Model predictive control for dynamic indoor conditioning in practice

doi:10.1016/j.enbuild.2022.112548

2022-12-01

PRE

Quirin Hamp *

Fabian Levihn

DOI：10.1016/j.enbuild.2022.112548

原文链接

原文求助

摘要

En 中文

The capability to dynamically plan, predict, and control indoor conditioning allows to adapt to individual preferences of inhabitants and enables demand side management. While former mainly improves thermal comfort of inhabitants so does latter unlock ecological and financial opportunities mostly for energy utilities. Commonly, dynamic indoor conditioning is based on piece-wise constant indoor temperature constraints. This paper's contribution is the presentation of additional constraints: in particular ones expressed relative to the nominal behavior of a hydronic heating system. This allows to simultaneously harness the relevant process variables in particular during the pre-loading and post-loading phases of a load reduction. The findings are based on data sets acquired on 10 inhabited, residential buildings in Stockholm over a whole year. One of the findings is that building models need to be adaptable if predictive control is applied in practice. This adaptability is assured by a novel concept, i.e. a so called model manager on which the control is relying for the selection of the most accurate model. Centralized optimal control of buildings connected to a district heating network is challenging in practice due to a high computational load. In order to reduce it, the herein presented method elaborates plans only every hour instead of at every control step for optimal control. Since these plans cannot be optimal due to the lack of regular update a hitherto unknown cascaded control logic has been developed that corrects planning errors and other disturbances. Capabilities are demonstrated and compared to conventional controllers in dynamic simulations of a multi-zoned building. The herein presented method is to our knowledge the first to provide all flexibility desired by energy utilities and inhabitants alike through harnessing the consequences of transitions. (C) 2022 Elsevier B.V. All rights reserved.

Keyword:

Model predictive control

Demand side management

Load reduction

Configurability

Building heating

Thermal storage

Thermal comfort

Demand forecasting

State space

District heating system

Modeling