The process of deriving a transfer function (and the more general process of dynamic modeling) is an art and science in and of itself and all too often neglected when teaching controls.

That said, here's the general approach:

1) determine the inputs and outputs. In general, inputs are variables that are independent of the system dynamics and should include your control variable.. Outputs can be thought of as variables you measure.  For each transfer function you have one input and one output.

2) Determine the 'lumps' of your system.  Since transfer functions are based on Ordinary Differential Equations (as opposed to PDE's), You need a lumped parameter modeling approach. Identify thermal masses for which you can assume uniform (but not constant) temperature distributions.

3)  Apply conservation of energy, conservation of mass and heat transfer laws to each lump and combine to get a single ordinary differential equation with your input and output.

4)  Take Laplace transform with zero initial conditions and manipulate to get your transfer function;.

Good luck!

REF:  Kulakowski, Gardner (ahem) and Shearer. "Dynamic Modeling and Control of Engineering Systems", Cambridge, 2007.

thankyou john.

Hello, Devishree! Transfer function is a ratio between output of system and its input written in Laplace transformation. In terms of thermal system, the outnumbered part of units can be presented as an aperiodic function, i.e. the response of thermal system on the input has exponential shape. The easiest (and quite popular in engineering) case is the first-order aperiodic function. The example of this function is shown in the picture attached below.

The denominator of function is a static gain, i.e. the number that show how much your system,that represented by transfer function, amplify the input. For example, if the transfer function presents the thermal behavior of resistor under the current passing through, than the number "10" in denominator means that temperature of resistor increase 10 times due to the current of 1 A. This tells us about steady state. In terms of transient modes, we have to have the time constant for the first-order function (how quickly the system responses to the input). In the case of the model attached below, we have value of the time constant equals 0.5. It means that  the transient time is about 1.5...2 s.

Also, I attached some links that could be helpful for you (from my point of view).

Good luck and best wishes for you.   

http://ece.gmu.edu/~gbeale/ece_421/xmpl-421-1st-order-01.pdf

ftp://ftp.dca.fee.unicamp.br/pub/docs/vonzuben/ea619_2s12/transfer_function_block_diagram.pdf

Sorry, Devishree, in the previous answer I committed an error in the sentence: "...the number "10" in denominator means that temperature of resistor increase 10 times due to the current of 1 A.". It should be written as follow: the number "10" in denominator means that temperature of resistor increase on 10 K (or another unit of temperature) due to the current of 1 A. It means that if the current has value of 2 (3, 4 A, etc), than the temperature will increase on 20 (30, 40 K, etc). Another words, the output equals the input multiplied by the gain.

Sorry for my inaccuracy in the previous answer (it's about my English...)

This is a nice introduction to heat-transfer (free download): http://web.mit.edu/lienhard/www/ahtt.html

you can use black-box modelling, 

where you apply an input and measure it and record and measure the output>

in this case you have the input and the output and you can guess the system order and come up with a n approximate linear model

try to include all the system equipment in your sinulink diagram, from governor to turbine to generator to load, so that you should determine the parameters of governor time constant, turbine time constant, inertia constant, damping coefficient and torque gain. Thus you can obtain the transfer function between load power input and frequency output. This will have 3 blocks of first order system in each block

If your system is already modelled, you can use the linearization or identification toolboxes after specifying the imputs and outputs of the system.

thankyou Alex, Nadir, Emna, Wesam, Eielsen... 

this is an introduction about thermal transfer function identification :

https://www.sciencedirect.com/science/article/abs/pii/S0017931017308864

https://www.sciencedirect.com/science/article/abs/pii/S0017931017320574