适应波达时刻不确定性及较低采样率的新能源送出线路就地暂态量主保护判据

Local Transient-based Main Protection Criterion of New Energy Transmission Line Adapted to the Uncertainty of Arrival Moment and Low Sampling Rate

不对称接地故障占所有线路故障的90%以上,接地距离保护在应对此类故障方面发挥了不可替代的作用。随着新能源高比例渗透,各种传统单端工频量保护性能显著下降已成为共识。基于故障分量线模和零模波速差的保护判据理论上仅需利用到故障初始行波到达时刻信息,是一种原理简单可靠的单端量快速保护判据,已经在直流电网中成功实践。但在尝试将这类保护应用于交流电网时发现,受波头前陡较缓而难以精确定位波到时刻、依赖高采样率等诸多不利因素影响,存在过大的模糊判别区,除了特长线路外,对绝大部分线路几乎没有应用可行性。波到时刻的精准辨识是一个复杂的非线性问题,利用人工智能的方法进行辨识是一条可行的解决思路,对此,该文提出一种新的单端暂态量主保护判据。首先,分析波达时刻与波形关系,并指出这种关系能够采用机器学习来映射;其次,引入高斯过程回归(Gaussian process regression,GPR),在对初始行波数据进行预处理得到样本集后,输入GPR预测模型进行训练;然后,依据模型评估指标得到最优训练模型以输出高可信性的线-零模波达时差,据此实现了基于行波模量传输时间差的保护判据;最后,在利用PSCAD仿真验证所提保护判据有效性和普适性的基础上,进一步利用现场实测数据对判据进行测试,验证其实用性。该文工作为新能源交流系统下单端暂态量保护的性能提升提供新的解决思路。

Asymmetric ground fault accounts for more than 90%of all line faults,and ground distance protection plays an irreplaceable role in dealing with such faults.However,with the high penetration of new energy,it has become a consensus that the performance of various traditional single-ended power-frequency based protection has significantly declined.The protection criterion based on the line mode vs zero mode wave velocity difference of fault components only needs to use the arrival time information of the fault initial traveling wave theoretically,which allows it to become a simple and reliable single terminal quantity fast protection criterion.Until now,it has been successfully applied in the DC power grid.However,when trying to apply this type of protection to the AC power grid,it is discovered that numerous adverse factors,such as the steep wavefront,hinder accurate determination of the wave arrival time.The reliance on high sampling rates and other factors result in an excessively large fuzzy discrimination zone,rendering this protection unsuitable for most lines,except for extremely long ones.Accurately identifying the wave arrival time is a complex nonlinear problem,and employing artificial intelligence-based methods for identification presents a feasible solution..In order to solve the above problems,a new single-ended transient based main protection criterion is proposed in this paper.First,the relationship between the arrival time and waveform is analyzed,and it is pointed out that this relationship can be mapped by machine learning.Next,Gaussian process regression(GPR)is introduced,the sample set is obtained after preprocessing the initial traveling wave data,and the GPR prediction model is input for training.Then,according to the model evaluation index,the optimal training model is obtained to output the highly reliable linear zero mode wave time of arrival difference.By this means,the protection criterion based on traveling wave modulus transmission time difference can be realized.Finally,on the basis of PSCAD simulation tests,the validity and universality of the proposed protection criterion are verified.Furthermore,the criterion is tested using field measured data to verify its practicality.This work provides a new solution for the improvement of single-ended transient based protection under the new energy AC system.