Evaluation of methods to select representative days for the optimization of polygeneration systems.

The optimization of polygeneration systems considering hourly periods throughout one year is a computationally demanding task, and, therefore, methods for the selection of representative days are employed to reproduce reasonably the entire year. However, the suitability of a method strongly depends on the variability of the time series involved in the system. This work compares the methods Averaging, k -Medoids and OPT for the selection of representative days by carrying out the optimization of grid-connected and standalone polygeneration systems for a building in two different locations. The suitability of the representative days obtained with each method were assessed regarding the optimization of the polygeneration systems. Sizing errors under 5% were achieved by using 14 representative days, and the computational time, with respect to the entire year data, was reduced from hours to a few seconds. The results demonstrated that the Averaging method is suitable when there is low variability in the time series data; but, when the time series presents high stochastic variability (e.g., consideration of wind energy), the OPT method presented better performance. Also, a new method has been developed for the selection of representative days by combining the k -Medoids and OPT methods, although its implementation requires additional computational effort. • Comparison of representative days' selection methods for optimization. • Pros and cons of averaging, k -Medoids and OPT methods are evaluated. • Guidelines for applying adequate representative days' selection method are provided. • A new method, combination of k -Medoids and OPT methods, is proposed. [ABSTRACT FROM AUTHOR]