Adaptive Abnormal Condition Detection for Low-Voltage Distribution Network based on the Quantile Regression Gradient Boosting Decision Tree

With the continuous development of modern power system, new energy continues to pour in, which not only brings optimization to the energy structure, but also increases the uncertainty of power grid operation. As an important guarantee link of power supply reliability, the timely detection of abnormal working conditions of the distribution network is very important. An adaptive abnormal condition detection method for low-voltage distribution network based on quantile regression gradient boosting decision tree model is proposed in this paper. Through the fitting of the current value, the corresponding dynamic threshold is generated according to the quantile principle and brought into the corresponding detection process. Under the condition of using fewer data sets, the detection of abnormal situations with small degree of current fluctuations can be realized. Using the actual data collected in Tianjin, China, the simulation experiment verifies that the method can achieve better detection results.