针对光伏出力的间歇性、随机性且用传统方法难以准确预测的问题,提出一种基于花斑翠鸟优化算法优化核极限学习机(Kernel Based Extreme Learning Machine,KELM)的光伏短期功率预测模型。首先采用皮尔逊相关系数选取影响光伏发电功率的主要特征因素并基于相似日理论对天气类型进行聚类；接着,采用变分模态分解法对分类后的原始数据进行分解为多个子序列,从而减少随机波动分量对数据的影响；然后利用斑翠鸟在觅食过程中的自然行为来进行优化KELM模型的核函数参数和正则化系数；最后,利用KELM模型建立历史数据之间的时间特征关系,得到光伏发电功率预测结果。结果表明,所提方法能够达到0.99的决定系数,较其他算法有更好的拟合效果；模型的预测时长仅1mim,预测效率有很大提升。

Aiming at the intermittent and stochastic nature of PV power and the difficulty of accurate prediction by traditional methods, a PV short-term power prediction model based on Pied Kingfisher Optimizer (PKO) Kernel Based Extreme Learning Machine (KELM) is proposed. model. Firstly, the Pearson correlation coefficient is used to select the main characteristic factors affecting the PV power and the weather types are clustered based on the similar day theory; then, the Variational Mode Decomposition (VMD) is used to decompose the classified raw data into multiple sub-sequences, so as to reduce the influence of random fluctuation components on the data; and then, the Pied Kingfisher Optimizer (PKO) is used to optimize the KELM short-term power prediction model. The natural behavior of the spotted kingfisher in the foraging process is used to optimize the kernel function parameters and regularization coefficients of the KELM model; finally, the KELM model is used to establish the temporal characteristic relationship between the historical data to obtain the prediction results of photovoltaic power generation. The results show that the proposed method can achieve a coefficient of determination of 0.99, which is a better fitting effect than other algorithms; the training time of the model is only 1mim, and the prediction efficiency is greatly improved.

内蒙古自治区重点研发与成果转化项目(2022YFHH0019)；内蒙古自治区新型重要能源综合利用技术集成攻关大平台（2023PTXM001）