中文摘要：

大电网可靠性评估中，现有负荷模型大多采用参数密度估计方法，即凭经验事先假定负荷的随机变化服从某种已知分布（例如均值和方差未知的正态分布），并用历史负荷数据对密度函数进行参数估计。深入分析参数密度估计方法的不足，提出采用非参数多变量核密度估计实现节点负荷联合概率密度的近似估计，所提方法完全基于数据驱动，不需对节点负荷的概率分布和相依形式进行任何主观假设，并可有效揭示隐藏在节点历史负荷数据中的结构信息（不确定性和相关性）。通过对改进的RBTS可靠性测试系统的计算分析，验证了该方法的有效性及其应用价值。

英文摘要：

In traditional bulk power system reliability evaluation, the load models are usually based upon parametric density estimation, which is a rule of thumb with the assumption that the form of underlying density is known （e.g. normal distribution with unknown expectation and variance） and carries out parameter estimation for the probability density function （PDF） using historical load data. This paper firstly analyzed the drawbacks of the parametric density estimation, and then presented the nonparametric multivariate kernel density estimation approach to approximately calculate the joint probability density distribution of bus loads. The presented method is a kind of data-driven approach without any assumption for underlying form and dependence of the bus load probability distribution, and is capable of uncovering the structure information （uncertainty and correlation） hidden in the historical load data. The reliability assessment results from the improved RBTS reliability test system verify the validity and applicability of the proposed method.