【作者】 谢忠玉；

【导师】 王科俊；

【作者基本信息】 哈尔滨工程大学 ， 模式识别与智能系统， 2010， 博士

【摘要】 短期负荷预测是电力系统调度运营部门一项重要的基础工作,预测精度的高低直接影响到电网运行的安全性、经济性以及电能质量。由于电力负荷表现出复杂性、不确定性、非线性的特点,使得传统的电力负荷时间序列预处理技术得不到令人满意的效果。混沌是确定性的非线性系统中出现的类似随机的现象。随着非线性混沌动力学的发展,人们对时间序列的复杂性有了更深刻的认识,尤其是混沌时间序列的分析已经成为一个非常重要的研究方向,这给短期负荷时间序列分析与预测提供了科学的方法。本文基于混沌理论,对短期负荷的变化规律进行分析和预测。主要研究内容如下:1.从重构相空间理论出发,探讨了相空间参数对重构空间质量的影响,以及确定相空间嵌入维数和延迟时间各种不同的方法。在短期负荷重构参数的选取上,一方面,使用Cao方法弥补了伪最近邻域法进行短期负荷相空间重构不准确的缺陷;另一方面,引入改进的C-C方法求取嵌入参数。两种方法互相结合,以验证本文求得的延迟时间和嵌入维数的准确性。2.对电力小时负荷时间序列进行混沌性质识别。一方面,采用Cao方法从定性的角度分析短期负荷的混沌特性;另一方面,通过提取短期负荷时间序列的混沌特征量:饱和关联维数、Lyapunov指数和Kolmogorov熵,从定量的角度分析电力负荷时间序列的混沌特性。另外,由于G-P关联维数的计算结果易受多种因素的影响,通过引入非主观关联维以弥补G-P关联维数判断时间序列混沌特性的局限性。3.针对加权一阶局域法预测模型计算量较大,而且会产生累积误差这一问题,采用加权一阶局域法多步预测模型进行预测。对华北某电网1998年全年负荷时间序列进行相空间重构,并使用加权一阶局域法多步预测模型和基于最大Lyapunov指数的预测模型做对比实验,分别对一天和一周的负荷进行预测仿真实验,研究表明,两种相空间预测模型对电力负荷短期预测是有效的。4.在研究了混沌时间序列的可预测尺度问题的基础上,提出了一种运用RBF神经网络平均可预测尺度内对短期负荷进行直接多步预测的方法,通过对本文所用的短期负荷时间序列进行相应的预测研究,实例应用表明直接多步预测方法的预报精度较高,预报效果令人满意。更多还原

【Abstract】 Short-term load forecasting(STLF)is an important and basal component in the operation of any electric utility whose accuracy directly influence power system’s security, profit and quality. Because of the highly nonlinear, complexity and indeterminacy of the power load, means that the traditional load time series pretreatment technology can not be achieved satisfactory results. Chaos is a quasi-stochastic phenomenon appearing in deterministic nonlinear dynamic system. With the development of nonlinear chaotic dynamic, there are more profound recognitions to the complexity of time series. Especially the analysis of time series is becoming an important research aspect that provides a scientific basis for analysis and prediction of short load time series.Based on the chaos theory, characteristic of short-term load time series is analyzed and its forecasting methods are studied in this paper. The main contents discussed in this thesis are described as follows:1. The influence of phase space parameters on phase space quality and the methods for determining delay time & embedding dimension are discussed on the reconstruction theory. About selection of reconstruction parameters, on the one hand, False Nearest Neighbors method is improved by Cao method, and accurate phase reconstruction parameters of short-term load; on the other hand, improved C-C method is introduced into selection of reconstruction parameters. The results of the two methods were mutually supplemented and verified.2. For electric hourly load, on the one hand, short-term load is analyzed qualitatively be Cao methed; on the other hand, quantitative calculation about saturation correlation dimension, the largest Lyapunov exponent and Kolmogorov entropy of power load is used to identity their chaotic characteristics. In addition, it is found that the calculation of correlation dimension is seriously influenced by many factors. Through introducing non subjective approach, to avoid the limitations of judging whether time series has chaotic character or not by the invariants are pointed out.3. Large computational quantity and cumulative error are main shortcomings of add- weighted one-rank local-region method for prediction of chaotic time series. Local-region multi-steps forecasting model based on phase reconstruction is adopted for short-term load time series prediction. When using the method for the STLF of a certain power network in the north china, 1998. The chaotic characteristics of the load time series in this case are analyzed, and the phase space reconstruction parameters are deduced. Local-region multi-steps forecasting model and the largest Lyapunov method are used in the forecasting experiment for the load in one day and one week. The prediction results indicated that the chaotic model is effective for short term load forecasting.4. Studied on the predictable size of chaotic time series, a method of direct multi-step prediction of Short-term load is proposed, which is based on the average predictable size and radial basis functions neural networks. And short-term load that have been used in this paper is predicted by the method. Simulation results for direct multi-step prediction method could get high precision.更多还原