Good models for building thermal behaviour are an important part of developing building energy management systems that are capable of reducing energy consumption for space heating through model predictive control. A popular approach to modelling the temperature variations of buildings is grey-box models based on lumped parameter thermal networks. By creating simplified models and calibrating their parameters from measurement data, the resulting model is both accurate and shows good generalisation capabilities. Often, parameters of such models are assumed to be a combination of different physical attributes of the building, hence they have some physical interpretation. In this paper, we investigate the dispersion of parameter estimates by use of randomisation. We show that there is significant dispersion in the parameter estimates when using randomised initial conditions for a numerical optimisation algorithm. Further, we claim that in order to assign a physical interpretation to grey-box model parameters, we require the estimated parameters to converge independently of the initial conditions and different datasets. Despite the dispersion of estimated parameters, the prediction capability of calibrated grey-box models is demonstrated by validating the models on independent data. This shows that the models are usable in a model predictive control system.
Forecasting weather conditions is important for, e.g., operation of hydro power plants and for flood management. Mechanistic models are known to be computationally demanding. Hence, it is of interest to develop models that can predict weather conditions faster than traditional meteorological models. The field of machine learning has received much interest from the scientific community. Due to its applicability in a variety of fields, it is of interest to study whether an artificial neural network can be a good candidate for prediction of weather conditions in combination with large data sets. The availability of meteorological data from multiple online sources is an advantage. In order to simplify the retrieval of data, a Python API to read meteorological data has been developed, and ANN models have been developed using TensorFlow.
Building Energy Management systems can reduce energy consumption for space heating in existing buildings, by utilising Model Predictive Control. In such applications, good models of building thermal behaviour is important. A popular method for creating such models is creating Thermal networks, based cognitively on naive physical information about the building thermal behaviour. Such models have lumped parameters which must be calibrated from measured temperatures and weather conditions. Since the parameters are calibrated, it is important to study the identifiability of the parameters, prior to analysing them as physical constants derived from the building structure. By utilising a statistically founded parameter estimation method based on maximising the likelihood function, identifiability analysis can be performed using the Profile Likelihood method. In this paper, the effect of different sensor locations with respect to the buildings physical properties is studied by utilising likelihood profiles for identifiability analysis. The extended 2D profile likelihood method is used to compute two-dimensional profiles which allows diagnosing parameter inter-dependence, in addition to analysing the identifiability. The 2D profiles are compared with confidence regions computed based on the Hessian.
Obtaining accurate models that can predict the behaviour of dynamic systems is important for a variety of applications. Often, models contain parameters that are difficult to calculate from system descriptions. Hence, parameter estimation methods are important tools for creating dynamic system models. Almost all dynamic system models contain uncertainty, either epistemic, due to simplifications in the model, or aleatoric, due to inherent randomness in physical effects such as measurement noise. Hence, obtaining an estimate for the uncertainty of the estimated parameters, typically in the form of confidence limits, is an important part of any statistically solid estimation procedure. Some uncertainty estimation methods can also be used to analyse the practical and structural identifiability of the parameters, as well as parameter inter-dependency and the presence of local minima in the objective function. In this paper, selected methods for estimation and analysis of parameters are reviewed. The methods are compared and demonstrated on the basis of both simulated and real world calibration data for two different case models. Recommendations are given for what applications each of the methods are suitable for. Further, differences in requirements for system excitation are discussed for each of the methods. Finally, a novel adaption of the Profile Likelihood method applied to a moving window is used to test the consistency of dynamic information in the calibration data for a particular model structure.
The occurrence of slug flow is a common problem arising in the oil well riser pipeline. To eliminate such slug flow, various control structures along with state estimation are designed and compared in this paper. Nonlinear model based predictive scheme are compared with classical PI controllers for three different control structures. One of the control structure is based on controlling the mass of the liquid in the riser pipeline, for which, an Unscented Kalman Filter is designed to estimate the mass. The simulation results show that the model based controllers perform relatively better than the classical controllers. Although computationally expensive, the control algorithm used in this paper for model based control still makes it real time implementable.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.