一种基于堆叠门控循环单元的暂态稳定智能评估系统

A Stacked Gated Recurrent Unit Based Transient Stability Intelligent Evaluation System

随着大电网之间的互联,安全稳定问题日益突出。目前,基于数据驱动的电力系统自适应暂态稳定评估方法通过平衡时间和准确性在测试系统上取得了优异的效果,但是存在模型构建繁琐和参数复杂等问题。提出了一种基于堆叠门控循环单元(Stacked-GRU)的暂态稳定智能评估系统;构建了基于时间上参数共享和空间上堆叠的Stacked-GRU模型,利用电力系统故障后的时间序列数据进行离线训练,得到最优参数;在线应用时,从故障清除时刻开始,通过判断当前时刻的暂态稳定评估结果是否满足置信度,如不满足则将评估信息将继续提供给下一个时刻进行评估,直到评估结果满足置信度为止,从而实现自适应的暂态稳定智能评估系统。仿真结果表明,提出的模型具有评估时间短和准确度高的优点。

Along with the interconnection between large power grids,their safety and stability becomes increasingly prominent.Although excellent results have been achieved on the test system of the data-driven adaptive transient stability evaluation system by virtue of balance time and accuracy,there exist such problems as cumbersome model structure and complicated parameters.The intelligent transient stability evaluation system based on stacked gated recurrent unit(Stacked-GRU)presented in this paper established a Stacked-GRU model with parameter sharing in terms of time and stacking in space,whereby the time sequence data collected after power system fault was used in off-line training to obtain optimal parameters.In the case of online application,starting from the moment of fault clearance,it was judged whether the current transient stability evaluation result satisfied the confidence or not,otherwise,the evaluation information would continue to be provided to the next moment for evaluation until the evaluation result satisfied the confidence,so that the intelligent adaptive transient stability evaluation system could be achieved.Simulation results showed that the model presented in this paper had such advantages as short evaluation time and high accuracy.