Abstract: The load fluctuation range of battery electric commercial vehicles is large, and directly applying the control strategies for passenger vehicles fails to ensure the control effect. This paper proposes a software model based load estimation method. The load estimation model adopts an error back propagation (BP) neural network optimized by a genetic algorithm, with vehicle speed, acceleration, torque and ramp signals as input parameters, among which the ramp signals are collected by a gyroscope, and the output is the estimated value of the total vehicle mass. After training with big data, the model parameters are frozen and then integrated into the application layer model of the vehicle control unit (VCU), and the average estimation accuracy of the model reaches 2.9%. This method realizes the function by means of software, conforms to the mainstream development concept of "software-defined vehicles", and effectively reduces the overall vehicle cost while achieving the same function.

Key words: genetic algorithm; neural network; gyroscope; vehicle control unit; load estimation model

摘要： 纯电商用车载重波动幅度大，直接沿用乘用车控制策略难以保障控制效果。文章提出 一种基于软件模型的重量估算方法，该载重估算模型采用遗传算法优化的误差反向传播（BP） 神经网络，以车速、加速度、扭矩及坡道信号为输入参数，其中坡道信号由陀螺仪采集，输 出为整车总质量估算值。经大数据训练后冻结模型参数，并将其集成至整车控制器（VCU） 应用层模型，模型平均估算精度达 2.9%。该方法依托软件实现功能，契合“软件定义汽车”的 主流开发理念，在达成同等功能的前提下有效降低整车成本。

关键词: 遗传算法；神经网络；陀螺仪；整车控制器；载重估算模型