基于CSO-BP神经网络的电力系统短期负荷预测

发布时间：2018-06-07 04:39

本文选题：负荷预测 + 数据预处理　；参考：《西安理工大学》2017年硕士论文

【摘要】：电刀系统短期负荷预测意义深远,是系统优化运行的基础,负荷预测结果对系统的经济与安全有重要意义。目前有很多种方法可以进行负荷预测,但是人们仍然希望得到更高精度的预测结果。因此,探索研究快速、准确的负荷预测方法有很大意义。本文详细介绍了电力系统短期负荷预测的研究内容,总结了国内外负荷预测的方法,分析了各预测方法的优缺点。分析了传统的神经网络预测算法的不足,提出了猫群算法优化BP神经网络的组合算法,本文的内容主要包括以下个方面:分析了美国某地的负荷特性,总结了负荷的周期特点和节假日特性,分析了影响负荷的各种因素与负荷之间的关系。介绍了历史负荷数据中不良数据的成因与对负荷预测的影响,采用改良后的模糊c-均值聚类算法对日负荷曲线进行聚类,得到若干特征曲线,通过负荷的横向相似性找出不良数据;最后对不良数据进行修正。修正不良数据,消除负荷曲线毛刺。建立了考虑日特征相关因素的BP神经网络负荷预测模型。以历史时刻的温度、负荷、星期类型归一化后作为BP神经网络预测模型的输入变量。介绍了猫群算法,测试函数表明猫群算法相比与遗传算法收敛更快且不易陷入局部最优解。为克服神经网络收敛速度慢,网络初始值选取不当而陷入局部极小点的缺点,提高电力系统短期负荷预测精度,将猫群算法和BP神经网络结合,提出基于猫群-BP神经网络的短期负荷预测模型。用猫群算法优化BP神经网络的权值和阈值,避免神经网络初始权值选取的盲目性,训练BP神经网络预测模型求得最优解。算例表明,猫群优化的BP神经网络有效改善了 BP神经网络的缺点,提高了神经网络用于电力系统短期负荷预测精度的效率和精度。相比于遗传神经网络,该方法预测精度更高,收敛更快,具有一定的实用性。
[Abstract]:The short-term load forecasting of the electric knife system is of far-reaching significance and is the basis of the optimal operation of the system. The result of the load forecasting is of great significance to the economy and safety of the system. At present, there are many methods for load forecasting, but people still want to get more accurate forecasting results. Therefore, it is of great significance to explore fast and accurate load forecasting methods. This paper introduces the research contents of short-term load forecasting in power system, summarizes the methods of load forecasting at home and abroad, and analyzes the advantages and disadvantages of each forecasting method. In this paper, the shortcomings of the traditional neural network prediction algorithm are analyzed, and the combination algorithm of the cat swarm optimization BP neural network is proposed. The main contents of this paper include the following aspects: the load characteristics of a certain place in the United States are analyzed. The periodic and holiday characteristics of load are summarized, and the relationship between various factors affecting load and load is analyzed. This paper introduces the causes of bad data in historical load data and its influence on load forecasting. The improved fuzzy c-means clustering algorithm is used to cluster the daily load curve, and some characteristic curves are obtained. Through the lateral similarity of load to find out the bad data, and finally to correct the bad data. Correct bad data and eliminate burr of load curve. A BP neural network load forecasting model considering the correlation factors of daily characteristics is established. The temperature load and week type of historical time are normalized as input variables of BP neural network prediction model. The cat swarm algorithm is introduced. The test function shows that the cat swarm algorithm converges faster than the genetic algorithm and is not easy to fall into the local optimal solution. In order to overcome the shortcomings of slow convergence speed of neural network and improper selection of initial value of neural network, and to improve the accuracy of short-term load forecasting in power system, the cat swarm algorithm and BP neural network are combined. A short term load forecasting model based on cat swarm BP neural network is proposed. The weight and threshold of BP neural network are optimized by using cat swarm algorithm to avoid blindness of selecting initial weight of neural network, and BP neural network prediction model is trained to obtain the optimal solution. The examples show that the BP neural network based on cat swarm optimization can effectively improve the shortcomings of BP neural network and improve the efficiency and accuracy of the neural network used in short-term load forecasting of power system. Compared with genetic neural network, this method has higher prediction accuracy, faster convergence and better practicability.
【学位授予单位】：西安理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM715

【参考文献】