Epever says you need to address them differently and wire them together when they are used in parallel, presumably so that they don't have spikes like this.
It appears as though you haven't got them sharing data so I would expect the issues you are seeing.
I could be wrong, but I seriously doubt this answer.
When just connected together (addressed differently) via RS485/Modbus the charge controllers do not "share" any data with each other unless there is some kind of other (Modbus) "master" device also connected to the same (Modbus) bus. Otherwise the charge controllers only "listen" for requests from a (non-existent) "master" device.
The "solar station" desktop computer software or mobile phone apps can be used to send those "master" commands to the bus, but in my observation they only send (write) commands as a result of user actions - and (read only) queries to monitor the status of the devices.
You might be referring to connecting the units via the PAL-ADP-50AN...
I got a couple of the PAL-ADP-50AN to experiment with. Based on the product description and (Modbus) traffic I've captured between the PAL_ADP-50AN and the charge controllers, I believe this device only helps with a few specific cases;
- Setting parameters on all connected devices
- Coordinating equalization charge cycles
- Coordinating charge termination
- Load balancing between charge controllers (By reducing output current as part of coordinating charge termination)
Those are cool functionalities, but in most cases not necessary (With the possible exception of large flooded lead-acid banks that actually need that equalization charge coordination).
It's all low speed coordination - happening at the RS485 baud rate of either 9600 or 115200 with response times and timeouts measurable in at least the tens of milliseconds range. That is, there's no high speed coordination of the MPPT loops or anything like that going on. (contrast this with something like the Victron Multiplus inverters which have to do high speed coordination to synchronize their output waveforms.)
Maybe the functionality of load balancing between charge controllers could help a bit with
reducing the chances of spikes under changing cloud cover and/or rapidly varying system loads. However, I don't think it would
solve the problem if the Epever SCCs are incapable of dealing with a given level of rapidly changing irradiance or power demand.
Note: I'm no EE, just an interested amateur. That said, if you have information to contradict what I've said here, please cite your sources (where relevant) and explain what you expect to happen in terms of MPPT event loop cycles and the state of the SCC's internal (output side) PWM voltage regulation.
