Research On Stochastic Unit Commitment Problem Considering Uncertainty Of Electric Power Demand

Unit commitment (UC) problem is an important part of the economic dispatch in the electric power systems, and should be primarily sovled in the short-term generation scheduling to achieve remarkable economic benefits. Therefore, the UC problem is one of the most important issues in the electric power systems. Recent years, the stochastic unit commitment (SUC) problem considering uncertainties are receiving increasingly attention largely as a result of the rapid development of stochastic optimization theory in the mathematical field and the improvement of computer technology. So far, many mathematical models and solution methods have been proposed for SUC problem, among which the expected value mathematical model and Lagrangian relaxation (LR) algorithm based solution method are generally adopted. However, due to the presence of duality gap, the LR based solution method is hardly to obtain the optimal solution. Therefore, it is necessary to improve the existing solution methods or to develop new solution method.This thesis focuses on the mathematical model and the solution method of the SUC problem:Firstly, the research background is briefly introduced., and the overview of traditional unit commitment problem and the SUC problem is given. Secondly, the mathematical model of traditional unit commitment problem is established, and then solved by dynamic programming (DP) method. The test results show that the model and the solution method are effective. Thirdly, in order to consider the effects of uncertain electric power demand on unit commitment, the uncertainty of electric power demand is modeled using a set of scenarios, which are introduced by scenario analysis. A mathematical formulation of the expected value model of the SUC problem is established. Finally, a solution method based on improved genetic algorithm (GA) is proposed to sovle the SUC problem.The GA algorithm is improved by introducing a new method to generate the initial population, a new mutation operator, and a local search operator. The SUC problem is sovled using the improved GA, which can automatically satisfy the bundle constraints. Based on numerical examples, the optimal solution of the SUC problem is compared with the optimal solutions of the deterministic unit commitment problem under the perfect electric power demand information, and the maximum electric power demand scenario as well as the situation with consideration of spinning reserve. Test results show the feasibility of the mathematical model of the SUC problem and its improved GA-based solution method.