基于快速模型预测控制的超级电容城轨充电

A fast model predictive control based ultracapacitor urban rail charging

超级电容作为一种新能源设备已经应用到了城轨列车中。常规时储能式城轨超级电容仅需在停站的30 s内充满电就能保证列车到下一站的正常运行。而这也要求充电设备提供快速上升无尖峰的大功率充电电流。考虑到充电需求及超级电容模型复杂性,将该充电问题构建为一个模型预测控制问题,通过一种改进的内点法对所构建的目标函数进行求解。相较于传统的内点法,仅选用一个经过试验验证的固定障碍参数进行求解。同时引入一种暖启动机制,将每一时刻的最终解作为下一时刻的初始解。两种机制的引入大大简化了目标函数的求解过程,实现了模型预测控制在大功率快速充电控制中的高效应用,控制效果达到预期。仿真及试验验证了所设计算法的有效性。

Ultracapacitor has been utilized as main power in urban rail vehicle. When urban rail parks at the station, roof-mounted ultracapacitors stack needs to be charged rapidly within 30 seconds, which requires that the charging platform can offer a high power current without peak. Taking the charging requirement and the complexity of ultracapacitor into consideration, the mentioned charging problem is formed as a model predictive control（ MPC） problem in this paper. An extended interior point method is utilized to solve the formed objective function. Instead of using a sequence of barrier parameters, this method only takes a fixed one that is carefully selected by experiment. Moreover, a mechanism called warm start is taken to further simplify the solving process, which uses the previous optimal solution as the initial solution of current instant. These two mechanisms dramatically simplify the solving process of objective function, and lead to satisfactory performance when MPC algorithm is utilized in this high power charging application. Simulation and experiment verify the effectiveness of the proposed algorithm.