Restoration of Short Periods of Missing Energy Use and Weather Data Using Cubic Spline and Fourier Series Approaches: Qualitative Analysis

Abstract

The paper presents seventeen approaches that use cubic splines and Fourier series for restoring short term missing data in time series of building energy use and weather data. The study is based on twenty samples of hourly data, each at least one year long. In order to differentiate the approaches, two comparisons were carried out. The first comparison was made between the estimated and actual values, as time series and as cross check plots. The second comparison is based on the fraction of the total data that can be estimated by an approach within specific ranges or error. Thus for the cooling and heating data, the fraction of the data set estimated within 1%, 5%, and 10% of the measured values was determined. For the dew point and the dry-bulb temperature samples, the performance is based on the amount of data that are within 1, 3, 5 and 10 °F from the actual data. From the results of this analysis, it appears that linear interpolation is a better approach for filling gaps one to three hours long. The cubic splines approach gave better performance for gaps between four and six.