Co-Evolutionary Algorithm Based on Lagrangian Method for Hydrothermal Generation Scheduling

• Main Authors: Ming-Huei Ke, 柯鳴輝
• Other Authors: Ruey-Hsun Liang
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/93080292719922847798

碩士 === 國立雲林科技大學 === 電機工程系碩士班 === 90 === In this thesis, the co-evolutionary algorithm (CEA) based on Lagrangian method is proposed for three optimization problems with economic dispatch, unit commitment, and hydrothermal generation scheduling. At first, the main purpose of economic dispatch is to minimize the production cost and the constraints to be satisfied by dispatching the power outputs of the generating units, based on the curves of the generating cost, under a given electrical load. In addition, units with prohibited operating zone and line losses are considered for being similar practical power system. The next, the main purpose of unit commitment is to minimize the overall production cost and the constraints to be satisfied by finding a set of feasible on/off status combination of thermal units under the study periods and a given electrical load. Finally, the main purpose of hydrothermal generation scheduling is to minimize the overall production cost and the constraints to be satisfied by scheduling the power outputs of all hydro and thermal units under the study periods, a given electrical load, and limited water resource. In the algorithm, genetic algorithms are successfully put into the Lagrangian method. The proposed algorithm has two procedures. At first, Lagrangian function is formed from primal problem by Lagrangian method, and then defining as dual function. Secondary, the co-evolutionary algorithm employs two genetic algorithms to parallelly evolve control variables and Lagrange multipliers of dual function. The best advantage of genetic algorithm is to easily perform mathematical calculations. Regardless of the characteristic of the objective function, genetic algorithm does not have to modify the designing rules, and it possesses the ability of going over local solutions toward global optimal solution. Therefore, genetic algorithm can improve the disadvantage of traditional Lagrangian method, which updates Lagrange multipliers according to the degree of violating system constraint by gradient algorithm, and further searching out global optimal solution. Finally, practical applications of the proposed algorithm have been verified on the Taiwan power system for the effectiveness on economic dispatch, unit commitment, and hydrothermal generation scheduling. Numerical results show that the proposed co-evolutionary algorithm based on Lagrangian method is a very effective method for searching out global optimal solution.