Figure 6 – 15-minute-ahead probabilistic forecasts for SAG throughput (TPH) and discharge pressure with P10–P90 uncertainty bands (400 mixture samples); example of asymmetric predictive intervals. 30-minute horizon. Figure 7 presents 30-minute-ahead forecasts using 1000 samples for the uncertainty bands. As expected, the point forecast becomes smoother and tends to lose some short-term variability at longer horizons; nonetheless, the measured throughput trend remains largely covered by the P10–P90 interval, which is valuable for risk-aware decision support when comparing operational alternatives over longer lead times. Improving long-horizon sharpness can be addressed in future work by combining multi-step strategies (e.g., rollouts of shorter-horizon models with predicted actions) and by integrating anticipated control actions from existing expert logic and regulatory loops. Figure 7 – 30-minute-ahead probabilistic forecasts for SAG throughput (TPH) and discharge pressure with P10–P90 uncertainty bands (1000 mixture samples).
RkJQdWJsaXNoZXIy MTM0Mzk2