Estimation of PV Power Generation by Linear Regression Model Using Voltage and Current Data

Abstract

PV systems have the disadvantages of large fluctuations in power and of not being controllable due to external factors. In addition, small-scale PV plants rarely receive maintenance after installation. Managers thus need a monitoring system that predicts the power of the PV plant in order to maintain performance and facilitate O&M. Recently, methods using big data to predict PV plant power have been applied. In this paper, power was predicted through learning based on PV plant field data. Furthermore, the error of the estimated power was analyzed through accuracy evaluations, RMSE, and R analysis. As the learning method, linear regression analysis was applied among machine learning models. Existing linear regression models can immediately estimate power by learning irradiation data as input variables and power data as output variables. However, if the PV system malfunctions, the accuracy of the estimated power generation decreases. In this paper, in order to address this problem, power was estimated by learning irradiation data as input variables and voltage and current data as output variables rather than directly estimating the power. As a result, the RMSE of the proposed linear regression equation was 15.9235kw, yielding a better power estimate than the existing method (16.4241kw).