Harmonic Current Detection And Tracking Control Of Active Power Filter

With the rapid development of world economy, harmonic pollution problems in the power system has been increasing severe by the use of a large scale of nonlinear loads. Meanwhile, utilization of new technologies such as smart grid, distribution generation and renewable energy integration causes various types of power quality issues. These increasing complexities of power quality problems pose new challenges to solutions of harmonic suppression and other power quality issues, and require development of new power quality management products that can-provide considerable and equivalent functions.Harmonic suppression of power system is one of important measures for power quality management. Active power filter (APF) which adopts active compensation way can compensate harmonics components in real-time when frequency and amplitude vary with time. It doesn’t need energy storage devices and its compensation performance would not be affected by power grid impedance. Therefore the APF technology has received much attention recently and numerous related research results have been reported. Further research on the performance improvement of APF and its adaptability to the engineering actual situation is required. A three-phase and three-wire shunt active power filter (SAPF) prototype is studied and analyzed in the dissertation to improve the capabilities of harmonic current detection, robustness of compensatory current tracking and performance of harmonic suppression.In order to improve the detection accuracy and convergence speed of harmonic current detection method based on instantaneous reactive power theory, a novel algorithm of LMS/LMF adaptive filter is proposed, which combines a LMS algorithm with a LMF algorithm. The new algorithm can adjust self-adaptively the proportions of LMS and LMF in weight update. When the input sign of the adaptive filter change suddenly, the algorithm is mainly adopted by LMF algorithm. As the input sign of the adaptive filter is in steady state, the algorithm is mainly adopted by LMS algorithm. So, the LMS/LMF algorithm integrates the merits of both fast convergence speed of LMF algorithm and high stability precision of LMS algorithm. The convergence condition, time constant and misadjustment formula of the proposed algorithm are derived. The simulation and experiment results show that the proposed method can make the dynamic response of the low pass filter faster and its stability accuracy better, thus improve dynamic and steady performances of the harmonic current detection method based on instantaneous reactive power theory.To improve compensatory current tracking performance of SAPF based on double hysteresis current control strategy of voltage space vector, a novel4-level α-β hysteresis comparator is proposed. The control strategy is directly implemented on α-β coordinates. The4-level α-β hysteresis comparators are input by current error vectors and output4state values to determine the location region of current error vectors. The location region of the reference voltage vector can be determined rapidly by the approximate value of the differential reference currents. The optimum output switching vector at each instant is selected according to state values output of two4-level α-β hysteresis comparators and reference voltage vector. The compensatory current error is limited within the hysteresis width. The proposed control method can efficiently increase the utilization of DC voltage, decrease IGBTs switching number and achieve fast response in SAPF.To improve compensatory current tracking performance of SAPF, a linearized and decoupling control method is proposed based on neural network inverse (NNI). Firstly, the mathematical model of SAPF is derived and its reversibility is proved. Then the NNI of SAPF is designed, which combines with original system of SAPF to construct a pseudolinear system, the pseudolinear system can be simply controlled by a proportional plus integral controller. The Back Propagation (BP) neural network and Radial Basis Function (RBF) neural network are adopted in the proposed control method, which makes compensatory current tracking control linearized and decoupled, and it outperforms feedback linearization method. Meanwhile, it is proved that the inverse system control method which is adopted by BP neural network exhibits strong robustness when the system parameters change dramatically.In order to combine the control of the compensatory currents and the DC voltage of SAPF, and make SAPF more robust, a unified control strategy based on sliding mode variable structure control is proposed. At first, according to the mathematical model of SAPF in dq reference frame, the state variables id, iq, Udc is determined, and the linear combinations of the three state variables errors are used to define the sliding surface function of the unified control strategy. The control law is designed, and saturation function substitutes for sign function to eliminate chattering. Then the stabilization conditions of sliding mode and existence of equivalent control are derived. It is proved that the unified control strategy of SAPF can be achieved through suitable sliding surface function, and the K values selection of sliding surface function has to find a compromise between the stabilization and dynamic performance of sliding mode variable structure control. SAPF can work properly and be more robust when the load and device parameters change within a certain range.