An Enhanced Markov Chain Monte Carlo-Integrated Cross-Entropy Method with a Partially Collapsed Gibbs Sampler for Probabilistic Spinning Reserve Adequacy Evaluation of Generating Systems

作者：Wang, Yue^*

来源：Electric Power Components and Systems, 2017, 45(15): 1617-1628.

DOI：10.1080/15325008.2017.1404660

• spinning reserve
• cross-entropy
• Markov chain Monte Carlo
• partially collapsed Gibbs
• generating systems

Probabilistic adequacy evaluation of allocated spinning reserve is beneficial to economically regulating this foremost auxiliary service to counterbalance unforeseen generation-demand mismatches. As the time horizon for a probabilistic spinning reserve adequacy investigation task may vary from several to dozens of minutes, adaptive importance sampling methods, such as the classical cross-entropy method and its variants, are appealing instead of the classical non-sequential Monte Carlo to estimate desired reliability indices due to the rareness of demand-not-supplied contingencies. In this article, a new adaptive cross-entropy method is proposed, particularly, nesting a specially optimized partially collapsed Gibbs sampler to help in avoidance of locally trapped Markov chain samples which may be encountered by traditional cross-entropy methods. RTS-79 is utilized for illustrating the superiority of the proposed method, termed E-MICEM, against its parent method, i.e., the Markov chain Monte Carlo-integrated cross-entropy method. Two traditional indices including loss of load probability and expected demand not supplied are comparatively evaluated and the simulation results suggest that the E-MICEM is superior in the efficiency of estimating the two indices. Some advices are also given on the build-in-parameter regulation for the E-MICEM applicable to the systems of different dimensions.

• 出版日期2017
• 单位中国农业大学