基于神经网络预测校核的暂态稳定预防控制

Preventive Transient Stability Control Based on Neural Network Security Predictor

针对电力系统稳定预防控制在线计算的复杂性,提出一种基于神经网络预测校核的在线预防控制决策方法。利用神经网络构建有功出力控制变量与暂稳TSI指标之间的映射关系,基于发电机有功输入实现数据驱动下的暂态稳定快速预测,取代基于微分代数方程的求解。在预防控制最优潮流模型中,嵌入离线训练的神经网络暂稳预测模型作为稳定性约束。由于在适应度函数计算中引入神经网络,利用神经网络的快速预测取代传统方法中的时域仿真校验计算,结合智能算法迭代求解时既实现预想故障集下的暂态稳定性校核,又保证较高的求解效率以满足在线计算的需求。模型最终求解结果可作为有效的在线预防控制策略,以保证故障发生后系统的稳定性。最后通过新英格兰39节点测试算例,验证了方法的可行性和有效性。

Aiming at complexity of pbwer system transient stability computation,an online preventive control method embedded with neural network-based prediction model is proposed.Mapping relationship between active power variables and transient stability index,built with neural network,is used to determine transient stability of power system instead of solving differential-algebraic equations.An off-line trained neural network model for transient stability prediction is embedded in the optimal power flow model as stability constraint.Since the fitness function calculation is added with neural network prediction model,the transient stability of power system under different contingencies can be fast verified instead of performing time domain simulation during iterative solution of intelligent algorithm.Its high solving efficiency meets requirements of online computation.The proposed method eventually provides effective preventive control strategy for operators to maintain power system stability after disturbances.Case study on New England39-bus system demonstrates feasibility of the proposed method.