Improve PV pool power distribution based on current production

[llucax]

What's needed?

Currently, PV pools distribute available power based on inverter bounds rather than actual current production. This can result in suboptimal utilization, especially when grid limits are enforced and inverters have varying production capabilities.

For example, with three inverters and a grid limit of 30 kW, if two are producing ~12 kW and one only 5 kW, the actual output may be much less than the grid limit (e.g., 10+10+5 = 25 kW instead of ~12+12+5 = 29 kW).

A smarter distribution could get closer to the grid limit while respecting inverter capacities.

Proposed solution

A possible solution is an iterative or continuous redistribution algorithm that probes inverter current production and reallocates power accordingly.

For instance, initially split the requested power evenly according to bounds, then observe output and rebalance if some inverters under-produce and other produce close to the allocated power. Continue until reaching a stable, optimal distribution close to the requested power without exceeding it.

This could start with a naive approach and be refined over time.

Use cases

• Grid limiting scenarios with heterogeneous inverter production.
• Maximizing PV pool output without exceeding grid limit.
• Improving the responsiveness and adaptability of PV pool control to real-world inverter production fluctuations.

Additional context

• The issue was observed when testing the edge app with ugrid 0.18 and a setup with 3 inverters: 1,2, and 3.
• When grid limit was set to 30 kW, two inverters produced ~12 kW, one only 5 kW, yet all were assigned 10 kW, resulting in only 25 kW delivered.
• There is uncertainty about the availability of real-time production capacity info from the edge or inverter models, but if available it could help narrow bounds dynamically.
• This was not previously considered, as earlier implementations were made before real-world use cases emerged.
• A continuous, feedback-driven redistribution, e.g., try 11+11+8, then 12+12+6, stabilizing at 12.33+12.33+5.33, could improve results.