Unit commitment using a stochastic extended neighbourhood search

A simulated annealing is combined with a tabu search, to develop a robust and powerful optimisation technique for solving the unit commitment problem. The problem is broken down into a combinatorial subproblem in unit status variables and a quadratic programming subproblem in unit power output variables. The combinatorial subproblem is solved using the proposed method. In the hybrid algorithm, which is referred to as a stochastic extended neighbourhood search, simulated annealing is used as the main stochastic algorithm, and a tabu search is used as an extended neighbourhood search, to locally improve the solution obtained by simulated annealing. The neighbourhood search uses local domain-knowledge, which results in rapid convergence of the simulated annealing algorithm. The results obtained for several example systems illustrate the potential of the hybrid approach.