Problem-specific genetic algorithm for power transmission system planning

A hypothesis is proposed that the only way for Turing computers to solve efficiently NP complete problems and NP hard problems is to use randomization techniques. By comparison of the popular randomization techniques, it is concluded that genetic algorithm may be the best choice until now. To overcome the disadvantages of conventional genetic algorithms, problem-specific genetic algorithm is suggested. Then, a problem-specific genetic algorithm that is developed specially for power transmission system planning problems is proposed. The algorithm searches global optimum from local optimums instead of from all feasible solutions, while the local optimums are found by a more efficient linear iterative minimum-cost-flow algorithm. Furthermore, the network flow model for power transmission system planning is improved so that the capacities and locations of transmission lines, substations and power plants can be optimized simultaneously. Results from a comparative study have proved the reasonableness and efficiency of the proposed model and algorithm. (C) 2002 Elsevier Science B.V. All rights reserved.