Frequently Asked Questions
Q. Can I use the Solar Forecast Arbiter to analyze wind power forecasts?
A. Yes! When creating a new Site, select Power Plant and simply ignore the solar-specific information such as surface tilt. Then proceed to use the Arbiter as normal. We are exploring what would need to be done to better support wind forecasts in this issue.
Q. Can I hide the identity of a power plant?
A. Sort of. See our Anonymous Data blog post for a full explanation.
Q. Who do I contact if I have a suggestion or I found a bug?
A. We encourage you to create an issue on our GitHub repositories. We manage a few repositories but the
solarforecastarbiter-coreissue tracker is likely where you want to report. If don’t want to report the issue on GitHub then please email us at email@example.com.
Q. I’m having trouble accessing
dashboard.solarforecastarbiter.org. Any ideas?
A. Some users at utilities have reported similar problems. The best course of action is to ask your IT department as there is not a single solution that we are aware of.
Q. What’s the meaning of the parameter issue time of day for a forecast that is issued multiple times per day?
A. Issue time of day is defined only once even for forecasts that are issued multiple times per day. That situation, it defines a single issue time of day, then the
run lengthdefines all subsequent issue times. So it doesn’t really matter if you specify, say, 1 or 2 or 11 am for a forecast that is issued on the hour. But you could also do things like issue time of day = 0045, lead time to start = 75 minutes to represent the start of the CAISO real time market window. The flexibility is powerful but a bit confusing.
Q. Can I use Solar Forecast Arbiter for resource assessment?
A. Yes, if you’re willing to ignore a couple of word choices. The Solar Forecast Arbiter ultimately compares one time series against another. Typically, one time series is modeled (the forecast) and one time series is a measurement (ground irradiance or metered power). But if you consider the modeled irradiance to be a 0 lead time forecast then you can easily use the Arbiter to compute metrics against ground observations. Or you may choose to evaluate a forecast against another modeled irradiance product (e.g. NSRDB, SolarAnywhere, etc). Refactoring the forecast-specific data model to support a more generic modeled time series and thus avoid the word choice problems would be great but also a substantial amount of work.
Q. Can I instruct the Arbiter to create reference forecasts using NWP data or persistence models for my own sites?
A. Not currently, but we hope to add this feature for historical data. You can install
solarforecastarbiter-coreto run these models on your local machine and obtain true forecasts. We can also run custom reference forecasts in a trial.
Q. Is there a limit to the number of requests per hour that I can make?
A. The Arbiter API does not currently impose a limit. For requests of reasonable size (i.e. response time is much less than request frequency) you can hit the server as often as you want. However, the Auth0 API does impose a limit, so if you are obtaining a new token with every request then you may be limited by Auth0.
Q. Should I obtain a new token with every request?
A. Generally you should only obtain a new token when your current token is nearing or past its expiration. Obtaining a token more frequently than is necessary runs the risk of hitting Auth0’s rate limit.
Q. Can I upload data to multiple forecasts or observations at once?
A. No. A separate POST request must be made to each forecast or observation endpoint. The benefit is that the format is exactly the same for all uploads, so there is no need to worry about any nuance in file names, csv column names, or json keys.
Q. How do I associate my data with UUIDs?
A. A brute force answer is a hardcoded dictionary or other mapping with keys that are your e.g. PI names and values that are the SFA UUIDs. Another option is to record your own resource name in the SFA’s observation metadata extra parameters field, then search/filter the observations for the right one/ones, then pull the UUID from the found object.
Q. Do you have any examples?
A. The API documentation contains a number of request and response examples. For examples using the Python API wrapper, please see the data_upload_download notebook in our workshop repository. You can actually run that notebook using MyBinder if you want a very quick start.
Q. I’m having trouble connecting to the API or requesting an access token securely. What gives?
A. If you’re behind a firewall, there is a good chance that is causing problems with verifying the SFA or Auth0 certificates. If you’re using the SFA core library or the python
requestslibrary, you can get around this by copying the appropriate PEM chain certificate to your machine and setting the
REQUESTS_CA_BUNDLEenvironment variable to the path to the certificate. These certs can typically be found by visiting the offending url in your browser and clicking the lock on the left side of the search bar. See this issue for more context.
Q. Any tips for trying to get data streaming from PI and into the Arbiter?
A. If you’re able to use Python, we’ve heard good things about PIConnect. Combine that with our Python wrapper of the API and hopefully you’ll be able to establish data feeds relatively quickly.
Q. How do I install the core framework?
A. Please see the installation instructions.
Q. I’d like to create reference forecasts on my own machine. Do you have any recommendations for how to make that work?
A. If you’re only using persistence forecasts then it’s relatively straightforward. If you want to use NWP data then it’s much more complicated due primarily to the dependency on
wgrib. The most reliable solution is to run the NWP data fetch in the Arbiter docker image. Please see docker, the core cli documentation, and run
solararbiter fetchnwp -hfor help. Once the data is available locally you can process it into irradiance or power forecasts using a
solarforecastarbiter-coreinstallation in your native OS.
Q. I’m interested in conducting a forecast trial using the Solar Forecast Arbiter. What do I need to get started?
A. There are a few steps to organizing a trial on the Solar Forecast Arbiter.
- See the Trials Documentation and consider the specifications of the trial and contact firstname.lastname@example.org when ready. The administrators can help design the trial and reach out to vendors.
- If you’re planning to provide measurements to forecast vendors (this is preferred), see the notes on Data Upload and consider the IT requirements of providing data when considering the timeframe of the trial.
- Send a signed Data Use Agreement to email@example.com if your company has not signed it.
Q. The metadata is inconsistent with other records and/or with observed data. What should we do?
A. Please let the trial administrators know of the inconsistency, but note that trials are always evaluated against the observed data regardless of the accuracy of the metadata.
Q. Can I access the trial data using my own account or must I use the anonymous account created for me for the trial?
A. You must log in with your anonymous account. We do it this way so that we do not need to retain any record of who’s who in our system, thus minimizing the amount of trust you must place in us as the administrators. You are the only person that knows which anonymous account is yours and only you can choose to reveal that to someone. If we instead extend the read/write permissions to your personal account then in principal we could retrieve the association by inspecting the permissions in the database (we would never do that!).