GEK Wiki / Steady State Run With Wood Pellets
  • If you are citizen of an European Union member nation, you may not use this service unless you are at least 16 years old.

  • Stop wasting time looking for files and revisions. Connect your Gmail, DriveDropbox, and Slack accounts and in less than 2 minutes, Dokkio will automatically organize all your file attachments. Learn more and claim your free account.

View
 

Steady State Run With Wood Pellets

Page history last edited by bk 11 years, 4 months ago

Forum Discussion

 

Purpose

With datalogging capabilities available via the GCU, we are conducting runs to understand the dynamics of reactor behavior and develop recommendations for reactor internal dimensions for various flows and biomass types.

 

Abstract

Wood pellets were gasified in the GEK v3.0. Ejector input pressures were initialized at flow rates of 6, 4, 3, 2, and 0.5 m^3/hr to observe the temperature profile of the reactor and the changes in pressure over time using the data logging capabilities of the gasifier control unit (GCU). The dynamics of the reactor behavior will be characterized to offer insight to the relation between the feedstock, internal dimensions of the gasifier, flow rate, rate of reaction, and gas quality.

 

Feedstock

Dragon Mountain wood pellets

 

 

 

Packing characteristics; angle of repose: <45 degrees

Inorganic ash content: 1% [reported]

Particle size: diameter: 1.5 inches, length: .250-.285 inches

Moisture content: 5%

Bulk density: 40-46 lb/ft^3

Void space: 50% (jim did a very rough test.  should be done again)

 

 

 

 

Components     

The GEK v3.0 was set up with a 3x6x6 reduction cone, nozzle outlet diameter of 1/4in, and ejector outlet diameter of 5/64 in. Propane was hooked up through a regulator allowing for pressure compensation of rotameter (flowmeter) readings for both propane and air. The propane was set up to allow complete combustion of tarry wood gas in the initial stages of start up.

 

The GEK v3.0 was set up with the hopper/auger system without the drying heat exchanger. The GCU was set up to control the auger in an on/off mode by receiving feedback from an IR detection system. The IR emitter and receiver was placed at the most sensitive shut-off threshold port located to the nearest port to the end of the auger.

 

The filter was set up to receive the wood gas directly from the reactor. Charcoal was used as a filter medium with a particle size distribution of .25-2.5mm. 

 

Instrumentation

The Gasifier Control Unit (GCU) was used to control the auger motor and flare glow plug along with monitoring eight thermocouples, and four pressure sensors and the fuel level with IR detection.

 

The glow plug flare ignition system was run through two FETs on the GCU and connected directly to the burner.

Glow plug flare ignition system

 

The grounding of the glow plug to the GEK subjected a voltage float on the entire system. Erroneous thermocouple outputs were observed as a result of this set up, so this component was removed before continuing the experiment.

 

Thermocouples were set up to offer a temperature profile of the gasifier during operation. Pictured here are the positions of the thermocouples from left to right: bottom of reduction (bred), top of reduction (tred), combustion (comb), 1 inch, 2 inches, 4 inches, and 6 inches above the nozzles on center. The thermocouple nearest to the combustion zone is expected to read lower than true combustion because of its offset location and cooling from incoming air from the nozzles. The stir bar is pictured here, however this set up was not used for this experiment.

 

               Thermocouple locations for temperature profile

A thermocouple was set up two inches before the outlet to the flare burner to offer feedback of the combustion front location in conjunction with flame characteristics.

 

Qualitative tar sampling was set up to pull a specified quantity of wood gas from the reactor outlet port right before the entrance to the filter. 

 

Tar sampling apparatus

 

Procedure:

The reactor was filled with charcoal up to an inch above the nozzles. The ejector was set to a flow rate of 6 m^3/hr while the reactor was lit using an external glow plug. When the reaction was seemingly self sustaining, the reactor was sealed, and the auger was turned on fill up the reactor.

 

The vacuum was restored at 16 minutes and the wood gas sustained a flame without supplementing propane at 18:45 minutes.

 

The reactor was set to initial ejector inlet pressures of 6, 4, 3, 2, 1, and 0.5 m^3/hr in this order. After 15 minutes of independent operation, the ash grate was then shaken 20 times manually and the reactor was then allowed to operate after an additional 10 minutes. At the end of each 25 minute block, the ash grate was again shaken manually and the hopper was opened to refill with wood pellets. The ejector inlet pressure was then initialized for the next lower flow rate in sequence.

 

Results:

Note that increasing pyrolysis temperatures may have been caused by creeping pyrolysis moving up the thermocouple support. It is unlikely that >600°C temperatures would be seen 6 inches above combustion. Pyrolysis should be in the 300°C range.

 

The spikes in the flare temperature indicate when combustion entered the 1.5" tangential pipe. This shows potential as a means to control air premix to maintain combustion inside the pipe, which should lead to cleaner combustion (reduced CO emissions according to Crispin [cite first ejector run page]), barring use of a lambda sensor.

 

Temperature and Flow Plot

Images of the tar samples plotted in the top plot in relation to sample time. Samples were drawn from the 1/2" tap between the reactor and cyclone. One sample description was skipped, so the latter points are offset by one sample (good data collection protocol still being established). Samples were drawn with a 60 mL syringe through 1" insulating ceramic felt. Spot diameter: X.

T_bred: bottom of reduction

T_tred: top of reduction

T_comb: "combustion", set away from central combustion to avoid burnout, so not wholly representitive

T_[x]in: thermocouples at x inches above air nozzles

T_flare: thermocouple in the 1.5" pipe entrance to swirl burner

P_reactor: pressure taken from the manometer tap at the bottom of the reactor

Q_air: flow of air through the v3 air inlet (measured with an orifice flowmeter)

 

Plot was generated in R (r-project.org), look for RunPlot.r in the data archive for the graphing code.

Another way to plot temperatures in the reactor. Would be more informative with a better color mapping. See the HeatMap() function

There is a very clear linear relationship between the pressures at the reactor and filter (as one might expect). The slope of this curve should change as the filter media becomes clogged. This could be a valuable metric to determine required maintance. The data in this plot comes from the end of the run.

Plot of the pull on the reactor vs. the incoming air flow. Colored clusters are points during a specific steady state (grey 6, pink 4, blue, 2, etc), black points were not during a steady state, when the hopper was opened along with other ports, leading to very different scattered relationships. During the steady states flow decreases as vacuum increases due to packing of the reduction cone.

 

Note that diagonal movement in each cluster is likely from the change in flow resistance as the reduction bell clogs. Resistance will depend on char properties. Resistance should be additive and should be the sum of reduction + nozzle + preheat tube resistances. Values would shift if air leaks are present.

 

Further Improvements:

  • Record pressure differential across gas cyclone in relation to gas flow. Alternately record gas flow with an orifice flowmeter (however orifice flowmeters are dependent on gas density, and cyclone pressure drop would also likely have a density dependency). How big an error this could produce should be calculated. The advantage of an air input measurement is that air density should be relatively constant, and can be better determined with measurements of temperature and humidity if necessary. High precision may not be necessary for eventual control, but is important for determination of gasifier efficiency, gas energy content, etc. 
  • Put combustion TC at hot spot.  maybe 1" from front of nozzle
  • Make a better performing pyrolysis TC array.  heights of 1, 2, 3, 4, 6, and 8 above nozzles.  likely a profile assembly of separate TCs that scews into a 1/2" bung on lid.
  • Record air input temperature.
  • Record gas exit temperature at top of gas cowling
  • Try for gas flow rate measurement at output
  • Improve tar sampling device and methodology. Motorized moving strip would be ideal.
  • Consider pressure drop across cyclone for flow rate measurement
  • Move to control to allow better achievement of steady state flow.
  • Minimize air leaks in the system (primary concern is reactor lid and viewport).  Use high temp grease on grate rotation shaft
  • Develop gas quality testing equipment (continuous flow calorimeter, gas sensing, engine output [risks gumming it up with intentional operation in sub-optimal regimes). 

 

Data Archive (including R code)

RunArchive-20090610.zip

Comments (5)

seachanged said

at 9:48 am on Jun 12, 2009

What is the unit for the Y axis on the upper chart?

bk said

at 6:52 pm on Jun 12, 2009

The upper graph is in m3/hr. Expect better graphs next week (working on it). Tufte is shaking his head.

scott Christopher Romack said

at 8:00 pm on Jun 12, 2009

Hey, you said Tufte!
Nice reference. I love Edward because I'm a minimalist! www.romack.net

seachanged said

at 10:55 pm on Jun 12, 2009

Mmmmmm. Nice crunchy numbers.
Those two clogs in the airflow graphs are fascinating.
The first one looks like it was pre-ignition, and was physically cleared.
The second one looks like it burned clear.
Are those accurate presumptions?

bk said

at 1:39 pm on Jun 15, 2009

seachanged: The new graph should help explain things. Air flow through the air inlet will drop when the viewport is opened, which is how we lit the reactor. The increased flows seen toward the middle of each "steady-state" run was due to shaking the grate. We're obviously a little ways away from recording truly steady-state behavior.

Scott: indeed. Hopefully the new graph is a bit better. The R code used to generate it is available in the archive, hopefully once we iterate a bit, we'll have something that'll provide an informative visualization of a run, and can be simply used by others.

You don't have permission to comment on this page.