I think LEACH-C is a good choice, since LEACH protocol suffers from several problems, one of them is non-uniform clustering. Hence, better clusters may be produced using centrally controlled algorithm like LEACH Centralized .

"LEACH-C" protocol is better that "LEACH" :

Although the advantages of "LEACH" routing protocol that it reduce the energy consumption for the nodes , it  has some of disadvantages such as :

- it focus significantly on the cluster head rather than cluster member this will deal      sometimes with the failure cluster head problem

- the selection of the cluster head is randomly which does not take take energy consumption of different nodes within the cluster 

-it does not work well with the application that require large area coverage along with muti-hop inter cluster communication.

"LEACH-C" is used to improve "LEACH" protocol : the number of the cluster heads in each round is equals to a predetermined optimal value,  while in "LEACH" protocol the number of the cluster heads varies from round to another because of the lack of global coordination between different nodes in the network.

According to this comparison , I decide that using Kmeans  algorithm for "LEACH-C" rather than the "LEACH" itself

I think a simulation of this already exists in NetSim where the k means is done in MATLAB and the leach simulation in NetSim

Thank you so much for your answers friends.

I want to simulate my idea on NS2 simulator.

can anyone give me an idea about how to implement k means clustering to LEACH protocol?

Hi @Utsav.you can integrate the code of k-means in leach.tcl,, recompile, make, make install

i'm not working on tcl i'm working in matlab

you can see this link, it gives idea :

https://www.mathworks.com/help/stateflow/ug/procedures-for-simulation.html

good luck

@Hasan , do you know of clustering mechanisms other than K means?