在预估光伏－制氢系统运行备用容量需求时，受光伏－制氢系统运行数据不确定性的影响，预估结果的误差偏大，为此，提出基于最近邻聚类的光伏－制氢系统运行备用容量需求预估模型。引入最近邻聚类中的不确定自然最近邻机制，以近邻数数量为基础，按稠密点、稀疏点、噪声点的分类标准，将数据集中的不确定光伏－制氢系统运行数据对象进行划分处理；在使用不确定自然邻域搜索算法获取到不确定的自然稳定状态输出结果后，根据特征值的差异性去除噪声点，借助不确定自然邻域密度因子对光伏－制氢系统运行数据进行聚类；在预估模型构建阶段，将径向对称的高斯径向基函数(radial basis function，RBF)作为核函数，并将所有的RBF输出结果映射到同一个空间中，得到光伏－制氢系统运行备用容量需求结果。测试结果表明，所提方法对最大备用容量需求预估结果的偏差始终稳定在250 MW以内，对最小备用容量需求预估结果的偏差始终稳定在150 MW以内，有效降低了能量管理的成本开销。

When estimating the spare capacity requirement for the operation of photovoltaic-hydrogen systems, the estimation error is relatively large due to the uncertainty of operation data of photovoltaic-hydrogen production system. Therefore, a prediction model based on nearest neighbor clustering is proposed for estimating the spare capacity requirement for the operation of photovoltaic-hydrogen systems. In this model, the uncertain natural nearest neighbor mechanism in the nearest neighbor clustering is introduced to classify the data points based on their density, sparsity, and noise. The data set is divided into different groups of uncertain photovoltaic-hydrogen production system operation data objects. After obtaining the uncertain natural stable state output results using the uncertain natural neighbor search algorithm, the noise points are removed based on the difference of eigenvalues. Then, the photovoltaic-hydrogen production system operation data is clustered using the uncertain natural neighbor density factor. In the construction phase of the estimation model, the radial symmetric Gaussian radial basis function (RBF) is used as the kernel function, and all RBF output results are mapped to the same space to obtain the photovoltaic-hydrogen production system operational reserve capacity requirement results. The testing results show that the proposed method has a maximum estimation error of less than 250 MW for the maximum reserve capacity requirement and a minimum estimation error of less than 150 MW for the minimum reserve capacity requirement, effectively reducing the energy management cost.