Entwicklung eines Konzeptes zur Regelung von raumlufttechnischen Anlagen auf der Basis von Methoden der dynamischen Systeme

In this paper we will show a way to a paradigm shift in building automation. Air handling units (AHU) play an increasingly important role in the air conditioning of buildings and their energy consumption. Classical approaches in building automation use PID controllers to ensure stable, comfortable and energy efficient operation of AHUs. When using a PID controller, a disturbance is not compensated until it has already caused a measurable change, e.g. in the room temperature. The time from the occurrence of the disturbance to the detection of its influence on the room climate is decisive for the speed with which the control can compensate for the disturbance. The slower the detection, the greater the effect of the disturbance and associated with it, for example, the cost as well as the discomfort of overheating a room. We will show that it is possible to detect the intensity of a disturbance based on a thermal source without waiting for its effect on the room climate. An indicator will be used to estimate the intensity of the thermal sources, based on an evaluation of a high-frequency signal component of a temperature or air velocity measurement. In order to obtain this indicator and the necessary basic understanding, we investigate the relevant properties of thermal convection, as they can occur under comfortable conditions in room air conditioning, using the example of Rayleigh-Bénard convection (RBC). A rectangular geometry in 2D is heated from below and cooled from above at temperature differences that approximate the heat dissipation of people indoors. The goal is to determine whether structures can be identified in the resulting flow field that can be used to estimate the intensity of the heat sources driving the convection. The convection under the investigated parameters shows a deterministic chaotic behavior. In order to be able to consider the properties of this behavior prototypically and isolated from other disturbing or too specific influences, we use the Lorenz model. The focus is on the use of different approaches from the field of time series analysis and statistics to test which approach can be used validly and numerically inexpensive for the formation of the indicator. The use of the sample standard deviation will be established as an important first building block. In order to better assess the effects of geometry, measured quantity as well as evaluation time in the formation of the indicator, the RBC will be investigated in detail with a direct numerical simulation (DNS). It will be shown that the use of the sample standard deviation allows to estimate the intensity of the convection with a linear approach with a high statistical confidence. The theoretical findings from the Lorenz model and the DNS will be complemented by measurements under laboratory conditions. It is shown that a transfer of the evaluation methods to a three-dimensional space geometry is possible by using a suitable measurement technique. Qualitative agreements on the specific properties, especially in the difference between an air velocity and a temperature measurement to the simulated data, underline the goodness of this broadly performed analysis. Finally, we will show that people can also be detected as real heat sources with the indicator and that, under this condition, it is possible to demonstrate the effects on operational safety and the possible cost savings on a model. In the model, the indicator is applied as a basis for a new model predictive (MP) control concept. It is shown that an improvement of the control performance as well as a reduction of the operation management costs can be realized. The main innovation of the concept is the reduction of the degree of difficulty of the controlled system in case of disturbance. This approach offers a wide range of applications, the full potential of which still needs to be further developed.