摘 要
直流配电网作为未来智能电网的重要组成部分，其高效、可靠运行对能源转型和电力系统现代化具有重要意义。然而，由于直流配电网结构复杂且故障特征与传统交流系统存在显著差异，现有的故障定位方法难以满足其高精度和快速响应的需求。为此，本文针对直流配电网故障定位问题展开深入研究，旨在设计一种适应性强、计算效率高的新型算法。研究首先分析了直流配电网的拓扑结构及故障传播特性，结合实际应用场景提出了一种基于多源信息融合的故障定位框架。该框架通过整合电压、电流暂态信号以及网络拓扑参数，构建了能够全面反映故障特征的数学模型。在此基础上，本文引入深度学习技术，利用卷积神经网络提取故障特征并实现精准分类，同时结合遗传算法优化模型参数以提升定位精度。实验结果表明，所提算法在多种工况下均表现出优异性能，不仅能够准确识别故障位置，还能有效应对噪声干扰和数据缺失等问题。此外，相较于传统方法，该算法显著降低了计算复杂度，具备更强的实时性。本文的主要创新点在于将多源信息融合与人工智能技术有机结合，突破了传统故障定位方法的局限性，为直流配电网的安全稳定运行提供了有力支撑。研究成果可为相关领域的理论发展和技术应用提供重要参考。关键词：直流配电网；故障定位；多源信息融合；深度学习；遗传算法优化
Abstract
As an essential component of future smart grids, the efficient and reliable operation of DC distribution networks plays a significant role in energy transition and the modernization of power systems. However, due to the complex structure of DC distribution networks and the significant differences in fault characteristics compared to traditional AC systems, existing fault location methods struggle to meet the demands for high precision and rapid response. To address this issue, this study conducts an in-depth investigation into fault location problems in DC distribution networks, aiming to design a novel algorithm with strong adaptability and high computational efficiency. The research begins by analyzing the topological structure and fault propagation characteristics of DC distribution networks, proposing a fault location fr amework based on multi-source information fusion tailored to practical application scenarios. This fr amework integrates voltage and current transient signals along with network topology parameters to construct a mathematical model that comprehensively reflects fault characteristics. Building on this foundation, deep learning technology is introduced, utilizing convolutional neural networks to extract fault features and achieve precise classification, while genetic algorithms are employed to optimize model parameters for enhanced location accuracy. Experimental results demonstrate that the proposed algorithm exhibits superior performance under various operating conditions, not only accurately identifying fault locations but also effectively addressing issues such as noise interference and data loss. Moreover, compared to traditional methods, the algorithm significantly reduces computational complexity, offering stronger real-time capabilities. The primary innovation of this study lies in the organic combination of multi-source information fusion and artificial intelligence technologies, overcoming the limitations of conventional fault location methods and providing robust support for the safe and stable operation of DC distribution networks. The research findings offer critical references for theoretical development and technical applications in related fields..
Key Words：Dc Distribution Network；Fault Location；Multi-Source Information Fusion；Deep Learning；Genetic Algorithm Optimization
目 录
摘 要 I
Abstract II
第1章 绪论 2
1.1 直流配电网故障定位的研究背景与意义 2
1.2 国内外直流配电网故障定位算法研究现状 2
1.3 本文研究方法与技术路线 3
第2章 直流配电网故障特性分析 4
2.1 直流配电网的结构与运行特点 4
2.2 故障类型及其对系统的影响 4
2.3 故障暂态信号特征提取方法 5
2.4 故障定位的关键技术需求 6
第3章 基于信号处理的故障定位算法设计 8
3.1 故障信号采集与预处理 8
3.1.1 数据采样与滤波方法 8
3.1.2 特征量提取与选择 8
3.1.3 数据标准化与降维 9
3.1.4 信号质量评估指标 9
3.2 时间域分析方法的应用 9
3.2.1 故障行波传播特性 10
3.2.2 行波到达时间估计方法 10
3.2.3 双端测距原理与实现 10
3.2.4 时间误差校正策略 11
3.3 频率域分析方法的研究 11
3.3.1 故障谐波分量提取 12
3.3.2 频谱特征与故障位置关系 12
3.3.3 傅里叶变换在定位中的应用 12
3.3.4 频率分辨率优化方法 13
第4章 算法性能验证与优化改进 14
4.1 模拟仿真平台搭建 14
4.1.1 仿真模型构建方法 14
4.1.2 参数设置与边界条件 14
4.1.3 故障场景生成策略 15
4.1.4 数据集生成与标注 15
4.2 实验测试与结果分析 15
4.2.1 测试环境与设备配置 16
4.2.2 不同工况下的定位精度评估 16
4.2.3 算法鲁棒性测试方法 16
4.2.4 结果对比与误差分析 17
4.3 算法优化策略研究 17
4.3.1 提高计算效率的方法 18
4.3.2 减少误判率的技术手段 18
4.3.3 多源信息融合定位方法 18
4.3.4 实时性与准确性权衡策略 19
结 论 19
参考文献 21
致 谢 22
