From the point of view of automatic control, a Central Receiver System (CRS) plant may be decomposed, almost in a first approximation, in two subsystems. The first one is the Heliostat Field (HF) and the second one is the Receiver and Power System (RPS), sited at the top of the tower. Figure 1 shows both components interacting. The HF reflects incoming solar radiation in predefined aimpoints. Following this natural decomposition the control system of the CRS is divided into two functional blocks: the HF Control System (HFCS) and the RPS Control System (RPSCS).

The basic objective of the HFCS is to position coordinates all the time, depending on the PS demands. The global objective of the HFCS is the generation of a uniform time-spatial distribution of the temperature onto the volumetric receiver (in the case of the PSA CRS), by controlling the timed insertion of group of heliostats associated to predefined aimpoints in the receiver, by modifying the aimpoints coordinates and by changing heliostats from one group to another during operation. This is accomplished by applying an aimpoint strategy over the HFCS [1].

From the control viewpoint, due to the requirements of the operation, the HFCS can be considered a hard real time system and thus, it requires a real time support system [2] to guarantee the adequate positioning of the heliostats at due time. Moreover, the HFCS must permit easy integration with the RPSCS, without losing the real time capabilities. The work that is being developed at present in PSA-CIEMAT is oriented towards the development of a real time distributed control system for the HFCS. The RT-CORBA [3] technology helps to integrate heterogeneous computing platforms and has been selected to implement the communications between the different data acquisition and control sytems. In this case, the RPSCS has been implemented in LabView over Windows XP OS. This control system is described in [4] in more detail. The HFCS is being developed in C++ language using the ISO POSIX C 1003.1c API interface, over LynxOS operating system. The real time logical interface to the PSCS is being implemented using The Ace ORB (TAO) RT-CORBA implementation [5-8]. The TAO CORBA distribution is widely used for industrial and scientific purposes (http://www. cs. wustl. edu/ schmidt/TAO. html).

Preliminary tests have been performed at CIEMAT-PSA using a Cluster Beowulf platform based in a switched 100 Mb/s Ethernet network and low computing resources nodes. The experiments have successfully demonstrated the low overhead and latency that TAO introduces in two-way inter-objects calls. Depending of the test performed the total cost in time of the overhead introduced by CORBA layers (at client and server objects) and network delays are lower than 5 ms in the worst cases [9].

Using RT-CORBA in the HFCS software design, a real time logical interface to the physical field (heliostats) is provided to the rest of computers and systems connected to the network domain. These computers may interact with this logical interface with independence of the operating system and programming language, which augments and promotes the heterogeneity and overcomes the constraints imposed by proprietary technologies. An example of interaction promoted by RT-CORBA is the periodical request of coordinates from each heliostat of the HF from the LabView application in the RPSCS (figure 2).

The main objective of the RPSCS is to regulate the pressure and temperature of the steam generated with the heat collected in the solar receiver. Figure 3 shows the components of the RPS at the CESA facility at CIEMAT-PSA. Briefly it is composed of:

• Solar receiver.

• Storage tank.

• Air-water/steam heat exchanger (HEX).

• Process load (in the figure formed by a steam turbine and a condenser).

• Actuation components: K1 and K2: valves; G1 and G2: blowers; B1: water pump; V1: steam valve.

The system operates as follows: the concentrated solar radiation heats of the air flowing through the volumetric receiver. The air temperature gradient through the receiver can be controlled via the actuation over K1 and G1, so the output temperature of the receiver is regulated controlling the air mass flow rate m1 over the K1-G1-Receiver branch. This is the first control loop. The control algorithm implemented at present is an classical PID with anti-windup controller.

In a similar way, the air mass flow rate entering the HEX is controlled by acting on K2 and G2.

The second control loop must control the energy flux at the HEX inlet, by extracting/impelling air from/to the storage tank depending among other variables on the solar radiation conditions, receiver state, etc. This control loop is also based in a classical PID with anti­windup controller.

Even in the case of good control of the conditions of the inlet energy into the evaporator, variations in the outlet steam conditions in the water-steam circuit are produced. To control the outlet steam temperature/pressure it is necessary to control the pump B1 and valve V1. The control loop of the pump B1 is characterised by having a long delay (the length of the tubes is about 1330 m) and thus the actual control system based on a three state on/off controller is being changed by a Smith Predictor-based control loop [4].

• Mass flow rate through the blowers. This variable could be controlled by manipulating u1, u2, u3 and u4 represented in figure 3. This assumption is nearly true due to the almost ideal mass flow generator characteristic observed in the blower-valve sets in both loops: storage-receiver and storage-HEX mesh.

• Incoming inlet radiation at the solar receiver. This is a perturbation variable and the source of energy for the whole process.

• Mass flow rate through the water pump. The characteristic of this pump is being studied, and it seems to be highly nonlinear.

• Steam valve opening fraction.

The model output signals are all the variables that will be controlled: pressure and temperature at the inlet of the air-water heat exchanger; pressure and temperature at the inlet of the process load (main control objective), and the temperature of the storage tank. These controlled variables are also shown in figure 3.