|
Re: how to persist and instantiate models?
Thanks again, Si. Left some comments and suggestions on the pull request.
Phil
|
|
|
Re: how to persist and instantiate models?
That sounds good. We'll take care of it.
Thanks again, Si. Left some comments and suggestions on the pull request.
Phil
|
|
|
OpenEEWeather not getting updated data
Hello,
We're testing out OpenEEWeather but are getting data that's out of date.
The weather station is Downtown LA/USC Campus, USAF 722874. We tried requesting for the last 30 days, but the last date available was November 22 -- over a week ago. Attached is the script we ran it with and the results we got.
Is this normal?
If so is this the best or "standard" service for getting weather for the CalTrack methods? Or is CalTrack independent of the weather service used?
|
|
|
Re: OpenEEWeather not getting updated data
We were expecting the data to be back on line by this point (usually it is one or two days behind, and maintenance was scheduled to end 11/29), so I went ahead and checked the source data and it appears that it was last updated on 2019-11-24, which is indeed unusual. The source file can be downloaded from ftp://ftp.ncdc.noaa.gov/pub/data/noaa/2019/, which lists the date the data was last updated. I expect it will be back to normal within a day or two. CalTRACK does not require use of the OpenEEmeter or eeweather libraries, but it does encourage use of public weather sources for repeatability.
If it works for you, I think it would be best to continue this conversation and other conversations about possible bugs on GitHub Issues. I think it will be better to have one place for all the bug-related conversations, and I think GitHub is the most easily discoverable. (OpenEEmeter -> https://github.com/openeemeter/eemeter/issues, EEweather -> https://github.com/openeemeter/eeweather/issues). Phil
On Tue, Dec 3, 2019 at 9:19 AM Si Chen < sichen@...> wrote: Hello,
We're testing out OpenEEWeather but are getting data that's out of date.
The weather station is Downtown LA/USC Campus, USAF 722874. We tried requesting for the last 30 days, but the last date available was November 22 -- over a week ago. Attached is the script we ran it with and the results we got.
Is this normal?
If so is this the best or "standard" service for getting weather for the CalTrack methods? Or is CalTrack independent of the weather service used?
|
|
|
Re: OpenEEWeather not getting updated data
OK, thanks for getting back to me. We'll use the github issues trackers in the future too.
We were expecting the data to be back on line by this point (usually it is one or two days behind, and maintenance was scheduled to end 11/29), so I went ahead and checked the source data and it appears that it was last updated on 2019-11-24, which is indeed unusual. The source file can be downloaded from ftp://ftp.ncdc.noaa.gov/pub/data/noaa/2019/, which lists the date the data was last updated. I expect it will be back to normal within a day or two. CalTRACK does not require use of the OpenEEmeter or eeweather libraries, but it does encourage use of public weather sources for repeatability.
If it works for you, I think it would be best to continue this conversation and other conversations about possible bugs on GitHub Issues. I think it will be better to have one place for all the bug-related conversations, and I think GitHub is the most easily discoverable. (OpenEEmeter -> https://github.com/openeemeter/eemeter/issues, EEweather -> https://github.com/openeemeter/eeweather/issues). Phil
On Tue, Dec 3, 2019 at 9:19 AM Si Chen < sichen@...> wrote: Hello,
We're testing out OpenEEWeather but are getting data that's out of date.
The weather station is Downtown LA/USC Campus, USAF 722874. We tried requesting for the last 30 days, but the last date available was November 22 -- over a week ago. Attached is the script we ran it with and the results we got.
Is this normal?
If so is this the best or "standard" service for getting weather for the CalTrack methods? Or is CalTrack independent of the weather service used?
|
|
|
Re: Live version of OpenEEmeter available?
Hi Steve,
Welcome and glad to have you. I apologize for my late response to this. The short answer to your question is no, Recurve no longer hosts the version of the OpenEEmeter that you describe, as at the time it was getting too little use to justify the cost and difficulty of maintaining it. Perhaps times have changed? If you're hoping to run the OpenEEmeter now, the best way to do so is probably to follow the tutorial, linked below. The tutorial is admittedly less convenient, but it does contain the same content and is considerably more powerful as it shows the user what is happening in the code and allows them to modify it to suit their particular needs. The tutorial can be found at http://eemeter.openee.io/tutorial.html. Let us know how it goes!
On Wed, Nov 20, 2019 at 12:38 PM Steve Schmidt < steve@...> wrote: Just joined the group! (Didn't know I was missing out.)
I recall that Recurve used to host a version of the latest eeMeter code that could be run on test data for individual buildings. Is this still available somewhere? I can't seem to find it.
Thanks.
-Steve
|
|
|
Re: Live version of OpenEEmeter available?
Steve Schmidt
Thanks for the reply Phil.
...at the time it was getting too little use ... Perhaps times have changed?
As CalTRACK usage grows and OpenEEmeter is part of more contracts I expect such a public version would get more use from a diverse group of interested parties.
My specific interest would be to use it as part of a sample CalTRACK test harness used to track improvements over time against a standard set of static project data. See CalTRACK issues #122 and #129 for background.
Any chance you could lobby internally to bring it back?
-Steve
|
|
|
What are blackout start and end dates in tutorial for?
Hello, I'm looking at the tutorial http://eemeter.openee.io/tutorial.html and would like to know what the role of blackout_start_date and blackout_end_date are? They're set to 12/26/2016 and 1/4/2017. Is this a holiday period? Why does the fitting of the model end on blackout_start_date
# get meter data suitable for fitting a baseline model
baseline_meter_data, warnings = eemeter.get_baseline_data(
meter_data, end=blackout_start_date, max_days=365
)
and the calculation of metered savings start on blackout_end_date?
reporting_meter_data, warnings = eemeter.get_reporting_data(
meter_data, start=blackout_end_date, max_days=365
)
|
|
|
Re: What are blackout start and end dates in tutorial for?
toggle quoted message
Show quoted text
On Tue, Jan 7, 2020 at 6:16 PM Si Chen < sichen@...> wrote: Hello,
I'm looking at the tutorial http://eemeter.openee.io/tutorial.html and would like to know what the role of blackout_start_date and blackout_end_date are?
They're set to 12/26/2016 and 1/4/2017. Is this a holiday period?
Why does the fitting of the model end on blackout_start_date
# get meter data suitable for fitting a baseline model
baseline_meter_data, warnings = eemeter.get_baseline_data(
meter_data, end=blackout_start_date, max_days=365
)
and the calculation of metered savings start on blackout_end_date?
reporting_meter_data, warnings = eemeter.get_reporting_data(
meter_data, start=blackout_end_date, max_days=365
)
|
|
|
Re: What are blackout start and end dates in tutorial for?
Thanks! This makes sense -- The model is built up until the start of the black out period, and then run starting from the end of the black out period.
On Tue, Jan 7, 2020 at 6:16 PM Si Chen < sichen@...> wrote: Hello,
I'm looking at the tutorial http://eemeter.openee.io/tutorial.html and would like to know what the role of blackout_start_date and blackout_end_date are?
They're set to 12/26/2016 and 1/4/2017. Is this a holiday period?
Why does the fitting of the model end on blackout_start_date
# get meter data suitable for fitting a baseline model
baseline_meter_data, warnings = eemeter.get_baseline_data(
meter_data, end=blackout_start_date, max_days=365
)
and the calculation of metered savings start on blackout_end_date?
reporting_meter_data, warnings = eemeter.get_reporting_data(
meter_data, start=blackout_end_date, max_days=365
)
|
|
|
Graph of hourly and daily models?
Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
Steve Schmidt
Just my opinion, but I think a chart showing the hourly energy usage in three categories (base, heating, cooling) would be more useful than all the different model parameters.
-Steve
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
Thanks for letting us know.
Yes, I agree it would be nice to get energy use by category, but how do we get that information?
On Thu, Jan 23, 2020 at 8:55 AM Steve Schmidt < steve@...> wrote: Just my opinion, but I think a chart showing the hourly energy usage in three categories (base, heating, cooling) would be more useful than all the different model parameters.
-Steve
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
The hourly models don't provide as straight-foward a means of disaggregation, as they construct separate model parameters for different temperature bins. So unlike the daily models where there is a single "cooling", "baseload", and "heating" coefficient, the hourly models have coefficients across up to 5 temperature bins. Additionally, the coefficients are different for each month.
On Thu, Jan 23, 2020 at 12:21 PM Si Chen < sichen@...> wrote: Thanks for letting us know.
Yes, I agree it would be nice to get energy use by category, but how do we get that information?
On Thu, Jan 23, 2020 at 8:55 AM Steve Schmidt < steve@...> wrote: Just my opinion, but I think a chart showing the hourly energy usage in three categories (base, heating, cooling) would be more useful than all the different model parameters.
-Steve
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
Interesting. I've been curious about what else besides M&V the CalTrack models could be used for. For example if you had heating, cooling, and baseload parameters, could you use that to compare across different buildings to determine which ones are more or less efficient? Is that what you were thinking?
On Thu, Jan 23, 2020 at 9:34 AM Stephen Suffian < steve@...> wrote: The hourly models don't provide as straight-foward a means of disaggregation, as they construct separate model parameters for different temperature bins. So unlike the daily models where there is a single "cooling", "baseload", and "heating" coefficient, the hourly models have coefficients across up to 5 temperature bins. Additionally, the coefficients are different for each month.
On Thu, Jan 23, 2020 at 12:21 PM Si Chen < sichen@...> wrote: Thanks for letting us know.
Yes, I agree it would be nice to get energy use by category, but how do we get that information?
On Thu, Jan 23, 2020 at 8:55 AM Steve Schmidt < steve@...> wrote: Just my opinion, but I think a chart showing the hourly energy usage in three categories (base, heating, cooling) would be more useful than all the different model parameters.
-Steve
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
Steve Schmidt
To determine efficiency you also need to know the amount of conditioned space. My intent is more to determine the accuracy of results against something that can be measured (like actual cooling energy).
On Fri, Jan 24, 2020 at 3:27 PM Si Chen < sichen@...> wrote: Interesting. I've been curious about what else besides M&V the CalTrack models could be used for. For example if you had heating, cooling, and baseload parameters, could you use that to compare across different buildings to determine which ones are more or less efficient? Is that what you were thinking?
On Thu, Jan 23, 2020 at 9:34 AM Stephen Suffian < steve@...> wrote: The hourly models don't provide as straight-foward a means of disaggregation, as they construct separate model parameters for different temperature bins. So unlike the daily models where there is a single "cooling", "baseload", and "heating" coefficient, the hourly models have coefficients across up to 5 temperature bins. Additionally, the coefficients are different for each month.
On Thu, Jan 23, 2020 at 12:21 PM Si Chen < sichen@...> wrote: Thanks for letting us know.
Yes, I agree it would be nice to get energy use by category, but how do we get that information?
On Thu, Jan 23, 2020 at 8:55 AM Steve Schmidt < steve@...> wrote: Just my opinion, but I think a chart showing the hourly energy usage in three categories (base, heating, cooling) would be more useful than all the different model parameters.
-Steve
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
Interesting. Do you mean comparing the model's cooling energy versus actual energy use of the equipment, for example with a virtual meter?
On Fri, Jan 24, 2020 at 4:11 PM Steve Schmidt < steve@...> wrote: To determine efficiency you also need to know the amount of conditioned space. My intent is more to determine the accuracy of results against something that can be measured (like actual cooling energy). On Fri, Jan 24, 2020 at 3:27 PM Si Chen < sichen@...> wrote: Interesting. I've been curious about what else besides M&V the CalTrack models could be used for. For example if you had heating, cooling, and baseload parameters, could you use that to compare across different buildings to determine which ones are more or less efficient? Is that what you were thinking?
On Thu, Jan 23, 2020 at 9:34 AM Stephen Suffian < steve@...> wrote: The hourly models don't provide as straight-foward a means of disaggregation, as they construct separate model parameters for different temperature bins. So unlike the daily models where there is a single "cooling", "baseload", and "heating" coefficient, the hourly models have coefficients across up to 5 temperature bins. Additionally, the coefficients are different for each month.
On Thu, Jan 23, 2020 at 12:21 PM Si Chen < sichen@...> wrote: Thanks for letting us know.
Yes, I agree it would be nice to get energy use by category, but how do we get that information?
On Thu, Jan 23, 2020 at 8:55 AM Steve Schmidt < steve@...> wrote: Just my opinion, but I think a chart showing the hourly energy usage in three categories (base, heating, cooling) would be more useful than all the different model parameters.
-Steve
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
Steve Schmidt
Correct, but not with a virtual meter. Best to use a submetering device like this: https://efergy.com/engage-sub-metering-kit/
The three CalTRACK methods (monthly, daily & hourly) can produce different results among themselves, and when compared to other tools. Until we compare results to ground truth we have no way to know how good each model is, and no way to know which aspects of the models need to be improved.
-Steve
On Mon, Jan 27, 2020 at 9:40 AM Si Chen < sichen@...> wrote: Interesting. Do you mean comparing the model's cooling energy versus actual energy use of the equipment, for example with a virtual meter?
On Fri, Jan 24, 2020 at 4:11 PM Steve Schmidt < steve@...> wrote: To determine efficiency you also need to know the amount of conditioned space. My intent is more to determine the accuracy of results against something that can be measured (like actual cooling energy). On Fri, Jan 24, 2020 at 3:27 PM Si Chen < sichen@...> wrote: Interesting. I've been curious about what else besides M&V the CalTrack models could be used for. For example if you had heating, cooling, and baseload parameters, could you use that to compare across different buildings to determine which ones are more or less efficient? Is that what you were thinking?
On Thu, Jan 23, 2020 at 9:34 AM Stephen Suffian < steve@...> wrote: The hourly models don't provide as straight-foward a means of disaggregation, as they construct separate model parameters for different temperature bins. So unlike the daily models where there is a single "cooling", "baseload", and "heating" coefficient, the hourly models have coefficients across up to 5 temperature bins. Additionally, the coefficients are different for each month.
On Thu, Jan 23, 2020 at 12:21 PM Si Chen < sichen@...> wrote: Thanks for letting us know.
Yes, I agree it would be nice to get energy use by category, but how do we get that information?
On Thu, Jan 23, 2020 at 8:55 AM Steve Schmidt < steve@...> wrote: Just my opinion, but I think a chart showing the hourly energy usage in three categories (base, heating, cooling) would be more useful than all the different model parameters.
-Steve
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
Re: Graph of hourly and daily models?
Do you feel a virtual meter would not be accurate enough?
Would this be something that the CalTrack group has or should be researching?
On Mon, Jan 27, 2020 at 10:34 AM Steve Schmidt < steve@...> wrote: Correct, but not with a virtual meter. Best to use a submetering device like this: https://efergy.com/engage-sub-metering-kit/
The three CalTRACK methods (monthly, daily & hourly) can produce different results among themselves, and when compared to other tools. Until we compare results to ground truth we have no way to know how good each model is, and no way to know which aspects of the models need to be improved.
-Steve
On Mon, Jan 27, 2020 at 9:40 AM Si Chen < sichen@...> wrote: Interesting. Do you mean comparing the model's cooling energy versus actual energy use of the equipment, for example with a virtual meter?
On Fri, Jan 24, 2020 at 4:11 PM Steve Schmidt < steve@...> wrote: To determine efficiency you also need to know the amount of conditioned space. My intent is more to determine the accuracy of results against something that can be measured (like actual cooling energy). On Fri, Jan 24, 2020 at 3:27 PM Si Chen < sichen@...> wrote: Interesting. I've been curious about what else besides M&V the CalTrack models could be used for. For example if you had heating, cooling, and baseload parameters, could you use that to compare across different buildings to determine which ones are more or less efficient? Is that what you were thinking?
On Thu, Jan 23, 2020 at 9:34 AM Stephen Suffian < steve@...> wrote: The hourly models don't provide as straight-foward a means of disaggregation, as they construct separate model parameters for different temperature bins. So unlike the daily models where there is a single "cooling", "baseload", and "heating" coefficient, the hourly models have coefficients across up to 5 temperature bins. Additionally, the coefficients are different for each month.
On Thu, Jan 23, 2020 at 12:21 PM Si Chen < sichen@...> wrote: Thanks for letting us know.
Yes, I agree it would be nice to get energy use by category, but how do we get that information?
On Thu, Jan 23, 2020 at 8:55 AM Steve Schmidt < steve@...> wrote: Just my opinion, but I think a chart showing the hourly energy usage in three categories (base, heating, cooling) would be more useful than all the different model parameters.
-Steve
Glad to hear it! Unfortunately, I haven't yet had the chance to build visualizations for the hourly model. There are a few challenges that make this a little harder than with the daily models. 1) There can be up to 12 separate models for 12 months of data 2) there are way more independent variables and parameters to visualize. It will certainly be possible to make nice visualizations but it will take some effort or a dedicated volunteer to help us get these across the line.
On Wed, Jan 22, 2020 at 3:12 PM Si Chen < sichen@...> wrote: Hello,
We're almost done with our openeemeter integration. We were able to generate a nice plot of the daily model based on the tutorial. Is it possible also to generate a plot of the hourly model?
|
|
|
