基于对角递归神经网络的智能小区用电负荷预测研究

发布时间：2018-02-15 06:13

本文关键词： 负荷预测对角递归神经网络粒子群算法灰色关联分析主成分分析　出处：《东北石油大学》2017年硕士论文　论文类型：学位论文

【摘要】：负荷分析是实现当地电力规划的前提和基础。在国家发展智能电网的大背景下,智能小区将是未来人们住宅小区呈现的主要形式,因此研究小区的用电负荷变化趋势以及特点具有重要的意义。目前供电企业对智能小区用电负荷的分析以及预测结果的精度在实际应用中并不是非常理想,主要原因是负荷本身会受到多种因素的影响而发生变化,所以不考虑干扰因素的简单预测对预测结果和数据分析来说准确度并不会很高。为了提高精度,基于原有的数据统计信息的采集模式和硬件设备,通过改变处理数据的方式来达到理想的效果。本文在研究了目前主要负荷预测方法的基础上,通过对智能小区用电负荷的影响因素、数据维度进行分析处理并以此作为理论,提出了一种适合智能小区用电负荷预测的分析方法。首先我们分析了与智能小区负荷变化相关的一些影响因素的特点,建立主要影响因素对智能小区用电负荷影响的灰关联分析模型,得到了主要参数与用电负荷的灰关联度,分析比较了用电量与各个影响因素的影响程度及其权重,分析结果表明小区用电量与大部分影响因素存在较强关联性,并且每个影响因素的影响关联度随季节、地域、时间段变化都有着明显的不同;其次,为提升后续负荷预测数据的处理速度,提出基于主成分分析方法的历史数据预处理方法,可以降低数据维度,不仅减少数据量、剔除冗余数据,提高了运算速度,而且避免数据量过少,保证运算精度。在进行负荷预测前,对数据降维处理以及影响因素分析的基础上,本文提出以对角递归神经网络算法为核心的智能小区用电负荷数据处理方法,以季节、常用电器、人口等影响因素为输入量,以小区用电负荷为输出量,通过对大量历史数据进行训练,并用粒子群算法优化了神经网络的权值,从而预测目标时段小区用电量,再与实际统计值进行对比,从而判断该预测模型的预测准确度。经过MATLAB仿真验证表明,本文提出的方法可以对小区的负荷做出准确度较高的预测,有效性得到了证实。
[Abstract]:Load analysis is the premise and foundation of local electric power planning. In the context of the national development of smart grid, intelligent community will be the main form of people's residential district in the future. Therefore, it is of great significance to study the changing trend and characteristics of power load in residential areas. At present, the accuracy of power load analysis and prediction in intelligent residential areas is not very ideal in practical application. The main reason is that the load itself will be affected by a variety of factors, so a simple prediction that does not take into account interference factors will not be very accurate for forecasting results and data analysis. Based on the original data statistical information collection mode and hardware equipment, the method of data processing is changed to achieve ideal results. This paper studies the main load forecasting methods. Through the analysis and processing of the influencing factors and data dimension of the power load in the intelligent residential area, this paper takes it as the theory. This paper presents an analysis method suitable for power load forecasting in intelligent residential areas. Firstly, we analyze the characteristics of some influencing factors related to load changes in intelligent residential areas. The grey correlation analysis model of the influence of main factors on the power load of intelligent residential area is established, and the grey correlation degree between the main parameters and the power load is obtained, and the influence degree and weight of the power consumption and each influencing factor are analyzed and compared. The results show that there is a strong correlation between electricity consumption and most of the influencing factors, and the correlation degree of each influencing factor varies significantly with the seasons, regions and time periods. Secondly, In order to improve the processing speed of subsequent load forecasting data, a historical data preprocessing method based on principal component analysis (PCA) is proposed, which can reduce the data dimension, reduce the amount of data, eliminate redundant data, and improve the operation speed. And to avoid too little data, to ensure the accuracy of the operation. Before load forecasting, the data dimension reduction processing and impact factors analysis, In this paper, a data processing method based on diagonal recurrent neural network algorithm is proposed for power load data processing in intelligent residential area. The data processing method takes the influence factors such as season, commonly used electrical appliances and population as the input quantity, and takes the cell power load as the output quantity. By training a large number of historical data and optimizing the weights of neural network with particle swarm optimization algorithm, the power consumption of the target period is forecasted, and then compared with the actual statistical value. The MATLAB simulation results show that the method proposed in this paper can predict the load of the residential area with high accuracy and its validity is verified.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM715;TP183

【参考文献】