Research On Electrical Vehicle Induction Motor Parameters Estimation Based On Data Driven Method

The vector control of induction motors is widely used in industrial applications due to its reliability and fast response.Especially in recent years,indirect field orient control(IFOC)of induction motor has become more and more popular in the field of electric vehicles.Because IFOC adopts the torque control using the slip angle frequency,it is essentially a feedforward control method.The parameters of the motor will lead to the misalignment of the rotor flux linkage of the induction motor,which will seriously affect the output torque and output efficiency of the motor.Therefore,it is necessary to identify the motor parameters.The most of existing induction motor parameters estimation methods are mostly model-based methods,which have the feature of that first establish the system model of the motor,and then use the modern control method to estimate the parameters of the system model,including no-load and locked rotor method,model reference adaptation system method,extended Kalman filter method,sliding mode observer method,etc.The model-based parameter estimation method is not computationally cost and easy to implement,so it is suitable for the real-time motor control application.However,there are some problems in the model-based parameter estimation method.For example,such methods can only identify the parameters of the system model,the estimation accuracy is poor when the model is inaccurate,and the anti-disturbance is low in some extreme cases.Considering the above issues,this paper studies a new idea that is based on the premise of not establishing the system model,but completely relies on the real data to estimate the parameters of the electric motor induction motor,which is called the data-driven electric vehicle induction motor parameter estimation method.Since the data-driven parameter estimation method does not depend on the system model,the shortcoming of the model-based parameter estimation method can be overcome.At the same time,the data-driven parameter estimation method is a"reward"-driven method.Therefore,the estimated parameters are not the parameters of the motor system model,but some parameters that can make the motor run in optimal condition.This is the most significant feature that the data-driven methods are different from the model-based ones,and it is also an important reason why the data-driven method can be applied to the motor control of electric vehicles.Based on the above analysis,the contents of this paper are as follows:1.Firstly,several classical model-driven induction motor parameter estimation methods are reviewed,and their advantages and disadvantages are analyzed.2.An off-line data-driven parameters estimation method for electric vehicle induction motor based on Deep-Q-Learning is studied.The advantage of this method is that the estimated parameters can make the motor output optimal torque and maximum efficiency under any condition.At the same time,the method has high robustness to noise.3.An off-line data-driven parameter estimation method for induction motor induction motor based on Deep Deterministic Policy Gradient is studied.The parameters estimated by this method also have the feature that the motor can output optimal torque and maximum efficiency under any condition.At the same time,the parameter estimation accuracy is higher than the deep-Q-learning method.4.Aiming at the shortcoming that the above method can only be applied to offline parameter estimation,a data-driven online electric vehicle induction motor parameter estimation method based on deep reinforcement learning and random forest combination is studied.The advantage of this method is that the algorithm is divided into offline and the online stage,while the online stage is computationally small,which is suitable for real-time online running of the motor.At the same time,the parameters estimated by this method can output optimal torque and optimal efficiency at any condition when the motor is running in practice.5.In the data-driven electric vehicle induction motor parameter estimation method,the quality of the collected data directly affects the accuracy of the estimation result and the convergence speed of the algorithm.Due to the complicated operating conditions of electric motor motors,various outliers are often generated in actual operation.Therefore,the cleaning of data,especially the detection and rejection of outliers is particularly important.To address this essiue,this paper studies a method of outliers detecting method called virtual nodes isolation forest,which can effectively detect and eliminate outliers in motor operation,ensuring the purity of data source.