Filter

You might try one of these isolators...

http://www.radioworks.com/ct-4.html

Have you determined that it is a conducted emission from the refrigerator, or a radiated emission?

What is the nature of the noise on the radio, e.g. constant whistle indicating a continuous wave noise.

Is the refrigerator a thermo-electric, or compressor type?

Does this refrigerator have digital temperature control that is causing this?

Is there an inverter that converts the +24DC to AC power producing the noise?

My questions are to determine the nature of the noise. You need to figure out if it's conducted, then a power line filter may block it, but if radiated, then RF shielding is needed.

Ignator

The interference appears only when the compressor is turned on.

The refrigerator uses a compressor bike?

This refrigerator not have digital temperature control.

There is no inverter that converts a + 24DC to AC power, the power is directly from the battery, because the compressor is 24vdc.

If the noise "tracks" (increases or changes frequency) with the refrigerator load, then noise source is "commutation" noise from the DC-motor.

An oscilloscope will visually "show" you what the "noise" looks like which should then enable you to define and select the proper/necessary DC-filter.

Ignator

My idea is exactly that, ie I will use a ferrite bead, which will make two coils, each of enrrolando 5 enameled wire turns of 16AWG. Each of these coils will be in series with the power cables (one on the + 24V and -24V in the other. And in parallel with these coils put two capacacitores not polarized 470nF, one at the entrance and another at the exit. What youYou think of this?

I would give it a try. As long as your power source has a fuse, your home made pi filter can't hurt anything. If the .47 micro farad is not helping, try another one of smaller value in parallel (keep leads short as lead inductance makes small capacitors ineffective), and a large one to give some decoupling energy for the high current pulses. Long wires of the power are inductive. My rule of thumb is at 100Mhz each inch (~25mm) of wire is 25 ohms impedance. So long wires will let RF 'float' on them.

If you could ground the chassis of this refrigerator with a wide brass strap (20+mm), that would be good. Those flexible ground straps made from minions of small diameter wires, braided into a strap really are bad at RF frequencies, as the RF flows on the surface and skin depth of many parallel conductors, and looks like lots of high impedance wires in parallel. Good for DC, but useless otherwise.

Also don't trust that the radio power input is not directly coupling the noise from it's power connection. A pi filter there may be useful.

One thing you could do is disconnect the antenna cable from the radio at it's input, and see if you can tune channels and hear noise. This will determine if it's a radiated vs conducted noise problem.

Don't transmit with the antenna disconnected unless you know the manufacture has designed it's amplifiers to not be damaged by open connection.