| The automobile industry has made great progress in the past few decades.Advanced driving assistance system can improve driving comfort and safety,which depends on accurate lateral vehicle dynamics states.However,these states can not be measured directly in an economic way,and needs to be estimated online through appropriate algorithm.In different road conditions,the change of the interaction force between the wheel and the ground will affect the performance of steer-by-wire.The existing algorithms have strong robustness to this part of the disturbance torque,but there is no targeted compensation.The estimation and compensation of this part of the disturbance torque can help to improve the steering performance of the vehicle.Three-level inverters are used to replace two-level inverters to drive motor in electric vehicle.The existing algorithm does not consider the form of neutral point voltage oscillation of three-level inverter in detail,which can help to solve the neutral point potential balance problem of three-level inverter and improve the output current quality.In this paper,the control algorithm of neutral point clamped three-level inverter is designed firstly.Starting from carrier modulation,the influence of three-phase duty ratio on neutral point average current is explored,and the size and variation form of disturbance term are analyzed.It is proved that the neutral point voltage imbalance is the form of third harmonic oscillation,and the neutral point potential balance and control are realized by modifying the disturbance term.Then,the sliding mode observer is designed to estimate the load torque of the steering motor.Through the combination of the two error dynamics systems,the sign relationship of the error is transferred to the third error dynamics system,so that the error dynamics of the designed sliding mode observer has the asymptotic convergence in finite time.Then,the vehicle lateral dynamics model is extended in higher order.Through continuous sampling,the moment of inertia of vehicle,the lateral stiffness of front and rear wheels,and the distance between front and rear turns to the center of gravity of vehicle are estimated by using the least square method in a short time interval.In order to solve the influence of small change of matrix caused by sampling in a short time interval on the solution result of matrix,the first order extended least square method and the second order extended least square method are proposed,and control factors are introduced to actively adjust to obtain better estimation results.Finally,the simulation experiment is designed on MATLAB.The proposed three-level neutral point potential balance and control algorithm can eliminate the third harmonic oscillation of neutral point voltage,and has high-quality output current waveform.The total harmonic loss can be reduced to 0.67%.The sliding mode observer can estimate the load torque more accurately,and the load torque compensation in steering can effectively improve the tracking performance.Compared with the basic least squares,the problem of matrix solution error caused by continuous sampling in a short time is well solved.Compared with the existing estimation algorithm,the root mean square value of estimation error is reduced by 79.8%. |
