If you're only running one line 1500 feet with single-phase AC, then a variation of what X-10 did is about the best you can get with that chip (1ms pulse of signal is a ONE, lack of that pulse is a ZERO). I used the Philips NE5050 (similar tech) 15 years ago, and was underwhelmed with it's performance. Looks like they've corrected some of those old problems with the TDA5051, but it's still ASK at ~100KHz over a 1 to 5 ohm effective impedance (dirty) AC network. As long as you're able to deal with VERY low bit rate, you can get *maybe* a few bits out around the zero crossing each AC cycle. The AC lines are relatively quiet for about 2-3ms after the zero crossing, so ~240 to 360 bits per second. If the noise is too high (lots of triacs or other switching near the zero crossing), you might be limited to one bit per zero crossing / 120 bps raw, not counting overhead.
If you have multiple 'thingies' talking on that 1500 foot line then you'll want some sort of CSMA/CD protocol, where you listen to your outgoing signal to see if it's being stomped on by another sender, at which point you back off and wait for 50 to 200 cycles of AC before trying again.
You're probably locked into the TDA5051 due to time and money, but I think you'd probably have better luck over long-distance with the FSK chips. With X-10 there was a heck of a lot of redundancy to make sure that something happened at the far end... theirs was a 'dumb' network, where the slaves generally couldn't reply, only receive. That meant the error-proofing had to go in up front, as they couldn't ACK a good packet.
And jeezus... if you're already using an ESP32 to drive this beast, why in the heck not WiFi?? You're inventing a horse and buggy all over again, when the rest of us are flying at 80MPH. For two nodes under $5 each, you can't beat the price. Add half-wavelength antennas and they'd do 1500 feet.