Jamil Harun, Qayyum Faiza, Iqbal Naeem, Kim Do-Hyeun
Department of Electronics Engineering, Jeju National University, Jejusi 63243, Korea.
Department of Computer Engineering, Jeju National University, Jejusi 63243, Korea.
Sensors (Basel). 2022 Aug 25;22(17):6402. doi: 10.3390/s22176402.
The shift of the world in the past two decades towards renewable energy (RES), due to the continuously decreasing fossil fuel reserves and their bad impact on the environment, has attracted researchers all around the world to improve the efficiency of RES and eliminate problems that arise at the point of common coupling (PCC). Harmonics and un-balance in 3-phase voltages because of dynamic and nonlinear loads cause a lagging power factor due to inductive load, active power losses, and instability at the point of common coupling. This also happens due to a lack of system inertia in micro-grids. Passive filters are used to eliminate harmonics at both the electrical converter's input and output sides and improve the system's power factor. A Synchronous Reference Frame (SRF) control method is used to overcome the problem related to grid synchronization. The sine pulse width modulation (SPWM) technique provides gating signals to the switches of the multilevel inverter. A multi-layer feed forward neural network (ML-FFNN) is employed at the output of a system to minimize mean square error (MSE) by removing the errors between target voltages and reference voltages produced at the output of a trained model. Simulations were performed using MATLAB Simulink to highlight the significance of the proposed research study. The simulation results show that our proposed intelligent control scheme used for the suppression of harmonics compensated for reactive power more effectively than the SRF-based control methods. The simulation-based results confirm that the proposed ML-FFNN-based harmonic and reactive power control technique performs 0.752 better in terms of MAE, 0.52 for the case of MSE, and 0.222 when evaluating based on the RMSE.
在过去二十年中，由于化石燃料储备持续减少及其对环境的负面影响，世界正朝着可再生能源（RES）转变，这吸引了世界各地的研究人员提高可再生能源的效率，并消除公共耦合点（PCC）出现的问题。动态和非线性负载导致的三相电压谐波和不平衡，由于感性负载、有功功率损耗以及公共耦合点的不稳定，会造成功率因数滞后。这在微电网中也会由于缺乏系统惯性而发生。无源滤波器用于消除电力转换器输入和输出侧的谐波，并提高系统的功率因数。同步参考框架（SRF）控制方法用于克服与电网同步相关的问题。正弦脉宽调制（SPWM）技术为多电平逆变器的开关提供门控信号。在系统输出端采用多层前馈神经网络（ML-FFNN），通过消除训练模型输出产生的目标电压和参考电压之间的误差，使均方误差（MSE）最小化。使用MATLAB Simulink进行了仿真，以突出所提出研究的重要性。仿真结果表明，我们提出的用于抑制谐波的智能控制方案比基于SRF的控制方法更有效地补偿了无功功率。基于仿真的结果证实，所提出的基于ML-FFNN的谐波和无功功率控制技术在平均绝对误差（MAE）方面表现更好，为0.752，在均方误差（MSE）情况下为0.52，基于均方根误差（RMSE）评估时为0.222。
