基于驾驶员需求转矩预测的模型预测控制能量管理

以并联式混合动力车辆为研究对象，基于驾驶员需求转矩预测，采用线性时变模型预测控制算法对对象车辆进行能量管理控制. 根据驾驶员前一段时间内的需求转矩，可以预测下一时段内驾驶员的需求转矩. 与指数函数预测方法相比，自回归模型在预测步长200步之内的预测准确度比指数函数高. 根据纵向动力学公式，可以由预测获得的需求转矩序列计算获得预测的车速序列；采用线性化处理的方法，将具有非线性特性的车辆模型转化成线性时变模型，采用线性时变模型预测控制算法进行求解；将基于驾驶员需求转矩预测的模型预测控制算法和基于规则的控制算法、工况已知的模型预测控制算法进行对比. 对比结果表明：基于驾驶员需求转矩预测的模型预测控制算法与基于规则的控制算法相比，在NEDC、UDDS和WLTC这3个标准工况下的燃油经济性均有所提高，但是与工况已知时的计算结果相比有提升的空间.

Model predictive control energy management based ondriver demand torque prediction

The linear-time varying model predictive control was proposed for the vehicle energy management control strategy based on the driver demand torque prediction aiming at the parallel hybrid electric vehicle. The driver’s demand torque in the next period was predicted according to the driver's demand torque in the previous period. The prediction accuracy of the autoregressive model within the 200 prediction step length is higher compared with the exponential function prediction method. The predicted speed sequence can be calculated from the predicted demand torque sequence according to the longitudinal dynamic formula. Then the vehicle model with nonlinear characteristics was transformed into a linear time-varying mode. The linear time-varying model predictive control algorithm was used to solve the problem. The model predictive control algorithm based on driver demand torque prediction was compared with the rule-based control algorithm and the model predictive control algorithm with known operating conditions. The comparison results show that the fuel economy of the model predictive control algorithm based on driver demand torque prediction is better under the three standard operating conditions of NEDC, UDDS and WLTC compared with the rule-based control algorithm, but there is room for improvement compared with the results of known operating condition.