Grid structure

NOTE: This document is a draft, and it has been written by someone who may not have the required knowledge and experience in both PV-systems and electrical engineering. Feel free to correct mistakes, or to delete nonsensical content.

There are multiple ways to structure the micro-grid, some of which were mentioned in the D1 Micro-grid meet-up at CCCamp2019. These different structures depend on the scale of the grid (in terms of power, distance, and availability), but also on safety requirements and existing regulations, which determine what can be done practically.

Long-distance interconnections

Very low voltage DC can not be carried very far, as the voltage drop implies a huge loss in efficiency. Multiple solutions were mentioned for this problem :

• High(er) Voltage DC, around ~300V : tricky to work with (expensive equipment required), and not very safe (DC is hard to break, especially at such voltages), and requires a licensed electrician
• Standard 230V AC : hard to synchronize properly, but already standardized. Sync problems can be avoided by only using unidirectional point-to-point links. The interconnection would present itself as a DC consumer to one side, and as a DC provider to the other. This way, standard, off-the-shelf equipment can be used.
• "Sneaker-grid" : carrying batteries using mechanical/human power

It may be useful to exchange data (supply, load, battery level, etc) between these subgrids. In that case, standard 230V AC would be practical, as it allows PLC using off-the-shelf equipment.

Short-distance interconnections

Short-distance interconnections can be done directly using DC current. Because of this, it may be useful to use a standard voltage. 42V was used at SHA2017, but 48V is a good candidate too (multiple of 12, lower than 60V). Even with DC interconnection there are lots of variations possible :

• The current flow can be bidirectional, or restricted to a single direction using diodes.
• The connection can be passive, or active (opening the circuit when required, based on loads/supplies on each side of the connection)

Then, there are two ways to interconnect the sub-grids :

• Mesh-like structure : point-to-(multi)point links between the sub-grids
• Shared bus structure : point-to-point links between each sub-grid and a shared bus

Using this shared bus could help simplify the mental model and safety systems :

• Fault current calculation in a mesh-like structure can be tricky, and so designing the safety systems could be much harder (and more expensive).
• The shared bus can be viewed as battery (it can act as a load, or as a supply)
• Breaker/fuses can be placed on the interconnection between the sub-grid and the bus
• The bus would have to be sized to the maximum possible current flow between sub-grids (tricky to calculate), or have its own over-current protections.
• It should be possible to bypass this shared bus (directly interconnect two sub-grids, if the proper safeties are set up). However, requiring that connections are made to the bus could help ensure that the system is properly safe, and would probably simplify the fault current calculations.
• The problem with this approach is that the shared bus is a SPOF. A mesh-like structure could be much more resilient.

For active connections, it may be useful to exchange data between sub-grids, for which a transmission method has to be determined. During the meeting, it was said that DC Power-Line is harder to do safely, as one would have to separate the signal from the relatively high DC currents.

PV-system interconnection

The previous section covers short-distance subgrid interconnection, where each subgrid can be seen as a battery, but does not provide a solution for inter-grid PV-system interconnection (a concrete problem encountered by many at CCCamp 2019). Solutions to this problem were not discussed.

• A possible solution could be to avoid using batteries for each PV controller, and instead use a battery bank on the shared bus. Again, this creates a SPOF. This also makes the system "simpler", as the shared bus is already mentally modeled as a battery. This would require a load/supply controller at the batteries. Note that "power availability" cannot be solely determined from the actual load : if some batteries are charging, there is "available power", but it would not be apparent in the bus voltage. Because of this, this load/supply management circuit would need to communicate an "available power" metric to the possible loads/users. Also, this load/supply controllers would have to be able to control the generators (PV systems) in order to avoid excess supply when batteries are full, unless this excess supply can always be used by some Power-to-X load. It may be possible to use the bus voltage to communicate this situation to the generators, but the load/supply controller may be more efficient if it has a precise idea of the state of all generators (independently from the loads), which cannot be communicated by the shared bus voltage. The disadvantage of this solution is that it requires an additional battery charging circuit (+ the load/supply management system), because the existing PV-systems battery charging circuits can not be made to work with this shared bus.

• It should be possible to have multiple battery banks on the shared bus (eliminating the battery-bank-SPOF, but not the shared-bus-SPOF) if the load/supply circuits for each bank can properly cooperate. In that case, and if the load/supply controller communicates the excess supply situation to the PV-systems using the bus voltage, the voltage drop will have to be taken in account, and could even make this approach infeasible, requiring OOB communication between PV-systems and load/supply controllers.

• It should also be possible to have multiple battery banks in a mesh-like structure (eliminating both SPOFs), but the load/supply controllers would have to be aware of the mesh structure and of its capacity (the definition of "available power" becomes much more complex, because it has to take in account the limitations of the path. in a shared bus structure, the "available power" is a byzantine problem (everyone has the same view of it), while in a mesh-like structure, "available power" has a different definition for each node of the mesh).