TPV Tube Generators for Apartment Building and Industrial Furnace Applications Lewis M. Fraas, James E. Avery, Wilbert E. Daniels, Huang X. Huang, Enrico Malfa*, Matteo Venturino*, Giandomenico Testi*, Gianni Mascalzi *%, Joachim G. Wuenning** JX Crystals Inc., 1105 12 th Ave. NW, Suite A2, Issaquah, WA 98027, USA *ABB Service S.R.L., Via C. Arconati, 1-20135 Milano, MI, ITALY Presently with Centro Sviluppo Materiali S.p.A, Piazza Caduti 6 luglio 1944, 1 Dalmine (BG), ITALY % Presently with Environmental & Power Technologies, v.le Edison 50, Sesto S. Giovanni (MI), ITALY **WS Wärmeprozesstechnik GmbH, Dornierstrasse 14, D-71272 Renningen, GERMANY Abstract. Major changes in the regulation of electric and natural gas industries during recent years have forced energy companies to explore opportunities in small-size Combined Heat and Power systems. These differ fundamentally from the traditional model of central generation and delivery since small, modular electric generators can be located very close to end-users inside a building or a single house within an industrial area, combined with the production of heat and cold. In particular, interest is growing in the new technologies for sub- 100kWe units, including systems based on thermophotovoltaic (TPV) technology. TPV generator tubes can be inserted into hot furnaces to generate electricity and low-grade heat. In this generator tube, a water-cooled GaSb photovoltaic converter array inside the tube faces outward toward an infrared emitter liner mounted on the inside surface of the closed-end tube. Each tube can be sized to generate several kw and a given furnace can heat several tubes. We have conducted pilot experiments on key components in order to develop the concept just described. This includes a pilot scale array tested in an electrical furnace that heat a 3 diameter alumina tube with an infrared emitting liner. Also, a silicon carbide tube with a water-cooling system was tested in a ceramic fiber lined furnace equipped with a commercial 200 kw flameless regenerative burner, simulating a TPV generator tube in such a system. INTRODUCTION Large furnaces such as those used in glass or melting industrial operations or alternatively in apartment buildings already have efficient burners and large high temperature spaces. TPV generator tubes such as the one shown conceptually in figure 1 can be inserted into these hot furnaces to generate electricity and low-grade heat (as in figure 2). In this generator tube, a water-cooled photovoltaic array inside the tube faces outward toward an infrared emitter liner mounted on the inside surface of the closed-ended tube. Each tube can be sized to generate several kw and a given furnace can heat several tubes. FIGURE 1. Multi-kW TPV tube concept for industry 5 kw from a two meter tube.
Auto regenerative low Nox burner 1 4 5 2 3 1.5 m TPV Module Fi 1 = hot gas 2 = emitter 3 = GaSb cell array 4 = IR radiation 5 = cooling system 1 m FIGURE 2. ABB s concept for insertion of four tubes into a furnace with a regenerative burner. JX Crystals has a patent pending on the TPV tube generator design shown in fig. 1 and is developing these tubes. The three key TPV components used here are the GaSb infrared sensitive photovoltaic cell, antireflection-coated refractory metal emitters, and simple dielectric filters. These three key components and how they work together are described in more detail in a companion paper presented at this conference. These TPV generator tubes can potentially be incorporated into furnaces in apartment buildings as is shown in the figure 2 ABB furnace design. These concepts originated approximately three years ago. However, at that time, no work had been done to demonstrate this concept. ABB then funded JX Crystals to conduct some pilot demonstration experiments on the TPV converter tubes intended to lead to the apartment building combined heat and power application. JX Crystals also obtained funding for pilot experiments from the US DOE directed toward industrial applications of these concepts. Tests have been carried out by ABB to investigate the possibility to design a mini co-generator around a flameless regenerative burner. Both sets of experimental findings are presented in the following sections. PILOT EXPERIMENTS ON TPV CONVERTER TUBES In our pilot experiments, an electrical furnace is used to heat a 7.7 cm diameter alumina tube with an infrared liner. PV circuits are mounted on a water-cooled paddle and the paddle is inserted into the heated tube. Each circuit contains an array of GaSb cells. Antireflection-coated tungsten foil is used as the emitter liner. Dielectric filters are located between the cells and the emitter. Figure 3 shows a 3D drawing of our pilot experimental configuration. Over the course of a year, we fabricated five test paddles, four of which have live circuits (pictured in fig. 4). The fifth paddle is a gold plated reference paddle used for background absorption measurements. The paddles shown in figure 4 all contain circuits with 15 series-connected GaSb cells shingle mounted in rows. Each paddle in is 18 cm long, with an active circuit length of 16.5 cm, using 1.65 cm wide cells.
existing oven SiC glowbars water-cooled holder with two circuits insulation R -type thermocouples front end seal with fittings Alumina tube back end seal with fittings AR-W foil Ta foil FIGURE 3. Tube furnace with photovoltaic paddle - oven and tube cut out to show interior features. The paddles in figures 4a and 4d have 30-cell circuits mounted on their top and bottom faces, with gold plated side walls; each circuit has two rows of 15 cells mounted in parallel (note that figure 4d has filter cover slides over the circuits). The paddle in figure 4b has 30-cell circuits mounted on all four faces. Finally, the paddle shown in figure 4c has eight 15-cell single-row circuits mounted in an octagonal array. c a b d FIGURE 4. Paddles fabricated at JX Crystals: (a) two circuits, (b) four circuits, (c) eight circuits, and (d) two circuits with filter cover slides.
30 cell circuit test fill factor = 0.705 Voc = 7.08 Isc = 6.80 Vmax = 5.47 Imax = 6.20 Pmax = 33.94 FIGURE 5. Current vs. voltage for a 30 cell circuit, such as the ones used in 4a, 4b, and 4d above. Figure 5 shows a typical flash test result for a 30-cell circuit in the last paddle fabricated. In addition to the various paddles used in our experiments over the last year, we have also experimented with various emitter liner configurations as will be described in the next section. TPV EXPERIMENTAL RESULTS Over the course of a year, we conducted 61 high temperature furnace runs. Each run takes a full day with a slow furnace ramp up followed by a dwell for several hours at temperature and then a slow ramp down. In all cases, the tube was evacuated, the paddle was water-cooled with thermister probes located at the paddle inlet and outlet, the water flow was monitored including provision for back up water supply if required, and I vs. V curves were recorded rapidly and periodically at temperature. In the following, we first describe the results qualitatively. Then, toward the end of this section, we present quantitative results. Our work over the last year can best be described in terms of the identification and resolution of three different problems that were initially intertwined: 1) Slow contaminant deposit buildup on the cold paddle emanating from the hot side elements. 2) Cracks and steps between the cells and circuits producing additional absorption losses degrading efficiency. 3) Poor heat transfer to the emitter liner resulting in poor emitter temperature uniformity and low emitter temperature. Both poor uniformity and low temperature degrade system performance. In the following, we describe the origin of each of these problems, its solution, and the effect that it had on our experiments. As these problems are intertwined, we highlight our key experiments in historical order. In our first two-circuit paddle experiment (run 9), we measured an efficiency of 11.2%. This result repeated in runs 11 and 12 as witnessed by engineers from ABB. This result was obtained at an emitter temperature of 1150 C. A subsequent recalibration of our thermisters reduced the
efficiency number to 10.9%. This emitter temperature was derived from modeling based on the measured cell current since the thermocouple reading of the emitter temperature proved unreliable. We wanted a higher emitter temperature. Also, at this point, the four-circuit paddle had been built and was ready for test. So prior to run 16, we removed the AR/W liner. We planned to coat the backside of this foil with WC in an attempt to blacken it and improve heat transfer to it from the furnace heater elements. However, this first liner was tightly fitting in the tube and had become brittle through high temperature operation. So, it broke and we fabricated a new liner. The results for run 16 were disappointing and there was a pronounced contamination coating on the paddle. It turned out that unknown to us at the time, the WC source material we used was bound together using cobalt and this produced the contaminant. Unfortunately, both the 4- circuit paddle and the initial 2-circuit paddle were irreversibly degraded in runs 16 and 18 by cobalt film deposition. Run 19 started us on a search for the origins of contaminants. Fortunately, we found that we could monitor contaminant buildup by using the gold reference paddle and watching the rate at which heat input into this paddle increased with time at temperature. Also, fortunately after each run, the contaminant layer could be wiped off the gold paddle to return it to its original condition. The contaminant could also be analyzed chemically. In runs 20 through 48, we found and eliminated the contaminant sources. There were two. The first and more extreme came from the Cobalt and could be eliminated by eliminating the use of Cobalt bonded WC. The second and more minor contaminant source came from a reaction between the silica binder in the alumina tube and the backside of the W foil. The silica reacted with W to produce tungsten oxide and silicon. We eliminated this problem by welding the tungsten seam with a proprietary welding method. This produced a tube where nothing from the backside could find its way to the inside paddle. By run 48, the first problem, the contaminant problem, had been resolved. On runs 49 through 52, we collected data on the 8-circuit paddle. The measured conversion efficiency for this paddle of 5.6% was very disappointing. We now believe that this poor efficiency was a result of problems 2 and 3. We decided that we could attack problem number 2, i.e. absorption at cracks and steps, by building a 2-circuit paddle using planar filter-coated glass slides adhesive bonded to the circuits thereby covering the cracks and steps. A photograph of this last 2-circuit paddle is shown in figure 4d. A visual inspection of this photograph should convince the reader that problem 2 is now resolved. We measured the performance of our last 2-circuit paddle in run 61. Our results for runs 9 and 61 are summarized in table 1. The measured efficiency for run 61 was actually somewhat lower at 9.5% than we had measured in run 9. This was at first surprising since we believe that the elimination of cracks and steps should have produced a higher efficiency paddle. We believe that problem 3, i.e. heat transfer to the emitter foil, is now the remaining problem. As we noted earlier, the foil in run 9 was in good mechanical contact with the alumina tube whereas in run 61, the foil was a welded tube that could be slid in and out of the alumina tube. Hence in run 61, problem 3 was more pronounced. This issue is discussed in more detail in the next section.
TABLE 1. Paddle Runs 9 and 61 First paddle (Run 9) Second paddle (Run 61) Circuit Data Isc 7.8 A 7.5 A Voc 6.5 V 6.5 V FF 0.66 0.66 Pmax 33.3 W 32.3 W Paddle Data Pelectric 66.6 W 64.6 W Pwater 840 W 940 W Pgold 230 W 260 W Derived Temitter 1150 C 1150 C Tcell 50 C 50 C Pwater-Pgold 610 W 680 W Efficiency & Pelec./(Pwater-Pgold) 10.9% 9.5% Power Density Power Density (elec.) 0.76 W/cm 2 0.74 W/cm 2 TPV MODELING AND INTERPRETATION JX Crystals used TracePro software to develop a model that reproduces the measured results from run 61. TracePro is a 3D Monte Carlo Ray Trace program that can generate heat transfer predictions given geometry and surface and material properties. Figure 6 illustrates the setup procedures for TracePro model calculations. How can we explain the differences between runs 9 and 61? Our thesis is that the difference comes from enhanced non-uniformity in the emitter temperature profile. We begin with visual observation made during run 61. There is a window at the end of the alumina tube that allows us to look at the paddle and emitter during a run. In run 61, it was observed that the emitter was hotter at its ends than in the region over the circuit. Our thesis is that the emitter temperature is drawn down in its center by heat transfer to the circuit whereas at its ends where less heat is lost, it is hotter. This thesis is strongly supported by inspection of the actual furnace geometry shown in the vertical cross section drawing depicted in figure 7. Note that the 34 cm furnace hot zone is double the length of the 17 cm paddle. Note also that only four of the furnace heating elements (shown as circles) directly heat the emitter in the paddle region, while the other four heater elements heat the end zones. FIGURE 6. TracePro model of paddle and tube.
FIGURE 7. Test configuration with assumed temperature profile for modeling, accounting for both poor heat distribution and poor heat coupling to tungsten emitter foil. The thesis that the emitter foil temperature can be drawn down by heat loss is further supported by an additional observation made during a secondary experiment done during run 61. When we back filled with Argon, we noted an interesting transient. The circuit current went up from 7.5 A to 8.6 A for a short time and then settled back down at a new equilibrium value of 7.8 A. We explain this as follows. In vacuum, the coupling between the alumina tube and the foil is poor. So the foil temperature is initially appreciably lower than the alumina tube temperature. When Argon fills the space between the two tubes, the heat transfer rate improves and the foil temperature rises toward the alumina temperature. Because of the stored heat in the alumina tube, it takes a short time for the alumina tube temperature to fall to a new equilibrium value. After establishing the new equilibrium in Argon, the foil temperature is higher and the alumina temperature is a little lower than in vacuum. Calculation suggests a foil temperature of 1150 C for 7.5 A and a foil temperature of 1180 C for 8.6 A or a change in temperature of 30 C. All of the above serves as justification for the temperature profile shown in fig. 7, where the emitter foil ends are assumed to be at 1250 C and the center region of the foil is assumed to be at 1150 C. This profile then produces additional power losses in the end segments of the circuit as is shown in figure 8. Watts 40 35 30 25 20 15 10 0 5 10 15 FIGURE 8. Power distribution into 14 circuit segments in Watts per segment.
Since the cells with the lowest illumination level control the current from a circuit, the excess power coupled into the circuits at each end is simply an additional loss. Our model suggests that this is an additional loss of approximately 72 Watts. Our model results reconciling runs 9 and 61 are presented in table 2. It has to be noted that our arguments here are simply plausible but certainly not rigorously proven. TABLE 2. TracePro analysis of run 61 (see table 1) Emitter Center Temperature 1150 C Given Emitter End Temperature 1250 C Cell Temperature 50 C Prediction Pwater 936 W Pgold 252 W Isc 7.5 A Pelectric 65.2 W P(excess at circuit ends) 72 W Pwater-Pgold-Pend 612 W Efficiency 10.6% The fact remains that problem 3, i.e. poor heat coupling to the emitter foil, needs to be addressed. For our pilot experiments, the problem can be resolved in three steps. First, since it is easier to radiation-couple to a black tube than to a white tube, we propose to replace the alumina white tube with a SiC black tube. This will have additional advantages in that SiC will be the material of choice in real systems and SiC can be more rapidly cycled. Second, we need to bond the emitter foil to the SiC. This is where some research will be required. Perhaps silicon powder can be used as a bonding agent. In any case, we note that the thermal expansion coefficient for SiC is less than that of W or Mo. This means that the emitter foil will expand and lock into the SiC. This is not the case for the alumina tube where the thermal expansion coefficient for alumina is larger and the gap widens between the two tubes with an increase in temperature. Finally, third, we propose to build a longer paddle that is more nearly the same length as the furnace hot zone length. This will improve uniformity and decrease the importance of end effects. Returning to our modeling results, the fact that we are able to fit our measurements with our model allows us to use our model to make projections regarding potential future performance improvements. Table 3 presents some model predictions. The first row of this table simply represents the fact that the model agrees with experiment. Subsequent rows report the predicted efficiencies when specific changes are made in the model. In the second row, we add perfect end mirrors in the model, eliminating the hot end effects, and the efficiency rises to 12.4%. In the third row, we hypothesize an improvement in cell quantum efficiency. The effect is minor. A more significant improvement to 14.2% results from an assumed reduction in circuit series resistance. Finally, rows 5 and 6 indicate that an efficiency greater than 15% can result from an emitter temperature exceeding 1200 C. Our model predicts that a major gain in both power density and TPV efficiency will come with increases in emitter temperature as can be seen in figure 9.
TABLE 3. Path to Higher Efficiencies TracePro Model Efficiency 1) Modeled with emitter at 1150 C per table 10.6% 2) Emitter at 1100 C with paddle end 12.4% 3) #2 with improved cell QE 12.8% 4) #2 with lower circuit series resistance 14.2% 5) #4 with emitter at 1200 C 16.6% 6) #4 with emitter at 1300 C 18.9% 7) Cell only with emitter at 1300 C 23.0% 2.0 20 1.8 19 Power Density (w/cm^2) 1.6 1.4 1.2 1.0 0.8 0.6 0.4 18 17 16 15 14 13 12 Efficiency 0.2 11 1050 1100 1150 1200 1250 1300 1350 Emitter Temperature (C) FIGURE 9. Efficiency and array power density vs. temperature. PILOT FURNACE EXPERIMENTS A series of tests have been carried out at WS (Wärmeprozesstechnik GmbH) ceramic fiber lined furnace (figure 10) to investigate the possibility to design a miniature co-generator around a flameless regenerative burner. In particular, the Regemat 200kW manufactured and commercialized by WS for direct heating in industrial application has been selected. It is composed of a series of nozzles, surrounding a central NG gun. Each ceramic nozzle is fed alternatively with hot combustion gases and air, acting as regenerative air preheater. In this way the combustion air can reach temperatures of about 800-1000 C, required for the behavior. A series of valves provides the switching between suction and injection for each nozzle. The burner should start up in flame mode and, when the furnace temperature reach 850 C, the burner can be switched to mode.
The furnace is 2000 mm long and has a cross section of 1200x1200 mm for the front dimensions. During mode, part of the flue gases are exhausted through the front of the furnace after their pass through the regenerative heat exchanger. In the WS test rig during furnace operation, there are four stainless steel air-cooled pipes arranged along the furnace, as depicted in figure 10, that usually provide heat extraction. In order to reproduce the operating condition for the TPV generator tube, a SiC tube has replaced one of these cooling pipes with a water-cooling probe inside it (figure 10). FIGURE 10. Furnace experimental set-up. The SiC tube selected, the commercial one used in WS radiant burner (C200/1650), has an external diameter of 195 mm (internal 185 mm) and a active length inside the furnace 1600 mm. The external surface has been equipped with thermocouples to measure the temperature distribution. Surface temperature has been recorded along the tube axis following three lines on the cylindrical pipe. The SiC tube surrounds the external pipe (φ o =141.3 mm, active length 1400 mm) of the water cooling system. FURNACE TEST RESULTS The goals of the tests were to verify the impact of operation mode on temperature level and uniformity along the SiC water cooled pipe at different furnaces operating conditions and the corresponding percentage of available heat that it is possible to transfer from the fuel to the SiC tube. During the test the burner efficiency and the pollutant emissions have also been recorded. Several tests have been performed by varying the furnace temperature starting from 850 C to 1250 C at 100% of burner power. In order to increase the furnace temperature the heat extraction has been adjusted reducing the flow rate to the aircooled pipes. The furnace operation has been tested also at partial load and measurements have been recorded for 50% burner load. Table 4 shows the furnace test results.
TABLE 4. Combustion Tests Matrix Flame 900 950 1000 1100 1200 1250 1200 Flame 1200 Test number 2 3 4 5 6 7 8 9 Nominal load % 100 100 100 100 100 100 100 100 NG Power input kw 200.2 191.2 200.2 202.2 200.2 203.2 199.2 204.2 Total Power input kw 202.2 193.2 202.2 204.3 202.1 205.2 201.3 206.4 Fuel-to-emitter η - 52% 79% 78% 75% 72% 67% 73% 41% Furnace Temperature C 906 957 1009 1109 1207 1259 1207 1162 Nox @ 3% O2 ppm 53 14 25 52 55 80 55 62 Surface temperature along the SiC tube containing the internal water pipe showed an average temperature gap with the furnace in the range of 80-110 C. No appreciable T difference was detected from flame to modes. The main reason for that is in the high mixing of flue gases even during flame mode promoted by the reduced length of the furnace and the high momentum for the flame. The dimensionless (T/T ref ) temperature profiles along the tube axial coordinate (x TC /x ref ) for both and flame mode varied from 82% to 94%. Figure 11 shows surface temperature profiles along the silicon carbide tube in mode. T / Tref [dimless] 1.10 1.05 1.00 0.95 0.90 0.85 0.80 0.75 0.70 gen. 1 gen. 2 gen. 3 Water cooling system length 0 0.2 0.4 0.6 0.8 1 X TC/X ref [dimless] FIGURE 11. Surface temp. profiles along the SiC tube in mode (furnace temp. = 1200 C). CONCLUSIONS Herein, we have described a new TPV geometry suitable for use in industrial furnaces or central heating furnaces in large buildings. In this geometry, the TPV array is located inside an emitter tube and the tube is heated from outside. As it turns out, this geometry is also very suitable for pilot experiments where the interactions between full size circuits and full size emitters can be explored. In the experiments reported here, we have measured a TPV efficiency of over 10% for the first time using complete circuits and full size emitters. We have also developed a model that quantitatively fits our measurements and allows us to predict efficiencies over 15% after additional development. The test campaign conduced at WS has confirmed that the autoregenerative burner operating in mode can provide suitable conditions in the combustion furnace for the development of a mini-chp TPV unit.