A New Sparsity Method to Large-scale UC Problem

Exploiting sparsity has long been a classic and practical approach to addressing complex problems. Inspired by the sparsity of the node admittance matrix and its associated methods, this paper explores the sparsity of the UC problem and proposes a new method to exploit this sparsity. Firstly, a procedure for calculating the sparse probability of unit commitment variables is devised. Secondly, this sparse probability is exploited both in the population initialization strategy and genetic operators. The experiments demonstrate that the proposed sparsity method exhibits competitive convergence rates in small-scale systems. Moreover, large-scale systems (up to 800-unit) are effectively solved, where traditional evolutionary algorithms (EA) are incapable. The benefit of considering sparsity is demonstrated.