Improvement of representative days selection in power system planning by incorporating the extreme days of the net load to take account of the variability and intermittency of renewable resources

•A way to include the extreme days of the net load in representative days.•The suitable number of representative days is found by two simple methods.•Dispatchable capacity is underestimated if the extreme days are not included.•The importance of assigning proper weights to the extreme days.•Several years of the net load data are used to select representative days. As the share of variable renewable energy in power systems increases, incorporating the operational constraints in generation expansion planning has become necessary to address flexibility issues. Accordingly, to preserve chronology and to make the technically detailed planning model tractable, the model should be solved for some representative periods. Different approaches have been introduced to select representative periods; however, most of them ignore the importance of the net load. In this research, a method is proposed to include the extreme days with higher and lower levels of the net load in representative days by using self-organizing map clustering. Further, the impacts of the extreme days, with different weighting approaches, on planning and operational decisions are analyzed for single-year and multi-year planning. A simple method is also devised to find the proper number of representative days. These methods are applied to the NREL-118 and the Belgian power systems by deploying a planning model based on the unit commitment problem. It was shown that investment costs and the required dispatchable capacity are greatly underestimated if the extreme days of the net load are not considered. Besides, using self-organizing map to weight the extreme days according to their importance in the historical data resulted in 45% improvement in approximating operating costs, a better estimation of operating variables, and a more reliable expansion for multi-year planning, demonstrating the significance of assigning proper weights to extreme days. The method for the number of representative days also performed well, especially when the extreme days were included.