Long- and Short-term Photovoltaic Power Prediction Based on the Combination of Multi-objective Improved Multivariate Universe Algorithm and Multi-layer CNN-LSTM-Attention Models
The intermittent, variable and stochastic nature of photovoltaic power generation poses a major challenge to grid integration. To address this problem, a combined prediction model CNN-LSTM-Attention (CLA) combining convolutional neural network (CNN), long short-term memory (LSTM) and Attention mechanism was proposed, which can effectively capture the potential patterns in time series. An improved multivariate universe optimization algorithm was used to optimize the LSTM parameters to enhance the predictive ability and robustness of the model. It was worth mentioning that the model design did not need to rely on complex data preprocessing steps, demonstrating good versatility and practicality. The introduced Attention mechanism empowered the model to adaptively focus on key historical information, thus significantly improving the prediction accuracy. The experimental results show that the model outperforms other models in terms of prediction accuracy and stability, with a coefficient of determination (R2) of 0.978, which exceeds the traditional models of 0.032 and 0.039, and when compared with the recently developed optimization models, it is 0.028 and 0.023 higher than the sparrow search algorithm-CNN-LSTM(SSA-CL) and dung beetle optimizer-CNN-LSTM(DBO-CL)models, respectively, and the error is significantly smaller than that of the other models. The application potential of this method in real PV power forecasting is verified, providing a strong technical support for future grid scheduling and operation.
DANG Hao-yan, WU Zhen-kui, SUN Yang, ZHANG Ji-hong, ZHANG Xin.
Long- and Short-term Photovoltaic Power Prediction Based on the Combination of Multi-objective Improved Multivariate Universe Algorithm and Multi-layer CNN-LSTM-Attention Models[J].
Science Technology and Engineering, 2026, 26(24): 10424-10434 DOI:10.12404/j.issn.1671-1815.2505849
目标函数均方误差(mean squared error,MSE)在式(23)中详细描述,计算预测的光伏功率输出和模型输出的实际值之间的平方差的平均值。目标函数的值越低越理想,而最小化均方误差可以提高预测精度和模型拟合能力。对于模型性能评估,采用三个参数:均方根误差(root mean squared error,RMSE)和平均绝对误差(root mean squared error,MAE),这两个参数均用作预测偏倚指标,量化实际值和预测值之间相关性的R2如式(26)。在回归分析中,R2接近1表示自变量的解释能力强于因变量。
ZangPengfa, WangKeqi, ZhangZhongwei, et al. Research on ultra-short-term photovoltaic power generation forecasting based on multi-feature extraction[J]. Laser & Optoelectronics Progress, 2025, 62(5): 333-343.
DaiZhaohui, ChenHao, LiuShenyi, et al. Short-term photovoltaic power forecasting based on the hybrid model of K-medoids-GBDT-PSO-LSTM[J]. Acta Energiae Solaris Sinica, 46(1): 654-661.
LiuShipeng, NingDejun, MaJue. LSTformer model for photovoltaic power generation forecasting[J]. Computer Engineering and Applications, 2024, 60(9): 317-325.
LiGuomin, ZhangJun, LiGengyin, et al. Wind and solar power generation forecasting based on dual-channel temporal convolution and residual attention mechanism[J/OL]. (2025-05-14)[2026-05-15]. https://kns.cnki.net/kcms/detail/10.1999.TM.20250514.1014.002.html.
BaiJiaqing, ZhangDong, LiuQuan, et al. Short-term photovoltaic power forecasting based on similar days and EM-ELM[J]. Journal of Shenyang Normal University (Natural Science Edition), 2024, 42(4): 364-371.
SongYu, XuYe, LiuFengping, et al. Research on a solar power forecasting model based on composite similar day selection using SO-CNN-LSTM[J]. Acta Energiae Solaris Sinica, 2025, 46(4): 301-312.
[18]
MohamadA H, SyedM S B, MuhammadH Z, et al. COA-CNN-LSTM: coati optimization algorithm-based hybrid deep learning model for PV/wind power forecasting in smart grid applications[J]. Applied Energy, 2023, 349: 121638.
[19]
FanY, MaZ, TangW, et al. Using crested porcupine optimizer algorithm and CNN-LSTM-Attention model combined with deeplearning methods to enhance short-term power forecasting in PV generation[J]. Energies, 2024, 17: 3435.
[20]
LiuL, LiY. Research on a photovoltaic power prediction model based on an IAO-LSTM optimization algorithm[J]. Processes, 2023, 11: 1957.
[21]
IwehC D, AkupanE R. Control and optimization of a hybrid solar PV—hydropower system for off-grid applications using particle swarm optimization (PSO) and differential evolution (DE)[J]. Energy Reports, 2023, 10: 4253-4270.
ZhuBing, XiaTian. Multiverse optimization estimation of SOC and SOH for lithium-ion batteries[J]. Battery Bimonthly, 2024, 54(5): 688-692.
[24]
Al-Ja’afrehM A A, MokryaniG, AmjadB. An enhanced CNN-LSTM based multi-stage framework for PV and load short-term forecasting: DSO scenarios[J]. Energy Reports, 2023, 10: 1387-1408.
[25]
GaoB, HuangX, ShiJ, et al. Hourly forecasting of solar irradiance based on CEEMDAN and multi-strategy CNN-LSTM neural networks[J]. Renew Energy, 2020, 162: 1665-1683.
TangXiaole, KangYanting, LuHao. BiLSTM photovoltaic power forecasting based on double decomposition and improved multi-objective raccoon optimization algorithm[J]. Acta Energiae Solaris Sinica, 2026, 47(3): 635-643.
WenShuang, MaYicheng, SunZhiqiang. Short-term photovoltaic power generation forecasting based on GWO-EEMD-BP neural network[J]. Journal of Central South University(Science and Technology), 2022, 53(12): 4799-4808.
LiZheng, ZhangJie, XuRuosi, et al. Short-term photovoltaic power forecasting based on similar day clustering and PCC-VMD-SSA-KELM model[J]. Acta Energiae Solaris Sinica, 2024, 45(2): 460-468.
FengPeiru, JiangGuifen, XuJiayin, et al. Short-term photovoltaic power forecasting based on CEEMD-BiLSTM-RFR[J]. Science Technology and Engineering, 2024, 24(5): 1955-1962.