Improving the accuracy of solar irradiance forecasting has become crucial since the use of solar energy power has become more accessible due to increased efficiency and decreased costs associated with its production. Data quality and availability are essential to producing accurate solar irradiance forecasts. In this article, we focus on the estimation of missing values in minutely, hourly, daily, and weekly solar irradiance series using an extensive number of imputation methods. We compare the accuracy of 36 imputation methods for solar irradiance series over a real dataset recorded in Australia under 16 experimental conditions. The experiments are run in a semi-Monte Carlo setting, in which missing values are randomly generated in the solar irradiance series. Our results identify the most reliable and robust approaches for the imputation of solar irradiance for each of the mentioned frequencies. While linear and Stineman interpolations and Kalman filtering with structural model and smoothing are found accurate for minutely and hourly series, weighted moving average gives the highly precise imputations for daily and weekly solar irradiance.