Unexpected behaviour with the `weighted_sum` decision method [BUG]

[tsmbland]

Describe the bug

I've been getting some unexpected behaviour using multiple objectives with the weighted_sum decision method, so I just thought I'd run a few sanity checks.

Firstly, for reference, here's how the default model looks with LCOE as a single objective:

I would think that using the weighted_sum decision method and adding a second 'dummy' objective with a weight of zero shouldn't have any impact on the simulation, but it actually changes the results quite substantially:

Similarly, having two LCOE objectives with 0.5 weight also has the same effect on the results:

Am I being overly simplistic to think that these two tests should be identical to the default example, or is there something weird going on here?

Context

Please, complete the following to better understand the system you are using to run MUSE.

• Operative system (eg. Windows 10): windows 10
• MUSE version (eg. 1.0.1): dev
• Installation method (eg. pipx, pip, development mode): dev
• Python version (you can get this running python --version): 3.9

[tsmbland]

@dalonsoa

[dalonsoa]

This is indeed odd. I think it might be related to that normalisation [0, 1] that the weighted_sum method is doing.

I would suggest your debug the problem starting by checking the objectives and decision objects in the following method in both scenarios:

MUSE_OS/src/muse/agents/agent.py

Line 291 in 7261daa

decision = self.decision(objectives)

To start with, if you have two LCOE objectives, for example, the objectives objects should reflect two of the same thing, and both identical to the case of having just one LCOE. Assuming that is the case, then dig into the weighted_sum to understand how those two are transformed and why the result is different than a single LCOE.

Unless I'm also being naive and weighted_sum is actually expected to do something different, I think this should point us to what is wrong here.

[ahawkes]

Yup, this sounds very odd, would appreciate some digging!

…

________________________________ From: Diego Alonso Álvarez ***@***.***> Sent: Thursday, May 30, 2024 5:26:02 PM To: EnergySystemsModellingLab/MUSE_OS ***@***.***> Cc: Subscribed ***@***.***> Subject: Re: [EnergySystemsModellingLab/MUSE_OS] Unexpected behaviour with the `weighted_sum` decision method [BUG] (Issue #317) This email from ***@***.*** originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list<https://spam.ic.ac.uk/SpamConsole/Senders.aspx> to disable email stamping for this address. This is indeed odd. I think it might be related to that normalisation [0, 1] that the weighted_sum method is doing. I would suggest your debug the problem starting by checking the objectives and decision objects in the following method in both scenarios: https://github.com/EnergySystemsModellingLab/MUSE_OS/blob/7261daa746dd28e9814d9ea3c48e3cdc31530399/src/muse/agents/agent.py#L291 To start with, if you have two LCOE objectives, for example, the objectives objects should reflect two of the same thing, and both identical to the case of having just one LCOE. Assuming that is the case, then dig into the weighted_sum to understand how those two are transformed and why the result is different than a single LCOE. Unless I'm also being naive and weighted_sum is actually expected to do something different, I think this should point us to what is wrong here. — Reply to this email directly, view it on GitHub<#317 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AC37JLMMH5ZRODCOJLL55VDZE4ZHVAVCNFSM6AAAAABIPE3IVOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMZZGY4DMMRRHA>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>