Efficient SPF approach based on regression and correction models for active distribution systems

Abstract

This study proposes efficient methods for sequential power flow (SPF) analysis of distribution systems with intermittent photovoltaic (PV) units and fluctuated loads. The proposed methods are based on machine learning techniques; more specifically, they use a regression trees (RTs) algorithm to construct a model for voltage estimation. This model is trained using synthetic data generated by a number of PV generation and load demand scenarios. The SPF methods that utilise iterative techniques have a high computational burden. In turn, the proposed method, which is called SPF-RT, is fast and accurate. Furthermore, the authors combine SPF-RT with a correction method to develop a new method, called SPF-RTC, which significantly reduces the estimation error of the RT model. The proposed methods are tested using a 33-bus distribution test system interconnected with two PV units. To assess the performance of the proposed methods, they conducted several experiments at different resolutions of day/year data. The proposed methods are compared with the iterative SPF methods and validated using the OpenDSS software. The simulation results demonstrate that the proposed methods outperform the other methods in terms of the computational speed. The SPF-RT and SPF-RTC methods are useful for real-time assessment of distribution systems with PV units.