Methodology for generating high time resolution typical meteorological year data for accurate photovoltaic energy yield modelling

Abstract

Accurate energy yield prediction is of utmost importance for commercial scale photovoltaic systems. One key parameter crucial to the prediction accuracy is the quality of solar radiation data. Most energy yield prediction models rely on Typical Meteorological Year data with maximum temporal resolution of one hour. In this work we develop a methodology to generate Typical Meteorological Year data with much higher time resolution using gap filling methods that aim to maintain high-quality solar radiation data for photovoltaic yield modelling. We demonstrate our method using ground-based one-minute solar radiation measurements available in Australia and find that hourly averaging reduces the share of irradiance values exceeding 1000 W/m2 by 4.1% to 10.8% compared to our one-minute resolution dataset. Such high irradiance values usually result from the cloud enhancement effect, which is filtered out by averaging. We estimate that the hourly averaging can lead to an underestimation of inverter clipping losses by 0.4% to 2.2% and an overestimation of the performance ratio by on average 1.1% for a common DC-to-AC ratio of 1.2. One potential issue is the limited availability of high-resolution radiation data with broad geographic coverage. However, new satellite-based irradiance products and stochastic models can overcome this limitation.