改进BP神经网络在负荷动静比例确定中的应用

APPLICATION OF IMPROVED BACK PROPAGATION NEURAL NETWORK FOR IDENTIFICATION OF THE PERCENTAGE OF DYNAMIC COMPONENT IN COMPOSITE LOAD

指出了电力系统负荷动、静组成比例在实际电力系统分析、计算中的重要性,并应用一种改进的BP算法--Levenberg-Marpuardt 反向传播算法来对神经网络进行训练,进而利用人工神经网络(ANN)来确定电力系统综合负荷动、静组成比例为β=F[Y(t),Y(t-1),…,Y(t-n),U(t),U(t-1),…,U(t-n)].其中,β为动态负荷在综合负荷中所占的比例,Y=[P,Q]T,U=[V,f]T.该算法改进了BP神经网络学习速度慢的缺点.应用该方法对仿真数据、动模实验数据和现场实测数据进行了测算,得出了其相应的动、静组成比例.测算结果验证了在确定负荷动、静比例时可以忽略频率的变化,证明了BP神经网络用于确定负荷动、静组成比例的有效性.

The percentage of dynamic load component (PDLC) is of great significance in power system dynamics analysis. The paper proposes an artificial neural network (ANN) based approach to determine PDLC through field collected data. Levenberg-Marquardt which is a kind of improved back propagation (BP) algorithm is adopted to train ANN. PDLC is obtained from the nonlinear mapping function F'(Y(t), Y(t-1),…, Y(t-n); U(t), U(t-1),…, U(t-n)) which is formed by the well trained ANN, where Y=[p,Q]~T, U=[V,f]~T. The approach can eliminate the low-train-speed disadvantage of BP networks. Several validations are conducted to identify PDLCs from simulation data, dynamics test data and field data. That small frequency deviation can be neglected is revealed from the results. And the effectiveness of the approach to determine PDLC is well-proved.