基于数据学习的最大频率偏差概率评估

发电技术

• • 下一篇

基于数据学习的最大频率偏差概率评估

温佳鑫，卜思齐，陈麒宇，周博文

1．香港理工大学电机工程学系，香港特别行政区九龙区 999077； 2．中国电力科学研究院有限公司，北京市海淀区100192； 3．东北大学信息科学与工程学院，辽宁省沈阳市 110819

Data-driven Probabilistic Assessment on Max Frequency Deviation

WEN Jiaxin, BU Siqi, CHEN Qiyu, ZHOU Bowen

1. Department of Electrical Engineering, The Hong Kong Polytechnic University, Kowloon District, Hong Kong SAR 999077, China; 2.China Electric Power Research Institute, Haidian District, Beijing 100192, China; 3. College of Information Science and Engineering, Northeastern University, Shenyang 110819, Liaoning Province, China

摘要/Abstract

摘要： 新能源正在逐步代替传统发电厂为用户提供电能，但同时也为频率的安全运行带来风险。因此，在规划阶段需要全面地对最大频率偏差越线风险进行概率评估。基于蒙特卡罗的规划方法效率很低，而神经网络可以通过对数据的学习做出快速有效的预测。因此，提出一种基于蒙特卡罗的神经网络算法以实现对区域最大频率偏差越线风险的快速评估。首先，产生大量随机扰动，仅需要对小部分扰动进行仿真，然后将这部分数据送入神经网络进行训练，并将剩余的大部分扰动送入训练好的神经网络进行输出预测，重复以上学习和预测的过程，将多次预测结果的平均值作为最终的预测输出，最终得到各个风险区间的概率分布情况。在IEEE 10机39节点的系统上验证了该规划方法的有效性。

关键词: 概率评估, 最大频率偏差, 蒙特卡罗-神经网络

Abstract: The renewable energy is gradually replacing synchronous generator to provide electricity to the consumers, meanwhile also brings potential risks on frequency stability. Thus, in operational planning stage, it is necessary to evaluate the risk of frequency violation in a probabilistic manner. There is a low efficient for Monte Carlo Simulation (MCS)-based planning while the artificial neural network (ANN) can predict the results effectively and efficiently via data learning. Therefore, this paper proposes an MCS-ANN to assess the risk of max frequency deviation. First, a large number of stochastic disturbances are generated, while only a small part of disturbances is used for MCS. Then, the simulation results as training data is used to train the ANN and the other large part of disturbances are sent into the trained model to obtain the predicted results. Repeat above steps to require an average value of the many predicted values as the final results. Final, according the forecasted results, the probabilistic distribution of concerned frequency intervals is obtained. The effectiveness and efficiency of the proposed MCS-ANN is verified on a modified IEEE 10-machine 39-bus.

Key words: probabilistic assessment, max frequency deviation, MCS-ANN

温佳鑫, 卜思齐, 陈麒宇, 周博文. 基于数据学习的最大频率偏差概率评估[J]. 发电技术.

WEN Jiaxin, BU Siqi, CHEN Qiyu, ZHOU Bowen. Data-driven Probabilistic Assessment on Max Frequency Deviation [J]. Power Generation Technology.