Micro grids decentralized hybrid data-driven cuckoo search based adaptive protection model
Micro grids decentralized hybrid data-driven cuckoo search based adaptive protection model
复制标题
DOI：
10.1016/j.ijepes.2021.106960
复制
发表时间：
2021
影响因子：
5.2
通讯作者：
J. Marín-Quintero;C. Orozco-Henao;J. Velez;A. Bretas
中科院分区：
文献类型：
--
作者：
J. Marín-Quintero;C. Orozco-Henao;J. Velez;A. Bretas
微电网保护给公用事业公司的工程师带来了巨大的技术挑战。为了在这些类型的网络上可靠运行，已经开发了几种保护方案，但是，这些方法强烈依赖于鲁棒的通信系统。本文提出了一种分散式自适应保护方案，并引入了一种数据驱动和无通信的方法。所提出的解决方案使用人工神经网络来训练智能电子设备作为故障分类器，并且使用布谷鸟搜索元启发式进行准最优调整。人工神经网络使每个分类器仅通过本地电压和电流测量即可检测故障。所提出的解决方案不假设设备之间进行通信，但是，每个设备为其相邻设备提供支持作为后备保护。此外，还考虑了电网的系统动态，例如拓扑的变化、微电网状态或分布式能源中断。由于数据驱动模型的每个故障标签都有自己的时间操作，因此时间协调简单且易于调整。所提出的方法在改进的 IEEE34 节点测试馈线上得到了验证。自适应保护方案的结果显示准确度高于 96%，可靠性达 99%。此外，该解决方案还揭示了每个智能电子设备的位置以及特征和超参数组合之间的相关性。该方法易于实现，没有难以设计的参数，并且突出了现实生活应用的潜在方面。
Micro-grid protection presents great technical challenges to utility company engineers. Several protection schemes have been developed towards a reliable operation on these type of networks, however, the methods are strongly dependent on robust communication systems. This paper presents a decentralized adaptive protection scheme and introduces a data-driven and communication-less approach. The presented solution uses an Artificial Neural Network to train Intelligent Electronic Devices as fault classifiers, also, it uses a cuckoo search metaheuristic to its quasi-optimal adjustment. The Artificial Neural Network enables each classifier to detect faults with only local voltage and current measurements. Presented solution does not assume communication between devices, however, each device brings support to their neighboring devices as back-up protection. Further, system dynamics on the electrical network, such as, changes of topologies, micro-grid status or Distributed Energy Resources outage are considered. The time coordination is simple and easily adjustable due to each fault label of the data-driven model has its own time operation. The presented method is validated on the modifiedIEEE34-nodes test feeder. The results of the adaptive protection scheme show accuracy values above of 96% and dependability of 99%. Also, the solution reveals a correlation between the location and the combination of features and hyper-parameters for each Intelligent Electronic Device. The method is development to be easy-to-implement, without hard-to-design parameters, and with highlights potential aspects for real-life applications.