电动汽车负荷预测方法适用性与应用研究

发布时间：2018-09-06 19:10

【摘要】：电动汽车行业作为清洁环保的新型交通出行载具,有着普通燃油车无法比拟的优势。近年来,在国家政策的扶持之下飞速发展。可以预期的是,在未来电动汽车充电负荷将成为电网用电负荷不可忽视的一个重要组成部分。然而电动汽车的充电行为会给电网造成很大的影响,出于电动汽车有序调度,能量管理和配网规划的考虑,对电动汽车的负荷预测提出了更多的预测需求和更高的预测精度要求。本分选取了快换式电动公交车、快充式电动出租车和快换式电动出租车三种预测场景样本数据,分别对基于灰色理论、概率模型和BP神经网络的三种电动汽车负荷预测方法的预测原理、预测数学模型和预测结果的适用性进行分析研究。对灰色预测模型的输入数据条件和短期负荷预测精度的关系进行考量,其一为输入数据量的不同与预测精度的关系分析,其二为输入数据的离散程度与预测精度的关系。应用基于灰色理论和BP神经网络的两种模型分析研究二者在电动汽车超短期和短期负荷预测时间尺度下的适用性,得到的结果表明在实际应用中,BP神经网络的预测效果优于灰色预测,尤其是在超短期负荷预测时间尺度下,可考虑预测点前一时刻的电动汽车负荷预测值的BP神经网络模型能有效的减少预测的平均误差和最大负荷相对误差。应用基于灰色理论和基于概率模型的电动汽车负荷预测方法对比分析二者在电动汽车中长期负荷预测时间尺度下的适用性。结果表明在中长期负荷预测尺度下,基于概率模型和基于灰色理论的电动汽车负荷预测方法各有预测侧重面。基于概率模型的负荷预测方法在原理之上更为适用于中长期负荷预测中考虑国家政策和未来电动汽车发展规模的典型日预测。基于灰色原理的负荷预测方法适用于电动汽车中长期负荷预测中日用电量和日最大负荷预测的应用。由于配网的规划需求电动汽车的时空负荷预测,本文利用电动汽车充电负荷与其空间分布的位置不相关这一特性,通过分别建立电动汽车时间维度的负荷预测模型和利用OD矩阵建立电动汽车空间负荷分配数学模型,结合二者得到时空联合分布的电动汽车充电负荷预测模型。
[Abstract]:Electric vehicle industry as a clean and environmental-friendly new type of transport vehicles, has an unparalleled advantage over ordinary fuel vehicles. In recent years, under the support of national policies, rapid development. It can be expected that the electric vehicle charge load will become an important part of the power grid load in the future. However, the charging behavior of electric vehicles will have a great impact on the power grid. Due to the consideration of the orderly scheduling, energy management and distribution network planning of electric vehicles, More forecasting demands and higher precision requirements are put forward for load forecasting of electric vehicles. This paper selects the sample data of three kinds of prediction scenarios, which are fast changing electric bus, fast charging electric taxi and fast changing electric taxi, respectively based on grey theory. The probabilistic model and BP neural network are used to predict the load of electric vehicle, and the applicability of forecasting mathematical model and forecasting results are analyzed and studied. The relationship between the input data condition of grey forecasting model and short-term load forecasting accuracy is considered. One is the analysis of the relationship between the difference of input data volume and the prediction accuracy, and the other is the relationship between the discrete degree of input data and the prediction accuracy. Two models based on grey theory and BP neural network are applied to analyze the applicability of the two models in the time scale of ultra-short term and short term load forecasting of electric vehicles. The results show that the prediction effect of BP neural network is better than that of grey forecasting in practical application, especially in the time scale of ultra-short-term load forecasting. The BP neural network model, which can consider the load forecasting value of electric vehicle at the previous moment, can effectively reduce the average error and the maximum load relative error. Based on grey theory and probabilistic model, the applicability of the two methods in the medium and long term load forecasting time scale of electric vehicles is compared and analyzed. The results show that under the medium and long term load forecasting scale, the probabilistic model and the grey theory based load forecasting method have different emphases. The probabilistic model based load forecasting method is more suitable for the typical daily forecasting considering the national policy and the future development scale of electric vehicles in the medium and long term load forecasting based on the principle. The method of load forecasting based on grey theory is suitable for the application of daily electricity consumption and daily maximum load forecasting in medium and long term load forecasting of electric vehicles. Because of the planning demand of distribution network, this paper makes use of the characteristic that the charge load of electric vehicle is not related to the location of its spatial distribution. The time dimension load forecasting model of electric vehicle and the mathematical model of space load distribution of electric vehicle are established by using OD matrix, and the charge load forecasting model of electric vehicle with time-space joint distribution is obtained by combining the two models.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM715

【参考文献】