In recent years, power system uncertainties have increased due to the growing integrations of intermittent renewable energy resources. It is imperative to introduce probabilistic load flow analysis in the study of power system operation and planning to adapt to the ever-increasing uncertainties. This paper proposes a scenariobased analytical method for the probabilistic load flow analysis, which takes advantage of both the scenario analysis method and the cumulant method. This method can not only consider various kinds of correlations among power inputs but also accurately represent the probability distributions of desired outputs with a reasonable computational burden. The performance of this method is evaluated on the IEEE 14-bus and 118-bus test systems. The accuracy and efficiency of the proposed method are validated through quantitative and graphical comparisons with Monte-Carlo simulation.

Introduction

In the past few years, renewable energy sources (RES) have experienced rapid development due to their numerous advantages. More and more uncertainties have been penetrating into the modern power systems, not only from load demands, network topology changes, outages of system components but also from the generations of RES, such as solar and wind power. Besides, due to complex meteorological processes, there are significant spatiotemporal correlations among the RES generation. Hence, assessing the behaviors of power systems with complex uncertainties becomes indispensable. Probabilistic load flow (PLF), firstly proposed in 1974 [1], has become the commonly used tool to analyze the influence of power system uncertainties. There are three mainstream PLF methods: numerical methods, analytical methods, and approximate methods [2]. As the most straightforward numerical method, Monte-Carlo simulation (MCS) firstly represents the uncertainties of input random variables (RVs) with a series of samples and then obtains the probability distributions of output RVs through a large number of deterministic power flow (DLF) calculations. The traditional MCS method with simple random sampling (MCS-SRS) [3] usually requires 104 –۱۰۶ trials to harvest accurate results. The massive computational burden hinders its applications in large-scale power systems. Hence, serval advanced sampling techniques, such as Latin supercube sampling [4], uniform design sampling [5], and Latin hypercube sampling (LHS) [6,7] are introduced to improve the computational efficiency. Besides, combined MCS and parallel computing [8] provides a promising approach for online PLF analysis. It achieves high accuracy at a low computational burden.