Research On Design And Application Of Optimization Algorithms In Scheduling And Control Problems

In recent years, the research and application of optimization algorithms in scheduling and control problems has got particular concern for its great impact in productional, economic and social activities.This paper focuses on the design and application of different categories of optimization algorithms accordingly to different optimization problems in scheduling and control problems, where simulation and test has been carried out to verify the effectiveness of each algorithm:At first, to tackle the growing concerned sports scheduling problem, this paper proposed a hybrid local search approach based on the combination of Tabu Search and Variable Neighborhood Descent meta-heuristic with an effective and comprehensive constructive algorithm to quickly obtain initial solution at a very high quality. Solutions of good competitiveness are obtained for benchmark instances.Secondly, for the optimization problems in data driven control, this paper proposed a control algorithms based on convex combination and perturbation optimization for linear system. The algorithm consists of 3 parts: The first part constructs a control sequence for a area in state space by linear combination of a steady control sequence starting from a certain initial state point. The second part defines a control objective function, and then search for the optimal control sequence within the space by combining different control sequences on the control state space boundary. The third part tries to find gratitude descent direction for control sequence by perturbation optimization method. The effectiveness of this algorithm has been proven by simulation.Finally, to solve convex quadratic programming problem in predictive control, this paper designed and developed quadratic programming solver software according to current popular algorithms. The software is programmed in C++ with a unified interface and consists of active set, interior point and Wolfe simplex methods, which can solve large scale quadratic programming problems over 500 dimensions. By testing benchmark problems and comparing results to commercial software, this paper has shown that the solver can handle practical convex quadratic programming problems in predictive control in high accuracy and effiency.