针对电动汽车充电需求的时空不确定性，电动汽车代理商如何参与电力市场并且如何实现运营利润最大化成为需要解决的问题。本研究首先建立基于多层深度学习算法的电动汽车出行预测模型，分别使用多层感知器和长短期记忆网络对电动汽车的出行数据和路况数据进行学习，通过训练好的预测模型预测次日的出行行为和路况。其次，考虑到路况的多变性对预测精度的影响，采用未来路径滚动优化方法和速度-能耗模型模拟次日电动汽车的出行行为，从而获取电动汽车更精确的时空状态和荷电状态。最后，考虑能量市场协同调度，通过日前市场的充放电调度模型对不同时段的电动汽车充放电行为进行规划，最大化电动汽车代理商的利益。为了证明所提预测方法的准确性，与预测常用的蒙特卡洛方法和拉丁超立方法进行对比，结果表明，本研究提出的深度学习算法具有更高的准确性。将模型应用于IEEE33节点测试系统中进行验证，实验结果表明，在电动汽车代理商的调度下能够有效降低电力系统的峰谷差，在系统出现阻塞的情况下，通过改变电动汽车的调度策略能够缓解系统线路阻塞问题。对代理商收益和用户出行成本进行分析，结果表明在代理商的调度下，不仅能够提高代理商收入还能够降低用户的出行成本，实现双赢。

In view of the time-space uncertainty of electric vehicle charging demand, how to participate in the electricity market and how to maximize the operating profit has become a problem that needs to be solved. Firstly, this study established an electric vehicle travel prediction model based on multi-layer deep learning algorithm. The multilayer perceptron and long short-term memory network were used to learn the travel data and road condition data of electric vehicles, and the travel behavior and road condition of the next day were predicted by the trained prediction model. Secondly, considering the influence of the variability of road conditions on the prediction accuracy, the future path rolling optimization method and the speed-energy consumption model were used to simulate the travel behavior of electric vehicles the next day, so as to obtain more accurate time-space state and charge state of electric vehicles. Finally, considering the coordinated scheduling of the energy market, the charging and discharging behavior of electric vehicles in different periods was planned through the charging and discharging scheduling model of the day-ahead market to maximize the interests of electric vehicle agents. In order to prove the accuracy of the proposed prediction method, it was compared with the commonly used Monte Carlo method and Latin hypercube method. The results show that the deep learning algorithm proposed in this study has higher accuracy. The model was applied to the IEEE33 node test system for verification. The experimental results show that the peak-valley difference of the power system can be effectively reduced under the scheduling of electric vehicle agents. In the case of system congestion, the problem of system line congestion can be alleviated by changing the scheduling strategy of electric vehicles. The agent’s revenue and the user’s travel cost were analyzed. The results show that under the agent’s scheduling, it can not only increase the income of agents, but also reduce the travel cost of users, and achieve a win-win situation.