However, the answers I did receive, especially the second one, were so much in line with what I saw on the weather file that it looks copied which I must affirm was not the case (see the answers at the bottom of this post).
So, the three things I found unusual about this weather file are, in order of ascending importance:
1. The year is given as 2005 throughout, but the comment says that it's a "typical year" file made from the time period 1970-2000.
2. Line 8 says that the first day of the weather file, i.e., Jan. 1, is set as a Sunday, but January 1 2005 was a Saturday.
3. The solar radiation appears to be shifted a half-hour ahead, probably because the original file was created in Europe, which tends to report the solar around the time step, e.g., -0:30 to +0.30, whereas in North America it's reported for the preceding time step, i.e., -1:00 to 0:00. Although that was the only criteria for why I chose this weather file (Legnica, Poland), I later found other more troubling aspects to the solar radiation that made me expand the discussion to QC'ing the solar on weather files in general.
I first became aware of the different conventions of reporting solar back in 1993 when I was involved in an IEA project on Low-Energy Cooling where one of the first items of business was to compare the climate conditions in the participant countries. When I tried to work with weather data from European colleagues, I was getting impossibly large spikes of Direct Normal Irradiance (DNI) for many sunrise hours due to this difference in convention, since I was using a US simulation program that follows the US convention and calculates the sun position at the midpoint of the preceding time step.
Although it might seem hard to tell which convention is being used in a weather file, it's actually quite noticeable if one were to compare the hourly profiles to those calculated by a Clear Sky Model. For example, this plot shows July 1 - 4, where the forward shifting of the solar on the weather file can be seen, particularly if you look for the sunrise and sunset hours. What really surprised me, though, was that the Global Horizontal Irradiance (GHI) and DNI seem switched on the weather file. The GHI profile is generally more smooth and the DNI more spiky, but here it's the reverse. Also, the GHI greatly exceeds the Clear Sky GHI on the morning of Day 3, which simply cannot happen.
The data for Jan. 1 -4 reveals more indications that the GHI and DNI on the weather file may be switched., with the GHI (?) again more spiky and on Jan. 2 almost three times as much as the Clear Sky GHI. The 30-minute shift is less visible owing to the small values, but can still be detected by looking at the sunrise and sunset hours.
So what does this say about QC'ing weather data? It seems that more emphasis has gone into QC'ing excursions in temperature than solar radiation. However, since the solar in almost all weather files is not measured but calculated, there should be all the more reason to regard it more carefully. Luckily, there are several simple, if not simple-minded, facts that can be used as reality checks, e.g., (1) the GHI should always be non-zero when the sun is above the horizon, making it possible to determine the hours of sunrise and sunset, (2) the GHI and DNI could never be greater than their Clear Sky values (don't worry about ground reflectance or atmospheric phenomena like cloud lensing since we're only dealing with calculated data), and (3) the Direct Horizontal ( DHI = DNI * arcsin of the solar angle) cannot be greater than the GHI.
I hope this answer has provided some insight into the contents of weather files. As in the previous contest, both respondents, Samuel Letellier-Duchesne and Michael Kummert, are declared winners and entitled to one free weather file of their choice from the WBT archive. Just let me know. (Full disclosure: Samuel appears to be Michael's graduate student :-))
Holiday greetings to all!
Joe
-------- Forwarded Message --------
| Subject: | Re: [Bldg-sim] Test your knowledge of simulation weather file formats Part 2: the EnergyPlus *.epw format |
|---|---|
| Date: | Fri, 15 Dec 2017 13:02:00 -0500 |
| From: | Samuel Letellier-Duchesne <samuel.letellier-duchesne@xxxxxxxxxx> |
| To: | Joe Huang <yjhuang@xxxxxxxxxxxxxxxxxxxxxxxx> |
Well, there is clearly something wrong with the sunup/sundown portion of the data as there is no systematically no transition for most days of the year. See image for comparison with us-based weather file. Will try to find more!
Polytechnique Montréal
Samuel Letellier-Duchesne
Ph.D. Candidate
Mechanical Engineering
Samuel Letellier-Duchesne
Ph.D. Candidate
Mechanical Engineering
2900 Edouard Montpetit Blvd,
Montreal, QC H3T 1J4 - Canada
Tel. +1-514-813-0056
On 12/19/2017 4:27 AM, Michaël Kummert
wrote:
Joe,
Here are my comments on the file that you asked us to examine.
Small details:
The EPW file appears to be a typical meteorological year but the year 2005 is given for all months, which is probably not correct. The file also has zeros in the minute field (5th field), while many typical EPW files have 60. I haven’t found a clear explanation of what that field is supposed to contain.
The main “problem”: solar radiation
Solar radiation data in this EPW file seems questionable:
- There is an apparent shift (roughly ½ h) in solar radiation reported in the file. This is apparent when comparing radiation in the file to calculated clear sky radiation
- The yearly average beam normal radiation is much lower than in other data sources (Meteonorm), but hourly values on some days are unrealistically high (in some cases twice the clear sky radiation)
- Calculated values for normal extraterrestrial (12th column) are not correct. That column probably shows another variable. On the other hand, calculated values for extraterrestrial horizontal seem correct
- This indicates a problem in the original data and/or in its interpretation and processing. If I had to guess, I would propose the following explanation:
- Let’s assume only total horizontal is available (estimated or measured). The shift in time causes total radiation in the morning to be very high. Some correlation similar to Erbs is used to estimate the diffuse radiation, and that correlation gives a small percentage of diffuse. Since the total radiation is very high, the calculated normal radiation on the horizontal is large, and the beam normal radiation is even larger.
- This could explain the large spikes in beam normal (mostly in the morning), but it does not explain why the beam normal radiation is too low in average.
Other variables (temperature, humidity, wind speed) seem to be usable. I didn’t check humidity calculations, I assumed the variables would be consistent.
I have put some graphs in a presentation which I may use one day in a class. I am attaching it in case you want to take a look at some of the strange solar radiation values vs clear sky radiation.
Michaël