基于GA-BP算法神经网络的电力系统稳定器研究

Research of Power System Stabilizer Based on GA-BP Neural Network

为解决常规电力系统稳定器中受控系统参数难以识别的问题,提出了一种可以根据电力系统运行工况自动调整控制参数的新型自适应电力系统稳定器(PSS),运用神经网络完成电力系统被控模型的精确在线辨识。通过在线测量同步发电机的有功和无功参数,自适应电力系统稳定器可按照相位超前补偿的设计原则实时自动调整稳定器参数,以达到最佳的抑制低频振荡的效果。为了避免BP算法在神经网络训练当中陷入局部小等一系列缺点,引入了一种GA和BP相结合的算法,将其用于人工神经网络的设计,同时将离线计算所得的PSS参数构成的样本对神经网络进行训练。通过在时域仿真结果表明自适应PSS能有效抑制电力系统低频振荡,极大的提高电力系统的动态和暂态稳定性。

The design and implementation of a neural- network- based power system stabilizer （PSS） are presented. Through measuring the parameters of active power and reactive power of the synchronous generator on - line, the parameters of PSS tuning can be achieved according to the environment in which it works through artificial neural network （ANN）, A hybrid of genetic algorithm （GA） and back - propagation （BP） algorithm is introduced which can find the correct global minima without getting stuck at local minima, and genetic algorithm is adopted to optimize the initial value in ANN. BP algorithm is utilized to search in local part and fast gets the matrix of the weight value and the threshold, The time domain simulation results indicate that the proposed adaptive FSS can damp low - frequency useillatious effectively over a wide range of operating conditions and significantly improve the dynamic and transient stability.