A Differential Evolution Approach for the Reliability Constrained Unit Commitment Problem

In this paper, an efficient approach is proposed to optimize the Unit Commitment Problem (UCP) considering the unreliability of the generating units and the load forecast uncertainty. Reliability indices such as the Loss of Load Probability (LOLP) and the Expected Energy Not Served (EENS) are included in the formulation of the UCP to implicitly assess the required spinning reserve of the system. The method is based on the Differential Evolution (DE) algorithm combined with a hereby proposed series of problem specific repair mechanisms, which enhance the algorithm's performance. The approach is tested on the IEEE Reliability Test System (IEEE RTS), which comprises 26 thermal units. The impact of the units' unreliability and of the load forecast uncertainty on the required reserve and on the total operation cost is evaluated. A benchmarking against previously proposed algorithms reveals that the proposed method provides consistently solutions of lower cost in competitive time. Moreover, the algorithm is applied on systems of larger size, demonstrating an efficient and robust performance.