You can make some assumptions like a typical FM receiver needs -110 dBm to work. Then assume you have an isotropic antenna in both cases because you didn't say anything about the antennas so we'll ignore the gain.
Next take a look at the path loss based on the 910 W (+59.6 dBm) power.
Your path loss can not exceed 59.6 + 110 = 169.6 dB (with loss dB and dBm are the same, but dB and dBm to watts is different).
The free space loss model is
$L_{fs}=32.45 + 20*log(d_{km}) + 20*log(f_{MHz})$
You'll need to solve this for $D_{km}$ and set $L_{fs}=169.6$ and $f_{MHz}=89.7$.
There are many other path loss models. NIST suggests several
- Free Space Model (shown above...least accurate)
- CCIR
- Hata
- Walfisch-Ikegami (WIM)
Having the antenna height will reduce your path loss some. In the popular WIM model in a line of sight (LOS) situation like yours where the base antenna is >30m high with no obstructions in the direct path then you can use a more realistic estimate:
$L_{wim-los}=42.64+26*log(d_{km})+20*Log(f_{MHz})$
Also the ITU has a specification ITU 1546 Method for point-to-area predictions for terrestrial propagation (pdf here). You can also find MatLab solutions and a lot of literature on their technique and field test results for broadcast FM. (FYI the models are all very similar in structure to NIST's models).