An adaptive off-line neural network inverse control system and method for an energy-feeding suspension

Abstract

The invention discloses an adaptive off-line neural network inverse control system and method for an energy-feeding suspension. The system has four input ports and one output port, including an off-line neural network inverse controller, an adaptive controller, and an adaptive algorithm. , pulse width modulator, adder, comparator and time delay module. Among them: the input signals of the four input ports are the terminal voltage of the energy storage element respectively U s, the actual output current of the electromagnetic damper i real and working speed v , and the given current of the electromagnetic damper i ref, the output signal of the output port is a pulse signal for controlling the DC-DC converter. The invention adopts the off-line neural network to inversely provide the steady-state control input required by the energy-feeding suspension, and the disturbance caused by system parameter perturbation and external uncertainty disturbance is compensated by the adaptive controller, which improves the robustness of the control system The anti-interference ability of the system and the system speed up the tracking of the actual current of the winding of the electromagnetic damper to the ideal given current.

Description

technical field [0001] The invention relates to an adaptive off-line neural network inverse control method of an energy-feeding suspension, belonging to the technical field of intelligent control of vehicle suspension systems. Background technique [0002] The energy-feeding suspension can recover the vibration energy generated by the suspension system due to the excitation of the road surface, which reduces the energy consumption of the suspension system and has a good application prospect. At present, the research focus of the energy-feeding suspension is on how to store the low energy-feeding voltage and coordinate the energy-feeding performance and vibration isolation performance of the suspension. Therefore, the design of the energy-feeding loop in the energy-feeding suspension system is very important. [0003] The DC-DC converter can be effectively applied to the design of the energy feeding circuit due to its advantages of realizing power conversion and adjusting the...

AI Technical Summary

Problems solved by technology

However, since the body, tires, etc. are themselves complex nonlinear elements, the electromagnetic damper and DC-DC converter in the energy-feeding suspension also have strong nonlinearity and time-varying properties, and the system parameters are often inaccurate and variable. , such as the nonlinear attenuation and fluctuation interference of the input voltage, the neglect of the capacitance, inductance and register resistance, and the uncertainty of the load will easily cause the mismatch between the controller parameters and the actual parameters, which will lead to the deterioration of the system performance; at the same time, the energy-feed suspension system The input is affected by the random excitation of the road surface, its working process is disturbed by the outside world, and its structural parameters will be perturbed to varying degrees. Therefore, when the DC-DC converter is controlled in real time, the dynamic response characteristics and resistance of the control system Interference ability requirements are particularly strict, the control system must be able to respond accurately and quickly, and be able to adaptively adjust the controller according to changes in input voltage and load, in order to ensure that the system has good dynamic response performance and robustness

[0004] Traditional DC-DC converter control methods rely on the explicit description of its mathematical model, such as sliding mode control, self-tuning control, robust control, etc., while the DC-DC converter is a high-order, strongly nonlinear, time-sensitive Variable or intermittent discrete circuits, so it is difficult to analyze them analytically with precise mathematical models

Images