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

Figure 11

The empirical coverage of (a) QGBRT, (b) QCNN, (c) QLSTM, and (d) our proposed model. Y-axis denotes normalized empirical coverage and X-axis represents different residences with 10 bars individually corresponding to 10 intervals. A bar with height close to 1 indicates a better empirical coverage.

(a)

(b)

(c)

(d)