coulomb wrote: Err, if that's the way it will work.
Duh. I think I have finally realised how they probably intend it to work, at least for the load shifting scenario.
I think they assume a common (but not universal) situation of a grid interactive inverter connected to a relatively high voltage string of PV panels, around 400 V. A small or zero feed in tariff is assumed, hence exporting to the grid is assumed to be of little benefit. The power wall would connect across the output of the PV panel, which is of course also the input to the grid interactive inverter. During the day, it would somehow know that conditions are right to draw some of the power from the PV panels to charge the battery. The presumed MPPT in the grid interactive inverter sees this as weaker PV output, so it generates less power during this time. Hopefully, this will be timed so that the battery gets charger mostly when there is excess PV power available, or when mains power is at least off-peak.
Later, either at night or possibly during the day when household demand is greater than PV supply, the power wall exports its power through its internal DC/DC converter. (Note: DC/DC, not DC/AC.) It exports its power through the same wires it imported power from, i.e. the connection to the PV panels and the inverter input. This raises the PV panel voltage. If it's at night when there is no solar input at all, the panel diodes prevent flow of power into the panels. The inverter sees this voltage, and assumes it's solar input. The DC/DC in the power wall will be designed to have a similar V-I curve to solar panels, so the inverter can't tell the difference, and will use all the power that the power wall makes available. So the grid interactive inverter generates power to support the load. Hopefully, it will not generate so much that power gets exported to the grid; this is to support the grid and reduce power imported from the grid.
So, using Kurt's example, if the power wall is set to generate a maximum of 2 kW, and ignoring losses for simplicity, then with a 2400 W load (e.g. typical kettle), then 400 W will be drawn from the grid, where it would have been 2400 W without the power wall. The fact that there is still some imported power doesn't matter much, because eventually the power wall will likely have used all its energy supporting the grid, and from then on, all power consumption will be imported anyway.
With the higher power limit from the power wall, it could cover such peaks, avoiding the importing of power till later, which may be an advantage. For example, the peak might happen before off-peak tariff comes in. Or you might be using less power that night, so it's better to use more of the energy available shaving peaks to zero, rather than leaving energy in the battery so you get less benefit from it.
Because the converter [ edit: was "inverter") in the power wall isn't directly supplying the load, it doesn't have to be big enough to supply your biggest peak load. The grid will supply any deficit. It also means that during a blackout, even though you have energy capable of running your house, you'll still have no power. That would explain why Tesla seem to consider the load shifting and the UPS (uninterruptable power supply) to be so different. I suspect most people don't realise this, even though it's similar to the situation of a daytime blackout: even though the sun is shining and you have plenty of solar power available, without the grid, it all goes to waste till the grid comes back again. (Again, assuming the typical grid interactive inverter installation.)
The power wall arrangement for load shifting sounds a little like the Enginer system for Prius cars. The extra 48 V battery, through a DC/DC converter, supports the hybrid battery, making more electrical power available to move the vehicle, but not necessarily proving all the power needed at any one point of time. Over a typical drive, all the energy from the 48 V battery ends up reducing petrol consumption. If you happen to take only a short trip, then possibly you can't take advantage of all of the electrical energy you have in the 48 V pack. The cheaper and hopefully greener electrical power reduces the use of the more expensive and less environmentally friendly petrol. With the power wall, the excess solar energy that would have been exported at a zero to low price per kWh is stored, and used to prevent importing at a higher cost per kWh.
How this gets implemented is an interesting question. Figuring out when the best time to charge or discharge the battery is a difficult decision. The optimum solution, if there even is one, depends on many factors: SOC of the battery, present power demand, the present cost of grid power, the likely availability of solar power later in the day (weather), and so on. If you have a discretionary load like a hot water heater or a pool pump, then it gets more complicated again; you don't want to run out of hot water and/or you want your pool to stay adequately cleaned. This is such a complex decision that there is a niche market for software that does this properly; it seems to be called Demand Charge Management (DCM). So Tesla are a clever company; they could do this well, or possibly allow some sort of interface to the DCM system of your choice. But that's more than a simple plug and play system; at the very least, it would need access to the current electrical demand of the house. Maybe this could be as simple as a wireless device that your electrician clips over the right cable inside your meter box. I don't know how it would access a weather forecast; perhaps it just ignores that potential source of information.
If this was obvious all along, I apologise for the unnecessary long post.
[ Edit : added a paragraph re size of DC/DC converter; added question at end re weather forecast; DC/DC inverter -> DC/DC converter ]
Nissan Leaf 2012 with new battery May 2019.
5650 W solar, 2xPIP-4048MS inverters, 16 kWh battery.
1.4 kW solar with 1.2 kW Latronics inverter and FIT.
160 W solar, 2.5 kWh 24 V battery for lights.
Patching PIP-4048/5048 inverter-chargers.
5kW and 7kW peak brings it back into the viable option field, 2 of the 7kW units in parallel would give a 14kW peak load and 14kW storage capacity, enough to think seriously about going off grid. A 5kW solar system could recharge a drained pack on a sunny winters day, all those figures are within the realms of reality for every house and using mortgage redraw it would be within the reach of many households.... glad I don't have shares in any of the grid based generation mobs 
