Analysis of conservation voltage reduction effects based on multistage SVR and stochastic process

Abstract

Summary form only given. This paper aims to develop a novel method to evaluate Conservation Voltage Reduction (CVR) effects. A multistage Support Vector Regression (MSVR)-based model is proposed to estimate the load without voltage reduction during the CVR period. The first stage is to select a set of load profiles that are close to the profile under estimation by a Euclidian distance-based index; the second stage is to train the SVR prediction model using the pre-selected profiles; the third stage is to re-select the estimated profiles to minimize the impacts of estimation errors on CVR factor calculation. Compared with previous efforts to analyze the CVR outcome, this MSVR-based technique does not depend on selections of control groups or assumptions of any linear relationship between the load and its impact factors. In order to deal with the variability of CVR performances, a stochastic framework is proposed to assist utilities in selecting target feeders. The proposed method has been applied to evaluate CVR effects of practical voltage reduction tests and shown to be accurate and effective.