The root problem here is that OP does not have net metering, where the utility charges 14 cents a kWH for power consumed but credits 14 cents a kWH for solar power generated. Utilities hate net metering because of the Duck Curve. They want to pay only 1-2 cents because of the general worthlessness of morning solar (at one point, California paid Arizona to take their surplus, since the reservoirs were dry and could not do pumped storage).
Given the anti-net-metering stand of the utility, you want to avoid selling power to the utility then buying it back. Thus the goal of directing solar to a storage load.
Is it net metering within the measurement interval?
If so, that's super easy.
But I don't believe this, and here's why. A typical house draws less than 100W most of the time, yet (in America) averages 1200W. That's because you have many large 3-6kW large loads banging on/off at intervals, waking up, running for 5 minutes and quitting. One or another is running several times an hour, so most windows would see their solar gobbled up by one of those periodic loads. So functionally, this would still be net metering, and that's exactly what the utility is trying to get away from*.
But OK. What if it is net metering within the 15, 30 or 60 minute sampling window?
Well, then, it's easy. Pick a higher frequency than that (say: 5 minutes) and do slow-format PWM - for 10% duty cycle turn the heater on for 30 seconds every 5 minutes. In any given 15 minute sampling window it'll have pulled 2000W for 1:30 and nothing for 13:50, exactly matching the 200W solar output in net.
You could use a contactor for that, but you could avoid all the clacking by simply using a triac and holding the triac open continuously for the ~30 second period. No sinewave chopping necessary; no power factor concerns.
If the meter is smarter than that...
A dimmer won't play with a quality meter (unless the power company wants to tolerate this). Already discussed here; let me hit the high points. Triac dimming is time-based. It sends power to the incandescent bulb some of the time.
So it's a 10 millisecond cycle instead of the 10 minute cycle in the straw-man above: but it's exactly the same problem. Electricity doesn't store for 10ms. You would simply be selling 200W to the utility for 90% of the "meat of the sinewave" and buying 1800W for the 10%.
The good news is, the solar inverter will not be degraded by the lousy power factor of the PWM controlled water heater, since its lowest impedance coupling is to the grid itself. The grid will see a weird PF, but whether that bothers them depends on how much the utility polices for that. I don't see them hunting you down for 2 kW :)
So what are the options?
The off-grid method: Dump terminals
To start with, off-grid solar people have been doing exactly this for a good long time. They have DC string solar charge controllers with a set of terminals called "DUMP". Excess solar that won't fit in the batteries gets fed out the DUMP terminals and they have a variety of ways to hook a water heater to that. (Careful with the water heater's thermostat; it's not rated to interrupt DC.)
So in this case, you would need to add an off-grid solar charge controller to replace your grid-tie inverter, a "token" battery merely to get the charge controller to function, and then attach the water heater to the "Dump" terminals. When the water heater's thermostat calls for heat, you'd switch the solar array from being connected to the grid-tie controller to the off-grid controller.
This has the advantage of being a COTS solution you can just buy; though skill will be required to make it work, as is so with all off-grid stuff. And it's not an option for "microinverter" solar panel setups.
Legit battery system with "Dump" to grid
Alternately, assuming your solar panels are DC, you can build a "legit" battery/off-grid solar system, same as you would for a remote cabin. Have the solar charge controller charge the battery, and hook up the water heater as a load driven off the battery (possibly through an inverter). The water heater would be fed exclusively from battery. It would draw normal currents, and would not need modification if used with an inverter.
You add some intelligence to control the water heater. When battery charge is above 90%, the water heater inverter spins up and runs the water heater until battery charge drops to 60%. Then it shuts off and lets the battery recharge. This repeats as long as there's solar.
* Varying cutoff percentage to suit the battery type.
Now, what do we do about the pesky utility grid? (ignore it, if it doesn't pay enough?) Well, that off-grid solar charge controller will have our old friend the Dump terminals, so this time, we direct Dump to the "Grid-Tie inverter". Thus, it sells power to the utility only when it doesn't have anything better to do with it.
The battery can be any size you want or find practical; the size of the battery will only influence the cycle interval A very small battery might have the water heating running 30 seconds every 5 minutes; a larger battery perhaps 18 minutes every 3 hours.
Again, all the heavy lifting is being done by COTS equipment you can just buy; your only original craft will be the battery monitoring that turns on the water heater as the battery nears full.