NAME: identification expand the application field beyond the scope


ID: 14002264

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!

order now



System identification in a nutshell….

It is basically the task of precisely constructing a dynamic model
that can be used to predict the outputs of a dynamical system. Models of
dynamic systems are typically described by differential or difference
equations, transfer functions, state-space equations, and pole-zero-gain


According to the IEEE Control Systems Magazine, early work in the
field of system identification was started or developed by the statistics and
time series communities. It can widely be seen or portrayed in the work of
Fisher (1912), Gauss (1809) and the theory of stochastic processes.

The start of the model based control era was around the 1960’s
with the help of Kalman’s key papers. This lead to the development of model based
theory for prediction, filtering and control. A growing pressure to apply
modern techniques to areas where models are not available from physics was need
thus the Need for system identification.


It is mostly used to estimate a grey or black model of a system which
is dynamic based on observing the input-output from experimental data.

Thus, system identification can be used to solve problems in
diverse fields because reliability and availability of the design techniques of
system identification expand the application field beyond the scope of
industrial applications.

So, system identification helps to:

Design control strategies for a system after
identification, example can be in optimizing an electrical microgrid operation.

Forecast the evolution for a particular system
(e.g. the future climate prediction according a to IPCC downscaling model).

Improve the internal knowledge of the system
which is to be identified.

Distinguish concealed components impacting a
system for example sun spots in the karst spring.

To recognize the interaction between coupled
systems (for example glaciers and climate).

Learn and analyse the properties of an
identified system.

System identification techniques

Maximum likelihood

Prediction-error minimization (PEM)

Subspace system identification


It is possible to create linear and
nonlinear dynamic system models from measured input or output data with the
help of System
Identification Toolbox™ in MATLAB.

The system identification toolbox provides Simulink blocks as well
as MATLAB functions and an application for constructing mathematical models of
dynamic systems. It gives you a chance to make and utilize models of dynamic
frameworks not effectively displayed from first standards or details. You
can utilize time-area and recurrence space input-yield information to
distinguish constant time and discrete-time exchange capacities, process
models, and state-space models. The toolbox additionally gives algorithms for
embedded online parameter estimation.

The toolbox can also perform grey-box identification for
estimating parameters of a client characterized model. System response
prediction and plant modelling of the identified model can be used in Simulink.
The toolbox supports time-series forecasting and time-series data modelling.


There are basically two different applications of system
identification which are control and analysis.

Control – in system identification you would want to know the
dynamics of the plan and you would want to design a controller to improve the
performance of the plan or system.

Analysis – the analysis part is basically aims to understand the
functioning of the plan or system

Predicting future climate change-

Researchers have built up a few PC run
recreations, or models, that consolidate and express in mathematical form what
we know about the procedures that control the atmospheric and hydrologic frameworks.
The models used are called General Circulation Models(GCM). These models are also
used by scientists to try and predict the effects or impacts of increased
greenhouse gas concentration in the atmosphere.

Prediction of energy consumption in buildings-

The models used will predict the systems
performance in terms of energy consumption using the measured input and output.
To train and test the models, data is taken or acquired from existing buildings.
Nonlinear, State space, and polynomials models based mathematical functions are
tested with different parameters such are temperature, time, and dew
point.  The outcomes demonstrate that the proposed models can yield or
output similar energy results. The created model can be utilized for vitality
evaluation and analysis.


System identification Process

Experiment planning

Selection of model procedure

Parameter Estimation

Model validation