基于机器学习的电力系统暂态稳定评估

发布时间：2018-05-25 22:42

本文选题：粗糙集理论 + 最小二乘支持向量机　；参考：《华北电力大学》2017年硕士论文

【摘要】：随着中国经济快速的增长,电力系统也在不断的发展,电网正向着远距离、特高压方向发展,运行中的电力系统规模将会变得更加庞大、物理变化过程也会更为复杂。在环境保护和经济可持续发展的压力下,电网的安全运行将面临更加严峻的考验。由于广域测量系统和机器学习技术的快速发展,利用数据挖掘和机器学习方法进行电力系统暂态稳定评估得以实现。本文将分别采用两种方法对电力系统进行暂态稳定评估,主要做了如下工作:(1)本文利用仿真软件进行仿真运算,采集故障数据,并编写特征量采样程序得到暂态稳定评估的原始样本集。在此基础之上,运用粗糙集理论对暂态稳定特征量进行属性约简,减少特征量的维数;然后以最小二乘支持向量机作为分类器,得到了良好的分类效果,并最后通过算例验证了所提方法的有效性。(2)考虑到故障特征与故障本身属性关系紧密,且还有人工参与因素,很大程度上增加了提取与优化的不确定性,也增加了分类器分类识别的难度,降低了机器学习的智能性。本文将深度学习的概念引入,并选择其中的深度信念网络应用于电力系统的暂态稳定评估中。深度信念网络是一种通过组合低层特征形成更高层抽象表示,并发现数据分布式特征表示的学习网络,本文根据它的特点,提出直接从原始样本集出发对故障后的状态进行分类的新方法,并通过算例验证了所提方法的有效性。(3)随着接入的风电容量不断增大,电力系统的稳定性将会发生改变。发电机转子角能够直接反映系统的暂态稳定性,本文将故障切除前的转子角轨迹特征作为深度信念网络的输入,对电力系统暂态过程中的发电机转子角的变化趋势进行预测研究。
[Abstract]:With the rapid growth of Chinese economy and the continuous development of electric power system, the power grid is developing in the direction of long-distance and ultra-high voltage. The scale of power system will become larger and the process of physical change will be more complicated. Under the pressure of environmental protection and economic sustainable development, the safe operation of power grid will face more severe test. With the rapid development of wide-area measurement system and machine learning technology, power system transient stability assessment is realized by using data mining and machine learning methods. In this paper, two methods are used to evaluate the transient stability of power system. The main work is as follows: 1) in this paper, we use the simulation software to carry out the simulation operation and collect the fault data. The original sample set of transient stability assessment is obtained by programming characteristic sampling program. On this basis, using rough set theory to reduce the dimension of transient stability feature, and then using least square support vector machine as classifier, a good classification effect is obtained. Finally, an example is given to verify the effectiveness of the proposed method. (2) considering that the fault features are closely related to the attributes of the fault itself, and there are human participation factors, the uncertainty of extraction and optimization is greatly increased. It also increases the difficulty of classifier recognition and reduces the intelligence of machine learning. In this paper, the concept of deep learning is introduced, and the depth belief network is selected for power system transient stability evaluation. The deep belief network is a kind of learning network which forms a higher level abstract representation by combining the lower level features and finds the data distributed feature representation. A new method is proposed to classify the state of fault directly from the original sample set. The effectiveness of the proposed method is verified by an example. The stability of power system will change with the increasing of the wind capacity. The rotor angle of the generator can directly reflect the transient stability of the system. In this paper, the characteristics of the rotor angle trajectory before the fault removal are taken as the input of the depth belief network. The change trend of generator rotor angle during transient process of power system is predicted.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM712

【参考文献】