Call frequency of CVL_ICONTROLLER

[cflenker]

Describe the bug

In former versions of the driver, the IController algorithm was called every time the manage_charge_voltage_linear() was called. As far as I know this is something link one call per second. In the latest version of @mr-manuel fork https://github.com/mr-manuel/venus-os_dbus-serialbattery it is also calculated at every call of manage_charge_voltage_linear() but self.control_voltage is only set every LINEAR_RECALCULATION_EVERY (default setting 60seconds) by using the set_cvl_linear-Function. For the behaviour and the stability of an discrete IController the call-frequency is essential. So by reducing the call-frequency from 1s to 60s has made the IController 60 times slower and less stable. I do not see any sense to reduce the call-frequency of the IController to 60s because the calculation effort is done every second anyway. The inital calibration (CVL_ICONTROLLER_FACTOR=0.2) is also tuned to fit a call frequency of 1/second.

How to reproduce

Charge to CVL

Expected behavior

IController is calculated every second.

Driver version

1.3.20240417dev

Venus OS device type

Cerbo GX

Venus OS version

3.30

BMS type

JKBMS / Heltec BMS

Cell count

16

Battery count

1

Connection type

Serial USB adapter to RS485

Config file

not relevant

Relevant log output

not relevant

Any other information that may be helpful

To achieve the old behaviour of the IController I changed LINEAR_RECALCULATION_EVERY from 60s to 1s. But this causes much more cpu load.

[mr-manuel]

Could you test with the latest nightly?

[cflenker]

Hello @mr-manuel, Thank you for your fix! The code looks fine now. I just updated my own system to v1.3.20240428dev. It will be sunny enough to reach 100% tomorrow. So I will be able to give you a feedback tomorrow evening. If it works fine we can close this issue and https://github.com/Louisvdw/dbus-serialbattery/issues/1038. Thank's a lot!

[cflenker]

It works. But not as smooth as with the old version 1.0.20240102beta. The CVL-controller does not lower CVL enough and so the cell voltages rise much more. I did not change anything in the config.ini I will try to find the difference.

[cflenker]

It's propably the rounding. An I-Controller increases/decreases only very little every timestep. The rounding seems to kill these increase/decrease if the cell voltage is only a little above MAX_CELL_VOLTAGE. I changed the rounding to 6 (instead of 2) digits. Let's see if it works better today.

[cflenker]

Now it works fine. The rounding has to be removed:

              if utils.CVL_ICONTROLLER_MODE:
                    if self.control_voltage:
                        control_voltage = round(
                            self.control_voltage
                            - (
                                (
                                    self.get_max_cell_voltage()
                                    - (
                                        utils.SOC_RESET_VOLTAGE
                                        if self.soc_reset_requested
                                        else utils.MAX_CELL_VOLTAGE
                                    )
                                    - utils.CELL_VOLTAGE_DIFF_KEEP_MAX_VOLTAGE_UNTIL
                                )
                                * utils.CVL_ICONTROLLER_FACTOR
                            ),
                            2,
                        )
                    else:
                        control_voltage = self.max_battery_voltage

[mr-manuel]

That means a rounding to 6 digits is not enough?

[cflenker]

That would be enough. But does it make sense? The display in Cerbo UI is rounded to 2 digits anyway.

[mr-manuel]

Yes, often there are useless roundings after calculations like 62.1000000001.

[cflenker]

Ok. Then 6 digits is fine. I tested with 6 digits.

[mr-manuel]

Fixed with v1.3.20240510dev.