Integrated Online Monitoring and Adaption of Process Model Predictive Controllers

Abstract

This paper addresses the design of an event-triggered, data-based, and performance-oriented adaption method for model predictive control (MPC). The performance of such a strategy strongly depends on the accuracy of the prediction model, which may require online adaption to prevent performance degradation under changing operating conditions. Unlike existing methods that continuously update model and control parameters from data, potentially leading to catastrophic forgetting and unnecessary control modifications, we propose a novel approach based on statistical monitoring of closed-loop performance indicators. This framework enables the detection of performance degradation, and, when required, controller adaption is performed via reinforcement learning and identification techniques. The proposed strategy is validated on a high-fidelity simulation of a district heating system benchmark.

Figures (3)

• Figure 1: Schematic of the AROMA DHS and its main variables.
• Figure 2: Time series trajectories of key variables for cases 1, 2, and 3. Time instants where changing conditions and parameter updates occur are indicated with arrows in the bottom plot. The long purple arrow in case 3 indicates when sysID is performed.
• Figure 3: Slices of feature space for case 2. Blue crosses are data points in $\mathcal{D}$. Circles show moving averages of features over 40 time steps after adaption has been triggered (black for early data, transitioning to yellow for new data).

Theorems & Definitions (4)

• Definition 1
• Definition 2
• Remark 1
• Remark 2