Multidimensional Optimization Theory And Application For Hydropower System Under Uncertain Enviroment

There are many multidimensional optimal problems with uncertainty in forecasting and operation of hydropower systems. To develop practical methods to solve these problems is very important to economical and effective operation of hydropower. Hydropower operation and hydrologic model calibration are two kinds of representative optimization problems. In this paper, started with these problems, the multidimensional optimization problems impact by uncaetain factors are studied. The main contents and research progress are as follows:(1) Towards the problem that single objective calibration can not reflect all differences between simulated and observed data, Multi-objective Shuffled Complex Evolution (SCE-UA) Algorithm and Genetic Algorithms (Gas) are presented and applied to continuous series simulation and flood events simulation. In continuous series simulation, some statistical values are taken as objectives. In flood events simulation, the qualificatory ratios of total runoff volume, peak time and peak discharge are taken as odjectives. The results show that multi-objective SCE-UA and Gas can deal with the uncertainty of calibration objectives and show high efficiency and stability in calibration.(2) Hydrologic model calibration is a hard computation task that costs lots of time. On consideration of the inartificial parallelism of arithmetics, paralleled GAs algorithem (PGA) and paralleled SCE-UA algorithem (PSCE-UA) for the calibration of Xinanjing model are presented and compared. The key idea of the two algorithems is the parallelism of Gas and PSCE-UA to increase calibration efficiency. The algorithems are written in Java and use JPVM library in communication, executed in a cluster of PCs. The results show that paralleled arithmetics can increase the efficiency and quality of calibration. And the PGA with multi populations can much more increase the efficiency and alleviate the problem of premature convergence.(3) A Generalised Likelihood Uncertanty Estimation (GLUE) and fuzzy optimization based uncertainty analysis method is presented for the dual uncertain problem of hydrlogica models and evaluation criterias. In this method,the basic theory of GLUE method is applied to Xinanjiang model uncertainy analysis, taking the qualificatory ratios of total runoff volume, peak discharge and peak time as criterias, using fuzzy membership as likelihood measure. The method is applied to Shuangpai reservoir, and compared to those of continues simulation. Results show that this method can reveal the uncertainy and sensitivity of model parameters,...