Voltage variation during day are common, this depends on load and on transformer selected tap (where i work, i see this every day on our 220kV like).

As told by Medved, tap changing is made in HV-MV substations, but voltage variation depends on transformer tap numbers.
Substation transformers doesn't have only 2-3 tap, but can have many combination, so 5% is quite high as voltage variance (as i can see, where i work we have 1% steps on taps).
However, all depends also on designing choices and maintenance.

Edit: have you checked neutral - earth voltage? It could also be an issue with neutral connection.

Since my house is not grounded, I can't measure the voltage between phase and ground, I can only between phase and neutral. I don't think that this is caused by a failing neutral connection.

Since my house is not grounded, I can't measure the voltage between phase and ground, I can only between phase and neutral. I don't think that this is caused by a failing neutral connection.

There is no ground connection available at all..? That's unbelievable and would never be permitted here

What do you do for appliances that need it, such as anything with exposed metal..? 

在中国，标准电压220V，正常情况下在220-230V之间徘徊，最高时出现过235V，最低时出现过214V，电工说电压在198-234V之间属正常。

The secondary of the distribution transformer (HV to low voltage) have Earthed Neutral, that is the case in all distribution systems used for distribution for homes (except few oddball delta systems without a Neutral). This connection would be right on the pole with the transformer, with a wire going to the Earth

So, you definitely can measure voltage between Phase and Earth, just pick an Earth connection yourself instead of using the house Earth (if it is not installed) : Exposed building rebar, water pipe going underground, any wire or metal object buried for some depth in wet soil outside, etc...

You should read nearly your normal L-N voltage between L and this Earth connection, and no more than few Volts between Neutral and this Earth

If you dont, this is a sign of either :

 - Failing Neutral connection (the N-E voltage drop you measure is on the bad connection in the Neutral)

 - Stray currents in the Earth from a nearby fault (Phase current reaching to Earthed object nearby)

...water pipe going underground,...

I would not refer to any piping as an opportunistic grounding anymore. With all new constructions since mid 80's, the distribution piping ses to be made plastic, the same reaching older buildings parts with any piping reconstruction. So today only fraction of only older buildings have the piping from the distribution metalic. Even when the piping within the building is "good old steel" (there is really a reason for the quotas), the piping is in reality not grounded at all. Even when the electrical code requires that.
The problem is, the electrical codes allow the metallic water piping to be used as grounding electrode, but the water work is not done nor inspected (nor required to be inspected) by any electrician, so mostly no grounding straps are installed after such piping reconstruction, making the building to effectively loose grounding electrode.

It is meant as a reference to measure L-"PE" voltage with a multimeter, not to provide PE to the house. Unless it is both disconnected from Earth completely and have some leakage towards it from L inside the house, it gotta be sufficient

It is meant as a reference to measure L-"PE" voltage with a multimeter, not to provide PE to the house. Unless it is both disconnected from Earth completely and have some leakage towards it from L inside the house, it gotta be sufficient

It is another isolated conductor leading to the house, so it ay carry in quite a nonzero potential (from some phase-water electrical leak somewhere) and cause false alarms.
Normally the plastic distribution piping shall be "accompanied" by a dedicated grounding strip, which is supposed to lead into each property (water meter well). But because "it is electrical and not piping", water installers often do not touch it and so it is left unconnected.
This strip would be very good ground, but you would have to be sure it is properly connected.

Update: The voltage went down to 200V in my hostel now, and my Osram/Ledvance LED filament lamp vibrates in brightness.

Update:

Since my house is not grounded, I can't measure the voltage between phase and ground, I can only between phase and neutral. I don't think that this is caused by a failing neutral connection.

Pick a long metallic screwdriver, put it in outside garden ground (wet), measure voltage between live and the screwdriver with a multimeter and insulated terminals (certified CAT II or III).
It won't be like a proper grounding, but for measurements, it will be better than some random pipes.

If is a failing neutral, you will notice nearly constant voltage between live and screwdriver, but undervoltage or overvoltage between neutral and line.

However, if your home doesn't have any grounding, it's better to call some qualified electrician to renovate electrical system in your house.

Maybe that voltage variation is causing  that my halogen bulbs for 220V are lasting nothing. 

Maybe that voltage variation is causing  that my halogen bulbs for 220V are lasting nothing. 

Well, if you operate a 220V rated incandescent (even more a high temperature one) at 10% overvoltage, there is no wonder the life halves.
And with the halogens the local overheating at warmup, normally causing quite strong cycle time limitation, becomes way more damaging too...

A Variac is really useful for things like this. I run all the 220V lamps I have on mine, it obviously works for all voltages up to our line voltage of 245V as well. Well worth the investment.

If that's the application then no need in a Variac, just 230V:24V isolation transformer for 1/10th of the power rating of the lamps you want to run...

If that was the only application then possibly, but we have lamps of many different voltages from 12V up to 240V and it is much better to wind up the voltage gradually, especially on old/rare lamps so as not to subject the filament to too much shock on power-up.