基于神经网络的分布式电源统一等效建模及其在PSASP中的应用

Unified Equivalent Modeling of Distributed Generation Using Artificial Neural Network and Its Application in PSASP

随着分布式电源(distributed generation,DG)在电网中渗透率的日趋增大,分布式电源的统一等效建模成为目前电力系统仿真领域的基础研究,人工神经网络(artificial neural network,ANN)以其强非线性映射能力和自适应自学习能力为此问题的研究提供了有效途径,综述了各种神经网络模型在电力系统负荷建模中的应用,并比较其优劣。对Elman神经网络的反馈支路进行了改进,使之适应分布式电源统一等效建模的需要,提出了一种改进Elman神经网络模型接入电力系统分析综合程序PSASP的思路,仿真实例表明改进Elman神经网络能有效描述各种分布式电源的外特性,能够满足工程仿真的需要。

With increasing penetration in power grid, distributed generation（DG） unified equivalent modeling becomes basic research in field of power system simulation. Artificial neural network（ANN） features strong nonlinear mapping ability and self-learning ability, providing an effective access to problem research. This paper overviews application of various neural network models in power system load modeling, and compares their merits and defects. Feedback branch of Elman neural network is improved to meet needs of unified equivalent modeling of distributed generation. Combination ideas between improved Elman neural network and power system analysis software package（PSASP） are proposed. Simulation examples show that improved Elman neural network can effectively describe external characteristics of variety of distributed generations and meet needs of engineering simulation.