The Research On The System Of The AC Asynchronous Electrical Dynamometer And Its Soft Sensor Model

The dynamometer is an important equipment in the power testing system for the drive machine and transmission machine. The AC asynchronous electrical dynamometer with squirrel-cage rotor is used widely since it has a simple structure, good precision, widely working range and good control performance. Under the control of transducer, the AC asynchronous electrical dynamometer can not only proved the load torque for power testing, but also drag the machine when it need cold running-in. Because the output power of the AC asynchronous electrical dynamometer can be feedback to the power grid through inverter feedback, the energy feedback for the system is realized and the electric circle is achieved easily. This paper studies on the system of the AC asynchronous electrical dynamometer which include constitute of the system, the control strategy for the system and the soft sensor technical in the system.First, this paper discuses the operation principle and stability condition of the AC asynchronous electrical dynamometer based on a motor engine test bed that developed by the author. After a simple math model for the transducer is proposed, the math model for the whole system is build. To the control strategy for the test bed, the constantly speed controller base on fuzzy PID is put forward in order to resolve the shortcoming of tradition PID controller. The result of experiment proves the system with the constantly speed controller has good adjust ability that can satisfy the requirements of the power testing.Second, a soft sensor model for the system of the AC asynchronous electrical dynamometer that can be used in the occasion where the speed is known is proposed. This model is based on an unknown input observer which is deduced form the model of the AC asynchronous electrical dynamometer. In this unknown input observer, the load torque is considered as an unknown input. The stability of the observer is proved by the Lypunov theory. To the system that the speed is unknown, a soft sensor model based on the unscented Kalman filter (UKF) is put forward. The simulation and the experiment prove that both the soft sensor model can estimate the quantity and has several errors due to their sensitive to the influence of the parameters and losses of the AC asynchronous electrical dynamometer.In order to overcome the parameters influence and losses, a soft sensor model based on the T-S fuzzy model is proposed. The model is build based on the fact that the load torque of the AC asynchronous electrical dynamometer is rising with the rising of the stator's temperature. The inputs of the model are the virtual value of the stator's voltage, the stator's current and the stator's temperature. The virtual value of the stator's voltage and the stator's current are computed by the improved fast Fourier method. In order to improve the precision of the model, the linear style in the tradition fuzzy rule is changed into the poly style and an adjustment for the clustering and membership degree is put forward based on the error of clustering's trivial. The experiments prove that the proposed model is effective. It can estimate the speed and the torque which has no sensitive to the parameters influence and losses when the temperature rising.In order to improve the estimate precision of the speed and torque farther, the normal function forms for the speed and the torque which consider the losses are studied. A soft sensor model based on these function forms using the Support Vector regression (SVR) which can be used to approximate nonlinear function in the whole region is put forward. The experiment prove that the precision of estimated is development than the soft sensor model based on the T-S fuzzy model. In order to development computer velocity and enhance the robust of the SVR soft sensor model, a nonlinear robust partial least-squares algorithm (NIRPLS) is proposed. The NIRPLS include the partial least-squares algorithm, SVR and the robust principal components regression based on principal sensitivity vectors (RPPSV). Since the SVR is used to realize the nonlinear map between the inner variable, the NIRPLS can be used in the AC asynchronous electrical dynamometer system which contains many nonlinear relations. The RPPSV in the NIRPLS is used to detect the outlier effectively. The computer velocity for the NIRPLS is development because the dimension of the input is decreased by picking up inner variable. The training result and the experiment prove that the soft sensor model based on the NIRPLS has good ability to estimate the speed and the torque of the AC asynchronous electrical dynamometer system. It also has good robust performance.At the last in this paper, the applied foreground of each soft sensor model is discussed according to the national standards. The facts which lead the errors are analyzed also. For the further research, the method to decrease the errors is proposed.