Zone MPC with guaranteed identifiability in presence of predictable disturbances

摘要

In this paper, the task of ensuring sufficiently excited data for the subsequent re-identification of a model used within the model predictive control framework is tackled. It introduces two algorithms that are developed specifically for zone MPC for systems with external disturbances. Both of them work in two steps and build on the original zone MPC solution. The trade-off between the data informativeness and degradation of the control performance is given by a user-defined threshold for the performance degradation. Moreover, a new optimization criterion quantifying data informativeness is introduced. The proposed algorithms are described in detail and their theoretical properties are discussed. Two different verification scenarios are presented, the first one with an artificial example and the second one considering a real-life building climate control task using a high-fidelity testbed model. The results show that they significantly outperform the classic zone MPC in data informativeness while maintaining acceptable zone violation and energy consumption. (C) 2019 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.