Research On Several Key Issues Of High Performance Hybrid Shunt Active Power Filter

The improvement of power quality,which has great significance in engineering practice,is one of the key issues in the electrical engineering research.On one hand,the widely use of power electronics provides a high speed,high efficiency and energy-saving control method for industrial equipment,on the other hand,the widely applied power electronic devices inevitably brought power quality problemsandcause unnecessary economic losses.In recent decades,power electronics and computer technology are getting more and more mature,which make the active power filter one of the most important methods for power quality adjustment and control in the future.This dissertation mainly researches from three aspects,which includes harmonic current detection and prediction methods,the current tracking control and the low loss technology for active power filter.The study will lay a foundation for the promotion of practicality of shunt hybrid active power filter.In order to improve the precision and dynamic response characteristics of harmonic current detection and acquire a better compensation performance,a harmonic current detection algorithm based on variable forgetting factor recursive least-squares(RLS)algorithm is proposed.The occurrence of the load current dynamic process is identifiedby the judgment condition which is 'given by the algorithm,and then the forgetting factor is assigned dynamically.Hence the convergent speed is significantly improved.The algorithm not only overcomes the detrimental impact of low-pass filters in p-q and ip-iq algorithms on dynamic response,but also alleviates the conflicting requirements on forgetting factor value between steady process and dynamic process in traditional RLS algorithm.By analyzing the effect of the delay of each components of the digital active power filter system on the harmonic compensation,the necessity of the harmonic prediction is obtained.The problem of the traditional LMS harmonic current prediction algorithm is constant step size,which can not reconcile the contradiction between the requirements of the convergence speed and steady-state error,a variable step size LMS harmonic current prediction algorithm is proposed to realize the step size and prediction model weights updating,in which,the step size varies with the prediction error,and the system internal noise and the timeliness of old errors are also considered.The improved deadbeat harmonic current tracking control proposed in this paperneed to predict of harmonic reference current in advance 2 steps.In contrast,previous prediction methods are usually repeated prediction or extrapolation polynomial method.However,,if the load is changing suddenly when the repeated prediction method is adopt,instruction data will have a great deviation.Longer time is needed to reach a new balance.As aresult,the dynamic characteristic is comparatively poor.When the extrapolation polynomial method is adopt,the prediction accuracy depends on the order,the higher the order,more accurately it predicts,the larger of calculation amount however.So it is difficult to select a proper order to reconcile the contradiction.A mixed prediction method is proposed,which combines two kinds of prediction methods,in which repeated prediction is adopt,when the load is stable,identifying the dynamic change of the load by a certain set of conditions the adaptive linear prediction will be effective,so that both advantages can be played.An improved dead-beat control algorithm is proposed in this work to slove the problem that the conventional dead-beat control is actually a one-step-delay control affected by the signal sampling,processing and executing.The algorithm will predict the active filter current reference two sample instants in advance with respect to the measurement instant and the inverter output current one sample instant in advance,and then the "real" dead-beat control will be achieved.The prediction of current reference adopt the above mixed prediction strategy.The mathematical model for a hybrid active power filter(HAPF)with series resonant branch,which includes the passive part and an active part,is established firstly,.After reducing the order of the model,a Lyapunov function was constructed,and then an appropriate control law was selected to make its leading function is negative definite,so that the steady tracking for harmonic current was realized.From the robustness and tracking performance perspective,the turning range and the optimization goal of the control gain was given by derivation.An auto-disturbance-rejection proportional integral(ADR-PI)control is adopted in outer DC bus voltage control loop which can efficiently enhance the ability to resist load disturbance.On the basis of the analysis of the switching loss of active power filter,two kinds of low loss strategies are studied,which are the crossover compensation strategy and the DPWM modulation strategy.A new topology is proposed,which comprised of a low frequency inverter unit and a high frequency inverter unit,to realized dividing frequency interleaved compensation of harmonic current.In the strategy,the low frequency inverter unit with low switching frequency,is in charge of compensation of low order components of load harmonic current,reactive current and maintaining the DC side voltage.Meanwhile,high frequency load compensation with high switching frequency is in charge of compensation of high order components of load harmonic current.A harmonic dividing frequency given method based on sliding DFT is studied which can separate the above mentioned components.An improved DPWM modulation strategy is proposed aiming at the characteristics of output current in the APF application.The simulation results show that the switching loss of the proposed algorithm is superior to that of hysteresis modulation and SVPWM modulation.In order to verify above-mentioned algorithms and strategies,an APF experimental platform is designed and constructed.Firstly,the design of main circuit is accomplished,including power devices and the driver circuits selection and design,analysis and optimization design of inductance and capacitance.Secondly,the current and voltage sampling circuit and the PLL module are designed.Finally,the control system algorithms processing flow chart is drawn and translated into program with C language,and hardware and software debugging is also actualized.Based on the platform,experimental researches are carried out to verified the validity of harmonic current detection and prediction and its tracking control strategy.