【摘要】：隨著社會與經濟的不斷發(fā)展,電能成為制約社會與經濟可持續(xù)發(fā)展的瓶頸因素之一。水電作為清潔能源,已成為我國能源結構中非常重要的組成部分,而且變得越來越重要。為了提高水能資源利用率,提高水庫發(fā)電的調度水平是重要途徑之一。近年來,隨著氣象預報技術及計算機技術的不斷發(fā)展,利用數(shù)值降雨預報信息來延長徑流預報的預見期和提高預報精度成為可能性。如何根據(jù)現(xiàn)有的預報信息提高水庫發(fā)電調度水平是現(xiàn)階段亟待解決的問題。 為此,本文選擇渾江梯級水庫群為例,采用美國全球預報系統(tǒng)(GFS)發(fā)布的降雨預報信息作為資料開展考慮預報信息的水庫群發(fā)電調度研究。首先,基于GFS數(shù)值降雨預報信息對渾江梯級水庫群的徑流進行預報,以此研究結果作為本文研究的信息輸入。然后在此基礎上,圍繞耦合預報信息的水庫發(fā)電調度模型展開了研究,分別從隱隨機、顯隨機和模擬-優(yōu)化的角度作為研究的切入點,并分別建立考慮預報信息的調度模型。最后,基于預報和調度兩方面的研究成果,建立渾江梯級水庫群發(fā)電預報調度系統(tǒng)作為系統(tǒng)的輸出。本文研究成果主要有以下幾個方面： (1)由于水文預報模型的結構、參數(shù)和初始值等誤差均會對預報結果產生影響,因此,采用多個不同類型的水文模型對中期徑流進行預報,并以聯(lián)邦濾波算法將多個模型預報結果進行融合,以達到模型缺點互補的效果。結果表明該組合預報模型能夠提高預報精度和穩(wěn)定性,而且組合預報模型對模型結構、參數(shù)和初始值帶來的預報誤差有較好的校正效果。 (2)在中期徑流預報研究成果基礎上,采用隱性的方式挖掘尋求具有普適性的發(fā)電調度規(guī)則,使水庫群發(fā)電調度耦合徑流預報信息。首先建立粗糙集和決策樹兩種數(shù)據(jù)挖掘算法；然后以確定性優(yōu)化算法獲得的發(fā)電調度理想過程為基礎,挖掘調度數(shù)據(jù)集中蘊含的發(fā)電調度規(guī)律；最后依據(jù)數(shù)據(jù)挖掘模型對具有相似性的調度決策合并同化為具有普適性的調度規(guī)則,從而獲得具有普適性的發(fā)電調度規(guī)則。 (3)為了比較水庫群發(fā)電調度模型效率,采用模擬-優(yōu)化的思想制定考慮預報信息的發(fā)電預報優(yōu)化調度圖。預報優(yōu)化調度圖中以旬為時間尺度將蓄水量轉化為流量量綱,蓄水量同徑流量耦合后構成調度圖的調度控制空間(OCS),并以OCS為預報調度圖的狀態(tài)變量建立水庫群的發(fā)電預報優(yōu)化調度圖,從而降低了調度圖的優(yōu)化“維數(shù)”并保持了調度圖簡明的結構。 (4)在水庫群發(fā)電調度模型中,采用顯性的方式考慮徑流預報信息。首先采用“聚合—分解”的思想,將梯級水庫群來水量和庫容進行聚合構造虛擬的“單一水庫”,然后以聚合來水量和聚合庫容作為變量建立聚合分解貝葉斯隨機動態(tài)規(guī)劃模型(AD-BSDP),并通過AD-BSDP模型優(yōu)化計算獲得耦合徑流預報信息的水庫群發(fā)電總出力決策圖,最后在實時調度階段,以棄水量最小和蓄水量最大為目標將總出力優(yōu)化分配至各水庫。研究表明,AD-BSDP在具有較大不確定性預報信息條件下,比傳統(tǒng)調度圖和其他驗證模型具有更高的發(fā)電效益和穩(wěn)定性。 (5)針對發(fā)電調度隨機動態(tài)規(guī)劃模型中,徑流預報信息不確定性對發(fā)電效益和穩(wěn)定性有較大影響的問題,本章研究建立了滾動時域發(fā)電調度模型,在此基礎上將徑流預報信息不確定性對水庫發(fā)電調度效益和穩(wěn)定性的影響進行了定量化的評價。 (6)基于預報信息不確定性對發(fā)電效益和穩(wěn)定性的影響評價,將10天徑流預報信息劃分為1-5天和6-10天兩段,前時段預報信息具有較高的預報精度,后時段預報信息具有較大不確定性,但仍有重要參考價值。為此,采用貝葉斯理論將前5天和后5天預報徑流信息進行耦合,建立短、中期徑流預報信息相套接的聚合分解貝葉斯隨機動態(tài)規(guī)劃模型(TS-BSDP)。模擬結果表明TS-BSDP模型的發(fā)電效益和穩(wěn)定性優(yōu)于AD-BSDP及其他驗證模型。 最后對全文做了總結,并對有待于進一步研究的問題進行了展望。
[Abstract]:With the development of society and economy, electric energy is one of the bottleneck factors that restrict the sustainable development of society and economy. As a clean energy source, hydropower has become an important part of our energy structure, and it becomes more and more important. In order to improve the utilization of water energy resources, it is one of the important ways to improve the dispatching level of reservoir power generation. In recent years, with the continuous development of the meteorological forecast technology and the computer technology, the forecast period of the runoff forecast and the possibility of improving the prediction accuracy are extended by using the numerical rainfall forecast information. How to improve the power generation dispatching level of the reservoir based on the existing forecast information is a problem to be solved at the present stage. In order to do this, this paper selects Hunjiang cascade reservoir group as an example, and uses the rainfall forecast information issued by the United States Global Prediction System (GFS) as the data to carry out the research on the power generation and dispatching of the reservoir group considering the forecast information. in that first place, the runoff of the river cascade reservoir group is forecast based on the data rainfall forecast information of the GFS value, and the result is the information input in this paper. Then, on the basis of this, the study on the reservoir power generation dispatching model about the coupling forecast information is carried out, from the perspective of the hidden random, the explicit random and the simulation-optimization as the starting point of the study, and the scheduling model considering the forecast information is established respectively. Finally, based on the research results of the two aspects of forecasting and scheduling, the power generation forecasting and dispatching system of the Hunjiang cascade reservoir is established as the input of the system. The results of this paper are as follows: Surface: (1) Because of the structure, parameters and initial value of the hydrological forecasting model, the prediction results can be affected. Therefore, a number of different types of hydrological models are used to feed the medium-term runoff. the prediction of the line is carried out, and a plurality of model prediction results are fused in a federal filter algorithm to achieve the complementation of the model defects. The results show that the combined forecasting model can improve the forecasting precision and stability, and the prediction error of the model structure, parameters and initial value of the model is better than that of the model structure, parameters and initial values. (2) On the basis of the research results of the medium-term runoff forecast, the power generation dispatching rule with universal applicability is excavated in a recessive way, so that the power generation dispatching coupling diameter of the reservoir group is made The paper first sets up two data mining algorithms for rough set and decision tree, and then based on the ideal process of power generation scheduling, which is obtained by deterministic optimization algorithm, mining and dispatching data in the dispatching data set. and finally, merging the scheduling decisions with the similarity into a scheduling rule with universality according to the data mining model, so as to obtain the scheduling rule with universality. (3) In order to compare the efficiency of the power generation dispatching model of the reservoir group, the simulation-optimized thought is used to develop the power generation considering the forecast information. Forecasting and optimizing the dispatching map. In the forecast and optimization schedule, the water storage volume is converted into a flow quantity class on the time scale of ten days, and the water storage quantity is coupled with the runoff quantity to form the dispatching control space (OCS) of the scheduling graph, and the power generation of the reservoir group is established by using the OCS as the state variable of the forecast dispatching map. and the optimal 鈥淣umber of dimensions鈥，

本文編號：2384744