A New Approach for Selecting Representative Patterns for Investment Planning in Local Energy Communities

Emerging investment strategies in assets for Renewable Energy Communities give rise to new formulations of the Generation Expansion Planning problem, accounting for local market and pricing mechanisms. The investment horizon spans over decades, with quarter hourly electricity consumption and production patterns, increasing the computational burden of the underlying optimization problems. In this work, we propose a new ex-ante procedure to compute a limited number of representative patterns for the optimal sizing model. We compare our approach with manual and automatic segmentation techniques relying on standard clustering validation indices, as well as a benchmark operating on the whole initial dataset. Results are demonstrated on a real 10 -members industrial community in Belgium. We show that our method limits the computational burden while maintaining accuracy of investment decisions.