电动汽车智能配电网短期负荷预测研究

发布时间：2018-04-04 16:46

本文选题：电动汽车　切入点：负荷预测　出处：《东南大学》2015年硕士论文

【摘要】：电动汽车是国家十二五期间重点引导和扶持的产业,针对未来电动汽车大规模普及的趋势,本文结合电动汽车充电自身的特点,从计算规模化电动汽车入网充电负荷时空分布和提高电动汽车功率预测精度两个方面展开研究。以推动电动汽车产业化进程,同时为电动汽车大规模接入电网安全稳定运行提供理论依据,降低大规模电动汽车并网带来的影响。本文的主要工作分为三个部分,总结如下：研究电动汽车规模化发展后其充电负荷在时间和空间上的分布是实现电动汽车与电网良好互动的基础。本文从两个方面着手：一方面,全面考虑不同类型电动汽车具有的不同行车规律、不同充电方式等,针对用户行为、充电模式、动力电池特性等各种因素分别进行讨论,研究各项因素对电动汽车充电功率特性的影响,保证功率计算模型输入参数的合理性,并通过综合算例分析复杂环境下电动汽车充电功率在时间上的分布。另一方面,由于电动汽车发展的社会属性,电动汽车的渗透率、充电电价和政府的引导策略等均会影响电动汽车充电负荷,且不同区域的电动汽车充电负荷特性也不同,综合上述影响因素分析电动汽车的集群效应并计算不同区域的电动汽车充电负荷在空间上的分布。在电动汽车时空分布已知的基础上,本文设计了一套多时间尺度负荷预测方案,分为日前预测、时前滚动修正预测和实时预测。日前预测结合待预测前多日电动汽车功率计算或实际结果,在满足预测精度的约束下对下一周每日24小时各时段功率进行预测。时前滚动修正预测结合当日新增数据,通过最新的电动汽车功率实际值来对日前预测在该时刻下的结果进行修正,降低日前功率预测误差对当日负荷预测值的影响。实时预测通过当时段新增负荷数据预测下一时刻(10或15mmin)负荷值,修正滚动预测在该时刻下的负荷预测误差。提高电动汽车功率预测精度是电动汽车规模化入网后参与电网优化调度与运行的关键,电动汽车负荷序列在时空上的分布是无序且随机的,本文在传统电力预测方式的基础上提出了适用于电动汽车预测的由改进GA算法、GA-BP算法和改进灰色算法组成的多时间尺度电动汽车决策型预测模型,通过层次分析法与专家经验的结合确定改进GA算法、GA-BP算法和改进灰色算法三种算法在决策型算法中的权重,充分发挥三种算法在日前预测、滚动预测、实时预测上的各自最大优势,一定程度上对三种算法进行取长补短,提高预测精度,同时提高电动汽车与电网互动的水平,降低电动汽车充电的波动性与随机性对电网的影响,为后续电动汽车相关研究提供数据支持与理论支撑。
[Abstract]:Electric vehicle is a key guiding and supporting industry during the 12th Five-Year Plan period. In view of the trend of large-scale popularization of electric vehicle in the future, this paper combines the characteristics of electric vehicle charging itself.The research is carried out from the aspects of calculating the space-time distribution of charging load and improving the precision of power prediction for large-scale electric vehicles (EVs).In order to promote the industrialization process of electric vehicles and provide the theoretical basis for the safe and stable operation of large-scale electric vehicles connected to the power grid, reduce the impact of large-scale electric vehicles connected to the grid.The main work of this paper is divided into three parts, which are summarized as follows: the research on the distribution of charge load in time and space after the development of electric vehicle scale is the basis of realizing good interaction between electric vehicle and power grid.This paper starts from two aspects: on the one hand, considering the different driving rules and charging modes of different types of electric vehicles, and discussing the various factors such as user behavior, charging mode, power battery characteristics, etc.The influence of various factors on the charging power characteristics of electric vehicles is studied to ensure the reasonableness of input parameters of the power calculation model, and the distribution of charging power of electric vehicles in time under complex environment is analyzed by a comprehensive example.On the other hand, because of the social attribute of the electric vehicle development, the electric vehicle's permeability, the charge price and the government's guiding strategy will all influence the electric vehicle's charge load, and the electric vehicle's charge load characteristic will be different in different area.Based on the above factors, the cluster effect of electric vehicle is analyzed and the spatial distribution of charging load of electric vehicle in different regions is calculated.Based on the known temporal and spatial distribution of electric vehicles, this paper designs a multi-time scale load forecasting scheme, which is divided into pre-day prediction, time-front rolling correction prediction and real-time forecasting.Combined with the power calculation or actual results of electric vehicle in many days to be forecasted, the power of 24 hours a day in the next week is forecasted under the constraint of prediction precision.Combined with the new data on the same day, the time front rolling correction forecast is used to correct the results of the day forecast at that time through the latest actual value of electric vehicle power, so as to reduce the influence of the day power forecast error on the day load forecast value.In real time forecasting, the load prediction error of rolling forecasting at the next time is corrected by adding load data to forecast the load value of the next time, 10 or 15 min.To improve the precision of electric vehicle power prediction is the key to participate in the optimal dispatching and operation of electric vehicle after it is put into the network on a large scale. The distribution of load sequence of electric vehicle in time and space is disordered and random.Based on the traditional power forecasting method, this paper presents a multi-time scale decision model for electric vehicles, which is composed of improved GA algorithm and improved grey algorithm, which is suitable for electric vehicle prediction.Through the combination of AHP and expert experience, the weight of the improved GA algorithm GA-BP algorithm and the improved grey algorithm algorithm in the decision-making algorithm is determined.To a certain extent, the three algorithms can be used to complement each other, improve the prediction accuracy, improve the level of interaction between electric vehicles and power grid, and reduce the impact of electric vehicle charging volatility and randomness on the power grid.To provide data support and theoretical support for the subsequent research on electric vehicles.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM715

【相似文献】