Statistical Load Forecasting Based on Optimal Quantile Regression Random Forest and Risk Assessment Index
In order to support the daily operation of the smart grid,the symmetrical prediction interval(PI)constructed by the probability density function analyzes the random load behavior and realizes the symmetrical interval prediction.However,this method lacks expected risk information such as weather conditions and load changes.A novel statistical load forecasting(SLF)model is proposed by applying quantile regression random forest(QRRF),probability plots,and risk assessment index(RAI)to obtain a practical illustration of the predicted risk of load demand curves.In order to understand the actual load conditions,the proposed SLF model is constructed with accurate point prediction results,and the PI is established from each quantile using QRRF.To correlate the uncertainty of external factors with the actual load,a probability plot is used to calculate the most like-ly quantile of occurrences within the training range.Based on current inputs,the RAI is used to calculate the expected risk of PI.The proposed SLF model is validated with New England power system data and compared with benchmark algorithms and Winkler scores.The results show that the proposed method can model more accurate load PI and risk assessment than existing benchmark models.
万方数据
维普
