The main objective of a solar furnace is to test samples of different materials by following prescheduled temperature profiles using the sun as energy source. Figure 5 shows the inside and outside views of the solar furnace of the PSA-CIEMAT.

As the primary energy source (solar radiation) cannot be manipulated, the energy entering the solar furnace is controlled and modulated using a shutter, which is the main control variable. The Solar Furnace Control System (SFCS) is composed of a data acquisition system to measure the sample temperature in different points, the solar radiation, shutter aperture, etc. (the heliostats automatically track the sun). Thus, the main objective of the SFCS is to control the temperature profile of the sample in spite of changes in solar radiation.

From the control viewpoint, the solar furnace is a system which presents various interesting characteristics which make the control problem a difficult task [11]:

• The characteristics of the samples are quite different depending on their nature (steel, cupper, etc.). Obtaining a fixed parameter controller to allow different samples to be controlled becomes a difficult task.

• The dynamic characteristics of each sample greatly depend on the temperature and introduce a high nonlinearity in the control system, which makes the behaviour of the controlled system change with the operating conditions.

• The control specifications are quite severe (rate of temperature increase, rate of temperature decrease, variable step changes, etc.) and have to be achieved with small errors.

• The system suffers from strong disturbances caused by solar radiation variations (slow variations due to the daily cycle or fast and strong variations due to passing clouds), which make the exact reproduction of the conditions of a determined test impossible.

• Limitations exist in the maximum temperature achievable by the materials and different constraints (nonlinearities) in the actuator (amplitude, slew rate, etc.).

The control algorithms that are being developed take into account these aspects. So, the main control candidates are gain scheduling or adaptive control [11], to account for changing dynamics, predictive control to take into account amplitude and slew rate constraints, robust controllers to accommodate modelling uncertainties and fuzzy logic control to include experts knowledge (or a combination of these techniques, that have been successfully used in distributed solar collector fields [12]). Most of these techniques require dynamic models of the system. It is important in this case to use a modelling technology that allows reuse and easy modification, to avoid coding a new model each time a sample of a new material has to be tested. Again, the object oriented modelling language for physical systems Modelica [10] has been selected. These models are going to be used both for simulation and control purposes, to allow pre-evaluation of the developed control algorithms and a real-time implementation of the models, for instance, as prediction models within a model predictive control framework. Figure 7 shows an interface of the developed models and figure 8 an example of the evolution of the temperature profile in the sample under constant solar radiation conditions and after a step in the shutter aperture at the beginning of a test.

Once the control strategies have been developed, it is also important that the SCADA (Supervisory Control and Data Acquisition) system allows easy integration of the controllers. The SCADA of the SFCS has been developed using Labview-DCS, that permits the easy integration of controllers and models developed using Matlab, Simulink, Dymola/Modelica, C++, etc. and can communicate with other systems using OPC (Ole for Process Control). Figure 9 shows a snapshot of the main screen of the SCADA system.

1. Conclusions

This work has presented the main activities and research lines that are being carried out within the scope of the specific collaboration agreement between the PSA-CIEMAT and the "Automatic Control, Electronics and Robotics” research group of the Universidad de Almerla. An overview of the decisions made in the selection of SCaDa systems, real-time distributed control systems, modeling and simulation environments has been included, as far as some ideas about the main control algorithms suitable for controlling this kind of plants.

2. Acknowledgements

This work has been performed within the scope of the specific collaboration agreement between the Plataforma Solar de Almerla and the Automatic Control, Electronics and Robotics (TEP197) research group of the Universidad de Almerla titled "Development of control systems and tools for thermosolar plants” and the projects financed by the MCYT DPI2001-2380-C02-02 and DPI2002-04375-C03. The authors would like to acknowledge many people who are involved in the mentioned agreement and projects, mainly Jose Domingo Alvarez (CRS systems) and David Lacasa (solar furnace).