Perspectives on system identification

Abstract

Abstract:System identification is the art and science of building mathematical models of dynamic systems from observed input-output data. It can be seen as the interface between the real world of applications and the mathematical world of control theory and model abstractions. As such, it is an ubiquitous necessity for successful applications. System identification is a very large topic, with different techniques that depend on the character of the models to be estimated: linear, nonlinear, hybrid, nonparametric etc. At the same time, the area can be characterized by a small number of leading principles, e.g. to look for sustainable descriptions by proper decisions in the triangle of model complexity, information contents in the data, and effective validation. The area has many facets and there are many approaches and methods. A tutorial or a survey in a few pages is not quite possible. Instead, this presentation aims at giving an overview of the "science" side, i.e. basic principles and results and at pointing to open problem areas in the practical, "art", side of how to approach and solve a real problem.

Model-predictive control for non-domestic buildings: a critical review and prospects

“…In the case where their MPC optimisation failed to yield a solution, Sturzenegger et al (2013) used the strategy of applying the second, normally discarded, element of the control sequence from the previous optimisation. Potentially, MPC should also serve as a diagnostic tool to detect malfunctions/degradation in the HVAC plant or sensors (Ljung, 2010). If the optimisation is consistently unable to produce a feasible solution, this should trigger an investigation in advance of occupant complaints about conditions.…”

“…Possibly a different sort of hybrid method where appropriate physics-based models of individual items of plant (e.g. a single fancoil unit) are embedded in a black-box framework may be usefully pursued, as suggested by Ljung (2010), although such an approach may lack the robustness to accommodate degradation due to wear-and-tear of the white-boxmodelled components. As a counter-argument, Vá a et al (2014, p.792) contend that grey-box modelling is to be preferred as black-box modelling "usually spoils the real system structure" (although quite what these authors mean by this is unclear).…”

“…Non-linear models are, nonetheless, established in the wider MPC literature (Camacho & Bordons, 2004), and have been successfully used by Bengea et al (2014) on a centralised supervisory system controlling a mid-sized building. Ljung (2010) has discussed the research challenges in the identification of non-linear models for control, in general. We consider that the necessity, or otherwise, of using non-linear plant models for buildings should be investigated, and compared to Hammerstein models.…”

On the use of watermark-based schemes to detect cyber-physical attacks

“…This situation can be quantified using mean squared error (MSE) [26,27]. In a nutshell, 12 Numeric simulation results using the single-and multi-watermark detection schemes with the same performance loss, and different adversary window size for eavesdropping the data channel before the attack.…”

Identification of single-input–single-output quantum linear systems

“…System identification theory [32][33][34][35][36] lies at the interface between control theory and statistical inference, and deals with the estimation of unknown parameters of dynamical systems and processes from input-output data. The integration of control and identification techniques plays an important role, e.g., in adaptive control [37].…”