Research On Optimal Scheduling Of Cascaded Reservoirs Based On Estimation Of Distribution Algorithm

The scheduling of cascaded reservoirs has been one of the hotspot academic and engineering researches. With China’s hydropower energy development process to speed up, the large-scale cascade hydropower stations are completed and put into operation. The hydraulic and electrical relationship among cascade reservoirs become more complex, with the increasing demand for co-ordination between different power source, which bring great challenges to the management of the cascaded reservoirs. The cascaded reservoirs scheduling considers runoff process, water supply and power generation characteristics, is a kind of high-dimensional, non-convex, nonlinear and complex constrained optimization problem. How to solve this kind of problem is a key issue of the cascade reservoirs scheduling. With the expansion of the scale of the cascaded reservoirs, the constraints of the optimization problem become more complicated, the traditional optimization methods can not meet the needs of cascaded reservoirs scheduling development. Therefore, many scholars utilize intelligent optimization algorithms to solve the optimal scheduling of cascaded reservoirs problems. However, intelligent optimization algorithms are easy trapped into local optimum and lack of constraint handling mechanism, which lead algorithm cannot obtain the global optimal solution when solve cascaded reservoirs optimal problems. A more effective optimization method and constraint handling strategy are still needed to deeply research for solving cascaded reservoirs optimal problems. In this thesis, combined with intelligent optimization algorithm theory, the theory of systems engineering optimization theory and other related theories, a global search algorithm named estimation of distribution algorithm is utilized to solve cascaded reservoirs optimal problems and hydrothermal optimal problems. Some conclusions with theoretical and practical value are obtained. The main achievements are as follows:(1) Estimation of Distribution Algorithms in a new evolutionary algorithm has better global search capability. The research based on analysis the performance of the algorithm and the effect of control parameters in the evolutionary process. In view of the critical shortcomings such as the scope of the search down too fast which lead to reduced diversity of the population and caused local convergence, a hybrid Univariate Marginal Distribution Algorithm (HUMDA) was proposed. In the proposed method, a two-stage dynamic parameters control strategy is used to control the mean and variance parameters in order to preserve the diversity of the population at the beginning of algorithm and improve the local search capability of the algorithm at the end of the execution. In addition, the chaotic search strategy is adopted to enhance the precision of solution and search efficiency. The HUMDA is tested by standard test functions by analyzing the sensitivity of the parameters and the performance of the local search strategy. The results show that, the algorithm has better global convergence ability and search accuracy. Furthermore, on basis of HUMDA algorithm, a cultured estimation of distribution algorithm (CEDA) which combines cultural framework control parameters strategy is proposed to improve the robustness of the algorithm. The CEDA has better performance use several standard function tests the CEDA, the results show that CEDA has better performance and optimization ability. Research on the improvement of the estimation of distribution algorithm, effectively improve the search performance of the algorithm and provide an effective means for solving cascaded reservoirs optimal problems.(2) By analyzing the characteristics of the solution space in reservoir optimal scheduling model, the proposed HUMDA is applied to solve reservoir optimal scheduling problems. Considering that local search may violate the constraints which lead to reduce the efficiency of search. A chaotic local search strategy combines with POA algorithm is proposed to improve the efficiency of local search. Applying the strategy to solve Long-term reservoir scheduling problem, the results show that the algorithm has better performance in solving reservoir scheduling problem. This research provides a reference for study cascaded reservoirs optimal scheduling with more complex constraints.(3) In order to reduce the affect the performance of the algorithm caused by constraint handling strategy which producing a large randomness variable in solving cascaded reservoirs optimal scheduling problem. An equality constraints handling strategy which considers the hydraulic connection between reservoirs is proposed to combine with CEDA to improve the efficiency of the algorithm to solve cascaded reservoirs optimal scheduling problem. Applied to the reservoir system which contains10reservoirs, the results show that the use proposed algorithm can obtain better solution.(4) By analyzing the characteristics of the solution space in the hydrothermal optimal scheduling model and the advantages and disadvantages of the existing constraint handling strategy, a combination of two-stage constraint handling strategy and CEDA is proposed. Using CEDA to solve dynamic economic dispatch in thermal power generation has better performance. Utilizing CEDA combined constraint handling strategy in thermal and hydropower to solve hydrothermal optimization problems. The results show that the CEDA combined with proposed constraint handling strategy has better performance for solving hydrothermal optimization problems. The research provides a new effective method for solving hydrothermal optimization problems.