基于神经网络的电力系统负荷预测问题研究

发布时间：2018-11-06 13:00

【摘要】：随着对电能需求的增加，电力系统的发展及改进变得更加重要。电力系统的负荷预测对系统调度的自动化十分重要，并且对于电力系统的安全稳定及经济运行具有重要意义。负荷预测的精度直接影响着电网的安全稳定，其预测结果为发电机组的运行提供帮助，为电厂的燃料供应计划提供依据，同时能够提高对系统的控制。预测结果不准确或误差过大会影响发电部门的燃料合理配置，减少其收益。研究具有精度高且实用性强的负荷预测方法对于电力的市场化及智能电网的发展是非常必要的。本文通过查阅相关文献，介绍了电力系统负荷预测的研究现状，分析并比较现有的不同预测方法的特点，具体研究人工神经网络方法的原理及学习算法，它通过对人脑基本特性抽象和模拟，形成一种自适应的并行信息处理方法，具有自学习和非线性映射等特点，对于电力系统的负荷预测有重要的应用价值。文中详细介绍了误差反向传播（Back Propagation，BP）神经网络和径向基函数（Radial Basis Function，RBF）神经网络的模型结构及其学习算法，分别建立了基于BP神经网络和RBF神经网络的电力负荷预测模型。建立模型的过程，对输入的原始数据进行预处理，去除不良数据并补充缺失数据；为避免神经元饱和，对输入样本作归一化处理；对于模型的初始权值及学习参数的选取也进行了分析。对建立的两个模型进行比较，BP神经网络模型的所需的学习训练时间较长，收敛性差，容易陷入局部极小情况；RBF神经网络模型的训练速度较快，收敛性好，，对于电力系统的负荷预测具有更大的优势。然后介绍了模糊控制理论，模糊理论控制方法不必建立精确的数学模型，便可以实现对复杂系统的控制。具体介绍模糊控制器的结构及其设计过程，包括输入变量选取及模糊推理和判决。利用模糊控制理论对RBF神经网络模型进行调整改进，提高其收敛速度，减少训练的时间，建立基于RBF神经网络与模糊控制相结合的电力系统负荷预测模型。利用建立的BP神经网络、RBF神经网络模型及RBF神经网络与模糊控制结合的模型，对某地区的实际负荷进行了预测，并对结果进行了误差分析与比较。这几种方法所得的预测结果的精度都能够满足电力部门的实际要求，说明了他们的有效性及实用性。应用RBF神经网络与模糊控制相结合的模型所得的结果误差最小，预测效果更好，说明该方法对于电力系统的负荷预测具有实用意义。
[Abstract]:With the increase of power demand, the development and improvement of power system becomes more and more important. The load forecasting of power system is very important to the automation of power system dispatching, and it is of great significance to the safety, stability and economic operation of power system. The accuracy of load forecasting directly affects the safety and stability of power grid. The forecasting results provide help for the operation of generating units, provide the basis for the fuel supply plan of power plants, and improve the control of the system at the same time. Inaccurate prediction results or errors affect the rational allocation of fuel in the power generation sector and reduce its benefits. It is necessary to study load forecasting methods with high accuracy and practicability for the development of power market and smart grid. In this paper, the current situation of load forecasting in power system is introduced, the characteristics of different forecasting methods are analyzed and compared, and the principle and learning algorithm of artificial neural network are studied in detail. By abstracting and simulating the basic characteristics of human brain, it forms an adaptive parallel information processing method, which has the characteristics of self-learning and nonlinear mapping, and has important application value for power system load forecasting. In this paper, the model structure and learning algorithm of error back propagation (Back Propagation,BP) neural network and radial basis function (Radial Basis Function,RBF) neural network are introduced in detail. Power load forecasting models based on BP neural network and RBF neural network are established respectively. In order to avoid neuronal saturation, the input sample is normalized by pre-processing the input original data, removing the bad data and supplementing the missing data. The selection of initial weights and learning parameters is also analyzed. Compared with the two models, the BP neural network model needs long learning and training time, poor convergence, easy to fall into the local minimum; The training speed of RBF neural network model is fast and the convergence is good. It has more advantages for power system load forecasting. Then the fuzzy control theory is introduced. The fuzzy theory control method does not need to establish the accurate mathematical model to realize the control of the complex system. The structure and design process of fuzzy controller are introduced in detail, including input variable selection, fuzzy reasoning and decision. The fuzzy control theory is used to adjust and improve the RBF neural network model to improve its convergence speed and reduce the training time. A power system load forecasting model based on the combination of RBF neural network and fuzzy control is established. Using the established BP neural network, RBF neural network model and RBF neural network combined with fuzzy control model, the actual load in a certain area is forecasted, and the error analysis and comparison of the results are made. The accuracy of the prediction results obtained by these methods can meet the practical requirements of the power sector, which shows their effectiveness and practicability. The model based on RBF neural network and fuzzy control has the smallest error and better prediction effect. It shows that this method is of practical significance for power system load forecasting.
【学位授予单位】：沈阳工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM715;TP183

【参考文献】