一种三段式序贯交叉熵重采样方法及其在电力系统短期可靠性评估中的应用

A Three-stage Sequential Cross-entropy Importance Sampling Method and Its Application in Composite Power System Short-term Reliability Evaluation

在短期可靠性的评估中,通常需要预测系统在未来一段相对较短的时间内,极端事件发生的概率或期望损失。由于短期评估时间跨度较短,引起系统失效的关键状态发生概率通常较低,导致原始序贯蒙特卡洛方法对系统状态进行评估的效率较低。首先以最小化Kullback-Leibler距离为目标推导了序贯仿真机制下元件重采样偏移转移率的解析表达式,基于交叉熵思想和重采样技术提出一种三段式序贯交叉熵重采样评估方法,解决原始状态转移率相对较小时采样效率较低的问题。通过对两个典型系统平均不可用率的仿真验证了算法的无偏性和高效性。最后,通过对IEEE-RTS79系统短期可靠性评估说明了算法可极大地提高短期可靠性评估效率,有一定的实用价值。

In the composite power system short-term reliability evaluation, it is a common practice to extrapolate the probability or expected loss of extreme hazard event during a relative short interval. Due to the short interval considered, the probability of critical event resulting in system failure is usually small, as a result, the crude sequential Monte Carlo method is lack of efficiency. In this paper, the analytical solution of importance sampling transition rate associated with sequential mechanism was found by minimizing the Kullback-Leibler distance. Then, a new three-stage sequential cross-entropy importance sampling evaluation method based on cross-entropy idea and importance sampling technique was proposed for the purpose of dealing with the low efficiency problem when the original transition rate was relative small. The proposed method was proven to be unbiased and of high efficiency through the simulation of average unavailability executed on two typical cases. Finally, the proposed method was applied to composite power system short-term reliability evaluation which was tested on the IEEE-RTS79. It is suggested that the proposed algorithm can tremendously improve the short-term evaluation efficiency and is of practical use.