Integrated Online Monitoring and Adaption of Process Model Predictive Controllers

TL;DR

This paper introduces an event-triggered, data-driven method for adaptively updating model predictive controllers using statistical performance monitoring, reinforcement learning, and identification, validated on a district heating system simulation.

Contribution

It presents a novel adaptive MPC framework that avoids continuous updates, reducing unnecessary control changes and preventing catastrophic forgetting.

Findings

Effective detection of performance degradation.

Successful controller adaptation via reinforcement learning.

Validated approach on a district heating system simulation.

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.