Abstract
This study attempts to algebraically decompose the periodic and random components of monthly total water consumption data metered at desalination plants in Kuwait. The decomposition approach is based on filtering the data in the frequency domain by selecting a threshold amplitude value below which harmonics are assumed to be random. The significant harmonics are used in Fourier series to develop a periodic model, and the residuals obtained by subtracting the periodic model from the original water data are tested by an autocorrelation function. The results show that the amount of random noise in the water consumption data represents a small portion of the overall pattern. The data can thus be considered periodic in nature and people's water consumption behaviour can be interpreted based on the significant periodicities detected by amplitude spectrum plot. This procedure allows application of different modifications for the decomposed periodic and random components, yielding more natural synthetic time series data generation.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call