Unified Quantile Regression Deep Neural Network with Time-Cognition for Probabilistic Residential Load Forecasting

Figure 10

Heat map of the 80% PINAW for four models. QGBRT, QCNN, QLSTM, and our proposed model in X-axis correspond to PINAW, respectively, while I_QGBRT, I_QCNN, and I_QLSTM denote PINAW improvement ratio compared with our method over other models. Y-axis represents different residences. Results of each residence are compared in the same row, where darker green shades imply a better performance.