Condenser re-engineering improves plant performance20 April 2000
The performance of most power plant condensers can be improved, but is it economically viable? This is a difficult decision but computational fluid dynamics can help. Two recent plant case studies illustrate how improved performance can be achieved. N Rhodes and J D Castro, Mott Macdonald, Croydon, UK; R Keller, SCE, San Onofre, USA; J F Lund, TVA, Chattanooga, USA; R J Bell and C D Hardy, Heat Exchanger Systems, Weymouth, MA, USA
Modifying a power plant condenser can be one of the most valuable and cost-effective ways of improving output or reducing fuel costs. In some cases it is possible to exploit any conservative features built in by the designer to achieve a performance target. In others it is a question of effectively tuning the condenser to give near perfect thermal performance, exploiting all the possibilities for heat transfer. The difficulty is in identifying if and where such possibilities exist.
The use of computational fluid dynamics (CFD) techniques makes this re-evaluation possible.
The techniques involve the development of a three-dimensional numerical model of the flow and heat transfer processes in a condenser. The model is based on a solution of the fundamental equations of fluid motion. The realism of the solution is only limited by the resolution of flow which the model can represent and our present knowledge of the heat- and mass-transfer mechanisms. Validation of the models is by way of plant data and, ideally, data such as CW outlet temperature maps obtained by instrumenting the outlet water box. When confidence in the correct setting up of the model is achieved it can be used to provide an assessment of the as-built condenser, examine the underlying causes of poor performance and be used as a test bed for possible modifications which can be proposed to overcome any weaknesses.
Two recent projects, one at Southern California Edison’s San Onofre plant (2x1100 MWe PWRs) and the other at Tennessee Valley Authority’s Watts Bar (2x1200 MWe PWRs), show how CFD has been used to assess condenser performance and determine potential improvements. In the San Onofre case, the objective was more to examine performance details such as the effect of stopping one of the CW pumps and whether tube vibration might result in this mode of operation.
At Watts Bar, a re-tube was in prospect with alternative tube materials which would render a lower heat transfer performance. In both studies, subtle weaknesses in the condenser were identified and modifications were designed to overcome them.
The results in both cases have been improved performance. San Onofre carried out a partial implementation of the recommended modifications, which brought about a substantial decrease in the oxygen levels in the condensate. This was felt to be due to a reduction in air binding in the condenser. Because the Watts Bar study was done prior to a retube, full implementation of our prescribed modifications was possible, and an improvement in back pressure of about 4 mbar was achieved.
The basic approach
In the application of CFD techniques, the region of interest, in this case the condenser, is divided into a number of small volumes or cells. The equations of fluid motion are solved numerically in each cell to yield values of velocity in each co-ordinate direction, pressure and steam/air concentration.
The computational model for the flow of steam through the condenser includes factors such as inundation (the increase in water-film thickness over the tubes caused by condensate falling from tubes higher in the nest) and the frictional pressure drop due to the tube-nest. Individual tubes are not usually represented individually in the mesh. Instead, a standard correlation is used to calculate the frictional pressure drop in each cell based on the number of tubes present and the Reynolds number of the flowing steam. In the present study, however, sub-models have been developed which predict the flow around individual sections of tube nest to provide detailed design of modifications.
Before any physical modifications to a condenser could be proposed it is essential to establish considerable confidence in the model. The basic scientific framework has been validated by application to many condensers, so what is in question in any given project is to ensure that the geometry and operating conditions are understood and correctly represented in the model.
Models are developed which represent the tube nest, tube support plates, obstructions within the condenser such as feed-heaters, drip trays and other significant structural elements. Simulations are then performed for plant operating conditions. In the model, this involves setting the steam inflow, CW inlet temperature and flow rate, and with a representation of the air offtake equipment. The model then predicts everything else, and reviewing the overall turbine exhaust pressure and the distribution of parameters through the nest, velocities and air concentrations, provide new insights into the efficiency of the condenser.
The knowledge which is thus obtained is used to devise modifications which might be made to the plant. Various considerations apply, particularly the difficulty of engineering the modification, so if a re-tube is planned, then more elaborate changes can be made. These are tested with the model to determine the performance improvements, and this information contributes to the financial case for the change.
The San Onofre case
The condenser unit at San Onofre consists of three separate condensers: two low pressure sections and a higher pressure section sandwiched between them. Cooling water flows from each end to the outlet cooling water boxes in the centre of the condenser. There are two inlet water boxes and eight tube bundles at each end of the condenser. Figure 1 shows a diagram of the full model of the condenser. A finer-grid model of one-quarter of the condenser unit was later developed to design modifications which would address shortcomings in performance
The modelling study was carried out to evaluate the performance of the condensers, to identify areas of air-blanketing and regions where tube vibration problems are apparent in the event of circulating water bumping – when the CW supply to one of the water boxes is turned off. Plant data and CW outlet temperature measurements were available to validate the model, which represented the 3D flow behaviour in all three condensers. The model included the tube-plates, tube-regions, central water boxes, baffles and all the air-offtake points. The bottom of the model was taken to be the normal water level indicated on the design drawings.
Measured operating conditions were as follows:
(one condenser) (lb/h) 2 585 165 (LP)
2 585 125 (HP)
Cooling water inlet
temperature (°F) 59
flow rate (gpm) 820 000 (205 000 per box)
Air in-leakage (lb/h) 10.915 (12.5 cfm)
Using the model back-pressures were predicted and compared with measured values, with the following results (inches of Hg):
LP1 HP1 LP2
Measured 1.76 2.17 1.62
Predicted 1.70 2.12 1.70
These simulations confirmed that the model represented the plant reasonably well. The steam/air mixture predicted at the offtake point indicated a high steam flow from the upper off-take, well in excess of the normal 2.3 lb of steam per lb of air, indicating severe steam leakage.
Figure 2 shows the velocity pattern in one quarter of the tube nest in the plane of the air offtake section. The high steam flow directly towards the air cooling section can be seen, and also that the steam flow into each bundle is far from uniform. There is a strong flow between the two bundles and at the symmetry plane on the right of Figure 2. Adjacent to the side wall the steam flows down next to the upper bundle and upwards next to the lower.
The cooling water temperature rises predicted by the model matched the experimental data fairly well. Predicted temperatures are on the high side in the upper nest. However, the general trends in temperature throughout the nest are well represented, particularly below the level of the air-offtake ducts, where air blanketing is predicted by the model.
Having established confidence in the model studies were first carried out to assess the case of the bumped waterbox. In this event, the predicted pressures for each condenser rose to the following values (inches of Hg):
LP1 pressure 4.21
HP pressure 4.08
LP2 pressure 1.76
This result represents the steady state conditions that would be expected to be achieved in the event of bumping. The air removal equipment was shut off for the bumped waterbox, with the steam load for each of the LP and HP condensers fixed at their normal values.
Studies of the velocities in the tube nest showed little likelihood of tube vibration problems, largely because of the rise in pressure of the bumped side and consequent increase in steam density and lower velocity.
A series of possible modifications were then studied with the model to determine whether the normal operation of the condenser could be improved. To stop the direct steam flow to the air off-take, the use of baffle plates above the nest, with some plates going into the nest on either side, were tried. These were found to be effective in reducing the steam bypassing, but resulted in higher pressures in the condenser. Sealing the gaps around the air offtake hoods was found to be more effective, but less easy to implement in practice. A combination of these approaches was taken on site, with additional plates and some sealing of the lane close to the off-take hoods.
Plant measurements are not yet available, but initial observation indicate that the condenser pressure was lower and that there was less dissolved oxygen. Both of these observations point to a confirmation that the steam leakage is reduced and the sizes of the air bubbles in the condenser are smaller.
The Watts Bar case
The TVA Watts Bar condenser units are divided into three zones: low (A), medium (B) and high (C) pressure. The sections are divided by baffle plates which run down to the level of the false bottoms in Zones A and B and the reheating trays near the bottom of the condenser in Zone C. The tubes in each condenser section are supported by a number of tube support plates, which are themselves attached using supports to the sides and floor of the condenser shell. Figure 3 shows an outline of the modelled geometry.
Steam enters the condenser via exhaust openings in the top of the condenser– one for each zone. Air being removed from the condenser is free to move along the length of the unit through the air-offtake hoods to any one of the four air-offtake points at the low pressure end (in Zone A). Holes in the division plates between the zones allow air from Zones B and C to travel the length of the condenser to the offtake point.
The study was prompted by a forthcoming re-tube of the condenser from 90-10 Cu-Ni to Sea-Cure tubing, this being done to remove copper from the system. However, the Sea-Cure tubing has a lower thermal conductivity and so a worsening of performance would be inevitable. A reduction in pressure of about 0.06 inches of Hg was calculated as a result of this.
Simulations were performed for summer and winter measured operating conditions. The runs reported here were based on the following summer conditions:
Steam flow (lb/h) 2 508 968 (Zone A)
2 473 254 (Zone B)
2 489 921 (Zone C)
Cooling water flowrate (gpm) 411 724
Cooling water inlet temperature (°F) 85.21
Again, the model was used to predict pressures and these were compared with the measured values (inches of Hg):
Zone A Zone B Zone C
Measured 2.57 3.5 4.78
Predicted 2.73 3.73 5.03
For this case also, a very high steam/air ratio was predicted, in excess of 10 lb steam/lb air. This result prompted measurements at the plant which confirmed a high level of bypassing directly to the offtake points.
Figure 4 shows the velocity field for this condenser. In this case there is also a high steam flow from the top of the condenser down towards the offtake hood. In addition, there is a high flow on both sides of the diagonal drip trays.
Numerous modifications were then tried in the model to reduce the steam bypassing, while maintaining the present pressures in the condenser. The best option developed at this stage was one where blockages were made outside air-offtake hoods and along diagonal drip trays, as illustrated in Figure 5. The resulting predicted pressures (inches of Hg) for this case were:
Zone A Zone B Zone C
2.77 3.77 5.14
There is little difference in the back-pressures, apart from in Zone C, where a small rise is observed. The air-offtake steam concentrations, however, fell sharply, indicating a steam/air flow ratio of 1.96 lb steam/lb air, below the design value of 2.2. This indicates that the pumps would be able to work at a lower pressure, bringing the back-pressures down in the condenser (particularly in Zones A & B). Calculations based on the minimum attainable pressure for the offtake pump show that the expected improvement would be of the order of 0.1 in of Hg. Calculations were then carried out which showed that the modification only had to be carried out in Zone A to produce almost all of the benefit and this had the advantage of bringing down the Zone C pressure to its former value. Limiting the modifications to Zone A was of immense value in terms of the re-tube operation, since less work needs to be done.
Reviewing the study results with the team, alternative implementations were reviewed and further very detailed models of the flow around specific baffle shapes were studied to provide further optimisation of effect and simplicity in application. The resulting baffle is similar to that shown in Figure 6, a flow visualisation from one of these studies, and was made from 13 gauge stainless steel, and in a form which could be readily inserted on site.
The modification was implemented in the condenser during replacement of the tubes. The measured performance is summarised below (pressure in inches of Hg):
Zone A Zone B Zone C CW temp (°F)
Pre-mod 2.57 3.47 4.78 85.2
Post-mod 2.42 3.46 4.65 84.4
The post modification operating data have shown that the modifications installed in Zone A have improved the condenser performance over that which would have been expected by the change in tube material from 90-10 copper nickel to Sea-Cure stainless steel.
Confirming the business case
The application of CFD techniques provides a fundamentally based method for the prediction of condenser performance. The model provides a flexible way of understanding the detailed flow behaviour within the condenser, enables the identification of shortcomings in the performance, and provides a means of testing and optimising modifications before they are implemented on site.
In the two case studies presented above, substantial improvements in condenser performance were made. The modifications required are different in each case, reflecting not only the design and operation of the units, but also the feasibility of different re-engineering concepts.
At the TVA’s Watts Bar condenser, where post-modification results are available, the benefits of the modification not only recovered the shortfall due to new tube material, but also gained an additional performance margin.
As with any plant modification there has to be a justifiable financial case. The model provided vital information on the magnitude of condenser improvements without which the cost benefits could not have been estimated. The realisation in practice of the benefits predicted by the model confirmed the business case.
Since these projects were undertaken, studies for Drakelow power station in the UK and Manitoba Hydro's Brandon station have been completed, and a study is currently being carried out for Amergen's Clinton station in the USA.