应用人工神经网络进行电力系统动态安全评价的新方法

本文提出了一种应用人工神经网络进行电力系统动态安全评价的新方法。把故障前系统的稳态运行参数作为特征量,摇摆过程中发电机转子间的最大相对摇摆角作为系统稳定性的量度,并证明了它们之间呈连续映射关系。随后表明了用三层前向网络实现这类映射的可行性。为减少神经网络的训练时间,本文提出了网络“分解—集结”方法和映射与分类相结合的思想。6机22节点系统的试验结果表明了本文提出方法的有效性。

A new method for power system dynamic Security assessmentusing artificial neural networks

This paper presents a new method for power system dynamic security assessment using aritificial neural networks, the oprating conditions of pre-fault system are taken as features, and the measurement of system stability is maxinum relative swing angle of generator rtors in swing process. The continous mapping significance between them is proved. The following is to show the feasibility to realige the mapping using three-layer forward networks. In order to reduce training time of neural networks, a network "Decomposition-aggregation" technique and an idea to combine mapping and classify are proposed. The test result of 6-generators, 22-buses system shows the effectiveness of the proposed method.