【摘要】 【目的】随着我国光伏发电占比不断提高，光伏发电受气象因素的影响较大，其输出功率因气象特征的复杂多变表现出强烈的间歇性和波动性，对未来光伏发电功率预测的准确度将直接影响电网的稳定安全运行。针对光伏发电功率的预测精度提升问题，提出了一种融合局部离群因子算法、遗传算法与高斯过程回归（GA-LOF-GPR）的预测模型。【方法】首先，本文挖掘发电功率与气象特征的关系，并采用特征权重K均值聚类对气象类型进行分类；其次，将局部离群因子算法与高斯过程回归模型结合，构建局部异常因子加权的高斯过程回归（LOFGPR）的预测模型；最后，通过运用遗传算法优化加权高斯过程回归模型的超参数。【结果】通过对2018年澳大利亚的光伏发电数据的仿真预测，验证了该预测模型的有效性和准确性。更多还原

【Abstract】 【Purposes】The proportion of photovoltaic(PV) power generation has been increasing in China. PV power generation is greatly affected by meteorological factors and its output power shows strong intermittency and volatility because of the complexity and variability of meteorological characteristics. Thus, the accuracy of future PV power prediction will directly affect the stable and safe operation of the power grid. 【Methods】For improving the prediction accuracy of PV power, in this paper, a prediction model that integrates the local outlier algorithm, genetic algorithm and Gaussian process regression(GA-LOF-GPR) was proposed. First, the relationships between power generation and meteorological features were explored, and feature weight K-means clustering was used to classify meteorological types. Second, the local outlier factor algorithm was combined with the Gaussian process regression model to construct the prediction model of local outlier factor-weighted Gaussian process regression(LOF-GPR). Finally, the hyper-parameters of the weighted Gaussian process regression model were optimized by applying the genetic algorithm. 【Results】The effectiveness and accuracy of the prediction model are verified by simulation prediction of the PV power generation data in Australia in 2018.更多还原