Back in December I stopped by the Boston Museum of science and checked out the public display of data from their small wind turbine experiment, and I posted some of the publicly available data here in this blog. Some of the turbines appeared to be working better than others, and I reported my best guesses about what was going on. I subsequently learned (I think in an email from David Rabkin, the project leader at the MOS) that the manufacturers of the Windspire turbine were sending a new one to replace the original, which had malfunctioned, and some time later I went by the museum and noticed that there was a Windspire that seemed to be operating normally. I also noticed that the Swift wind turbine (which I had never seen operating before) was sometimes running, so I assumed some repair had been made to it.
This morning I had the opportunity to attend a great conference put on by David Rabkin and his project team, where they presented their data, gave a summary of their experience with each turbine, and answered questions from the large (seemed like well over 100 people) and energetic audience. First, here is the updated data from the public display, in the same spreadsheet I used before. Because of the issues they have had with the various turbines, I have considered the cumulative data suspect, and for my comparisons (energy per swept area and approximate best case Cost Of Energy at this site) I have used just the data from the most recent month. Note that these numbers are based on my estimate of the cost of a ground-mount installation, not the MOS data for their rooftop installation, which I will include below. (Click for a readable version.)
The inferred COE from Windspire, Skystream, and Proven is slightly higher in this data than in the December data, probably due to lower winds compared to December. The inferred COE for the Aerovironment is slightly lower, perhaps reflecting a seasonal shift from NW winds – which the constrained yaw of these turbines can’t face – to SW winds, and for the Swift dramatically lower, probably reflecting the repair which was mentioned by MOS staff in the conference. Once again, the Skystream comes out on top, both in terms of energy production per swept area and inferred best-case COE, which was $0.41/kWh at this site when using my assumptions (no repairs or service costs, 20 year life). The graph at the top shows the power curves as displayed at the public display on the ground floor of the museum. The effect of the anemometer problem on the Proven turbine can be clearly seen as a rough cloud of points with unrealistically high windspeeds compared to the other anemometers.
Now, starting from the northwest end of the building (Cambridge side), here are the comments the MOS team had about each installation, as presented by Marian Tomusiak, who was in charge of the data:
The Windspire (7.4 sq. meters, 1.2kW nameplate) suffered unspecified mechanical problems that kept it from spinning up; a replacement turbine was installed on March 2 of 2010. The new turbine is able to come up to speed, but the MOS team reported that it has a safety feature that causes it to shut down for 30 minutes whenever the windspeed exceeds 27 mph – this cuts into production in higher windspeeds. Since replacement the Windspire has produced 26 kWh per month. As best I can tell, I was partially wrong in my assessment of the original Windspire at the MOS – while it was indeed malfunctioning and electrically braked by the alternator, the main shaft apparently was not bent. The disconcerting low-speed wobble I observed at around 20-30 RPM is a normal aspect of the turbine operation; it smooths out as it comes out of stall. The machine appeared to be running up to speed yesterday and it looked good.
The MOS crew reported no problems with the Skystream turbine (10.9 sq. meters, 2.4 kW nameplate). The turbine has produced 138 kWh per month. Over the entire period since the recording began in October, the energy production has been 939 kWh, which corresponds to about $0.36/kWh according to my assumptions given above. This is more than 2x lower than the nearest competitor (Proven). As I’ve disclosed here before, while I don’t have any financial stake in Southwest Windpower, I worked for the company for a few years around 2003, and contributed to the engineering design of Skystream – so I’m pleased that our work is holding up well in this experiment.
According to the MOS, the biggest issue with the Swift turbine (3.6 sq. meters, 1kW nameplate) is that the turbine is sited in a strong eddy, as confirmed by CFD modeling donated by ANSYS of Lebanon NH. They would like to move this forward to a better spot, but there is a conflict with pre-existing HVAC equipment. There were also issues with the controller or inverter of the turbine – I did not see it spin for the first few months I kept an eye on the museum. Apparently Cascade Engineering made some unspecified repair to the controller on January 15th. Since then the turbine has been running, and it has produced 15 kWh per month.
The array of AeroVironment machines (5 units, aggregate 13 sq. meters, 5 kW) has also had some issues – to quote the MOS presentation: “The 5 AeroVironments present issues not yet understood: Inverter occasionally shuts down with errors; Turbines stop producing power for a day or 2, often after a period of high winds, then start again. One tail shroud is broken off. Aerovironment has scheduled a repair for mid-May. MOS power curve is significantly lower than published data.” The five unit array has produced 18 kWh per month. MOS reports that Aerovironment has stopped selling these units, though they still service them.
The Proven turbine (23.6 sq. meters; 6kW nameplate) initially had grounding problems with its anemometer install which resulted in spurious edges, leading to a poor power curve – though the raw power output for comparison with other turbines would not have been affected. The anemometer was fixed December 4 of 2009. Even after this repair, the MOS reports the output of this machine falls below the published curve by 35%. The turbine produced 238 kWh per month.
I noted once again that all of the anemometers were located below hub height, which will artificially inflate the power curves of the turbines; in answer to a question the MOS team reported that they basically ran out of money and couldn’t design and erect tall anemometer masts. I talked to an engineer who was attending as a representative from the National Wind Technology Center in Colorado; he said to really trust the power curves in such a complex site, the anemometers would have to be located at hub height and closer to the turbines, and that direction sensors would have to be used for each turbine to set up directional exclusion zones to throw out data when the anemometer was in the wind shadow of the turbine. Still, the overall sense that I got was that the MOS had done a good project and made a real contribution.
Towards the end the MOS team presented on the financial aspects of the project, and they did not shy away from the conclusion – as implemented none of the turbines could produce energy cost-effectively. Installation of the five turbine systems cost a total of about $360,000. The largest single component was design/fabrication/installation of custom structural steel to interface the turbine towers with the structure of the museum building. In giving his overall assessment, Rabkin said (not an exact quote, as my stenography skills suck): “Most of the cost is in building integration – more experience and standardization will reduce costs… Small turbines may make economic sense when ground mounted in good wind regimes.” Here is the MOS data on cost of energy, given as cost per kW of first year’s production, which gets around the open question of how long the various machines will last:
Andy Brydges of the Massachusetts Clean Energy Center gave a presentation outlining the generous incentives offered by the CEC, which amount to nearly $4 per installed watt, approximately half of which is paid out after a year based on actual production. This incentive based on actual production is a fantastic move, and Andy gave some indication that they were wavering in their commitment to the delayed incentive, and I would strongly encourage them to keep it, as the data Andy reported suggested that their policies seem to be pushing capacity factors upward towards 10% (for reference, the capacity factor of the SkyStream turbine at the MOS is approximately 9%).
He also described a software tool available on the web called CWEST that would attempt to correct standard wind map data for user-input local obstacles. This is a step in the right direction, though when I downloaded a version (for the MOS site) it did not work quite as I expected – there was no input for hub height, and it wasn’t clear whether the obstacles I was entering were referenced to ground level or to hub height (the time-honored advice of experts is that a turbine should be at least 20 feet taller than any obstacle within 400 feet, or minor variations on that concept). The effect of the spreadsheet tool was to decrease the wind map estimate for 30m reference height from 5.3m/s to 5.1m/s. Many photos of recent turbine installs on the web show hub heights well below surrounding obstacles, which are commonly 50-100′ high in the east, and generally speaking, 100′ towers are not recommended standard by these turbine manufacturers, so it would be helpful if the tool provided a bit more granularity and gave citizens considering an installation a clearer picture of the energy cost of siting a turbine below surrounding obstacles.
All in all, a great project by an energetic team, providing a wealth of information for turbine manufacturers and parties interested in installing wind turbines in a complex urban environment. The MOS reported that they will publish their data and presentations at
– at this point the link redirects to the conference announce. My own conclusions: wind turbines need a lot of real-world field testing to achieve a reasonable level of reliability, and they need substantial towers to produce a reasonable amount of energy, whether they are mounted on buildings or not.