Evaluating the Effectiveness of Conservation Voltage Reduction with Multilevel Robust Regression

Conservation voltage reduction (CVR) can effectively reduce electricity consumption and peak demand by keeping the customer voltages in the lower half of the permissible range. To facilitate widespread adoption of CVR, a reliable and robust CVR performance evaluation methodology is in critical need. However, it is difficult to accurately estimate the load reduction impact of CVR in practice. The data quality issues in supervisory control and data acquisition and advanced metering infrastructure make it challenging to distinguish a few percentage of load reduction from measurement errors and bad data. This paper develops a multilevel robust regression model within the framework of statistical experimental design to address the data quality issues. The proposed model is capable of providing robust and reliable estimates of load and voltage reduction from CVR at both distribution feeder and substation levels. The effectiveness of the proposed methodology is validated with field CVR demonstration data provided by a major California electric utility.