About data reduction techniques and the role of outliers for complex energy systems

Abstract

Abstract Optimal design and scheduling of energy systems with a high share of renewables is a complex and computationally demanding task. The mismatch of supply and demand of energy requires the consideration of timeseries with a granularity of a few minutes, which is in contrast to the lifetime of the system of multiple decades. This paper proposes an algorithm for systematically reducing the input data and computational effort in mixed integer linear programming (MILP) of energy systems. Unlike the state- of- the-art, the influence of different numbers of typical periods is not examined on the on the quality of the clustering algorithm but on the objective function and the integer decisions. The issue is addressed by exploiting the two-stage nature of the optimal design and planning of the system by sequentially performing k-medoids clustering. The demonstration of the proposed algorithm shows that very few typical periods are sufficient to achieve near optimal decisions. The proposed approach is outperforming algorithms for time series aggregation (TSA) in this field by reducing CPU time by more than 40 %. The inclusion of the integer decision in the algorithm allows the application to multi objective optimization (MOO). The case study demonstrates that the runtime of the MOO can be reduced by approximately 90 %, while diverting less than 2 % on Pareto optimal solutions. Outliers have no impact on the techno-economic analysis but may lead to significant electricity peaks in energy systems with a high share of renewables.