【作者】 张铭；

【导师】 张勇传； 李承军；

【作者基本信息】 华中科技大学 ， 水利水电工程， 2008， 博士

【摘要】 在水电站水库调度过程中存在着许多不可预见因素,这些不可预见因素的存在必然会对水库的运行调度带来风险。主要包括:来水风险、决策放水风险、发电效益风险。而来水风险起决定性作用。风险决策分析的目的就是采用相关方法对产生风险的各种可能的误差进行分析和评估,找出风险效益匹配或风险偏好最合适的决策方案。在本论文中,主要针对水库时段来水风险及电站调度期发电效益风险进行研究。对水库调度期面临时段而言,调度风险主要来自于入库径流的不确定性。由于时段入库径流的不确定性,导致决策结果的不确定性。目前得到较多应用的水文预报模型,预报制作单位都把它们视作确定性的,作为它们输出的预报值一般以确定值的形式发布给用户。然而,水文过程的发生与发展取决于气象因素和地理因素,是一个复杂的动态过程;水文预报模型接受水文、气象等多种输入,运用概化的模型参数,只是客观水文过程的仿真。这些复杂的因素导致了水文预报(入库径流预报)必然存在不确定性。基于马尔可夫过程的水库发电优化调度,受到广泛研究的优化准则是收益期望值最优。然而,期望准则对风险是不敏感的,不适宜需要直接对风险有所反映或限制的优化问题。在水电站水库优化调度中,为保证一定水平的发电收益,调度策略在尽可能使年期望总发电量大的同时,还要求平时发电量低于某个给定值的概率不得大于某个预定值。仅限于以收益期望值为优化目标是不够的,有必要选择其他的优化目标来反映实际问题的风险及决策者的意愿或风险偏好。在阅读和研究大量水库调度领域文献的基础上,研究了中长期优化调度以及发电调度风险评估问题。论文主要内容及创新成果包括如下几部分:第一部分(第三章)介绍了一种水库中长期补偿优化调度方法。对其建模原理、特点及求解方法进行了详细评述,并对实例的优化结果进行了详细分析。第二部分(第四章)介绍了一种水库调度时段风险分析法。采用贝叶斯概率水文预报理论,将实时气象因子与历史水文数据序列相结合,综合利用各种信息以提高入库径流预报精度。采用因子灰关联预报模型对流域中长期降水进行预报,建立流域降水径流确定性预报模型和贝叶斯概率水文预报模型,并以概率分布形式定量地描述水文预报的不确定度。以隶属东北电网的丰满水库为例,对所建立的模型进行了检验。第三部分(第五章)则给出一种水库优化调度收益最小风险模型。该模型以事件风险概率为优化目标,寻求最优策略。具体可描述为:对预给定的调度期发电总收益水平,寻求最优的水库调度控制策略,使得收益低于此预收益值的概率风险最小。此模型有利于决策者根据自己风险承受能力及风险偏好,选择不同的控制策略。最后,对全文进行了总结与展望。在对全文研究工作进行总结的基础上,对今后有待于深入研究的问题及研究方向进行了展望。更多还原

【Abstract】 There exist many unpredictable factors during the hydro scheduling process, which cause many kinds of risk for the actual operation such as: inflow risk, decision-making risk, power generation revenue risk. The reservoir inflow risk is the potential one and plays a decisive role. Risk analysis for the decision-making is to take related methods to analyze and assess various possible errors and identify the most suitable decision-making with risk-benefit or risk preference matching. The inflow risk and generation revenue risk are studied.For the facing period of the whole regulation horizon, the main scheduling risk comes from the runoff uncertainty, which causes the uncertainty of the decision-making results and expectations. For the general adept hydrological forecasting model in applications, they are treated as certainty ones. The output of the forecast variable is always released to the users as a determinant value. However, the occurrence and development of the hydrological phenomenon is a complex dynamic process depending on the meteorological and geographical factors. Hydrological forecasting model accepts hydrological, meteorological and other inputs, using overviewed parameters and those are only the simulations of the objective hydrological process. All those complex factors lead the hydrological (runoff forecasts) forecasting to be uncertainty.For the reservoir generation operation based on the Markov process, the extensively applied optimization criterion is the optimal expectations revenue. However, the expectations criterion is not sensitive to the risk. and is not suitable for the optimal problems which need to reflect or limit the risk directly. For the optimal operation of hydropower, in order to guarantee a certain level of generation revenue, the operation strategy adopted should ensure the probability of the generation revenue beyond a given one to be lager than some given value while pursuing the expectations revenue to be more lager. Limited only to the expectation revenue as the optimization objective is not enough, it is necessary to choose other optimal goal to reflect the actual risk or the policy-makers’will or risk preferences.Based on reading and studying large number of literatures about reservoir scheduling, the long-term optimization scheduling problem and the scheduling risk assessment are studied. The main thesis content and innovative achievements include the following parts:A long-term compensation optimization scheduling method is introduced in chapter 3. The principles and characteristics and solving method of the model are introduced in detail. A detailed analysis on the results of the example is given too.One risk analysis method for the period scheduling in introduced in chapter 4. Bayesian statistic forecasting theory is adopted to formulate the reservoir long-term runoff forecasting model, which describes the uncertainty of hydrological forecast by distribution function. Gray correlation prediction model of meteorological factors is presented to count the uncertainty of the input factor. Real-time meteorological information and history hydrological data are coupled effectively, which breakthrough the limitations on information utilization and samples learning of the determined forecast method and improve the accuracy of hydrological forecast. The established model is tested on the runoff forecast of the Feng-man reservoir.A minimum risk control model for the reservoir long-term generation optimization is presented in chapter 5. The optimal scheduling criterian is the probability that the expected generation of the whole period not exceeding the pre-set generation target to be smallest. Compared with the generally used guidance of the largest expectation power generation model, this model is fitted for the decision-making in which the risk is needed to be limited to reflect the risk preference of the policy makers.Finally, a summarization of the dissertation is given in the last part (Chapter 6), and also proposes problems that need to be settled in the future and some suggestions for further study.更多还原