基于DRNN自整定准PR控制的光伏并网系统研究

发布时间：2018-07-25 11:45

【摘要】：以配置LCL滤波器的单相光伏并网逆变系统为研究对象,针对其非线性、时变特性,提出一种基于准比例谐振(quasi proportional resonant,QPR)控制的直流电压外环、并网电流内环和电容电流内环三闭环控制策略,并给出准PR控制器的离散化方法。针对准PR控制器参数固定无法适应电网参数变化,难以达到系统控制性能最优的不足,提出一种基于动态对角递归神经网络(diagonal recurrent neual network,DRNN)自整定准PR控制器并给出其控制算法,DRNN采用基于二阶梯度的递推预报误差算法(recursive prediction error,RPE),其较一阶梯度反向传播算法(propagation,BP)具有更快收敛速度。仿真对比分析和实验结果表明,采用上述控制策略的单相光伏并网逆变系统具有良好的输出电流质量和快速的动态响应性能,可有效避免并网电流谐振并实现并网电流高功率因数。
[Abstract]:Based on the nonlinear and time-varying characteristics of single-phase photovoltaic grid-connected inverter system with LCL filter, a DC voltage outer loop based on quasi-proportional resonance (quasi proportional resonance is proposed. The three-loop control strategy of grid-connected current inner loop and capacitive current inner loop is presented, and the discretization method of quasi-PR controller is given. The fixed parameters of the quasi-PR controller can not adapt to the change of the power network parameters, so it is difficult to achieve the optimal control performance of the system. A self-tuning quasi-PR controller based on dynamic diagonal recurrent neural network (diagonal recurrent neual neural network) is proposed and its control algorithm is presented. The recursive prediction error algorithm based on second-order gradient, (recursive prediction error-error algorithm, is presented, which is better than that of the first order gradient back-propagation algorithm (Propagation BP). Faster convergence rate. The simulation and experimental results show that the single-phase photovoltaic grid-connected inverter system with the above control strategy has good output current quality and fast dynamic response performance. It can avoid grid-connected current resonance and realize high power factor of grid-connected current.
【作者单位】：北华大学电气信息工程学院;国网湖北省电力公司荆州供电公司;
【分类号】：TM615

【相似文献】