【作者】 王嵩；

【导师】 肖建；

【作者基本信息】 西南交通大学 ， 电气系统控制与信息技术， 2013， 博士

【摘要】 电力牵引系统是铁路机车的动力来源,为满足铁路系统的发展,目前机车电力牵引采用高性能的交流调速技术,因此高性能的矢量控制方法及直接转矩控制方法成为电力牵引传动系统的首选。感应电机无速度传感器控制增加了系统的鲁棒性并能降低成本。电力牵引控制系统目前需解决的关键问题有粘着控制、牵引电机实时的状态估计及重要参数的实时辨识等。不同于传统的数字系统中对所有信号采用同步等间距采样的策略,多采样率系统中可以对系统中不同信号采用不同的采样策略。输入多采样率控制系统通过增加输入采样速率,增加了数字控制器的控制能力,可以实现许多单采样系统不具备的控制功能；输出多采样率使得控制器能获得更多被控对象的信息,使控制系统有更强大的控制能力。随着电机运行工况的不同,电机参数值会发生变化。转子时间常数是矢量控制感应电机系统中最敏感的重要参数,同时其准确性在很大程度上影响牵引控制系统的性能。针对电机的转子时间常数难以测量且随时间变化的情况,利用多采样率控制系统具有传统单采样率控制系统更强的控制性能等特点,将多采样率方法与MRAS方法相结合,提出了基于多采样率MRAS的转子时间常数参数辨识方法。在硬件在环试验系统中进行实时实验,并与传统MRAS的参数辨识结果进行了对比。结果表明,多采样率MRAS能实时对参数进行辨识,得到了满意的效果。感应电机状态估计的精度直接影响电机系统性能,同时状态估计值中包含的噪声会影响电机系统正常运行。针对这个问题,将多采样率控制理论与传统的扩展Kalman滤波(EKF)状态估计方法结合,提出了电机状态估计的多采样率EKF方法,分别得到了输入多采样率EKF方法及输出多采样率EKF方法。通过硬件在环试验系统进行了试验,通过对比输入多采样率EKF与不同帧周期下EKF算法对于感应电机状态估计的性能,表明多采样率EKF方法是一种适合实时应用的电机状态估计方法,同时算法能有效消除噪声、提高状态估计性能,并能降低成本。相对于EKF方法,强跟踪滤波方法(STF)具有更强的关于模型不确定性的鲁棒性和关于突变状态的跟踪能力的特点。针对机车电力牵引系统中牵引电机负载转矩的变换较快的问题,提出了基于多采样率STF的状态估计方法。通过扩展建立的电力机车单轴单电机模型,得到了与真实机车对应的多轴多电机的电力机车模型,通过与机车真实空转特性的对比验证了模型准确性。通过建立的电力机车多轴模型进行仿真试验,采用多采样率STF方法对发生空转的轮对的牵引电机进行状态估计,得到了良好的状态估计性能。电力机车的牵引和制动依赖于其轮轨间的粘着,若牵引力或制动力大于轮轨间可用粘着力,则会出现空转或滑行,制约牵引力或制动力的发挥,同时也会造成轮轨擦伤、轮箍发热损坏、甚至脱轨等安全隐患。电力机车采用粘着控制的方法提高机车的粘着力,而粘着控制的前提是空转或滑行状态的快速识别。为得到及时准确且迅速的空转识别,基于电力牵引系统中电气量响应速度远快于其机械量的思想,针对目前广泛用于电力机车粘着控制系统中的利用轮轴转速等机械量识别空转的方法的不足,首次提出了完全基于电气量的空转识别方法。通过某型电力机车反复空转过程中采集的大量数据和多采样率EKF牵引电机负载转矩方法,建立了基于电气量的空转识别规则。通过建立的模型进行了试验,结果表明：与传统的利用转速等机械量空转识别方法相比,该规则的应用大大降低了空转识别时间,避免了因空转导致的牵引力大幅丢失,提高了粘着控制系统的性能和电力机车的牵引效率。更多还原

【Abstract】 Electric traction system is the power source for railway locomotives. The locomotive electric traction adopts high performance AC speed adjustment technology to meet the development of the railway system, thus the high performance field-oriented control (FOC) and direct torque control are the first choices of electric traction drive system. Speed sensor-less control of induction motor promotes the system robustness and reduces system cost. The key issues to be resolved of electric traction control system including:adhesion control, real-time state estimation of traction motor, real-time identification of important parameters.Different from the traditional digital system adopting a single sampling rate for all different signals, multi-rate system adopts different sampling strategies for different signals. Input multi-rate system increases the sampling speed of input, thus ability to control the system is improved, whereas the system could realize many control functions which single-rate system cannot. Output multi-rate system enables the controller to get more information of controlled object, thus system has a better control ability.Motor parameters may vary during motor operation. Among these parameters, rotor time constant is the most sensitive variable in FOC induction motor system. Its identification accuracy greatly affects the performance of traction control system. In view of the rotor time constant is time varying and difficult to measure, Using the principle that multi-rate control system has stronger control performance than traditional single-rate control system, combining multi-rate control method and model reference adaptive system (MRAS) method, the multi-rate MRAS parameter estimation method is proposed. Using the experiment set-up to validate the method, the comparison of parameter identification performance between multi-rate MRAS and traditional MRAS verified the proposed method is real-time and the proposed method has satisfactory dynamic and static estimation performance.Induction motor state estimation accuracy directly affect the performance of the system, at the same time the noise contained in state estimated value will affect the normal operation of motor system. Aiming at this problem, combining the extended Kalman filter (EKF) method and multi-rate control theory, the multi-rate EKF algorithm including input and output algorithms is proposed. A Hardware-in-the-loop (HIL) experiment set-up is proposed to validate the proposed method, comparing the state estimation performance between input multi-rate EKF method and traditional EKF methods under different frame period, the comparison verified the multi-rate EKF induction motor state estimation method is real-time, the method can effectively eliminate noise and improve the performance of state estimation and reduce the cost.Comparing with the EKF method, strong tracking filter (STF) has better robustness on model uncertainty and better tracking ability about abrupt state changes. In electric traction drive system, the load torque of traction motor changes quickly. Aiming at this problem, the multi-rate STF algorithm is proposed. By extending the single-axle locomotive model, a multi-axle model is proposed to simulate the real locomotive operation, the comparison between the model and actual slipping features verified the accuracy of the model. Using the proposed multi-axle model to validate the method, the method is applied to estimate the state of traction motor which is connected to the wheel-sets occur slipping phenomena, the result show that the multi-rate STF method has satisfactory estimation performance.The traction and braking of the electric locomotive is realized by adhesion force between wheel and rail. If the traction/braking force is greater than the available adhesion force between wheel and rail, the wheel-set will slipping/skidding, restricting the utilization of traction/braking force. It also will cause safety hidden trouble such as abrasion of wheel and rail, damage of wheel rim, even train derailment. Adhesion control method is adopted to improve the locomotive adhesion force. The slip/skid detection is the precondition of the adhesion control. To get accuracy rapid and effective slip detection, based on the principle that the response speed of electrical quantities is much faster than mechanical quantities in electric traction system, a slip detection method only based on electrical quantities is firstly proposed. The method is aiming to solve the disadvantage of present widely used adhesion control system in locomotive which is based on mechanical quantities such as wheel axle rotation speed. The electrical quantities based slip phenomena detection rules are designed based on the actual data obtained from the HXD2locomotive and the traction motor load torque multi-rate EKF state estimation method. Using the proposed model to validate the rules, the result show that the rules greatly reduced the slip detection time, avoid the severe loss of traction force, give full play to the locomotive traction power, and improve the overall performance and traction efficiency of the locomotive.更多还原