Problem Solving Tools in Waste-To-Energy Systems

Transcription

1 14th North American Waste to Energy Conference May 1-3,2006, Tampa, Florida USA NAWTEC Problem Solving Tools in Waste-To-Energy Systems Mark White, Manager of Combustion Engineering Greg Epelbaurn, Principal Boiler Engineer Covanta Energy Corp. 40 Lane Road Fairfield, NJ mwhite\g)covantaellcrgv.com (973) gepelbau(a)covantaellergv.com (973) Abstract Covanta is using a multifaceted approach to problem solving in Waste-to-Energy systems which combines several types of computer modeling with physical cold flow models, field testing, and engineering experience. This problem-solving approach is applied to boiler corrosion, gas and particulate flow patterns, reagent injection, and APC system issues. Our goals are to bring the most appropriate tools to each issue and incorporate results back into the engineering approach in order to continually improve our technical capabilities. Several types of computer modeling are used. A commercially available energy balance program is used for steam cycle evaluations and boiler energy balance and heat transfer calculations. Computational Fluid Dynamics (CFD) models are developed to investigate temperature and flow patterns where local conditions must be understood in detail. We have made extensive use of cold flow models to improve performance of APC systems, and to evaluate overfire air mixing in furnaces, and flow distribution through tube banks in boilers. Field testing is used to investigate temperature fluctuations and distributions, flow stratification, corrosion rates, and to validate modeling or analytical results. Each of these approaches has its own set of advantages, disadvantages, and limitations, and must always be combined with a healthy dose of operating and engineering experience. Analytical work is done by, or in close cooperation with, our operations and engineering staff with many years of experience operating, designing, and modifying boilers, APC systems, and related equipment. This integrated approach has yielded significant improvements in many cases and is being used in increasingly complex applications. Introduction The Waste-to-Energy (WTE) industry shares many features with utility power generation, but many issues are unique to WTE. Approaches to solving problems which have been very successful in utility or other industrial applications may not be successful, or may not even be applicable, in the WTE industry. As a result, WTE operators often find it necessary to develop solutions to problems. This paper addresses some of the approaches and tools that Covanta Energy uses to understand and solve operational problems in our 31 domestic WTE facilities. The approaches and tools discussed include: Field testing - Direct measurement of actual conditions in the facility. Process modeling - Use of computer models to calculate equipment performance, using various levels of sophistication. These models include boiler models with mass and energy balances and heat transfer rates, and. models of steam cycle components. Cold flow plastic scale models - These are physical scale models which allow visual indication and direct measurement of flow conditions. The models are relatively inexpensive, but cannot model heat transfer or combustion reactions. CFD models - These programs are able to model all relevant phenomena, including flows, heat transfer, and chemical reactions. However, they are expensive, and require significant effort to develop and understand. 13 1

2 Field Testing Field testing is probably the oldest and most common approach to problem solving. Will a new material stand up to the environment in a waste-toenergy furnace? It seems straightforward to put a test coupon in and find out. What is the temperature at the inlet to the superheater? Install a thermocouple and measure it. While the concept is simple enough, there clearly is a lot more involved. The corrosion rate of an uncooled metal sample in the furnace may be completely unrelated to the corrosion rate of the same material applied to a cooled boiler tube, while thermal radiation, slag buildup, and other effects can introduce serious discrepancies between the temperature of a thermocouple and the actual gas temperature. Covanta relies heavily on field testing for many purposes including acceptance testing, performance evaluations on major equipment such as turbines and cooling towers, SNCR system optimization, establishing baseline operating conditions for equipment modifications, and validation of modeling work. In all cases the field testing must produce meaningful and reliable results to support further work and ensure valid conclusions. The company has invested significant effort and cost in developing in-house expertise and evaluating measurement instruments and techniques to assure that results are correct. Two important field measurements that are frequently made are furnace temperatures and air flows. When it is important to have an accurate gas temperature measurement inside a furnace, Covanta uses a High Velocity Thermocouple (HVT), also known as a suction pyrometer. These devices provide an accurate measurement, but are very time- and labor-intensive. The integrity of the thermocouple and radiation shield must be verified before and after each test, and the port has to be kept clear. Typically the instrument is moved to different depths in the furnace over a period of time, and data interpretation requires that temperature differences be evaluated both in terms of probe location and varying furnace conditions. As waste-to-energy combustion is generally much more dynamic than other common combustion processes, the furnace may undergo significant variation in temperature during the test period. The dwell time at each probe position must be sufficient to allow a clear distinction between temperature differences due to probe position and temperature differences due to fluctuations in the combustion process. Because of the effort necessary to use the HVT probe, Covanta has investigated alternative gas temperature measurement instruments which allow comparable accuracy with less manual involvement. At this point, the instrument-of-choice is the IR pyrometer utilizing a wavelength at the CO2 emissivity peak between 4 and 5 microns. A significant number of tests have been performed using side-by-side comparison of these IR instruments with HVT probes which consistently show close agreement. These instruments are able to measure temperature based on radiation from the gas itself, thus providing a more accurate indication than other instruments which base the temperature on the radiation emitted by particles in the gas stream. Figure I illustrates temperature measurements by three different devices: an HVT, an IR gas pyrometer, and the permanently installed plant thermocouple. The close agreement between the IR and HVT is clear, as is the varying responsiveness to temperature fluctuations, with the IR most responsive and the plant TC least. We have had mixed results with particle-based systems. Our belief is that two issues contribute to inaccuracy in these instruments. First, particles are not always in thermal equilibrium with the gas stream, with larger particles more likely to be at different temperatures than smaller ones. Second, in narrower furnaces the particles do not completely fill the field of view, and the instrument is able to "see" the wall on the opposite side. Air flow measurements are normally made with station instrumentation, but occasionally need to be verified for various purposes. This is typically done using a pitot tube and liquid manometer, with a series of measurements made at different points in the duct. Flows can then be calculated from these measurements and the duct geometry, provided the pitot tube measurements are made accurately. As with any measurement technique, errors can creep into the procedure and invalidate the results. Sometimes the erros come from unexpected sources. For example, we are aware of one case in which a socalled "digital" manometer was substituted for the standard liquid-filled device, and very unexpected results were obtained. The device's calibration was verified and it was tested against other instruments, but appeared to be functioning correctly. Ultimately it was determined that transmissions over the plant's radio communications devices were interfering with the digital manometer, causing a distorted reading each time a nearby microphone was keyed. This experience illustrates the importance of properly selecting the instrumentation for critical measurements. 132

3 Another area of field testing involves corrosion research. Covanta is continually evaluating alternative materials in field tests throughout the company, with a dozen or more tests underway at any given time. Most of these tests involve application of the test material to boiler tubes, with success or failure determined by inspection of the materials during subsequent outages. We have also constructed a cooled corrosion test probe, which allows test coupons to be inserted or removed while the boiler is in operation. The system measures and controls the temperature of the test coupon to duplicate the temperature seen on actual boiler parts. This arrangement allows material samples to be tested quickly and inexpensively. All of the field testing techniques require accurate, calibrated equipment and knowledgeable, experienced personnel. Specialized equipment such as flow meters, gas analyzers, and data acquisition equipment may be required. Testing can be expensive, and results may only be applicable for the period in which testing was actually performed. Ultimately, however, field testing is indispensible as a tool in areas such as acceptance testing, SNCR system optimization, turbine performance and materials testing. Process Modeling Process modeling uses commercially available software packages to simulate all or part of an industrial process. Covanta uses this approach in several areas, including optimization of steamcycle operation, evaluation of boiler modifications, and assessment of boiler cleaning methods. Process modeling software typically uses a set of predefined modules for plant equipment, such as boilers, heaters, pumps, and turbines. These components are linked together to represent the part of the system being investigated. Detailed design information is provided for each component, and inlet and outlet streams are defined. The program performs very detailed calculations within each module, and iterates to converge on a complete solution of the system. Setting up a model is typically done by a process engineer, and may take a few days to a few weeks, depending on the level of detail required. Model output must then be validated by comparing with actual process conditions. Frequently, validation will require more information than is available through the plant's instrumentation and data acquisition equipment. In many cases we have installed additional temperature sensors and temporary data collection devices to monitor intermediate points in system. Some "tuning" of the model may be required to account for variable such as fouling factors or equipment deficiencies before it can be used reliably. Once the model has been developed and validated, various cases can be run to evaluate the impact of process changes. Several weeks of training and working with the program are necessary to develop a proficiency in setting up models, but much less experience is needed to run the model. For this reason, these process models can be developed by the corporate engineering department, and transferred to the plant operations staff for day-to-day use. Predictions from a validated model are compared with actual process measurements such as flows, temperatures, and pressures. Differences between predicted and measured values can indicate equipment deficiencies such as fouling or steam leaks, which can assist in maintenance planning. Weather changes (ambient temperature, barometric pressure, humidity) can also affect power plant performance, but operators may not always recognize the impact. With the same operating regime, the plant may appear to generate a few extra megawatts on some days, and underperforrn on other days. This may lead the operators to spend hours chasing nonexisting problems as they try to get the performance back up. A working plant process model allows operators to evaluate the impact of weather on plant performance in less than an hour. In a more advanced application, Covanta is currently evaluating the use of process models to measure the effectiveness of boiler cleaning. This application involves modeling the overall boiler as a series of separate boiler modules, with flue gas and water/steam flows tied together appropriately as shown on Figure 2. Based on known operating conditions, the model calculates heat release in the combustion process, and heat transfer and flue gas temperatures within each section of the boiler. Fouling factors are determined for each boiler component for the specific time accosiated with the data collection period. This allows us to track fouling of boiler components over time and establish the point when cleaning is required. Applying this method before and after cleaning, can become a valuable engineering tool in evaluating the effectiveness of the specific cleaning techniques used in each boiler area. Process modeling has proven to be an effective tool for quickly evaluating the impact of operational 133

4 changes such as taking a feed water heater out of service, switching from electric feedwater pump to steam driven, varying dearator pressure, or changing main steam temperature. A modest effort is required to develop a "base model" for a facility, but once constructed the model becomes a valuable tool for operational "what-if' scenanos and helps optimize plant operation. Cold Flow Models Flow modeling with physical models is a proven and long-used approach to understanding and improving flow-related problems. Over a number of years, Covanta has had great success in several areas and continues to use these models today. The most successful applications have been in the air pollution control (APe) systems, where temperature changes are small and poor flow conditions are causing problems. When applied in the right areas, physical models have proven to be very successful and cost-effective tools. Figure 3 illustrates a scale model of a complete boiler from grate to economizer. This is typical of the physical models that Covanta has used, being constructed of clear plastic and wood at a 1/6 scale. The transparent walls allow observers to visualize flow conditions by introducing smoke or light particles to the process. These tests can be videotaped to provide a pennanent record of the initial flow conditions and the affects of various proposed modifications. Construction and operation of these models requires the specialized knowledge and equipment of a specialty testing company. One of the strong advantages of cold flow models is the ease and low cost of testing alternate arrangements. Elements such as baffles, turning vanes, and diffusers can be easily installed, moved, and modified until an optimal configuration is found. Quantitative results are obtained by several means. These include measuring gas velocities on a multipoint grid at various cross-sections of interest, measuring pressures at various points in the model to assess pressure drops, and through use of tracer gases, such as S02 or NH3 which are introduced into the model and measured at various points using an appropriate gas analyzer. These models can clearly identify regions with poor flow conditions, such as very high velocities or flow recirculation. Most of Covanta's tests relied on visual observations to evaluate mixing effects, and on velocity profiles to evaluate the uniformity of gas flow at important points in the system. Using these velocities, the flow distribution can be determined and compared for various alternatives. A highly successful series of tests was conducted beginning in scrubbers I to improve gas flows in wet These tests showed very non-uniform flow distributions inside several scrubbers, with very high velocities in the central region, surrounded by a large stagnant or recirculating region. This caused a number of operational problems, including heavy buildup on the scrubber walls and very short residence time in the scrubber vessel, resulting in carryover of wet particles into the baghouse which caused bag blinding. Modifications to the scrubber inlet, and in some cases the outlet, were modeled in the test and shown to greatly improve flow conditions in the scrubbers. Based on these tests, significant improvements have been made to the plant equipment at 9 Covanta facilities. Direct validation of physical models is difficult because the primary quantitative result of the model is a velocity profile, yet velocity is extremely difficult to measure in the high temperature, high particulate flow streams present in a waste-to-energy system. Thus, validation of physical models is generally based on correlations between actual plant performance and flow patterns noticed in the model. For example, do the areas of greatest superheater tube damage correspond to an area of the model with a clear flow anomoly, such as very high gas velocities? In the scrubber model, recirculating flows occurred in the model in the same areas where heavy buildup was seen on the actual scrubber's walls, thus providing a clear and convincing explanation for the buildup. Similarly, the high gas velocities and short residence. times within the scrubber vessel explained the presence of wet particles at the scrubber outlet. Covanta has also used physical modeling to improve performance of overfire air systems in the lower furnace. The lower furnace is a very complex and dynamic region with high velocities, large temperature differences between gas streams, and complicated chemical reactions occuring. The primary goal of cold modeling in this region is to assess the extent of mixing by the overfire air and identify any areas where gases from the grate are able to get past or between the overfire air nozzles, giving an opportunity for incompletely burned gases to excape the lower furnace. However, Covanta's success in modeling of this area has been limited. Modifications have been made to the overfrre air systems on several boilers based on physical model 134

5 predictions, but the expected improvements have not always been seen. Covanta currently views physical models as one of the primary tools for evaluating and improving flow conditions in regions with minimal temperature change, with particular success in APC systems. Computational Fluid Dynamics (CFD) Models CFD modeling is a very powerful tool when appropriately applied. Unfortunately, it is relatively easy to put together a simple model and generate impressive-looking pictures which have little relation to reality. As a result, its capabilities are over-rated by some, under-rated by others. Our technology partner, MARTIN GmbH, has utilized CFD modeling for a number of years with different experience. Figure 4 is an example of results that a research institute obtained from three different CFD models of the same boiler using identical inputs. boiler, but cannot be applied in the lower furnace. It is unclear how valid these results are in between these regions, for example in the upper furnace. The first attempts to include detailed combustion calculations in a CFD model were based on a separate combustion modeling program, with the output of the grate model used as the input to the CFD model This quickly became very complicated because the grate model ended below the secondary air nozzles, yet secondary air has significant impacts on primary combustion. As a result, it was necessary to couple the models using an iterative approach in which the initial combustion results were input to the CFD model which calculated a solution including radiation onto the combustion bed. These results were introduced back into the combustion model for recalculation, and the output again introduced into the CFD model. Several iterations were typically necessary before reasonable agreement was obtained. Even with this effort, there were always discrepancies in the energy balance at the interface between the two models, and uncertainties about the accuracy of the results in the lower furnace - the key region being investigated. There are several possible reasons for such a pronounced difference in the results. One which frequently occurs is computational time related to convergence criteria. Figure 5 shows side by side two temperature profiles computed using the same model of the same boiler with identical inputs. The only difference between the left and right profiles is the number of iterations before the model terminated to output the results. CFD modeling requires a lot of computer power and time, which may not always be available. Attempting to speed up the progress by terminating calculations too early, or by relaxing convergence criteria can lead to erroneous results. Quality of the mesh ("brain cells" of CFD models) is also very critical for developing an accurate and compehensive CFD model and generating credible results. A finer mesh generates more accurate results, but increases the cell count which requires more computer power and iteration time. Achieving the delicate balance between mesh resolution and computer capabilities IS an important key to a successful CFD project Some of Covanta's early experiences with CFD modeling utilized simplistic combustion modeling at the grate. These models were based primarily on UF A distribution, along with estimates about the combustion process, but did not actually attempt to model combustion of refuse. These models can provide meaningful results downstream in the The next step in modeling was to develop a proprietary combustion model within the CFD program so a single model could represent the entire system. This approach, which is currently in its final development stage, integrates MSW or RDF combustion into a single boiler model and provides full interaction between primary (undergrate) and secondary (overfue) air streams with proper simulation of very complicated mechanisms of chemical reactions and heat transfer in the lower furnace. Similar to the Process and Cold Flow modeling, Covanta sees CFD model validation as the most critical part of project development. Typically, very extensive data from field testing is collected in order to validate a CFD model from the technical standpoint, and to achieve a high confidence level to make serious boiler modifications. Figure 6 compares CFD results with HVT and infrared temperature detector measurements in one boiler model. In one project, serious modifications were made to the superheater at the Essex County Resource Recovery Facility. Results from conventional process and boiler models differed from those predicted by the CFD model, primarily because the CFD model is able to accurately represent variations in gas flow and temperature within the boiler, whereas conventional models assume that the gas flow and temperature are uniform within any cross- 135

6 sectional area. Upon completion of the modifications, field tests found very good agreement between the model predictions and actual performance2. Figure 7 illustrates the nonuniform temperature distributions predicted by the CFD model. Generating meaningful results with CFD requires a significant commitment over many years to develop, validate, and continually improve the models. It requires a thorough understanding of both detailed modeling techniques and WTE operations. Covanta management has made this commitment with a full-time PhD level employee devoted to CFD modeling, with significant support from others in the technical operations group, and two dedicated computers with commercial CFD software. The goal is to use CFD modeling as a useful tool for areas where other approaches are not successful, particularly the primary and secondary combustion processes in the lower furnace. Conclusion Clearly, there are many tools available to assist in solving problems in Waste-to-Energy systems. Successfully resolving a problem requires first that the correct tools be applied, second, that they be applied appropriately, and finally, that the results of the evaluation be properly interpreted. We believe there is a synergistic value in having this variety of engineering tools within one Process Engineering Group, as we have in Covanta. Combining Field Testing with Process Modeling helps tremendously in setting up and validating a CFD model. Commercially available process modeling software, with all its capabilities, can only represent average values at any point in the system. Deviations from these average values, for example higher gas velocities and temperatures in certain areas of the furnace, cannot be evaluated with process models. Additionally, many of the process modules rely heavily on empirical data obtained primarily from utility type boilers, rather than WTE experience. As a result, CFD models in combination with field testing become extremely important ld cross validation of boiler process models. Each of the tools identifed in this paper can be quickly applied to a situation by someone with an incomplete understanding of either the tool or the system being investigated, with the result generally being unproductive. Making effective use of these tools requires an understanding of their capabilities and limitations, a committed effort to evaluate and understand unexpected results, and sufficient experience and judgement to accurately interpret results in view of the real-world process. It is ultimately the last step which determines whether or not a tool is useful. Results of any investigation must make sense when viewed by the operating and engineering professionals within the company. Within Covanta we have had numerous discussions with vendors and consultants over results obtained by applying tools and methods developed in other industries to WTE applications. In many cases we have concluded that despite working very well in other applications, these approaches cannot simply be transferred to the WTE industry as-is. In many cases the technique may ultimately be well-suited to our industry when properly modified to account for differences, but this is not always the case and as WTE professionals we need to be aware of these concerns. I Gorsky, Nikita and Bowen, C. F. Peter, "Improving Semi-Dry Scrubber Performance through Gas Flow Modeling," NAWTECI , Epelbaum, Greg, "Corrosion Study and Modification of Superheater Tubes in a Large Mass Burn WTE Boiler," NAWTECI ,

7 Furnace Temperature Data Comparison of HVT, IR, and Plant Thermocouple u a 1800 f! Q) Co 1600 I- IR HVT Plant Thermocouple :37 14:47 14:57 15:07 15:17 15:27 15:37 15:47 Time, hh:mm 15:57 Figure 1- Temperatures measured by High Velocity Temperature (HVT) Probe, IR pyrometer, and station thermocouple II '.:at :\j.ovl Air SH I V &J '..,, '., '.M,""". Water1Stllam _ 2 ij Bottom Ash Feedwater U ' I lui I ::: I ' ",".l, Boiler Exit t '"d Figure 2 - Process Model for Mass Burn WTE Boiler 137

8 Figure Cold flow model of a complete WTE boiler 138

9 : : fl 1 Figure 4 - Results of three CFD models based on the same boiler design and operating conditions 1.80e e e e e e e e e e e e e ( l n ( Ic:: : 3.00e+02. "1 Figure 5 - Results from the same model using different convergence criteria 139

10 RlghtWaII r: "'... --=.:::':.:.'" -:. :::. :. :::.: :':"IR.77 e 03.i4e n.72,' 03.70e C3. h+c3.tqe 03.C20 01 He O,.57e OJ.;.. 03.S2e1-u3.50e 0,).47c.. 03."!S.e.. (:J.42e+OJ.4 Dc" ,3,f;' (d.32e 01.30e '0] Local Temperature "2 ft n " 4 ft" " 6 ftn " 8 ftn Area from RW from RW from RW from RW Peak HVT Testing IR CFD Figure 6 - Modeled and measured flue gas temperatures in a mass burn WTE boiler 140

11 1.20e+O 1.18e+O 1.16e+O 1.15e+O 1.13e+O 1.11e+O 1.0ge+O 1.0ae+O 1.06e+O 1.04e+O 1.02e+O 1.01 e e+O 9.70e+O 9.53e e+O 9.17e+O 9.00e+O 8.82e+O 8.64e+O 8.47e+O 8.2ge et e+O 7.76e+O 7.58e+O 7.41e+O, 7.23e+O 7.05e+O 6.88e+O r--x 6.70e+O Z l.20e+o l.18e+o l.16e+o l.15eto l.13eto l.lle+o l.0geto l.08e+o 1.06e+O 1.04e+O l.02e+o l.ole+o 9.88e+O 9.70e+O 9.53e+O 9.35e+O 9.17e+O 9.00e+O 8.82e+O 8.64e+O 8.47e+O 8.2ge e e+O 7.76e e+O 7.41e+O 7.23e+O 7.05e+O 6.88e+O rx 6.70e+O Z Contours of Static Temperature (fl Feb 20,2004 FLUENT 6.1 (3d, segregated, spe6, rkel Contours of Stat i c Temperature (fj Feb 20, 2004 FLUENT 6.1 (3d. segregated. spe7. rkel Mass-Avg. Before SH AfterSH Temperature, F Modification Modification T CFD Model Process Model TIC and IR (Not Mass-Avg.) Figure 7 - CFD Predictions showing non-uniform temperature distributions at superheater outlet before and after modification, and comparison with process model and field measurements 141