Modeling Of Switched Reluctance Motor Based On Intelligent Algorithm In Small Sample

Switched reluctance motors have the advantages of simple structure,low cost,good speed regulation performance,good adaptability to harsh environments,etc.,and are widely used in electric vehicles,textile industry,household appliances and other fields.However,its bi-salient pole structure and magnetic saturation make it difficult to establish current and torque models,and these models are the basis of motor performance evaluation and control.Because of the nonlinearity of switched reluctance motors,numerical models of switched reluctance motors are generally established based on experimental or simulated data.The amount of data required for modeling often determines the time cost of modeling.Therefore,it is a research focus to study the current and torque modeling of switched reluctance motors in the case of small samples.Accurate modeling data is the cornerstone of building high-precision models.The paper introduces two methods for measuring the flux linkage value of switched reluctance motors,one is the rotor fixing method,and the other is the fast measurement method based on torque balance.After a series of post-processing,the fast measurement method can obtain accuracy similar to the rotor fixing method at several specific positions.The step-by-step modeling method is a modeling method based on a smart algorithm that uses a fast measurement method to obtain a small number of training samples and still works well.The step-by-step modeling method,like many current or torque modeling methods based on intelligent algorithms,treats the modeling problem as a black box,and there is a certain improvement interval.Therefore,this paper mainly studies to improve the model accuracy by increasing the prior knowledge.This article observes the flux linkage model based on BPNN and SVM,and finds that the model accuracy is high in the region of the straight line of ? with ? curve.Therefore,this paper proposes a mapping function based on the change law of ?-? curve to make the three-phase 6/4 motor i fixed The ?-? curve at [0 °,30 °] is basically a straight line.The accuracy of the model obtained by training the mapped data at [0 °,30 °] is significantly improved.In order to improve the accuracy of the flux linkage model in other angle intervals,this paper further proposes a mapping function from the perspective of balancing the linearity and nonlinearity between training samples.Using the model trained by the samples after the two functions are mapped,the mean square error(Mean Squared Error MSE)is reduced by tens of times compared with the original model.By applying the improved flux linkage model to the step-by-step modeling method,the accuracy of the step-by-step modeling method in measuring samples has been significantly improved.The step-by-step modeling method consists of a flux model based on SVM and a current model based on BPNN.Such a structure is more complicated.Based on the i-? curve,this paper analyzes the importance of training samples at different positions to establish a high-precision current model.According to the reason why the training data is important between(30 °,45 °),a mapping function is proposed.The changed rule of the i-? relationship after the mapping becomes simple,which reduces the difficulty of learning the current model.According to the BPNN training process,it is difficult to fit the region that increases rapidly with the increase of ?,and the activation function of some network layers is modified to an exponential function.If samples less than 15 ° are removed from the training set and test set,the accuracy of the current model based on BPNN is very high,indicating that the obtained model has learned that the spacing between adjacent ?-i curves in the [15 °,30 °] interval is basically the same One rule.Therefore,this paper proposes a mapping function based on this phenomenon combined with the improvement of the flux model,so that the training set and the test set with training samples smaller than 15 ° still maintain high accuracy.Combining the two methods of mapping function and modifying activation function,this paper builds a current model with similar accuracy to the step-by-step modeling method based on the samples of four specific positions that can be obtained by BPNN using the flux measurement method,but only one is based on BPNN Current model.Finally,based on the Simulink environment,the CCC control and APC control models of the switched reluctance motor are built.The experimental results further verify the effectiveness of the improved scheme.