solarEngineering Model

[dpinney]

SPEC LINK See @dpinney.

STARTUP

• @drdanley writes initial spec.
• @dpinney will turn spec into task list and then find implementor.
• Implementor reviews spec questions david anywhere things are unclear. Issue passed back to David.
• Hello world interface.
• Add the rest of the inputs to the html template.
• Implement table with rows that can be added/deleted for PV interconnections.
• Get all inputs to disk in model directory.

PROOF OF CONCEPT FOR TIME TRAVEL CHART

• Get a pre-calibrated feeder.
• Add a group_recorder, check for correct specs.
• Run gridlab, get all output into a dict.
• What's the data volume requirement for PNGs? 2 MB per day. One voltdrop is 80kB. Let's say we do 24 hours. 24*80 kB = 2 MB. Average page size on the internet is over a MB, so this is not good.
• Do we send all pngs at once? No, too big, use send_from_directory.
• How big is too big for a PNG data set? 480 hours. We've got 2 models at 30MB and 45MB. Let's say 40 MB, which is 20 days.
• Figure out how to make the data structures fit together--translate to voltTable format then run most voltPlot code unchanged.
• Switch the time to 24 hours max.
• Add a switch to name the glm in the GLD solver. It's clearn for single-GLD-run models.
• Need specs of voltTable and allOutputData['xVoltDump.csv'].
• Re-architect to break on nodeVolts data structure. Shorter, cleaner implementation.
• Need +x+yj parsing function. Built in complex works!
• Draw first PNG. Takes about 8 minutes for laying out the graph but subsequent PNGs should be fast.
• Make 24ish PNGs. Loops are cool.
• The calibrated feeder has no geospatial data. Problem? No, because it does have the coordinated saved from a previous layout in the browser.
• zfill those strings. And fix the existing output.
• HTML interface.
• Chart title with timeStamp.

PORT TIMETRAVEL TO THE MODEL

• Current status? Runs fine.
• Restart, since we definitely need background processing? Yes, because we will have to merge less going in the other direction.
• Copy GridlabMulti to restart development, clean strings related to gridlabMulti.
• Remove multi-feeder options.
• Re-enable output charts.
• Fix hide/show.
• Insert group recorder instead of volt dumps.
• Switch to background running.
• Add calibration code. Not an issue.
• Switch imports to model style.
• Draw correctly via web server.

FINISHING

• Small changes.
  • Remove "Scada File Name" input field.
  • Move Study Details section to bottom.
  • Move "Model Runtime Statistics" content into right side of Study Details.
  • Add little finger from multiGridlab that lets us jump to the feeder editor.
• Voltage map slider control should have steps = allInputData.simulationLength.
• Attach solar in locations users input. Deprecated. Solar attachment in gridEdit now.

OTHER CHARTS

• Which ones are definitely needed? Powerflow, energy balance, cap and reg, climate. Take everything from GridlabMulti and narrow down from there.
• Write energy balance.
• Write cap and reg chart.

EXTRAS

• Enforce a maximum on the number of PNGs created, otherwise the model file size will be too large. 20 day = 240 hour max on PNG generation?
• Better color map for time travel chart. It should make clear where the ANSI band is (114-126 V equivalent).
• Switch user input from sqft to kw. solar attachment issue deprecated.

[dpinney]

Okay, we've got the spec version 1.0. See the first comment for location.

[dpinney]

Passing to @jinhw1989.

I realize you've got a lot of issues assigned to you now, so let me know if you need to send some back to me.

[cshjin]

@dpinney I am attending the symposium about grid on my campus those days, so I got a bit behind of schedule. Don't worry, I will catch it up. :watch:

[dpinney]

I'm not worried. I trust you. Thanks for the note.

[cshjin]

@dpinney please update date the SPEC LINK, I cannot get access to it.

[dpinney]

Sorry I screwed that up. Link in first comment is fixed now.

[dpinney]

Assigned back to me because the spec isn't clear enough.

[cshjin]

I can push _solarEngineering.py and _solarEngineering.html after I got the frame looks good.

[dpinney]

Uh oh, I just pushed models with those same names.

I renamed mine, so feel free to push yours with the names you gave. We can compare mine and yours today at 4. I suspect yours are better.

[cshjin]

Several questions when I throw it in Gridlab-D:

• What kind of nodes can add PV, adding a invertor to triplex_node and then a solar?
• The SCADA load data should be similar with cvrDynamic when integrate to gridlabd?
• The time travel
  • meter voltage of every node should be colored?
  • how to calculate the power flow in or not? I don't quite understand.
  • Should the time travel have graphs everty hour? static graph or dynamic?

[dpinney]

What kind of nodes can add PV, adding a invertor to triplex_node and then a solar?

Both node and triple_node objects should be available for users to attach solar to. You're correct about the inverter--it needs to have its parent set to the node/triplex_node, and then the solar needs to have its parent set to the inverter.

The SCADA load data should be similar with cvrDynamic when integrate to gridlabd?

Yes, the SCADA process should be the same as in cvrDynamic. It'll have to run through calibrate.

The time travel... meter voltage of every node should be colored?

Yes. In the same way they're colored in the voltageDrop model. This example shows it well.

how to calculate the power flow in or not? I don't quite understand.

Add a currdump object to the Gridlab model. Every line with a negative current = reverse powerflow, should be colored red. We should probably test this out with a model we know has reverse powerflow: the 90 percent solar Gridlab file which is part of this model has reverse flow.

Should the time travel have graphs everty hour? static graph or dynamic?

Yes, every hour. Static.

[dpinney]

Call Notes:

Attaching panels to nodes: spec asks for kW input, but GLD object takes sqft.

We should take sqft from the user and ignore kW for now.

How to get voltages for all nodes over time?

Two options: voltdump or recorders on every node. Let's use voltdump for now. runtime parameter on voltdump allows us to change time at which voltages are calculated. Let's use that to sample once each hour for 24 hours.

How to get tap changes for regs and caps out of GLD?

We can attach a recorder to a reg or cap, and then record the tap_a parameter (also _b, _c, etc.). We should also record the tap_A_change_count because we should compare that to the counts we calculate.

How to integrate calibration?

Don't run calibration as part of the model execution process. To test the model, use a feeder which has time-varying load objects already on it: Olin Barre GH.json.

[cshjin]

For hourly sampled volt of each node, I created voltdump like : {"object":"voltdump","filename":"voltDump_"+str(index)+".csv", "runtime": 24}, but it was unable to get the hourly data. How to write the voltDump or recorders structure in .glm? How to write those object properties?

object voltdump {
    filename voltDump.csv;
    runtime 24;
};

object collector {
    interval 3600;
    group class=node;
    file collector.csv;
    property voltage_A,voltage_B,voltage_C;
}

object recorder {
    interval 3600;
    parent CAP134;
    file recorder.csv;
    property voltage_A,voltage_B,voltage_C;
}

?

[dpinney]

Hmm. Test glm location?

[cshjin]

Here is the workspace: https://www.dropbox.com/sh/lh7my5x6lgwl6se/AABJkcLTySEhZep8QgPFFi-ba?dl=0

[dpinney]

The solution is to add a voltdump object for every hour you want to capture.

An example of how to do this for 2 hours is in /omf/omf/scratch/voltdumpDebug/.

The relevant section of the GLM:

clock {
    timezone PST+8PDT;
    starttime '2000-01-01 00:00:00';
    stoptime '2000-01-02 00:00:00';
};

//A WHOLE BUNCH OF OBJECTS HERE

object voltdump {
    filename voltdump1am.csv;
    runtime 2000-01-01 00:01:00;
};

object voltdump {
    filename voltdump6am.csv;
    runtime 2000-01-01 00:06:00;
};

[cshjin]

Got it. Thanks.

[dpinney]

My pleasure. Answering that made me feel useful.

[jcfuller1]

Another option is to use a group_recorder -- it will have its own issues with data, as it only records one property, but does record as a time series - may be easier to use, depending on your application. Example:

object group_recorder { group "class=node"; property voltage_A; interval 3600; file mine.csv; }

Output:

<

> Timestamp, node1, node2, ... 2001-01-01 00:00:00,120+0j,120+0j,... 2001-01-01 01:00:00,120+0j,120+0j,... ...

[dpinney]

group_recorder is superior to voltdump.

[dpinney]

Ownership to me. Proof of concept of time travel chart done.

[dpinney]

Latest progress updated in top comment.

[dpinney]

Updated top comment with Pooja's progress, a few new todos.

[dpinney]

Time travel chart in.

[PoojaSan]

• Voltage map slider control should have steps = allInputData.simulationLength. -> Completed • Enforce a maximum on the number of PNGs created, otherwise the model file size will be too large. 20 day = 240 hour max on PNG generation? ->Completed ,validation added for 1-240 hours

[PoojaSan]

• Better color map for time travel chart. It should make clear where the ANSI band is (114-126 V equivalent). –> changed the colormap to spectral from jet.Tried a couple of other colormaps as well. The change works only after changing the simulation Length and refreshing the page.Working on this.

[PoojaSan]

Hi David, I am working on attaching solar for the nodes specified by the user. I understand the concept of attaching recorder objects to capacitor/regulator but I am not clear about how would we attach solar object to nodes. I tried to figure out from available examples on gridlabd forums. Tried running a sample glm as follows

object node { name BROMLEY; phases ABCN; bustype PQ; voltage_A +7559.9995+0j; voltage_B -3779.9998-6547.1514j; voltage_C -3779.9998+6547.1514j; nominal_voltage 7200; object solar { groupid Solar_System; name Solar_House_A_1220; power_type DC; INSTALLATION_TYPE ROOF_MOUNTED; panel_type SINGLE_CRYSTAL_SILICON; generator_mode SUPPLY_DRIVEN; generator_status ONLINE; efficiency 0.18; area 20m^2; orientation FIXED_AXIS; }; };

But this didnt work out. I am posting my query on gridlabd forum. Meanwhile, can you shed some light on my queries? thanks,

[jcfuller1]

In pseudo-code (since SourceForge appears to be down again):

object node { name node1; ... }

object meter { parent node1; name meter1; ... }

object inverter { parent meter1; name inv1; ... }

object solar { parent inv1; ... }

[PoojaSan]

Hello jcfuller1, thank you for the response. that was helpful.

[PoojaSan]

Hi Jason,

In reference to my previous query -

after your response, I am trying to create a sample GLM file(just 1 node,1 meter,1 inverter and 1 solar) and run it through gridlabd. The moment I run the file, it stops working. It says gridlabd.exe stopped working. A problem caused the program to stop working correctly. Windows will close the program and notify you if a solution is available. I have no clue whats going wrong and your help is highly appreciated. Please find GLM code below.

module tape; module residential; module climate; module generators; module powerflow { solver_method FBS; NR_iteration_limit 50; };

clock { timezone PST+8PDT; starttime '2010-06-05 00:00:00'; stoptime '2010-06-05 00:03:00'; }

object climate { name "climate"; interpolate QUADRATIC; tmyfile "climate.tmy2"; };

object node { name node_meter_house; phases AS; nominal_voltage 120; object recorder { property voltage_A; file "measured_node_house.csv"; interval 1800; }; };

object meter { phases AS; name meter_test; parent node_meter_house; nominal_voltage 120; }; // Inverter object inverter { groupid Inverter_System; name inv_House_A_1220; parent meter_test; phases AS; generator_mode CONSTANT_PF; generator_status ONLINE; inverter_type PWM; rated_power 5000; // Nominal Power power_factor 1.0; inverter_efficiency 0.95; }; // Solar System object solar { groupid Solar_System; name Solar_House_A_1220; parent inv_House_A_1220; power_type DC; INSTALLATION_TYPE ROOF_MOUNTED; panel_type SINGLE_CRYSTAL_SILICON; generator_mode SUPPLY_DRIVEN; generator_status ONLINE; efficiency 0.18; area 20m^2; orientation FIXED_AXIS; }; Thanks, Pooja

[jcfuller1]

Well, that shouldn't happen. Can you send me the file (and the TMY file)? jason dot fuller at pnnl dot gov

[jcfuller1]

I'm trying to track down why this fails so "explosively" (i.e., silent failures don't make me happy).

The issue is "node" vs. "triplex_node" (or "meter" vs. "triplex_meter"). The former supports phase A, B, C, D, N, while the latter supports A, B, C, S, N. Basically, anything marked "triplex" assumes that the phasing is 120/240 instead of 3-phase.

So, you have a couple ways to fix it:

//  Change everything to the 3-phase system, or phase A
object node {
    phases A;  // A "node" is a 3-ph object, so does not support S-phase
    ...
}
object meter {
    phases A;  // A "meter" is a 3-ph object, so does not support S-phase
    ...
}
object inverter {
    phases A; // to align with it's parent meter
    ...
}

or

//  Change everything to the S-phase system, but still on phase A
object triplex_node {
    phases AS;  // A "triplex_node" can support the split-phase system
    ...
}
object triplex_meter {
    phases AS;  // A "triplex_meter" can support the split-phase system
    ...
}
object inverter {
    phases AS; // to align with it's parent meter
    ...
}

[PoojaSan]

Yes. that makes sense. I changed triplex node to node and this works fine. I could run the GLM properly. Thank you for your help Jason. I really appreciate it.

-Pooja

[dpinney]

Demo works. Closing.