基于小波-BP神经网络的贝叶斯概率组合预测模型及其在预报调度中的应用

发布时间：2018-08-06 11:35

【摘要】：中长期径流预报方法一直是国内外研究的热点和难点,从传统的成因分析方法、水文统计法、时间序列分析方法等,发展到现代的人工神经网络、小波理论、灰色系统和混沌理论等,各方法因其机理与适用环境不同而各具优势。另外,随着水电站在电网系统的作用日益显著,以及水电站在电网系统的调度与运行日益复杂,继续深入研究中长期径流预报方法、补充和完善相关理论与方法,以合理、有效地提高中长期径流预报的精度,并在此基础上形成指导水库运行的调度策略,具有重要的理论意义和应用前景。本文主要完成如下两部分工作：(1)采用一元线性回归模型模拟贝叶斯分析的先验分布和似然函数,建立了基于小波-BP神经网络的贝叶斯概率组合预测模型,将其应用于老挝Namngum水库月径流量预测中。该模型有效提高了预测精度；此外,同时相对于确定性水文预报方法而言,组合预测模型可定量地、以分布函数形式描述水文预报的不确定度,为后续水库调度提供了更多、更全面的信息。(2)以Namngum水电站为研究实例,以组合预报结果为依据,建立以发电量最大为目标函数的优化调度模型,并采用POA算法进行求解；将调度结果同现有运行方式下的结果进行对比,结果表明,应用WA-BP-BY模型预报结果可在原有基础上进一步提高Namngum水电站水库的发电效益,可为今后水电站水库发电计划制定提供参考依据。
[Abstract]:Long-term runoff forecasting method has been a hot and difficult point in domestic and international research. From traditional cause analysis method, hydrological statistics method, time series analysis method and so on, it has developed to modern artificial neural network, wavelet theory, etc. The grey system and chaos theory have their own advantages because of their different mechanism and applicable environment. In addition, with the increasingly significant role of hydropower stations in the power network system, and the increasingly complex operation and operation of hydropower stations in the grid system, the long-term runoff forecasting methods are further studied to supplement and improve the relevant theories and methods in order to be reasonable. It is of great theoretical significance and application prospect to improve the precision of medium and long term runoff forecasting and to form the dispatching strategy to guide reservoir operation on this basis. The main work of this paper is as follows: (1) A Bayesian probability combination prediction model based on wavelet BP neural network is established by using a linear regression model to simulate the prior distribution and likelihood function of Bayesian analysis. It is applied to forecast monthly runoff of Namngum reservoir in Laos. In addition, compared with the deterministic hydrological forecasting method, the combined forecasting model can quantitatively describe the uncertainty of hydrological forecast in the form of distribution function, which provides more for the subsequent reservoir operation. (2) taking the Namngum hydropower station as an example, based on the combined forecast results, the optimal dispatching model with the maximum generating capacity as the objective function is established and solved by using the POA algorithm; By comparing the operation results with those under the existing operation mode, the results show that the application of WA-BP-BY model forecast results can further improve the power generation efficiency of the Namngum hydropower station reservoir on the basis of the original prediction results. It can provide reference basis for future hydropower station reservoir power generation plan formulation.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TV697.1;TV124

【参考文献】