【作者】 吴宏晓；

【导师】 侯志俭；

【作者基本信息】 上海交通大学 ， 电力系统及其自动化， 2007， 博士

【摘要】 短期负荷预测是电力系统负荷预测的重要组成部分,是电力系统调度和计划部门安排购电计划和制定运行方式的基础,对电力系统的可靠和经济运行意义重大。随着电力市场的发展,精度高、速度快的预测理论和方法越来越受到重视。本文围绕这一课题进行了研究和探讨。论文首先综述了短期负荷预测的意义、特点、影响因素及其主要研究方法的优缺点,明确了本论文的主要研究内容和研究方向。在第二章中,介绍了神经网络的基本原理,指出了神经网络在应用中存在的问题。论述了免疫算法的机理和研究现状,给出了免疫算法的一般步骤。本论文取得的主要成果如下:1.提出了一种免疫聚类径向基函数神经网络(ICRBFNN)模型来预测电力系统短期负荷。在ICRBFNN的设计中,根据共生进化和免疫规划原理,提出了共生进化免疫规划聚类算法,可以自动确定RBF网络隐层中心的数量和位置,并采用递推最小二乘法确定网络输出层的权值。对华东某市进行的电力系统短期负荷预测表明,与传统的径向基函数神经网络预测方法相比,ICRBFNN方法具有更高的预测精度。2.提出了一种协同进化免疫网络模型来预测电力系统短期负荷。在神经网络的设计中,根据协同进化和免疫算法原理,提出了协同进化免疫算法,对神经网络的结构和参数同时进行学习,形成了一种新型的神经网络学习算法。对斯洛伐克东部电力公司进行的电力系统短期负荷预测表明,与传统的径向基函数神经网络预测方法相比,协同进化免疫网络方法具有更高的预测精度。3.在对支持向量机(SVM)方法进行分析的基础上,提出了一种免疫支持向量机(IWSVM)方法来预测电力系统短期负荷,其中利用免疫规划算法优化支持向量机方法的参数。免疫规划算法利用浓度和个体多样性保持机制进行免疫调节,有效地克服了未成熟收敛现象,提高了群体的多样性。电力系统短期负荷预测的实际算例表明,与支持向量机方法相比,所提免疫支持向量机方法具有更高的预测精度。4.针对城市电力系统年用电量增长的特点,将灰色神经网络模型GNNM(1,1)引入城市年用电量预测。GNNM(1,1)模型是把灰色方法与神经网络有机结合起来,对复杂的不确定性问题进行求解所建立的模型。该模型通过建立一个BP网络,来映射GM(1,1)模型的灰色微分方程的解。GNNM(1,1)模型采用BP学习算法,网络经训练收敛后就可进行城市年用电量预测。算例计算表明,与灰色预测方法相比,GNNM(1,1)模型具有更强的适应性和更高的预测精度,适用于城市年用电量预测。论文最后对上述研究成果进行了总结,指出了短期负荷预测中还有待进一步研究的问题。更多还原

【Abstract】 Short-term load forecasting (STLF) is very important to the control, operation, and schedule of power system. Deregulation and competition of the power industry are now propelling the utilities to operate the system at an even higher efficiency. This trend further intensifies the concern as to the accuracy of STLF. The paper aims to improve the precision of STLF.In chapter 1, the significance and current situation of main researches in the paper are introduced. The original works completed by the author are also pointed out. In chapter 2, the basic principle of the neural network and immune algorithm is introduced.The main achievements of paper are1. An immune clustering RBF neural network(ICRBFNN) model is presented for STLF. In the design of the ICRBFNN, a novel clustering method based on the symbiotic evolutionary and the immune programming algorithm(SEIPCM) is proposed. The SEIPCM automatically adjust the number and positions of hidden layer RBF centers. The weights of output layer are decided by the recursive least squares algorithm. The proposed ICRBFNN model has been implemented based on the actual data collected from the East China Power Company and compared with the traditional RBF neural network(RBFNN) method. The test results reveal that the ICRBFNN method possesses far superior forecast precision than the RBFNN method.2. A cooperative coevolutionary immune network (CCIN) model is presented for STLF. In the design of neural network, a novel method based on the cooperative coevolutionary and the immune algorithm (CCIM) is proposed. The CCIM is used to evolve the structure and parameters of neural network. The proposed CCIN model has been implemented based on the actual data collected from the Eastern Slovakian Electricity Corporation and compared with the traditional RBF network (RBFN) method. The test results reveal that the CCIN method possesses far superior forecast precision than the RBFN method.3. An immune support vector machines(IWSVM) method is presented for STLF. The immune programming algorithm, inspired by the immune system of human and other mammals, is used to optimize the parameters of support vector machines. The algorithm has the advantage in preventing premature convergence and promoting population diversity. The forecasting results demonstrate that the proposed method has higher forecast precision for STLF.4. According to the speciality of electricity demand development in a city, the grey neural network model GNNM(1,1) is introduced into the field of city electricity demand forecasting. The GNNM(1,1) model is the combination of grey system and neural network, which can solve the complex uncertain problems. The GNNM(1,1) model builds a kind of BP neural network which can map the solution to the grey differential equation of GM(1,1) model, then the model is trained by using BP algorithm. City electricity demand is forecasted after the GNNM(1,1) model is convergent. The forecasting results demonstrate that the GNNM(1,1) model has higher adaptability and forecast precision for city electricity demand forecasting.The main works are summarized at the end of this paper. Further work to the research is pointed out.更多还原