatively stationary sub-sequences. However, due to its complexity, it is difficult to manually determine the two parameters. Therefore, the BWO algorithm began optimizing these parameters of VMD. On this basis, a comprehensive index combining sample entropy and VMD decomposition to reconstruct the errors of each order component is used as a fitness function. Through this method, a thorough search was conducted to identify the optima parameter combination. Then, the optimized VMD (OVMD) is utilized to decompose non-stationary power signals. Then, the decomposed sta-tionary subsequence is combined with carefully extracted optimal features and input into the TCN‒BiGRU combination model for prediction. This combination model aims to use the advantages of TCN and BiGRU to process data and make more accurate predictions. Finally, the pre-dicted values of each subsequence are sequentially stacked to obtain the result, which is expected to provide reliable predictions for wind power generation. Results and Discussions The RF algorithm is strategically employed to screen meteorological features and systematically rank their importance, enabling the accurate selection of features that significantly impact wind power forecasting. The experimental results indicate that wind speeds at vertical heights of 10, 30, and 50 m from the ground play an important role in influencing the accuracy of wind prediction. VMD is adopted to address the non-stationarity of wind power generation; however, manually determining its two parameters has proven to be challenging. Therefore, the BWO is proposed to optimize these parameters, with sample entropy and error reconstruction serving as key fitness function indicators. Compared to other optimization algorithms such as genetic algorithm (GA) and whale optimization algorithm (WOA), the BWO algorithm demonstrates significant per-formance, with faster running speed, stronger stability, and greater robustness. Then, the optimized VMD is utilized to decompose the non-stationary power signal, resulting in higher-quality subsequences and ultimately improving prediction accuracy. The dataset is carefully divided into a training set, a validation set, and a testing set to verify the accuracy of the model. This study selects a single model to compare the BiGRU network model and the OVMD‒TCN‒BiGRU combination model proposed in this study with other combination models for experimental a-nalysis. Through the graph, error evaluation indicators, and time indicators, the experimental results show that although the time of the proposed model is not optimal, its error evaluation indicator value is the smallest, highlighting its advantages. In addition, experiments are conducted not only on the main dataset but also extended to January and August data, which represent seasonal differences, for generalization to verify the re-liability and broad applicability of the model. The verification results indicate that the constructed model effectively handles various datasets and complex time series features, with strong robustness and generality, and can run stably and efficiently in various practical scenarios. Conclusions The decomposition of raw wind power data is primarily examined using the OVMD algorithm in this study. The meteorological factors selected by combining the decomposed sub-components with RF features are input into the TCN‒BiGRU model for prediction. Their respective advantages are integrated to enhance the accuracy and stability of the prediction. The experimental results indicated that applying this series of methods improves the prediction accuracy. Its capacity to manage complex time series data is demonstrated, and the advancement and innovation of algorithms and models are supported. New directions for technological progress and practical application in the field of wind power prediction are established.
XiaoYulong, ZouChongzhe, ChiHetian,et al.Boosted GRU model for short-term forecasting of wind power with feat-ure-weighted principal component analysis[J].Energy,2023,267:126503. doi:10.1016/j.energy.2022.126503
[2]
HuangLing, LiLinxia, ChengYu,et al.Short-term wind po-wer prediction based on SAM‒WGAN‒GP[J].Acta Energiae Solaris Sinica,2023,44(4):180‒188.
SunRongfu, ZhangTao, HeQing,et al.Review on key technologies and applications in wind power forecasting[J].High Voltage Engineering,2021,47(4):1129‒1143. doi:10.13336/j.1003-6520.hve.20201780
Krishna RayiV, MishraS P, NaikJ,et al.Adaptive VMD based optimized deep learning mixed kernel ELM autoencoder for single and multistep wind power forecasting[J].Energy,2022,244:122585. doi:10.1016/j.energy.2021.122585
[7]
LiGuoquan, GaoJianyu, BaiTianyu,et al.Wind power prediction based on improved crow search algorithm and support vector machine[J].Foreign Electronic Measurement Technology,2022,41(2):40‒45.
LahouarA, Ben Hadj SlamaJ.Hour-ahead wind power for-ecast based on random forests[J].Renewable Energy,2017,109:529‒541. doi:10.1016/j.renene.2017.03.064
[10]
LiuXing, WangYan, JiZhicheng.Short-term wind power prediction method based on random forest[J].Journal of Sy-stem Simulation,2021,33(11):2606‒2614. doi:10.16182/j.issn1004731x.joss.21-FZ0705
NiuZhewen, YuZeyuan, LiBo,et al.Short-term wind pow-er forecasting model based on deep gated recurrent unit ne-ural network[J].Electric Power Automation Equipment,2018,38(5):36‒42. doi:10.16081/j.issn.1006-6047.2018.05.005
Niksa‒RynkiewiczT, StommaP, WitkowskaA,et al.An intelligent approach to short-term wind power prediction using deep neural networks[J].Journal of Artificial Intelligence and Soft Computing Research,2023,13(3):197‒210. doi:10.2478/jaiscr-2023-0015
[17]
GuoLing, XuQingshan, ZhengLe.A forecasting method for short-term load based on TCN‒GRU model[J].Electric Power Engineering Technology,2021,40(3):66‒71.
WuHuijun, GuoChaoyu, SuChengguo,et al.Combined prediction method for short-term wind power based on EEMD‒GRU‒MC[J].Southern Power System Technology,2023,17(2):66‒73.
XuWu, LiuYang, ShenZhifang,et al.Short-term wind pow-er prediction based on VMD‒GRU model with improved local self-attention mechanism[J].Power System and Cl-ean Energy,2023,39(3):83‒92.
DingTong, FuXiaojin, LiuMingwang.Research on wind power predition based on GA‒VMD‒BiLSTM algorithm[J].Journal of Yangzhou University(Natural Science Edition),2022,25(4):44‒49.
WangXiaodong, LiShanshan, LiuYingming,et al.Ultra-short-term wind power prediction based on variable feature weight[J].Acta Energiae Solaris Sinica,2023,44(2):52‒58.
HeKunmin, WangXiao, YangJing,et al.Fault diagnosis of wind turbine gearbox based on RF feature optimization and WOA‒ELM[J].Electronic Measurement Technology,2023,46(5):57‒64.
ZengLiang, LeiShumin, WangShanshan,et al.Ultra-short-term wind power prediction based on OVMD‒SSA‒DE-LM‒GM model[J].Power System Technology,2021,45(12):4701‒4710.
LuPeng, YeLin, PeiMing,et al.Short-term wind power forecasting based on meteorological feature extraction and optimization strategy[J].Renewable Energy,2022,184:642‒661. doi:10.1016/j.renene.2021.11.072
WangRui, XuXinchao, LuJing.Short-term wind power prediction based on SSA optimized variational mode decomposition and hybrid kernel extreme learning machine[J].Information and Control,2023,52(4):444‒454.]
LiLei, LinShan, JiaJiehui.Short-term load forecasting based on TCN‒attention neural network[J].Electric Po-wer Information and Communication Technology,2023,21(3):10‒16.
ZhaoXingyu, WuQuanjun, ZhuWei.Short-term power loa-d forecasting based on CEEMDAN and TCN‒LSTM mo-del[J].Science Technology and Engineering,2023,23(4):1557‒1564.
ZhouJianguo, WeiSiti.Hybrid carbon price prediction mo-del based on time convolution neural network and double-scale feature selection[J].Electric Power Science and Engineering,2023,39(4):41‒49.
SuLiancheng, ZhuJiaojiao, LiYingwei.Short-term wind power prediction based on temporal convolutional network residual correction model[J].Acta Energiae Solaris Sinica,2023,44(7):427‒435.
ZhuRuijin, LiaoWenlong, WangYusen.Short-term prediction for wind power based on temporal convolutional network[J].Energy Reports,2020,6:424‒429. doi:10.1016/j.egyr.2020.11.219
[46]
LiangLu, LiuYuanlong, LiuShaohua,et al.Research on short-term load forecasting of power system based on E-CA‒TCN[J].Proceedings of the CSU‒EPSA,2022,34(11):52‒57.
WangChenyang, DuanQianqian, ZhouKai,et al.A hybrid model for photovoltaic power prediction of both convolutional and long short-term memory neural networks optimized by genetic algorithm[J].Acta Physica Sinica,2020,69(10):143‒149. doi:10.7498/aps.69.202191935
LiuHui, LingNingqing, LuoZhiqiang,et al.Power grid short-term load forecasting method based on TCN‒LSTM and meteorological similar day sets[J].Smart Power,2022,50(8):30‒37.
YangFei, LiuYang, ChangSuoliang,et al.Prestack seismic porosity prediction method based on bidirectional GRU and attention mechanism[J].Geophysical Prospecting for Petroleum,2024,63(3):598‒609.
JiaTaorong, YaoLixiao, YangGuoqing,et al.A short-term power load forecasting method of based on the CEEM-DAN‒MVO‒GRU[J].Sustainability,2022,14(24):16460. doi:10.3390/su142416460
[55]
ZengLiang, DiFeichao, LanXin,et al.Short-term wind po-wer prediction based on CEEMD‒CNN‒BiGRU‒RF model[J].Renewable Energy Resources,2022,40(2):190‒195.
WuJunyue, ZhaoErgang, GuoZengliang,et al.Ultra-short-term photovoltaic power multi-step prediction based on spearman coefficient and TCN[J].Acta Energiae Solaris Sinica,2023,44(9):180‒186.