Generating two-parameter models with EEmeter


Phil Ngo
 

Hi Ethan!

I'll point you to the code here - unfortunately this is baked in. It wouldn't be too tricky to add, though- happy to review a pull request.

This line shows the candidate model selection - you'd need to make sure the candidates included your two-parameter models:
https://github.com/openeemeter/eemeter/blob/2f99fdada349f2f5090089e534d8137ee6a82431/eemeter/caltrack/usage_per_day.py#L1844

This line shows how you might add a two-parameter model candidate - you'd have to tweak this formula and add another function to generate the model candidate:
https://github.com/openeemeter/eemeter/blob/2f99fdada349f2f5090089e534d8137ee6a82431/eemeter/caltrack/usage_per_day.py#L1460

Hope that helps,

Phil

On Tue, Jul 5, 2022 at 2:11 PM Ethan Goldman <ethan.goldman@...> wrote:
Here's a question (or perhaps a feature request) for others who might be using EEmeter for gas meter modeling: there is a flag to force EEmeter to ignore cooling energy (fit_hdd_only), but can you force it to generate a "two-parameter" model with just slope and intercept?

Here I mean a more traditional X-intercept, not a base load. For background, I'm modeling run-time data from a gas furnace (metering the gas solenoid valve), so I'm quite sure there is no non-heating "base" load represented in my data! (I'm using eemeter v.3.1.0, FWIW) 

Currently, EEmeter is giving me a model that adds about 20% more energy in the annual base load to a typical heating season for this area. It's also likely picking the wrong balance point, might be skewing the slope a bit, and (most importantly) doesn't actually produce a model that matches the physical reality of the data. 

I realize this is a bit of an edge case since I'm technically modeling sub-meter data, but it's pure HVAC and definitely weather-dependent, so it seemed like a reasonably good fit for EEmeter. Does anyone have an easy fix, or would I need to add some new functionality to the code to get this behavior to work?

Thanks,
Ethan



--

Phil Ngo (he/him)

VP of Engineering

801.244.9860 LinkedIn

 

Recurve.com Newsletter LinkedInAngelListTwitter


Ethan Goldman
 

Here's a question (or perhaps a feature request) for others who might be using EEmeter for gas meter modeling: there is a flag to force EEmeter to ignore cooling energy (fit_hdd_only), but can you force it to generate a "two-parameter" model with just slope and intercept?

Here I mean a more traditional X-intercept, not a base load. For background, I'm modeling run-time data from a gas furnace (metering the gas solenoid valve), so I'm quite sure there is no non-heating "base" load represented in my data! (I'm using eemeter v.3.1.0, FWIW) 

Currently, EEmeter is giving me a model that adds about 20% more energy in the annual base load to a typical heating season for this area. It's also likely picking the wrong balance point, might be skewing the slope a bit, and (most importantly) doesn't actually produce a model that matches the physical reality of the data. 

I realize this is a bit of an edge case since I'm technically modeling sub-meter data, but it's pure HVAC and definitely weather-dependent, so it seemed like a reasonably good fit for EEmeter. Does anyone have an easy fix, or would I need to add some new functionality to the code to get this behavior to work?

Thanks,
Ethan