Heuristic algorithm for the piecewise linear segmentation of multiple time-series for solar thermal systems inverse modelling

Abstract

This paper presents a novel algorithm for the piecewise linear segmentation of multivariate time-series and proposes its application to the analysis of hot water thermal solar systems (TSS). The ISO 9459-5:2007 norm describes a non-intrusive dynamic test for the performance assessment of TSS. This allows to characterize the system heat losses and the thermal stratification properties, as well as to predict its long-term performance. The application of this norm requires an inverse modeling approach where the parameters of a simplified plug flow storage model, based on simulation runs, are determined through an optimization procedure aiming at the adjustment of the predicted results to those obtained by a predefined experimental test sequence (3-5 days). This paper proposes a new method to decrease the computation time required for the model simulation, which is based on the segmentation of the multivariate time-series into a piecewise linear approximation, where the number of segments is critically selected. An illustrative example is presented consisting in the simulation of a real 3-day experimental dataset with 26873 points and a 15 s sampling rate.