Research On Application Of PI-Repeat Control Strategy In BP Neural Network Of Micro-grid SAPF

With the large investment in many power electronic devices in the micro-grid,the power quality of the micro-grid is seriously polluted,which causes serious pollution to the power quality of the micro-grid.Therefore,it is imperative to control harmonics and losses of the power grid.This paper introduces the basic principle and topology structure of SAPF in detail.To improve the ability of SAPF to suppress harmonics and compensate power quality in micro-grid operation,a novel micro-grid parallel APF topology is proposed and analyzed.The p-q and ip-iq detection methods propose a novel ip-iq detection method that uses a decoupled phase-locked loop circuit and a positive-sequence fundamental wave extractor in a synchronous coordinate system instead of a traditional phase-locked loop to obtain a sine and cosine signal.It not only solves the problem that the phase difference of the positive sequence fundamental wave signal is not generated when the extractor generates three-phase voltage unbalanced condition,but also has the problems of poor anti-interference ability,delay lag,and the like during the subsequent coordinate transformation.This method can accurately detect the harmonic current and greatly improve the detection accuracy.Through comparative analysis of the simulation results,the correctness of the improved method is verified.The current limiting requirements of the DC bus voltage side in the APF unit are introduced.When the non-linear load suddenly increases,a short circuit occurs or an open circuit fault occurs,an overcurrent fault and an overmodulation fault of the SAPF will result.Therefore,an improved current-limiting control method based on amplitude and phase correction is used.When the modulation voltage value exceeds half of the DC bus voltage(400V),the modulation voltage is cut off and kept within 400 V to achieve the system synthesis.Protective effects.Aiming at the problem that traditional PI control system can't adjust the PI control parameters online in real-time under the running state,it will lead to the problems of low control accuracy and insufficient stability.A BP neural network recursive integral PI-repetition suitable for active power filter is proposed.In the control strategy,the PID neural network is used to continuously learn the real-time tuning of the PI control parameters to obtain real-time RMS values;using the recursive integration function,M PI control can be simultaneously responded in one cycle,instead of each sampling point in each The point-by-point integration is performed during the corresponding period,which greatly improves the rapidity and accuracy of the tracking error,and parallels with the repeated control to improve the ability to track thesteady-state error to ensure the stability of the system operation.Compared with the traditional PI control,the optimized control strategy not only improves the response speed and compensation accuracy of the system operation,but also greatly improves the system stability.A corresponding experimental platform was built to verify the results of the experiment,data show that the improved detection and control strategy can realize grid connected APF regulating micro power quality function better.