基于GAN模型与随机森林算法的保护系统智能状态评价与预警

Smart status evaluation and early warning approach for highly-reliable protection systems based on GAN model and random forest algorithm

为提高继电保护系统状态评价的准确性,提出一种基于GAN模型与随机森林算法的智能状态评价方法。首先,结合现场情况与专家意见,建立系统状态指标集,并针对继保设备状态数据不平衡的问题,提出基于生成对抗网络的状态数据生成方法;然后,建立基于随机森林的继保系统综合评价模型;最后,结合设备的历次状态评价结果,给出设备的健康指数变化曲线及其劣化趋势,提供相应的状态预警。基于真实数据的实验结果表明,该方法能较准确评价系统状态,对合理安排检修周期、制定检修计划具有参考价值。

To improve the status evaluation accuracy of protection systems,a smart status evaluation method based on GAN model and random forest algorithm is proposed.Firstly,a system state indicator set is established in combination of the field conditions and expert opinions.To address the problem of the imbalance of relay protection equipment state data,a state data generation method is proposed based on the generation countermeasure network.Then,a comprehensive evaluation model of protection systems based on random forest is established.Finally,combining with the preceding state evaluation results,the health index curve of the equipment and its deterioration trend are given to provide corresponding state early-warning.The real-data experimental results show that this method can more accurately evaluate the system status,and has reference value for rationally arranging the maintenance cycle and formulating the maintenance plan.