The construction of a novel steam generation system for concentrating solar power (CSP) plants, developed in Germany by Balcke-Duerr, has been aided by using a two-phase flow dynamic simulation programme.

Thibault Henrion, Vienna University of Technology, Austria & Vitali Tregubow, Balcke-Duerr GmbH, Germany

In the development process of a new type of steam generation system for a concentrating solar power plant (CSP), a dynamic simulation was used to assess the transient behaviour of the system. Because the solar boiler will be started up and shut down on a daily basis its start-up time is one of the key factors influencing the profitability of the unit.

It is therefore crucial to determine during the start-up phase that the water/steam circulation begins correctly and in what direction. Stagnation or a start-up with a reverse flow could lead to instability in the water circuit and increase the start-up time. In order to check the ability of the system to begin operating as soon as the sun delivers its heating power or its response to rapid changes in solar radiation levels (i.e. cloud formation or during a storm), the model has to take into account the thermal inertia of the materials.

Boiler Description

The principle of a CSP plant is well known, i.e. solar radiation is reflected by mirrors and focused on an absorber pipe through which thermal oil flows. It is heated up to approximately 400 °C, and this hot oil is used to generate steam through a heat exchanger. The main feature of the Balcke-Duerr patented advanced solar boiler is its innovative bundle design.

The design of the advanced solar boiler differs from other concepts because it integrates the three different heat exchanging areas of the steam generation system (i.e. the economizer, steam generator and superheater) into a single cylindrical frame. Also, a cross-counter-current flow design is used for the bundles (see Figure 1).

Figure 1: Schematic of the steam generation system

The feedwater flows to the economizer area of the boiler, where it is heated to just below its boiling point. The geometry of the system, equipped with a steam drum, has been designed to allow a natural water circulation of the water/steam mixture in the four steam generator sections.

The resulting steam flows from the drum to the superheater, where it reaches its final temperature of 380 °C. At the same time, oil flows counter-current to the water stream and cools down from 400 °C to 300 °C.

Establishing the Model

The unit is modelled using the APROS1 simulation programme, which is a tool used to provide dynamic simulations of both thermal and nuclear power plants. It allows the modelling of a two-phase flow and considers each phase separately (six-equation model2). The possibility of varying the model’s boundary conditions and modelling elaborated control systems makes this programme ideal for the purposes of the investigation.

The different assumptions and hypotheses taken into account for the simulation are discussed below.

APROS features a database of different fluids that can be simulated; it includes two different predefined types of oil. This oil model is thought to have been develeoped for studying combustion rather than heat transport.The oil database of APROS allows adjustment to be made to the properties of the oil. The oil properties calculations are based on the density and the dynamic viscosity at 15 °C and 100 °C. The APROS calculations of fluid properties were reproduced in order to adjust the thermal oil parameters (i.e. enthalpy, viscosity and density) so that they remain within an acceptable error margin within the relevant temperature range of 300–400 °C.

Despite adjustments made, errors remain significant for heat capacity and heat conductivity, so their influence on the heat transfer coefficients calculations is taken into account in the modelling of the heat exchangers.

The standard module of the APROS ‘counter-current heat exchanger’ is used to model the heat exchanger’s areas. Primarily, one heat exchanger is used to model one bundle; it is discretized into 30 nodes to achieve a high accuracy in the heat transfer calculations. The pressure losses coefficients are calculated and given as model inputs.3, 4

Following several experiments, the steam generator model accuracy is increased by improving the discretization of the bundle. The latter is divided into seven separate heat exchangers, discretizising its height. This allows for a better modelling of the flow inside the headers and gives better results concerning the cycle criterion of the steam generators.

The piping geometry is accurately reproduced in the model so that the natural circulation of the boiler is simulated with a high definition. Diameter, length, elevation differences and pressure loss coefficients are significant inputs for the different modelling components, i.e. downcomers, risers, steam drum, etc. However, a few simplifications were made.

The steam drum internal components cannot be reproduced with APROS, therefore the elevation of the riser pipes is placed above the water level of the drum so that no steam is entrained in the downcomer pipes. An ideal separation of water and steam is simulated in the steam drum; all four downcomers are connected to the steam drum at the same point; and all four risers are connected to the steam drum at the same point.

The numerical simulation of the transient processes also requires a precise modelling of the unit control system. Three control loops were reproduced for the control of the solar boiler during the start-up phase.

The boiler feedwater supply is realized in the model with a pump located before the economizer. Its speed is controlled in order to adjust the water mass flow so that the water level in the drum reaches its set point. In order to regulate the steam drum pressure, a valve controls the steam outlet after the superheater. This modelled piece of equipment is comparable with a turbine inlet pressure control valve.

The blow-down of the steam drum is also simulated. A valve regulates the water mass flow that is evacuated from the steam drum at a constant set-point value.

To validate the boiler design parameters, the unit was first simulated with constant boundary conditions equivalent to full load parameters and brought to a steady state. On the oil side, the input temperature, mass flow and pressure are given under constant boundary conditions. Among the relevant boundaries of the steam/water side, a value is fixed for the feedwater temperature and the steam drum pressure set-point.

During the simulation of the start-up phase, the above cited parameters are variable. This can be easily reproduced with APROS because of boundary condition modules. A table of values is given for each boundary that varies with the elapsed time of the start-up phase simulation. APROS interpolates linearly between the given points.

A characteristic ramp, from the initial values to the final steady state values, is given as input for each four variable boundaries: oil mass flow and temperature, steam drum pressure set-point and feedwater temperature. The blow-down mass flow is kept constant during the whole simulation.

The steam drum water level is controlled (along with the speed controlled feedwater pump) due to a PID control loop. Experience shows that appropriate control parameters only work for a minimum mass flow of the pump. Therefore, in addition to the previously mentioned boundaries, the feedwater mass flow is given as an additional boundary condition during the beginning of the start-up phase.

As a consequence, the steam drum water level control loop is off-line during that time.

The feedwater mass flow profile is chosen so that the fluctuations in the water level stay within acceptable limits. Approximately halfway through the start-up phase, the pump overtakes the control of the steam drum water level.

Model Validation

The validation of the model is, combined with the model build-up, an iterative process. In order to reach optimal simulation conditions, the model results are compared with the desired reference data (oil outlet temperatures, preheated feedwater and steam flows and superheated steam temperature, as well as the cycle criterion).

Differences are therefore identified and their causes analyzed. Model parameters like discretization precision and heat exchange efficiency (heat transfer coefficient) are responsible for these differences and are adjusted in order to provide a more accurate model.

First of all, the design conditions of the unit are programmed into the APROS model. This simulation corresponds to a full load operation case.

Initially each heat exchange area (steam generators, superheater and economizer) is simply modelled in APROS, with one heat exchanger module with a maximum of discretization nodes. Simulation revealed noticeable differences between the simulated and design steam generator’s cycle criterion values.

Subsequent simulation showed that the influence of the pressure loss coefficients of the downcomer and riser are not significant enough to explain the differences. The size of the header is identified as the main factor influencing natural convection flows. The accuracy of the modelling was therefore increased to separate the steam generator into several layers, each simulated by one heat exchanger module.

In this way, the steam/water mass flow inside the headers within the bundle can be more accurately calculated. With increasing elevation, the downcomer header mass flow decreases and the riser header mass flow increases. With the new descretization, the behaviour of the headers is better taken into account and the resulting pressure losses are assessed with greater accuracy.

Table 1 shows significantly improved values for the downcomer water flow and the cycle criterion of the modelled heat exchangers with the improved discretization.

The average errors between the design values and the original model results are 13 per cent for the cycle criteria and 11.5 per cent for the downcomer water flows. Between the design values and the improved model, errors of 10.5 per cent for the cycle criterion and 7.2 per cent for the downcomer water flow were calculated. This comparison shows a significant improvement in the accuracy of the model.

Differences, however, remain between the real oil properties and the modelled ones so a correction factor is applied to the oil side heat transfer coefficients of the APROS heat exchanger modules. In order to assess the error induced by the properties deviations, two heat transfer coefficients were calculated separately with the real oil data and the APROS simulated ones. Equation 1(3, 5) and Equation 2(6) were used for evaluating the Nusselt characteristic number giving the bundle heat transfer coefficient.

Both calculations give comparable differences between the heat transfer coefficients.The calculated differences with the Gnielinski correlation are used as correction factors (see Table 2).

The heat transfer coefficients on the water side are always higher than those on the oil side, thus the oil side coefficients determine the heat transfer. With these ratings of the simulated heat exchangers, only a small deviation is observed between the simulation results and the design values, as shown in Table 1 and in Table 3. However, the header dimensions are a limiting factor for the natural circulation in the evaporators.

The next step consists of a comparison between the start-up simulation results of the boiler and the scaled data from another existing unit but with differences in construction (i.e. reference data).

Prior to the unit starting up, no oil or water flows and both are in thermal equilibrium at approximately 200 °C. In the steam/water circuit, saturation conditions are simulated. The heat exchanger’s wall material is also in thermal equilibrium with the fluids at 200 °C. As mentioned previously, the variable boundaries are applied across the whole start-up phase, and the results of the simulation are then compared to the reference data.

Figure 2 shows a strong convergence of the predicted oil outlet temperature between the simulation and the reference data (±5 °C), except for two areas at 1:50 and 2:40. Inconsistencies in the reference data are responsible for these differences. In Figure 3, a good alignment of the model with the reference is also observed – differences in the overheated steam temperature do not exceed ±5 °C.

Figure 2: Oil outlet temperature
Figure 3: Superheated steam temperature

The profiles of the simulated water and steam mass flows in Figure 4 also closely follow the reference data. The difference at about 2:40 is due to inconsistencies in the reference data, while the control loop of the feedwater supply is responsible for the small fluctuations in the APROS water flow between 3:00 and 4:00.

Figure 4: Steam/water mass flow

An iterative process was also necessary to find the optimal feedwater flow profile, so that the steam drum level remains within an acceptable range during the start-up phase, i.e. before the feedwater pump regulation loop is activated. The adjustment in the feedwater pump control loop parameters was also made iteratively.

The validation results overall show high convergence between the simulation and reference data, with any small differences explained by either non-optimal control loop parameters or inconsistencies in the reference data. Taking these results into consideration, the simulation model should be able to provide reliable results regarding to the solar boiler start-up time.

Simulation Findings

The data from the start-up simulation (see Figure 5) show that boiling occurs in the economizer area of the system during the start-up phase.

Figure 5: Economizer parameters during start-up

In Figure 5, a vapour mass fraction reaching 1 per cent in the flow coming out of the economizer can be observed. This phenomenon lasts from around 2:15 to 3:05, and is not unusual for a steam generation system.

In addition, the simulation also shows that, under normal operation, water cools by 6 °C as it leaves the economizer (see Table 3), which is a conventional value for such a unit.

These findings indicate that attention should be paid to the steam drum design. The steam produced in the economizer during this transient phase should not flow into the downcomer pipes because it could affect the natural circulation of the boiler. In order to avoid this undesirable effect, internal components can be build within the steam drum to enhance the vapour separation, and/or a certain elevation and space gap should be maintained between the downcomer and the economizer pipes’ drum connections.

The observations of the water and steam mass flows in the downcomers entering the steam generators and in the risers flowing out of them allow the circulation inside the steam generators to be analyzed.

During unit start-up, it is observed that the circulation begins oscillating in the first steam generator (see Figure 6). It started at approximately 2:00, but thereafter the circulation becomes stable during the rest of the start-up phase. When abrupt disturbances in the oil heat input occur (i.e. a fall in the oil mass flow followed by a steep increase in the oil mass flow combined with a steep increase in the steam drum pressure at about 2:40), the simulation shows that the circulation is not seriously affected (see Figure 6). This highlights a robust natural circulation.

Figure 6: Natural water circulation in steam generator 1

Figure 7 represents the water flows in the downcomers of all four steam generators. It shows a more detailed picture of the natural circulation start. Circulation begins with oscillations in steam generators 1 and 2 when the oil mass flow remains low (low heating power), but once the oil heating power rapidly rises, at 1:58, it stabilizes quickly.

Figure 7: Natural circulation start

Also in Figure 7, in steam generator 4 a small temporary negative circulation mass flow is observed. In order to investigate this short phase more precisely Figure 8 focuses on the circulation start-up of the fourth steam generator.

In both riser and downcomer a very small water mass flow appears for a short period of about 180 seconds (see Figure 8). This reverse flow, however, does not disturb the operation of the unit for the following reasons:

  • Only the water is flowing in a reverse direction.
  • Steam flow is recorded in the riser after the flow returns to the positive direction (approximately 2:01:30).
  • No steam flow is recorded in the downcomer.

Figure 8: Natural circulation start by steam generator 4

The conclusion from the simulation results is that the natural circulation is stable as soon as the heat flux is sufficient. For the simulated start-up phase, the natural circulation begins rapidly in the steam generator heat exchangers, within 15 minutes of the start of the oil circulation.

Small negative water flow can transitionally occur in the last steam generator. However, this mass flow reverses as soon as steam is produced and does not represent any problem for the proper operation of the unit.

Conclusions and Perspectives

With the help of the APROS dynamic simulation programme, it has been possible to build up and validate a dynamic simulation of a thermal oil steam generation system, enabling an analysis of the equipment behaviour during its start-up phase prior to its manufacture. The results are consistent with design calculations and allow for its further optimization.

These numerical investigations reveal that boiling occurs in the economizer during the start-up phase. This phenomenon can be controlled through simple construction precautions.

The APROS simulation highlights that for this integrated boiler design the natural circulation begins within a short time, and also shows that this boiler concept could fulfil the operation requirements of a CSP plant. Following the construction of the analyzed steam generation system, the model could be further improved by validating it with process data in order to achieve greater simulation accuracy.

It would also be possible because of APROS’ automation tools conduct safety analysis (i.e. the simulation of an emergency shutdown), tests and optimization on some control loops parameters, or even create a unit simulator for operator training.



2. M. Hänninen, J. Ylijoki. The constitutive equation of the APROS six-equation model. 2007.

3. VDI Wärmeatlas, 9. Überarbeitete und erweiterte Auflage, 2002.

4. Heinz Zoeble and Julius Kruschik. Stroemung, durch Rohre und Ventile. Springer Verlag Wien – New York, 1982.

5. Peter von Böckh. Wärmeübertragung Grundlage und Praxis, 2. bearbeitete Auflage. 2004.

6. R. Kneer. Wärme und Stoffübertragung. Skriptum, Aachen. 2006.

The authors would like to thank their co-authors René Schimon and Karl Ponweiser from the Institute of Energy Systems and Thermodynamics, Vienna University of Technology, Austria, for their invaluable contribution to the article.

More Power Engineering International Issue Articles
Power Engineering International Archives
View Power Generation Articles on