Abstract: Residential power load forecasting is mainly used in the power outage scheme in the power dispatching to improve power supply reliability and residential customer satisfaction. Because of the large amount of power data and the strong uncertainty factors,it is often difficult to predict the load.The existing power load forecasting methods cannot obtain the degree of freedom of power data,resulting in poor stability of load forecasting process and low accuracy of prediction results.To this end,a residential power load forecasting method based on depth conditional probability density function is proposed in this paper. By introducing the fourth power kernel function,the variable relationship between the observed value and the predicted value of residential power load data with time change is obtained. The predicted vector values conform to normal distribution by gaussian regression equation. The optimal degree of freedom of the predicted value is extracted by the crossvalidation method,the quantile is determined by the degree of freedom,and the prediction data quantile of the next random variable is determined according to the comparative analysis results,so as to realize the prediction of residential power load.The simulation results show that the power load fluctuation results obtained by the proposed method are consistent with the measured values,and the error can be controlled within 0.001MW-0.437 MW. The experimental data suggests that this method has high prediction accuracy and can provide effective help for power decision-making.