Recently, several people have brought a published paper by Dr. Daryoush Allaei, inventor of the Invelox ducted wind generator, to my attention. Here’s the opening statement of the published paper to get a sense of why it might be worth questioning:
A new concept in wind power harnessing is described which significantly outperforms traditional wind turbines of the same diameter and aerodynamic characteristics under the same wind conditions and it delivers significantly higher output, at reduced cost.
The paper makes the following claim:
Fig. 10 shows the daily energy production improvements of INVELOX with respect to the traditional WTG system. The results show INVELOX generated 80–560% more electrical energy than the traditional WTGs. P-Day 8 means partial data was collected on the eighth day. The total average energy production improvement of INVELOX over 8 days is about 314%.
This is an extraordinary claim. Modern horizontal axis wind turbines achieve 80% of Betz’ Limit. 80% more energy would imply exceeding Betz’ limit by an extraordinary amount, and 560% more electrical energy would imply exceeding the total energy available in a volume of wind by a considerable amount. Extraordinary claims require extraordinary evidence.
I’ve written briefly about the Invelox before in my ongoing assessment of good and bad bets in wind generation innovative technologies. The Invelox didn’t fare well:
This product and company score eight out of thirteen red flags. They’ve racked up a couple of million in grants and investment as well, and are still getting press as if they were somehow promoting something that hadn’t been failed 90 years ago per Robert W. Righter’s book “Wind Energy in America: A History.”
The odds: bad bet
However, that assessment is a fairly straightforward set of thirteen questions useful in identifying red flags about new wind generation technologies based on fairly quickly assessable items such as claims to exceed Betz’ Limit and principals with no backgrounds in wind generation.
The Sheerwind had never made any of their data and assessments public, so there was no meat to chew on, just hyperbolic statements in the popular press not dissimilar to the oddly aggrandizing statement that opens their paper. This published paper finally provided the meat.
There was a challenge however. Most of the paper deals with computational fluid dynamics modelling (CFD), an area where I don’t pretend expertise, but merely know a couple of simple rules of thumb which are useful. The paper was developed with a Professor of Energy Research of the Department of Mechanical Engineering of the City College of New York, Yiannis Andreopoulos. Thankfully, one of my regular, if anonymous, correspondents is an expert on CFD having done helicopter blade design CFD studies, as well as using CFD to assess wind generation approaches. He prefers to fly under the radar personally, but agreed to let me publish his thoughts on the paper with the proviso that I position them as a set of questions related to the paper, as if he were peer-reviewing it.
All content between the dashed lines is from my correspondent. I’ve edited his emailed remarks slightly for minor typos and formatting, so assume any errors were introduced in this process and are not attributable to him.
1. CFD Tool Assessment/Methods used for Invelox.
ANSYS and COMSOL are generally accepted tools for CFD analysis. Much of the effort in CFD revolves around screen-mesh sizing (higher meshing results in higher accuracy of estimation), method used for CFD (this is the actual aerodynamic priniciples expressed formulaically in code) and parameters set for high-speed computing (eg, wind-speeds, mass flow rate, surface friction, heat dissipations, etc). Generally a RANS solver (Reynolds Averaged Navier Stokes) is used to handle the high powered aero calculations which was used in this study.
Good things about the approach: ANSYS and COMSOL are excellent programs to use for CFD and the method – RANS — is standard. They also used Lewis Panel Method for axisymmetric potential flow modelling. That’s fair to do. A turbulence model was used, k-epsilon, which is very commonly used. So nothing crazy there, but I would have used the Spalart-Allmaras turbulence model, because k-epsilon is simplistic and has a host of uncertainty build into it. As far as the meshing is concerned they did model up to 3 million points, so that’s fair, but a higher point mesh for this machine (simply the size of it) would have been preferred.
a. Constant input velocity. Bad simulation. Real wind, close to terrain boundaries (ground) is erratic; it is not constant generally. The CFD should have been modeled with Spalart-Allmaras to include in the parameter turbulent free stream flow field with shifting wind-speeds and cross-winds for the site of the turbine. Although Betz also used constant velocity and pressure distribution across the actuator disc in his test, this is an ideal and never replicable for real-world operation. ANSYS and COMSOL are designed to check fluctuating input velocity and it was not setup to do so based on this report.
b. Steady state flow condition without a turbine inside is also bad for this CFD. And that is noted in the report for the CFD test in ANSYS and COMSOL. This will lead to uber-rosy results for any duct or venturi tube without blockage (prop, actuator disk, etc). They basically modeled air flowing real fast at a constant speed through an empty tube with the ground helping out for the concentration effect. Red Flag. One must do the opposite for 3-D simulations like these.
c. All the CFD was done without a turbine inside. From the report:
The CFD results are based on a steady-state formulation and therefore the model does not include the unsteady motion of the INVELOX system or turbine. Furthermore, these simulations did not involve any rotating turbine. The only meaningful comparison between CFD and experiments (i.e. field data) are the velocity distributions, speed ratios, and mass flow rates in the absence of the turbine.
At the very least, they should have used a screen mesh to model the rotors, and to provide pressure drop references for the rotors. I would accept that as a substitute for rotors. But if they are running ANSYS, it would have been wise if they had actually modelled spinning rotors in the CFC because ANSYS is set up to do that. That’s what ANSYS does. It tells you what to expect when you tell it what it is actually solving for. Red Flag.
2. Field Data Comparison to CFD Analysis
From the report:
In order to compare the field data with those generated by the CFD models, we collected wind speed data when the turbine was not placed inside the Venturi section of INVELOX.
and then in the very next sentence, it says:
Fig. 8a shows the measured free stream and Venturi wind speeds for 24 data sets with Sunforce turbine inside the Venturi section of INVELOX.
This is confusing. But nonetheless, they are presenting real data (or that’s what they call it) that shows agreement to the simplistic and challenged CFD analysis they performed. So I am skeptical based on this statement from the report:
In this paper, a small sample of the results is presented here.
I have no way of knowing how many days (is it 2 years, 1 year) of actual data and monitoring that was done to even begin to determine if the “small sample” size is significant enough to extrapolate. I would say no. The sample size presented in this report of real world data is not big enough. We need AEP and 12 months of data to assess any benefit long-term or short term of this machine. I have no way of knowing if the test data presented for real world results is true or not, since ANSYS and COMSOL were done without an operating turbine and power output extrapolated.
The other question I have is they use the phrase “when the turbine is inside the Venturi or Invelox”. Is the turbine actually spinning inside the Invelox. Is it producing power and what is the power output? It could be that the turbine may not actually be spinning inside the INVELOX during this test on Jan 2, 2013, but literally just sitting in there to block flow so they can align the test to CFD analysis (which was done without modelling a turbine). They do not show the wake flow velocity and pressure downstream of the turbine, and that is essential for power analysis. I have too many questions about the turbine ‘sitting” inside the tube as opposed to “operating” inside the tube. Nonetheless, it does appear that they demonstrated mass flow augmentation, which really is no big deal since mass flow augmentation is well known and previously documented over the last 50 years for ducted machines. What I am skeptical about is the actual benefit of mass flow on an operating turbine in the Invelox absent any wake flow data. They say that they did, but wake flow data is not presented and that’s the missing piece to determine if any useful power is being created. It’s capacity factor and LCOE that matter for modern wind energy.
The paper does not show the results for the un-ducted turbine either. So how can any comparison be made by casual on-lookers like ourselves? There is no data presented for the un-ducted turbine. The report data just shows a % increase between Invelox and the micro-wind turbine. What it does show, if the test data is true, although grossly incomplete, is that Invelox just made the most expensive and biggest 600 watt turbine ever. Look at the size of that thing … 600Watts! If they want more power out of the Invelox, then they will have to make a bigger duct system. The un-ducted machine can simply extend its blades at a fraction of the cost it would take Invelox to make their entire duct work bigger to increase power or better yet, just add solar panels to the un-ducted machine and the Invelox is beaten (if their test data isn’t rigged). Take note also, the un-ducted micro-turbine is on a 10m tower and the Invelox is at 18m – 8 meters higher. That’s not a good comparison. The un-ducted machine is also material efficient orders of magnitude than Invelox.
Invelox is still not displacing Big Wind, anytime soon, or anytime at all.
I will clarify something regarding my correspondent’s comments. The appropriate comparison is between the swept area of the duct entry and the swept area of the unducted turbine. That’s an apples-to-apples comparison of generation from a given volume of air. So while my correspondent pointed out clearly the disparity in height of the openings, he only implied the challenge of not comparing actual swept areas.
My correspondent’s extensive remarks are the type of thing that would have been good to have been caught in peer review, especially in a journal like Energy. I had hoped that he would be amenable to submitting them as a published critique in the journal, but he prefers not to.
Based on the weaknesses in their methodology and the challenges with their inappropriate comparisons, I cannot agree with the author’s conclusion that:
INVELOX has a strong potential and is worthy of further development.
In the hopes of stimulating a useful discussion, I am extending an invitation to Drs. Allaei and Andreopoulos to provide their responses to my correspondent’s expert critique here, just as many of the leading researchers in the area of airborne wind energy provided extensive comments on my assessment of the engineering compromises involved in that area of wind generation innovation.