基于DTW-SC与Bi-LSTM网络的电动公交短期充电负荷预测

对电动公交车进行较为精准的短期充电负荷预测，有利于电网合理调度，从而在一定程度上缓解大规模电动公交车接入对电网冲击的影响。因此，提出一种基于DTW-SC与Bi-LSTM网络的电动公交短期充电负荷预测方法，以提高负荷预测精度。首先，结合电动公交车充电负荷具有的间歇性和波动性特点，提出基于动态时间规整(dynamic time warping, DTW)的改进谱聚类(Spectral Clustering, SC)方法，对公交车日充电负荷曲线进行聚类；其次，对每类负荷综合考虑时间、日类型、温度及历史负荷值等影响因素，利用双向长短期记忆(bi-directional-long short-term memory, Bi-LSTM)构建电动公交车短期充电负荷预测模型；最后，利用某市实际天气数据和历史负荷数据进行仿真验证，并与其它预测方法进行对比分析。实验结果表明，所提方法能提高短期充电负荷预测准确度。

Short-term Charging Load Forecasting for Electric Buses Based on DTW-SC and Bi-LSTM Networks

A more accurate short-term charging load forecasting for electric buses is conducive to reasonable grid dispatching, thus alleviating the impact of large-scale electric bus access on the grid to a certain extent. Therefore, an electric bus short-term charging load prediction method based on DTW-SC with Bi-LSTM network is proposed to improve the load forecasting accuracy. Firstly, the improved spectral clustering (SC) method based on dynamic time warping (DTW) is proposed to cluster the daily charging load curve of buses by combining the intermittent and fluctuating characteristics of electric bus charging load. Secondly, a short-term charging load forecasting model for electric buses is constructed using Bi-LSTM neural network for each type of load considering the influencing factors such as time, day type, temperature and historical load values. Finally, the actual weather data and historical load data of a city are used for simulation and validation, and are compared and analyzed with other forecasting methods. The experimental results show that the proposed method can improve the accuracy of short-term charging load forecasting.