基于KPCA和XGBoost算法的非侵入式负荷辨识方法

非侵入式负荷监测(Non-Intrusive Load Monitoring，NILM)通过采集用户侧智能电表的电气特征数据，进行数据挖掘与分析，能够有效的实现负荷辨识。在家用电器功率、电流、电压波形及各次谐波特征的数据中，采用核主成分分析方法（Kernel Principal Components Analysis，KPCA），解决非线性特征提取与降维，最大限度抽取特征信息。再利用一维卷积核提取时序特征并压缩后输入到XGBoost模型，得到负荷辨识结果。在实验数据集上进行了验证，证明文中提出算法的泛化性和执行效率方面有较大优势。

Non-intrusive load identification method based on KPCA and XGBoost algorithm

Non-intrusive Load Monitoring (NILM) can effectively achieve Load identification by collecting the electrical characteristics data of the user side intelligent electricity meter and carrying out data mining and analysis. In the data of power, current, voltage waveform and each harmonic characteristic of household appliances, Kernel Principal Components Analysis (KPCA) is adopted to solve nonlinear feature extraction and dimension reduction, and extract feature information to the maximum extent. One dimensional convolution kernel is used to extract time series features and compress them into the XGBoost model to obtain load identification results. It is verified on the experimental data set, which proves that the proposed algorithm has great advantages in generalization and execution efficiency.