The use of optical flame scanners for combustion process analysis in OP-650 power boiler

Transcription

1 Archivum Combustionis Vol. 35 (2015) No 2 The use of optical flame scanners for combustion process analysis in OP-650 power boiler Nabagło D. a*, Podgórski P. b*, Janda T. a, Sokołowski P. b a EDF Polska S.A., Dział Badań i Rozwoju, ul. Ciepłownicza 1, Cracow, Poland b IP&S Sp. z o.o. Lubiczów, ul. Warszawska 77C Stare Babice, Poland Abstract Optical flame scanners are commonly used in flame supervision systems (FSS) for the purpose of protection against of flame loss inside the combustion chamber. The presence or absence of flame is indicated by the scanner according to analysis of electromagnetic radiation changes in the wavelength range nm emitted by supervised flame. The use of digital signal processing technology in IR spectrum allows to recognize the supervised flame with very high ability and allows to use the signals for combustion process analysis. Electromagnetic field disturbances in the infrared due to changes in the flame, are determined by the combustion process properties in burner zone i.e. the burner heat load, excess air ratio, temperature of PF mixture, the amount and swirl of primary and secondary air. The ability of use the information provided by the scanners - in this unique parameters supervised flame i.e. flame temperature, power spectral density and the radiation amplitude changes - became the basis for the initiation of studies regarding the optical flame scanners use for combustion process analysis. The paper presents the first results of work on the use of optical flame scanners in the analysis of combustion process in 650 t/h live steam power boiler in EDF Power Plant in Rybnik. The boiler is supplied by 5 coal mill units, each of coal mills supplies 4 pulverized coal burners (PF burners). The boiler has a start-up installation consisting of 12 heavy oil burners placed in PF burners and individually controlled by 12 flame scanners. Based on the analysis of the measuring data obtained from scanners an assessment of the quality of the combustion process for 2 coal mill units (8 PF burners) have been done. Keywords: combustion process analysis, flame scanner, pulverized coal burner, power boiler, optical research of flame 1. Introduction Optimization of the combustion process is an important subject of scientific research due to the changing legal and market conditions forcing highly flexible operation of the power production units based on hard coal combustion. Coal-fired plants in Polish Energy System currently are facing a considerable challenge in adapting its assets to work in peak demand regime which means more frequent outages, start-up and dynamic load changes dictated by momentary power demand in the system. Such difficult working conditions for power units bring operational problems causing lifespan and efficiency reduction. In order to counteract a number of research and development activities were launched, which are focused on combustion process optimization in the context of these changes. The control of coal combustion process in power * Corresponding authors: Daniel Nabagło. tel.: ; address: daniel.nabaglo@edf.pl Przemysław Podgórski. tel.: ; address: p.podgorski@ipands.pl

2 The use of optical flame scanners for combustion process analysis in OP-650 power boiler boiler is characterized by large inertia which results in difficulties within the optimum adjustment of the quantity of air and fuel fed into the boiler. Taking into account the need to ensure rapid load changes, it appears that the control systems based on traditional PID controllers combined with the slow measurement systems in control loop pose a serious limitation of power boilers exploitation in such regime. Science and industry are looking for solutions to solve this problems - on the one hand amongst faster and more precise control algorithms (eg. control systems based on artificial neural networks algorithms or immunological), on the other hand amongst the modern measurement technology [1]. The best example of such solutions are flame scanners commonly used in pulverized fuel (PF) boiler for monitoring heavy oil burners [2]. Flame scanners based on modern sophisticated opto-electronic systems are able to monitor in real-time the parameters flame [3,4,5,6]. It means that the flame scanner is the first and fastest measuring system, which is able to identify parameters of the flame in burner level zone inside combustion chamber [7]. So far, the effects of mills load changing or issues with incorrect combustion process in the combustion chamber were observed with a significant delay by O 2 and CO concentration measurement downstream of the boiler or by the live/reheat steam temperature measurements [2]. Currently usage of flame scanner give the possibility in real time diagnosis of problems with combustion process for each individual burner equipped in own scanner. 2. Flame scanners characteristic The flame scanners are widely used in PF boilers start-up systems for monitoring the flame, in so-called flame supervision systems (FSS), whose task is to confirm the presence or absence of flame in the combustion chamber via a relay to cut off the startup fuel supply in a time not greater than that specified in the technical norms. The functionality of flame scanners is based on the detection of electromagnetic radiation of the flame. Depending on the type of burned fuel the flame emit electromagnetic radiation in a typical range of the whole spectrum. Fig. 1 shows a comparison of the electromagnetic spectrum from flames of the burning gas, oil and coal. Fig. 1. Common fuels radiation intensity related to wavelength [2] In order to obtain explicit information about supervised flame, the accurate discrimination from other flames is crucial. For this purpose, the learning process of scanners is performed, where the 100

3 Nabagło D., Podgórski P., Janda T., Sokołowski P. unique characteristics of the flame is recorded in device memory. Cutting-edge flame scanners based on microprocessor technology use two separate optical sensors dedicated for measurements and discrimination. An example is PARAGON 105F1-1 scanner from FIREYE Inc. used in the combustion process analysis in EDF Polska S.A. Power Plant in Rybnik (Rybnik PP). This scanner is dedicated for supervising of flames comes from liquid and solid fuel burners. It has an integrated double function sensor of electromagnetic radiation, which is used for supervision of the flame and flame temperature measurement. Operation of the scanner is based on real-time comparison of amplitude-frequency characteristics of flame blinking with characteristic stored during the learning procedure. Amplitude-frequency characteristics of the flame is obtained by means of a Fast Fourier Transform (FFT) of spectrum [7]. Momentary changes of flame characteristic attest about changes taking place in the supervised flame. On the basis the identification of the flame which characteristics differ from the other flames can be done, and thus in real time identification of the burner that burns differently in relation to the other burners can be performed. For full compatibility with Distributed Control Systems (DCS) the Paragon flame scanner software generates signals describing the amplitude - frequency characteristic: Average Amplitude (AA), Power Spectrum Density (PSD) and Flame Temperature (FT) [7]. AA is the average amplitude of the electromagnetic radiation changes interpreted as the intensity of the flame. If value is higher, the combustion process is more stable. PSD represents the intensity of flame blinking (vibrations). The signal value is correlated with geometrical position of supervised flame. If value is higher, the flame is closer to the burner outlet [7]. Due to the fact that supervised flame parameters given by the scanner depends on the conditions of combustion process, it was necessary to start dedicated research on this subject, which will confirm the correlation and dynamics of these parameters value changes according to combustion process conditions. 3. Research facility Combustion process analysis was performed for power boiler no 1 (K-1) type OP-650 (nominal capacity 650 t/h of live steam) in EDF Polska S.A. Power Plant in Rybnik (EDF Rybnik). K-1 is a front-fired pulverized fuel (PF) with 20 coal burners. According to Fig. 2 the PF burners are located on 4 levels. Additionally K-1 is equipped with 2 levels of OFA nozzles 6 on front wall and 10 on rear wall. Each of 20 PF burners has adjustable secondary air dampers which are controlled by unit operator. Start-up installation of boiler K-1 consist of 12 heavy oil burners which are marked on Fig. 2. Each oil burner has FSS which base on Paragon 105F-1 flame scanner. All scanners have the same geometrical position in relative to supervised flame. Each scanner was undergone learning procedure for distinguish the supervised flame from the background. The amplitude-frequency characteristic pattern were set-up separately for oil and coal flame for unique identification of coal burners flame with simultaneous supervision of oil burners. 101

4 The use of optical flame scanners for combustion process analysis in OP-650 power boiler Fig. 2. Burners arrangement on front wall of boiler K-1 4. Research program Combustion process analysis were made for 8 PF burners which are supplied from 2 coal mills. For tests the B1 B4 and C1 C4 burners were chosen where parameters of the mill no 2 (B) and mill no 3 (C) were controlled. Each of those PF burners has individual flame scanner Paragon 105F-1 with the same geometrical installation so the measuring conditions are comparable. The research were performed as dedicated test, where some exploitation parameters were forced: 1) Unit net power P net MW e and 200 MW e, 2) Mill B and C capacity W - 8 t/h and 20 t/h, 3) Primary air for mills F - 35 knm 3 /h and 50 knm 3 /h, 4) Coal-air mixture temperature T C and 115 C, 5) Secondary air dampers V - 0% and 100%. The conditions mentioned above combined into complex research plan, where for each unit net power and mills capacity were carried out 8 individual tests. In total the research plan include a 64 tests, which were focus on examination of combustion process condition changes impact on parameters measured by flame scanners. Table no 1 shows the tests configurations for each net power and mill capacity. Test no 1 Test no 2 Test no 3 Test no 4 Test no 5 Table 1. Parameters forced during tests Tests parameters T F V Test no Test no 7 35 Test no

5 Nabagło D., Podgórski P., Janda T., Sokołowski P. Each test lasted about 45 minutes and test were carried in time period one after the another. At the beginning of test no 1 and at the end of test no 8 the fuel sample analysis were performed to eliminate the possible impact caused by physical-chemical parameters deviation of fuel. 5. Results The analysis was divided into two parts. First one was focused on time domain analysis of measured parameters by scanner and second shows the results of statistical analysis. The measuring data form scanners were analyzed as momentary data with 5-second interval. The power unit and mills exploitation data were collected in DCS but parameters measured by scanners were collected in dedicated software where analysis of supervised flame can be performed. Taking into account the breadth of the research results in the paper only the chosen effects and summary for unit power P net =200 MW e are presented Secondary air impact Mill B (W=20t/h, F=50kNm 3 /h, T=115 C) Fig. 3 and Fig. 4 present the results of PSD and AA measurements at mill B during test no 5 and test no 6 where the secondary air dumpers position has been changed from 100% to 0%. Fig. 3. PSD value changes during Test no 6 and Test no 5 where V was changed from 100% to 0% on mill B For secondary air dampers open on 100% the PSD signal for burners B1, B2 and B3 was stable in the range (Fig. 3). Similarly, the AA signal was stable and the fluctuation was within the range of (Fig.4). There is a clear difference in the PSD and AA signal on burner B4 relative to the other burners. The lower value of the PSD leads to the conclusion that the flame is shifted from the burner nozzle, and high volatility suggest a worse stability of the flame. After the complete closure of the secondary air dampers on burners B2, B3 and B4 the PSD and AA signal value increase, which indicate that flame has been shifted closer to the burner outlet. In case of B4 the closure of secondary air damper cause the clear stabilization of PSD and AA signals which mean that combustion process is stabilized. For a 100% open 103

6 The use of optical flame scanners for combustion process analysis in OP-650 power boiler of secondary air dampers too large amount of air was fed in relation to amount of coal causing that the combustion process was unstable. Fig. 4. AA value fluctuation during Test no 5 and Test no 6 where V was changed from 100% to 0% on mill B Fig. 5 and Table 2 present the PSD and AA statistical data during tests no 5 and no 6 for mill B. Mean Standard Deviation Variance Minimum Median Maximum Table 2. Statistical data Parameter AA PSD Burner B1 B2 B3 B4 B1 B2 B3 B4 Test no Test no Δ,% Test no Test no Δ,% Test no E+6 6.6E+6 6.2E+6 1.1E+7 Test no E+6 9.4E+6 5.1E+6 7.4E+6 Δ,% Test no Test no Δ,% Test no Test no Δ,% Test no Test no Δ,%

7 Nabagło D., Podgórski P., Janda T., Sokołowski P. Fig. 5. PSD and AA statistical data as box chart Mill C (W=20t/h, F=50kNm 3 /h, T=115 C) Fig. 6 and Fig. 7 present the results of PSD and AA measurements at mill C during test no 5 and test no 6 where the secondary air dumpers position has been changed from 100% to 0%. Fig. 6. PSD value changes during Test no 5 and Test no 6 where V was changed from 0% to100% on mill C Completely closed secondary air dampers causes retraction of the flame from the nozzle in all four (C1, C2, C3, C4) burners. The PSD signal is relatively low <5000 (Fig. 6). For burner C1 the flame is out of the supervised range. Complete opening of the secondary air dampers improves the combustion dynamic - PSD signal value increases, but this is not an optimal condition, because the PSD and the AA are low in relation to the burners fed from mill B (Fig. 6 and Fig. 7). The flame of the burner C1 shows no clear change, which confirms that for the entire mill C conditions are not optimal and the ratio of the amount of air to fuel should be corrected. 105

8 The use of optical flame scanners for combustion process analysis in OP-650 power boiler Fig. 7. AA value changes during Test no 5 and Test no 6 where V was changed from 0% to100% on mill C Fig. 8 and Table 3 present the PSD and AA statistical data during tests no 5 and no 6 for mill C. Mean Standard Deviation Variance Minimum Median Maximum Table 3. Statistical data Parameter AA PSD Burner B1 B2 B3 B4 B1 B2 B3 B4 Test no Test no Δ, % Test no Test no Δ, % Test no E+5 5.7E+6 8.0E+6 7.1E+6 Test no E+5 8.6E+5 2.1E+6 1.0E+7 Δ, % Test no Test no Δ, % Test no Test no Δ, % Test no Test no Δ, %

9 Nabagło D., Podgórski P., Janda T., Sokołowski P. Fig. 8. PSD and AA statistical data as box chart In comparison with results at mill B the dynamic of PSD and AA changes for mill C are much higher than for mill B. Simultaneously the mean an median values for mill C are lower than for mill B. This can be concluded that the flame is too far from burners C1 C4 or the flame scanners are in incorrect position Mill capacity impact Mill B (V=100%, F=50kNm 3 /h, T=115 C) Fig. 9 and Fig. 10 present the results of PSD and AA measurements at mill B for 20 t/h and for 8 t/h capacity. Fig. 9. PSD value changes during Tests no 5 for two different mill capacity W=20 t/h and 8 t/h on mill B 107

10 The use of optical flame scanners for combustion process analysis in OP-650 power boiler Fig. 10. AA value changes during Tests no 5 for two different mill capacity W=20 t/h and 8 t/h on mill B Changing the feeder load on mill B with 20 t/h to 8 t/h resulted in a slight decrease in the signal PSD for B1 and B2 burners (Fig. 9) but still the value was >10,000, and the combustion process was stable. The PSD and AA have been changed for B3 and B4, where for B3 the PSD decrease - change the amount of fuel deteriorated the combustion process, whereas for B4 the load change has stabilized combustion process - the flame was moved closer to the burner and does not show significant fluctuations as in the case of mill load 20 t/h (Fig. 9 and Fig. 10). Fig. 11 and Table 4 present the PSD and AA statistical data during tests with 20 t/h and 8 t/h of mill B feeder load. Fig. 11. PSD and AA statistical data as box chart Statistical data confirms that the strongest reaction on feeder load changes were observed for burner B1 and B4 but the direction of PSD and changes is different for burner B1 the PSD and AA is proportional with feeder load and for B4 the PSD and AA is inversely proportional. This different reaction can be connected with PF mass flow distribution which depend on mill load, especially if the simple PF splitters are used. 108

11 Mean Standard Deviation Variance Minimum Median Maximum Nabagło D., Podgórski P., Janda T., Sokołowski P. Table 4. Statistical data Parameter AA PSD Burner B1 B2 B3 B4 B1 B2 B3 B4 20 t/h t/h Δ, % t/h t/h Δ, % t/h E+6 6.6E+6 6.2E+6 1.1E+7 8 t/h E+6 5.9E+6 5.6E+6 6.3E+6 Δ, % t/h t/h Δ, % t/h t/h Δ, % t/h t/h Δ, % Mill C (V=100%, F=50kNm 3 /h, T=115 C) Fig. 12 and Fig. 13 present the results of PSD and AA measurements at mill C for 20 t/h and for 8 t/h capacity. Fig. 12. PSD value changes during Tests no 5 for two different mill capacity W=20 t/h and 8 t/h on mill C 109

12 The use of optical flame scanners for combustion process analysis in OP-650 power boiler Fig. 13. AA value changes during Tests no 5 for two different mill capacity W=20 t/h and 8 t/h on mill C For mill C the load change from 20 t/h to 8 t/h resulted in a significant change of PSD and AA parameters on all burners (Fig. 12). For 20 t/h mill load the flames were significantly moved away from burners nozzles - PSD signal value was <2000 while no variation in signal AA confirms the very low intensity at the measurement range (Fig. 13). A low value and low dynamic of PDS and AA parameter changes mean that flame was moved beyond the optimum combustion zone. The reduction of fuel rate with the same primary air flow caused the flames shift near to the burner outlet. Changing the air/fuel ratio improved the combustion process in burners zone which was observed by increasing the PSD and AA signal. Fig. 14 and Table 5 present the PSD and AA statistical data during tests with 20 t/h and 8 t/h of mill C feeder load. Fig. 14. PSD and AA statistical data as box chart 110

13 Nabagło D., Podgórski P., Janda T., Sokołowski P. Table 5. Statistical data Parameter AA PSD Burner B1 B2 B3 B4 B1 B2 B3 B4 8 t/h Mean 20 t/h Δ, % Standard Deviation 8 t/h t/h Δ, % t/h E+4 8.7E+4 1.1E+5 1.1E+6 Variance 20 t/h E+5 8.6E+5 2.1E+6 1.0E+7 Δ, % t/h Minimum 20 t/h Δ, % t/h Median 20 t/h Δ, % t/h Maximum 20 t/h Δ, % The direction of PSD and AA value changes is the same for all PF burners supplied from mill C. The highest changes were observed for burner C1 and C4. Those differences between burners indicates the non-balanced PF flow and air distribution Primary air impact Mill B (W=20 t/h, T=100 C, V=0%) Fig. 15 and Fig. 16 present the results of PSD and AA measurements at mill B during primary air changes. Tests with a primary air flow change from 35 knm 3 /h to 50 knm 3 /h on mill B showed little effect on the values of the PSD and AA parameters. The combustion process was stable on burners B1, B2 and B3 both at 35 knm 3 /h and at 50 knm 3 /h. The burner B4 had the same problems with flame stability which has been shown during tests with mill load and secondary air changes. 111

14 The use of optical flame scanners for combustion process analysis in OP-650 power boiler Fig. 15. AA value changes during test no 1 and no 4 where F was changed from 35 knm 3 /h to 50 knm 3 /h on mill B Fig. 17 and Table 6 present the PSD and AA statistical data during tests primary air changes at mill B. Fig. 16. PSD and AA statistical data as box chart The primary air changes caused the non-uniform response on flame scanners. For B2 and B4 the PSD and AA changes direction were different than for B1 and B3. It was also caused by nonproportional changes of fuel/air ratio for each burner. Mean Table 6. Statistical data Parameter AA PSD Burner B1 B2 B3 B4 B1 B2 B3 B4 Test no Test no Δ,% Standard Test no

15 Nabagło D., Podgórski P., Janda T., Sokołowski P. Parameter AA PSD Burner B1 B2 B3 B4 B1 B2 B3 B4 Deviation Test no Variance Minimum Median Maximum Δ,% Test no E+6 8.0E+6 6.3E+6 1.3E+7 Test no E+6 5.5E+6 4.5E+6 1.4E+7 Δ,% Test no Test no Δ,% Test no Test no Δ,% Test no Test no Δ,% Mill C (W=20 t/h, T=100 C, V=0%) Fig. 18 and Fig. 19 present the results of PSD and AA measurements at mill C during primary air changes. Fig. 17. PSD value changes during test no 1 and no 4 where F was changed from 35 knm 3 /h to 50 knm 3 /h on mill C 113

16 The use of optical flame scanners for combustion process analysis in OP-650 power boiler Fig. 18. AA value changes during test no 1 and no 4 where F was changed from 35 knm 3 /h to 50 knm 3 /h on mill C For mill C the change of primary air amount from of 35 knm 3 /h to 50kNm 3 /h caused a significant decrease in the value of the PSD signals and AA (Fig. 18 and Fig. 19). This means that the flame has moved away from the burner outlet can thus be concluded that the air/fuel had been negatively changed. Reaction for air change was the same for all burners of mill C. Fig. 20 and Table 7 presents the PSD and AA statistical data during tests primary air changes at mill C. Fig. 19. PSD and AA statistical data as box chart Table 7. Statistical data Parameter AA PSD Burner B1 B2 B3 B4 B1 B2 B3 B4 Test no Mean Test no Δ,% Standard Test no

17 Nabagło D., Podgórski P., Janda T., Sokołowski P. Parameter AA PSD Burner B1 B2 B3 B4 B1 B2 B3 B4 Deviation Test no Variance Minimum Median Maximum Δ,% Test no E+6 7.3E+6 1.4E+6 1.0E+7 Test no E+4 8.7E+4 1.1E+5 1.1E+6 Δ,% Test no Test no Δ,% Test no Test no Δ,% Test no Test no Δ,% For mill C the PSD and AA value changes were very strong during test with primary air changes. For each PF burner the response had the same direction the PSD and AA value decrease with primary air decrease. In comparison with mill B the reaction on primary air change is stronger for mill C than for mill B. This indicate that for mill C the fuel/air ratio was not optimal and should be corrected for combustion process optimization. 6. Summary The combustion process analysis showed a clear reaction on the flame scanners parameters PSD and AA by changing combustion conditions in the area of burners. This reaction is not unambiguous for all burners, which can be used to diagnose problems with a single burner. For valid measurement results the flame scanners have to be directed into burners zone with high accuracy and with comparable angle. Differences in the inclination angle of scanner viewfinder tube can also cause results ambiguity that can be misinterpreted. The performed tests also showed different responses PSD and AA signals for the burners on different levels of the windbox. The possible reason could be the pressure conditions in the combustion chamber and in the windbox, where for K-1 boiler depends on several factors (Over-Fire Air flow, flue gas fans load) and are difficult to control. Combusted fuel during the test did not differ significantly from standard one what suggests that the tests carried out were not affected by fluctuations in physical-chemical parameters of fuel (Fig. 21). In the fuel supply system of boiler the flame scanner is the first and fastest-reacting system for measurement and diagnostic of combustion process, which is much faster than "classic" measurements to provide information about the combustion process. This enables to use flame scanners as basic measuring system for the co-operation with automatic control systems for air and fuel flow optimization. As research has shown it is possible to optimize the coal mill units operation based on the measured parameters by flame scanner, where the flame scanner responds on: mill load changes, primary and secondary air flow changes. 115

18 The use of optical flame scanners for combustion process analysis in OP-650 power boiler Fig. 20. Fuel characteristic combusted during tests on mill B and C in comparison with standard fuel for boiler K-1 The presented results concern the power unit operation with a maximum or near maximum load and the impact of different mill units operation configurations was eliminated (all 5 mills were in operation). High differences in the PSD and AA parameters during tests on mill no. C confirmed the strong impact of changes in the quantity of primary air and mill load. Air/fuel ratio changes in burners zone are manifested by an increase or decrease in the PSD and AA value. For mill no. B there was a clear difference in the work of the burners, where B4 had a problem with the stability of combustion. Fluctuations of PSD and AA value were significant when the amount of fuel have changed. After mill load changes the stabilization of the flame were also observed. Measurement capabilities of modern flame scanners provide unique information about the combustion process within the supervised flame. To use scanners in combustion control systems require to conduct further studies to build a dependence between the PSD and the AA and the conditions prevailing in the combustion the chamber. In addition the tests carried out showed that changing of fuel and air flow cause the disproportions between the signal values of the PSD and AA. On this basis we can conclude that coal-air mixture distribution optimization can be done by use of flame scanners, but the long-term tests are needed. References [1] Wojdan K, Świrski K, Warchoł M, Milewski J, Miller A: A Practical Approach to Combustion Process Optimization Using an Improved Immune Optimizer. Sustainable Research and Innovation Proceedings, Vol 3; [2] Adynowski J, Sokołowski P, Podgórski P, Kalinowski K. Aid to the optimization of the combustion process in pulverized coal-fired burners by use of flame safeguard systems. Conference Proceedings, 12 th International Conference on Boiler Technology; [3] Ballester J,Garcı a-armingol T. Diagnostic techniques for the monitoring and control of practical flames. Progress in Energy and Combustion Science 36 (2010) [4] Chi T, Zhang H,Yan Y, Zhou H, Zheng H. Investigations into the ignition behaviors of pulverized coals and coal blends in a drop tube furnace using flame monitoring techniques. Fuel 89 (2010) [5] González-Cencerrado A, Gil A.,Peńa B. Characterization of PF flames under different swirl conditions based on visualization systems. Fuel 113 (2013) [6] Zhou H, Tang Q, Yang L, Yan Y, Lu G, Cen K. Support vector machine based online coal identification through advanced flame monitoring. Fuel 117 (2014) [7] TYPE 105F1-1 Integrated Flame Scanner and Temperature Analyzer: User Manual, CU-108; March