Unit commitment by annealing-genetic algorithm

Abstract

This paper presents the application of the annealing-genetics (AG) algorithm to solve the unit commitment (UC) problem. By exploiting the similarity between a minimization process and the cooling of a molten metal, the simulated annealing (SA) guides the search point toward a global minimum with probability 1 using a devised move generation strategy and cooling schedule. But the SA usually takes much computation time in order to arrive at a near-global minimum. The genetic algorithm (GA) is a general-purpose optimization technique based on the principle of natural selection and natural genetics. The GA is a fast algorithm but usually has the inferior solution quality compared to the SA. The AG algorithm incorporates the genetic algorithm into the simulated annealing to improve the performance of the simulated annealing in solving combinatorial optimization problems such as the unit commitment problem. Numerical results on two cases including a realistic Taiwan power (Taipower) system and comparisons with results obtained using the AG, the SA, the GA, the dynamic programming (DP) and the Lagrangian relaxation (LR), show that the features of easy implementation, fast convergence, and highly near-optimal solution to the UC problem can be achieved by the AG.