Semi-automated Thermal Envelope Model Setup for Adaptive Model Predictive Control with Event-triggered System Identification

TL;DR

This paper introduces a semi-automated, knowledge graph-based framework for setting up model predictive control in building HVAC systems, reducing manual effort and improving efficiency through event-triggered system identification.

Contribution

It presents a novel semantic-assisted control framework that automates MPC setup and employs event-triggered system identification for building energy management.

Findings

Effective automatic MPC setup demonstrated in simulations

Improved computational efficiency with event-triggered SI

Enhanced accuracy of system dynamics learning

Abstract

To reach carbon neutrality in the middle of this century, smart controls for building energy systems are urgently required. Model predictive control (MPC) demonstrates great potential in improving the performance of heating ventilation and air-conditioning (HVAC) systems, whereas its wide application in the building sector is impeded by the considerable manual efforts involved in setting up the control-oriented model. To facilitate the system identification (SI) of the building envelope as well as the configuration of the MPC algorithms with less human intervention, a semantic-assisted control framework is proposed in this paper. We first integrate different data sources required by the MPC algorithms such as the building topology, HVAC systems, sensor data stream and control settings in the form of a knowledge graph and then employ the data to set up the MPC algorithm automatically. Moreover, an event-triggered SI scheme is designed, to ensure the computational efficiency and accuracy of the MPC algorithm simultaneously. The proposed method is validated via simulations. The results demonstrate the practical relevance and effectiveness of the proposed semantics-assisted MPC framework with event-triggered learning of system dynamics.